Formation of the Compact Confomer of Kinesin Requires a COOH-terminal Heavy Chain Domain and Inhibits Microtubule-stimulated ATPase Activity*

Full-length Drosophila kinesin heavy chain from position 1 to 975 was expressed in Escherichia coil (DKH975) and is a dimer. The sedimentation coefficient of DKH975 shifts from 5.4 S at 1 m NaCl to ∼6.9 S at <0.2m NaCl. This transition of DKH975 between extended and compact conformations is essentially identical to that for the heavy chain dimer of bovine kinesin (Hackney, D. D., Levitt, J. D., and Suhan, J. (1992) J. Biol. Chem. 267, 8696–8701). Thus the capacity for undergoing the 7 S/5 S transition is an intrinsic property of the heavy chains and requires neither light chains nor eukaryotic post-translational modification. DKH960 undergoes a similar transition, indicating that the extreme COOH-terminal region is not required. More extensive deletions from the COOH-terminal (DKH945 and DKH937) result in a shift in the midpoint for the transition to lower salt concentrations. DKH927 and shorter constructs remaining extended even in the absence of added salt. Thus the COOH-terminal ∼50 amino acids are required for the formation of the compact conformation. Separately expressed COOH-terminal tail segments and NH2-terminal head/neck segments interact in a salt-dependent manner that is consistent with the compact conformer being produced by the interaction of domains from these regions of the heavy chain dimer. The microtubule-stimulated ATPase rate of DKH975 in the compact conformer is strongly inhibited compared with the rate of extended DKH894 (4 s−1 and 35 s−1, respectively, for k cat at saturating microtubules).

Kinesin is an ATP-dependent motor protein that is involved in movement of membranous vesicles along MTs 1 (see Refs. 1 and 2). The NH 2 -terminal ϳ340 amino acids of the heavy chain forms a globular motor domain (head) that has MT-stimulated ATPase activity (see Fig. 2B). The motor domain is followed by a long central coiled-coil stalk region and a small nonhelical domain at the COOH-terminal. The central region contains several positions at which the coiled-coil propensity is low and these likely represent hinges in the stalk. The first coiled-coil region that extends from the motor domain is designated the neck region. Constructs that contain the head and the COOHterminal part of the coiled-coil neck form dimers (3). There is a likely hinge at position ϳ400 that marks the boundary between the neck and the stalk. Peptides from the neck (4,5) and stalk (6) have been demonstrated to interact in a coiled-coil manner by several criteria. The crystal structure of a dimeric head plus neck construct has been recently determined (7) and it directly demonstrates the coiled-coil interactions in the neck region.
Native kinesin is a heterotetramer composed of a dimeric heavy chain core with two light chains attached in the COOHterminal region (8,9). At high salt kinesin exist in an extended conformation with an s 20,w value of ϳ6 S, but adopts a more compact conformation at low salt concentration with an s 20,w value of ϳ9 S. This global conformational transition is readily reversible and can be observed by electron microscopy (10,11). A species lacking detectable light chains is also obtained from bovine brain and undergoes a corresponding salt-dependent transition between extended (ϳ5 S) and compact (ϳ7 S) conformations (11). We report here that Drosophila heavy chain dimers also undergo a 7 S/5 S transition. Furthermore, specific regions of the tail and head/neck interact in a salt-dependent manner that is consistent with this interaction being responsible for producing the compact conformation.
Isolated head domains have a high MT-activated ATPase activity when prepared by limited proteolysis (12) or as fusion proteins (13). The rate of ϳ40 s Ϫ1 at saturating MT concentration that is observed with dimeric head constructs (see Ref. 3) is consistent with tight coupling of 1 ATP per step during motility (40 s Ϫ1 per head ϫ 2 heads per dimer ϫ 8 nm per tubulin dimer along protofilament predicts a sliding rate of ϳ600 nm/s that is equal to the observed sliding rate of single molecules, see Ref. 14, for discussion). The maximum ATPase rate of bovine kinesin in the compact conformer, however, is only ϳ1 s Ϫ1 for the 9 S heterotetramer and ϳ7 s Ϫ1 for the 7 S dimer (15) and this suggests that the inhibition of full-length kinesin versus free head domains is due to interaction of the heads with tail domains in the compact conformer. Myosin from smooth muscle and nonmuscle cells undergoes a similar conformational transition at low salt concentration that produces a compact form with a highly inhibited rate of actinstimulated ATP hydrolysis (16,17).
It has not, however, been possible to directly test the inhibition of ATPase of kinesin in the compact versus the extended conformers because the transition could only be produced at high ionic strength. MT-stimulated ATPase and MT-stimulated ADP release are strongly inhibited by even moderate salt concentrations (18,19) and it would be difficult to separate the direct inhibitory effect of salts from their possible secondary activation linked to the 7 S/5 S transition. We report here that removal of a small COOH-terminal region blocks formation of the compact conformation. This allows comparison of the MTstimulated ATPase rate of full-length kinesin in the compact conformation to that of this slightly shorter extended construct at low salt where both are potentially active. Under these conditions, the extended construct has the same high ATPase activity as short dimers of heads, whereas the compact construct is strongly inhibited.

MATERIALS AND METHODS
All reactions were performed at 25°C in A25 buffer as described previously (3). Centrifugation on sucrose density gradients and gel filtration on Bio-Gel A-5m were performed using standard proteins as described previously (11). SDS-PAGE was performed by the method of Laemmli (20) using a 4% stacking gel and a 11% separating gel, except for the separation of the long constructs (Fig. 1B) which used a 6% separating gel.
Construction of Expression Plasmids-All kinesin constructs were derived from the original Drosophila cDNA clone of Yang et al. (21). pGST864 -975 encodes the tail domain region between positions 864 and 975 as a fusion protein with GSTase. It was obtained by ligation of the PvuII fragment of kinesin that contains the COOH-terminal into the SmaI site of pGEX-3X (Pharmacia). pGST864 -893 was obtained by cleavage of pGST864 -975 with NcoI and EcoRI, treatment with DNA polymerase (Klenow), and religation. This procedure removes the coding region between amino acids 894 and the COOH terminus and results in a fusion protein with an extension of EFIVTD beyond position 893. pGST893-975 was obtained by insertion of the NcoI/EcoRI fragment of kinesin that contains the COOH-terminal into a modified pGEX-2T (Pharmacia) that contains an NcoI site following the BamHI site of the multiple cloning site. The resulting construct contains pGEX-2T sequence through the thrombin cleavage site terminating in PRGS, followed by PA and then kinesin sequence from Met 893 to the COOH-terminal. This was confirmed by DNA sequencing. pGST613-975 was similarly obtained by insertion of the NcoI/EcoRI fragment of kinesin that was not cut at the internal NcoI site at 893. Constructs terminating at 910, 937, and 960 were obtained by polymerase chain reaction using pDK613-975 (see below) as the target and with T7 promoter primer in the forward direction and reverse primers that introduced a stop codon at the appropriate position followed by an EcoRI site. These polymerase chain reaction products were cleaved with XhoI and EcoRI and inserted into a backbone derived from pGST613-975 by cleavage with XhoI and EcoRI. This produced a set of constructs in pGEX with kinesin sequence from 613 to the new stop positions. pGST893-910, pGST893-937 and pGST893-960 were produced from these constructs by cleavage with NcoI and religation of the backbone to remove the 613-892 region.
pDKH975 contains the entire coding region of Drosophila kinesin heavy chain from position 1 to 975 as a non-fusion protein. It was derived from pDKH392 (22) by cleavage with NsiI and EcoRI and insertion of the NsiI to EcoRI fragment of kinesin. pDKH894, pDKH910, pDKH927, pDKH937, pDKH945, and pDKH960 code for amino acids from position 1 to the indicated stop position. They were produced by insertion into pDKH975 of either the NcoI/EcoRI fragment from pGST893-910, pGST893-937, and pGST893-960 or of new polymerase chain reaction fragments. In the final constructs, the only coding region that is polymerase chain reaction-derived is between the NcoI site at Met 893 and the stop codon and this was confirmed by DNA sequencing. pDKH614 was derived from pDKH894 by cleavage with NcoI and religation of the backbone. In pDKH614, Ala 614 is converted to Glu. pDKH560 was derived from pDKH614 by cleavage with NsiI and NcoI and insertion of the NsiI/AflIII fragment of kinesin. In pDKH560, Ser 560 is converted to Glu. pDK613-975 and pDK893-975 contain the indicated kinesin regions as non-fusion proteins in pET-21d (Novagen).
Protein Expression and Purification-Kinesin head domains were prepared as described previously (3,14). The long heavy chain constructs, DKH894-DKH975, were expressed and processed through the P-11 chromatography step as described for the head domain constructs except that induction with isopropyl-1-␤-D-galactopyranoside was performed at Յ25°C. MgATP was included at Ն0.1 mM in all solutions. Solid ammonium sulfate was added to the pooled P-11 fractions to 36% and centrifuged to collect the kinesin in the pellet. Both bovine kinesin and long heavy chain constructs of Drosophila kinesin precipitate at this low concentration of ammonium sulfate. Shorter Drosophila kine-sin constructs, as well as most Escherichia coli proteins, remain soluble and this step provides considerable further purification. After dissolving the ammonium sulfate pellet in minimal A25 with 1 M NaCl, it was subjected to gel filtration on Bio-Gel A-5m in 1 M NaCl. The peak fractions were dialyzed versus KA buffer (14), loaded on a Bio-Gel DEAE column, and eluted with 150 mM NaCl in KA buffer. The pooled fractions were dialyzed against A25 buffer with 50 mM KCl in 50% glycerol with 2 mM dithiothreitol and stored in small aliquots at Ϫ80°C.
Cells containing induced GSTase fusion proteins were lyzed and clarified essentially as described for DKH340 except that the buffer was 50 mM Bicine/NaOH, pH 8, with 10 mM EDTA. The clarified extract was applied to a glutathione-Sepharose column, washed with 20 mM Bicine/ NaOH, pH 8, containing 1000 mM NaCl and 2 mM EDTA. The protein was eluted with 20 mM Bicine/NaOH, pH 8, containing 50 mM NaCl, 2 mM EDTA, and 5 mM glutathione. Control GSTase was obtained by expression of pGEX-2T (Pharmacia).
Inclusion Bodies-When the long non-fusion constructs (DKH894-DKH975) are induced at 37°C, the pool of soluble construct is vanishingly small and essentially all of the expressed protein is present as inclusion bodies. For isolation of inclusion bodies, cells that had been induced at 37°C were lysed and centrifuged as described above. The pellet was washed by resuspension and centrifugation at 10,000 ϫ g successively in MgM buffer (14) with 0.5 M NaCl, 0.5 M NaCl and 1% Triton (twice), and finally with 0.05 M NaCl (twice). The washed pellets were solubilized by heating to 100°C in SDS-PAGE sample buffer and used as size markers for unproteolyzed constructs.
Proteolysis-Constructs containing the extreme COOH terminus are highly susceptible to proteolysis during isolation and the particular preparations used here have some proteolytic cleavage as indicated in Fig. 1. For the GSTase fusions, the largest species is the size expected for a full-length construct in each case (Fig. 1A). The proteolytic cleavages are not likely to have occurred in the GSTase domain because GSTase itself is resistant to proteolysis and because all of the proteolytic fragments continue to bind tightly to glutathione-Sepharose. Also, in preparations that have been more extensively proteolyzed, the ladder of fragments shifts down toward the mass for GSTase, but not below, as expected for proteolytic removal of the kinesin-derived COOH-terminal region. The longer fragment (PF-I) results from cleavage at Յ960 as this fragment of GST893-975 comigrates slightly faster than uncleaved GST893-960. The short fragment (PF-II) results from cleavage at ϳ940 as this fragment of GST893-975 and GST893-960 is slightly larger than GST893-937. In more extensively proteolyzed preparations such as that used in Fig. 5, the fragment terminating at ϳ940 is the major species. Constructs lacking the 940 -975 region are not significantly proteolyzed.
Full-length DKH975 is also partially proteolyzed to species corresponding to cleavage at the PF-I and PF-II sites as indicated in Fig. 1B. Proteins obtained from inclusion bodies were used for size standards as they have likely precipitated before undergoing proteolysis. Although there is some unproteolyzed protein, the major band in a typical preparation of DKH975, as indicated in Fig. 1B, corresponds to cleavage at the PF-I site. There is also a significant band corresponding to cleavage at the PF-II site. Some preparations of DKH975 are more highly proteolyzed with the major band corresponding to PF-II and these preparations have not been included in the work presented here. Preparations of DKH960 show much less proteolysis and appear predominately as a single band of the expected size. Although it cannot be excluded that some proteolysis of DKH960 has occurred to produce the PF-I product, which is approximately the same size as DKH960, there is no major accumulation of the PF-II product as observed with DKH975. All of the shorter constructs (DKH894-DKH945) migrated as single bands at the same position as the corresponding proteins isolated from inclusion bodies.

Salt Dependence of Sedimentation of Long Heavy Chain Constructs-
The sedimentation coefficients of the full-length Drosophila kinesin heavy chain construct, DKH975, and truncated DKH960 are 5.4 S in A25 buffer with 25 mM KCl and 1 M NaCl, but shift to 6.7-7.0 S at low salt concentration with a midpoint for the transition of ϳ0.5 M NaCl ( Fig. 2A). The behavior of these long Drosophila constructs is essentially identical to that of the bovine kinesin heavy chain dimer (11). The elution position of DKH975 and DKH960 during gel filtration on Bio-Gel A-5m at high and low salt (not shown), as well, is essentially identical to that of the bovine heavy chain dimer (11) with an increase in Stokes radius at high salt. The similarity of the sedimentation and diffusion properties to those of bovine dimer indicates that DKH975 and DKH960 are also dimers that undergo a transition from a compact conformation at low salt to an extended, highly asymmetric, conformation at high salt concentration. A possible model for this transition is presented in Fig. 2B as a framework for discussion.
The sedimentation coefficient of DKH894, DKH910, and DKH927 remain Յ5.6 S ( Fig. 2A), even at low salt and thus removal of the COOH-terminal 48 amino acids inhibits formation of the compact conformer. Deletions of intermediate length, DKH945 and DKH937, still form the compact conformation at low salt, but the midpoint for the transition is shifted to lower salt concentration. The midpoints for DKH945 and DKH937 are similar, but DKH937 is shifted slightly to lower concentrations.
In the preparations of DKH975 that were used, the major component was a proteolysis fragment, PF-I, of approximately the same size as DKH960 (see "Materials and Methods"). Thus the observed sedimentation behavior of DKH975 preparations is dominated by the PF-I species and does not necessarily represent the behavior of true full-length DKH975. At 300 mM NaCl, DKH960 should be in the compact conformation, but DKH945 and shorter should be mainly extended. Chromatography of DKH975 preparations at 300 mM NaCl indicates that full-length DKH975 comigrates with the PF-I species at the position expected for the compact conformer, and thus it is likely that full-length DKH975 does behave similarly to DKH960. The smaller PF-II species elutes earlier at the position expected for the extended conformer, consistent with its length of ϳ940.
Localization of the Site of Interaction in the NH 2 -terminal Region-The inability of DKH894-DKH927 to from the compact conformer at low ionic strength suggests that formation of the compact conformer results from interaction of the COOHterminal region with regions closer to the head domain. A GSTase fusion protein containing the kinesin heavy chain from amino acids 864 -975, designated GST864 -975, was used to determine if it could interact with head domains in a salt-dependent manner. GST864 -975 was loaded onto a column containing glutathione-Sepharose and head domains were passed down the column in low salt buffer to determine if binding could occur. As indicated in Fig. 3A, DKH365, DKH381, and DKH392 were retained by the column, whereas DKH340 and DKH357 passed through. Subsequent elution of the columns with high salt buffer released the bound DKH365, DKH381, and DKH392.
Localization of the Site of Interaction in the COOH-terminal Region-In order to further localize the site of interaction in the tail region, GSTase fusions containing smaller parts of the COOH-terminal region were tested for interaction with DKH392. As indicated in Fig. 3B, the region between 893 and 975 was sufficient for salt-dependent interaction, but the region between 864 and 893 was not sufficient for tight interaction. Column binding tests were also performed with GSTase fusions containing kinesin tail sequences from 893 to 937 and 893 to 960 as indicated in Fig. 4. Under these conditions, none of the three head domains binds to a GSTase control, GST893-910 (not shown), or GST893-937. Extension of the tail domain Only the region corresponding to the full-length species is shown. The long constructs differ only slightly in relative size and tall gels (Hoefer SE280) with 6% acrylamide were required for even partial separation. Also the time for electrophoresis was increased past the time required to elute the tracking dye so that the long constructs migrated closer to the bottom of the gel. The proteins were loaded at a low level that was just detectable by sensitive staining with colloidal Coomassie G-250 (31). Even moderate increases in loading produced bands that were too wide for resolution and migrated faster as illustrated by the two loadings of DKH960 at the right of the gel. The specific interaction that is indicated for the compact 7 S conformation has part of the COOH-terminal non-helical domain interacting with the motor domain and the adjacent potentially helical region of the tail interacting with the coiled-coil neck. This detailed interpretation is speculative, but is consistent with the localization of the regions that are required for formation of the compact conformer as developed under "Discussion." to 960 (GST893-960) does produce salt-dependent binding of DKH365 and DKH405, but not DKH346. However, the binding of DKH405 is stronger than that of DKH365, as DKH365 is rapidly eluted by 150 mM salt, whereas higher concentrations of salt are required to rapidly elute DKH405.
Comigration of Head and Tail Regions during Centrifugation at Low Salt Concentration-Sucrose density gradient centrifugation of DKH392 alone in the absence of added salt (Fig. 5A) gives a sedimentation coefficient of 5.9 S compared with 5.2 S observed previously in 25 mM KCl (22). The self-aggregation of DKH392 is more extensive at lower salt (3) and this shift in s 20,w value likely results from some reversible aggregation under these conditions. Sedimentation of GST893-975 alone gives a sedimentation coefficient of 4.4 S that is consistent with a dimeric protein of its subunit molecular weight and is similar to the value of 4.4 S obtained for the parent dimeric GSTase (not shown). The preparation of GST893-975 used here is more extensively proteolyzed than the sample of GST893-975 in Fig.  1A. The proteolytic fragment PF-II is the most abundant species and only a minor amount of PF-I is present, but considerable full-length GST893-975 is still present.
When GST893-975 is mixed with an excess of DKH392 before sedimentation, full-length GST893-975 comigrates with DKH392 at a higher sedimentation coefficient of ϳ7 S (Fig. 4C). Most of the major proteolytic fragment (PF-II) of GST893-975 does not migrate more rapidly in the presence of DKH392, but some of the proteolytic fragment does shift as indicated by streaking into fractions further down the gradient. Weaker interaction of PF-II with DKH392 is consistent with the size of the COOH-terminal truncation to ϳ940 (Fig. 1B) and the shift in the midpoint for DKH937 and DKH945 that is observed in Fig. 2. The parent GSTase protein shows no shift in sedimentation coefficient when centrifuged under these conditions in the presence of excess DKH392 (not shown).
The interaction observed between GST893-975 and DKH392 is likely to be weak and reversible under these conditions and is only observed in Fig. 5C 4. Binding of heads to GST893-937 and GST893-960. Columns of glutathione-Sepharose (1 ml bed volume) were loaded with GSTase fusion proteins (2.5 mg). A mixture of DKH346, DKH365, and DKH405 was loaded on the columns (1 ml of 0.2 mg/ml each) in A25 buffer with 10 mM NaCl and washed with a total of 1.5 ml of buffer with 10 mM NaCl. The combined eluents were pooled as the run-through fraction (RT). The column was then washed sequentially with increasing salt concentrations of 10, 150, and 1000 mM NaCl (2.5 ml total for each). Fractions were analyzed by SDS-PAGE. Loading of each fraction was adjusted so that the protein concentration would equal that of the preload if the recovery was 100% in that fraction. salt are sufficient to inhibit association. GST893-975 also fails to comigrate with DKH392 in the absence of added salt when the concentration of DKH392 is reduced (not shown), indicative of a weak interaction that requires high levels of DKH392 to maintain a significant fraction of the GST893-975 as the complex.
ATPase Rates-The initial rates for ATP hydrolysis was determined for compact DKH975 and extended DKH894 in A25 buffer with 50 mM KCl (Fig. 6). The k cat and K 0.5(MT) values were 35 and 3.8 s Ϫ1 and 1.9 and 0.27 M, respectively, for DKH894 and DKH975. DISCUSSION The heavy chain dimer of bovine kinesin has previously been shown to undergo a transition from an extended conformer at high salt concentration to a more compact conformer at low salt concentration (11). The bovine dimer, however, likely results from proteolysis of the light chains during isolation of the native bovine heterotetramer (11,23) and the heavy chain dimer could potentially still contain small light chain domains that are not easily detected by SDS-PAGE. The ability of the Drosophila heavy chain dimer obtained by expression in E. coli to form the compact conformer at low ionic strength indicates that the heavy chain itself possesses the capability to form a compact conformation without the involvement of light chains or eukaryotic post translational modification.
Two types of mechanisms could produce the compact conformer. One possibility is that the stalk is bent back on itself at the likely hinge at position ϳ600 and that the interaction of the front and back half of the stalk along their lengths provides the major stabilization of the compact conformer. In the other type of model, as indicated in Fig. 2B, formation of the compact conformer is driven by a favorable interaction between a domain in the NH 2 -terminal head region and a domain in the COOH-terminal tail region that can occur because there is sufficient flexibility in the stalk for the two ends of kinesin to come into contact. The break in the predicted coiled-coil of the stalk around position 600 is a candidate hinge for either model. The inability of DKH927 to form the compact conformer at low salt ( Fig. 2A) indicates that the COOH-terminal region between 927 and 960 is required and, conversely, that the stalk domain itself cannot produce a stable compact conformation. This suggests that the compact conformer is not produced by interactions within the stalk, but rather by interaction between specific segments in the NH 2 -and COOH-terminal regions as indicated in Fig. 2B. Such a direct interaction is supported by the observation that separately expressed tails and heads interact in a salt-dependent manner (Figs. 3-5). This does not totally preclude involvement of antiparallel interactions between the front and back half of the stalk domain in the  (36)). The indicated sequence is that for Drosophila with a "*," indicating absolute conservation in all 5 species and a ":," indicating only conservative changes. The strong conservation observed at the beginning of this region extends into the area on the NH 2 -terminal side of position 881, but there is no strong conservation on the COOH-terminal side of position 950. Positively and negatively charged residues are indicated by ϩ and Ϫ, respectively, on the line above the sequence. The heptad repeat positions a and d are also indicated above the sequence for heptad frames 7N-3 and 7N. B, prediction of coiled-coil tendency for Drosophila kinesin heavy chain determined by the program MacStripe2.0 ((37), http://www.york.ac.uk/depts/biol/units/coils/ coilcoil.html) implementing the COILS2 algorithm of Lupas (38). Solid line is for a window size of 28 residues with a shift in predicted heptad frame from 7N to 7N-3 at position 912. The dashed line is for the more stringent case of a window size of 21 residues and it emphasizes the weaker strength of the prediction for the region around Pro 883 and for the region between 912 and 930. The arrow indicates the end of the Drosophila sequence at 975. compact conformer, but does indicate that any such interactions within the stalk are not the major energetic driving force for formation of the compact conformer.
Analysis of the tendency for coiled-coil formation indicates (Fig. 7) that the region between 850 and 910 is likely to be coiled-coil in heptad frame 7N with a weak area around Pro 883 . The region between 910 and 930 has a weaker predicted tendency for coiled-coil formation in heptad frame 7N-3. Detailed inspection (Fig. 7) indicates that there is a reasonable heptad repeat in both heptad frames 7N-3 and 7N, that could facilitate a 4-helix or antiparallel coiled-coil interaction between the ϳ883 and 930 region of the tail and the helical region of the neck.
The ability of DKH937 to form the compact conformer at low salt ( Fig. 2A) and the ability of GST893-960 to bind to DKH365 (Fig. 4) indicate that the tail region from 893-937 and the head region of 1-365 are sufficient for tight interaction, although interaction of additional regions may provide further stabilization. The NH 2 -terminal half of the coiled-coil neck is likely to play a critical role as even partial deletion (DKH357 versus DKH365) strongly inhibits interaction with the tail domains. These regions of both the neck and tail are highly charged (16 charged for 340 -365 and 15 charged for 912-936) and this could account for the high salt dependence of the interaction. The whole region between 883-936, however, has an abundance of charged groups and is it not possible to definitely predict the alignment of the tail and neck regions in the complex. Although the NH 2 -terminal half of the neck is likely to be important for the binding of the tail domain, several constructs containing the neck and adjacent regions, in the absence of complete head domains, have been tested for binding to the tail domain and none have been found to bind strongly. 2 It is thus likely that strong binding to the tail domain requires interaction over a more extended region than just the NH 2 -terminal half of the coiled-coil neck. Although the ability of DKH937 to form a compact conformation at low salt indicates that the region beyond 937 is not absolutely required, the shift in the midpoint for the transition with longer constructs indicates that the region beyond 937 can stabilize the interaction. If the region of ϳ345-365 interacts with ϳ910 -930 in an antiparallel configuration as in Fig. 2B, then the region of the tail beyond 930 would be juxtaposed to the head domain and in a position for potential interaction. Conventional kinesins show strong sequence conservation in this region up to position 950 (Fig. 7) and the conserved region (940 -950) surrounding the IAK at positions 942-944 is thus a candidate for this additional area of interaction with the head. This antiparallel alignment would also juxtapose the COOH-terminal half of the neck with the region between ϳ883 and 910 and interaction of these regions could be responsible for the stronger interaction of DKH405 versus DKH365 with tails (Fig. 4).
The highly charged NH 2 -terminal half of the coiled-coil neck has a gap in the hydrophobic heptad repeat and this region was postulated to be capable of reversibly uncoiling, while the dimer remained anchored by the stronger interactions in the better heptad repeats of the COOH-terminal half of the neck (24). Recent work with peptides from the neck region has established that such uncoiling of the NH 2 -terminal half of the neck can occur (4,5). Although the exact oligomeric state of the complex between the tail and head/neck is not known, the interaction between monomeric DKH365 and GST893-975 suggests that formation of a 4-helix bundle is not necessary, at least for the NH 2 -terminal half of the neck. An alternative possibility is that strand displacement may occur with uncoil-ing of the NH 2 -terminal half of the neck to allow formation of a pair of antiparallel coiled-coils with the region from ϳ910 to 930 of the tail. The relatively poor prediction for a parallel coiled-coil interaction of this region of the tail with itself may be reflective of an actual role in antiparallel interaction with the neck.
Fungal kinesins are relatives of conventional kinesin (25) that have lost the highly charged nature of both the neck and tail regions that are homologous to those implicated above in formation of the compact conformation. Either these regions of fungal kinesins do not interact, or both regions have co-evolved to maintain an interaction that is not as highly dependent on charge. Interestingly, the IAK region is, however, conserved between conventional and fungal kinesins.
Entropic effects are likely to be the major cause of the weaker apparent interaction of domains on separate constructs relative to domains that are covalently linked. For example, DKH975 remains predominantly in the compact conformation at 0.3 M salt, whereas interaction of GST893-975 and DKH392 during cosedimentation is weak and readily reversible even in the absence of added salt. When the domains are on separate constructs, they are free to diffuse away from each other following dissociation, whereas the dissociated domains of DKH975 are restrained to remain near each other because they are covalently linked through the stalk. The failure of DKH405 or DKH365 to bind to GST893-937 in the column binding assay (Fig. 4) while DKH937 can still form the compact conformer, likely also reflects the more stringent requirements for association of fragments that are not covalently linked. Additive contributions from the entire extended binding region including the IAK region that is present in GST893-960 are apparently required for binding to be tight enough to be observed under these more stringent assay conditions. Entropic considerations also predict that the column binding experiments (Figs. 3 and 4) can better detect weak interactions compared with cosedimentation experiments (Fig. 5). The sedimentation velocity of an equilibrating mixture is directly related to the fraction associated; but in the column binding assay, head domains must interact sequentially with many bound GST-tail fusions before they can be eluted.
The analysis of ATPase rates of heavy chain constructs that are close to full-length is complicated by the tendency for formation of asters of MTs in the presence of ATP as originally reported by Urrutia et al. (26). The heavy chain constructs used here can also produce such asters during the course of an ATPase assay at high concentration as verified by electron microscopy. 3 The ATPase results in Fig. 6 were obtained as initial rates at low kinesin concentration (30 nM) in order to minimize the contribution from kinesin molecules in asters. Extended DKH894 has a high k cat at saturating MT concentration that is similar to that of fully active shorter constructs (3). The k cat of DKH975, however, is 10-fold lower, indicating inhibition in the compact conformer. The true rate of DKH975 in the compact conformer may be even lower than the observed rate of 3.4 s Ϫ1 as the K 0.5(MT) of 0.27 M is unexpectedly low and part of the observed activity likely represents catalysis by the contaminating PF-II species which has a high ATPase rate and a low K 0.5(MT) value. 2 With bovine kinesin, inhibition of MTstimulated ATPase activity results from inhibition of the ability of MTs to stimulate ADP release from the compact conformer that is present at low ionic strength (15). This inhibition could result directly from steric blockage by the tail of the MT interaction site on the head or indirectly by a conformational change in the head or neck that is induced by binding of the tail 2 M. Stock and D. Hackney, unpublished observations. 3 D. Hackney, unpublished observations. domains. If the site on the tail domain to which the head domain bound was overlapping with the site on the tail to which the vesicle receptor bound, then competition between these two binding modes would provide a natural basis for regulation by mass action (11) as illustrated by the model of Fig. 8. Even if the receptor-binding site is on the light chains, the binding of tails to heads versus receptors could be still be competitive as long as steric constraints prevented both interactions from occurring at the same time. When free in solution, kinesin would be in the compact conformer and be inhibited, while binding of the membrane receptor to the tail would release the heads in their active form. If the binding of the tail to the receptor was stronger than binding of the tail to the heads and if kinesin was in excess, then all the receptors would be occupied with active kinesin and any excess kinesin would be in the compact conformer and thus inhibited. Such binding-induced activation of kinesin has been reported with nonphysiolgical surfaces (27)(28)(29) and is likely to also be responsible for the high motor activity of adsorbed kinesin in motility assays, but activation by physiological receptors remains to be demonstrated. The occurrence of receptor-induced activation by induction of the extended conformer would not preclude additional or alternative regulation by other means such as phosphorylation (30).