Interaction of Kinesin Motor Domains with α- and β-Tubulin Subunits at a Tau-independent Binding Site

Interaction of rat kinesin and Drosophila nonclaret disjunctional motor domains with tubulin was studied by a blot overlay assay. Either plus-end or minus-end-directed motor domain binds at the same extent to both α- and β-tubulin subunits, suggesting that kinesin binding is an intrinsic property of each tubulin subunit and that motor directionality cannot be related to a preferential interaction with a given tubulin subunit. Binding features of dimeric versus monomeric rat kinesin heads suggest that dimerization could drive conformational changes to enhance binding to tubulin. Competition experiments have indicated that kinesin interacts with tubulin at a Tau-independent binding site. Complementary experiments have shown that kinesin does not interact with the same efficiency with the different tubulin isoforms. Masking the polyglutamyl chains with a specific monoclonal antibody leads to a complete inhibition of kinesin binding. These results are consistent with a model in which polyglutamylation of tubulin regulates kinesin binding through progressive conformational changes of the whole carboxyl-terminal domain of tubulin as a function of the polyglutamyl chain length, thus modulating the affinity of tubulin for kinesin and Tau as well. These results indicate that microtubules, through tubulin polymorphism, do have the ability to control microtubule-associated protein binding.

Interaction of rat kinesin and Drosophila nonclaret disjunctional motor domains with tubulin was studied by a blot overlay assay. Either plus-end or minus-enddirected motor domain binds at the same extent to both ␣and ␤-tubulin subunits, suggesting that kinesin binding is an intrinsic property of each tubulin subunit and that motor directionality cannot be related to a preferential interaction with a given tubulin subunit. Binding features of dimeric versus monomeric rat kinesin heads suggest that dimerization could drive conformational changes to enhance binding to tubulin. Competition experiments have indicated that kinesin interacts with tubulin at a Tau-independent binding site. Complementary experiments have shown that kinesin does not interact with the same efficiency with the different tubulin isoforms. Masking the polyglutamyl chains with a specific monoclonal antibody leads to a complete inhibition of kinesin binding. These results are consistent with a model in which polyglutamylation of tubulin regulates kinesin binding through progressive conformational changes of the whole carboxyl-terminal domain of tubulin as a function of the polyglutamyl chain length, thus modulating the affinity of tubulin for kinesin and Tau as well. These results indicate that microtubules, through tubulin polymorphism, do have the ability to control microtubule-associated protein binding.
The various structures and functions of microtubule (MT) 1 networks are mediated by many different structural and motor microtubule-associated proteins (MAPs) (for review see Refs. [1][2][3][4]. To coordinate these functions, the major problem in cells is to regulate not only the expression of the different MAPs but, more precisely, their differential interactions with MTs. Regulation of MAP binding through phosphorylation has been extensively documented (5)(6)(7), although little information is as yet available concerning motor proteins (8 -11). However, several reports have clearly shown that part of this regulation had also to be supported by the MTs themselves. For instance, it has been shown that in lobster axons, only a subset of MTs was competent for vesicle transport (12), suggesting that the MT surface available for interaction with motors was not homogeneous along all of the axonal MTs. The same relationship between distinct MT subsets and specific binding of kinesinlike proteins has also been observed in the mitotic spindle apparatus (13)(14)(15) and in the Chlamydomonas flagellum where Klp1 was shown to interact selectively with the central pair C2 MT (16). The chemical diversity of the MT surface could provide such specific positional and binding information in cells for many MT-interacting proteins. The large genetic and posttranslational heterogeneity of the ␣and ␤-tubulin subunits (for review see Refs. [17][18][19] can represent the molecular bases for such information. Indeed, we recently showed that polyglutamylation, the oligomeric post-translational modification of tubulin (20 -23), was involved in the regulation of Tau and MAP2 interaction with tubulin through a modulation of affinity as a function of the polyglutamyl chain length (24). Because, on one hand, the extension of the polyglutamyl chain from one to six units is thought to induce conformational changes in the carboxyl-terminal domain of both tubulin subunits, leading to the observed modulation of affinity for structural MAPs, and, on the other hand, all of the known MT-based motors (kinesin, kinesin-related proteins, and cytoplasmic and flagellar dyneins) have been reported to bind also in this carboxyl-terminal domain (25,26), it follows that their binding could also be under the same control as the structural MAPs.
We have analyzed the binding of different kinesin head motor domains (conventional rat kinesin (RK) and the Drosophila minus-end-directed kinesin-related protein (ncd, for nonclaret disjunctional)) to tubulin and investigated the influence of the degree of polyglutamylation of the various ␣and ␤-tubulin isoforms on motor binding. In this paper, we report that the two types of motors bind in vitro to both ␣and ␤-tubulin subunits and that this interaction appears to be controlled by polyglutamylation in the same manner as for Tau or MAP2. In addition, competition experiments carried out with several monoclonal antibodies and with kinesin heads and heat-stable MAPs indicate that the respective binding sites of structural and motor MAPs are independent from each other, although located in a sufficiently close vicinity on the tubulin molecule to undergo the same conformational control by polyglutamylation. produced in our laboratory (27).
Recombinant Proteins-The recombinant motor domain constructs of rat kinesin (RK375 and RK329) and Drosophila ncd (HA-N333) were kindly provided by E.-M. Mandelkow and F. Kozielski (Max-Planck Unit for Structural Molecular Biology, Hamburg, Germany). RK375 and RK329 correspond to amino acids 1-379 and 1-329 of the rat kinesin sequence, respectively. HA-N333 corresponds to amino acids 333-700 of the Drosophila claret segregational protein sequence.
Protein Purification-Tubulin was prepared from a 100,000 ϫ g supernatant of adult mouse brain by two cycles of assembly disassembly and further purified by phosphocellulose column chromatography (P11, Whatman) in MEM buffer (50 mM MES, pH 6.8, 2 mM EGTA, 2 mM MgCl 2 ) containing 1 mM Mg-GTP and a mixture of protease inhibitors (10 g⅐ml Ϫ1 aprotinin, 10 g⅐ml Ϫ1 leupeptin, 1 mM phenylmethylsulfonyl fluoride). Heat-stable MAPs (essentially Tau and MAP2) were routinely prepared from once cycled MTs, which were depolymerized at 4°C, cleared by centrifugation, brought to 0.5 M NaCl, and boiled for 5 min.
Subtilisin Digestion-Purified brain tubulin (1 mg⅐ml Ϫ1 ) was digested in MEM buffer with 1.25 or 5 g⅐ml Ϫ1 subtilisin (Carlsberg, Sigma; 7-15 units⅐mg Ϫ1 ) for 45 min at 37°C, leading to a limited or a total digestion of tubulin, respectively. Reaction was stopped by the addition of 1 mM phenylmethylsulfonyl fluoride, and aliquots were processed by one-dimensional PAGE.
One-dimensional and Two-dimensional PAGE-Protein separation by one-dimensional or two-dimensional PAGE and immunodetection was carried out as described previously (24,27) except that ampholytes 5-6 (Serva) were used in the isoelectric focusing dimension. Antibody binding was revealed by the ECL system (Amersham Corp.), and autoradiographs were quantified by scanning using a Vernon integrating densitometer.
Blot Overlay Assay-Binding of kinesin heads to tubulin separated by one-dimensional or two-dimensional PAGE and transferred onto nitrocellulose membranes (Hybond C, Amersham Corp.) was performed essentially as described previously (24). Briefly, after monitoring of blotted tubulin by Ponceau red staining, strips corresponding to lanes of one-dimensional gels (or rectangles corresponding to tubulin regions of two-dimensional gels) were cut from the blots and placed into the grooves of a hand-made, Plexiglas incubation device adjusted to the size of the strips. Nitrocellulose pieces were blocked overnight in overlay buffer (OV ϭ MEM buffer containing 1 mM dithiothreitol, 0.1% (v/v) Tween 20, and 0.1% (w/v) gelatin), incubated for 1 h at 20°C with the overlaying protein fraction, and then washed 5 ϫ 10 min. Protein interactions were then stabilized with formaldehyde (24) before the blots were equilibrated in TBS-T (20 mM Tris, pH 7.5, 136.8 mM NaCl, 0.1% Tween 20) and processed for immunodetection.
Kinesin Binding to Coated Tubulin Heterodimers-Phosphocellulosepurified tubulin (1 g in 100 l/well) was coated onto enzyme-linked immunosorbent assay multiwell microtiter plates (Nunc) in MEM buffer at 37°C. We checked by immunoassay that a constant amount of 500 ng was actually coated in each well according to the manufacturer. Wells were equilibrated and blocked overnight in OV then incubated with various concentrations of kinesin heads in 100 l for 60 min at 25°C. Wells were rinsed with OV (5 ϫ 5 min), and kinesin-tubulin interactions were stabilized with formaldehyde (24). After equilibration in TBS-T, tubulin-bound kinesin was detected with anti-HIPYR 2 h at 37°C and a ␤-galactosidase-conjugated secondary anti-rabbit IgG (Biosys, France) 1 h at 37°C, using o-nitrophenyl-␤-D-galactoside (Sigma) as chromogen. Reactions were quantified automatically at 490 nm.
Microtubule Sedimentation Assay-MTs were polymerized from a whole brain supernatant in the presence of 20 M Taxol and washed with 0.5 M NaCl to remove bound MAPs. MTs were pelleted, resuspended in MEM buffer, and used in micro-assays of kinesin head binding. MTs (60 pmol) were mixed with kinesin heads (50 pmol) in the absence or the presence of different nucleotides and salt in a final volume of 25 l. After centrifugation (10 min at 25 p.s.i. in the A100 Airfuge rotor, Beckman), the whole pellet and supernatant proteins were analyzed by one-dimensional PAGE.
Size Exclusion HPLC-RK329 and RK375 were suspended in MEM buffer (with or without 500 mM NaCl), cleared 10 min at 12,000 ϫ g, and analyzed (separately or mixed) by gel filtration HPLC (Waters) on a TSK G3000 column (7.8 ϫ 300 mm) equilibrated in MEM buffer and calibrated with standard proteins. Protein elution (1 ml⅐min Ϫ1 ) was monitored at 280 nm. Proteins from each collected fraction were analyzed by one-dimensional PAGE, detected by Coomassie blue staining and immunoprobed with anti-HIPYR antibodies.

Interaction of Rat Kinesin Heads (RK375) with ␣and ␤-Tu-
bulin Subunits-Increasing concentrations of RK head were overlaid onto a constant amount of tubulin separated by onedimensional PAGE and immobilized onto nitrocellulose. Fig.  1A shows that RK head binding, detected on both ␣and ␤-tubulin subunits with the specific anti-kinesin peptide (HIPYR) antibody, increased proportionally with RK concentration up to a saturation value. RK375 binding to both ␣and ␤-tubulin was FIG. 1. RK375 and ncd head domain binding to ␣and ␤-tubulin subunits. A, phosphocellulose-purified tubulin (2 g) was separated by one-dimensional PAGE, blotted onto nitrocellulose, and overlaid with increasing concentrations of RK375 (from 0 to 10 g⅐ml Ϫ1 ) in 1 ml of OV buffer. Kinesin bound to separated ␣and ␤-tubulin subunits was detected with anti-HIPYR antibody, and the resulting signals (upper panel) were quantified by densitometric scanning (expressed in arbitrary units, lower panel). B, to check the specificity of RK375 binding to tubulin, 2 g⅐ml Ϫ1 of RK375 (the concentration used in the following experiments) were overlaid onto whole supernatant proteins from brain (S o ) separated by one-dimensional PAGE. In these conditions, kinesin binding, revealed by anti-HIPYR, was limited to tubulin (dashes, right panel). In addition, when increasing concentrations of competitive free tubulin were added to the overlay solution (containing 2 g⅐ml Ϫ1 RK375), a concomitant inhibition of RK binding to both tubulin subunits was observed (left panel). C, same experiment as in A except that the motor head of the minus-end-directed kinesin-related protein ncd was substituted to RK375. progressively inhibited when increasing concentrations of free tubulin were added to the overlay solution ( Fig. 1B, left panel). A selective binding to both tubulin subunits was also observed when RK375 was overlaid onto whole supernatant proteins from brain (Fig. 1B, right panel). These results indicate that the tubulin-kinesin interactions occurring during the blot overlay procedure were specific. Likewise, the reverse, minus-enddirected motor head of ncd also bound to both ␣and ␤-tubulin subunits (Fig. 1C), suggesting that the directionality of a kinesin motor protein along a MT is not determined by its selective interaction with a given ␣or ␤-tubulin subunit.
To assess the in vitro binding assay used in the preceding experiments, RK head binding was tested in the presence of nucleotides. Kinesin binding was progressively inhibited by increasing concentrations of GTP or of its hydrolyzable form, Mg-GTP ( Fig. 2A). The effect of nucleotides on RK head binding was also tested using heterodimeric tubulin coated onto multiwell enzyme-linked immunosorbent assay plates and taxolstabilized MTs by a sedimentation assay. Fig. 2 (B and C) shows that in both cases the binding of RK heads was weakened by GTP, indicating that monomeric tubulin in the blot overlay assay behaves as heterodimeric or MT-assembled tubulin with respect to kinesin head binding.
Formation of RK head-tubulin complexes was also readily prevented by raising the salt concentration in the incubation medium ( Fig. 2, C and D). No interaction could be detected beyond 75 mM NaCl. Formation of kinesin-tubulin complexes was still effective, however, in the presence of 1 or 2 M urea (data not shown). These results indicate that kinesin-tubulin interactions occur mainly through electrostatic bonds.
It is interesting to note that when MES was substituted by PIPES, the binding efficiency of kinesin to tubulin dropped severalfold. At 100 mM PIPES, a widely used buffer molarity, however, binding was faintly detectable, at levels 5-10 times lower than with 50 mM MES (data not shown).
In all of these experiments where nucleotides, salt, chaotropic agent, or buffers were tested, the variation in the binding of RK heads was always similar for ␣and ␤-tubulin subunits, indicating that the same basic binding features for kinesin were shared by both subunits, which in turn suggests the involvement of related sequences in orthologous domains in both subunits.
Binding of a Truncated Rat Kinesin Motor Domain-When a truncated derivative of RK head, RK329, was overlaid onto immobilized tubulin (Fig. 3A), the resulting binding was estimated to be 5-10-fold lower than that obtained with the whole RK375 head. Although less efficiently, RK329 bound to both tubulin subunits at similar levels. The same difference in tubulin binding was observed when increasing concentrations of RK375 and RK329 were added to microwells coated with heterodimeric tubulin (Fig. 3B). These results suggest that amino acids 330 -375 of RK are important for enhancing the interaction with tubulin.
RK329 and RK375 were then analyzed by gel filtration chromatography to check the assembly state of each motor. Fig. 3, (C and D) shows that RK329 (calculated mass, 36.8 kDa) and RK375 (calculated mass, 42.7 kDa) each eluted as a single peak (retention times, 9.3 and 7.8 min, respectively) corresponding to globular polypeptides of M r 37,200 and 120,000, respectively. Given the steric occupancy of a nonspherical, two-headed protein, these values are compatible with a dimerized state of RK375. When the column was run in 500 mM NaCl, RK375 eluted at the same position (Fig. 3C, dotted line). The salt resistance of RK375 dimers suggests that ␣-helical coiled-coil interactions could be involved between amino acids 330 -375 of kinesin sequences (see Ref. 28 and "Discussion").
Regulation of Kinesin-Tubulin Interaction by Polyglutamylation-Because polyglutamylation was previously shown to modulate the affinity of tubulin for the structural MAPs Tau and MAP2 (24), we researched whether it could also be involved in the control of the binding of kinesin. Brain tubulin was separated by resolutive two-dimensional PAGE to physically space the different isoforms carrying an increasing number of glutamyl units (22,24) and transferred onto nitrocellulose. The different isotubulins were then probed for their ability to interact with kinesin. By comparison with the whole set of tubulin isoforms available on the nitrocellulose membrane (Fig. 4A), as detected with general anti-tubulin antibodies, Fig. 4B shows that RK375, immunodetected with anti-HIPYR, does not interact with the same efficiency with all of the different forms of tubulin. As previously observed for Tau (24), kinesin interacted principally with moderately modified ␣and ␤-isotubulins. The unmodified, more basic, primary translation products as well as the fully modified, more acidic isoforms only bound weakly, if at all, to kinesin. Because the type of response was similar for ␣and ␤-tubulin and involved post-translationally derived tubulin isoforms, polyglutamylation appeared as the only obvious candidate responsible for this differential kinesin binding; it is the only modification that occurs on both tubulin subunits (Fig. 4C), and its oligomeric structure, which confers one to six negative charges to tubulin (20 -23), can explain the progressive variation of RK binding observed as a function of tubulin acidification. It appears that addition of the first three glutamyl units increases the affinity of tubulin for kinesin, whereas further lengthening of the posttranslational chain decreases it.
Competition with Monoclonal Anti-tubulin Antibodies-To explain the influence of polyglutamylation on kinesin binding, an attempt was made to block the post-translational chains with a monoclonal antibody raised in our laboratory, GT335, which specifically recognizes the polyglutamylated motif on ␣and ␤-tubulin (27). As a control, four other anti-tubulin mono-clonal antibodies were used in parallel to block their respective epitopes, all located also in the carboxyl-terminal domain (Fig.  5F): the anti-tyrosinated ␣-tubulin YL1/2, the anti-␣-tubulin YOL1/34, and two general anti-␣and anti-␤-tubulin, DM1A and DM1B, respectively (29,30). Purified brain tubulin was separated by one-dimensional PAGE, transferred onto nitrocellulose, and progressively saturated by increasing concentrations of each monoclonal antibody. After extensive washing to eliminate unbound antibodies, membrane strips were overlaid with RK375, and tubulin-bound kinesin was detected with the polyclonal anti-HIPYR antibody and a peroxidase-labeled antirabbit IgG, to avoid any cross-reaction with the previously bound mouse IgG (Fig. 5, A-E).
Increasing fixation of GT335 on the polyglutamyl chains progressively and rapidly inhibited kinesin binding on both tubulin subunits (Fig. 5A). A similar binding inhibition by GT335 was previously observed for Tau (24). Fixation of YL1/2 on its carboxyl-terminal tyrosylated epitope of ␣-tubulin did not significantly affect the binding of RK375 (Fig. 5B). Fixation of DM1A and DM1B interfered with kinesin binding on the ␣and the ␤-tubulin subunit, respectively (Fig. 5, C and D), DM1A being much more efficient than DM1B in binding inhibition. In the presence of YOL1/34, a significant inhibition of RK binding to ␣-tubulin was also observed (Fig. 5E). In these experiments, binding of RK to ␤-tubulin in the presence of YL1/2, YOL1/34, or DM1A and to ␣-tubulin in the presence of DM1B served as internal controls.
Binding of Kinesin to Subtilisin-cleaved Tubulin-With antibody blocking experiments, it is difficult to know at first whether the observed inhibition is a direct effect, that is, a coincidence of the ligand binding site and the antibody epitope, or is due indirectly to a steric hindrance of the large IgG   FIG. 3. Binding of RK329 versus RK375. A, 2 g of tubulin separated by one-dimensional PAGE and blotted onto nitrocellulose were overlaid with 2 and 10 g⅐ml Ϫ1 of RK329 or RK375, and the resulting signals (insert) were quantified. B, increasing concentrations (0 -5 g in 100 l) of RK329 or RK375 were added to wells of microtiter plates coated with 500 ng of tubulin. In both types of experiment, kinesin head binding to tubulin was revealed with anti-HIPYR antibody and quantified (expressed in arbitrary units). C, 50 g of each RK329 and RK375 were mixed in MEM buffer and separated by size exclusion chromatography. Protein elution (1 ml⅐min Ϫ1 ) was monitored at 280 nm (solid line). After equilibration of the column with MEM containing 500 mM NaCl, 75 g of RK375 in MEM-NaCl were rechromatographed (dotted line). D, aliquots of fractions collected from the first run were analyzed by one-dimensional PAGE. molecule, which can block the accessibility of adjacent binding sites. For the inhibition observed with GT335, at least, the corresponding epitopes can be easily removed by digesting tubulin with subtilisin (24,31,32). In this case, after blot overlay with RK375 of partially or totally digested tubulin (Fig. 6), it appeared that kinesin still bound to cleaved (␣S and ␤S) tubulin. Densitometric integration of the autoradiograms indicated that (i) RK375 interacted with the same efficiency with intact ␣and ␤-tubulin, (ii) ␣S and ␤S bound less kinesin than uncleaved ␣and ␤-subunit, respectively, and (iii) ␣S bound approximately two times less kinesin than ␤S. Although kinesin binding efficiency was decreased with subtilisin-cleaved tubulin, it is clear that RK375 did still bind to the carboxyl terminus-deleted tubulin, indicating that the kinesin binding sites are not located in the extreme carboxyl-terminal domain of tubulin but rather upstream of or even close to the subtilisin cleavage sites (Asp 438 for ␣-tubulin and Gln 433 for ␤-tubulin (32)). Altogether, the results from Figs. 4, 5, and 6 suggest that polyglutamylation of tubulin controls kinesin binding through an indirect, conformational change of the carboxyl-terminal domain of tubulin, as previously reported for Tau or MAP2 binding (24).
Moreover, it is relevant that compared with intact tubulin, kinesin binding is less affected than Tau binding by subtilisin cleavage (data not shown). Combined with the observations that DM1A prevented kinesin but not Tau binding and that both are under the same control by polyglutamylation, it can be concluded that the binding sites for kinesin and Tau are distinct, although sufficiently close together within the carboxylterminal domain. The Tau binding site would be downstream from that of kinesin, ending, at least, in the vicinity of the subtilisin cleavage site.
Binding Sites for Structural and Motor MAPs-To get information about the respective binding sites for structural and motor MAPs on the tubulin molecule and to know whether the different MAP classes can compete with each other, the following experiments were carried out. Constant amounts of tubulin were separated by one-dimensional PAGE and transferred onto three nitrocellulose membranes that were incubated with increasing concentrations of heat-stable MAP from brain (Fig. 7). As control, tubulin-bound Tau was detected and quantified on blot 1 using the monoclonal antibody Tau-1; as shown in Fig. 7 (dotted line), the Tau binding sites were effectively saturated from 4 g⅐ml Ϫ1 onwards. On blot 2, a second overlay was performed with a constant amount of RK375; tubulin-bound kinesin was detected and quantified with anti-HIPYR. As il-  Fig. 7 (dashed line), the binding level of RK375 was constant and independent of the amount of Tau previously bound to tubulin, indicating that kinesin head interaction was not impeded by bound Tau. To rule out the possibility that bound Tau could have been displaced by RK375, blot 3 was processed as blot 2 except that tubulin-Tau interactions were stabilized with formaldehyde between the first overlay with Tau and the second with RK375 (Fig. 7, solid line). The comparison of the binding curves shows that stabilization of the tubulin-Tau complexes did not modify the constant interaction of RK375 with Tau-saturated tubulin. Altogether, these results strongly suggest that the binding sites for Tau and kinesin on the tubulin subunit molecules are independent.

Interactions between Tubulin Subunits and Kinesin-Kine-
sin is a plus-end-directed microtubule motor that binds to MTs with a longitudinal periodicity of 8 nm, corresponding to that of a tubulin dimer (33,34), and moves along a single protofilament track (35, 36) with 8-nm steps (37). In agreement with the binding stoichiometry and the step increments, three-dimensional map structures of MT-bound kinesin reconstructed after electron microscopy and image analysis have shown recently that kinesin steps over two adjacent tubulin monomers along a protofilament axis, the motor domain being closer to one monomer than to the other (38 -41). Upon interaction and nucleotide binding, large scale conformational changes were also revealed in tubulin (39) and in kinesin (38). To characterize the tubulin subunit(s) targeted by kinesin, cross-linking experiments have identified either the ␤-tubulin (34,42) or both the ␣and ␤-tubulin (43). These different results could mean that the kinesin motor domain sits well on both tubulin subunits; it is closer to ␤-tubulin, however, in agreement with the structural data.
In this paper, we used a quite different approach by blot overlay to study the interaction between tubulin and kinesin. We showed that the kinesin head domain binds to both ␣and ␤-tubulin monomers with the same apparent affinity. Under the conditions used here, tubulin monomers are quite separated on blots; they play their own part and cannot receive conformational information from the other subunit(s), as is the case in the heterodimeric structure and especially in a MT. It remains, however, that our results show that each ␣or ␤-tubulin monomer has the intrinsic property to interact with kinesin. Moreover, all of the effectors tested in the binding assays modulate in the same way and at the same extent the binding to both tubulin subunits, indicating that both monomers share the same binding features for kinesin. Inhibition of kinesin binding on ␣and ␤-tubulin by competitive free tubulin added in the overlay solution as well as the selective binding of RK to tubulin in a whole protein extract (see Fig. 1B that RK binding was specific. Interaction of Tubulin with Plus-or Minus-end-directed Kinesin-Comparison of binding properties of the conventional RK motor domain with those of the reverse, minus-end-directed Drosophila kinesin-related protein ncd showed no significant quantitative or qualitative difference. In particular, both tubulin subunits were bound at similar levels by the two motors. Moreover, when analyzed by blot overlay after separation of tubulin by two-dimensional PAGE, the ncd motor bound differentially to the diverse isotubulins as a function of their degree of polyglutamylation (data not shown) following a pattern quite similar to that observed with RK375. Thus, the vectorial specificity of the motors within the kinesin superfamily, moving toward the plus-or the minus-end of a MT, does not seem to be related to their preferential binding to a given ␣or ␤-tubulin subunit (which could have been sufficient to confer to the motor protein a specific position on the MT and hence a given polarity of translocation). Furthermore, because it has been reported that changing the motor domain from its normal amino-terminal position to an ectopic carboxyl-terminal location does not reverse the direction of its movement along the MT (44) and that both plus-and minus-end-directed motors compete for the same binding site on tubulin (45), it appears that the directional specificity of a kinesin motor has to be researched subtly in the kinesin amino acid sequences and/or in the fine tune interaction with tubulin.
Binding of the Truncated Form of Kinesin Head, RK329 -The net decrease in the ability of the truncated motor domain RK329 to bind tubulin points to the amino acids 330 -379 peptide segment. The presence of this segment could help the upstream region, described as the boundary of the MT-binding domain (46), to adopt the right conformation. More precisely, this region could be involved in the dimerization of head domains. As previously reported (28), the high ␣-helical propensity observed in the stalk, which extends up to position 340, is compatible with a coiled-coil structure just after the head domain. Because a 340-or 357-amino acid kinesin construct is monomeric and a 392-amino acid one is dimeric (28), the 379amino acid rat kinesin construct used here might also be dimerized in solution. Practically, when RK329 and RK375 were analyzed by HPLC gel filtration, RK329 was eluted later than RK375, with elution times compatible with a monomeric state for RK329 and a dimeric state for RK375 (see Fig. 3C). In theory, the difference in the binding of RK329 versus RK375 in the overlay or enzyme-linked immunosorbent assay-type experiments, as revealed by antibody detection, should thus have been in a ratio of 1:2. Instead, the 5-10-fold decrease we observed suggests that dimerization could drive a conformational change in the tubulin binding site of the kinesin head domain, increasing its affinity for tubulin.
Tubulin Binding Sites for Structural and Motor MAPs-Several experiments reported here indicate that Tau and kinesin have distinct binding sites on the tubulin molecule, confirming previous reports (47)(48)(49)(50)(51). In competition experiments, Tau and kinesin bound independently to tubulin (see Fig. 7). In addition, antibody blocking experiments using DM1A and DM1B led to quite different results; although neither antibody affected the binding of Tau (24), they significantly interfered with that of kinesin; DM1A has a more pronounced effect than DM1B (see Fig. 5).
Although independent, these binding sites must be sufficiently close to each other, however, within the carboxyl-terminal domain of tubulin. Both sites are under the control of tubulin polyglutamylation, which is thought to alter the conformation of the whole carboxyl-terminal domain as a function of the post-translational chain length (see Fig. 4 and Ref. 24).
Moreover, it has been reported that the projection domain of MAP2, and eventually that of Tau, could interfere with kinesin binding (47,50,52), suggesting that if the binding sites are distinct, they are not far from each other.
The extreme carboxyl-terminal sequences of ␣and ␤-tubulins carrying the polyglutamyl chains can be cleaved with subtilisin. In this case, there is a 2-fold decrease in kinesin binding and a more pronounced effect in Tau binding, suggesting that the Tau binding site is closer to the subtilisin cleavage site and hence downstream relative to the kinesin binding site. Moreover, it has been reported that taxol-stabilized MTs cut with subtilisin (instead of soluble tubulin dimers, as in this paper) retained the property to bind kinesin but lost completely (48,51) or partially (52,53) the property to bind Tau, due to the cleavage of the ␣-subunit in the latter case.
It remains difficult, however, to assign more precise locations for the kinesin and Tau binding sites on linear representations of tubulin molecules (Fig. 8A). All of the results from the literature point to region II of the carboxyl-terminal domain of the tubulin subunits described by Goldsmith et al. (26), corresponding to sequences highly conserved within each tubulin subunit (Fig. 8A). Secondary structure predictions (54,55) have suggested that most of the sequences in region II form an anionic ␣-helix (from amino acids 413-418 to amino acids 434 -436 in ␣-tubulin and from amino acid 407 to amino acid 426 in ␤-tubulin), which likely corresponds to the ϳ25-Å-long density observed by Nogales et al. (55). It is interesting to note that in agreement with the putative structure we published recently (24), the DM1A and YOL1/34 epitopes are superimposed on the external face of this helix (Fig. 8B), leaving enough room for protein binding sites on the internal, anionic face. We suggest that the accessibility of these internal binding sites is controlled by the polyglutamyl chains located downstream (anchored at residues Glu 445 in ␣-tubulin and Glu 435 in class II ␤-tubulin; see Refs. 20, 32, and 56). In this view, the accessibility of this anionic face would be blocked in the nonglutamylated tubulin by the interaction with the cationic face of an upstream ␣-helix (amino acids 385-403 in ␣-tubulin and amino acids 371-391 in ␤-tubulin, Fig. 8A) through ionic bonds. This binding inhibition would then be released by the negatively charged polyglutamyl chains, which would compete with the ionic bonds and move the two helices apart from each other, giving thus the accessibility to the MAP binding sites (see also Ref. 24).
It is relevant that GT335, which readily inhibits both Tau and kinesin binding, has also been shown recently to strongly inhibit dynein-based flagellar motility (57), suggesting that the polyglutamyl chains could play a crucial role in controlling the whole carboxyl-terminal domain of tubulin. Experiments are in progress to test and complete the data with other structural and motor MAPs.