A new look at the microtubule binding patterns of dimeric kinesins¹

Abstract

The interactions of monomeric and dimeric kinesin and ncd constructs with microtubules have been investigated using cryo-electron microscopy (cryo-EM) and several biochemical methods. There is a good consensus on the structure of dimeric ncd when bound to a tubulin dimer showing one head attached directly to tubulin, and the second head tethered to the first. However, the 3D maps of dimeric kinesin motor domains are still quite controversial and leave room for different interpretations. Here we reinvestigated the microtubule binding patterns of dimeric kinesins by cryo-EM and digital 3D reconstruction under different nucleotide conditions and different motor:tubulin ratios, and determined the molecular mass of motor-tubulin complexes by STEM. Both methods revealed complementary results. We found that the ratio of bound kinesin motor-heads to αβ-tubulin dimers was never reaching above 1.5 irrespective of the initial mixing ratios. It appears that each kinesin dimer occupies two microtubule-binding sites, provided that there is a free one nearby. Thus the appearances of different image reconstructions can be explained by non-specific excess binding of motor heads. Consequently, the use of different apparent density distributions for docking the X-ray structures onto the microtubule surface leads to different and mutually exclusive models. We propose that in conditions of stoichiometric binding the two heads of a kinesin dimer separate and bind to different tubulin subunits. This is in contrast to ncd where the two heads remain tightly attached on the microtubule surface. Using dimeric kinesin molecules crosslinked in their neck domain we also found that they stabilize protofilaments axially, but not laterally, which is a strong indication that the two heads of the dimers bind along one protofilament, rather than laterally bridging two protofilaments. A molecular walking model based on these results summarizes our conclusions and illustrates the implications of symmetry for such models.

Introduction

The cellular, biochemical and structural information on the kinesin superfamily has reached a level which allows to draw a rather detailed picture of their functions and properties (for reviews see: Hirokawa 1998, Bloom and Golstein 1998, Mandelkow and Johnson 1998, Mandelkow and Hoenger 1999, Sheetz 1999, Endow 1999). The interaction of kinesin-like motor domains with microtubules has been extensively studied by several groups using electron microscopy and computational 3D analysis Hirose et al 1995, Hirose et al 1996, Hirose et al 1998, Hirose et al 1999, Hoenger et al 1995, Hoenger et al 1998, Hoenger and Milligan 1997, Kikkawa et al 1995, Aranl et al 1996, Arnal and Wade 1998, Sosa et al 1997a, Sosa et al 1997b, Rice et al 1999, Kikkawa et al 2000. The atomic structures of monomeric kinesin and ncd motor domains have been solved by X-ray crystallography Kull et al 1996, Sablin et al 1996, Sack et al 1997. More recent X-ray structures of dimeric motor constructs Kozielski et al 1997, Sablin et al 1998 revealed further insight into the motor neck regions and the dimerization properties of the α-helical coiled-coil within the stalk. Finally, a 3D map of the αβ-tubulin heterodimer at near atomic resolution has been solved by electron crystallography from zinc-induced tubulin sheets (Nogales et al., 1998). These data are the basis for molecular docking experiments, which combine intermediate resolution data from 3D cryo-electron microscopy with high-resolution data from electron or X-ray crystallography. This now enables structural interpretations of large macromolecular assemblies such as intact microtubules (Nogales et al., 1999) and microtubule-motor complexes to near-atomic detail Sosa et al 1997a, Hoenger et al 1998, Kozielski et al 1998, Hirose et al 1999, Rice et al 1999, Kikkawa et al 2000.

Cryo-electron microscopy 3D reconstructions revealed a consensus view on the overall arrangement of dimeric ncd motor domains complexed with microtubules Sosa et al 1997a, Hirose et al 1998. Ncd dimers interact with tubulin through one of their heads with the second ones protruding outward. Slight conformational changes upon nucleotide exchange have been reported within the unattached heads (Hirose et al., 1998). However, in the case of dimeric kinesin the situation is much more confusing (Hirose et al 1996, Hirose et al 1999, Aranl et al 1996, Arnal and Wade 1998, Hoenger et al 1998; see also Figure 1). There is a general agreement on the densities related to one strongly bound head domain per αβ-tubulin heterodimer. However, significant differences exist in the interpretations of the volumes that relate to the second head of the dimer (Figure 1; compare Hoenger et al., 1998 with Hirose et al 1996, Hirose et al 1999, Hirose et al 1998). Some maps were interpreted to show loosely bound heads, analogous to the second head of ncd, but containing only a fraction of the expected mass. This lack of mass was interpreted as the result of an intrinsic flexibility of these so called “loosely bound” heads at this position, which, nevertheless, was considered to reflect a relevant conformational state of the kinesin dimer Hirose et al 1996, Hirose et al 1999, Arnal and Wade 1998. The “loose” head had a different orientation than that of ncd, which was interpreted to reflect the different directionalities of the two motors (see also Endow, 1999).

Here we used cryo-electron microscopy (cryo-EM) and 3D image reconstruction, scanning transmission electron microscopy (STEM) mass determinations and molecular modeling to clarify the properties of dimeric kinesin-microtubule complexes. Our findings address two key-issues which are different in our interpretation from earlier ones: (1) How accurate are the current 3D reconstructions of microtubules decorated with dimeric kinesin under conditions where both heads potentially may interact with the microtubule surface at the same time? (2) How well defined and unique are the conformational states of apparently loosely-bound (supposed not to contact the microtubule surface) heads of kinesin dimers under near-physiological conditions?

Given that the apparent electron densities of kinesin dimers on microtubules revealed inconsistent results among different groups, the attempts to correlate them with the X-ray structure led to different docking models. Two of these models show one tightly and one loosely bound motor head per αβ-tubulin dimer, however, both reveal a different orientation (Kozielski et al 1998, Hirose et al 1999; Figure 1(b) and (c)). In addition, our previous results showed a separation of heads and docking onto two different β-tubulin subunits (Hoenger et al., 1998; Figure 1(a)) a similar model was implied by kinetic arguments (Romberg et al., 1998). Here, we show additional evidence that the two heads bind along one protofilament (Figure 5). We now believe that, in contrast to dimeric ncd (Sosa et al., 1997a) the appearance of the loosely bound kinesin heads does not reflect a physiologically relevant state. Our results indicate that the conformations of dimeric kinesins on helically averaged 3D maps are induced by an oversaturation of available microtubule binding sites (see Figure 6). The additional mass takes the appearance of loosely tethered but periodic heads as a result of the helical averaging procedure. Finally, by proposing our walking model we argue that a hand-over-hand mechanism cannot be achieved without the involvement of at least two alternative pathways on how the trailing motor domain switches into the leading position (see also Cross 1995, Howard 1996; Figure 9).