Cryo-electron tomography of microtubule-kinesin motor complexes

Abstract

Microtubules complexed with molecular motors of the kinesin family or non-motor microtubule associated proteins (MAPs) such as tau or EB1 have been the subject of cryo-electron microcopy based 3-D studies for several years. Most of these studies that targeted complexes with intact microtubules have been carried out by helical 3-D reconstruction, while few were analyzed by single particle approaches or from 2-D crystalline arrays. Helical reconstruction of microtubule-MAP or motor complexes has been extremely successful but by definition, all helical 3-D reconstruction attempts require perfectly helical assemblies, which presents a serious limitation and confines the attempts to 15- or 16-protofilament microtubules, microtubule configurations that are very rare in nature. The rise of cryo-electron tomography within the last few years has now opened a new avenue towards solving 3-D structures of microtubule-MAP complexes that do not form helical assemblies, most importantly for the subject here, all microtubules that exhibit a lattice seam. In addition, not all motor domains or MAPs decorate the microtubule surface regularly enough to match the underlying microtubule lattice, or they adopt conformations that deviate from helical symmetry. Here we demonstrate the power and limitation of cryo-electron tomography using two kinesin motor domains, the monomeric Eg5 motor domain, and the heterodimeric Kar3Vik1 motor. We show here that tomography does not exclude the possibility of post-tomographic averaging when identical sub-volumes can be extracted from tomograms and in both cases we were able to reconstruct 3-D maps of conformations that are not possible to obtain using helical or other averaging-based methods.

Figure 1. Comparison between cryo-electron tomography and helical 3-D reconstruction of frozen-hydrated microtubules decorated with monomeric kinesin (Eg5) motor domains

(A) A 4.5nm slice from a tomogram of microtubules decorated with monomeric Eg5 motors. Microtubules which are slightly tilted to the image plane expose a sequence of their top, lumen and bottom regions as they cut through the slice. The B-lattice pattern emphasized by the motor decoration shows a different angle along the Bessel -2 helix according to the sketch. Since this is a left-handed helix, red lines show the angular orientation of the helical path according to the top view (towards the observer), while the green lines indicate the course of the helix at the bottom of the microtubule. Occasionally a lattice seam, typical for non-helical microtubules with B-lattices, can be seen directly. The inset magnifies the framed area and reveals a bottom view of the seam, here made visible by the motor decoration. The seam is visible due to the shift of one protofilament by 4nm in the axial direction relative to the other (interruptions in green lines). (B) Cross-section through a 9nm projection of a volume-average obtained by PEET combining 218 particles along one microtubule. This microtubule is composed of 14 protofilaments, which can be clearly seen despite the obvious loss of resolution along the Z-axis (here top to bottom) due to anisotropy caused by the missing wedge of data. (C) The surface rendering of the average clearly shows the helical path of the microtubule as well as the lattice seam typical for a 14-protofilament microtubule. (D–F): Helical averaging of microtubules is only possible if they are truly helical. This is the case for a 15-protofilament microtubule (red frame in F) but not for the more common 13-protofilament microtubules (yellow frame in F). Fourier filtering (D & E) of the microtubules marked by frames reveals strikingly different 2-D projection patterns. 13-protofilament microtubules (D) are perfectly aligned with the tubular axis while 15- (E–I), or 16-protofilament microtubules (Figs. 5 & 6) reveals a super-twisted arrangement. The supertwist is visible in the 3-D reconstruction in F as a deviation of the protofilaments (dotted green line) from the microtubule axis (solid green line). Kinesin motor decoration reveals a Fourier filtered image as shown in G. The helical diffraction pattern (H) shows the helical layerline pattern where the layerlines around Bessel -2 (axial motor-tubulin dimer repeat) and -4 (axial α–β–α–β tubulin repeat) form a cluster of 3–4 strong layerlines due to the convolution of the -2 and -4 helices with the protofilament supertwist. (I) The 3-D reconstruction reveals the added kinesin heads as yellow densities, marking each αβ-tubulin dimer.

Figure 2. Cryo-electron microscopy of microtubules decorated with Kar3Vik1 heterodimers

(A) Frozen-hydrated microtubules decorated with a dimeric Kar3Vik1 construct reveals obvious cooperative microtubule-binding properties. The cooperative binding creates a heterogeneous decoration pattern where some microtubules show full decoration on one side while the opposing side is free of motor decoration (red frame). (Insets in A) Crystal structures of the Vik1 motor-homology domain (left), and the Kar3 motor domain (right) are shown for comparison. Some structural similarities such as the central β-sheet surrounded by three α helices on each side are obvious. However, despite the similarity, Vik1 lacks a nucleotide-binding site. (B) Magnified and contrast enhanced region of the frame in A. The picture contrast has been inverted and boosted by an “embossing” filter to emphasize the visibility of the tubulin-Kar3Vik1 dimer complexes. Since the motors have been flushed with AMP-PNP, we believe that the domain in contact with the microtubule surface is the Kar3-MD while the tethered domain is Vik1-MHD. However, this remains to be determined experimentally

Figure 3. Dissection of microtubules partially decorated with Kar3Vik1 by cryo-electron tomography

(A) A 9nm slice through a tomogram of microtubules decorated with Kar3Vik1 heterodimers reveals features similar to those discussed in Figure 1. Kar3Vik1 decorates microtubules in a cooperative fashion leaving empty patches besides fully decorated areas. (B) Three individual 4.5nm sections through the top, center, and bottom of the tomogram of the microtubule boxed in A. The inset in the center panel shows an end-on view of the microtubule. While the bottom region shows strong striations every 8nm corresponding to a fully motor-decorated surface (green lines) the top region reveals empty microtubule protofilaments running axially. A striking advantage of tomographic 3-D reconstruction is observed in the center panel in that small single events such as individual missing motors (red circles) are revealed with much better clarity than in 2-D projections. These events would be lost after helical averaging due to the symmetry constraints. (C) An isosurface representation from a PEET average of 146 particles selected from the microtubule in B gives a clear view of how the top of the microtubule is free of motor decoration while the bottom and sides of the microtubule are completely decorated by KarVik1.

Figure 4. Cryo-electron tomography of Kar3Vik1 dimers connecting between adjacent microtubules

(A) 18nm x-z (top) and x-y (bottom) slices through a tomogram showing the cross section through two adjacent microtubules (top) and the Kar3Vik1 motors extending toward each other between the two microtubules (bottom). (B) An 18nm projection and surface rendering of an average obtained by PEET from 44 subvolumes selected from the tomogram in A. The bridging motors appear to bend towards each other in a manner different from the regular radially extending conformations shown in Figures 3 and 5. Since this is a highly artificial in vitro situation, the biological significance of these bridging motors is not clear. Rather, this exemplifies the way we are able to use cryo-electron tomography and subsequent volume averaging to analyze structural details that would not be possible by other averaging methods.

Figure 5. Cryo-electron tomography and subsequent volume averaging of a fully decorated Kar3Vik1-microtubule complex

(A) 3nm thick tomographic x-y slice of a microtubule decorated with Kar3Vik1 overlaid with the averaged volume shown in B–E and Figure 6A and 6B. (B–D) 4nm x-y slices through the top and center of an average of 99 particles selected from the tomogram in A according to the planes indicated in E. The center slice in D cuts along the missing wedge, noticeable as reduced resolution compared to the x-y slice above. The central x-y slice (C) through the microtubule clearly shows densities corresponding to the α- and β-subunits of tubulin and to the two globular domains of Kar3 and Vik1 extending out from the microtubule. (E & F): End-on view (x-z) of the 3-D map before (E) and after rotational averaging over all protofilaments (F). The rotationally averaged map now looks indistinguishable from a helical reconstruction map at the same resolution. While the missing wedge effect is clearly visible before rotational averaging as the strong densities and smearing out along the z-axis, rotational averaging eliminates the missing wedge completely.

Figure 6. Rotational averaging around the tubular axis effectively eliminates the missing wedge

(A & B) Surface renderings of the averaged 3-D map are shown in Figure 5A–E. The cross-section reveals the number of protofilaments of this microtubule to be 16, which produces a perfectly helical microtubule (Bessel order -2) without seams (see B). (D & E) Surface rendered 3-D map after rotational averaging of the map in A (see also Fig. 5F). The missing wedge effect is eliminated. By rotationally averaging over all 16 protofilaments the asymmetric unit changes from axial slices along the microtubule axis to one αβ-tubulin–motor domain complex. Hence, the theoretical maximum number of asymmetric units included in the average is increased from 99 to 1584 (16 × 99). However, our average is based on the 1242 particles with the best correlation to the reference because this subset gave the highest resolution. However, we have included particles with the highest correlation scores only up to the point where the Fourier shell correlation continues to improve; thus, the number of particles used in the rotationally averaged volume is 1242. (C & F) For Fourier-shell correlation calculations, the datasets were split in half on a random basis and the two halves were correlated against each other. Fourier-shell correlation graphs obtained from the maps in A (C) and D (F) reveal resolution limits of 3.8nm and 3.2nm respectively based on the 50% correlation criterion. For FSC calculations a dataset is split in half on a random basis and correlated against each other. Accordingly the number of asymmetric units in each group is maximally 49 before rotational averaging, and 621 afterwards.