The motor domain of cytoplasmic dynein in the unprimed conformation adopts two orientations on the EM grid under our negative staining conditions, giving two distinct ring-like views (B and 1C). The most common view (B) is similar to previous images of an identical motor construct (Samso and Koonce, 2004
). We refer to this here as the “top” view because it corresponds to the view looking directly along the channel axis in their 3D reconstruction. With the image oriented so that the stalk emerges at the 11 o'clock position, well-defined stain-excluding lobes are visible around the left side of the head, while density on the other side is stronger but less distinct. The second view (C) resembles the “right” view of axonemal dynein-c (D) (Burgess et al., 2003
). The right view similarly oriented (C) shows pronounced stain-excluding lobes around the upper and right side of the head and less differentiated density on the left side. A spike and a groove (C, arrowheads) are pronounced in the cytoplasmic dynein. However, the main conclusion from these first images of right views of cytoplasmic dynein is that the head structures of axonemal and cytoplasmic isoforms are strikingly similar.
Location and Structure of the Stalk in the Unprimed Motor
Image classification reveals that the MT-binding stalk emerges at a range of angles from a fixed position in the head (; see Movie S3
available online). In top views it emerges from a prominent lobe of density, whereas in right views it emerges between two adjacent lobes (see also C). The stalk is about 2 nm wide, consistent with the prediction of a coiled-coil structure. The visible coiled coil is 10.4 nm long. With the ~4 nm distal MTBD this is very similar to the length of dynein-c's stalk (Burgess et al., 2003
). The coiled coil of dynein-c in the unprimed conformation has a bend about two-thirds along its length (Burgess et al., 2003
), which may correspond to a proline residue within the outward α helix (Yagi et al., 2005
). By contrast, the stalk of cytoplasmic dynein lacks this proline and is straight, except for an occasional kink (to the right) at its distal end immediately adjacent to the MTBD ().
Stalk Structure in the Unprimed Motor
Dynein's Six AAA+ Modules Alone Form a Ring
To investigate the structure of the head domain, in particular the contribution made by the AAA+ region, we engineered truncation constructs lacking the C sequence (ΔC), the N sequence (ΔN), and both these flanking sequences (ΔNΔC; see A). Functional assays showed that the C sequence is not required for basal ATPase but is required for MT-binding and motile activities, whereas the N sequence is required for all these activities (Table S1
). Analytical ultracentrifugation showed that all three truncation constructs retain a low frictional coefficient similar to the intact motor domain, indicative of a compact fold (Table S1
Structural Impacts of Truncation on the Dynein Motor
EM and image processing reveal that the ΔC construct, which lacks all 406 residues C-terminal to AAA6, is strikingly similar to the motor domain in the right view (B). The head has the same overall asymmetric ring shape and the stalk is present in the same place (C). The main difference is the appearance of reduced density to the right of the central stain pool and a loss of the spike on the left margin of the head (B and 3C). The ΔC construct does not show a gap in the head corresponding to a missing peripheral domain that would be expected from heptameric models.
A subset of ΔC molecules (~8%) has a different appearance. An extended lobe of material protrudes from the head opposite the stalk (arrowhead, B) and is variable in position. This structure has appropriate dimensions to be the linker previously identified in axonemal dynein-c (Burgess et al., 2003
). Such images suggest that deletion of the C sequence can destabilize linker-head interactions to favor linker undocking.
ΔN molecules, which lack the 542 residues N-terminal to AAA1, show a striking new “ring” appearance, rather than top or right views (B). This new appearance is more symmetrical, with wedge-shaped densities defined by radial lines of stain. The stalk is intact (C) and emerges from one of the wedge-shaped densities.
ΔNΔC molecules, which lack both N and C sequences, also show a ring appearance. The ΔNΔC ring is surprisingly similar to the ΔN ring (B), despite the loss of the C sequence, equivalent in length to two AAA+ modules. The main difference between ΔNΔC and ΔN is weaker density at the ~8 o'clock position (as orientated in B), and increased variability in this region, as indicated by less sharply defined density here (C). Overall, the ring appears complete and the stalk is intact (C).
These truncation constructs reveal several new aspects of the organization within the dynein head. The C sequence is not an integral part of the ring as proposed in heptameric models. Instead, the C sequence may stabilize closure of the ring because its removal causes structural variability opposite the stalk. Removal of the N sequence reveals a more symmetrical ring, recalling images of other ring-shaped AAA+ proteins (Mocz and Gibbons, 2001
). Together, these results show that dynein's six AAA+ modules alone form a ring structure.
Mapping Sites within the Motor Domain Using GFP-Based Tags
To determine how the heavy-chain sequence maps onto the morphology of the motor domain, we used EM to examine fusion proteins in which GFP and BFP were inserted at seven different locations (A; Figure S1
). These tagged constructs show robust MT-sliding activity (Kon et al., 2005
), including those newly engineered in this study (data not shown). Images of tagged motors (B) appear substantially similar to the untagged motor (B and 1C), indicating that insertion of the tags does not perturb the overall fold of the heavy chain.
Mapping the Locations of the GFP-Based Tags in Dynein Fusion Proteins in the Unprimed Conformation
We used two methods to establish the location of the tags: difference mapping (Figure S2
) and a novel image classification procedure to scan systematically positions around the perimeter of the head (Movies S1 and S2
). The results from these two methods are consistent. We then applied image classification to the regions identified by the first two methods, to show the tags in more detail (B). The tags appear as globular densities consistent with the β barrel structures of GFP and BFP. To locate the tags accurately and without bias, we used an automatic detection procedure (see legend, Figure S2
) and calculated their mean positions (C), which we describe in detail below.
The N Terminus of the Unprimed Motor Lies near the Stalk Base
The GFP tag fused to the motor N terminus is close to the periphery of the head near the base of the stalk (B; Figure S2
). This is observed in both top and right views. Scanning classification confirms that N-terminal GFP is absent from other positions around the perimeter of the head (Movies S1 and S2
). This finding is contrary to an earlier suggestion (Meng et al., 2006
) that the motor N terminus is randomly orientated around the head and lies at a high radius (see legend to Movie S2
for discussion). We conclude that in the unprimed conformation the N terminus lies near the base of the stalk. This location is close to the linker-tail junction in axonemal dynein-c (D), suggesting that a similar linker exists in cytoplasmic dynein.
AAA1 Is Opposite the Stalk and the N Sequence Spans the Head
The B1 tag, inserted 20 amino acids downstream of the main catalytic AAA+ module (AAA1), has a peripheral location opposite the stalk in both top and right views (A and 4B). Because GN and B1 tags lie on opposite sides of the head, the polypeptide chain between them must span the head. Within this sequence, ~240 amino acids are predicted to form AAA1 and the downstream sequence to B1 (A), leaving the remaining ~550 amino acids upstream of AAA1 to span ~14 nm across the head (C). This fits the model in which the N sequence includes the linker domain (Numata et al., 2008
), the mechanical lever originally proposed by Burgess et al. (2003)
AAA2, AAA5, and AAA6 Fit a Counterclockwise Arrangement of AAA+ Modules
Having established that AAA1 lies opposite the stalk and N terminus, we next determined the direction of AAA+ modules around the head. The B2 tag, inserted within AAA2, is positioned counterclockwise from the B1 site in top and right views (B). This indicates that both views show the same face of the AAA+ ring. Confirming this, the B5 tag inserted 68 amino acids downstream of AAA5 lies counterclockwise of the stalk in both views. The B6 tag, inserted within AAA6, lies counterclockwise from the B5 site in top views (B). Thus, in the views shown, AAA1, AAA2, AAA5, and AAA6 are arranged counterclockwise around the ring (C). The close proximity between AAA1 and AAA6 fits our finding that the core of the motor is a hexameric ring of AAA+ modules.
The C Sequence Spans between AAA6 and near the Stalk Base
The finding that the C sequence does not close the ring raises the question: where is it located within the head? To investigate this we located the B7 tag, inserted about one-third through the C sequence (corresponding approximately to the naturally truncated C terminus of fungal dyneins). Difference mapping shows that the B7 tag has an internal position within the head in top and right views (B) in contrast to the other tags. In both views, B7 is located within ~6 nm of the base of the stalk. To map where the C sequence terminates, we imaged a new construct with BFP fused at the C terminus of the motor (BC). The BC tag lies on the head periphery between the B5 and B6 tags (B). Thus, the C sequence spans from AAA6 toward AAA5 and the base of the stalk in its first one-third and then returns toward AAA6 (C).
Movement of the Linker during the Priming Stroke
To investigate dynein's motile mechanism, we located the position of the GFP tag attached to the linker N terminus in the unprimed and primed conformations. To generate the primed conformation, we treated the motor with ATP and vanadate to trap the ADP.Vi complex (Kon et al., 2005
). Most strikingly, in ADP.Vi motors, GFP is shifted toward AAA2 (A). This is seen in both right view (4% of motors) and top view (96%), although the distribution of GFP positions differs in these two views. In right view, all ADP.Vi motors show GFP close to AAA2 (B). In top view, the distribution is bimodal: 44% show GFP near AAA2, while 56% show GFP near the stalk base, coinciding with its unprimed location (B). This leads us to speculate that in the ADP.Vi motor (1) the linker exists in a poised equilibrium between primed and unprimed conformations and (2) the equilibrium position is altered by the orientation of the molecule on the EM grid. This might be analogous to the situation in other motors, where crystallization conditions are thought to shift conformational equilibria of myosin's converter domain and kinesin's neck linker (Vale and Milligan, 2000
). Our 3D analysis (below) shows that in the top view, but not the right view, the linker swing has a large component in the z direction, suggesting that the top view is the more likely of the two to have been influenced by the EM grid. The distal portion of dynein's linker is occasionally revealed in top views of ADP.Vi motors, as a rod ~2 nm thick (A, arrows) connecting the N-terminal GFP to the head (close to the B2 site). Together, these observations suggest that during the priming stroke, the motor N terminus moves from near the stalk base and AAA4 toward AAA2 by a swing of the linker.
Structural Changes between Unprimed and Primed Conformations
The mean displacement of GFP during the priming stroke is 19 nm in right views (B). Measurement in top views is complicated by the broader distribution of GFP in ADP.Vi motors. Based on Gaussian fits to the bimodal distribution of GFP angles around the head (B), we segregated the ADP.Vi motors into two subpopulations (, legend). We define the motors in the subpopulation nearer AAA2 as the primed conformation. Using the mean GFP position of this subpopulation, the displacement of GFP during the priming stroke is 13 nm in top views (B). In both views, the direction of the linker swing is almost parallel to the long axis of the stalk.
Magnitude and Direction of the Linker Swing
Right views of ADP.Vi motors show a prominent accumulation of stain at the base of the stalk (C, arrow), not seen in apo/ADP motors. This change at the stalk base likely occurs because of movement of the linker N terminus. This accumulation of stain suggests that the stalk coiled coil bifurcates at the junction with the head (C, arrow), as reported for dynein-c (Burgess et al., 2003
Tilting of the Stalk between Weak and Strong MT-Binding States
The ADP.Vi motor binds to MTs with weaker affinity than the apo/ADP motor (Imamula et al., 2007
). Between these states, we find that the angle of the stalk changes relative to the head (C). In both states the stalk angle is variable (Movie S3
), with similar standard deviations (C, legend). The distributions overlap but there is a shift in their mean angles. From apo/ADP to ADP.Vi, the stalk tilts clockwise: in right view by 16° and in top view by 2° (C), the former displacing the center of the MTBD by 5 nm. Thus, relative to the head, the stalk undergoes a small nucleotide-dependent tilt.
Three-Dimensional Movement of the Linker N Terminus during the Priming Stroke
To determine the positions of the various tags in 3D, and the 3D movement of the linker N terminus, we calculated the angular relationship between top and right views (Figures S3 and S4
). Superficially, top and right views look like reflections of one another in a vertical mirror, suggesting they may be related by a rotation of ~180°, but the similar emergence points of the stalk and the positions of the tags, most notably B2 and B5, rule this out (C). To obtain the axis of rotation between top and right views, we used the positions of tags in the unprimed motor (Figure S3
). The axis of rotation obtained (A) is also compatible with the segregation in top views of primed and unprimed linker conformations as defined above (B). We then used geometric constraints to establish that top and right views are related by a rotation of between 50° and 116° about this axis (see Figure S4
). This is consistent with our earlier interpretations of left, side, and right views of dynein-c (Burgess et al., 2004b
; see Figure S5
). Based on this angular range, we calculated the 3D positions of the five tags (C and S4
D; see also Movie S4
) including the 3D position of GFP in the primed conformation (D; Movie S4
). This analysis shows that during the priming stroke the majority of GFP movement occurs in the plane of the right view, with a large component perpendicular to the plane of the top view (D). The 3D displacement of GFP is ~19–21 nm (Movie S4
). Accounting for the size of GFP, we infer that the distance moved by the linker N terminus is ~16–18 nm.