[email protected] and [email protected] into the structural domains that make up an individual dynein heavy chain have been gained from electron microscopy of negatively stained molecules and single-particle image processing. The low-resolution model of dynein’s architecture that has emerged allows us to consider how the domains move and communicate to produce directed motion. 4.1 INTRODUCTIONThe dynein heavy chain is a ~0.5 MDa molecular machine that links ATP hydrolysis to cycles of microtubule binding, movement, and release. Dynein heavy chains are found widely in eukaryotes and fall into two principal classes. Cytoplasmic isoforms transport diverse cargo in cells [49] and assemble into homodimers that can coordinate hundreds of successive steps along microtubules without detachment. Axonemal isoforms, as heterotrimers, heterodimers, and monomers, form cross-bridges (termed inner and outer “arms”) between microtubule doublets in cilia and flagella and generate the bending motions of these cellular appendages [14]. The movement of all characterized dynein heavy chains shows a strong bias toward the minus end of microtubules. Despite the diverse cellular functions and limited sequence identity among dynein heavy chains, striking early electron microscopy (EM) images showed that cytoplasmic and axonemal isoforms have a shared Handbook of Dynein Edited by Keiko Hirose and Linda A. Amos Copyright © 2012 Pan Stanford Publishing Pte. Ltd. www.panstanford.com

overall architecture [1, 18, 48] (see Fig. 1.3). More recent EM images of motor domains from different isoforms further encourage the view that substantial mechanistic similarities exist across the dynein family of enzymes.Dynein’s structure and mechanism have remained a frontier in cell motility research, owing to the large size and flexibility of the heavy chain. With the exception of the microtubule-binding domain, high-resolution structural information is currently lacking for the molecule’s ~4,600 amino acids. However, structural and functional studies are converging on a low-resolution model of dynein’s architecture and motor mechanism that is the subject of this chapter. In the first part, we discuss each of the structural domains within the heavy chain. In the second part, we consider how these domains move and communicate to produce directed motion. 4.2 THE FORM AND FUNCTION OF THE HEAVY CHAINEM established that each dynein heavy chain folds to form a globular, ring-like “head” with two elongated structures, the “tail” and the “stalk,” emerging from it (Fig. 4.1B). In earlier literature, the tail and stalk were referred to as the stem and B-link respectively [18]. 4.2.1 The Tail DomainThe N-terminal ~1,300 amino acids of the heavy chain form the tail (Fig. 4.1A). This was shown by proteolytic digestion of the heavy chain [32] and reinforced by molecular genetics and EM [28]. In cytoplasmic dynein, the tail mediates heavy chain dimerization and binding of associated protein chains that regulate dynein and link it to cargo molecules (see Chapter 13). In axonemal dyneins, the tail also mediates oligomerization and binding of associated subunits (see Chapter 14) and anchors the heavy chain to the A-microtubule of one doublet microtubule against which it can exert force relative to the B-microtubule of the adjacent doublet. The most detailed structural data show that the tail of a monomeric inner-arm dynein folds into a ~25 nm long structure that tapers from its base into a thin (~2 nm), flexible neck near the head [6] (see Fig. 4.1). Planar flexibility has been observed in the neck region, indicating that it may be a compliant element, capable, for example, of storing strain. Torsional flexibility in the neck has also been observed [7], suggesting that there may be a degree of swivel between the tail and head. The tails of different dynein isoforms contain numerous predicted short a-helical motifs, but the structure and topology of the polypeptide within the tail remain uncertain. Multimeric dyneins of both axonemal and cytoplasmic origins show the heads emerging from the same end of the tail complex, implying that the tails assemble in a parallel rather than antiparallel

binding sites respectively and constitute the minimal motor unit of ~3,300 amino acids (~380 kDa). 4.2.2 The Stalk DomainThe stalk is an elongated domain that was originally visualized as the point of contact between axonemal dyneins and the B-microtubule (hence its earlier name; the “B-link”) [17]. A similar stalk structure was later discovered in cytoplasmic dyneins [1, 13]. The stalk is now known to be an intramolecular, antiparallel coiled coil structure, with a globular domain at its tip that binds and releases microtubules in a nucleotide-sensitive manner. A crystal structure of cytoplasmic dynein’s microtubule-binding domain and distal coiled coil has recently been solved [8] (Chapter 6), representing the first high-resolution structure of any part of the heavy chain. Little is known about the structure of the stalk at its junction with the head, although EM images hint that the two strands of its coiled coil may bifurcate (Fig. 4.2A).One critical role of the stalk coiled coil is to relay information between the microtubule-binding domain at its distal tip and the ATPase site(s) within the head, allowing coordination between track binding and ATP hydrolysis. The coiled coil is a curious and fascinating fold for this role. Recent evidence suggests that communication through the ~10 nm long coiled coil is achieved by changes in the alignment between its outward (CC1) and return (CC2) a-helices (see Sections 3.5 and 6.3) [8, 15, 27], though alternative models have been proposed. Consistent with a structural change occurring within the stalk, its stiffness and curvature have been seen to vary in a nucleotide-dependent fashion in axonemal dynein-c [6]. In cytoplasmic dynein, a small nucleotide-dependent shift in the stalk angle relative to the head has also been observed (Fig. 4.2A) [39]. These may be different mechanical “readouts” of the same helix sliding mechanism, produced by differences in stalk amino acid sequence between these isoforms. Images of engineered cytoplasmic dynein motors in which the stalk coiled coil is cross-linked in different alignments support the view that CC1 and CC2 can accommodate relative sliding corresponding to at least two turns of a-helix (one heptad repeat of coiled coil) [27]. Because the alignment between CC1 and CC2 appears to be in dynamic equilibrium [27], it is possible that the stalk is sensitive to mechanical strain, allowing dynein to alter its microtubule binding and ATPase activities in response to tension. A second key role of the stalk appears to be to orient the head domain such that it produces movement toward the microtubule minus end [8]. This might occur via angled binding of the stalk, so that its long axis is not perpendicular to the microtubule long axis, but tilted toward the microtubule plus end (see Chapter 5) [44]. Two highly conserved proline residues in CC1 and CC2 introduce a kink in the coiled coil immediately adjacent to the microtubule-

binding domain, observed in both the crystal structure and EM images (Fig. 4.2A,B). This kink presumably steers the exit angle of the stalk from the microtubule. Interestingly, some monomeric axonemal dynein isoforms

contain an additional proline residue in the middle region of CC1 (Fig. 4.2D). The stalk of dynein-c, which contains this proline, has a bend midway along its length in the absence of nucleotide (Fig. 4.2C; arrowhead), whereas the stalk of cytoplasmic dynein is straight in this region.An untested possibility is that this proline bends the stalk in those axonemal isoforms in which it occurs (Fig. 4.2D) and plays a role in the unusual ability of these isoforms to generate torque against the microtubule during sliding (visualized in in vitro motility assays as microtubule rotation) [24, 47]. Thus, it is conceivable that the stalk structure has diverged among dynein isoforms to produce different types of movement or strain sensitivity suited to their particular biological function. 4.2.3 The Head DomainThe bulk of the motor domain sequence forms the head, which has a ring-like appearance with a diameter of about 13 nm. In our current view, the head can be further subdivided into the AAA+ ring, the C-terminal sequence, and the linker domain which together give rise to the appearance of seven lobes of density seen in EM averages [39, 40, 44]. The AAA+ ringSequence analysis shows that the head contains six AAA+ modules (termed AAA1-AAA6), placing dynein in the AAA+ superfamily of ATPases [35]. The AAA+ module is a ~200-250 amino acid fold typically containing several characteristic motifs involved in ATP hydrolysis. Most AAA+ proteins contain one AAA+ module per polypeptide and assemble into homo-hexameric ring structures. Some contain two AAA+ modules per protomer and assemble into double-layered rings. Many AAA+ rings exert vectorial work by threading a substrate through their axial pore; for instance, unfolding protein substrates at the gates of ATP-dependent proteases, or acting as nucleic acid helicases. While the ring-like appearance of dynein’s head is consistent with its AAA+ relatives, the fusion of its six AAA+ modules into a single polypeptide is a special feature of the dynein lineage. Only dynein’s closest known relative-midasin, which functions in ribosome biogenesis-appears to share this feature [12, 45] (see Section 1.3).Fusion of dynein’s six AAA+ modules within the heavy chain has enabled them to take on different functional roles (reviewed in [37], Section 3.3). Only the first four (AAA1-AAA4) contain nucleotide binding and hydrolysis motifs and among these AAA1 is the primary site of ATP hydrolysis linked to motor activity. AAA2-AAA4 appear to be regulatory in function. In particular, AAA3 plays an important part in the mechanism, as mutations that prevent

nucleotide binding and hydrolysis here cripple dynein’s ability to release from microtubules [9, 25, 42]. It is known that microtubule-based movement by some dyneins either requires ADP, or is accelerated by it, consistent with regulatory ADP binding at one or more sites [51]. The utilization of multiple nucleotide-binding sites distinguishes dynein from myosin and kinesin and discovering the precise roles of AAA2-AAA4 in the mechanism is an exciting challenge for future investigation. The stalk domain is located between AAA4 and AAA5 and between the other AAA+ modules are inserts of unknown fold. Crystal structures of several oligomeric AAA+ enzymes in different nucleotide states have been solved (see [10, 20, 43] for reviews). Each AAA+ module comprises large and small subdomains connected by a hinge (Fig. 4.1E). The large subdomain has an a/b fold and bears Walker-A and Walker-B motifs involved in nucleotide binding and hydrolysis, respectively. The small, C-terminal subdomain is predominantly a-helical and forms a partial lid over bound nucleotide. The small subdomain can undergo rigid-body movements with respect to the large subdomain [3, 16, 50] and in many AAA+ proteins it contains amino acids that detect the contents of the nucleotide-binding site (e.g., via interaction of the Sensor II motif with the

g-phosphate). On ring assembly, the AAA+ architecture enables residues from an adjacent large subdomain to participate in the preceding subunit’s active site (e.g., via insertion of an “arginine finger”), providing a means for inter-module communication. The nature of AAA+ module interfaces also means that changes in orientation between large and small subdomains in one AAA+ module can be propagated to movement of adjacent modules in the ring [16].Because the nucleotide-binding sites are formed by residues from more than one AAA+ module it is reasonable to expect in dynein that the six AAA+ modules along the polypeptide chain are arranged structurally around the ring in the same order (1-6), although the sequences intervening the AAA+ modules are in principle long enough to allow other arrangements. EM visualization of GFP-tagged dynein motors has confirmed that AAA+ modules 5, 6, 1, and 2 are located adjacent to one another, strongly supporting the sequential organization of AAA+ modules within the ring (Fig. 4.3B). Hence, AAA1 is positioned in the ring diametrically opposite the stalk, placing dynein’s main site of ATP hydrolysis ~25 nm from the microtubule-binding domain at the end of the stalk. To put this distance into context, the ATPase and track-binding sites in the other cytoskeletal motors myosin and kinesin are an order of magnitude closer together.A homology model for dynein’s AAA+ modules has been built with a ring organization based on an oligomeric AAA+ hexamer (Fig. 4.1D) [33]. This model has proved a useful guide for engineering dynein and its hexameric

shape shows an overall similarity to EM images of dynein constructs comprising only the AAA+ region (though the EM images show greater asymmetry around the ring; Fig. 4.1C). While it is difficult to generalize about mechanisms in AAA+ proteins, especially for dynein which is an early branch of the superfamily, the evolutionary connection between microtubule-based motility and substrate remodeling enzymes is intriguing. The C-terminal sequenceC-terminal to AAA6 is a region of ~400 amino acids in most dynein heavy chains, which contains a mixture of predicted a-helices and b-sheets and is thought to form part of the head structure. The exact role of the C-sequence in dynein’s mechanism is currently mysterious, but it is known to regulate dynein ATPase and microtubule-binding activity [13, 21, 39]. Structurally, the C-sequence had been envisioned for some time to form an integral part of dynein’s ring, forming a seventh compact domain that bridges AAA1 and AAA6 in a heptameric ring. However, the finding that dynein’s AAA+ region alone forms a ring structure [39] has led to a revised model, in which the C-sequence interacts with the AAA+ ring rather than forming an intimate part of it (Fig. 4.3B). This is consistent with the fact that fungal dynein isoforms show robust motility despite having a naturally foreshortened C-sequence (~130 amino acids long) [11, 38]. Moreover, controlled trypsin proteolysis of the rat cytoplasmic dynein motor results in scission at a site equivalent to the fungal dynein C-terminus and the resulting fragments can be separated by gel filtration [21]. Thus, evidence suggests that the core of dynein’s motor is a ring of six AAA+ modules and the C-sequence forms an additional, interacting subdomain (or subdomains).Constraints for the location of the C-sequence within the head have come from mapping of BFP tags inserted in its sequence. These suggest that the C-sequence spans a region of the ring from AAA6 toward AAA5 and AAA4 (near the base of the stalk) and terminates near AAA6 (Fig. 4.3B). Truncation of the entire C-sequence causes increased structural variability in this region of the ring, suggesting that it may stabilize the structure [39]. The linker domainIn recent years, evidence has emerged for a rod-like structural element within the ~550 amino acids N-terminal to AAA1. Termed the “linker,” this domain connects the tail to AAA1 by taking a path across the AAA+ ring, likely interacting with one or more of the AAA+ modules (Fig. 4.3B). Accumulating data implicate the linker as dynein’s main mechanical element that amplifies conformational changes originating in the AAA+ modules (see Section 4.3).