Cytoskeletal motors include myosins, kinesins and dyneins. Myosins move along tracks of actin filaments, whereas kinesins and dyneins move along microtubules. Many of these motors are involved in trafficking cargo in cells. However, myosins are mostly monomeric, whereas kinesins are mostly dimeric, owing to the presence of a coiled coil. Some myosins (myosins 6, 7 and 10) contain an SAH (single α-helical) domain, which was originally thought to be a coiled coil. These myosins are now known to be monomers, not dimers. The differences between SAH domains and coiled coils are described and the potential roles of SAH domains in molecular motors are discussed.
There are three groups of cytoskeletal motors: myosins, kinesins and dyneins. Myosins move along tracks composed of actin filaments, whereas kinesin and dynein move along microtubule tracks. In most interphase cells, microtubule tracks radiate out from a central structure, the centrosome, and grow towards the plasma membrane. Actin filaments have a more complex organization, with a dense and com-plex organization close to the plasma membrane, and sparser organization elsewhere. They can also be bundled into stress fibres or stress-fibre-like structures towards the basal surface, and underneath the nucleus in crawling cells. Generally, vesicles and other cargo need to move out towards the plasma membrane from the Golgi, close to the nucleus, and inwards from the plasma membrane towards the nucleus. Long-distance inward transport is co-ordinated by dynein, and outward movement by kinesins. Myosins are generally involved in shorter range trafficking at the plasma membrane or in other actin-enriched areas such as the Golgi.
Some molecular motors contain regions of coiled coil in their amino acid sequence. Coiled coils are a well-described structural feature found in many proteins. They are composed of two α-helices each of which contain a series of heptad repeats (a-b-c-d-e-f-g)n in which the ‘a’ and ‘d’ residues tend to be apolar or hydrophobic (e.g. leucine, isoleucine or valine), and the remaining residues favour helix formation. To form the coiled coil, the two α-helices self-associate with bury the hydrophobic residues at the centre of the molecule. In molecular motors, such as myosins and kinesins, coiled coils have a particular significance in that the coiled coil is the main structural feature known to dimerize the motors. In the resultant dimeric protein, there are two motor domains giving the molecule the potential to take multiple steps as each motor interacts with its track in sequence. It is therefore important to understand coiled coils in these motor proteins, and how they dimerize the motors.
In humans, there are 39 genes that code for myosin, organized into 12 different classes (Figure 1A; ). Of these, only three classes contain an α-helical coiled-coil domain (myosin 2, myosin 5 and myosin 18). In these classes, the myosins have been shown to form dimers. In the case of myosin 2, patterns of alternating charge distributed along the coiled coil additionally allow this class of myosin to form filaments. Three further classes (myosin 6, myosin 7 and myosin 10) were originally thought to contain a coiled-coil domain, based on sequence analysis only. Coiled-coil prediction programmes such as pepcoil or coils, predict that there is a region of coiled coil just distal to the motor and lever in these myosins. However, the coiled-coil prediction algorithms mistake sequences rich in charged residues for coiled coils . In the case of these three classes of myosins, a close inspection of the sequence revealed that they are all highly enriched in charged residues and contain very few hydrophobic residues, inconsistent with coiled-coil formation .
The myosin and kinesin family trees for humans
It has now been established that the predicted coiled-coil domains of myosins 6, 7 and 10 do not form coiled coils, but instead contain stable SAH (single α-helical) domains, and that all three of these myosins are monomeric. Myosin 10 was first shown to contain an SAH domain instead of a predicted coiled coil . A peptide (37 amino acids long) from the N-terminal of the predicted coiled coil was shown to be almost 100% helical and monomeric in solution. In addition, a construct of myosin 10 containing the motor, lever and the entire predicted coiled-coil sequence was also shown to be mostly monomeric in vitro. Importantly, even in the rare dimers, dimerization involved only the extreme C-terminus of this construct. An SAH domain, approximately 15 nm long (equivalent to 100 residues at a rise of 0.15 nm per residue, as observed for α-helices) was observed in both monomers and dimers . It was also suggested in this report that the predicted coiled coil of myosins 6 and 7 were also likely to be SAH domains from sequence homology between the predicted coiled-coil domains of these myosins. We now know that purified full-length molecules of myosins 6, 7 and 10 are indeed monomeric in vitro [4–6], and that the predicted coiled-coil domain of myosin 6 is an SAH domain .
SAH domains are unusual in that these domains are highly stable in solution in isolation. In contrast, other α-helices are usually not stable in isolation, but need to be stabilized by their interaction with other parts of the same protein or by binding to other proteins [as is the case for the α-helical lever in all myosins, which is stabilized by binding to light chains/CaM (calmodulin)]. The stability of SAH domains originates in ionic bonds formed between oppositely charged side chains in successive turns of the α-helix (Figures 2A and 2B). These both act as ‘staples’ to resist unfolding of the helix and to screen the hydrogen bonds between the polypeptide backbone amide groups from attack by water. The side chains of arginine and lysine residues preferentially bond with glutamate side chains four residues downstream (i,i+4) or three residues upstream (i,i−3), whereas glutamate preferentially interacts with arginine and lysine residues at (i,i+3) and (i,i−4). The distributed charge on the guandinium headgroup of arginine is probably especially effective as it can bond simultaneously with glutamate residues downstream and upstream. H bonds between glutamate and glutamine residues at (i,i+4) and can also add stability to the SAH domain (reviewed by ). The SAH domains of myosins 6, 7a and 10 are rich in these stabilizing interactions (Figures 2A and 2B).
SAH domains in myosins 6, 7a and 10
The position of the SAH domain, just after the motor and lever in myosins (Figures 2C and 2D) raised the possibility that it could extend the length of the lever. This would allow myosins with SAH domains to take longer steps than possible with the canonical lever arm only consisting of CaM bound to IQ motifs in an α-helix. This was tested by replacing part of the lever in myosin 5 with an equivalent length of SAH domain to determine if this chimaeric construct could walk processively along actin as in the wild-type construct . The SAH domain was able to functionally replace the lever. The chimaeric myosin 5 (containing two IQ motifs and a length of SAH domain equivalent to the remaining deleted four IQ motifs) takes similar-sized steps to wild-type myosin 5 as it moves along actin both in TIRFM (total internal reflection fluorescence microscopy) assays and in the optical trap. However, unlike wild-type myosin 5, the ATPase of the chimaeric myosin 5 is not gated, suggesting that the SAH domain does not sufficiently transmit strain between the two motor domains to force gating of ADP release. Calculations based on a small reduction in step size measured for the chimaera in the optical trap were consistent with an estimated 10-fold lower bending stiffness for an SAH domain compared with a canonical myosin lever (IQ+CaM) . This reduction in bending stiffness may help to account for the lack of gating in the chimaera. In SAH domain-containing myosins, which are monomeric, the lack of gating is not problematic, as there is no requirement for strain to be transmitted between two motor domains to co-ordinate their movement. In dimeric motors, gating is important to co-ordinate the activity of the two motor domains, and it reduces the probability of non-productive steps, where the lead head detaches and rebinds at the same position on actin.
A second major family of molecular motors in cells are the kinesins, which in humans are encoded by 41 genes, organized into 14 classes (Figure 1B). In contrast with myosins, a coiled coil dimerizes the majority of kinesins. Only some isoforms, KIF26A and KIF26B (kinesin-11) and KIF24 (kinesin-13) do not appear to contain a region of predicted coiled coil in their sequences. The kinesin-11 class is highly divergent, and its motor domain lacks ATPase activity . Kif24 is involved in ciliogenesis . As well as homodimers, some kinesins form tetramers (kinesin-5) or heterodimers (kinesin-2). The dimeric nature of many, but not all of the kinesins has been demonstrated (e.g. kinesin-1, kinesin-2 and KIF10). In others, the dimeric state is only predicted by sequence analysis (e.g. Kif16B).
Remarkably short sequences of coiled coil are sufficient to dimerize some kinesins (Figure 3). Conventional kinesin-1 contains an extended coiled-coil sequence. However, a short sequence of coiled coil just after the motor domain (residues 337–369) in a motor domain construct (residues 1–379) is sufficient to dimerize the motor domain . An isolated peptide of the N-terminal sequence of this short region of coiled coil (residues 332–349) dimerizes weakly (Kd 9.6 mM). The subsequent C-terminal sequence (350–379) dimerizes more strongly (Kd 60 μM), while a peptide containing both regions (332–369) dimerizes most strongly (Kd 62 nM) . The N-terminal sequence is rich in charged interactions similar to those found in SAH domains, and while it mostly contains hydrophobic or apolar residues in ‘a’ and ‘d’ positions, it does not contain any of the preferred residues (leucine, isoleucine or valine). Given the weakness of the hydrophobic seam, it has been suggested that the interactions between glutamate and lysine residues outside the hydrophobic seam could stabilize the two α-helices, allowing the coiled coil to separate into its two individual helices in this region without collapsing to random coil . The C-terminal sequence does contain the preferred isoleucine and leucine residues in a and d positions respectively and, although this region of coiled coil is relatively short (five and half turns), it is able to dimerize a motor domain construct.
The coiled-coil domains adjacent to the motor domain in three classes of kinesin: kinesin-1, -2 and –3
The start and end of this coiled coil in kinesin-1 also contains N- and C-capping boxes that help prevent fraying at the ends of the coiled coil. The N-capping box motif consists of seven residues (N″-N′-Ncap-N1-N2-N3-N4, where residues N1–4 belong to the helix, Ncap is the boundary residue and N′ and N″ precede the helix) . The C-capping motif consists of 10 residues (C4-C3-C2-C1-Ccap-C′-C″-C3′-C4′-C5′; , where residues C1–4 are inside the helix, Ccap is the boundary residue and residues C′ to C5′ are outside the helix. Capping boxes at the start and end of helices tend to contain a high frequency of specific residues. At the N-terminus, the side chain of the residue in the Ncap position hydrogen bonds with backbone NH residue of N3 downstream, and conversely the side chain of the N3 residue hydrogen bonds with backbone NH of the Ncap residue [13,16]. Similarly C-capping boxes contain common Ccap residues , and C″, C3′ or C4′ form hydrophobic interactions with C3 or C4 inside the helix. Glycine and asparagine are both common residues that mark the end of the helix. Hydrophobic staples can also be present  and help stabilize the coiled coil.
Intriguingly, Unc104 (kif1A) in the kinesin-3 class was originally thought to be monomeric, but is now known to dimerize via a short region of coiled coil . Inspecting the sequence of the short region of coiled coil in one isoform of kinesin-2 (KIF3A) (Figure 3B) and in four different kinesin-3 isoforms (Figures 3C–3F), and comparing them with the well-characterized coiled coil of conventional kinesin (kinesin-1 or Kif5; Figure 3A) shows some interesting conserved features (Figure 3). First, the a positions tend to be occupied by isoleucine, valine or leucine, and the d positions by leucine. Kif3A contains a less favoured phenylalanine in one of the d positions, but this region of coiled coil also contains many charged residues outside the core of the coiled coil that could promote helix formation. N- and C-capping motifs as described  that prevent fraying of the ends of the α-helices can be identified in all of the kinesin isoforms. In addition, oppositely charged residues in e and g positions, which can also help stabilize coiled coils, are present. Thus even though the regions of coiled coil in the kinesin-2 and in the different kinesin-3 isoforms are short, they contain many features that potentially promote and stabilize coiled coil formation.
Given that short stretches of coiled coil can dimerize kinesins, is it possible that short stretches of potential coiled coils might be present in myosins 6, 7 and 10 could still dimerize these molecules? This has been suggested to occur for myosin 6  via a short stretch of potential coiled coil between the end of the three helix bundle [or LAE (lever arm extension)] and the start of the SAH domain (Figure 2). The LAE is devoid of any charged interactions, and the boundary between the LAE and the SAH domain is clear, as the position where charged interactions start to be observed (Figures 2A and 2B). The potential region of coiled coil that has been postulated for myosin 6 is at the start of the SAH domain (Figure 2B). This region is rich in charged residues (glutamate, lysine and arginine) that would stabilize a single α-helix. However, in this sequence, there is only one preferred residue (leucine) in either an a or d position in the two and a half turns of coiled coil. The remaining a and d residues are glutamine, alanine and methionine, which are statistically much less common and less well favoured in coiled coils. Therefore it seems unlikely that this short stretch of potential coiled coil will dimerize, except perhaps at very high concentrations. The sequence of this region of postulated coiled coil is somewhat similar to the weakly dimerizing region at the start of the coiled coil in kinesin-1 (Figure 2A), and therefore one might expect that the Kd of the postulated region of coiled coil in myosin 6 is similarly low (millimolar).
The potential presence of a region of very weakly dimerizing coiled-coil in myosin 6 could explain why it can be induced to dimerize at very high concentrations, when it is bound to actin. Only approximately 10% of molecules dimerize under these conditions, and the dimers quickly dissociate . This also raises the question as to the physiological relevance of the apparent unfolding of the LAE domain in myosin 6 dimers  (Figure 2D), if dimerization is so rare. It will be interesting to determine if the LAE in myosin 6 can unfold in the monomer and whether this uncouples the SAH domain from the motor and canonical lever affecting the working stroke of myosin 6. Given the low concentrations of myosin 6 in a cell, it is unlikely that myosin 6 would dimerize via a short weak coiled coil in vivo. Myosin 6 does not appear to be required to dimerize for its cellular function . It may cluster and dimerize/oligomerize on vesicles via an interaction of the cargo domain with both PtdIns(4,5)P2 and Dab2, although Dab2 does not dimerize myosin 6 .
The SAH domain of myosin 7a does not appear to contain any significant stretch that would favour dimerization. However, the SAH domain of myosin 10 contains a very short stretch of potential coiled coil at its C-terminal end, where a short hydrophobic seam containing leucine and alanine residues is surrounded by a region of highly charged helix-stabilizing interactions, reminiscent of the weakly dimerizing N-terminal region in kinesin-1. Note, however, that the e and g residues in this region are all acidic, which would destabilize this as a dimer. Most recently, it has been shown that full-length myosin 10 is monomeric , and its activity regulated by a head–tail interaction as described for myosin 7 [5,24]. Intriguingly, this recent report suggests that myosin 10 is induced to dimerize when it binds to PtdIns(3,4,5)P3 via its PH (pleckstrin homology) domains, although the mechanism by which it does so was not made clear .
In conclusion, it is important to determine empirically whether myosins dimerize or not. Regulation of dimerization can influence the activity of myosins in vivo as a monomer cannot walk processively along actin filaments, whereas dimers have the potential to do so. However, multiple monomeric myosins on a vesicle are effective in moving a vesicle (or bead) long distances . Many myosins function as monomers in the cell (e.g. myosin 1), and there is no suggestion that these molecules might dimerize. Unlike kinesins, it appears unnecessary for the majority of myosins to dimerize in order to perform their cellular functions, including short-range vesicle trafficking along the actin network. This may be because these myosins have less far to travel, and have a complex actin network to navigate, compared with the long straight microtubule tracks used by kinesins. It may also be related to the regulation of these molecules. A single kinesin molecule could traffic its cargo a long way, whereas monomeric myosins need to work in small teams. The ability to regulate the number of myosin molecules attached to a cargo, might work as an effective switch between tethering (small numbers of myosins) to trafficking (many molecules) at least for monomeric myosins. In conclusion, while it is now clear that most myosins work effectively as monomers, we still have much to learn about their regulation and function in vivo.
Cellular Cytoskeletal Motor Proteins: A Biochemical Society/Wellcome Trust Focused Meeting held at Wellcome Trust Genome Campus, Hinxton, Cambridge, U.K., 30 March–1 April 2011. Organized and Edited by Folma Buss (Cambridge, U.K.) and John Kendrick-Jones (MRC Laboratory of Molecular Biology, Cambridge, U.K.).
I am much indebted to Professor Peter Knight for reading and commenting on an earlier version of this paper.
The work related to SAH domains was funded by the Biotechnology and Biological Sciences Research Council.