Lucky 13 - microtubule depolymerisation by kinesin-13 motors

Microtubules (MTs) are polar polymers built from αβ-tubulin heterodimers that polymerise in a head-to-tail fashion to form protofilaments; these associate laterally to form the wall of the hollow MT polymer (Desai and Mitchison, 1997) (Fig. 1). Although MTs may be viewed as static structures around which cell biology moves, their dynamics are important for cellular function. MTs exhibit dynamic instability and undergo cycles of random growth and shrinkage. The transition from growth to shrinkage is called catastrophe and that from shrinkage to growth is called rescue. αβ-tubulin binds GTP and hydrolyses it to GDP. Whether GTP or GDP is bound controls MT dynamics and it is convenient to think of this in terms of heterodimer structure. GTP-tubulin favours MT growth because it has a straight conformation that enables incorporation into the lattice (Wang and Nogales, 2005), whereas GDP-tubulin favours catastrophe and depolymerisation because of its curved conformation (Ravelli et al., 2004). However, tubulin dimers are held within the MT lattice even when GDP is bound. Consequently, MTs only exhibit dynamics from their ends and it is here that cellular regulators exert their control.

Members of the kinesin superfamily of ATP- and MT-dependent motor proteins not only move cargo along MTs but also regulate MT dynamics and their best-described function in this capacity is MT depolymerisation. Kinesin-8, kinesin-14 and kinesin-13 all have depolymerisation activity, but kinesin-13 motors are the best characterised. Kinesin-13 class members were initially identified in screens for motors involved in spindle function (Wordeman and Mitchison, 1995; Walczak et al., 1996) and in neurons (Aizawa et al., 1992). Depletion/overexpression experiments revealed the importance of these motors as catastrophe factors (Walczak et al., 1996; Maney et al., 1998) and a seminal paper in 1999 characterised their MT-depolymerising function (Desai et al., 1999). Kinesin-13 activity can be required either simply to deconstruct MTs at specific cellular sites or to enable coupling of MT depolymerisation to movement of the cellular cargo to which the motor is attached (e.g. a chromosome). Here, we discuss recent work that has shed light on the depolymerisation mechanism and the roles of kinesin-13 motors in various cellular processes.

Kinesin-13 proteins are homodimeric catastrophe factors in which the kinesin motor domain, which contains the MT- and ATP-binding sites, lies in the middle of the amino acid sequence (Fig. 2A). Consequently, this class was originally termed Kin I (I for internal) or M-type (M for middle) kinesins (Vale and Fletterick, 1997; Miki et al., 2001; Lawrence et al., 2004). Sequence analysis of the kinesin-13 motor core shows it has high sequence similarity to all kinesins, and elements such as the nucleotide-binding motifs are well conserved (Fig. 2B). High-resolution structure determination confirms that this sequence conservation is reflected in the 3D structure of the motor core (Sablin et al., 1996; Ogawa et al., 2004; Shipley et al., 2004) (Fig. 3A) and electron microscopy (EM) studies demonstrate that there is a conserved interaction between the MT lattice and kinesin motors (Moores et al., 2003). N-terminal to the motor core is a kinesin-13 class-specific stretch of ∼60 highly charged residues known as the neck. N-terminal to the neck region is a 25-30 kD domain involved in subcellular localisation (Maney et al., 1998; Wordeman et al., 1999; Walczak et al., 2002; Kline-Smith et al., 2004). Constructs comprising this domain alone appropriately localise to the site of kinesin-13 function in vivo, independently of MTs. The C-terminus contains a predicted coiled-coil region but data indicate that both N- and C-termini are required for dimerisation (Maney et al., 2001).

The basis of kinesin-13 motor depolymerisation activity is the production of the bent, catastrophe-favouring conformation of tubulin at MT ends (Desai et al., 1999). Theoretically, this could occur either by selectively stabilising a bent tubulin conformation or by actively catalysing a straight-to-bent tubulin transition. MTs stabilised by paclitaxel or the non-hydrolysable GTP analogue GMPCPP have been used in many kinesin-13 studies because unstabilised (and therefore dynamic) MTs are difficult to work with (e.g. Desai et al., 1999; Maney et al., 2001; Moores et al., 2002; Hunter et al., 2003; Hertzer et al., 2006). That kinesin-13 motors can readily depolymerise these stabilised substrates supports the argument that kinesin-13 motors actively deconstruct their MT substrate rather than passively relying on inherent conformational fluctuations in the dimers found at MT ends (the `terminal tubulins'). However, whether this mechanistic conclusion can be applied to the dynamic MTs that are the in vivo substrates for these motors remains an open question.

The key step in the kinesin-13 MT-depolymerisation mechanism occurs as ATP binds, demonstrated by the unpeeling of stabilised MTs in the presence (and only in the presence) of the non-hydrolysable ATP analogue, AMPPNP (Desai et al., 1999; Moores et al., 2002; Shipley et al., 2004; Moores et al., 2006a). These bent tubulin structures form because the kinesin-13 motor is trapped in a pre-hydrolysis state that actively bends the underlying terminal tubulin. This supports the idea of direct coupling between ATP binding and terminal tubulin bending (Desai et al., 1999; Moores et al., 2002; Moores et al., 2006a). ATP hydrolysis is not required for tubulin release (Desai et al., 1999; Moores et al., 2002; Hunter et al., 2003) but catalytic depolymerisation requires that the ATPase cycle be completed for the motor to be released.

Because of the high degree of sequence conservation in the motor core, the location of the kinesin-13 motor within the middle of the protein sequence was initially thought to be responsible for depolymerisation activity. However, more recent experiments have shown that this property is intrinsic to the kinesin-13 motor core (Maney et al., 2001; Moores et al., 2002; Ovechkina et al., 2002). The kinesin-13 motor core, like all kinesin core structures, is shaped like an arrowhead (Sack et al., 1999; Shipley et al., 2004) (Fig. 3A). The motor core is built around a central eight-stranded β-sheet with three α-helices on each face. As the kinesin motor sits on the MT surface, the arrowhead tip points towards the MT plus end while the nucleotide-binding pocket near the base of the arrowhead faces towards the MT minus end (Fig. 3B).

Examination of the structure of the kinesin-13-MT interface suggests that the α4 helix of the motor - its principal energy-transducing element - is directly involved in depolymerisation because it sits directly over the tubulin intra-dimer interface and could deform it (Niederstrasser et al., 2002; Moores et al., 2003; Ogawa et al., 2004) (Fig. 3A,C). Mutation of kinesin-13-specific residues on or near the α4 helix simply weakens the motor-MT interaction. Mutation of conserved residues in loop 2 of the motor, by contrast, does not diminish MT binding, but these mutants are unable to perform the ATP-dependent tubulin straight-to-bent step of the depolymerisation cycle (Ogawa et al., 2004; Shipley et al., 2004) (Fig. 3C). It is probably the cooperation of kinesin-13-specific residues throughout the motor core that leads to depolymerisation activity.

Although the motor core defines the depolymerisation ability of the kinesin-13 class, it is a poor depolymeriser in comparison with larger constructs and does not appear to work at all in vivo (Ovechkina et al., 2002; Ogawa et al., 2004; Hertzer et al., 2006). By contrast, a monomeric neck+motor construct has depolymerisation activity similar to that of the full-length dimeric motor both in vitro and in vivo (Maney et al., 2001; Ovechkina et al., 2002; Hertzer et al., 2006). The neck region has a high positive charge, and point mutations that maintain this retain depolymerisation function (Ovechkina et al., 2002). The motor-proximal third of the neck sequence has been visualised at atomic resolution and forms an α-helix along the side of the motor domain, pointing towards the MT surface (Ogawa et al., 2004) (Fig. 3C). Recent work has demonstrated that the neck is likely to be important during the tubulin deformation step. We observed that neck+motor constructs do not release tubulin from MT ends on ATP binding - as is seen for the motor core (Moores et al., 2002) - but that tubulin protofilaments, nevertheless, form bent curls around the motor (Moores et al., 2006a). Our current hypothesis is that the neck acts as an additional tether to the MT wall but its exact contribution to depolymerisation efficiency in the context of the full-length dimer remains to be determined.

All kinesin-13 constructs from every species examined depolymerise MTs at both plus and minus ends, which implies an inherent apolarity in their mechanism. Given their depolymerisation activity, these motors must discriminate against binding sites on the MT lattice and preferentially associate with MT ends, their true substrates (Desai et al., 1999; Hunter et al., 2003). The mechanism of end recognition is unknown but, presumably, terminal tubulins have increased flexibility and distinct binding surfaces that facilitate binding (Fig. 4). In vitro experiments have shown that the full-length motor quickly finds MT ends (see below) but the kinesin-13 motor core is not efficient at this and readily decorates the MT lattice (Moores et al., 2003). The experiments that characterise lattice binding by the core are instructive, however, since they demonstrate that (at least for the core) the ATPase activity of the kinesin-13 motor is largely inhibited when the motor is bound to the lattice. Indeed, ADP release is induced on binding of the motor core to the lattice (Moores et al., 2003) but ATP cannot then bind to the motor. Inhibition of ATP binding has been observed kinetically (Hekmat-Nejad and Sakowicz, 2004) and is consistent with the absence of conformational change in the motor-MT complex on binding to AMPPNP (Moores et al., 2003). This supports the idea that ATP binding is coupled to bending of the terminal tubulin. End-specific ATPase activity might be attributable to the class-specific tertiary structure of the motor core, in which the convex kinesin-13 MT-binding surface seems to complement the presumed flexibility of terminal tubulins (Ogawa et al., 2004; Shipley et al., 2004) (Fig. 3C). Neck+motor constructs target MT ends but they do so less efficiently than the full-length motor (Hertzer et al., 2006), which again suggests that additional elements in either the N- or C-terminal domains enhance end recognition.

The ATPase activity of kinesin-13 motors is stimulated by individual αβ-tubulin heterodimers (Hunter et al., 2003; Moores et al., 2003; Hertzer et al., 2006). Dimeric tubulin is unlikely to be a true substrate for kinesin-13, so this probably reflects the importance of a flexible tubulin conformation within the ATPase cycle of the motors. In fact, the kinesin-13 motor core does not discriminate at all between MT ends and dimeric tubulin and its ATPase activity is stimulated equally by both (Moores et al., 2003). By contrast, the ATPase activity of both neck+motor and full-length constructs is stimulated approximately ten times more effectively by MT ends than by tubulin heterodimers (Hunter et al., 2003; Hertzer et al., 2006). This once again reinforces the idea that components that contribute to depolymerisation efficiency lie outside the motor core, residues at the extreme C-terminus of the full-length motor being strongly implicated in end discrimination (Moore and Wordeman, 2004a).

Kinesin-13 dimers move rapidly along the MT lattice by 1D diffusion (Hunter et al., 2003; Helenius et al., 2006) (Fig. 4). This movement occurs in the post-hydrolysis, ADP.Pi state of the motor. It does not require additional rounds of ATP binding and hydrolysis and is, of course, more efficient than 3D diffusion in allowing the motor to get to MT ends. Importantly, this 1D diffusion is not directional and therefore does not favour one end of the polar MT. However, because of this non-directionality and because the motor can change direction during a single diffusive interaction, the motor will not necessarily ever find the end of longer MTs. The kinesin-13 N-terminal domain targets the protein to specific sites in the absence of MTs in vivo, so localisation to MT ends could occur independently of such lattice-based diffusion (Maney et al., 1998; Walczak et al., 2002). However, kinesin-13 molecules have been observed along MTs in cells and cell extracts (Walczak et al., 1996; Wordeman et al., 1999; Moore and Wordeman, 2004a), which suggests that transient interaction with the lattice is important in vivo.

Why are these motors dimeric? Dimer formation allows motile kinesin-1 motors to step processively along MT tracks, enabling long-range cargo transport (Vale and Milligan, 2000). By contrast, why kinesin-13 needs to be dimeric is unclear. A recent study showed that the affinity for tubulin heterodimers of a monomeric neck+motor construct is higher than that of the dimeric full-length motor (Hertzer et al., 2006). Dimeric kinesin-13 motors might therefore be optimally efficient depolymerisers because they can more readily detach from the products of depolymerisation (tubulin dimers). However, it is also plausible that the availability of two heads enables processive kinesin-13-catalysed depolymerisation - i.e. removal of successive tubulin dimers by a single kinesin-13 molecule (Hunter et al., 2003; Helenius et al., 2006). Processive dimeric kinesin-13-catalysed depolymerisation of GMPCPP- and paclitaxel-stabilised MTs has been observed, although the number of ATP molecules required per terminal tubulin depends on the means of stabilisation (Hunter et al., 2003). These observations again speak to the question of whether artificially stabilised MTs reflect these enzymes' true, dynamic substrate - for example, in a dynamic MT there might be a stabilising cap of only one to two GTP-tubulins per protofilament and so only these would need to be removed by kinesin-13 to initiate catastrophe. Processivity may also vary between different kinesin-13 motors. However, processivity would certainly have the mechanistic advantage that the motor remains attached to its substrate and presumably allows any cellular cargo (e.g. chromosomes) to maintain contact with the MT too.

Aspects of the MT substrate have also been examined. The acidic tubulin C-terminal tails (CTT) are important for processivity of motile kinesins and for dynein (Wang and Sheetz, 2000) and they are also important for various aspects of kinesin-13-catalysed depolymerisation. The β-tubulin CTT is required for tubulin bending by the motor core but removal of these residues does not affect lattice binding (Moores et al., 2002). CTTs are also involved in the depolymerisation mechanism of full-length kinesin-13 but their importance for the full-length motor is less clear. 1D diffusion and, consequently, depolymerisation is slowed when the CTTs (α and/or β) are removed (Helenius et al., 2006), whereas other work has shown that the CTTS are essential for depolymerisation by the dimeric motor (Niederstrasser et al., 2002). Whether the requirements for the CTTs by different constructs reflect related aspects of kinesin-13 function is uncertain.

In many cases, the reported effects of overexpression and/or depletion of a kinesin-13 on other MT regulators (and vice versa) are indirect and occur because the finely balanced forces present in vivo are disrupted (see below). However, several direct effectors/antagonists of kinesin-13 function have been characterised in simpler in vitro systems. The MCAK activator ICIS (inner centromere Kin I stimulator) localises to the centromere in a kinesin-13-dependent fashion (Ohi et al., 2003). ICIS itself binds to MTs and stimulates the already potent depolymerising activity of kinesin-13 in vitro through an unknown mechanism. In vivo, its function may be linked to the specific localisation and regulation of depolymerising activity (see below). By contrast, the proposed MT-stabilising activity of XMAP215, a well-conserved MT-associated protein (MAP), has been shown to counter kinesin-13 depolymerisation, and the combined activities of XMAP215 and kinesin-13 alone can generate MT dynamics similar to those seen in vivo (Tournebize et al., 2000; Kinoshita et al., 2001; Holmfeldt et al., 2004; Noetzel et al., 2005). The nature of the XMAP215-MT interaction is complex and this may be reflected in its ability to balance rather than completely block kinesin-13 activity.

The classical neuronal MAP tau inhibits kinesin-13 (Noetzel et al., 2005). Tau is thought to stabilise MTs by binding along protofilament crests (Al-Bassam et al., 2003; Santarella et al., 2005) and therefore tau and its relatives probably competitively block the kinesin-13-binding site. Whether this occurs at MT ends, where the effect on kinesin-13 depolymerisation will presumably be direct, or tau just inhibits the motor's diffusive search for MT ends along the lattice is not yet known. Tau requires tubulin CTTs for a strong interaction with MTs (Santarella et al., 2004), which further reinforces the overlap of binding requirements with kinesin-13. By contrast, the neuronal MT-stabilising protein doublecortin does not block kinesin-13 activity, probably because it binds in between the MT protofilaments and so does not impede the MT-kinesin-13 interaction (Moores et al., 2004; Moores et al., 2006b).

The generic depolymerisation mechanism discussed above is thought to be broadly applicable to all kinesin-13 motors, but more detailed comparisons will probably reveal mechanistic refinements for individual class members (Ogawa et al., 2004; Mennella et al., 2005; Moore et al., 2005). Indeed, when one considers the cellular roles of these kinesins, the picture becomes very complicated. Roughly equivalent class members can be identified in different organisms (Table 1) but the functional parallels frequently break down because the experimental systems and the roles of the individual kinesin-13 motors, even among vertebrates, are not identical. Nevertheless, the evidence for the involvement of kinesin-13 motors in both global and local control of MT dynamics is clear, and we outline this, in general terms, below.

The MT cytoskeleton undergoes a dramatic transformation when cell division begins. The characteristic interphase radial MT array completely disappears and the highly dynamic bipolar mitotic/meiotic spindle forms (Fig. 5). Kinesin-13 motors are found, in part, as a soluble pool in cells. Perturbation of this pool disrupts the spindle, which is consistent with a role for kinesin-13 in global regulation of MT dynamics (Walczak et al., 1996; Maney et al., 1998; Kline-Smith and Walczak, 2002; Rogers et al., 2004). The spindle is maintained by the balance of many forces. Its complexity makes functional dissection difficult, because many compensating mechanisms prevent failure of this important cellular structure. Nevertheless, key and, in some cases, conserved functions have been assigned to kinesin-13 class members.

Probably the most clear-cut role is at spindle poles and reflects the importance of these in initiation of spindle bipolarity. Depletion of kinesin-13 motors in multiple organisms results in formation of monopolar spindles or unusually long, MT-dense bipolar spindles (Walczak et al., 1996; Goshima and Vale, 2003; Gaetz and Kapoor, 2004; Ganem and Compton, 2004; Rogers et al., 2004; Goshima et al., 2005; Zhu et al., 2005; Laycock et al., 2006). These phenotypes are consistent with the need for regulated MT growth at spindle poles. Pole-based kinesin-13 motors may also act in anaphase to depolymerise the minus ends of kinetochore MTs, contributing to poleward flux and chromatid separation (Rogers et al., 2004; Rogers et al., 2005). However, the role of poleward flux in anaphase is very system-dependent and, consequently, the generality of this aspect of kinesin-13 function is unclear (Ganem et al., 2005).

At the other end of spindle MTs, kinesin-13 motors are also found at the kinetochore and several roles are ascribed to them at this location. The kinetochore consists of distinct domains - the centromeric, DNA-proximal region, the inner kinetochore and the outer kinetochore plate, where spindle MTs attach (reviewed in Maiato et al., 2004; Moore and Wordeman, 2004b; Chan et al., 2005; Parra et al., 2006). Kinesin-13 motors are believed to be active at all these sites. Depletion of centromeric kinesin-13 (by overexpression of a construct containing only the N-terminal localisation domain) results in several distinct phenotypes. In mammalian cells (hamster), MCAK depletion causes a trailing-chromatid defect in anaphase, which supports a model in which MCAK progressively depolymerises spindle MTs from their plus ends and drags the chromatid to which they are attached towards the spindle poles during anaphase (the `Pac-Man' model) (Maney et al., 1998; Sharp and Rogers, 2004). Rogers et al. have proposed a similar function for KLP59c on the basis of the effects of its depletion in Drosophila (Rogers et al., 2004), in which the rate of chromatid separation during anaphase is reduced.

A role in correction of inappropriately attached pre-metaphase chromosomes has also been assigned to centromeric kinesin-13 motors (Kline-Smith et al., 2004). Depletion of kinesin-13 from centromeres generates lagging chromatids, despite normal rates of chromatid movement (rat kangaroo MCAK in PtK2 cells). In these depleted cells, correlated light and EM studies showed that incorrectly attached MTs ran laterally along the surface of kinetochores on unaligned chromatids. This suggests that kinesin-13 at the centromere prunes away inappropriately attached MTs (merotelic and syntelic) and allows correct chromosome congression and subsequent accurate anaphase. Some aberrant MT attachments do not significantly perturb the tension across the sister chromatids and, therefore, are not detected by the spindle checkpoint. Anaphase proceeds even when such incorrect attachments are present so this error correction by kinesin-13 is particularly important.

Regulation of kinesin-13 motors (specifically MCAK) at the kinetochore by the aurora B kinase is also a focus of attention (Andrews et al., 2004; Lan et al., 2004; Ohi et al., 2004; Parra et al., 2006). Aurora B has a general role promoting chromatid biorientation and it phosphorylates MCAK on several serine/threonine residues, including within the neck, thereby inactivating it. Aurora B activity is required for localisation of MCAK to the kinetochore and controls its movement between kinetochore subdomains through prophase and metaphase. Unaligned chromatids that are not under tension have a rapidly exchanging population of phosphorylated and dephosphorylated MCAK at their centromeres, whereas dephosphorylated MCAK is located towards the outer kinetochore in aligned chromatid pairs under tension. Thus, MCAK activity is dynamically controlled and localised. These MCAK sub-populations presumably perform specific activities, such as MT attachment error correction at the centromere during prometaphase and force generation during anaphase at the kinetochore. Such subtleties of regulation may also be controlled by ICIS (Ohi et al., 2003). All that is certain is that the known regulatory mechanisms controlling kinesin-13 location and activity represent the tip of the iceberg.

Within the spindle, the different kinesin-13 activities do not operate independently; indeed, different compensatory effects have been observed when DNA- or pole-based kinesin-13 activities are perturbed. For example, in Drosophila, KLP10A and KLP59C work co-ordinately at either end of kinetochore MTs to bring about accurate chromosome segregation during anaphase (Rogers et al., 2004). If both Kif2a and MCAK are removed in vertebrate cells, by contrast, poleward flux is eliminated but mitotic progression is unperturbed. However, lagging chromatids are observed in anaphase, presumably because the centromere-based, error correction activity of MCAK has also been eliminated from these spindles (Ganem and Compton, 2004; Ganem et al., 2005). Thus although the individual roles of kinesin-13 motors are gradually being revealed, when we consider the combination of force-generators within the spindle, the picture is still very complex and varies from organism to organism.

Attention has recently turned to the activities of kinesin-13 motors during interphase. Their role in control of global MT turnover has already been mentioned, but there are also more subtle regulatory mechanisms at work and this is a fast-developing area. A recent study, for example, has revealed that a kinesin-13 (KIF2A) plays an essential role in brain development in vivo by controlling neurite outgrowth (Homma et al., 2003). Additionally, recent studies have highlighted the association of kinesin-13 motors with the plus ends of growing MTs as they approach and interact with the cell periphery (Menella et al., 2005; Moore et al., 2005). This apparent paradox seems to occur through recruitment of kinesin-13 by MT plus-end tip tracking proteins (+TIPs) (Banks and Heald, 2004; Akhmanova and Hoogenraad, 2005; Menella et al., 2005; Moore et al., 2005). Presumably, kinesin-13 motors ride growing MT plus ends in an inactive state until the appropriate cellular cue activates them. Individual kinesin-13 motors appear to be regulated in different ways, and it will be interesting to discover whether their different roles reflect genuine differences in their enzymatic activity or are the consequence of modulation by other MT-binding proteins.

As mentioned above, MT depolymerisation activity has also been observed in the case of kinesin-8 and kinesin-14 classes (reviewed in Ovechkina and Wordeman, 2003). Depolymerisation by kinesin-14 motors, traditionally seen as minus-end-directed motors, has been described from one or other MT end but is inhibited on stabilised MTs (Endow et al., 1994; Chu et al., 2005; Sproul et al., 2005). The depolymerisation activity of kinesin-8 (shown to have plus-end-directed motility) (Pereira et al., 1997) was proposed on the basis of the effect of knockout and knockdown experiments on cell division in vivo. Here, the reduction of kinesin-8 activity has a deleterious effect on spindle bipolarity (Goshima and Vale, 2003; Zhu et al., 2005) and chromosome congression and separation (West et al., 2001; Garcia et al., 2002; West et al., 2002). Detailed studies, such as those published recently (Gupta et al., 2006; Varga et al., 2006), are required if we are to understand the role of depolymerisation by these motors and, in particular, how they coordinate directional motility and depolymerisation. Data concerning kinesin-13 motors provide a framework for functional and mechanistic comparison. What is particularly tantalising about depolymerising kinesins is that yeast genomes (S. cerevisiae and S. pombe) encode both kinesin-8 and kinesin-14 motors but not kinesin-13. Do kinesin-8 and kinesin-14 motors perform the functions in yeasts that kinesin-13 motors do in higher eukaryotes, or are the MT cytoskeletons of higher eukaryotes sufficiently different that kinesin-13 motors have evolved specifically to perform the necessary roles? Future studies will address these important questions.

Ultimately, the key to defining the roles of kinesin-13 in vivo will hinge on understanding the context in which they operate and the impact of other MT-associated factors on their activity. In the meantime, there is plenty still to uncover about their molecular mechanism, particularly the role of the dimer in the depolymerisation mechanism. While it is now more than 20 years since the activity of kinesin motors was first defined (Vale et al., 1985), it is less than 10 years since the depolymerising kinesin-13 motors were characterised (Desai et al., 1999) and it is encouraging to see how far we have come.

We thank our kinesin-13 collaborators in Linda Wordeman's lab (University of Washington, Seattle, WA) and at Cytokinetics Inc. (San Francisco, CA) for exciting experiments and stimulating discussions. C.A.M. is a BBSRC David Phillips Fellow and R.A.M. is supported by an N.I.H. grant (GM52468).