y. Cell Sci. Suppl. 5, 121-128 (1986)Printed in Great Britain © The Company of Biologists Limited 1986

121

THE ROLE OF MICROTUBULE POLARITY IN THE MOVEMENT OF KINESIN AND KINETOCHORES

T . J . M IT C H IS O NNational Institute for Medical Research, London NW 7 1AA, UK

S UMMARYMicrotubules are important for organizing and directing many types of intracellular motility.

Recently progress has been made in the analysis of two types of motility at the molecular level: the movement of axonal vesicles driven by kinesin, and the movement of chromosomes driven by the kinetochore. Both require ATP for movement in vitro. Kinesin-driven movement is unidirectional, towards the microtubule plus end, while movement of the kinetochore is bidirectional. These similarities and differences are discussed and incorporated into a new model for the kinetochore- microtubule interface.

I N T R O D U C T I O NThe internal contents of cells are in a state of constant motion, but their overall

predictable morphology and asymmetry implies that this movement must be subject to spatial organization over large distances. Microtubules are linear elements that follow rather straight courses over long distances through cytoplasm, and are thus suited to coordinating intracellular transport. Since their discovery, it has been appreciated that microtubules play an important role in directing the movement of chromosomes during mitosis, and of other organelles throughout the cell cycle (Porter, 1966). Until recently the only system in which microtubule-dependent motility could be analysed at the molecular level was the ciliary axoneme where the favourable spatial organization and high local concentration permitted the identi­fication of dynein as an ATPase that produces force against microtubules (Gibbons, 1981). More recently, the combined application of video-enhanced differential interference contrast microscopy (Allen et al. 1981) and biochemistry to squid axoplasm has permitted the isolation of a novel force-producing ATPase, kinesin (Vale et al. 1985a,c). The classic problem of chromosome movement has also come under renewed attack, and although we are still far from a molecular analysis, it is now possible to discuss the movement of the kinetochore along the microtubule at a mechanistic level. In this article I will review recent progress on the ways in which kinesin and kinetochores move along micro tubules, and in particular the influence of the polarity of the microtubule lattice. The comparison may suggest some new approaches to these mechanistically related but biologically very different problems.

122 T. J. MitchisonM I C R O T U B U L E P OLARI TY

Microtubules are formed by helical polymerization of their subunit protein, tubulin (reviewed by Kirschner, 1978). Such a helical polymer is necessarily polar, because the asymmetric subunits are all aligned in the same direction along the lattice. In fact tubulin is a dimer of similar a and ¡3 subunits, so the micro­tubule could be pseudo-bipolar if a and ¡3 were arranged head-to-head in the dimer. However, the strong 4nm repeats in the microtubule lattice, evident from X-ray diffraction (Cohen et al. 1971) and electron microscopy (Amos et al. 1976; Mandelkow et al. 1986) suggest that cvand /? are arranged in the same direction in the lattice, and head-to-tail in the dimer. Thus if one thinks of the amino acids exposed by each subunit on the micro tubule surface as forming a molecular arrow, then all the arrows are pointing in the same direction. As well as being expressed on the surface lattice, this polarity means that the two ends of the microtubule are different, probably with all a subunits exposed on one end, and all ¡5 on the other; though which is the plus end is not yet known.

Assays for the polarity of microtubules have utilized both differences at the ends of the lattice, manifest as different growth rates (Bergen & Borisy, 1980), and visual­ization of polarity in the surface lattice by the asymmetric binding of further tubulin oligomers (hook decoration; Heidemann & McIntosh, 1979), or dynein (Haimo et al. 1979). The hook decoration assay has been widely applied to microtubules in cells, showing that under most circumstances the plus, or fast-growing end, is distal to the cell centre (Euteneuer & McIntosh, 1981a). The uniform polarity of cytoskeletal microtubules means that they can determine not just the tracks that organelles follow, but also the direction of movement (McNiven et al. 1984). The importance of direction is well illustrated by the axon, where completely different classes of organelles are transported towards and away from the cell body (Grafstein & Forman, 1980). Microtubules running parallel to the axon, with uniform polarity (plus ends distal to the cell; Heidemann et al. 1981), probably direct this traffic (see below).

The mechanism that controls the polarity of microtubules within cells is not known, but specific nucleating structures probably play an important role. The centrosome nucleates microtubules with their plus end distal in vitro (Bergen et al. 1980). The reason for this is unknown, but it seems likely that a component of the centrosome must recognize and effectively cap the minus end. The kinetochore can also nucleate in vitro (Telzer et al. 1975), but apparently lacks a specific end inter­action since microtubules grow out with both polarities (Mitchison & Kirschner, 1985a). Mechanisms for determining polarity other than nucleation at centrosomes must exist, however, to account for its specification in microtubules not attached to the centrosome, such as those in some axons (Chalfie & Thomson, 1979). In addition, the reorganization of microtubule polarity in severed melanophore arms (McNiven et al. 1984) suggests that the microtubule cytoskeleton may be most stable with plus ends distal to the cell centre even in the absence of conventional nucleating sites.

Movement ofkinesin and kinetochores 123K I N E S I N : D I R E C T I O N A L I T Y OF A P U R I F I E D TRA NSL OC AT OR

Kinesin is a protein that was purified from squid axoplasm on the basis of its ability to promote latex bead movements along taxol-stabilized microtubules and movement of the microtubules relative to the glass substrate (Vale et al. 1985a). The properties of kinesin are most easily explained by a model in which the molecule can bind non- specifically to negatively charged surfaces, and specifically to the surface of the microtubule through an active site. In the presence of ATP the molecule can then walk along the micro tubule lattice, presumably hydrolysing the nucleotide, although this has not yet been demonstrated.

Kinesin-promoted movement appeared to be unidirectional, since all beads moved in the same direction along a single microtubule, which itself moved in the opposite direction along the substrate (Vale et al. 1985c). In order to determine the direction of movement relative to the polarity of the lattice, it was necessary to observe it on a microtubule substrate of known polarity. This was provided by astral arrays nucleated in vitro by centrosomes. Conditions for forming arrays of known polarity had been previously established (Mitchison & Kirschner, 1984a), but some modi­fication was required to make unfixed, taxol-stabilized arrays. At achievable centro- some concentrations (107 ml-1) only a tiny fraction (<0-2 %) of input tubulin can be converted into nucleated micro tubules. Thus asters must be separated from free tubulin before taxol addition to avoid production of large numbers of free micro­tubules. This was accomplished by sedimenting asters through glycerol-containing cushions, which also served to stabilize the unfixed micro tubules against depolym­erization prior to taxol addition. This technique was a modification of the one previously used in preparing asters for electron microscopy (Evans et al. 1985).

Kinesin, ATP and latex beads were added to the immobilized astral arrays, which were observed by video-enhanced differential interference microscopy. Bead move­ment was observed to be uniformly centrifugal, i.e. towards plus ends. At the same time distal microtubule segments moved centripetally, driven by kinesin on the substrate (Vale et al. 19856). Thus kinesin is a plus-end-directed translocator. This specificity reinforces the idea that a kinesin active site makes a stereospecific protein— protein interaction with repeated sites on the microtubule lattice. Since directionality comes from the kinesin-microtubule interface it is not influenced by factors such as the orientation of kinesin on the bead surface. Presumably active movement requires more than one binding site, so the molecule remains attached to the microtubule while reaching out for a new attachment. Whether a single kinesin molecule (which contains more than one copy of the predominant 120X103Mr polypeptide; Vale et al. 1985a) can move on its own is not known.

The introduction of the directionality assay led to the discovery in axoplasm of a second motile activity, acting towards minus ends (Vale et al. 19856). This differed from kinesin by both pharmacological and immunological criteria. The potent in­hibition of minus-end-directed motility by vanadate and V-ethylmaleimide suggests a relationship to axonemal dynein, which is also thought to exert force towards minus ends (Sale & Satir, 1977).

124 T. J. MitchisonI M P L I C A T I O N S FOR T H E P H Y S I O L O G I C A L ROLE OF K I N E S I N

The polarity results suggest that if kinesin is indeed a motor for fast axonal transport, then it is responsible for anterograde movement. At present the relation­ship between the bulk of kinesin, which is present in the soluble cytoplasmic pool (Vale et al. 1985c), and vesicle-bound forms is unknown, though preliminary work suggests a receptor protein on vesicles (Vale et al. 1985c). This is a very interesting direction for future research since it may hold the key to the specification of the direction of vesicle transport. At present the simplest model to account for the directionality of axonal transport invokes two different receptors, one for kinesin and one for the retrograde factor. One class of organelles would bind kinesin and move anterograde along the unipolar axonal microtubules, while another would bind the retrograde transporter and move back towards the cell body.

Kinesin was purified initially from axons, but is also present in other cell types such as sea urchin eggs (Scholey et al. 1985) and fibroblasts (R. D. Vale, personal communication). It seems likely to play a general role in organelle transport towards the cell periphery, and a balance between kinesin and minus-end-directed trans­location could be important for positioning organelles within cells. The apparent lack of specificity for transport substrates in vitro suggests that kinesin could also generate tension between microtubules and other cytoskeletal elements. The polarity results suggest, for example, a role in pulling kinetochore microtubules polewards (J. R. McIntosh, personal communication), and perhaps stretching intermediate filaments towards the periphery of the cell.

C H R O M O S O M E M O V E ME NT I N VIVOThe movement of chromosomes on the spindle can be considered a special case of

an organelle moving with the help of a microtubule, in which movement is intimately coupled to microtubule dynamics. Chromosomes are attached to the spindle by microtubules with their plus ends terminating at the kinetochore (Euteneuer & McIntosh, 19816). When chromosomes move towards or away from the pole, kinetochore microtubules must add or lose subunits, while remaining attached (for reviews, see Nicklas, 1971; Pickett-Heaps et al. 1982). We probed these dynamics by microinjecting biotin-labelled tubulin subunits into living cells at metaphase, and following their incorporation into microtubules by immunoelectron microscopy (Mitchison et al. 1986). Subunits incorporated at the kinetochore, giving rise to kinetochore microtubules with labelled segments proximal to the kinetochore, and unlabelled segments distal. The labelled segments elongated with time while spindle length remained constant, suggesting a continuous polewards flux of subunits during metaphase. This treadmilling (Margolis & Wilson, 1981) of the kinetochore micro­tubules corresponds to plus-end-directed motility of the stationary kinetochores. When cells injected in metaphase went into anaphase, labelled segments proximal to the kinetochore were lost, suggesting that kinetochore microtubules shorten by disassembly at the kinetochore. Thus anaphase movement corresponds to minus- end-directed motility of the kinetochore, and this structure is capable of bidirectional

Movement of kinesin and kinetochores 125movement with respect to the micro tubule lattice, in apparent distinction to kinesin and dynein. However, these data do not permit us to draw direct conclusions about force generation by the kinetochore, because of the complexity of the spindle, and the possibility of multiple force-generating mechanisms. Movement of the kinetochore relative to the microtubule lattice is likely to be at least in part a response to external forces acting on the microtubules, and in order to clarify the situation the kinetochore-microtubule interaction must be studied in isolation.

C H R O M O S O M E M OV EM ENT I N V IT R ODuring an analysis of the interaction of microtubules with isolated chromosomes

in vitro, we observed that proximal microtubule assembly could occur at the kinetochore in the presence of ATP (Mitchison & Kirschner, 19856). Furthermore, in the presence of ATP the kinetochore could move along the surface lattice of taxol- stabilized microtubules, apparently towards the plus end, at a maximal rate of 2-3|Ummin_1 (about 1/10 the rate of kinesin-induced bead movement in vitro). These observations were made by following the position of a biotin-labelled micro­tubule segment relative to the kinetochore by fixing, at time intervals and analysing populations. This method is less satisfactory than real-time observation for analysing movement; for example, rapid, intermittent motion could not be distinguished from slow, continuous motion. Thus there is an urgent need for reanalysis of in vitro chromosome movement by video-enhanced microscopy, pending which the con­clusions, particularly concerning the polarity of movement, must be treated with caution. Nevertheless, the results strongly suggest that the kinetochore-microtubule connection is mediated by an ATPase that binds to the surface lattice.

S T R U C T U R E OF T HE K I N E T O C H O R E - M I C R O T U B U L E I N T E R F A C EThe ability of the kinetchore to hold on to the microtubule end while subunits are

added or lost necessitates multiple contacts to the lattice. It seems unlikely that 13 contacts to the protofilaments at the end of the lattice would be sufficient, and the most plausible structure seems to be a sliding collar, making multiple contacts with the surface lattice, a structure first graphically suggested by Margolis & Wilson (1981). This structure helps explain both the in vivo and in vitro data, but the exact nature of the individual binding sites in the collar remains to be addressed. Do they mediate specific protein-protein interactions with the lattice, respecting its polarity as with kinesin, or is the interaction less specific? Is their attachment and detachment from the lattice a simple reversible binding reaction, or part of a unidirectional enzymic cycle?

The bidirectional movement of the kinetochore in vivo argues against its being an active directional translocator like kinesin or dynein, and the translocation observed in vitro was if anything in the wrong direction to explain anaphase movement. I think it is most likely that the flux at the kinetochore in vivo is imposed by external forces on the microtubule. Assembly (plus-end-directed movement) could be driven by

126 T. jf. Mitchison

polewards pulling on kinetochore microtubules (perhaps by kinesin, see above). Such forces would achieve congression if they acted along the length of micro­tubules, so that total force was dependent on kinetochore-to-pole distance. Dis­assembly (minus-end-directed movement) could perhaps be driven by an elastic element pulling on the chromosome (Pickett-Heaps et al. 1982), but I prefer the idea that the thermodynamic drive to disassembly alone is sufficient to move chromosomes polewards (Inoue & Sato, 1967). GDP-liganded microtubules have a strong thermodynamic drive towards disassembly, even while other, possibly GTP-capped microtubules, continue to assemble (Mitchison & Kirschner, 19846).

Hill (1985) has described a model kinetochore with multiple binding sites to the lattice, and a simple activation energy per site for slipping from one position on the lattice to a neighbouring one. Using a small amount of binding energy per site (1 -5 kcal m o P ') and proportionally reasonable activation energies for slipping, the model kinetochore will stay firmly attached as it follows a shrinking microtubule. In such an equilibrium binding model the binding energy per site is limited by the need for the kinetochore not to slow down the rate of microtubule disassembly too much: anaphase movement (1—3,ummin_1; Nicklas, 1971) is not very much slower than free plus-end disassembly in vitro (12/im m in-1 ; Mitchison & Kirschner, 19846). If the equilibrium model is correct, then the kinetochore-microtubule connection is likely to be mediated by multiple weak bonds that have low activation energies for sliding, such as electrostatic interactions. Conceivably, the microtubule-kinetochore

1/Fig. 1. Model for the kinetochore-m icrotubule interaction. The kinetochore acts as a sliding collar on the microtubule, with many individual sites that interact with the lattice. Individual binding sites are ATPase molecules that have a high affinity for the lattice in the ADP (D) conformation, and dissociate on ATP binding (T ). T he plus end of the microtubule is free to add and lose subunits (squares). In the absence of external force, GDP-liganded subunits will tend to dissociate, and the kinetochore will slip down the disassembling microtubule in order to maximize the num ber of sites still bound to the lattice. If an external force pulls the microtubule polewards (left in figure) while the kinetochore is fixed, subunits will add to the end, since assembly within the kinetochore collar is favoured by formation of bonds with the ATPase sites. T hus the collar acts as a reversible transducer between assembly dynamics and force, even though the ATPase cycle is unidirectional.

Movement of kinesin and kinetochores 127interaction could be that of a negatively charged rod in a positively charged hole. In this case individual binding sites would be very different from kinesin, and might not notice the polarity of the lattice. The in vitro movement data, however, suggest a different model, in which the individual sites are in fact more like kinesin, that is ATPase molecules that bind tightly to the lattice when liganded with ADP, and dissociate on ATP binding (Fig. 1). Unlike kinesin the kinetochore ATPase would not be strongly directional, that is it would tend to rebind to the lattice at the same position from which it dissociated. Thus the kinetochore on its own would not be an active translocator in the presence of ATP (though it might display a weak plus-end- directed bias as detected in vitro). In the absence of ATP, a rigor complex would be formed that caps the microtubule against disassembly (Mitchison & Kirschner, 19856). This model kinetochore should act as a transducer between assembly dynamics and force in the same way as Hill’s equilibrium model does, except that the constraints on the rate constants are removed. Binding is strong in the ADP-liganded state, mediated by specific protein-protein interactions, and weak in the ATP- liganded state, as is probably the case for kinesin. Disassembly and sliding both occur when an individual site is in the low-affinity ATP state, so their rates are independent of the binding energy in the high-affinity state. Thus the model combines the advantages of easy sliding and rapid disassembly with those of a high- affinity interaction, which respects the polarity of the lattice. Selective interaction with microtubules of the correct polarity (plus-end proximal) could be important for correct spindle morphogenesis.

I thank all the friends in the mitosis field who have contributed useful discussion, Lynn Williams for drawing Fig. 1, Pat Magrath for preparing the manuscript and the Medical Research Council for financial support.

R E F E R E N C E SA l l e n , R. D ., A l l e n , N. S. & T ravis , J. L. (1981). Video enhanced contrast differential

interference microscopy (AVEC-DIC): a new method capable of analysing microtubule related motility in the reticulopodal networks of Allogromia laticollaris. CellMotil. 1, 291-302.

A m o s , L. A ., L in c k , R. W. & K l u g , A . (1976). Molecular structure of cytoplasmic microtubules. In Cell Motility (ed. R. Goldman, T . Pollard & J. Rosenbaum), pp. 847-868. N Y : Cold Spring Harbor Laboratory Press.

Be r g e n , L. G. & Borisy , G. G. (1980). Head to tail polymerisation of microtubules in vitro.J . Cell Biol. 84, 141-150.

B e rg e n , L. G ., K u riy a m a , R. & B o risy , G. G. (1980). Polarity of microtubules nucleated by centrosomes and chromosomes of Chinese hamster ovary cells in vitro. J . Cell Biol. 84, 151-159.

CHALFIE, M. & T h o m so n , J. N. (1979). Organisation of neuronal microtubules in the nematode Caenorhabditis elegans.J. Cell Biol. 82, 278-289.

C o h e n , C . H a rriso n , S. C . & St e p h e n s , R. E. (1971). X-ray diffraction from microtubules.J . molec. Biol. 59, 375-380.

E u t e n e u e r , V. & M cI n t o sh , J. R. (1981a). Polarity of some motility related microtubules.Proc. natn. Acad. Sci. U.S.A. 78, 372-376.

E u te n e u e r , V. & M c I n to s h , J. R. (19816). Structural polarity of kinetochore microtubules in PtK j cells. J . Cell Biol. 89, 338-345.

E v a n s , L ., M itc h is o n , T . J. & K ir s c h n e r , M . W. (1985). Influence of the centrosome on the structure of nucleated microtubules. J . Cell Biol. 100, 1185-1191.

G ib b o n s , I . R . (1981). Cillia an d flagella of eukaryotes. J . Cell Biol. 91, 107-124.

128 T. J . MitchisonG r a f s t e i n , B. & F o rm a n , D. S. (1980). Intracellular transport in neurons. Physiol. Rev. 60,

1167-1282.H aim o , L. T . , T e l z e r , B. R. & R o sen b au m , J. L. (1979). Dynien binds to and cross bridges

cytoplasmic microtubules. Proc. natn. Acad. Sci. U .SA. 76, 5759-5763.H e id e m a n n , S. R., L a n d e r s , J. M. & H am b o rg , M. A. (1981). Polarity orientation of axonal

microtubules. J . Cell Biol. 91, 661-665.H e id e m a n n , S. R. & M cI n t o sh , J. R. (1979). Visualisation of the structural polarity of

microtubules. Nature, Land. 286, 517-519.H i l l , T . L. (1985). Theoretical problems related to the attachment of microtubules to

kinetochores. Proc. natn. Acad. Sci. U .SA. 82, 4404-4408.I n o u e , S. & S a to , H. (1967). Cell motility by labile association of macromolecules. The nature of

spindle fibres and their role in chromosome movement, jf. gen. Physiol. 50, 259-292.K irsch ner , M. W. (1978). Microtubule assembly and nucleation. Int. Rev. Cytol. 54, 1-70.M a n d e lk o w , E. M ., S c h u l th e is s , R ., R app, R . & M a n d e lk o w , E. (1986). On the surface lattice

of microtubules: Helix starts, protofilament number, seam and handedness. J . Cell Biol. 102, 1067-1073.

M a r g o l i s , R. L. & W ils o n , L. (1981). Microtubule treadmills - possible molecular machinery. Nature, Lond. 293, 705-711.

M c N iv e n , M. A., W a n g , M. & P o r t e r , K. R. (1984), Microtubule polarity and the direction of pigment transport reverse simultaneously in surgically severed melanophore arms. Cell 37, 753-765.

M ttc h is o n , T . J., E v a n s , L ., S c h u lz e , E . S. & K ir s c h n e r , M . W. (1986). Sites of microtubule assembly and disassembly in the mitotic spindle. Cell 45, 515-527.

M itc h is o n , T . J. & K ir s c h n e r , M. W. (1984a). Microtubule assembly nucleated by isolated centrosomes. Nature, Lond. 312, 232-236.

M itc h is o n , T . J . & K ir s c h n e r , M . W. (19846). Dynamic instability of microtubule growth. Nature, Lond. 312, 237—241.

M it c h iso n , T . J. & K ir s c h n e r , M . W. (1985a). Properties of the kinetochore in vitro.I. Microtubule nucleation and tubulin binding. J . Cell Biol. 101, 755-765.

M it c h iso n , T . J. & K ir sc h n e r , M . W. (19856). Properties of the kinetochore in vitro.II . Microtubule capture and ATP dependent translocation. J . Cell Biol. 101, 766-777.

N ic k la s , R. B. (1971). Mitosis. Adv. Cell Biol. 2, 225-297.P ickett-H e a ps , J. D ., T ip p e t , D. H . & P orter, K. R. (1982). Rethinking mitosis. Cell 29,

729-744.P orter , K. R. (1966). Cytoplasmic microtubules and their functions. In Ciba Fdn Symp.

Principles o f Biomolec. Org. London: Churchill.S a le , W. S. & S a u r , P. (1977). The direction of active sliding of microtubules in Tetrahymena

cilia. Proc. natn. Acad. Sci. U .SA. 74, 2045-2049.S c h o le y , J . M ., P o r t e r , M . E ., G r is s o n , P. M . & M c I n to s h , J . R . (1985). Iden tifica tion of

k inesin in sea u rc h in eggs, an d evidence for its localisation in th e m ito tic sp ind le . Nature, Lond. 318, 483-486.

T e l z e r , B. R ., M o se s , M . J. & R o sen b au m , J. (1975). Assembly of microtubules onto kinetochores of isolated mitotic chromosomes of HeLa cells. Proc. natn. Acad. Sci. U .SA. 72, 4023-4027.

Va l e , R . D ., R e e s e , T . S . & Sh e e t z , M. P. (1985a). Identification of a novel force generating protein, kinesin, involved in microtubule base motility. Cell 42, 39-50.

V a le , R . D ., S c h n ap p , B. J., M itc h is o n , T. J., S t e u e r , E., R e ese , T . S. & S h e e tz , M . P. (19856). Different axoplasmic proteins generate movement in opposite directions along microtubules in vitro. Cell 43, 623-632.

V a le , R . D ., S c h n ap p , B. J., R e ese , T . S. & S h e e tz , M. P. (1985c). Organelle, bead and microtubule translocations promoted by soluble factors from the squid giant axon. Cell 40, 559-569.