1: \documentclass[prl,aps,superscriptaddress]{revtex4}
2: %\documentclass[preprint,aps,superscriptaddress,showpacs]{revtex4}
3: \usepackage{graphicx}
4: \usepackage{epsfig}
5: \usepackage{amsmath}
6: \usepackage{amssymb}
7: \usepackage{tabularx}
8: \usepackage{multirow}
9: \usepackage{array}
10:
11: \begin{document}
12:
13: \title{Modelling Dual Pathways for the Metazoan Spindle Assembly
14: Checkpoint}
15:
16: \author{Richard P. Sear}
17: \affiliation{Department of Physics, University of Surrey, Guildford,
18: Surrey GU2 7XH, UK.}
19:
20: \author{Martin Howard}
21: \affiliation{Department of Mathematics, Imperial College London, South
22: Kensington Campus, London SW7 2AZ, UK.}
23:
24: \date{\today}
25: \begin{abstract}
26: Using computational modelling, we investigate mechanisms of signal
27: transduction
28: %where a small subcellular structure must communicate with
29: %distant locations across a large cell volume. As an example we focus
30: focusing on
31: the spindle assembly checkpoint where a single unattached
32: kinetochore is able to signal to prevent cell cycle progression. This
33: inhibitory signal switches off rapidly once spindle microtubules have
34: attached to all kinetochores. This requirement tightly constrains the
35: possible mechanisms. Here we investigate two possible mechanisms for
36: spindle checkpoint operation in metazoan cells, both supported by
37: recent experiments. The first involves the free diffusion and
38: sequestration of cell-cycle regulators. This mechanism is severely
39: constrained both by experimental fluorescence recovery data and also
40: by the large volumes involved in open mitosis in metazoan cells. Using
41: a simple mathematical analysis and computer simulation, we find that
42: this mechanism can generate the inhibition found in experiment but
43: likely requires
44: %%% NEW
45: a two stage signal amplification cascade.
46: %%%
47: The second mechanism involves spatial gradients of a short-lived
48: inhibitory signal that propagates first by diffusion but then
49: primarily via active transport along spindle microtubules. We propose
50: that both mechanisms may be operative in the metazoan spindle assembly
51: checkpoint, with either able to trigger anaphase onset even without
52: support from the other pathway.
53: \end{abstract}
54:
55: \maketitle
56:
57: \noindent
58: Keywords: Signal transduction, kinetochore, spindle assembly
59: checkpoint, mathematical modelling
60:
61: \vspace*{0.8cm}
62:
63: The question of how a signal emanating from a small, compact structure
64: in a cell can be amplified and propagated to an entire cell is
65: fundamental to cell biology \cite{nasmyth05}. An excellent example is
66: provided by the spindle assembly checkpoint (SAC) \cite{musacchio02},
67: which regulates cell cycle progression from metaphase to anaphase
68: during mitosis. The segregation of sister chromatids that occurs
69: during anaphase is permitted only after all the kinetochores are
70: attached via microtubules to the mitotic spindle. Even a single
71: unattached kinetochore can signal to the rest of the cell and prevent
72: cell cycle progression \cite{rieder95,rieder94}. A fundamental issue
73: is how a relatively small structure, such as a kinetochore, can
74: generate sufficient signal to robustly communicate with distant
75: subcellular locations \cite{nasmyth05}. Moreover, this signal must
76: switch off rapidly, within a period of minutes, after complete
77: kinetochore attachment to spindle microtubules \cite{rieder94}. These
78: requirements strongly constrain the possible signal transduction
79: mechanisms. In this paper, we focus
80: particularly on the SAC in cases where the nuclear envelope breaks
81: down prior to SAC activity (open mitosis), as in metazoan cells. In
82: this context, we examine two distinct models: a diffusive
83: sequestration model and a model involving active signal transport
84: along spindle microtubules. We believe that both of these pathways
85: may be in simultaneous operation in the metazoan SAC.
86: %However, we
87: %emphasise that our analysis is quite general and may apply to other
88: %signal transduction systems where a small structure must reliably
89: %communicate to distant subcellular locations.
90:
91: The metaphase/anaphase transition is triggered by an intricate
92: sequence of events centred around the proteins securin, cyclin B and
93: separase. The first step is the ubiquitination of securin and cyclin
94: B by the Anaphase Promoting Complex/Cyclosome (APC/C) \cite{castro05},
95: a process that tags securin/cyclin B for destruction via the 26S
96: proteasome. This degradation allows separase to cleave the cohesin
97: complex that tethers sister chromatids together. Once the cohesin
98: complex has been cleaved, the sister chromatids aree pulled apart to
99: opposite poles by the mitotic spindle. In order to prevent premature
100: entry into anaphase, the SAC must prevent securin/cyclin B
101: ubiquitination by the APC/C until proper attachment of {\it all}
102: chromosomes to the spindle. Evidence has accumulated for a number of
103: overlapping, and therefore possibly redundant, mechanisms for SAC
104: operation. The APC/C is known to be stimulated by Cdc20 binding;
105: hence, a plausible way to achieve APC/C inhibition is to inhibit the
106: ability of Cdc20 to bind to the APC/C. One possibility is that Cdc20
107: is held, sequestered, in an inactive form via binding to Mad2, with
108: the production of Cdc20-Mad2 being promoted by unattached kinetochores
109: \cite{howell04,deantoni051,deantoni052}. Key proteins identified at
110: unattached kinetochores include Bub1, Mad1, Mad2, BubR1 (Mad3 in
111: budding yeast), Bub3 and Cdc20. Moreover, FRAP experiments have
112: revealed that some of these proteins, including Mad2, BubR1 and
113: Cdc20, turnover rapidly at unattached kinetochores
114: \cite{howell04,shah04}. Furthermore, the available evidence suggests
115: that Mad2 exists in two forms: open (O-Mad2) and closed (C-Mad2)
116: \cite{sironi02,luo02}, with the closed form adopted when bound to
117: Cdc20. Production of C-Mad2-Cdc20 may be catalysed via the
118: kinetochore bound C-Mad2-Mad1 complex. Intriguingly,
119: Refs.~\cite{deantoni051,deantoni052} propose that C-Mad2-Cdc20 away
120: from the unattached kinetochore can convert further cytosolic O-Mad2
121: and Cdc20 into their bound C-Mad2-Cdc20 state. In this way the
122: relatively weak signal coming from an unattached kinetochore can be
123: amplified leading to comprehensive Cdc20 sequestration throughout the
124: cell. Recent experiments have further implicated a protein called
125: p31$^{\rm comet}$ in switching off this signal after complete
126: kinetochore attachment \cite{deantoni051,xia04,mapelli06}.
127:
128: However, the above ``Mad template'' model is not the only proposed
129: mechanism for APC/C repression. BubR1 and Bub3 are also known to bind
130: Cdc20 and thus repress the APC/C \cite{tang01}. Indeed, BubR1 appears to
131: be a more potent inhibitor of Cdc20 than Mad2, and both BubR1 and Mad2
132: may mutually promote each other's binding to Cdc20 \cite{fang02}.
133: Furthermore, Bub1 is believed to phosphorylate Cdc20, possibly also
134: repressing the APC/C \cite{tang04}. In addition, the overall copy
135: number of Cdc20 is down-regulated until all the kinetochores are
136: attached \cite{pan04}. Clearly, reducing the overall number of Cdc20
137: will impair the effectiveness of the APC/C prior to anaphase. Moreover,
138: not only does Cdc20 form complexes with Mad2 and BubR1, but it is also
139: believed to form a separate complex called the Mitotic Checkpoint
140: Complex (MCC), consisting of Mad2, BubR1, Bub3 and Cdc20
141: \cite{sudakin01}. This complex also appears to be a potent inhibitor
142: of the APC/C \cite{sudakin01}. In addition, the APC/C is regulated by
143: phosphorylation through the kinase Cdk1 \cite{kraft03}. Lastly, the
144: APC/C itself has intriguing localization properties, localizing, for
145: example, to unattached kinetochores \cite{acquaviva04}. This
146: positioning may also have implications for the mechanism of APC/C
147: inhibition.
148:
149: Recently, a pioneering paper by Doncic {\it et al.}~\cite{doncic05}
150: introduced a careful mathematical analysis to the question of how the
151: SAC functions. Focusing on yeast, they analysed various simplified
152: models of how an unattached kinetochore can signal to the rest of the
153: cell. Yeast undergoes closed mitosis, where even at the onset of
154: anaphase the nuclear envelope is still complete. As a consequence, a
155: diffusive signal from an unattached kinetochore is only required to
156: propagate within the nuclear volume of a few cubic microns. However,
157: %%% NEW
158: one important difference between yeast and metazoan cells is that the
159: latter
160: %%%
161: undergo open mitosis, where the nuclear envelope has
162: broken down by the time the SAC is active; hence any freely diffusing
163: signal from an unattached kinetochore must propagate throughout the
164: cytoplasm, a volume a thousand times larger than the nucleus of a
165: yeast cell.
166: %%% NEW
167: Furthermore a considerable amount of data is available concerning the
168: metazoan SAC in the form of flourescent bleaching data and estimates
169: for the copy numbers of the relevant molecules. In this paper, we
170: therefore use computational modelling to analyze the SAC in metazoan
171: cells. We develop a new model that incorporates two step signal
172: amplification, a mechanism with close similarities to previously
173: investigated multistep signal cascades
174: \cite{heinrich02,kholodenko06}. This similarity points towards a close
175: connection between the SAC and other cell signalling systems, such as
176: MAPK cascades. Importantly, we find that our model generates robust
177: inhibition in metazoan cells, unlike models without an amplification
178: step.
179: %%%
180:
181: As discussed above, the most popular models of APC/C inhibition
182: involve transport solely by diffusion. However, these models are hard
183: to reconcile with experiments of Rieder {\it et
184: al.}~\cite{rieder97}. Using cells with two independent spindles,
185: Rieder {\it et al.}~found that unattached kinetochores on one spindle
186: did not block anaphase onset in a neighbouring complete spindle. This
187: experiment appears to indicate that any ``wait anaphase'' signal would
188: have to be limited to a single spindle, and {\it not} diffuse
189: throughout the cytosol. This finding is in clear contradiction with
190: the mechanisms introduced above. For this reason we develop a second
191: model where a short-lived signal from an unattached kinetochore first
192: diffuses to a nearby microtubule, before being actively transported
193: along spindle microtubules. In this way the signal is restricted to
194: the spindle. Moreover, we propose that both the active transport and
195: diffusive mechanisms may be in simultaneous use in metazoan cells. We
196: suggest that either of the signalling mechanisms may individually be
197: able to initiate anaphase, even without support from the other
198: pathway. With this assumption, our models are then fully consistent
199: with both the experiments of Rieder {\it et al.}~\cite{rieder97} and
200: with a diffusive sequestration model \cite{deantoni051}.
201:
202: \section*{Results}
203:
204: \section*{Mechanisms with diffusive transport}
205:
206: %We consider the situation where a small structure inside a cell must
207: %signal its state to the rest of the cell.
208: %In particular,
209: We analyze
210: the case where the kinetochores control the concentration of a freely
211: diffusing species $c$. When a kinetochore is unattached and so is
212: signalling, a large majority of the species $c$ is in the $c^*$ state,
213: but when the last kinetochore itself switches off, the $c^*$ species
214: rapidly decays to the $c$ state, thus communicating the switch-off
215: (attachment) to the rest of the cell.
216: %We refer to the signalling
217: %structure as a kinetochore, however, our model and conclusions apply
218: %to any situation where a relatively small structure is required to
219: %control the state of a freely diffusing species throughout a
220: %compartment or cell.
221: We will sometimes refer to the $c^*$ species as being in the
222: inhibiting state as it is this state that prevents securin/cyclin B
223: ubiquitination by the APC/C.
224: %However, in other contexts the $c^*$ may
225: %be activating a process, as in the response to DNA damage.
226: %At present, insufficient data exists to model the metazoan SAC system
227: %in full detail; instead we investigate outline models in order to
228: %understand which mechanisms are {\it in principle} capable of
229: %producing a functional SAC \cite{doncic05}.
230:
231: \
232:
233: \noindent
234: {\bf The concentrations of the freely diffusing species are almost
235: uniform} The timescale for diffusion across the cell is
236: $\tau_D=r_c^2/D$, whereas the mean time for a molecule to collide with
237: the kinetochore is $\tau_C=r_c^3/(Dr_k)$. Here $D$ is the diffusion
238: constant for the protein; $r_c$ is the distance across the cell
239: (typically a few tens of microns); and $r_k\simeq 0.2\mu$m
240: \cite{mcewen98} is the radius of the kinetochore.
241: %%% NEW2
242: If we assume that the lifetime for the inhibiting $c^*$ species is
243: much longer than $\tau_D$ then any gradient in its concentration will
244: clearly be small. Furthermore, the timescale for diffusion across the
245: cell is much smaller than that for collisions with the kinetochore,
246: $\tau_D/\tau_C=r_k/r_c \ll 1$. Hence, each molecule of the $c$
247: species criss-crosses the cell many times between kinetochore
248: reactions and so gradients in its concentration are also small in most
249: of the cell. If, on the other hand, the lifetime of the inhibiting
250: species is short with respect to $\tau_D$, then its concentration will
251: no longer be uniform and instead density gradients will form. This
252: scenario will arise later on when we consider models with active
253: transport.
254: %However, for longer lifetimes, the concentration of $c^*$, and hence
255: %the inhibition, will be almost uniform throughout the cell.
256: However, for longer lifetimes, we can model the inhibition using
257: simple ordinary differential equations.
258: %These are a function of time
259: %only; there is no explicit space dependence. We have also performed
260: %explicit calculations using partial differential equations (see also
261: %Ref.~\cite{doncic05}) which do explicitly include space. However, as
262: %expected, we find only small gradients that do not effect our
263: %conclusions.
264: %Note that the ratio of the two timescales
265: %$\tau_D/\tau_C=r_k/r_c$ is independent of the diffusion constant $D$
266: %and so gradients will be small whatever the value of $D$.
267: %%%
268:
269: \
270:
271: \noindent
272: {\bf Inhibitor production only at a kinetochore} In the simplest
273: possible model, the inhibiting $c^*$ species is produced only at
274: unattached kinetochores. We denote the steady state rate of
275: production of $c^*$ at the final unattached kinetochore by $J_{\rm
276: off}$. In order to allow the inhibition to be switched off at the
277: beginning of anaphase, the inhibiting $c^*$ species must be
278: unstable. We model this instability by a first-order decay $c^*\to c$,
279: with rate constant $\alpha$. Cdc20 is known to activate the APC/C and
280: so trigger anaphase. Since both Mad2 and BubR1 are known to bind to
281: Cdc20, we can tentatively identify $c$ as Cdc20 while $c^*$ is a
282: Mad2-Cdc20 or BubR1-Cdc20 complex. As the kinetochore is such a small
283: structure, we now address the question of whether the flux of $c^*$
284: produced {\it only} at a single unattached kinetochore is sufficient
285: to maintain inhibition.
286: %%% NEW
287: At steady state the fluxes on and off a kinetochore must be the same,
288: i.e. $J_{\rm on}=J_{\rm off}$. Furthermore, at steady state, the rate
289: of production of $c^*$ at the last unattached kinetochore must equal
290: its rate of decay, i.e. $J_{\rm off}=\alpha N_c^*$, where $N_c^*$ is
291: the number of inhibiting $c^*$ molecules. Experimental data, largely
292: on marsupial PtK$_1$ and PtK$_2$ cells, gives estimates for many of
293: the model parameters \cite{howell00,howell04}. From this data we will
294: be able to estimate the value of $N_c^*$ and see whether good
295: inhibition can be obtained.
296: %%%
297:
298: Experimental evidence for the flux $J_{\rm off}$ is available from
299: FRAP (Fluorescence Recovery After Photobleaching) data for the
300: recovery of fluorescence of checkpoint proteins at unattached
301: kinetochores. This data provides a direct measure of dissociation
302: rates. Together with an estimate of the copy number of each
303: kinetochore bound protein, we can then derive the off flux $J_{\rm
304: off}$. For BubR1 the (fast phase, see \cite{howell04}) FRAP half-life
305: ($t_{1/2}$) is 3s, whereas for Mad2 the half-life is $10-20$s
306: \cite{howell04,shah04}, with a kinetochore occupancy $N_k$ of about
307: 1000 molecules for both \cite{howell04}. This gives an off flux of
308: about $J_{\rm off}=N_k \ln 2/t_{1/2} \sim 1000\times\ln 2/3\sim
309: 200~s^{-1}$ for BubR1 and $J_{\rm off}\sim 50~s^{-1}$ for Mad2.
310: %%% NEW
311: Note that these fluxes are actually the maximum possible, as not all
312: of the released molecules will be in the inhibiting form.
313: %%%
314: The lifetime of the inhibiting species $\alpha^{-1}$ has not been
315: directly measured. However, anaphase begins about 20 min after the
316: last kinetochore becomes attached \cite{rieder94}. Clearly, the number
317: of inhibiting $c^*$ molecules must decrease dramatically over a period
318: of no more than about 10 min, giving an upper bound of
319: $\alpha^{-1}\sim 10$min, a value which is also consistent with the
320: data of Ref.~\cite{clute99}. If we sum the fluxes of Mad2 and BubR1
321: and set $\alpha^{-1}=10$ min, the number of molecules $N_{c*}=J_{\rm
322: off}\alpha^{-1}\simeq 150,000$. The total number of Cdc20 molecules
323: is approximately $800,000$ \cite{howell04}. Thus with the experimental
324: fluxes and even assuming a decay rate as slow as is reasonable we find
325: that at steady state a single kinetochore can only sequester
326: approximately 20\% of the Cdc20.
327: %%% NEW
328: Furthermore, if the lifetime of the inhibiting $c^*$ molecules is
329: shorter than supposed by our upper bound $\alpha^{-1}\sim 10$min, or
330: if the flux $J_{\rm off}$ is lower, then an even smaller fraction will
331: be in the inhibiting state.
332: %%%
333: Hence, the formation of Mad2/Cdc20 and BubR1/Cdc20 complexes solely at
334: an unattached kinetochore cannot maintain good inhibition.
335:
336: The experimental FRAP data used in the above analysis was not
337: specifically derived from the {\it final} unattached
338: kinetochore. Potentially the final unattached kinetochore might have a
339: higher turnover rate, thereby generating greater inhibition. We
340: therefore develop a complementary approach to determine whether a
341: reaction only at a single unattached kinetochore is {\it in principle}
342: sufficient to sequester most of the Cdc20. We estimate the inhibition
343: that would be generated if the on flux $J_{\rm on}$ at the kinetochore
344: were maximised, i.e., diffusion limited with a diffusion constant $D$
345: at the top of the range of the values measured {\it in vivo}.
346: Diffusion constants in the cytoplasm of metazoan cells have typically
347: yielded values in the range $1$ to $30\mu$m$^2$s$^{-1}$
348: \cite{wojcieszyn81,lang86,seksek97,schmiedeberg04}. Also, Howell {\it
349: et al.} \cite{howell04} found that a bleached spot of cytoplasmic
350: GFP-Cdc20 $0.8\mu$m across recovered its fluorescence in less than
351: $0.2$s, implying a diffusion constant $\gtrsim 5\mu$m$^2$s$^{-1}$. If
352: the on flux to the unattached kinetochore is diffusion limited, then
353: using a plate geometry appropriate for a kinetochore, the on flux at
354: steady state is given by $J_{\rm on} \sim 4 D r_k N_c/v_c$
355: \cite{berg93}. Here, $N_c$ is the number of $c$ molecules,
356: $v_c=6000\mu$m$^3$ is the cytoplasmic volume \cite{howell04}, and we
357: use a fast diffusion constant $D=30\mu$m$^2$s$^{-1}$. At steady state
358: the flux $J_{\rm on}$ must be balanced by the $c^*\to c$ flux,
359: $J_{c*\to c}=\alpha N_{c*}$. Equating these two fluxes and defining
360: the constant total number of molecules as $N=N_c+N_{c*}$ (neglecting a
361: negligible number at the kinetochore), we have $N_{c*}/N=(1+\alpha
362: v_c/4Dr_k)^{-1}\approx 3/4$. Hence, even in an optimistic scenario,
363: only about three-quarters of the Cdc20 can be
364: sequestered. Furthermore, to reach this limit, high flux rates are
365: required, much higher than found experimentally for BubR1/Mad2 (see
366: above). For these reasons, it is unlikely (though still theoretically
367: possible) that the metazoan SAC could function through this method.
368:
369: The fundamental difficulty with this mechanism can be presented in an
370: alternative way that brings out the essential role played by the cell
371: volume. The time each molecule spends in the $c^*$ form is on the
372: order of $\alpha^{-1}$. The time taken for each molecule in the
373: $c$ state to find the unattached kinetochore is about
374: $v_c/k_k$, where $k_k$ is the unattached kinetochore binding
375: rate. This time scale is proportional to the cell volume and hence
376: becomes longer as the cell size increases. As the volume increases,
377: for fixed lifetime $\alpha^{-1}$, each molecule spends less and less
378: time in the inhibiting state and eventually the mechanism
379: fails.
380:
381: \
382:
383: \noindent
384: {\bf Autocatalysis} One way to try to increase the number of inhibiting
385: molecules would be for the reaction that sequesters Cdc20 to occur not
386: just on the kinetochore but also off the kinetochore. If $c^*$ itself
387: catalyzes an off kinetochore $c\to c^*$ reaction then the reaction is
388: autocatalytic. Autocatalysis was considered by Doncic {\it et al.}
389: \cite{doncic05}, where it was found to be unsatisfactory for the case
390: of closed mitosis in yeast. The reason is that, for good inhibition,
391: the off-kinetochore autocatalysis will likely have a large reaction
392: rate, as the on-kinetochore reaction is on its own insufficient to
393: maintain good inhibition (see above). Consequently, when the
394: kinetochore reaction is switched off, the result is only a weak
395: perturbation of an autocatalytic reaction. In other words
396: %%% NEW2
397: inhibition of a ``go
398: anaphase'' signal either cannot be switched off, or switches off only
399: very slowly (see Fig.~1).
400: %%%
401: Interestingly, the Mad2 template model
402: proposed by Ref.~\cite{deantoni051} is essentially identical to this
403: mechanism, where the closed form of Mad2 catalyzes the
404: conversion of open to closed Mad2 both on and off the
405: kinetochore. Thus an unattached kinetochore inhibits anaphase by
406: generating C-Mad2-Cdc20 which in turn generates more C-Mad2-Cdc20 off
407: the kinetochore via an autocatalytic reaction. However, by itself this
408: autocatalytic mechanism does not allow switch off of the metazoan SAC
409: for the same reason it does not for the SAC in yeast.
410:
411: Autocatalysis could play a role in the inhibition, provided an
412: additional process switches off the autocatalytic reaction once the
413: last kinetochore attaches to a microtubule. It has been suggested
414: that the protein p31$^{\rm comet}$ could play a role in this context
415: \cite{deantoni051,xia04,mapelli06}. Potentially p31$^{\rm comet}$
416: could be upregulated to abruptly switch off the SAC. However, the
417: question is then how this signal could be turned on so rapidly after
418: microtubule attachment to the final kinetochore.
419: %%% NEW
420: If p31$^{\rm comet}$ competes with O-Mad2 for C-Mad2 binding
421: \cite{mapelli06}, then this is no easier than regulating the
422: concentration of free Cdc20, our original problem. Potentially,
423: p31$^{\rm comet}$ could be upregulated using a two step reaction
424: process; however such a scheme would have to be conceptually similar
425: to that discussed in the next section. We emphasize that p31$^{\rm
426: comet}$ could also play other important roles, for example in
427: switching off kinetochore signalling following microtubule attachment
428: \cite{vink06}.
429:
430: \
431:
432: \noindent
433: {\bf Model with an amplification step}
434: %Models with only on-kinetochore
435: %inhibition and models incorporating off-kinetochore autocatalysis both
436: %have difficulty agreeing with experiments.
437: We now turn to a
438: model with an off-kinetochore, but non-autocatalytic, reaction. This
439: involves the species: $e$, $e^*$ and $c^*$. We first assume that an
440: $e$ species cycles through the unattached kinetochore, where it is
441: converted to the inhibiting $e^*$ form. We further assume that the
442: $e^*$ is able to convert the $e$ form to a second
443: inhibiting $c^*$ species,
444: %%% NEW2
445: giving a second step to the signal amplification.
446: %%%
447: Importantly the $c^*$ form cannot convert further $e$
448: into the $c^*$ or $e^*$ forms. This form of amplification ensures that
449: only molecules that have passed through the
450: unattached kinetochore participate in amplifying the inhibitory
451: signal. As the inhibitory signal is not produced autocatalytically, it
452: can switch off rapidly after the final kinetochore-microtubule
453: attachment. The reaction processes are:
454: \begin{eqnarray}
455: e+K&\stackrel{k_k}{\rightarrow}&e^*+K \qquad
456: e^*~\stackrel{\alpha_e}{\rightarrow}~e \nonumber \\
457: e+e^*&\stackrel{k}{\rightarrow}&c^*+e^* \qquad
458: c^*~\stackrel{\alpha}{\rightarrow}~e, \label{twospec}
459: \end{eqnarray}
460: where $K$ is an unattached kinetochore. It is natural to associate
461: the $e$ species with free Cdc20, and the $e^*$ and $c^*$ species with
462: complexes of Cdc20 such as C-Mad2-Cdc20 and BubR1-Cdc20. We are
463: assuming that one form of the complex ($e^*$) can catalyze the
464: production of further complexes ($c^*$) which differ in that the $c^*$
465: cannot participate in manufacturing further complexes. This
466: distinction between the two forms is a clear prediction of our
467: modelling.
468: %%% NEW
469: However, we cannot definitively identify the difference between
470: the $c^*$ and $e^*$ species, which could, for example, involve
471: phosphorylation or a conformational change. Clearly, our reaction
472: scheme is schematic; more complex schemes based on the same principles
473: are certainly feasible.
474: %%% NEW2
475: One possibility is that there could be more than two steps to the
476: amplification process. Alternatively,
477: %%%
478: p31$^{\rm comet}$ could be
479: rapidly upregulated using a two step reaction, as discussed
480: previously. Another possibility is using a two step reaction to
481: modulate the decay rate $\alpha$. Both possibilities could in
482: principle lead to both good inhibition and rapid checkpoint switch
483: off, but are fundamentally similar to the scheme of
484: Eq.~{\bf\ref{twospec}}. The key prediction of the model
485: of Eq.~{\bf\ref{twospec}} is
486: for at least two species allowing non-autocatalytic amplification off
487: the kinetochore and hence strong sequestration.
488:
489: The above two-step process has close similarities to other multi-step
490: signalling cascades such as for MAPK
491: \cite{heinrich02,kholodenko06}. However, there are differences, for
492: example, the $e$ species participates in both steps of the above
493: amplification process. In MAPK cascades, on the other hand, a
494: separate $c$ species is converted to the $c^*$ form in the second
495: amplification reaction. Nevertheless, the principle of using more
496: than one step to provide robust amplification but with rapid response
497: times is similar and is likely conserved across many different
498: signalling systems. However, the difficulty of robust signalling in
499: the SAC is particularly acute, since the initial signal emerges from
500: such a small region (a single unattached kinetochore).
501: %%%
502:
503: For the parameter values of the reactions in Eq.~{\bf\ref{twospec}},
504: we use $\alpha^{-1}=\alpha_e^{-1}=5 {\rm min}$. These lifetimes are
505: shorter than used previously: the two step nature of the reaction
506: mechanism now dictates that shorter lifetimes are needed for switch
507: off within the appropriate timeframe of less than $20$ min. Despite
508: these short lifetimes, a two step reaction cascade ensures that robust
509: signal amplification is still achieved. In dilute solution {\it in
510: vitro}, rate constants for diffusion-limited protein-protein
511: association are around $10^{-3}$ to $10^{-2}\mu$m$^3$s$^{-1}$
512: \cite{northrup92}, with the exception of some large rate constants
513: where there is significant electrostatic attraction between the
514: proteins. We take an {\it in vivo} rate constant at the top end of
515: these values, with $k=10^{-2}\mu m^3 s^{-1}$. For the rate constant
516: $k_k$, we use the diffusion limited value $k_k\sim 4 D r$
517: \cite{berg93}. Assuming $D=20 \mu$m$^2 s^{-1}$, i.e. fairly fast
518: cytoplasmic diffusion, we find $k_k\sim 20 \mu $m$^3 s^{-1}$.
519:
520: Since the gradients in the concentrations are small (see above), we
521: can determine the time dependence and steady-state values of the three
522: species from ordinary differential equations:
523: \begin{eqnarray}
524: \frac{{\rm d}N_{e*}}{{\rm d}t}&=&\frac{k_k}{v_c}N_e-\alpha_e N_{e*},
525: \qquad \frac{{\rm d}N_{c*}}{{\rm d}t}=\frac{k}{v_c}N_{e*}N_e
526: -\alpha N_{c*},
527: \label{modeld2}
528: \\
529: \frac{{\rm d}N_{e}}{{\rm d}t}&=&-\frac{k_k}{v_c}N_e-\frac{k}{v_c}
530: N_{e*}N_e + \alpha_e N_{e*}+\alpha N_{c*},
531: \label{modeld3}
532: \end{eqnarray}
533: where $N_x$ is the number of molecules of species $x=e,e^*,c^*$. At
534: steady state (ss), we have:
535: \begin{equation}
536: N_{e*}^{ss}=\frac{k_k}{v_c\alpha_e}N_e^{ss}, \qquad
537: N_{c*}^{ss}=\frac{k k_k}{v_c^2 \alpha \alpha_e}\left(N_e^{ss}\right)^2.
538: \end{equation}
539: Defining the (constant) total number of molecules as
540: $N=N_e+N_{e*}+N_{c*}$, we find:
541: \begin{equation}
542: N=\left(1+\frac{k_k}{v_c \alpha_e}\right)N_e^{ss}+
543: \frac{k k_k}{v_c^2\alpha \alpha_e}\left(N_e^{ss}\right)^2.
544: \end{equation}
545: For $N=800,000$, and using the above parameters, we find
546: $N_e^{ss}\approx 40,000$, $N_{e*}^{ss}\approx 40,000$ and
547: $N_{c*}^{ss}\approx 720,000$. Hence around $95$\% of the molecules are
548: in the inhibiting state. For the single unattached kinetochore, we
549: find that the diffusion limited on rate onto the kinetochore is about
550: $J_{\rm on}\sim 4 D r N_e^{ss}/v_c\sim 100s^{-1}$. Assuming a
551: kinetochore population of around $1000$, as found experimentally, this
552: gives a half-life for the kinetochore bound population of
553: $5-10$s. This is roughly consistent with the observed Mad2, BubR1
554: kinetochore half-lives \cite{howell04,shah04}. As shown in
555: Fig.~\ref{dplot}, we also find that the signal switches off quite
556: quickly after microtubule attachment to the final kinetochore. After
557: 10 min the fraction of the $e$ molecules has increased from 5\% to
558: 24\% and after 20 min 60\% is in the $e$ form. Moreover the
559: switch-{\it on} of the checkpoint is even quicker, with good
560: inhibition being reestablished within 1 min of even a single
561: kinetochore detachment, see Fig.~\ref{dplot} inset. This is in good
562: agreement with cyclin B1 data from Ref.~\cite{clute99}, which suggests
563: that switch-on is essentially an order of magnitude faster than SAC
564: switch-off.
565:
566: \begin{figure}[t!]
567: \begin{center}
568: \epsfig{file=fig1.eps,width=7.0cm}
569: \end{center}
570: \caption{
571: \lineskip 2pt \lineskiplimit 2pt Plot, for the model of
572: Eq.~{\bf\ref{twospec}}, showing the fraction of molecules $N_e/N$
573: (solid line), $N_{c*}/N$ (dashed line), $N_{e*}/N$ (dotted line) as a
574: function of time, with the final kinetochore attaching at time $t=0$,
575: i.e.~initial steady-state concentrations, but $k_k$ then set to zero
576: at $t=0$. Inset shows the same fractions, but starting with all $N$
577: molecules in the $e$ form and with $k_k$ set to $20\mu$m$^3$s$^{-1}$
578: at $t=0$. For comparison, in the main plot, the fractions $N_c/N$ for
579: the production only at kinetochores and autocatalytic models are
580: plotted as grey curves. They should be compared with the solid black
581: curve.
582: %%% NEW2
583: The grey solid curve is with a reaction only at a kinetochore,
584: and the grey dot-dashed curve is with autocatalysis.
585: %%%
586: Note that with a
587: reaction only at a kinetochore the initial inhibition is weak whereas
588: with autocatalysis inhibition is near total even when $k_k$ is set to
589: zero.
590: %%% NEW2
591: The values of $k$ (for the autocatalytic rate), $\alpha$ and
592: $k_k$ are the same in the single species models
593: as for the model of Eq.~{\bf\ref{twospec}}.
594: %%%
595: \label{dplot}
596: }
597: \end{figure}
598:
599: We therefore conclude that the above model is compatible with
600: experiments. However, we did use a diffusion-limited value for $k_k$
601: and a value of $k$ near the top end of the range of reaction-rate
602: constants for typical proteins in a dilute solution
603: \cite{northrup92}. Reducing $k$ by an order of magnitude weakens the
604: level of inhibition, although over $85\%$ of the molecules are still
605: in the inhibiting state. However, the model is not consistent with
606: experiments if the reaction rates are further reduced or if diffusion
607: is substantially slowed. Thus we predict that if the mechanism of the
608: SAC is diffusive amplified sequestration then measurements of the
609: reaction rates will reveal rather fast reactions,
610: %%% NEW2
611: or possibly more than two steps to the amplification process.
612: %%%
613:
614: For completeness we also analyze an alternative two species model
615: previously proposed by Doncic et al \cite{doncic05}, for the smaller
616: volumes involved in the yeast SAC. The reaction scheme
617: for their model is:
618: \begin{eqnarray}
619: e+K&\stackrel{k_k}{\rightarrow}&e^*+K \qquad
620: e^*~\stackrel{\alpha_e}{\rightarrow}~e \nonumber \\
621: c+e^*&\stackrel{k}{\rightarrow}&
622: c^* \hspace{1.475cm} c^*~\stackrel{\alpha}{\rightarrow}~c+e.
623: \label{twospecdoncic}
624: \end{eqnarray}
625: Note that this scheme does not catalytically amplify the inhibitory
626: signal. Here, one $e^*$ molecule can only interact with one $c$ molecule
627: while in our previous model a single $e^*$ molecule can convert many
628: molecules into the inhibiting form, thereby producing
629: amplification. For good inhibition we require
630: $N_{e^*}^{ss}+N_{c*}^{ss}\approx 800,000$. Using
631: $\alpha^{-1}=\alpha_e^{-1}=5{\rm min}$, we find that the flux off the
632: unattached kinetochore must be well over $1000s^{-1}$, faster than the
633: diffusion limited maximum, even for high levels ($10^5$ copies) of the
634: $e$ species. This finding is, of course, not unexpected: the
635: lack of amplification means that the flux of inhibitory molecules off
636: the unattached kinetochore must be higher than in the catalytic model
637: proposed above. We therefore believe that this model is probably not
638: able to account for the SAC in metazoan cells.
639:
640: \section*{Mechanisms involving active transport}
641:
642: As shown above, our model with an amplification step is able to
643: explain many features of the metazoan SAC. However, it is not
644: consistent with the experiments of Rieder {\it et al.}
645: \cite{rieder97}. They observed that an incomplete spindle did {\em
646: not} inhibit another complete spindle 20$\mu$m away.
647: %%% NEW
648: Furthermore an unattached kinetochore was found to inhibit anaphase
649: onset everywhere within its local spindle even those parts more than
650: 20$\mu$m away. These findings are clearly incompatible with models
651: where inhibition propagates purely diffusively away from incomplete
652: kinetochores into the cytoplasm.
653: %As discussed previously, variations in concentration
654: %produced by this mechanism are small, meaning that the inhibition
655: %signal will be approximately equally strong throughout the
656: %cytoplasm. Hence, it is not possible for a diffusive signal to inhibit
657: %anaphase progression on one spindle, while allowing a second spindle
658: %to progress.
659: %%%
660:
661: These observations motivate us to consider mechanisms in which the
662: spindle itself plays an active role. If the signal is propagated
663: within the spindle itself, and not spread throughout the cytoplasm,
664: then the obvious transport mechanism is via molecular
665: motors. Propagation via active transport along microtubules is fast;
666: motors can move at speeds of microns per second \cite{howard01}. As
667: the typical spindle length scale is approximately 10$\mu$m, then
668: transport across the spindle takes only seconds at that speed,
669: consistent with rapid checkpoint switch-on/off. Of course, by
670: definition, an unattached kinetochore is not connected via
671: microtubules to a spindle pole. Hence that kinetochore must first
672: produce a freely diffusing species to carry the signal as before.
673: This inhibitory species, which we denote by $g^*$, initially diffuses
674: through the cytoplasm, but only until it either encounters a minus-end
675: directed microtubule-bound molecular motor, or decays to an inactive
676: form, $g$. If the molecule encounters a microtubule-bound motor, this
677: binding then stabilises the active $g^*$ form and transports the
678: inhibitory signal to a spindle pole.
679:
680: However, before we can conclude that this model with active transport
681: is consistent with the experimental data of Rieder {\it et
682: al.}~\cite{rieder97}, we need to demonstrate that it is possible to
683: find a lifetime for the $g^*$ that is long enough to allow it to
684: encounter a motor but short enough to prevent more than a small
685: fraction diffusing 20$\mu$m away. If $g^*$ is manufactured at a rate
686: $J_{g*}$ at unattached kinetochores and decays at a rate
687: $\alpha_{g*}$, then the concentration $c_{g*}({\bf r},t)$ satisfies
688: the partial differential equation
689: \begin{equation}
690: \frac{\partial c_{g*}}{\partial t}=D\nabla^2 c_{g*}-
691: \alpha_{g*} c_{g*}+J_{g*}\delta^3(r).
692: \label{rd}
693: \end{equation}
694: We solve Eq.~{\bf\ref{rd}} at steady state, after assuming spherical
695: symmetry around the source (the kinetochore) and negligible
696: concentration of $g^*$ at large distances from the source. The
697: solution is
698: \begin{equation}
699: c_{g*}(r)=(J_{g*}/4\pi) \left(\lambda/r\right)\exp(-r/\lambda),
700: \end{equation}
701: where $\lambda=\sqrt{D/\alpha_{g*}}$ and $r$ is the distance from the
702: kinetochore. Of course, the assumption of spherical symmetry is a
703: gross approximation, especially as the kinetochore is a plate shaped
704: structure. However, we are only interested in qualitative results for
705: which this approximation will be reasonable at large distances. If we
706: set the cytoplasmic $g^*$ lifetime to be $\alpha_{g*}^{-1}=0.5$s, then
707: even with a large diffusion constant of $D=20\mu$m$^2$s$^{-1}$,
708: $\lambda\sim 3\mu$m. Hence the signal $20\mu$m away will be greatly
709: attenuated.
710: %%% NEW
711: Due to the short lifetime of the $g^*$ form, the inhibitor forms a
712: steep gradient inside the cell. Hence we predict that subcellular
713: concentration gradients, already believed to be important for
714: microtubule growth and kinetochore capture, also play an important
715: role in checkpoint function \cite{wollman05}.
716: %%%
717:
718: When we considered diffusive sequestration in the previous section we
719: found kinetochore fluxes on the order of 100 molecules/s. If we
720: assume a similar flux $J_{g*}=100/$s then the concentration of $g^*$
721: $1\mu$m away from the source is approximately $20\mu$m$^{-3}$. At this
722: concentration the reaction rate per motor is $20k$, where $k$ is the
723: reaction rate between a pair of proteins. If, as in the previous
724: section, we take $k=10^{-2}\mu$m$^3$s$^{-1}$ \cite{northrup92}, then
725: we have a rate per motor complex of $0.2$s$^{-1}$. Thus a motor will
726: pick up a $g^*$ molecule within a few seconds if it is close to an
727: unattached kinetochore. If we assume a motor density of
728: $10\mu$m$^{-1}$, moving at $1\mu$m s$^{-1}$, then we expect a rate on
729: to a spindle pole of perhaps $5$s$^{-1}$ per microtubule. Metazoan
730: cells will have large numbers of spindle microtubules in the vicinity
731: of an unattached kinetochore, increasing the on-rate still
732: further. Even if some of the signal is lost in transit to the spindle
733: pole, the flux is adequate to communicate the state of the
734: kinetochore.
735: %%% NEW
736: Note that the localization of the inhibitory signal means that less
737: amplification is needed: the flux off the kinetochore together with
738: directed transport are by themselves sufficient to produce good
739: inhibition.
740: %%%
741: The next question is how the pole processes this information.
742: %Clearly
743: %the $g^*$ molecules cannot continuously accumulate at the spindle
744: %pole, as this would lead to the pole increasing in size which is not
745: %observed.
746: One attractive possibility is that the $g^*$ inhibit the spindle pole
747: and are subsequently released back into the cytoplasm in the inactive
748: $g$ form. When the last kinetochore attaches, the active transport
749: ``wait anaphase'' signal is switched off, releasing the inhibition and
750: allowing the spindle pole to broadcast a ``go anaphase'' signal. This
751: signal could be actively transported by
752: plus end directed motors to communicate with connected kinetochores on
753: the same spindle. However, in the experiments of Ref.~\cite{rieder97},
754: once one spindle had entered anaphase, the other spindle also
755: progressed to anaphase regardless of whether it contained unattached
756: kinetochores. This finding suggests that a final ``go anaphase''
757: signal is transmitted via diffusion.
758:
759: For active transport models there is little relevant experimental
760: data. As a result, our modelling has inevitably been more speculative
761: and less detailed than for diffusive sequestration models. In
762: particular it is not clear what the signalling molecule $g^*$ might
763: be. Presumably it cannot involve BubR1-Cdc20 or Mad2-Cdc20, as we
764: require a short lifetime in the cytoplasm. Furthermore these molecules
765: are not known to bind to minus-end directed motors. The active
766: transport model nevertheless predicts that the inhibitory signal is
767: propagated away to the spindle pole by a minus-end directed motor. It
768: is tempting to associate the motor protein dynein with this role;
769: however, this assignment is problematic. If dynein were performing
770: this function then inhibition of dynein would effectively switch off
771: the inhibitory active transport ``wait anaphase'' signal, resulting in
772: anaphase progression. However, experiments have revealed precisely
773: the opposite effect: inhibition of dynein leads to inhibition of cell
774: cycle progression \cite{howell01}. Furthermore this block was not due
775: to a more general effect of dynein inhibition, as injecting Mad2
776: antibodies in dynein inhibited cells still led to rapid anaphase entry
777: \cite{howell01}. Hence, either dynein is not sufficiently inhibited
778: in this experiment, implying that some inhibitory signal can still
779: leak through, or else other motors are involved. Dynein is already
780: known to transport kinetochore components
781: \cite{howell01,howell00}. However, this transport is associated with
782: the removal of Mad2 binding sites at kinetochores once a microtubule
783: has attached \cite{howell01}.
784: %%% NEW
785: Without these binding sites, Mad2 cannot cycle through a
786: kinetochore. When this removal occurs at the final attached
787: kinetochore, Cdc20 can no longer be sequestered by Mad2, and instead
788: will be free to trigger anaphase progression.
789: % In other words, after
790: % microtubule attachment, active transport is involved in releasing the
791: % inhibition of a diffusive ``go anaphase'' signal.
792: % In our modelling, this is the process by which the flux onto the
793: % kinetochore $J_{\rm on}$ is reduced to zero once a kinetochore is
794: % attached to a microtubule.
795:
796: In summary, our active-transport model is consistent with
797: the experimental data of Rieder {\it et al.}. We can easily find a
798: lifetime for the inhibitory species $g^*$ that is long enough for the
799: $g^*$ to reach an adjacent microtubule, thereby communicating the
800: state of the kinetochore to the spindle pole, but short enough for
801: strong attenuation at another spindle $20\mu$m away.
802:
803: \section*{Discussion}
804:
805: In this paper, we have shown that two models with quite different
806: mechanisms: the diffusive reaction cascade model, and a
807: model with active transport, are both possible signalling mechanisms
808: for the SAC. These two models are schematically illustrated in
809: Fig.~\ref{schem2}.
810:
811: \begin{figure}[t!]
812: \epsfig{file=fig2.eps,height=6.0cm}
813: \caption{
814: \lineskip 2pt \lineskiplimit 2pt
815: A schematic of (A) the diffusive two step reaction model and (B)
816: the active-transport model. Kinetochores are shown in green, red
817: dashed lines denote diffusion, and solid red arrows denote motion via
818: active transport. Sister chromatids are shown in blue, while the
819: black lines are microtubules.
820: \label{schem2}
821: }
822:
823: \end{figure}
824:
825: \
826:
827: \noindent
828: {\bf Dual Pathways for the Metazoan SAC?} One attractive possibility
829: to reconcile the above models and the experimental data is that, in
830: metazoan cells, both mechanisms are used. Interestingly, cells with
831: unattached kinetochores microinjected with Mad2 antibodies prematurely
832: entered anaphase \cite{gorbsky98}.
833: %%% NEW
834: This procedure will flood the cell with unsequestered Cdc20, while
835: unattached kinetochores will still be signalling a ``wait anaphase''
836: signal via the active transport mechanism. The fact that the cell
837: still enters anaphase
838: %%%
839: indicates that an active transport mechanism is probably not essential
840: for checkpoint function: the purely diffusive pathway suffices. On the
841: other hand, the experiments of Rieder {\it et al.} \cite{rieder97}
842: show that an inhibitory diffusive pathway can be
843: overruled by a second pathway, which, as we have seen, is likely based
844: on active transport. We therefore propose that {\it either} mechanism
845: can trigger anaphase onset, even without support from the other
846: pathway.
847: %%% NEW
848: The switch off of an active transport based ``wait anaphase'' signal,
849: or the release of sufficient freely diffusible Cdc20, by switching off
850: an efficient sequestration apparatus, are each separately capable of
851: generating cell cycle progression.
852: %%%
853: With this assumption, our modelling is then entirely
854: consistent with experiment.
855:
856: \
857:
858: \noindent
859: {\bf Future work} Although some of the key principles used by the
860: metazoan SAC are starting to become clear, there is still much that
861: remains to be understood. Even for models of signalling via
862: diffusion, which have been more actively pursued, further quantitative
863: measurements would be very useful. For example, if the {\it in vivo}
864: diffusion constants were found to be small then this would eliminate
865: some possible models. Also, we predict that amplification without
866: autocatalysis is likely to be required, which implies the existence of
867: at least two inhibiting species ($e^*,c^*$). Of course, the {\it in
868: vivo} checkpoint dynamics will likely be much more complicated than
869: our simple outline model, but the requirement for amplification will
870: likely remain. Better characterization of the BubR1/Mad1/Mad2/Cdc20
871: protein dynamics should therefore allow these predictions to be
872: tested.
873: %%% NEW2
874: We would also like to emphasize the close connection between our
875: diffusive two step reaction and other signal cascades, such as for
876: MAPK.
877: %%%
878:
879: For the alternative active transport pathway, a first goal would be to
880: directly observe and image its components. For example, it would
881: instructive to search for the transport of checkpoint proteins along
882: spindle microtubules {\it prior} to microtubule
883: attachment. Furthermore, disruption of appropriate minus end motors
884: may be able to generate premature anaphase onset by disrupting the
885: active transport inhibitory signal. As we have discussed previously,
886: it is also important to examine how any such signal is processed by
887: the spindle pole to provide inhibition.
888:
889: In general, future experimental work will need to measure more of the
890: model parameters before we can make reliable quantitative predictions
891: for intracellular signalling. Nevertheless, as we have shown,
892: computational models can play a useful role in discriminating between
893: viable and inviable mechanisms of checkpoint function.
894:
895: We would like to thank Fred Chang, Alexey Khodjakov, Yinghui Mao and
896: Kim Nasmyth for very useful discussions. MH acknowledges funding from
897: The Royal Society.
898:
899: \begin{thebibliography}{99}
900:
901: \bibitem{nasmyth05} Nasmyth, K. (2005) {\it Cell} {\bf 120}, 739-746.
902:
903: \bibitem{musacchio02} Musacchio A. \& Hardwick, K. G. (2002) {\it
904: Nat. Rev. Mol. Cell Biol.} {\bf 3}, 731-741.
905:
906: \bibitem{rieder95} Rieder, C. L., Cole, R. W., Khodjakov, A. \& Sluder,
907: G. (1995) {\it J. Cell Biol.} {\bf 130}, 941-948.
908:
909: \bibitem{rieder94} Rieder, C. L., Schultz, A., Cole, R.
910: \& Sluder, G. (1994)
911: {\it J. Cell Biol.} {\bf 127}, 1301-1310.
912:
913: \bibitem{castro05} Castro, A., Bernis, C., Vigneron, S., Labb\'e,
914: J.-C. \& Lorca, T. (2005) {\it Oncogene} {\bf 24}, 314-325.
915:
916: \bibitem{howell04} Howell, B. J., Moree, B., Farrar, E. M., Stewart,
917: S., Fang, G. \& Salmon, E. D. (2004) {\it Curr. Biol.} {\bf 14},
918: 953-964.
919:
920: \bibitem{deantoni051} Antoni, A. D., Pearson, C. G., Cimini, D.,
921: Canman, J.C., Sala, V., Nezi, L., Mapelli, M., Sironi, L., Faretta,
922: M., Salmon, E. D. \& Musacchio, A. (2005) {\it Curr. Biol.} {\bf 15},
923: 214-225.
924:
925: \bibitem{deantoni052} DeAntoni, A., Sala, V. \& Musacchio, A. (2005)
926: {\it Phil. Trans. R. Soc. B} {\bf 360}, 637-648.
927:
928: \bibitem{shah04} Shah, J. V., Botvinick, E., Bonday, Z., Furnari, F.,
929: Berns, M. \& Cleveland, D. W. (2004) {\it Curr. Biol.} {\bf 14},
930: 942-952.
931:
932: \bibitem{sironi02} Sironi, L., Mapelli, M., Knapp, S., Antoni, A. D.,
933: Jeang, K.-T. \& Musacchio, A. (2002) {\it EMBO J.}
934: {\bf 21}, 2496-2506.
935:
936: \bibitem{luo02} Luo, X., Tang, Z., Rizo, J. \& Yu, H., (2002) {\it
937: Mol. Cell} {\bf 9}, 59-71.
938:
939: \bibitem{xia04} Xia, G., Luo, X., Habu, T., Rizo, J. \& Yu, H. (2004)
940: {\it EMBO J.} {\bf 23}, 3133-3143.
941:
942: \bibitem{mapelli06} Mapelli, M., Filipp, F. V., Rancati, G.,
943: Massimiliano, L., Nezi, L., Stier, G., Hagan, R. S., Confalonieri,
944: S., Piatti, S., Sattler, M. \& Musacchio, A. (2006) {\it EMBO J.}
945: {\bf 25}, 1273-1284.
946:
947: \bibitem{tang01} Tang, Z., Bharadwaj, R., Li, B. \& Yu, H. (2001) {\it
948: Dev. Cell} {\bf 1}, 227-237.
949:
950: \bibitem{fang02} Fang, G. (2002) {\it Mol. Biol. Cell} {\bf 13}, 755-766.
951:
952: \bibitem{tang04} Tang, Z., Shu, H., Oncel, D., Chen, S. \& Yu,
953: H. (2004) {\it Mol. Cell} {\bf 16}, 387-397.
954:
955: \bibitem{pan04} Pan, J. \& Chen, R.-H. (2004) {\it Genes Dev.} {\bf
956: 18}, 1439-1451.
957:
958: \bibitem{sudakin01} Sudakin, V., Chan, G. K. T. \& Yen, T. J. (2001)
959: {\it J. Cell Biol.} {\bf 154}, 925-936.
960:
961: \bibitem{kraft03} Kraft, C., Herzog, F., Gieffers, C., Mechtler, K.,
962: Hagting, A., Pines, J. \& Peters, J.-M. (2003) {\it EMBO J.} {\bf
963: 22}, 6598-6609.
964:
965: \bibitem{acquaviva04} Acquaviva, C., Herzog, F., Kraft, C. \& Pines,
966: J. (2004) {\it Nat. Cell Biol.} {\bf 6}, 892-898.
967:
968: \bibitem{doncic05} Doncic, A., Ben-Jacob, E. \& Barkai, N. (2005) {\it
969: Proc. Natl. Acad. Sci. USA} {\bf 102}, 6332-6337.
970:
971: \bibitem{heinrich02} Heinrich, R., Neel, B.G. \& Rapoport, T.A. (2002)
972: {\it Mol. Cell} {\bf 9}, 957-970.
973:
974: \bibitem{kholodenko06} Kholodenko, B.N. (2006) {\it
975: Nat. Rev. Mol. Cell Biol.} {\bf 7}, 165-176.
976:
977: \bibitem{rieder97} Rieder, C. L., Khodjakov, A., Paliulis, L. V.,
978: Fortier, T. M., Cole, R. W. \& Sluder, G. (1997) {\it
979: Proc. Natl. Acad. Sci. USA} {\bf 94}, 5107-5112.
980:
981: \bibitem{mcewen98} McEwen, B.~F. Ding, Y. \& Heagle, A.~B. (1998)
982: {\it Chromosome Res.} {\bf 6}, 123-132.
983:
984: \bibitem{howell00} Howell, B. J., Hoffman, D.~B.,
985: Fang, G., Murray, A.~W. \& Salmon, E. D.
986: (2000) {\it J. Cell Biol.} {\bf 150},
987: 1233-1249.
988:
989: \bibitem{clute99} Clute, P. \& Pines, J. (1999) {\it Nat. Cell Biol.}
990: {\bf 1}, 82-87.
991:
992: \bibitem{wojcieszyn81} Wojcieszyn, J. W., Schlegel, R. A., Wu,
993: E.-S. \& Jacobson, K. A. (1981) {\it Proc. Natl. Acad. Sci USA}
994: {\bf 78}, 4407-4410.
995:
996: \bibitem{lang86} Lang, I., Scholz, M. \& Peters, R. (1986) {\it
997: J. Cell Biol.} {\bf 102}, 1183-1190.
998:
999: \bibitem{seksek97} Seksek, O., Biwersi, J. \& Verkman, A. S. (1997)
1000: {\it J. Cell Biol.} {\bf 138}, 131-142.
1001:
1002: \bibitem{schmiedeberg04} Schmiedeberg, L., Weisshart, K., Diekmann,
1003: S., Hoerste, G. M. Z. \& Hemmerich, P. (2004) {\it Mol. Biol. Cell}
1004: {\bf 15}, 2819-2833.
1005:
1006: \bibitem{berg93} Berg, H. C. (1993) {\it Random Walks in Biology},
1007: (Princeton University Press, Princeton), pp. 28.
1008:
1009: \bibitem{vink06} Vink, M., Simonetta, M., Transidico, P., Ferrari, K.,
1010: Mapelli, M., De Antoni, A., Massimiliano, L., Ciliberto, A.,
1011: Faretta, M., Salmon, E. D. \& Musacchio, A. (2006) {\it Curr. Biol.}
1012: {\bf 16}, 755-766.
1013:
1014: \bibitem{northrup92} Northrup, S.~H. \& Erickson, H.~P.,
1015: (1992) {\it Proc. Natl. Acad. Sci. USA} {\bf 89}, 3338-3342.
1016:
1017: \bibitem{howard01} Howard, J. (2001)
1018: {\it Mechanics of Motor Proteins and the Cytoskeleton}
1019: (Sinauer Associates, Sunderland, MA).
1020:
1021: \bibitem{wollman05}
1022: Wollman, R., Cytrynbaum, E. N., Jones, J. T., Meyer, T., Scholey,
1023: J. M. \& Mogilner, A. (2005) {\it Curr. Biol.} {\bf 15}, 828-832.
1024:
1025: \bibitem{howell01}
1026: Howell, B. J., McEwen, B. F., Canman, J. C., Hoffman, D. B., Farrar,
1027: E. M., Rieder, C. L. \& Salmon, E. D.
1028: (2001) {\it J. Cell Biol.} {\bf 155}, 1159-1172.
1029:
1030: \bibitem{gorbsky98}
1031: Gorbsky, G. J., Chen, R.-H. \& Murray, A. W.
1032: (1998) {\it J. Cell Biol.} {\bf 141}, 1193-1205.
1033:
1034: \end{thebibliography}
1035:
1036: \end{document}
1037:
1038:
1039: