Protein phosphatase 2A controls the order and dynamics of cell cycle transitions
Introduction
How
the S-phase and M-phase of the eukaryotic cell cycle are ordered, and what
makes transitions between them irreversible, are fundamental questions. Much
data suggests that cyclin-dependent kinase (CDK) activity is the key element of
this control. In fission yeast, a single mitotic CDK-cyclin complex
(CDK1-cyclin B) can promote an ordered cell cycle, implying that different CDK1 activity thresholds ensure that S-phase
and M-phase do not overlap (Coudreuse and Nurse, 2010; Fisher and Nurse, 1996). Mathematical modelling predicted that changes in
CDK1 activity may be dictated by hysteresis, characterized by different cyclin B thresholds for activation and
inactivation (Novak and Tyson, 1993), which has been confirmed experimentally (Novak and Tyson, 1993; Pomerening et al., 2003; Sha et al., 2003;
Solomon et al., 1990). Hysteresis is a property of bistable systems, and it
allows them to function as a robust switch, generating irreversible transitions
between two mutually exclusive stable states. Bistability requires positive
feedback loops, in which CDK1 inhibits its negative regulator, Wee1 (Mueller et al., 1995) and stimulates its activator, Cdc25 (Kumagai and Dunphy, 1992), but also requires ultrasensitivity (Kim and Ferrell, 2007). Mechanisms such as multisite phosphorylation and
substrate competition can partially explain ultrasensitivity, but cannot
account for it entirely (Kim and Ferrell, 2007; Trunnell et al., 2010).
Albeit
overshadowed by CDKs, phosphatases have long been known to regulate the cell
cycle (Kinoshita et al., 1990). We propose that mitosis-regulating protein
phosphatase activities (MPPs), that dephosphorylate Cdc25 and Wee1, should
contribute to bistability, since inhibition of the MPP has been predicted to
abolish hysteresis (Novak and Tyson, 1993). This is consistent with experiments in egg extracts
using phosphatase inhibitors such as okadaic acid (OA), which renders histone
H1 kinase activity proportional to added cyclin concentrations (Solomon et al., 1990). Moreover, phosphatase inhibition can also
inappropriately trigger mitosis (Izumi and Maller, 1995). The identity of the MPP is not known, although PP2A is a likely candidate as it restrains CDK1
activation and is required for dephosphorylation of CDK1 substrates including
Cdc25 and Wee1, the balance of which controls mitosis onset in interphase egg
extracts (Clarke et al., 1993; Felix et al., 1990; Lorca et al., 2010; Margolis
et al., 2006; Solomon et al., 1990). PP2A exists as a heterotrimer of catalytic (“C”),
structural (“A”), and regulatory (“B”) subunits. In fission yeast, the product
of the PP2A C gene, ppa2, is
responsible for OA sensitivity, and acts in concert with Wee1 and in opposition
to Cdc25 to inhibit mitotic CDK1 (Kinoshita et al., 1993). In budding yeast, PP2A-Cdc55 is required for
dephosphorylation of the Cdc25 and Wee1 homologues (Wicky et al., 2011). In Xenopus egg extracts, PP2A is active in
interphase and inhibited in mitosis, in a Greatwall kinase-dependent manner,
which is required to maintain CDK1 substrates in a phosphorylated state (Castilho et al., 2009; Mochida et al., 2009; Vigneron et al., 2009; Yu
et al., 2006). As Greatwall activity, in turn, depends on CDK1,
this raises the possibility of a mutual antagonism between CDK1 and PP2A.
However, it is possible that other phosphatases could contribute to MPP
activity. PP1, for example, is regulated by CDKs, can dephosphorylate CDK1
substrates (Wu et al., 2009), and a fission yeast PP1 mutant enters mitosis prematurely
(Kinoshita et al., 1990).
A
simple way of ensuring that S-phase and M-phase do not overlap would be to
employ the CDK1/MPP balance not only to promote mitosis, but also to prevent
DNA replication. High CDK1 activity inhibits replication licensing, preventing
inappropriate DNA replication (Mahbubani et al., 1997). Whether or not the CDK1/MPP activity ratio might
affect other steps of DNA replication is unknown, although PP2A, like S-phase
CDKs, is required for formation of replication pre-initiation complexes (Chou et al., 2002; Lin et al., 1998; Murphy et al., 1995; Zou and
Stillman, 1998).
Based
on these considerations, we reasoned that
it should not be the level of CDK1 activity per se, but rather the bistable CDK1/MPP activity ratio that orders
the cell cycle. Here, we show that elimination
of MPP activity abolishes bistability in the CDK1 activation network and allows
entry into mitosis at very low CDK1 activity levels, bypassing even the mitotic
requirements for Cdc25. Secondly, the CDK1/MPP ratio regulates not only mitotic
onset, but also DNA replication, and the mitotic state, with high CDK1 and low
PP2A activity, also prevents post-licensing steps of DNA replication, including
the chromatin-loading of S-phase CDKs. Thirdly, the major MPP is PP2A-B55. We
demonstrate these findings theoretically and experimentally, using the Xenopus
egg extract system that facilitates study of core cell-cycle system components.
We propose a simple mathematical model in which PP2A is required for
bistability of CDK1 activity and contributes to a robust separation of S-phase
and mitosis.
Results
Bistability
in the CDK1 control network requires PP2A activity
The mitotic CDK1 control network contains several
regulators connected by feedback loops, whereby CDK1 stimulates Cdc25 and
inhibits Wee1 (Figure 1A). The network generates two stable steady states with
low and high CDK1 activities, respectively, separated by unstable steady states
(Novak and Tyson, 1993; Pomerening et al., 2003). Transitions between these states are switch-like and
occur at different cyclin thresholds for CDK1 activation and inactivation,
creating S-shaped balance curves. Bistability requires the CDK1-Cdc25 / Wee1
feedback loops, but earlier modelling predicted that it also requires a protein
phosphatase (which we name MPP) to counteract effects of CDK1 on Wee1 and Cdc25
(Novak and Tyson, 1993). Recent experiments suggest that PP2A might be such
an MPP (Mochida et al., 2009). Thus, Wee1 and Cdc25 are phosphorylated by CDK1 and
would be dephosphorylated via the CDK1-regulated MPP, PP2A, creating coherent
feed-forward loops in the network. This should increase the distance between
the two cyclin thresholds and make the bistable switch more robust (Novak et al., 2010). Using this information, we have developed a model of the bistable switch controlling mitotic
entry which includes CDK1-dependent regulation of the MPP (Figure 1,
Supplementary methods). To
analyse the network behaviour, we calculated steady-state levels of cell cycle
regulators (Figure 1B-F). The low CDK1 and high MPP state constitutes S-phase
while the high CDK1 and low MPP activity state corresponds to M-phase (Figure
1B, C). Entry into mitosis depends on mitotic cyclin accumulation. The S-phase
steady state with low cyclin levels and CDK1 activity is replaced by the
mitotic steady state at cyclin levels of about 0.6 (Figure 1D & E). Between
cyclin levels of 0.3 and 0.6 the control system is bistable (Figure 1D &
E). The CDK1/MPP activity ratio changes drastically between the two steady states,
which translates into a switch-like change in the phosphorylation state of
generic CDK substrates (Figure 1F, bold line). We then analysed the behaviour
of the network in response to MPP inhibition (Figure 1B-F, +OA). With enough inhibition, bistability is lost, together
with the S-phase steady state allowing DNA replication (Figure 1C). This is
confirmed by time-course simulations of the model, where the mitotic state is
induced at low cyclin levels with S-phase timing (Figure 1G & H). Interestingly even lower CDK1 activity than normally
present in early interphase (for example, after CDK1 inhibition) should allow mitotic entry, although with a delay
(Figure 1I).
To
experimentally test these predictions, we used early interphase Xenopus egg
extracts, that have low CDK activity (Krasinska et al., 2008a), supplemented with sperm chromatin, and cycloheximide
to prevent cyclin synthesis. If MPP activity is mainly due to PP2A, OA
inhibition should lead to inhibition of S-phase and induction of mitosis in the
absence of cyclin translation. Addition of 1mM OA prevented DNA replication (Figure 2A), as expected (Chou et al., 2002; Lin et al., 1998; Murphy et al., 1995), and also provoked chromatin condensation, nuclear
membrane breakdown and disassembly of lamina in all nuclei (Figure 2B). OA
addition also caused dephosphorylation of CDK1Y-15 and the electrophoretic
mobility shifts of Cdc25, Wee1 and Gwl, a hallmark of their mitotic
hyperphosphorylation (Figure 2C) that always correlated with nuclear breakdown
and was subsequently used as a mitotic marker. Addition of recombinant cyclin B
lacking its N-terminal 90 residues (cyclin B∆90) reproduced the effects of OA
(Figure S1A and B), in accordance with mathematical modelling (Figure S1C).
Fostriecin, a potent PP2A-family inhibitor with 10,000-fold selectivity over
PP1 (Walsh et al., 1997), also abolished DNA replication while inducing
mitotic onset (Figure 2D). OA generated three-fold higher CDK activity than is
normally present in early interphase egg extracts, that was sufficient to
trigger mitosis despite being much lower than that obtained by cyclin B∆90
addition (Figure 2E). Partial PP2A inhibition reduced the concentration of
cyclin B needed to promote complete phosphorylation of CDK1 substrates and
therefore mitotic onset (Figures 1F, 2F, and S1D). From these data we conclude,
firstly, that PP2A is a major determinant of the threshold CDK activity
required for mitotic entry; secondly, that inhibiting PP2A unmasks the mitotic
potential of the CDK activity already present in early interphase. This
activity is probably due to both cyclins A and B, as CDK1-cyclin A is active in
interphase Xenopus egg extracts (Krasinska et al., 2008a; Strausfeld et al., 1996), and we observed low levels of cyclin B2 (Figure
S1E).
Our
mathematical model predicts that if PP2A is inactivated by OA, even a
significant reduction in CDK activity should only delay, but not abolish,
mitotic onset (Figure 1I). To test this, we investigated CDK requirements for
mitotic onset when PP2A is inhibited in early interphase. In Xenopus egg
extracts CDK2 activity promotes CDK1 activation by phosphorylating Cdc25 (Margolis et al., 2003). However, CDK2 depletion had no effect on mitotic
entry induced by 1mM OA (Figure 3A), indicating that its role is
bypassed, as expected if MPP is PP2A. Depletion of the majority of CDK1,
however, delayed but did not prevent mitosis (Figure 3A). Additionally removing
CDK2 using the small CDK-binding protein p13Suc1 did not accentuate
the delay, indicating that roles of CDK1 and CDK2 are different (Figure 3B). We
obtained similar results using specific inhibitors NU6102 and RO-3306, which in
extracts inhibit CDK2 and CDK1, respectively (Krasinska et al., 2008a; Krasinska et al., 2008b). Increasing concentrations of the two inhibitors
progressively reduced CDK activity back to interphase levels while increasingly
delaying, and, at the highest concentrations, preventing OA-induced mitotic
onset (Figure 3C). The same result was obtained when we added recombinant p21Cip1/waf1
protein, which inhibits both CDK1 and CDK2 (Figure 3D). These results confirm
that the balance of PP2A and CDK1 activities determines the dynamics of mitotic
entry. We imagined that loss of switch-like activation of CDK1 might lead to
asynchrony of mitotic onset. Indeed, whereas with OA alone mitotic onset was
synchronous, upon decreasing CDK activity both mitotic and interphase nuclei
were present at some time points (Figure S2).
Entry
into mitosis normally requires positive feedback in which CDK1 activates Cdc25
and inhibits Wee1. Mathematical modelling predicts that MPP inhibition should
reduce requirements for Cdc25 in mitotic onset, because Wee1 should still be
inhibited (Figure 4A). We therefore compared effects of depleting Cdc25, and/or
inhibiting ERK1/2, as it was reported that ERK1/2 kinases can phosphorylate and
stimulate Cdc25 activity (Wang et al., 2007). UO126, which inhibits ERK1/2 activation, only
marginally reduced CDK1 activity and did not prevent OA-induced mitotic onset
(Figure 4B). In contrast, immunodepletion of around 90% of Cdc25 strongly
reduced CDK1 activity. Nevertheless, as predicted, Cdc25 depletion had no
effect on mitotic onset induced by OA, even when combined with ERK1/2
inhibition (Figure 4B). These results reinforce our conclusion that CDK1
activity requirements for mitosis are very low when MPP is inhibited, and show
that ERK1/2 cannot substitute for CDK1 activity to trigger mitotic onset in the
presence of OA.
Because the Cdc25/ Wee1 balance is a target of
the checkpoint that restrains mitotic entry during DNA replication (Dasso and Newport, 1990), we then asked whether the checkpoint would affect
MPP requirements. Mathematical modelling predicted that unreplicated DNA could
increase the cyclin threshold for CDK1 activation by increasing MPP activity (Novak and Tyson, 1993; Pomerening et al., 2003). If so, decreasing the amount of unreplicated DNA
should allow entry into mitosis with lower OA levels. This was confirmed
experimentally: at nuclear concentrations that mimic early embryonic
development (1400 nuclei/ml), 0.5mM OA is insufficient to promote mitosis (Figures 2F,
4C) but it is sufficient at 140 nuclei/ml (Figure
4C). Note that the biochemical state
characteristic of mitosis does not require the presence of DNA. Therefore,
unreplicated DNA alters the balance between CDK1 and PP2A activity as expected,
providing further evidence that PP2A functions as the MPP.
The S-phase
stable steady state: replication licensing and initiation stages are
differentially sensitive to PP2A inhibition
In early interphase, PP2A is active, CDK1 activity is
low, and DNA synthesis can take place. The mathematical model predicts that
this state should be robust with respect to small increases in cyclin and PP2A
levels, which therefore should not affect DNA replication. We verified these
predictions experimentally: replication rates are neither reduced by addition
of recombinant cyclin B∆90 (Figure S3A) at levels that do not provoke mitosis entry (Figure
S1D) nor enhanced by addition of purified PP2A (Figure S3A). We next
investigated the mechanisms of the block to DNA replication upon PP2A
inhibition. Chromatin loading of pre-replication complex (pre-RC) proteins,
including Cdc6, Cdt1 and MCMs, occurred very early, but Cdc6 later disappeared
from chromatin, coinciding with electrophoretic retardation of cytosolic Cdc6
and other pre-RC proteins, indicating the mitotic state (Figure 5A). Thus,
mitotic hyperphosphorylation "unlicenses" origins, preventing their
conversion to pre-initiation complexes (pre-ICs). We next asked whether the
mitotic state is required for the inhibition of DNA replication. In agreement
with simulations (Figure S3B), low concentrations of PP2A inhibitor decreased
replication rates but did not prevent replication (Figure S3C). Interestingly,
when we reduced PP2A to intermediate levels (50-100mM fostriecin or 0.7mM OA), onset of mitosis did
not occur, but DNA did not replicate (Figure 5B and S3D). Replication also did
not occur when we prevented OA from inducing mitosis by CDK inhibition (Figure
S3E). These results suggest that either there is an intermediate steady state
in which the CDK1/MPP activity ratio is too low for mitotic onset but too high
for S-phase, which is not expected from the model, or PP2A promotes replication
by counteracting other kinases along with CDK1. Either way, this could provide
a buffer ensuring robust separation of S-phase and mitosis. To investigate the
nature of this state, we analysed pre-RC and pre-IC formation when PP2A is
partially inhibited. Figure 5C shows that licensing can occur, but pre-IC
components are absent from chromatin. Pre-IC loading requires S-phase CDKs (Mimura and Takisawa, 1998; Zou and Stillman, 1998), which in Xenopus is fulfilled by CDK2-cyclin E, and
cyclin E and CDK2 were absent from chromatin at intermediate PP2A activity
levels (Figure 5C, E, F). We could detect PP2A subunits on chromatin early
during DNA replication where they directly interacted with cyclin E (Figure
S3F, G). These results suggest that PP2A acts on chromatin to control S-phase
CDK loading. Indeed, OA no longer had any effect when added at the point where
CDKs become dispensable for DNA replication (Figure S3H). One possibility is
that an intermediate CDK1/PP2A ratio prevents loading of S-phase CDK onto
chromatin and thus also prevents DNA replication, an unexpected variation on
the simple bistability model. If so, then modelling predicts that sufficiently
reducing CDK1 activity should restore DNA replication (Figure 5D). When we
titrated OA to minimal levels that inhibit replication, we could partly restore
replication by inhibiting CDK1 with RO-3306 (figure S3I).
PP2A
counteracts additional mitotic kinases to allow cyclin E-CDK2 loading onto
chromatin in S-phase
One explanation for the incomplete replication rescue
by CDK1 inhibition is that additional kinases opposed by PP2A ensure that DNA
replication and mitosis do not overlap. Given that chromatin binding of cyclin
E is inhibited by phosphorylation (Furstenthal et al., 2001), and that cyclin E can be phosphorylated by multiple
kinases including GSK3, ERK1/2, CDK1, and CDK2 itself (Furstenthal et al., 2001; Welcker et al., 2003), we suspected that PP2A counteracts not only CDK1 but
also another kinase to allow cyclin E loading and replication origin firing. To
test this idea, we used staurosporine, a broad-spectrum kinase inhibitor. In
the absence of OA, 2mM staurosporine reduced replication efficiency (Figure
S3J), presumably because it also partly inhibits S-phase CDKs. Nevertheless,
whereas 0.7mM OA alone abolished DNA replication, addition of 2mM staurosporine restored cyclin E-CDK2 loading onto chromatin and
allowed DNA replication in some nuclei (Figure 5E). Inhibition of GSK3 with
AR-A014418 had no effect (data not shown). Preventing ERK1/2 activation with
UO126 restored most cyclin E loading (Figure 5F), although had little effect on
replication (figure S3K). Combining UO126 and RO-3306, to counteract both
ERK1/2 and CDK1, fully restored cyclin E levels on chromatin, and around 40% of
DNA was replicated (Figure 5F). Further adding staurosporine to the inhibitor
cocktail, however, was not effective (Figure S3K), perhaps because CDK activity
is overly reduced. We then added a recombinant CDK2-cyclin E complex in which
we mutated all ERK/CDK consensus phosphorylation sites (supplementary
information). Although active, this did not rescue the block of DNA replication
due to partial OA-inhibition (Figure S3L). It has also been reported that
caffeine, which inhibits ATM/ATR checkpoint kinases, could restore DNA
replication in the absence of PP2A (Petersen et al., 2006). In our hands, caffeine alone did not affect the
block to DNA replication by OA (Figure S3M), but, once the OA block is overcome
using UO126 and RO-3306, caffeine slightly boosts replication rates (Figure
S3N). Finally, we noticed that partial PP2A inhibition provoked strong chromosome
condensation and histone H3 phosphorylation on serine 10, even in the absence
of mitosis (Figure 5B and S3O), providing a potential hindrance to replication
fork progression. Although inhibition of Aurora kinases with ZM4474439 could
abrogate the increase in H3S10 phosphorylation, it could not restore
replication. We conclude that there are multiple additional kinases, superimposed
onto the bistable network and not considered in our mathematical model, which
restrain replication when PP2A is partially inhibited, and that replication and
mitosis are incompatible at many levels.
PP2A-B55
complexes are the major PP2A complexes involved in control of DNA replication
and mitosis
Inhibition of the PP2A-B55d complex is required for mitotic entry, but its depletion does not
provoke mitotic entry in interphase extracts without cyclin synthesis (Castilho
et al., 2009; Mochida et al., 2009), unlike complete inhibition by OA or
fostriecin (Figure 2). Therefore, other PP2A complexes suffice to maintain the
interphase state, or the low levels of PP2A-B55d remaining after depletion are sufficient. To distinguish between these
possibilities we attempted to deplete the four B subfamilies using antibodies
raised against B55d, B56e, B56g and B'/PR48, and assayed DNA replication. The B55d antibody only slightly reduced abundance of the two bands recognised.
Figure 6A shows that none of the depletions significantly affected DNA
replication. Multiplex depletions were not efficient, so we used microcystin
covalently immobilised onto agarose beads, reported to specifically remove PP2A
from Xenopus extracts (Maton et al., 2005). To determine the composition of any phosphatase
complexes removed, we analysed the material isolated by the control and
microcystin agarose depletions by semi-quantitative tandem mass spectrometry
(Figure S4). The major phosphatase identified was indeed PP2A, although PP1,
PP4, PP5 and PP6 subunits were also present, to a lesser extent. Interestingly,
the predominant B subunits of PP2A identified were B55a and B55d, although several B56 subunits were also present, at
lower levels (Figure S4B). Western blotting analysis showed that microcystin
depletion led to nearly complete depletion of PP2A catalytic and structural
subunits (Figure 6B). The PP2A family members PP4 and PP6 were also
significantly depleted, but PP1 and PP5 less so. Both bands recognised by the
PP2A-B55d antibody, which we infer from proteomics data to be
the highly related B55a and B55d, were depleted, but no change was detected for the other B subunits. The
result of the depletion was either mitotic onset from interphase, or, in other
experiments, inhibition of replication without mitotic onset (Figure 6C, left
and right panels respectively), as expected. Unlike inhibition of PP2A by OA or
microcystin depletion, inhibition of PP1 by tautomycin, which has a similar IC50
for PP1 as OA for PP2A, or inhibitor-2 protein, a specific PP1 inhibitor (Cohen, 1989), did not affect DNA replication nor induce mitosis
(Figure 6D); on the contrary, PPI2 delayed cyclin B∆90-induced mitotic onset
(Figure 6E). At low concentrations, due to its high affinity essentially all OA
will be complexed to PP2A. We therefore titrated OA to the minimal
replication-inhibitory concentration for individual extracts, and then added
purified PP2A complexes. This partially rescued replication (Figure 6F).
Finally, to definitively demonstrate the role for PP2A-B55, we exploited the
recent discovery that Greatwall-phosphorylated Arpp19/endosulfine is a specific
inhibitor of PP2A complexed with B55 subunits (Gharbi-Ayachi et al., 2010; Mochida et al., 2010). To reduce specific phosphatase activity to a level
allowing inhibition by recombinant proteins, we primed the extract with 0.1mM OA, which had no phenotype alone. We immunoprecipitated Greatwall
kinase from a mitotic extract and used it to thiophosphorylate recombinant
Arpp19, which we then titrated into the extract. High or intermediate Arpp19
levels inhibited DNA replication and caused mitotic entry, whereas low levels
blocked replication in interphase (Figure 6G). Therefore, while we cannot
formally rule out redundancy with PP2A in complex with other B subunits, and/or
PP1, PP4, PP5 or PP6, we show that PP2A-B55 complexes are essential in
promoting replication and preventing mitosis.
All stages
of DNA replication are incompatible with mitosis
Our results indicate that the cell cycle has several
mechanisms preventing overlap of S- and M-phase. Nevertheless, one report found
that mitotic chromatin can replicate if a concentrated S-phase nucleoplasmic
extract is added (Prokhorova et al., 2003). We therefore asked whether, once initiated,
replication can continue in mitosis. First, we allowed chromatin to form with
or without OA for the time normally taken to form active pre-ICs, then transferred
it into control extracts, or extracts rendered mitotic either by OA treatment
or addition of cyclin B∆90. Replication in the donor and recipient extracts was
monitored by two-colour immunofluorescence microscopy and quantified by
incorporation of 33P-dCTP (Figure 7A). Pre-ICs formed in control
extracts replicated only briefly in mitotic extracts, which probably reflects
the lag before the nucleus breaks down and chromatin becomes mitotic, before
arresting (Figure 7A, B, S5A, B). To formally exclude the possibility that
replication forks can progress in the mitotic state, we first allowed origins
to fire by incubating chromatin in an egg extract containing aphidicolin to
prevent any elongation. Chromatin was then transferred to a mitotic extract
with aphidicolin until it became mitotic. This mitotic post-initiation
chromatin was reisolated and transferred to a third extract, also mitotic but
without aphidicolin, thereby potentially allowing elongation, which was
quantified (Figure 7C). Elongation did not occur (Figure 7C). Taken together,
these results show that all steps of DNA replication, pre-RC assembly, pre-IC
formation, and elongation, are incompatible with mitosis, ensuring a robust
ordering of the cell cycle.
Discussion
PP2A appears to act as a central modulator of
phosphorylation to generate two stable states, interphase and mitosis. To
illustrate this, an analogy is presented in Figure 7D, which depicts a
traditional Japanese fountain known as a shishi
odoshi. A water flow generates oscillations between two states, due to the
cylinder filling up until it overbalances, whereupon it empties and swings back
again. In our analogy, the flow represents phosphorylation due to CDK1. PP2A is
represented by the cylinder, which continuously dephosphorylates mitotic CDK1
substrates. In interphase, PP2A prevails, preventing mitotic onset and allowing
S-phase (Figure 7D, left). But as a result of increasing CDK1 activity, which
activates Greatwall kinase to generate a PP2A inhibitor, PP2A becomes
overwhelmed and phosphorylation of mitotic CDK1 substrates can occur, bringing
about mitosis while simultaneously preventing DNA replication (Figure 7D,
right). Several points of this analogy, supported by our experimental
observations, are worth highlighting. Firstly, S-phase and mitosis are
inversely regulated by mitotic CDK1-phosphorylation and never overlap, whether
PP2A is active or not. Secondly, artificial inhibition of PP2A (the cylinder
forced right) should allow even low-level CDK1 activity to bring about mitosis,
whose timing should depend on CDK1 activity. Thirdly, since CDK1 substrates are
involved in the feedback loops, PP2A should be a critical component of these
loops. As such, it is likely to contribute to the ultrasensitivity of the
regulatory network to increasing CDK1 activity. Several mechanisms for
ultrasensitivity have been proposed, involving multisite phosphorylation of the
regulatory enzymes, including allostery and competition among phosphorylation
sites, but ultrasensitivity found in extracts could not be recapitulated with a
minimal network in vitro (Kim and
Ferrell, 2007; Trunnell et al., 2010). Although inclusion of a CDK1-regulated
phosphatase could theoretically generate ultrasensitivity (Novak et al., 2010),
so far this has not been experimentally proven; addition of purified PP2A
catalytic subunit alone did not restore the ultrasensitivity found in extracts
(Trunnell et al., 2010). The most likely reason is that in this case, the
phosphatase activity is not regulated by CDK1. We nevertheless find that
inhibiting PP2A eliminates bistability in the extract. We suggest that to
reproduce physiological ultrasensitivity in
vitro, it would be necessary to reconstitute a network containing PP2A-B55,
the PP2A inhibitor Arpp19/ENSA, Greatwall kinase, and the currently unknown
components allowing Greatwall to be regulated by CDK1.
We
find that mitosis and DNA replication are fundamentally incompatible due to
opposite regulation by phosphorylation. One aspect of this incompatibility
might be the requirement for a nucleus to replicate. In yeast, unlike
metazoans, the nuclear membrane remains intact during mitosis, suggesting that
replication and mitosis might in some experimental circumstances become
compatible. To date, although the present and many other studies performed in
different model systems have found that the mitotic state can be experimentally
induced before completion of DNA
replication, only one report, to our knowledge, describes ongoing replication
in mitosis (Prokhorova et al., 2003). Interestingly, in this study, replication
was induced using a concentrated nucleoplasmic extract, which bypasses the need
for a nuclear membrane (Walter et al., 1998). We find by serial transfer of post-initiation
chromatin that replication elongation does not occur in mitotic extracts. Taken
together, these observations suggest that addition of concentrated
nucleoplasmic extracts must also somehow modify the mitotic state – perhaps by
restoring some PP2A activity, although this would be expected to cause exit
from mitosis.
In our model, the CDK1/PP2A activity ratio determines whether S-phase or
M-phase will occur. There is an interesting parallel in budding yeast. A
distinct phosphatase, Cdc14, reverses CDK1 phosphorylation at the exit from
mitosis, and has recently been found to be required for DNA replication, where
it acts to permit a sufficient accumulation of replication factors in the
nucleus (Dulev et al., 2009). Xenopus Cdc14 homologues also have roles in control
of mitosis and can dephosphorylate Cdc25 (Krasinska et al., 2007), although whether they are required for S-phase is
not clear. If so, they could potentially provide another barrier to DNA
replication within mitosis. Our shishi
odoshi analogy also suggests a mechanism for "phase-locking" of
the Cdc14 autonomous oscillator to the CDK oscillator, as proposed recently (Lu and Cross, 2010). If Cdc14 rather than PP2A oscillations are
represented by the cylinder, it can be seen that they indeed do not require
oscillation of the water flow; nevertheless, the water flow itself is required,
just as Cdc14 oscillations disappear if all CDK1 activity is abolished.
Furthermore, the frequency of oscillations will depend on the water flow-rate.
The phase-locking comes about because of the double-negative feedback between
the oscillator and the water flow. Although not represented in the analogy, in
reality both Cdc14 and PP2A activities lead to CDK1 inactivation – in other
words, the water flow is not constant, but increases and decreases with the
same frequency as the autonomous oscillator, locking the two oscillation
frequencies together. Certainly, closer study of the relationships between
Cdc14 and PP2A in yeast and vertebrates will be important in order to build a
universal model of how phosphatases control the cell cycle.
Post Comment
No comments