The Eukaryotic Cell Cycle: Molecules, Mechanisms and Mathematical Models
Sep 5, 2023
April 3, 2013. BioEngineering Seminar Series. University of Illinois Urbana-Champaign
"The Eukaryotic Cell Cycle: Molecules, Mechanisms and Mathematical Models"
Dr. John Tyson, University Distinguished Professor, Department of Biological Sciences, Virginia Tech
Thank you very much, Ting, for that nice introduction, and thanks to the
Department of Bioengineering for the invitation, and thank you all for coming
out. This is a lovely audience, and I hope I can entertain you for the next 50
minutes or so with tales of eukaryotic cell cycle control in yeast. So I should
give some credit to NIH, General Medical Sciences, for supporting this work pretty
faithfully for the last 10 years or so. So let me start with physiology, what the
cell cycle is all about, what we're trying to understand, and then I will
talk a little bit about the molecular biology that underpins this physiology,
and then we'll talk about the general principles of cell cycle control and
mathematical modeling. So the cell cycle is the sequence of events by which a
growing cell replicates all of its components and divides them more or less
evenly between two daughter cells, so that the daughter cells inherit the
machinery and the information that's necessary to repeat the process. As you
probably know, the cell cycle is central to everything in biology because all
cells come from previously existing cells by this process of cell
replication and cell division. So all living organisms are made of cells and
all cells come from previously existing cells by the process of the cell cycle.
So understanding the molecular control of the cell cycle is truly a central
problem in the biological sciences, and it's also of great importance in the
health sciences because, as you know, cancer is a set of
diseases which are all related to misregulation of cell growth and cell
division and cell death pathways. But one may ask, if we're really interested
ultimately in understanding cancer cells and how to treat cancer cells, why would
we start with budding yeast cells and spend so much time understanding budding
yeast? Because budding yeast cells don't get cancer, budding yeast cells don't pay
big hospital bills and things like that. However, you're all engineers and you
know if you wanted to be a car mechanic, where would you start? Would you start
with an old Beetle or would you start with a hundred thousand dollar Audi
sports car, right? Of course you'd start with the old Beetle and you'd tinker
with that and understand how it works and the sports car is a kind of direct
lineal descendant and all the basic power train and control systems are
there in the Beetle. That's what you want to understand and that's why we want to
understand what budding yeast cells are doing. They're a wonderful type of cell
for doing experiments, for gathering great data sets, and what we learn from
understanding control in budding yeast cells carries over quite directly to
mammalian cells, but I'm not going to talk about mammalian cells at all today.
Okay, there are two aspects of cell cycle that I want to focus on in this lecture.
The first is what we call the alternation of DNA synthesis and
mitosis. Okay, so the most important molecules in the cell that need to be
replicated and divided to the daughter cells are the DNA molecules, the
hereditary molecules, the genome, and during the eukaryotic cell cycle, each
DNA molecule is replicated during a particular phase of the cycle, the phase
of DNA synthesis or S phase of the cell cycle. So in the cartoon here, let's see,
how do I get them? Okay, in the cartoon that would be the cell membrane and the
nuclear membrane and two chromosomes in the cell, a blue chromosome and a red
chromosome, and during S phase, those chromosomes are replicated and that was
probably an important piece that just fell out, right? It's under your foot. I
think that's what, okay, that's for changing them. Yeah, for changing. Oh, I
don't need that, do I? Okay, just take it. Okay, where was I? So during S phase of the
cell cycle, the DNA is replicated. Each double-stranded DNA molecule is copied,
so you get two identical double-stranded DNA molecules. They're called
sister chromatids. So at the end of S phase, there are two blue DNA molecules,
there are two red DNA molecules. They're identical molecules and they're held
together by glue. That's the yellow stuff in between. Okay, so the sister chromatids
are glued together during S phase. Then later, after a gap, the cell goes into
mitosis and in mitosis, those sister chromatids are to be partitioned to the
two daughter cells. Each daughter cell gets one and only one copy of each DNA
molecule and the way that works, you probably know, is the cell builds a
mitotic spindle. So there are two spindle poles and from the spindle poles come
threads called microtubules and those microtubules capture the replicated
chromosomes at particular locations in the center called kinetochores. The
microtubules capture the chromosomes and the idea is that one sister chromatid is
to be attached to one pole of the spindle and the other to attach to the
other pole of the spindle. That's a proper alignment of the replicated
chromosome on the metaphase plate. When all of the chromosomes are properly
aligned on the metaphase plate, then a signal is set out to continue through
the cell cycle and the next thing that happens is that the glue that's holding
the chromosomes together, the glue is dissolved and then the microtubules can
separate the sister chromatids and they're carried to opposite sides of the
cell and new nuclear membranes are built around those chromosomes and the cell
divides down the center and you get two daughter cells that are ready to repeat
the process. So it's very important that this process of copying the DNA
and then partitioning the sister chromatids, the two daughter cells, that
those processes alternate as you go through the cell cycle. It's just
like you have a Xerox machine and a collator. So you put your
paper into the Xerox machine, you get two copies of every page, then you put
it into the collator and the collator separates it so you get one complete
book. And those occur at different phases in the cell cycle. So this is the
Xerox machine and this is the collator. And you have to go around the
cycle, first copying and then collating, then copying then collating. It's very
important that you don't, for instance, slip back. So you go from the G1 phase
where the chromosomes are unreplicated through S phase the chromosomes are
replicated. You don't want to slip back and do DNA synthesis again because then
you'll get four copies of every chromosome. And that's not a good idea.
It's not a fatal mistake, but cells don't like to be polyploid like that.
Even worse would be if you go into mitosis and separate the chromosomes and
come into G1 and then you slip back and you try and go into mitosis again
without doing DNA synthesis. That's a fatal mistake and the cells will die if
they do that. You can mutate cells so that they make these mistakes and
they're very unhappy or they're dead, but it's very important for the normal
control system to assure that there's irreversible progression around the
cell cycle. G1S, G2M, G1S, G2M. So that's one thing I want to understand, why this
irreversible progression. Yes, question.
Yeah, so trisomy 21 actually happens during meiosis, not during mitosis. Meiosis
is this different process where the homologous chromosomes have to be
separated as you form the germ cells. It's a very similar, but it's but it's
also different, okay? So if there's a mistake there, then you get the trisomies.
Yeah, okay. The other question I want to address is you need a molecular control
system that's going to organize these events in time and make sure that they
get this alternation and you get these irreversible transitions, but that
molecular control system has to function reliably in the face of unavoidable
molecular noise because, let's face it, a yeast cell is very small and it doesn't
have very many copies of the molecules that have to implement this control
system, okay? So it's not like a chemical reactor with a mole's worth of
substances. You only have one copy of every gene, you have a handful of copies
of messenger RNAs, you might have a hundred or a thousand copies of the
proteins, so there's bound to be stochastic fluctuations in this control
system and we want to make sure that it operates reliably in the face of that
kind of noise. So that'll be the second half of the talk. Any other questions to
start? Okay, so that's the basic physiology. That's what we would like to
understand. Now we have to know learn a little bit about the molecular biology.
What are the underlying proteins that control the events of the cell cycle? In
all eukaryotic cells that have been studied in any detail, it's always the
same group of proteins. These proteins are called cyclin-dependent kinases, CDK,
cyclin-dependent kinases. These are protein kinases. That means they are
enzymes that phosphorylate other proteins, okay? So if I'm a protein kinase
and you're a target protein, I add a phosphate group to you and that phosphate
group carries a lot of negative charge. It causes the phosphorylated proteins
to change their shape and to change their activity. So when the target
protein gets phosphorylated, it can be turned on or it can be turned off, okay?
So the way the cyclin-dependent kinases control events in the cell cycle is by
phosphorylating proteins involved in DNA synthesis and later phosphorylating
proteins involved in mitosis. And whatever is done in one phase of the
cell cycle has to be undone later because it's a repetitive process. So
wherever there's a kinase putting phosphate groups on, there's a
phosphatase enzyme later taking the phosphate groups off, okay? The
cyclin-dependent kinases, as their name suggests, require a partner in order to
be active. Those partners are called cyclins, and in yeast the cyclins are
called CLIN2, CLIB5, and CLIB2, okay? Those are the cyclin partners of the
kinase. These partners drive the very earliest event of the cell cycle, which
is called start. It's when the yeast cell commits itself to a new round of DNA
synthesis and division. CLIB5-dependent kinase is important in driving the
events of DNA synthesis, and later CLIB2-dependent kinase drives entry
into mitosis and blocks exit from mitosis, blocks the metaphase to anaphase
transition when the glue is dissolved, okay? So it's the rise in activity of
these cyclin-dependent kinases that drives the cell out of G1 into S phase
and then into M phase and causes the replicated chromosomes to align on
the metaphase plate. In order to finish the cycle, we have to get rid of the
cyclin-dependent kinases and dephosphorylate all the proteins, and
there we need a different set of components. So there's a protein
complex called the anaphase-promoting complex. The anaphase-promoting complex
essentially promotes the dissolution of the glue, which allows the sister
chromatids to be pulled apart. That's called anaphase. So it promotes anaphase
by dissolving the glue, and it promotes exit from mitosis by degrading the B-type
cyclins, CLIB2 and CLIB5, okay? So it gets rid of the things that brought you
into mitosis to allow you to get out of mitosis. The anaphase-promoting complex
needs a partner to direct its activity to the right places. That partner is
called CDC20, which appears very abruptly as cells exit from mitosis. And
then later, CDC20 is replaced by a different partner called CDH1, but
basically they do the same thing. And finally, in G1, there's a
stoichiometric inhibitor of the cyclin-dependent kinases, which can bind
the cyclin-dependent kinases and keep them inactive. I'll say a little bit more
about these in just a second, but are there any questions so far? Okay, a lot of
new friends that you're just meeting. I hope you get the basic idea. Cyclin-
dependent kinases drive you into S-phase, up to metaphase, and then you have to get
out so you can start over again. Okay, what did I do? Sorry. Now, we know a lot
about the molecular biology of the interactions amongst these proteins,
these regulatory proteins, and this is how we represent that interaction
network. If you were an electrical engineer, it would be a wiring diagram
for a VCR or something like that. We're bioengineers. Our components are
these different protein complexes. The solid arrows are chemical reactions, so
this describes the synthesis of Glyb2. It combines with the kinase to form an
active protein kinase, and this is the degradation of Glyb2 here, and the
degradation of Glyb2 is driven by these two components, Cdc20 and Cdh1, and then
there are phosphorylation and dephosphorylation reactions. For instance,
the Glyb2-dependent kinase can phosphorylate Cdh1 and inactivate it, and
there's a phosphatase called Cdc14 that can dephosphorylate Cdh1 and activate
it, okay? So that's how you read a diagram like this. These are each
individual biochemical components. They have reactions that produce them and
destroy them, and they're enzymes, so they catalyze other reactions in the network.
It's a very dense network with lots of feedback and feedforward, positive
feedback loops and negative feedback loops, and our job is to try to determine
whether this mechanism can account for the known features of cell cycle
regulation in budding yeast, okay? I don't think it's obvious from looking at the
network, and as a matter of fact, the network I think is so complicated that I
would be hesitant to rely completely on my intuition to figure out how a network
like this is going to behave, okay? And what we've been trying to do for many
years now is to convert mechanisms like this into mathematical models, be very
precise about our assumptions of how it works, and then do analysis and
computation on those models, compare to experiments, and see whether this network
really is consistent with the way yeast cells behave. Think of this as a
hypothesis, right? We don't know that this is true. We think it's true. How do you
tell whether it's true? Well, it needs to fit the facts. Harry, did you have a question?
Well, as you indicated, with a low number of these molecules in the cells, so there's a
potential for all of these reactions to occur, but there's a physical
coincidence in space that has to be present. Right, so we'll get to
that, okay? But first I want to do a deterministic model with differential
equations, and then we'll convert it into a stochastic model. Alex?
Is the translation of genes into these proteins and the regulation of the genes part of that method?
Yes, it is. So, for instance, MCM1 is a transcription factor, and that little
dashed arrow means that's the transcription factor that activates the
gene that encodes Clib2, okay? So there are a number of transcription factors
here that aren't, and there's messenger RNAs that I haven't even
indicated on this diagram, okay? So I want to start with deterministic models and
give credit to my collaborators, my longtime collaborators, Bela Novak and
Kathy Chen, who've been working on this project from the very beginning, and two
postdocs, they're now established scientists in their own right, Attila
Cekosnagy and Andrea Ciliberto, who were involved in this from the
beginning. Okay, so the way I want to describe how this network functions is
to divide it into kind of simpler modules and try to convince you that
there are little regulatory modules here that basically control the properties of
this regulatory network, and I want to focus in on Clib2-dependent kinase and
CDH1. Remember, CDH1 is involved in a proteolytic pathway that destroys Clib2
and therefore destroys the activity of the kinase, because the kinase has to
have a cyclin partner in order to be active, okay? On the other hand, Clib2 is
a protein kinase that can phosphorylate CDH1 and inactivate CDH1, okay? So CDH1
is an inhibitor of Clib2-dependent kinase, and Clib2-dependent kinase is an
inhibitor of CDH1. So these two protein complexes are involved in mutually
inhibitory connections, and you might guess that at least it's possible here
to get two different states of the control system. There could be a state in
which CDH1 is active and Clib2 is rapidly degraded, so you can never get
the kinase activity, and that we would call the G1 phase of the cell cycle. When
CDH1 is active and there's no kinase activity, you're in the G1 phase of the
cell cycle. But the other possibility is we can have lots of Clib2-dependent
kinase activity. CDH1 is phosphorylated and inactive, so Clib2 isn't destroyed,
and that would be kind of the SG2M phase of the cell cycle. So there's this basic
mutual inhibition that can create a bistable switch, and those two stable
steady states are the two basic phases of the cell cycle. So let's see how that
would work out mathematically. Here's a simple representation of the motif. You
have synthesis and degradation of Clib2, you have phosphorylation and
dephosphorylation of CDH1, and CDH1 is degrading Clib2. Now that looks a lot
simpler, and we can just translate that directly into a pair of
nonlinear differential equations for the amount of Clib2 dx dt and for the
activity of CDH1. And on the right-hand side, we have the synthesis of
Clib2, that's this arrow, and we have the degradation of Clib2, that's
this term, and the degradation depends upon the activity of CDH1. Similarly, the
activity of CDH1 depends upon the kinase, which is phosphorylating it, so
that's the negative term because it's inactivating CDH1, and there's a
phosphatase represented by A, which is catalyzing the reverse reaction, taking
the phosphate group off and activating CDH1. Now there are many ways to analyze
a pair of nonlinear ODEs like this. I'm just going to skip right to what a
mathematician would call a one-parameter bifurcation diagram and
what a cell biologist would call a signal response curve. The signal in
this case is the ratio of the activity of the transcription factor S that's
driving the synthesis of Clib2 to the activity of the phosphatase, which is
trying to activate CDH1. If I treat that as a parameter for these differential
equations and plot the steady-state activity of the Clib2-dependent kinase,
that's species X, because of this double negative feedback loop, I get a bistable
system. There's an off state to the control system, there's an on state to
the control system, there are two saddle node bifurcations, and there's a region
of bistability in between. So for certain values of the parameter, this
switch can be either off or on. So that's a basic property of this little
network. Now how do we induce the switches? We can induce a switch from the
G1 phase of the cell cycle into the SG2M phase of the cell cycle by increasing
this parameter, going past the saddle node bifurcation point, and then we'll
automatically turn the switch on. So that's starting the cell cycle, and
then to exit the cell cycle and go back to G1, we have to increase the activity
of the phosphatase, drive the parameter in the other direction, go past the other
saddle node bifurcation, and we can flip the switch off. So it's a bistable
switch, and we want to alternatively flip the switch off and flip the switch on.
Okay, just like the light switch on the wall is a bistable switch, and you come
in, you have to push on the lever, and you flip the lights on, and when the
seminar is over and you leave, you push it in the other direction, and you flip
the switch off. Two stable steady states. So if you're with me this far, then the
next question is, what are the mechanisms that flip the switch on and flip the
switch off? Yeah, Harry. It is true that there's a bistable situation with the
light switch, but I don't have to go all the way down to where one equilibrium
ceases to exist, which is what you show here. If I push hard enough, I can leave
in the middle of that bistable region. So why does that not happen? So I think what
Harry is saying is there are two ways to flip a switch. You can flip a switch by
pushing the parameter past the saddle node, and then it'll flip automatically.
You could get this switch to flip if you could perturb it past the unstable
steady state, and then it would flip, okay? Actually, that's not what you want
to happen in a cell, because that would be a kind of spontaneous flipping of the
switch due to molecular noise, and as a matter of fact, what I want to prove to
you later is that won't happen. Given the level of molecular noise in a yeast cell,
the switch won't flip by accident. You have to push the bifurcation parameter.
Okay, so back to the control network. We have the double negative feedback here
that creates the bistable switch. Now I need some interactions that will flip
the switch on. The switch is flipped on by Cln2 dependent kinase, because
Cln2 dependent kinase can also phosphorylate and inactivate CDH1. So the
Cln2 dependent kinase, the starter kinase, is an inhibitor of CDH1, and when it
pushes down CDH1, then Cln2 dependent kinase can come up, and Cln2
dependent kinase comes down here and inactivates the transcription factor
that's driving the production of Cln2, and Cln2 is a very unstable protein, so
when Cln2 turns off its production, Cln2 goes away. So Cln2 is a friend of
Cln2, but Cln2 is an enemy of Cln2, and the double negative feedback loop is
embedded in a three-component negative feedback loop. Minus, minus, minus. Now, what
are the consequences of that topology? Here's the motif. What are the
consequences of the topology? I want to consider Cln2 as essentially a
bifurcation parameter on the underlying bistable switch. When Cln2 is zero, I can
have two stable steady states with Cln2 dependent kinase low, with Cln2
dependent kinase high, but we start in G1 phase of the cell cycle. We start in
that steady state. In order to induce the switching of the control system, we
have to create some Cln2, and as Cln2 is produced, it drives the system past the
saddle node bifurcation, flips to the on state. Now Cln2 kinase activity is
high. It shuts off the production of Cln2. Cln2 goes away, but now we're
captured in the upper steady state, and now you see that the transition from G1
into S is an irreversible transition. You need a signal to flip the switch,
but once you flip the switch, the signal can go away, and the switch holds in the
upper steady state. That's why the transition is irreversible. Okay, you with
me? So that commits the cell to doing DNA synthesis and doing the beginning of
mitosis all the way up to metaphase when the chromosomes are aligned on the
metaphase plate. Now how are we going to get out of mitosis and back to G1? That's
the job of a different part of the control mechanism. In fact, it's the job
of this phosphatase, CDC14, because CDC14 takes the phosphate group off and
activates CDH1. So CDC14 is a friend of CDH1. Now how is CDC14 linked to Cln2?
I'll let you sort it out, but the end result is that high activity of Cln2
eventually activates the phosphatase CDC14, which activates the very thing
that kills Cln2. So again, we have a bistable switch embedded in a negative
feedback loop, but this negative feedback loop is responsible for the other major
transition of the cell cycle. Now again, a bifurcation diagram, but a different
bifurcation parameter. Now the activity of the phosphatase, and in order to get
out of G2M, we have to do a transient release of the phosphatase CDC14 so that
the cell can get driven past the saddle node bifurcation, flip to the lower state,
and CDC14 has done its job. It can go back into its hole and not bother me
anymore because now I'm back in the G1 state of the cell cycle. You with me?
That's how it works. Questions?
What if the small net feedback is not in the big loop? How does that change?
Well, if it's not embedded in the big loop, then the big loop can't affect it, right?
The function of the negative feedback loops is precisely to flip the switch
and then to go away. Once the switch is flipped, because it's a negative feedback
loop, the signal goes away. That's important. The bistability creates
the possibility of irreversibility, creates the two phases, the possibility
of irreversible switching. The negative feedback loops are just there to
flip the switch one way or the other. Let's put the two pieces together.
I put the bistable switch in the center and the two negative feedback loops on
one side or the other and put the two bifurcation diagrams back to back.
Here we start in G1. Well, let me say the kind of neutral condition is no
Cln2, no starter kinase, and no phosphatase, CDC14. The neutral position
is here in the center, and in the neutral position, we see the bistability.
You can be in either one or the other state, and we start in G1. There's
very low Cln2 dependent kinase activity. Now the cell cycle is just
moving around these hysteresis loops. We go from G1 into S and G2 because you
turn on the starter kinase. Then the starter kinase goes away. You go into
mitosis, and in order to get out of mitosis and to go into anaphase and
telophase and back to G1, you need to activate the phosphatase. That's
our picture, our theoretical picture of the cell cycle with the various phases
of the cell cycle mapped onto the bifurcation diagrams. Now you understand,
I hope, why you always get an alternation. You always do S phase, and then you do M
phase, and then you do S phase, and then you do M phase. It's an irreversible
progression built into the topology of the control network. Now if this is true,
then we should be able to get some direct experimental evidence for the
bistability of the control system in the neutral state. If we were to knock out
both Cln2 and Cdc14, in principle the remaining control system can be
either on or off, and we would like to design an experiment by which we can
control which state it's going to be in. And these experiments were done by
some of our yeast colleagues, yeast molecular geneticists. In the one case,
the geneticists used a different cyclin called Cln3, which they could
control. So now this is an experimental system where the
experimentalist is controlling Cln3, and the prediction is if you just expose
the cell to a small perturbation with Cln3, pushing it in this direction, and
then remove Cln3, it won't be enough to flip the switch. So you'll see Cln3
activity go up, and then you'll come right back to the original steady state
because you haven't gone far enough to flip the switch. But if you perturb the
cells a little bit more with Cln3 to get past the saddle node bifurcation,
then you'll see the switch flip. And that's the experiment that was done by
Fred Cross at Rockefeller University some time ago, 10 years ago, published in
Molecular Biology of the Cell. He did exactly that experiment and proved that
there are two stable steady states of the system, exactly as predicted by the
bifurcation diagram. And in fact, this experiment was done to test the
prediction of the model. In fact, the title of the paper is Testing a
Mathematical Model of the Cell Cycle. And in fact, it's the first time that a
molecular geneticist actually paid attention to what theoreticians were
saying and did the experiment that needed to be done. So we have great
respect and admiration for Fred Cross, a good friend. Similarly, you can test
bistability on the other side by transiently expressing CDH1 activity.
Remember, CDH1 destroys Clb2. It's not a phosphatase. It destroys Clb2. So if
you express CDH1 transiently, you destroy some Clb2. If you don't destroy
enough of it, you'll go out of mitosis and then come back into mitosis. So it'll
be a reversible process. And if you destroy a little bit more, then you go a
little bit further. You go past the separatrix, and then you'll make the flip
into G1, and you'll stay in G1. And that's the experiment that was done
by Frank Ullman at the Cancer Research UK in London and published in 2009 and
got exactly these results. You turn on a little bit of CDH1. It looks like
you've gone out of mitosis, and you turn it off, and the cells go right back into
mitosis. It's a reversible process. And you do it a little bit longer, a little bit
more CDH1, and then they can make the irreversible flip into G1. Okay, I'm not
going to describe the experiments to you. They're quite complicated. Questions?
The mathematical model here, again, has a saddle node, or are you referencing the separatrix? Is this slightly different?
It's the same model, but I should say that the logic here is just a
hand-waving logic. If it really is a bistable switch, it should have this
property. We can't actually calculate exactly how much you're going to have to
do. These are all hand-drawn pictures, but underneath are real
bifurcation calculations. Okay, you with me so far? So I've presented a kind of
picture of a yeast cell as a little control system, just like an engineer
would be used to, with positive and negative feedback loops, with
bistability, with bifurcations. And I tried to show you that the bifurcation
structure of this control system is completely consistent with the
absolutely crucial physiology of the control of the cell cycle in yeast cells.
Okay, and it's the bistability of this control system that accounts
for the irreversibility of the basic transitions in the cell cycle and
ensures, if it's working properly, that you'll always alternate S phase and M
phase, S phase and M phase. Now the next question I want to address is, will this
really work in the small confines of a yeast cell which has a volume of about
40 femtoliters, which is 40 times 10 to the minus 15 liters. That's a pretty
small drop. Okay, and it's not big enough to hold very many molecules, and
therefore, if these molecules are trying to implement this control system, there
are going to be unavoidable, there's going to be unavoidable molecular noise,
which can perturb the system. And the basic question is, will this bistable
switch hold in the face of these fluctuations, or will it flip
prematurely by the mechanism that Harry proposed a few minutes ago, okay, due to
these random fluctuations. In order to address that question, we have to convert
the deterministic model based on ODEs into a realistic stochastic model based
on individual molecules and propensities for reactions to occur. And we have to do
reliable stochastic simulations on that model. So that's what I want to do next,
but before I do it, let me say a little...

