Signal transduction by molecular switches

D. Molenaar
December 2022

This subject is part of the Learning Path Mathematics and Modeling.

VU bachelor course AB_1239: Cell Biology and Histology, December 2022
Version: CBH2022.01
Copyright © 2022 Vrije Universiteit Amsterdam
 signal transduction by molecular switches 2

The problem

Figure 1 is reproduced from Alberts “Essential Biology”, pg 541–542.

The textbook states that signal proteins that are phosphorylated and
dephosphorylated are often used as molecular switches. The story, as

Figure 1: The figure from Alberts,

“Essential biology”, 3 rd ed pg 542.
Important details: the rate of phos-
phorylation is controlled by an input
signal (“Signal in”) that sets the speed
at which the protein kinase phosphory-
lates the signaling protein. The protein
phosphatase is always active. The pic-
ture does not really explain how the
output signal (“Signal out”) is deter-
mined by the phosphorylated form of
the signaling protein.

it is told in Alberts is, however, incomplete. The characteristic prop-

erty of a switch is that it has two states, which we usually call “on”
and “off”. A single signaling protein molecule is, of course, either
phosphorylated or dephosphorylated, and you could propose the
naïve hypothesis that this constitutes the on and off states of a switch.
However, a cell does not contain just one copy, but a reservoir of
several ten-thousands to hundred-thousands of copies of a signaling
protein. This reservoir or pool of signaling proteins can be partially
phosphorylated when the incoming signal has a particular value. It
is unclear whether the cell should decide that the switch is on or off
when 10 %, 25 %, 50 % or 75 % of these protein molecules are phos-
phorylated. What do these intermediate values mean? The textbook
does not explain how a pool of signal proteins can act as a switch.
Hence, the question is whether there are mechanisms that allow
a pool of signaling proteins to act as a switch, for example, by let-
ting all signaling proteins in a pool to be either phosphorylated or
dephosphorylated when the input signal is above or below a certain
threshold value. This is what we are going to investigate.

Different signal transduction characteristics

Before we answer the question of the previous section, we will con-

sider how incoming and outgoing signals from a biological signal
transduction system are related when such a system would show
 signal transduction by molecular switches 3

signal out
100 %
switching behavior. In fig. 1 we see an incoming signal that con-
trols the switch and an output signal coming out of the system. The
output signal is the fraction of phosphorylated or dephosphorylated pro-
tein molecules. For a reservoir of signal protein molecules to display
ideal switching behavior then, we would need a signal transduc-
tion characteristic as is shown in fig. 2: below the threshold of the
incoming signal there is no phosphorylated signal protein, and above
the threshold all molecules are phosphorylated. A system with this signal in
threshold
property would act as a toggle switch, whose area of uncertainty is
Figure 2: Ideal on/off signal transduc-
negligibly small — try to let a toggle switch to “hang” in the middle: tion characteristic. The domain of the
with a good switch you will not succeed. In practice such behavior incoming signal in which switching
takes place (pink) is very small.
is not possible for a biochemical switch. However, something as
shown in fig. 3 should be possible. This is an S-shaped signal in/out
characteristic. This is also called ultrasensitivity.
To conclude: two properties make a signal transduction system

signal out
100 %
behave like a switch. These are 1) a threshold value for the incoming
signal below which the outgoing signal is virtually absent 2) a small
switching domain for the incoming signal, where the outgoing signal
takes on intermediate values. Both properties together cause a steep
or less steep S-shaped, or ultrasensitive input-output characteristic.

Mechanisms to achieve an ultrasensitive characteristic

signal in
threshold
There are different mechanisms to achieve an S-shaped, ultrasensitive Figure 3: On/off signal transduction
signal transduction characteristic. The best known are: characteristic attainable using biochemi-
cal mechanisms. The domain in which
Cooperativity. Cooperativity in a multimeric protein can give rise switching takes places is small, but
certainly larger than for an ideal on/off
to an S-shaped curve. For example, the amount of oxygen bound characteristic.
to hemoglobin shows an S-shaped dependency on the oxygen
concentration.

Decoy phosphorylation. Signal proteins often contain multiple,

seemingly non-essential phosphorylation sites which, due to a
higher affinity, are initially phosphorylated by the kinase. Their
function is possibly to prevent phosphorylation of the functional
phosphorylation site when kinase activity is low. They therefore
act as a decoy for the kinase. As a result of this, the functional
phosphorylation site is only significantly phosphorylated at a high
signal strength, resulting in a threshold for the incoming signal.
The principle is illustrated in fig. 4.

Multisite phosphorylation. This is similar to the previous mecha-

nism and occurs when the signal protein only gives a signal when
multiple phosphorylation sites contain a phosphate. For exam-
ple, if there are three phosphorylation sites on the signal protein,
 signal transduction by molecular switches 4

Figure 4: Generating ultrasensitivity

by decoy phosphorylation. Decoy sites
are gray, the functional site is dark-blue
and phosphorylated sites are indicated
by a red star. Only when the functional
site is phosphorylated is the protein in the
right conformation to pass on the signal,
shown as a red halo. As indicated by
this scheme, the two decoy sites are
increasing kinase activity
phosphorylated at low kinase activity,
in random order, due to their high
affinity for the kinase.
then at low kinase activity and low incoming signal, in general,
only one or two of the three sites will be phosphorylated, and
the signal protein will not be active. Only at high kinase activity,
will the third site also become phosphorylated. This effect causes
a threshold value for the incoming signal, and thereby leads to
ultrasensitivity, as illustrated in fig. 5.

Figure 5: Generating ultrasensitivity by

multisite phosphorylation. Only when
all sites are phosphorylated is the protein
in the right conformation to pass on
the signal. In contrast to the situation
for decoy phosphorylation there is no
large difference in affinity of the sites
for the kinase and hence, no preferred
order in which the three sites are being
phosphorylated.

increasing kinase activity

Zero-order ultrasensitivity. is a term that refers specifically to a

signal characteristic that arises when the protein kinase and /
or protein phosphatase have a high affinity — low K M — for
the signal protein. This mechanism will be investigated in the
computer lab.

Steady state: a balance between phosphorylation and dephosphory-

lation

It is essential to realize that the kinase and phosphatase are active

at the same time, and that the phosphorylation state of the signal
protein is a result of the balance between the activities of these two
enzymes. The cycle in fig. 1 is continuously turning. It is in a dynamic
equilibrium, so not a thermodynamic equilibrium1 . The characteristics 1
What is the definition of thermody-
as in fig. 3 that we investigate here, are a result of that dynamic namic equilibrium? And why is this not
a thermodynamic equilibrium? See also
equilibrium. We call such a dynamic equilibrium a steady state2 . The the questions in the lab.
last three mechanisms for ultra sensitivity in the above list will only 2
Each thermodynamic equilibrium is,
work in a dynamic equilibrium. strictly speaking, a steady state because
the concentrations are constant, but
each steady state is not a thermody-
namic equilibrium. With a “dynamic
equilibrium” we usually, and certainly
in the above case, mean a steady state
that is not a thermodynamic equilib-
rium.
 signal transduction by molecular switches 5

The mechanism of zero-order ultrasensitivity

Zero-order ultrasensitivity was described for the first time in a paper

by Goldbeter, and Koshland [4]. We will treat it here in a slightly sim-
plified version, and we do that mainly by using graphical instead of
mathematical solutions. The conditions for zero-order ultrasensitivity
are:

1. The protein kinase and protein phosphatase exhibit Michaelis-

Menten kinetics with respect to the concentrations of their sub-
strates, the dephosphorylated and phosphorylated protein signal,
respectively.

2. The total amount of signal protein is constant.

3. After a change of the incoming signal the balance between phos-

phorylation and dephosphorylation quickly restores to a steady
state after a change of the incoming signal.

These assumptions lead to a few equations. Let’s call the rate of

the kinase vk , the rate of the phosphatase v p , the concentration of the
phosphorylated signal protein [ M∗ ] and of the dephosphorylated pro-
tein [ M]. The dephosphorylated signal protein M is the substrate of
the kinase, and the phosphorylated protein signal M∗ is the substrate
of the phosphatase. According to the first assumption the following
holds:
Vk · [ M]
kinase: vk = (1)
Kk + [ M ]
Vp · [ M∗ ]
phosphatase: vp = (2)
K p + [ M∗ ]

These are just Michaelis-Menten equations for the kinase and phos-
phatase, where Vk and Vp are the maximal rates — “Vmax -es” — of
the kinase and phosphatase and Kk and K p their respective Michaelis
constants — K M ’s. According to the second assumption:

[ M] + [ M∗ ] = [ MT ] (= constant) (3)

The third assumption says that the system is virtually always in

steady state, which means that the rates of phosphorylation and
dephosphorylation of the signal transduction protein are equal:

vk = v p (4)

To make this a proper signal transduction system we assume that

the Vmax of the kinase (Vk ) depends on an incoming signal S. The
simplest assumption is that the rate of the kinase is proportional
to the incoming signal, or Vk = S · V, where V is a constant. The
 signal transduction by molecular switches 6

equations that we have proposed can be solved to obtain an equation

[ M∗ ]
that expresses the fraction of phosphorylated signal protein Fraction is a mathematical term, here
[ MT ] meaning the ratio of concentrations
as a function of S. The mathematical derivation is shown on pg. 12. of the phosphorylated signal protein
Instead of using these equations, however, you can obtain a good over that of the total amount of signal
protein.
impression of the mechanism of zero-order ultrasensitivity by graph-
ically studying the steady state of this signal transduction system.
The initial steps of this graphical derivation are demonstrated in
fig. 6. We use eqs. 1 to 4, but only graphically. The x-axis represents

rate Figure 6: Graphical estimation of the

kinase fraction of phosphorylated a signaling
protein at steady state. The rate of the
kinase ( ) and phosphatase ( ) are
displayed.
vk = v p
phosphatase

steady-state
← [ M∗ ] or [ M] → [ MT ]

the concentration [ M ] of non-phosphorylated signal protein. The

rate of the kinase increases with this concentration according to the
Michaelis-Menten enzyme model, and the maximum concentration
of [ M] equals [ MT ], the total amount of signal protein. To each value
of [ M] there belongs a value of [ M∗ ], because the total amount of
signaling protein is constant: [ M∗ ] = [ MT ] − [ M ] (eq. 3). Hence, [ M∗ ]
is equal to 0 when [ M] = [ MT ] and increases to the left on the x-axis.
So we can also plot the rate of the phosphatase in this graph. At the
intersection of both lines v p = vk . Drawing a line from that point per-
pendicular to the x-axis gives the steady-state concentration of [ M ] —
and hence of [ M∗ ]. The effect of a low or high input signal is that the
Vmax of the kinase is reduced or increased. This shifts the rate curve
of the kinase down or up, and shifts the intersection point (steady
state) to the right or left in the graph. This is indicated in fig. 7. If you
now plot the steady-state concentration [ M∗ ] as a function of the Vmax
of the kinase — fig. 8 — you obtain a curve that does not look like an
S-curve, which means that the system as it is simulated now is not
ultrasensitive. How can ultrasensitivity be achieved?
 signal transduction by molecular switches 7

rate Figure 7: The effect of increasing or

kinase decreasing the incoming signal is that
the Vmax of the kinase is increased ( )
or decreased ( ). Due to this, the
steady state concentration moves to
the left — lower [ M ], higher [ M∗ ] — or
phosphatase right.

← [ M∗ ] or [ M] → [ MT ]

[ M∗ ]

Condition for zero-order ultrasensitivity

Goldbeter, and Koshland [4] realized that an ultrasensitive signalling
system could be achieved when the affinity of the kinase and phos-
phatase would be very high, in other words when the K M ’s for their
Vmax of kinase
substrates — the non-phosphorylated and phosphorylated signalling
Figure 8: The steady state values of
protein — would be very low. The reason why this leads to ultrasen- M∗ from fig. 7 as a function of the
sitivity is subject of the computer lab. Vmax of the kinase, or equivalently,
as a function of the incoming signal.
Because of this high affinity of the kinase and phosphatase for
How do you conclude that this signal
their substrates, these enzymes operate under almost all conditions in/out characteristic displays no
of the signal transduction system at a rate near their Vmax , i.e. at a ultrasensitivity?

saturating substrate concentration. When the substrate of an enzyme

is saturating — i.e. much higher than the K M , then the rate is almost
independent of the substrate concentration. This is the reason why
this is called zero order ultrasensitivity. Under these conditions the
rate of the enzyme is proportional to the “zero-th order” of the
substrate concentration s, in other words proportional to s0 = 1, that
is — virtually— independent of s!
In the computer lab you will experience that ultrasensitivity is
relatively easy to understand by drawing similar graphs as figs. 7
and 8 but then for kinases and phosphatases with very low K M ’s.

Dynamic simulation

Until this point, we only considered the steady state. Steady states
are important to study, however, if it would take a very long time
before a system reaches the steady state, the steady state would be
quite irrelevant to cellular behavior. Therefore, we are also interested
in the rate at which steady state is reached. So, we have to study the
course of the changes in concentrations of phosphorylated and non-
phosphorylated signal protein in time, after a change in the signal
 signal transduction by molecular switches 8

strength. In the Excel sheet we do this by dividing time in very small

discrete steps, and by calculating the instantaneous changes in the
concentrations of both forms. These instantaneous rates are given by
the rate equations of the kinase and phosphatase. The mathematics behind this type of
simulation concerns the (numerical)
solution of differential equations. We
Literature will demonstrate how to do this when
discussing models of epidemics in the
course Microbiology and Toxicology in
There is a beautiful series of review articles about ultrasensitivity January.
written by James Ferrell, an authority in this field [1, 2, 3]. The lab
of James Ferrell combines experimental and theoretical research to
understand signal transduction pathways, including those involved
in the cell division cycle. He showed that also in this process, ultra-
sensitivity plays an important role. His research is an example of how
modelling of signal transduction leads to completely new insights.
The field of research that uses mathematical models to understand
biological systems is called Systems Biology.
 signal transduction by molecular switches 9

Computer lab

Please note that you have to provide an answer for all questions in the computer lab before you are allowed to sign off
the attendance list. If you do not know the answer, discuss the question with your peers or a lecturer. If you still don’t
know the answer then try to formulate at least what aspect of the question you don’t understand, or why you don’t
understand the question.

Question 1 Recap of Michaelis-Menten kinetics

In this exercise we recapitulate knowledge from the biochemistry course.
a) From that course you know that there is a relation between the K M and the Vmax . Namely: the K M
equals the substrate concentration at which the rate of the enzyme . . . ?
b) Sketch in one panel the graphs displaying the rate v — on the vertical axis — as a function of the sub-
strate concentration S — on the horizontal axis — of two Michaelis-Menten enzymes. One with a low
K M and one with a high K M . Assume that both enzymes have the same Vmax .
The computer lab is carried out using the excel file phosphorylation_switch.xlsx. Open this file now. The
file has two worksheets, steady state and kinetics. The steady state worksheet provides the steady
state solution of the degree of phosphorylation of the signal protein at different strengths of incoming
signal. You can change some parameters at the top left of the worksheet, i.e. V, Kk , K p and [ MT ]. V is a
constant that determines the Vmax for the kinase and for the phosphatase. However, the Vmax of the kinase
Vk can be changed by means of a signal S as well by the expression Vk = S · V. It is also important to
realize what the effect of these parameters is on the dynamics of phosphorylation of the signal protein. For
example, when initially all the protein is in the non-phosphorylated state, and the incoming signal is high,
it takes a while before the system reaches a steady state with a lot of phosphorylated signal protein. The
time course of concentrations [ M] and [ M∗ ] is simulated in the kinetics worksheet. How this time course
depends on the parameters and the initial situation (the amount of M∗ at the start). For the dynamic
simulation, parameters are taken from the steady state worksheet, and you can change the signal S and the
starting amount of [ M ], [M]_start at the top left. [M]_start should, of course, not be higher than [ MT ]!

Question 2 Low affinity kinase and phosphatase

Open the Excel file phosphorylation_switch.xlsx, worksheet steady state with the original parameter
values (V = 1, K = 20, K p = 20 en [ MT ] = 10).
a) With the current parameters, does the enzyme display the typical characteristics of a molecular switch?
Compare the graph in the steady state sheet with figs. 2 and 3. Explain why it does or why it doesn’t.
b) Open the worksheet kinetics. The cells should have the following values: Signal=2 and [M]_start=10.
Study the dynamics in the graph “long term time scale”. How long does it take before the system
appears to have reached steady state?
c) In the same worksheet, derive from the graph “Graphical derivation of the steady state concentration”
what the steady state concentration of the non-phosphorylated protein M will be at this signal strength.
Check whether this is the same as the value from the dynamic simulation.
d) At the bottom of the worksheet, look up the final concentrations of M and M∗ , at approximately 101
seconds. In the sheet steady state look up whether these are in agreement with the calculated steady
state concentrations at a signal strength of 2.
 signal transduction by molecular switches 10

e) There will be small differences in the calculations above between the dynamic and steady state predic-
tions. What is the cause of this difference?

Question 3 High affinity kinase and phosphatase

Open the worksheet steady state.
a) Now reduce the K M ’s of the kinase and phosphatase to 0.05. What happens with the Signal In/Out
characteristic? Again look at the figure in the steady state sheet.
b) What changes in the dynamics of the system — hint: look in the kinetics worksheet: how long does it
take to reach steady state? Can you explain that?
c) In the graphs “long time scale” and “short timescale” you observe that concentrations run in a straight
line to the final concentrations. They bend off only at the last moment. Explain why the concentra-
tion goes in a straight line to the final concentration — hint: it probably helps by you using the graph
“Graphical derivation . . . ” to imagine what happens to the rate of the phosphatase and kinase while the
concentrations of [ M ] and [ M∗ ] change.
d) What do you see changing in the graph “Graphical derivation . . . ”?
e) Reduce the “Signal” to 0.5 — in the yellow cell at the top of the worksheet. Describe changes that
you observe in the graphs “Graphical derivation . . . ” and “long term time scale”. What is the final
concentration of [ M]?
f) Why do the concentrations in the graph “long term time scale” hardly change anymore?
g) Change [M]_start to a number between 0 and 10 so that the concentration of protein [ M ] does signifi-
cantly change over time. Which value of [M]_start is needed to induce a significant change?

Question 4 Reconstruction of the ultrasensitive signal in/out characteristic

We see that the idea proposed by [4] works, and that high affinity — low K M — of the kinase and phos-
phatase for their substrates leads to an ultrasensitive signal in/out characteristic. But do we understand
why this idea works? To obtain a better understanding, we will reconstruct the ultrasensitive signal in/out
characteristic graphically as we did in figs. 7 and 8. We will use the graph “Graphical derivation . . . ” in the
kinetics worksheet for this.
Set the value of “Signal” to 0.5, 0.9, 1, 1.1, and 1.5, study the effect in the graph “Graphical derivation
. . . ”, and plot “Signal” on the x-axis and the corresponding steady state values for [ M∗ ] on the y-axis. Note
that you just need to make a sketch on paper, not a graph with numerically correct values. Now explain in
words how the high affinities cause the ultrasensitive signal in/out characteristic.

Question 5 Energy consumption by the signal transduction system

Set the value of the “Signal” cell back to 2, go to the steady state sheet and reduce V by a factor 5 to
V = 0.2.
a) Why does it take longer now until steady state is reached?
b) Does the signal characteristic change? Why/why not?
c) In the dynamic state ATP is consumed continuously (see fig. 1), also in the steady state! How does the
ATP consumption depend on V?
 signal transduction by molecular switches 11

d) What do you conclude from the above about the relationship between ATP consumption and rate of
signal transduction?
e) Why would a cell sometimes still opt for fast signal transduction?
f) At which signal strength will the consumption of ATP be high, and when will it be low?

Question 6 Steady states are dynamic equilibria

a) Why are the mechanisms that generate ultrasensitivity by “decoy” or “multi-site” phosphorylation
(pg. 3) only possible in a dynamic equilibrium, so when kinase and phosphatase are active simultane-
ously?
b) Explain why the steady states of these signalling pathways are not thermodynamic equilibria. Use fig. 1
in your explanation.

Question 7 Distinguishing decoy and multisite phosphorylation

Decoy and multisite phosphorylation are similar mechanisms to achieve ultrasensitive signal transduction.
Current molecular biology techniques allow you to inactivate specific phosphorylation sites by introducing
mutations in the DNA encoding the signal transduction proteins. How would this technique allow you to
experimentally distinguish the two mechanisms? In other words, how would the response of both signal
transduction by these mechanisms differ if you “knocked out” certain phosphorylation sites?

References

Below you will find the references to the literature. The so-called DOI’s or Digital Object Identifiers are identifi-
cation codes for online versions of the documents. In this syllabus these are printed as active links which, when you
click on them, open the documents in your browser (if you are connected to the internet). In some cases, you will
need access through your VU-account. When you are not logged in on the VU network you can get access by adjust-
ing your browser once. The manual describing how to do this can be found on the website of the university library:
www.ub.vu.nl, using the search term ‘Access outside the campus’.

(1) Ferrell, J. E., Jr, and Ha, S. H. (2014). Ultrasensitivity part I: Michaelian responses and zero-order
ultrasensitivity. Trends in Biochemical Sciences 39, 496–503, DOI: 10.1016/j.tibs.2014.08.003.
(2) Ferrell, J. E., and Ha, S. H. (2014). Ultrasensitivity part III: cascades, bistable switches, and oscillators.
Trends in Biochemical Sciences 39, 612–618, DOI: 10.1016/j.tibs.2014.10.002.
(3) Ferrell, J. E., Jr, and Ha, S. H. (2014). Ultrasensitivity part II: multisite phosphorylation, stoichiometric
inhibitors, and positive feedback. Trends in Biochemical Sciences 39, 556–569, DOI: 10.1016/j.tibs.
2014.09.003.

(4) Goldbeter, A., and Koshland, D., Jr (1981). An amplified sensitivity arising from covalent modification
in biological systems. Proc. Natl. Acad. Sci. U. S. A. 78, 6840–6844, DOI: 10.1073/pnas.78.11.6840.
 signal transduction by molecular switches 12

Appendix
Mathematical derivation of the “zero order ultrasensitive” model

This is for enthusiasts: the calculations are a bit tough but there is no complicated math involved. At the
end we have to solve a quadratic equation using the ABC formula. For the convenience we first define a
few new variables. We are interested in the fractions of phosphorylated and non-phosphorylated signaling
[ M] [ M∗ ]
protein, so [ M ] and [ M ] . We therefore define
T T

[ M] [ M∗ ]
m≡ and m∗ ≡
[ MT ] [ MT ]
And we define “relative Michaelis constants” (the reason will become clear below)
Kp Kk
kp ≡ and kk ≡
[ MT ] [ MT ]
Dividing both sides of eq. 3 by [ MT ] we obtain
[ M] [ M∗ ]
+ =1 or
[ MT ] [ MT ]
m + m∗ = 1 or
m = 1 − m∗

To write eqs. 1 and 2 in terms of m and m∗ we divide numerator and denominator of these equations by
[ MT ]. We obtain
[ M]
Vk · [ MT ] Vk · m
vk = =
Kk
[ MT ]
+ [[MM]] kk + m
T
[ M∗ ]
Vp · [ MT ] Vp · m∗
vp = =
Kp M∗ ]
+ [[M k p + m∗
[ MT ] T]

Substitution of these equations in eq. 4 yields

Vk · m Vp · m∗
=
kk + m k p + m∗
Substitution of m = 1 − m∗ yields
Vk · (1 − m∗ ) Vp · m∗
=
k k + 1 − m∗ k p + m∗
From this m∗ can be solved. We multiply both sides of the equation by the denominators k k + 1 − m∗ and
k p + m∗ , and divide both sides by Vp . To calculate the dependency on the signal S that controls the kinase
we also use the definition Vk = S · V, in which, for convenience, we set V = Vp . This yields

S · (1 − m∗ )(k p + m∗ ) = m∗ · (k k + 1 − m∗ )
Sk p + Sm∗ − Sm∗ k p − Sm∗2 = m∗ k k + m∗ − m∗2
m∗2 (1 − S) + m∗ S − Sk p − 1 − k k + Sk p = 0


 signal transduction by molecular switches 13

The last equation is a quadratic equation in m∗ , which can be solved for m∗ using the ABC formula:
√
∗ − B ± B2 − 4AC
m = where (5)
2A
A = 1−S
B = S (1 − k p ) − 1 − k k
C = Sk p

Equation 5 is, with a little variation, the formula used in the excel file to calculate the steady state m∗ .