TEACHING RESOURCE

C O M P U TAT I O N A L B I O L O G Y forward and backward rates will be equal at

a single point, and the system will be mono-
Bistability in Biochemical stable (Slide 10). Two system properties that
can lead to bistability are (i) “ultrasensitive”
Signaling Models positive feedback and (ii) a back reaction
that saturates with increasing amounts of
Eric A. Sobie* product (Slides 16 to 19) (3)
Bifurcation Diagrams to Illustrate
This Teaching Resource provides lecture notes, slides, and a student assign- Bistable Regimes
ment for a two-part lecture on the principles underlying bistability in biochemi- Systems that have the capacity for bistabil-
cal signaling networks, which are illustrated with examples from the literature.
ity frequently only exhibit this behavior un-
The lectures cover analog, or graded, versus digital, all-or-none, responses in
der particular conditions. For instance, if an
cells, with examples from different types of biological processes requiring each.
Rate-balance plots are introduced as a method for determining whether generic
external stimulus activates a participating
one-variable systems exhibit one or several stable steady states. Bifurcation dia- enzyme, it is possible that bistability will
grams are presented as a more general method for detecting the presence of only be present for intermediate values of
bistability in biochemical signaling networks. The examples include an artificial stimulus and that high or low levels of stim-
toggle switch, the lac operon in bacteria, and the mitogen-activated protein ki- ulus will produce monostability. In such a
nase cascade in both Xenopus oocytes and mammalian cells. The second part case, the value of stimulus where the sys-

Downloaded from http://stke.sciencemag.org/ on December 15, 2017

of the lecture links the concepts of bistability more closely to the mathematical tem switches from monostable to bistable
tools provided by dynamical systems analysis. The examples from the first part represents a bifurcation, and important sys-
of the lecture are analyzed with phase-plane techniques and bifurcation analysis, tem properties can frequently be inferred
using the scientific programming language MATLAB. Using these programs as a by plotting stimulus on the abscissa and a
template, the assignment requires the students to implement a model from the relevant dependent variable on the ordinate.
literature and analyze the stability of this model’s steady states. Examples of such bifurcation diagrams
from several generic reaction schemes are
Lecture Notes a commitment is made, small fluctuations presented in Slides 20 to 23.
Analog Versus Digital Responses in the initiating stimulus are insufficient to Phase-Plane Analysis of Steady States of
Biological processes can be graded or digi- cause a change in the output. In some cases, Biological Systems
tal (Slides 2 to 5). Most biological process- cell decisions are irreversible. The second method of analysis that can
es taught to undergraduates and beginning Rate-Balance Plots for Quantitative determine the presence of bistability is the
graduate students are analog and graded Analysis of Bistability plotting of trajectories in the phase plane
in nature. For instance, the following rep- Digital, yes-or-no cell decisions generally (Slides 24 and 25). Given a set of differen-
resent physiological examples of analog result from biochemical signaling networks tial equations defining the evolution of the
responses: (i) the increase in cardiac output that exhibit bistability. This means that ei- state variables of the system, a “nullcline”
during exercise, (ii) the secretion of insulin ther a low activity (“No”) state or a high is defined as the set of points at which the
by pancreatic beta cells in response to el- activity (“Yes”) state are possible, but inter- derivative of a variable is zero. For a system
evated blood glucose, and (iii) the increased mediate levels of activity are observed only with two variables, nullclines can be plotted
concentration of a reaction’s product when rarely (Slide 3). At the level of biochemical as a function of the two variables to illus-
the concentration of the catalyzing enzyme reactions, bistability typically results from trate important characteristics of the system
increases. In these cases, the output is a “mutual activation” or “mutual repression” (Slides 24 and 25). For instance, each time a
smoothly varying function of the input, and feedback loops (Slide 6) (1, 2). However, nullcline is crossed, the direction the system
the output returns back to the original level these arrangements do not by themselves travels in the phase plane changes (Slide
when the input is removed (albeit some- guarantee that a particular network will be 24). Examples illustrate that when the two
times with a delay) (Slide 4). bistable. A particular signaling network nullclines intersect at only a single point,
For some other biological processes, will frequently be monostable (one steady the system must have only one steady state.
however, such an analog response is inade- state) under some conditions and bistable This steady state may be stable, indicating
quate, and the cell (or organism) must make under others (Slides 7 and 8). Quantitative a system that always returns back to this
a digital, yes-or-no decision. Examples in- analyses are required to determine whether point, or it can be unstable, in which case
clude fertilization, differentiation, cell divi- a particular signaling network exhibits bi- the system oscillates continuously. How-
sion, apoptosis, and immune cell activation stability. One method for determining bista- ever, if the nullclines intersect three times,
(Slide 5). In these situations, the cell’s de- bility is the rate-balance plot (Slides 9 and the system can be bistable. These prin-
cision is difficult to reverse, such that once 10). The work of Ferrell and Xiong (3) is ciples are illustrated with proven bistable
used to illustrate how, when considering a biological signaling systems, including the
Department of Pharmacology and Systems
single variable, rate-balance plots provide artificial toggle switch built by Collins and
Therapeutics and Systems Biology Center New an intuitive, yet rigorous, means to estab- co-workers (Slide 26) (4); the mitogen-ac-
York, Mount Sinai School of Medicine, New lish the stability characteristics of a network tivated protein kinase (MAPK) cascade in
York, NY 10029, USA (Slides 10 to 19). If both the forward rate Xenopus oocytes (Slide 27) (5, 6); and the
*Corresponding author. E-mail, eric.sobie@ and the backward rate of a reaction depend MAPK cascade in mammalian cells (Slide
mssm.edu linearly on the quantity of substrate, then 28) (7, 8).
www.SCIENCESIGNALING.org 27 September 2011 Vol 4 Issue 192 tr10 1
TEACHING RESOURCE

Analysis and Simulation of Biological program that addresses the assignment. In ues of [A*], so this plot will be most clear if
Systems with MATLAB Part 2, the students implement a two-vari- you plot each steady-state value with a dis-
The second part of this lecture reinforces able model of the E. coli lac operon. This tinct symbol and do not connect the points.
the quantitative concepts introduced previ- system exhibits either one or two stable Lines 57 to 65 in the program accommodate
ously by presenting analysis and simulation steady states, depending on the value of ex- the situation in which one or multiple steady
scripts written in MATLAB, then illustrat- tracellular lactose. In addition to perform- states are possible.
ing the results obtained with these scripts. ing dynamic simulations with this model, 3. From your analysis in problem 2, how
This lecture begins with a brief review of the students perform a phase-plane analysis large does Kmb have to become before bi-
certain key points from the previous lecture to determine the stability characteristics. stability is no longer observed?
(Slides 29 to 35). Students are provided The second assignment is the more difficult 4. From your analysis in problem 2, why
with MATLAB scripts that (i) generate of the two. do small values of Kmb allow for bistability,
rate-balance plots and (ii) simulate the tem- The Supplementary Materials include whereas large values do not?
poral evolution of a generic two-variable two MATLAB programs: “ratebalance.m,” Student Assignment: Part 2
system that can potentially exhibit bistabil- which generates rate-balance plots for cases To complete this assignment, you must im-
ity. These scripts are used by the students with (i) no feedback or (ii) ultrasensitive plement a simple two-variable model of the
as templates for completing the Problem feedback, and “repression.m,” which con- E. coli lac operon (Slides 47 and 48) and
Set (see below). Explaining these scripts sists of an ordinary differential equation use nullclines to evaluate bistability (Slides
and allowing the students to run simulations (ODE) model of a generic two-variable sys- 25 and 38 to 42). The differential equations

Downloaded from http://stke.sciencemag.org/ on December 15, 2017

with them helps to reinforce the pertinent tem with mutual repression. governing this system are as follows:
concepts. For instance, the dynamical simu- Student Assignment: Part 1
dl
lations show that when a system is bistable, In a system with a single variable, rate- = βlext LacY –γl
the initial conditions determine whether the balance plots can be used to determine dt
“high” steady state or the “low” steady state whether bistability is present (see Slides 10
is eventually attained. Initial conditions to 20 for examples). The MATLAB script
dLacY l4
close to the unstable steady state can lead “ratebalance.m” plots these curves for two = δ + p 4 4 – σ LacY
to either outcome, and the time required simple cases: (i) a constant forward rate and dt l +l0
to reach steady state is typically quite long (ii) ultrasensitive autocatalytic feedback
with such initial conditions (Slides 36 to that causes the forward rate to be nonlin- For this question, assume that the following
38). Finally, a simple graphical method for ear. The variable of interest A* (Astar in the parameter values are fixed:
determining the stability characteristics of script) is assumed to be normalized to the
steady states in phase-plane plots is de- quantity [A]TOTAL, so this ranges from 0 to 1. β = 1, γ = 1, δ = 0.2,
scribed and applied to a two-variable ex- For several values of the stimulus S, the pro- l0 = 4, p = 4, σ = 1.
ample system (Slides 39 to 42). The result gram plots forward and backward rates, as
obtained with this method is shown to be well as the steady states where the forward The variable lext, representing extracel-
consistent with the answer obtained through and backward rates cross. lular lactose, is the variable that can vary
the classical method from dynamical sys- Bistability can be produced with ultra- and potentially cause the bacteria to change
tems, computing the roots of the character- sensitive feedback, as in the example shown states.
istic equation (Slides 43 and 44). in Slide 17. It is not possible with linear 1. Plot the nullclines of this model for lext
Conclusion feedback and a first-order reverse rate. = 2.5. Because of the l4 term appearing in the
The lecture ends with a review of some of However, bistability can be produced if lin- second equation, it is easiest to treat LacY
the biological systems that can exhibit bi- ear autocatalytic feedback is combined with as the dependent variable (ordinate) and l
stability (5, 7) (Slides 45 to 48), including a “back reaction” that becomes partially as the independent variable (abscissa). Your
the now-classic example of the lac operon saturated when the amount of product (A* nullclines should intersect in three places.
of Escherichia coli (Slides 47 to 48) (9, 10, in this case) increases (see Slide 19). 2. From examining the equations in
11). The latter system is the subject of the 1. Modify “ratebalance.m” to simulate problem 1, above the LacY nullcline (that
homework assignment, which is discussed the condition of linear autocatalytic feed- is, increased LacY), is LacY increasing or
in Slide 49. back with a partially saturated back reac- decreasing?
tion. You can assume that the stimulus [S] 3. From examining the equations in
Problem Set is zero. Start with the following constants: problem 1, above the l nullcline, is l in-
Introductory Details kminus = 5; Kmb = 0.1. creasing or decreasing? Draw arrows on
The homework assignment requires the 2. Assuming S = 0, test several different the LacY and l nullclines to illustrate these
students to apply the methods of analysis values of Kmb and determine which values results. From these arrows, determine in
taught in the lectures. In Part 1, students of Kmb produce bistability and which do not. qualitative terms the stability of the three
generate rate-balance plots for a system You will have to determine the “interesting” steady states.
with linear positive feedback but a back range of Kmb values to examine. Plot your 4. Modify the provided program
reaction that can become saturated. The as- results as a bifurcation diagram: Kmb on “repression.m” so that the new model now
signment includes a MATLAB script that the abscissa, steady-state values of [A*] on simulates the temporal evolution of l and
generates rate-balance plots for a simpler the ordinate. For the bistable values of Kmb, LacY. To reach steady state, a total simula-
system, and the students convert this into a there will be three possible steady-state val- tion time of 20 s is reasonable, and the solu-
www.SCIENCESIGNALING.org 27 September 2011 Vol 4 Issue 192 tr10 2
TEACHING RESOURCE

tion converges with a time step of 0.01 if  iscipline: Bacteriology, biochemistry,

D ers, toggles and blinkers: Dynamics of regulatory
and signaling pathways in the cell. Curr. Opin.
Euler’s method is used. bioengineering, biophysics, cell biology, Cell Biol. 15, 221–231 (2003).
5. With the modified program, for lext = developmental biology, education, enzy- 3. J. E. Ferrell, W. Xiong, Bistability in cell signaling:
2.5, plot the time evolution of LacY and l mology, reproductive biology, theoreti- How to make continuous processes discontinu-
for the following initial conditions: (i) l = 8, cal biology ous, and reversible processes irreversible. Chaos
11, 227–236 (2001).
LacY = 3; (ii) l = 3.2, LacY = 1.3; (iii) l = 3, Keywords: Dynamical systems, mathe-
4. T. S. Gardner, C. R. Cantor, J. J. Collins, Construc-
LacY = 1.2; (iv) l = 2, LacY = 1. Explain the matical modeling, differential equations, tion of a genetic toggle switch in Escherichia coli.
different results obtained. protein kinase signaling, fertilization, Nature 403, 339–342 (2000).
6. With the modified program, vary lext lac operon, rate-balance plot, MATLAB, 5. J. E. Ferrell Jr., E. M. Machleder, The biochemical
basis of an all-or-none cell fate switch in Xenopus
over the range 1 to 7 and compute the tem- nullcline, simulation oocytes. Science 280, 895–898 (1998).
poral evolution of l and LacY. For each sim- 6. W. Xiong, J. E. Ferrell Jr., A positive-feedback-
ulation, store the final value of LacY. Run Technical Details based bistable ‘memory module’ that governs a
cell fate decision. Nature 426, 460–465 (2003).
two sets of simulations: one in which the Format: PowerPoint (.ppt) 7. U. S. Bhalla, R. Iyengar, Emergent properties
initial conditions are l = 8, LacY = 3, and Size: 3.86 MB of networks of biological signaling pathways.
a second in which the initial conditions are Requirements: Microsoft PowerPoint Science 283, 381–387 (1999).
l = 2, LacY = 1. Generate a bifurcation dia- Format: MATLAB (.m) 8. U. S. Bhalla, P. T. Ram, R. Iyengar, MAP kinase
phosphatase as a locus of flexibility in a mito-
gram of how the final value of LacY varies Requirements: Mathworks MATLAB gen-activated protein kinase signaling network.
as a function of lext. Plot the results of the Format: PDF (.pdf) Science 297, 1018–1023 (2002).

Downloaded from http://stke.sciencemag.org/ on December 15, 2017

two sets of simulations in different colors. Requirements: Adobe PDF Reader 9. A. Novick, M. Weiner, Enzyme induction as an
all-or-none phenomenon. Proc. Natl. Acad. Sci.
7. From the analysis in problem 6, how U.S.A. 43, 553–566 (1957).
small must lext become before bistability is Supplementary Materials 10. W. K. Smits, O. P. Kuipers, J. W. Veening, Phe-
no longer observed? http://stke.sciencemag.org/cgi/content/full/ notypic variation in bacteria: The role of feed-
back regulation. Nat. Rev. Microbiol. 4, 259–271
8. From the analysis in problem 6, how sigtrans;4/192/tr10/DC1
(2006).
large must lext become before bistability is Slides. Bistability in biochemical signal- 11. E. M. Ozbudak, M. Thattai, H. N. Lim, B. I.
no longer observed? ing models. Shraiman, A. Van Oudenaarden, Multistability in
9. From the analysis in problem 6, ex- Problem set MATLAB code. Two MAT- the lactose utilization network of Escherichia coli.
Nature 427, 737–740 (2004).
plain in biological terms why bistability is LAB programs for computational analysis 12. Acknowledgments: The author thanks Y. S.
only observed at intermediate values of lext. of bistability. Lee for testing the problem set. Funding: The
Answer key. MATLAB code and PDF of development of this course was supported by a
Systems Biology Center grant (PS0GM071558)
Educational Details the answers are available upon request and a training grant in Pharmacological Sciences
Learning Resource Type: Lecture notes, (T32GM062754).
assignment, presentation References
10.1126/scisignal.2001964
Context: Graduate 1. J. E. Ferrell Jr., Self-perpetuating states in signal
Intended Users: Teacher, learner transduction: Positive feedback, double-negative
feedback and bistability. Curr. Opin. Cell Biol. 14,
Intended Educational Use: Learn, plan, 140–148 (2002). Citation: E. A. Sobie, Bistability in biochemical
teach 2. J. J. Tyson, K. C. Chen, B. Novak, Sniffers, buzz- signaling models. Sci. Signal. 4, tr10 (2011).

www.SCIENCESIGNALING.org 27 September 2011 Vol 4 Issue 192 tr10 3

Bistability in Biochemical Signaling Models
Eric A. Sobie

Sci. Signal. 4 (192), tr10.

DOI: 10.1126/scisignal.2001964originally published online September 20, 2011

Downloaded from http://stke.sciencemag.org/ on December 15, 2017

ARTICLE TOOLS http://stke.sciencemag.org/content/4/192/tr10

SUPPLEMENTARY http://stke.sciencemag.org/content/suppl/2011/09/23/4.192.tr10.DC1
MATERIALS

RELATED http://stke.sciencemag.org/content/sigtrans/4/190/tr2.full
CONTENT
http://stke.sciencemag.org/content/sigtrans/6/271/tr2.full

REFERENCES This article cites 11 articles, 4 of which you can access for free
http://stke.sciencemag.org/content/4/192/tr10#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science Signaling (ISSN 1937-9145) is published by the American Association for the Advancement of Science, 1200 New
York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive licensee American
Association for the Advancement of Science. No claim to original U.S. Government Works. The title Science Signaling is a
registered trademark of AAAS.