Content uploaded by Valerie M Weaver
Author content
All content in this area was uploaded by Valerie M Weaver on Sep 09, 2014
Content may be subject to copyright.
University of Pennsylvania
ScholarlyCommons
Departmental Papers (BE) Department of Bioengineering
1-1-2008
Integrin-mediated signalling through the MAP-
kinase pathway
Ka Lai Yee
University of Pennsylvania
V. M. Weaver
University of California
Daniel A. Hammer
University of Pennsylvania, hammer@seas.upenn.edu
Copyright 2008 IEEE. Reprinted from IET Systems Biology, Volume 2, Issue 1, January 2008, pages 8-15.
This material is posted here with permission of the IEEE. Such permission of the IEEE does not in any way imply IEEE endorsement of any of the
University of Pennsylvania's products or services. Internal or personal use of this material is permitted. However, permission to reprint/republish this
material for advertising or promotional purposes or for creating new collective works for resale or redistribution must be obtained from the IEEE by
writing to pubs-permissions@ieee.org. By choosing to view this document, you agree to all provisions of the copyright laws protecting it.
This paper is posted at ScholarlyCommons. http://repository.upenn.edu/be_papers/104
For more information, please contact repository@pobox.upenn.edu.
Integrin-mediated signalling through the MAP-kinase
pathway
K.L. Yee, V.M. Weaver and D.A. Hammer
Abstract: The mitogen activated protein (MAP) kinase cascade, leading to extracellular-regulated
kinase (ERK) activation, is a key regulator of cell growth and proliferation. The effects of ERK are
mediated by differences in ERK signalling dynamics, including magnitude and duration. In vivo,
ERK signalling is stimulated by both growth factors and adhesion signals. A model for adhesion-
mediated ERK activation is presented. Outputs of the model such as ERK and FAK activation, as
well as responses to different ligand densities, are compared with published experimental data. The
model then serves as a basis for understanding how adhesion may contribute to ERK signalling
through changes in the dynamics of focal adhesion kinase activation. The main parameters influen-
cing ERK are determined through screening analyses and parameter variation. With these par-
ameters, key points in the pathway that give rise to changes in downstream sign alling dynamics
are identified. In particular, oncogenic Raf and Ras promote cell growth by increasing the magni-
tude and duration, respectively, of ERK activity.
1 Introduction
The MAP kinase cascade is an important signalling pathway
associated with tumorigenic behaviour. In cancerous cells,
the pathways that regulate proliferation are disrupted so
that cells experience unchecked growth. A major factor in
proliferation is the increased activity of the MAPK
cascade, which can be caused by increased external
stimuli or mutations along the pathway that promote
pathway activation [1]. The MAPK involved in growth
regulation is extracellular-regulated kinase (ERK). Upon
double phosphorylation by mitogen activated protein
kinase (MEK), double phosphorylated extracellular-regu-
lated kinase (ERKPP) translocates to the nucleus and
induces gene transcription by phosphorylation of transcrip-
tion factors [2]. One of ERK’s targets is fos, a member of
the AP-1 transcription factor heterodimer [3]. AP-1 family
transcription factors induce the transcription of cyclin D1,
which promotes cell-cycle entry [4, 5]. Modelling this
pathway to gain an understanding of how specific elements
give rise to different ERK signalling dynamics is an import-
ant step towards regulating signal transduction, and
ultimately cell behaviour, in vivo.
The signalling pathway that culminates in ERK acti-
vation is initiated by extracellular stimuli such as adhesion
to the extracellular matrix or growth factors. Cell adhesion
activates ERK by binding of a5b1 integrins at the cell
surface to extracellular matrix proteins such as fibronectin
[6, 7]. Integrins activate ERK via a pathway involving
Shc, Grb2, Sos, the GTPase Ras and focal adhesion
kinase (FAK) [5, 6, 8]. Epidermal growth factor (EGF)
binding to its receptor is also known to activate ERK
through the same Shc–Grb2–Sos– Ras pathway as integ-
rins [9 –11]. Although integrins may activate ERK in the
absence of growth factors, cells in vivo are exposed to
both adhesion signals and growth factor signals. In addition,
there is a cross-talk between growth factors and
integrin-activated pathways to ERK activation at different
levels of the pathway [12–14]. This cross-talk likely
gives rise to distinctive ERK activation signals that can
regulate different behaviours [12]. For example, ERK
activity stimulated by growth factors and adhesion indepen-
dently is transient, whereas ERK activity in adherent,
growth factor stimulated cells is sustained. Moreover, the
cell-cycle protein cyclin D1 is only induced by sustained
ERK signals [5, 15] .
The MAPK cascade and activation by EGF have been
previously explored with deterministic, computational
models. A basic model of the Shc – Grb2 –Sos–Ras
pathway activated by EGF was described by Kholodenko
et al. [16]. A more comprehensive model of ERK activation
in response to EGF stimulation has also been examined
[17]. In that model, EGF receptor – ligand binding and
dimerisation initiate the traditional Shc–Grb2–Sos–Ras
pathway that leads to the MAPK cascade. More recently,
Chapman and Asthagiri [18] undertook an exploration of
MAPK cascade dynamics. Their work demonstrated that
adjusting rate constants and protein or phosphatase concen-
trations controls ERK activation dynamics. Similarly,
Hornberg et al. [19] applied control analysis to the model
developed by Schoeberl, identifying specific proteins that
are important in EGF-mediated ERK signalling.
As a first step towards developing a model of ERK acti-
vation by both EGF and integrins, a computational model
for integrin activation of ERK is presented in this paper.
In this model, the ERK signalling pathway is initiated by
a5b1 integrin binding to its extracellular ligand fibronectin.
The model successfully simulates experimental FAK and
ERK activation time courses. Semi-quantitative data on
FAK and ERK activation after adhesion [20] are used to
# The Institution of Engineering and Technology 2008
doi:10.1049/iet-syb:20060058
Paper first received 4th August 2006 and in revised form 16th April 2007
D.A. Hammer and K.L. Yee are with the Department of Bioengineering,
University of Pennsylvania, 240 Skirkanich Hall, 210 S 33rd St, Philadelphia
PA 19104, USA
V.M. Weaver is with the Department of Surgery, University of California San
Francisco (UCSF), 513 Parnassus Ave, Room S-321, San Francisco, CA
94143-0470, USA
E-mail: hammer@seas.upenn.edu
IET Syst. Biol., 2008, 2, (1), pp. 8–15
8
verify the results of the model. After comparing simulated
signals with experimental results, the pathways were
explored to gain insight into specific characteristics that
contribute to ERK activation dynamics. To systematically
explore the se t of parameters used in the model, a screening
method developed by Morris [21] was used to identify the
key parameters that affect different characteristics of the
output signal, such as overall dynamics, the maximum
value and time to reach the maximum.
2 Methodology
2.1 Biological considerations
The signalling events culminating in ERK activation are
modelled as a set of differential equations. The protein–
protein interactions included in the pathway were chosen
based on published pathways and experi mental data and
are illustrated in Fig. 1a. The full pathway and a list of reac-
tions and rate constants are provided in the Supplementary
Materials.
The integrin/FAK activation portion of the model deals
with receptor–ligand binding as a cell spreads on fibronectin.
The cell–surface contact area begins at zero and increases
according to an experimentally determined profile [22] (see
Supplementary Materials). The increase in contact area regu-
lates the transport of receptors and ligands into the contact
area where binding occurs. Integrin–fibronectin binding is
modelled according to the Bell model for receptor–ligand
binding [23]. Because integrin clustering is a complex
process involving the many cytoskeletal proteins and the
transduction for force, clustering is modelled as dimerisation
of receptor–ligand pairs (see Supplementary Materials). We
assume that only dimerised receptor–ligand pairs are able to
recruit FAK, which becomes autophosphorylated at tyrosine
397 (FAKP) [24].
Src activation occurs after cell adhesion, although it is
unclear how adhesion signals regulate the Src activatio n
machinery [25]. Activation begins with the dephosphoryla-
tion of cytoplasmic Src at tyrosine 527. The protein then
undergoes a conformational change and translocates to the
membrane [26]. To be fully active, Src undergoes an inter-
molecular autophosphorylation at Y416. However, fully
active Src can be re-phosphorylated and dephosphorylated,
thereby cycling it from active to inactive conformations
[27–29]. This module is linked to the rest of the pathway
by the ability of Src to phosphorylate its substrate FAK at
Y925 [30]. Both Src and FAK are capable of phosphorylat-
ing Shc [9, 11, 31]. This interaction leads to the same
sequence of signalling events the activated by the EGF
receptor, ultimately resulting in the activation of the
MAPK cascade [32, 33].
Evidence suggests that integrins may also activate ERK
through a FAK-independent pathway [9]. Such a pathway
likely involves Fyn, a member of the Src kinase family,
and the alpha integrin subunit [34]. Because of limited
experimental data on this pathway, this model focuses on
the FAK-dependent pathways to ERK activation.
2.2 Model parameters
Rate constants and protein concentrations were either found
in literature or estimated based on known constants for
similar reactions. For protein interactions without published
binding constants, K
d
, k
f
, and k
r
were estimated from avail-
able peptide binding data. For example, binding constants
for Src and FAKP in the Src activation module were esti-
mated with rate constants for Src SH2 domain binding to
the pYEEI peptide, which is similar to the autophosphoryla-
tion site for FAK [35, 36]. Most binding events in this
pathway involve SH2 – phosphotyrosine and SH3–PXXP
binding [37, 38].
During integrin activation of FAK, k
1
is the rate of dimer-
isation of receptor – ligand pairs, whereas k
p
and k
m
are the
forward and backward rate constants for FAK binding to
receptor–ligand dimers and subsequent autophosphoryla-
tion. k
1
is a phenomenological rate constant assigned a
value of 0.01 s
21
, which corresponds to maximum levels
of FAK activation. Decreasing k
1
will decrease the final
magnitude of FAK activation without significantly chan-
ging the kinetics. Values for k
p
and k
m
were estimated to
obtain a variety of time scales for FAK autophosphoryla-
tion, which are discussed in Section 3. The Src activation
module also contains the phenomenological rate constant
k
t
, which represents the rate of Src transport from the cyto-
plasm to the membrane by the cytos keleton [39]. Because
these steps are not easily modelled, they are lumped into
the transport rate k
t
; k
t
is set at 0.1 s
21
, at which the rate
of cytoskeletal transport is maximal. Rate constants for
the phosphatase and kin ase reactions in Src activation
were obtained from literature [40– 42]. Subsequent analyses
of FAK and downstream signalling focus on the relative
effects of different parameter values on kinetics and magni-
tude, so that the conclusions drawn are valid regardless of
the specific phenomenological constants used.
Fig. 1 Protein-Protein Interactions
a Signalling pathways for ERK activation by adhesion
b Ras cycle of GDP/GTP binding
IET Syst. Biol., Vol. 2, No. 1, January 2008 9
Initial protein concentrations were obtained from
Schoeberl et al. [17] and were varied over several orders
of magnitude to explore their effects on signal dynamics.
The constant concentrations of phosphatases were also esti-
mated based on these initial values. The initial conditions
specify protein concentrations for FAK, Src, Shc, Grb2,
Sos, MEK and ERK. All other protein species present in
the system at any given time are derived from these proteins
and have an initial concentration of zero.
The resulting set of equations with initial condition s was
solved in MATLAB 6.5 using the ODE15s solver.
3 Results
3.1 Kinetics of FAK and ERK activation
Asthagiri et al. [20] have shown that activation of ERK by
adhesion occurs before maximal FAK activation.
Simulations of ERK and FAK activation showed similar
kinetics (Fig. 2), which was also robust over large ranges
of parameter values (see Section 3.4.2). Over a range of
FAK recruitment rates ( k
p
), the ERK activation peak con-
sistently occurs ahead of full FAK activation. Since FAK
is the only means by which ERK is activated in this
model, early and therefore relatively low levels of FAK
activation are sufficient for stimulating the MAPK
cascade. For FAK to achieve sufficient levels of activation
to stimulate the MAPK cascade at early times, it was
found that the total level of FAK must be higher than that
of downstream effectors such as Sos, the activator of the
MAPK cascade. For example, a peak magnitude of
1.4 nM ERKPP at t ¼ 2800 s requires the amount of total
FAK to be nearly 22 times the amount of Sos. As a result,
it is assumed that the amount of FAK is in excess of other
signalling proteins in the system. Lastly, sustained FAK
activation does not necessarily result in sustained ERK acti-
vation, implying that FAK serves as an initial stimulus for
ERK activation, but the duration of ERK activation is also
regulated by downstream factors that will be discussed.
3.2 FAK and ERK activation in response to ligand
density
Fibronectin density is one of the most readily controllable
experimental parameters in the system and has been
shown to affect both the magnitude and kinetics of ERK
activation [20]. As ligand density increases, the magnitude
of the ERKPP peak also increases, whereas the peak
occurs faster. The effects of ligand density were success-
fully replicated in silico for similar fibronectin densities,
assuming a high receptor density of 7.89 10
10
mol-
ecules/cm
2
( 1.5 10
6
/cell) (Fig. 3a). Not surprisingly,
changes in ligand density had substantial effects on
ERKPP dynamics only when ligand density is non-
saturating (R
t
. L
t
), thus requiring a high assumed receptor
density. Furthermore, the linear correlation between the
ligand density and initial rates of FAK and ERK activation
observed by Asthagiri et al. [20] was also seen in simu-
lations (Fig. 3b). At saturating ligand densities, the linear
correlation fails.
3.3 Exploration of signalling dynamics
3.3.1 Relationship between FAK and ERK activation:
The effects of ligand density on ERK raise the question of
how the signal is transduced downstream. Since FAK is a
major mediator between adhesion-based signals and down-
stream signals, it is reasonable to examine the effects of
Fig. 2 FAK and ERK activation over a range of k
p
, assuming
that K
d
¼ k
m
/k
p
¼ 1066 is constant
Experimentally observed time courses (Asthagiri) are also plotted
Fig. 3 Effects of ligand density
a ERK activation in response to changes in fibronectin density
b Initial rates of FAK and ERK activation with ligand density
Initial rates were calculated from 470 to 480 s, which falls within the
linear phase of activation
Onset of the linear phase is subject to a time lag due to the clustering
process
IET Syst. Biol., Vol. 2, No. 1, January 200810
ligand density on FAK activation to understand how such
changes can affect ERK. Because FAK is in excess of
ligand and receptor binding sites, the magnitude of FAKP
is directly proportional to ligand density, whereas the time
required to reach the maximal FAKP is unchanged. As
shown in Fig. 3a, increases in ligand density correspond
to increases in ERKPP magnitude and decreases in the
time to reach the maximum. To determine whether such
effects are specific to changes in the magnitude of FAKP,
we considered the effects of changes in the time to reach
the maximal FAKP by changing the total amount of FAK.
Increased levels of FAK drive the rate of its activation,
resulting in increased kinetics without changing the magni-
tude of FAKP. Fig. 4a shows that equivalent increases in
ligand density and FAK decrease the time of the ERKPP
peak similarly. Fig. 4a also illustrates that ligand density
has a more pronounced effect on ERKPP magnitude than
does FAK. In order to better compare the effects of magni-
tude and kinetics directly, FAK activation dynamics are
expressed as an initial rate of activation (Fig. 4b). If the
effects of FAKP magnitude and kinetics are equivalent,
each initial rate calculated from either ligand density or
total FAK changes should correspond to a unique change
in ERK activation dynamics. However, the curves for
ERKPP dynamics differ depending on ligand density or
total FAK changes, indicating that the effects are parameter-
specific. In fact, ligand density, via changes in the magni-
tude of FAKP, appears to be a stronger regulator of
ERKPP dynamics than total FAK, which changes the kin-
etics of FAKP. Furthermore, the differing curves in
response to ligand density and FAK imply that the initial
rate of FAKP is not the sole determinant of ERKPP
dynamics and that other factors affected by these two par-
ameters are also playing a role. Moreover,
FAK-independent pathways are not likely the source of
these factors because their contributions to ERK activation
are only on the order of 10
219
to 10
218
M. Regardless of
these factors, the relationship between ERKPP and initial
rate is nonlinear and plateaus as the initial rate in creases.
Thus, the effects of ligand density and FAK will attenuate
as these two factors reach saturation.
3.3.2 Ras and Raf as regulators of ERK activity: To
better understand the behaviour of the system, parameters
that play a key role in regulating the dynamics of the
ERK signal were identified by parameter variations. The
duration of ERK activation is found to be particularly sen-
sitive to the concentrations of Ras and Raf. Testing Raf con-
centrations over an order of magnitude from 10 to 100 nM
produced ERKPP signals with a range of durations
(Fig. 5a). At or above a [Raf] of 100 nM, the ERKPP
signal is a sharp peak at approximately 5 min, which falls
off quickly by 30 min. As [Raf] is decreased, the ERKPP
signal duration increases until a sustained signal is obtained
at 10 nM. Between 20 and 60 nM, very small changes in Raf
concentration can give rise to significant changes in signal
duration. For example, a 25% concentration decrease from
40 to 30 nM results in a signal duration increase of nearly
45%. For high [Raf], the ERKPP signal is very similar to
that obtained by Asthagiri et al. [20]. Similarly, the duration
of the ERKPP signal increases with increasing levels of Ras
(Fig. 5b). Although the changes in ERKPP are not as sensi-
tive to small changes in [Ras], large increases in [Ras] do
yield significant prolongation of ERK activation.
The sensitivity of ERKPP to Raf and Ras originates at the
locus of intersection between the RasGTP/GDP cycle and
the MAPK cascade (Fig. 1b), where RasGTP binds and acti-
vates Raf, forming Raf
. The fast drop off in signalling that
forms the peak in ERKPP signals results from the depletion
of RasGTP available for interaction with Raf. When
RasGTP is depleted, no more Raf can be activated
causing the Raf
and subsequently ERKPP signals to fall
off, resulting in the formation of a peak. Therefore, as
[Raf] increases, the Raf – RasGTP binding reaction also
becomes faster, thereby increasing the rate at which
RasGTP is depleted. Similarly, Sydor et al. [43] have
shown that the Raf–Ras complex has an extremely short
lifetime, after which Raf is released for downstream signal-
ling. The resulting Ras species, denoted RasGTP
,isno
longer capable of activating Raf and will undergo GTP
hydrolysis catalysed by GTpase activating proteins
(GAPs), which are recruited to focal adhesions after integ-
rin–ligand binding [44, 45]. In the case of high [Raf], Ras
accumulates as RasGTP
, suggesting that the Raf –
RasGTP reaction is much faster than the action of GAPs
in recycling Ras. Thus, nearly 80% of Ras accumulates in
the GTP
form and is not recycled fast enough to the
Fig. 4 Effect of increase in ligand density and FAK on time
a Fold changes in ERK activation dynamics as a function of fold
changes in ligand density and total FAK
All fold changes are normalised to the starting values (53/mm
2
and
2 10
6
, respectively)
b Fold changes in ERK activation dynamics as a function of the initial
rate of FAK activation
Initial rates in response to changes in ligand density and total FAK
were calculated based on the dynamics of FAK activation from
t ¼ 470– 480 s, where the signal is linear
IET Syst. Biol., Vol. 2, No. 1, January 2008 11
RasGTP form to prevent depletion. As [Raf] is reduced, the
rate of RasGTP conversion to RasGTP
is more comparable
to the rate of recycling to RasGTP. Depletion of RasGTP is
delayed and an increasingly sustained ERKPP signal is
obtained. An increase in [Ras], conversely, increases the
amount of RasGTP available, prolonging the ERKPP signal.
3.4 Sensitivity analysis
3.4.1 Mo rris screening: To systematically identify key
parameters involved in ERK signalling, the Morris screen-
ing method [21] was applied to the model. The Morris
method allows for the effects of a particular parameter vari-
ation to be seen in the context of different values for every
other parameter, as opposed to fixing all other parameters.
The result of the model is a measure of the effect each par-
ameter has on the output signal. For the dynamics of ERK
activation, the effect of each parameter on the overall
ERK time course as measured by the norm of the change
in output, time to reach the maximum, the peak magnitude
and three specific time points was calculated. These values
were normalised by the relative change in input values.
However, because parameters were sampled over a log
uniform distribution, the fractional change remains constant
between parameters. Morris runs were performed until at
least the top 20% of parameters converged (n ¼ 160).
The main parameters involved in regulating ERK acti-
vation are rate constants associated with the MAPK
cascade. However, of the factors that could be readily
manipulated experimentally, Ras and Raf fall within the
top 15% of the key parameters for several measures of
ERK dynamics, confirming the effects on ERKPP signal
duration described above (Table 1). Out of 106 parameters,
Ras ranked second in importance for the time of the ERKPP
peak, 16th based on overall time course and 29th based on a
late time point that indicates signal duration. Similarly, Raf
ranks fourth in determining the time of the ERKPP peak and
seventh in early activation. Both Ras and Raf also fall
within the top 20% of the parameters affecting the magni-
tude of ERK activation, ranking 18th and 19th, respectively.
Interestingly, RasGAP ranks first out of all the parameters
for the time of the ERK activation peak, again highlighting
the importance of the RasGDP/GTP cycle in regulating
ERK activation dynamics. Receptor density also falls
within the top 15% of the key parameters affecting the
ERKPP signal. Moreover, integrin density and contact
area, which determine the total amount of fibronectin avail-
able, are ranked seventh and fiveth, respectively, as regula-
tors of the kinetics of ERK activation. FAK is also identified
as a key parameter in ERK activation kinetics, supporting
the idea that initial FAK signalling contributes significantly
to the activation of ERK. Integrin density, and not fibronec-
tin density, was likely identified in the Morris screening
because of the relative values of the two parameters used.
Depending on the values of receptor and ligand density,
the rate-limiting density will be identified as the most
important parameter. As a result, the importance of integrin
density also points to the importance of fibronectin density
in modulating the magnitude and kinetics of ERKPP.
3.4.2 One-at-a-time parameter variation: In order to
gauge the robustness of the model to changes in parameters,
one-at-a-time parameter variations were performed when
holding every other parameter fixed at their assumed
values. Parameters were varied over three orders of magni-
tude above and below their assumed values, and the range
over which the conclusions drawn in Section 3 held qualitat-
ively was determined (Fig. 6). The model is found to be
robust and tolerant of significant changes in many par-
ameters. Parameters 24 – 64 and 85–105 refer to reactions
involved in Src activation, Shc phosphorylation and recruit-
ment of both Grb2 and Sos. Even many of the MAP kinase
cascade reaction rates, including some that were identified
as the key regulators by Morris screening (ex. 81, 79, 74,
83, 82 and so on), do not limit the qualitative conclusions
drawn from the model.
4 Discussion
This model for the integrin activation of ERK successfully
replicated experimentally observed behaviour for FAK and
ERK activation, as well as changes in ERK activation in
response to fibronectin density. Application of the Morris
method to the model allowed the identification of the key
parameters contributing to ERKPP dynamics. Thus, the
model serves as a tool for understanding and identifying
characteristics of the pathway that contribute to different
ERKPP behaviour. The insights gained by such models
are important steps to ultimately controlling signalling path-
ways in vivo.
Fig. 5 ERK signals with a range of durations
a ERK activation in response to [Raf]
[Ras] is assumed to be 1.072 10
25
M
b ERK activation in response to [Ras]
Raf is assumed to be 3.76 10
27
M
IET Syst. Biol., Vol. 2, No. 1, January 200812
The ability of the model to replicate experimentally
measured time courses for FAK and ERK in response to
adhesion on fibronectin is an important foundation for
understanding the adhesion-specific parameters that might
contribute to ERK activation. The appearance of an early
peak for ERKPP is not surprising as long as there is an
excess of FAK, which serves as an early stimulus for
ERKPP. Factors affecting FAKP magnitude, such as fibro-
nectin density, are found to be stronger regulators of
ERKPP dynamics than those affecting only FAKP kinetics.
However, the effects of ligand density are not solely depen-
dent on the initial rate of FAK activatio n, but are likely
transduced downstream by additional pathways. Although
additional Western blo t time courses have been published,
we chose a semi-quantitative time course for the purpose
of comparing to simulated results. Deviations between the
experimental data used here and other available data may
be due to a myriad of differences including cell type and
experimental procedure.
The simpli fied integrin clustering model used is shown to
be sufficient for predicting the effects of ligand density on
ERK activation. The model based on dimerisation predicts
that increases in ligand density will increase the ma gnitude
of FAK activation, which has been shown by Garcia and
Boettiger [46]. Changes in the magnitude of FAK activation
are then translated into changes in the kinetics and magni-
tude of ERK activation. Although k
p
and k
m
may alter the
time at which FAK activation reaches its maximum, the
ultimate magnitude, as specified by ligand density, will
not be affected. Changes in k
1
may also affect the magnitude
of FAK activation, but the relative changes that result from
ligand density variations will remain the same. More gener-
ally, the magnitude of FAK activation increases with ligand
density regardless of the assumed value of k
1
or k
p
and k
m
.
As a result, changes in downstream signalling that originate
with ligand density will also remain the same qualitatively.
Future work should include a more detailed model of the
integrin clustering process so that the roles of additional
parameters involved in clustering can be tested.
Raf and Ras were identified as the key regulators of ERK
signalling magnitude and duration, in agreement with the
control analysis performed by Hornberg et al. [19].
Furthermore, the behaviour of proteins involved in
RasGDP/GTP cycling reveals that Raf and Ras may func-
tion as a key checkpoint in the MAPK cascade because of
the relative kinetics of RasGDP/GTP cycling and Ras –
Raf binding, which allow for the fo rmation of transient
and sustained ERKPP signals. However, the main effects
of the Raf oncogene may not be mediated by this particular
characteristic of the signalling pathway, as was suggested
by Hornberg. Increases in [Raf] result in increasing magni-
tudes of ERK activation, but increasingly transient acti-
vation signals. For cell-cycle progression and uncontrolled
growth, one would expect that increasing the levels of onco-
genic Raf would instead lead to sustained signalling. On the
other hand, this model shows that increases in the level of
the oncogene Ras result in more sustained ERK signals,
although larger changes in the Ras concentration are
required. This is consistent with the role of Ras as an onco-
gene that promotes tumorigenic behaviour. Moreover, the
Table 1: Experimentally controllable parameters
identified by Morris screening within the top 85th
percentile of parameters important for ERK activation
ERKPP behaviour Experimentally controllable
parameters within 85th
percentile
overall time course PP3 (2), ERK (3), PP2 (7), MEK (8),
PP1 (9), Raf (12), Rt (15), Ras
(16)
time to max RasGAP (1), Ras (2), MEK (3), Raf
(4), CA (5), PP2 (6), Rt (7), FAK
(11), Grb2 (16)
t
1
¼ 1500 s PP3 (2), ERK (3), PP2 (5), Raf (7),
PP1 (9), Grb2 (11), MEK (14)
t
2
¼ 4000 s PP3 (2), ERK (3), PP2 (5), MEK (8),
Ras (10), Rt (11), PP1 (13)
t
3
¼ 15000 s ERK (2), PP3 (4), MEK(6), PP2 (7),
PP1 (10)
maximum value PP3 (2), ERK (3), PP2 (6), PP1 (8),
Grb2 (10), MEK (11), Rt (12)
Each parameter’s rank out of the total 106 parameters is included
in parentheses. PP1, PP2 and PP3 refer to phosphatases for Raf,
MEK and ERK, respectively. CA denotes contact area
Fig. 6 Parameter ranges over which qualitative results hold
Parameters were varied one at a time over six orders of magnitude
when holding all other fixed
Changes in parameter are expressed as log
10
of the fold change
Parameters that correspond to each parameter number along the x-axis
are listed in the Supplementary Materials
Fig. 7 Effects of Raf and Ras concentrations on ERK activity
Concentrations of Ras and Raf are expressed as multiples of the
minimum plotted values (5 10
27
and 1 10
28
M, respectively)
Magnitude of ERK, expressed as a percentage of the maximal ERK
activation, is also plotted for different values of Raf (B)
Shaded regions represent an initial transient activation followed by
sustained mid-level signalling at 33– 66% of maximum (sustained),
sustained high-level signaling at 66– 100% of maximum (plateau), a
decrease to less than 33% of maximal signal within 6 h (transient) or
a decrease to less than 33% of maximum between 6 and 12 h
(transition)
IET Syst. Biol., Vol. 2, No. 1, January 2008 13
behaviour of the Raf/Ras checkpoint suggests that
mutations in Ras that affect the GDP/GTP cycling, such
as its affinity for GAPs or GTPases, may be a mechanism
for modulating cell behaviour. Therefore the oncogenic
effects of Raf may be mediated mainly by changes in the
magnitude of ERK activation, whereas the effects of Ras
are more likely to involve changes in signal duration.
This is consistent with Roovers and Assoian’s [5] sugges-
tion that cell-cycle progression requires a large transient
ERK activation that will induce p21cip expression in the
early G1 phase followed by a sustained mid-level activation
that promotes cyclin D1 expression and stops the induction
of p21cip by late G1 phase. It has also been shown that
p21cip is induced by high levels of Raf [47, 48]. Our
results would suggest that an ideal combination of Raf
and Ras levels could create cell-growth-promoting ERK
activities. To determine these levels, Raf and Ras concen-
trations were varied and their effects on the magnitude
and duration of ERK activation were observed (Fig. 7).
The concentrations plotted are model-specific, however
the trends hold regardless of the parameter values.
Increasing Ras concentration promotes transition of ERK
activity from transient to sustained behaviour. Here, transi-
ent and sustained behaviour refer to initial ERK activation
that falls to less than a one-third and between one-third
and two-thirds of initial activation, respectively. At low
[Raf], a plateau in ERK activation may form in addition
to transient and sustained behaviour for certain [Ras].
Also, the magnitude of ERK activity is found to be signifi-
cantly decreased with [Raf] so that cell-cycle progression
might be inhibited regardless of the signal duration. When
both Raf and Ras concentrations are high, ERK activation
exhibited a large magnitude initial peak that fell off to sus-
tained mid-level activation. This behaviour is consistent
with cell-cycle progression. This may correspond to cells
in which both Raf and Ras oncogenes are over-expressed
or contain activating mutations. Thus, both Raf and Ras
may act together to promote cell growth. Raf contributes
to the magnitude of the signal, bringing the initial transient
activation of ERK to levels that allow for induction of
p21cip, whereas Ras primarily extends the duration of
ERK, sustaining mid-level activation for long times.
It has been suggested that the peak in ERK activation
arises from negative feedback, which turns off the ERK
activation signal [49–51]. Feedback, either positive or
negative, has not been included in this system. Although
negative feedback may occur, we have shown that the
Raf/Ras checkpoint can also play a role in regulating
signal duration, particularly in controlling the fall-off of
the ERK activation sign al. In fact, for certain Raf and Ras
concentrations, the transient activity of ERK decreases
rapidly, but does not approach zero, instead remaining at
some lower level of ERK activation. In such cases, a nega-
tive feedback might be an important mechanism for shutting
down the remaining ERK activity.
Src and Shc represent parallel pathways by which the
MAPK cascade can be activated by FAK. As the model is
run now, the Shc pathway dominates signalling to ERK. It
is possible for Src to be the main contributor to ERK acti-
vation, but that assumption is not explored in detail in this
paper. However, the overall effects seen in this paper
were confirmed in a pathway in which the levels of Shc
are significantly reduced so as to allow the Src pathway to
dominate (data not shown).
Although our model is robust and sufficient for predicting
ERK activation by adhesion, a complete model of ERK acti-
vation should take into account the contributions of both
growth factors and adhesion. Cross-talk between growth
factor and adhesion signalling is vital for producing high-
magnitude, sustained ERK activation. In this paper, a
model for adhesion-based activation is presented. Future
work will involve the addition of the growth-factor-mediated
pathway. Coupled with the detailed model of clustering
described above, the model not only complements exper-
imental systems in which signalling in adhesi ve, growth
factor stimulated cells is measured, but also a means for
exploring the cross-talk between growth factors and
receptors.
5 Acknowledgments
This work was supported by a grant from Department of
Defense W81XWH-05-1-330 to D.A. Hammer and
V.M. Weaver.
6 References
1 Hanahan, D., and Weinberg, R.A.: ‘The hallmarks of cancer’, Cell,
2000, 100, (1), pp. 57–70
2 Aplin, A.E., and Juliano, R.L.: ‘Regulation of nucleocytoplasmic
trafficking by cell adhesion receptors and the cytoskeleton’, J. Cell
Biol., 2001, 155, (2), pp. 187–191 [Epub 15 October 2001]
3 Shaulian, E., and Karin, M.: ‘AP-1 in cell proliferation and survival’,
Oncogene, 2001, 20, (19), pp. 2390 –2400
4 Marshall, C.: ‘How do small GTPase signal transduction pathways
regulate cell cycle entry?’, Curr. Opin. Cell Biol., 1999, 11, (6),
pp. 732 –736
5 Roovers, K., and Assoian, R.K.: ‘Integrating the MAP kinase signal
into the G1 phase cell cycle machinery’, Bioessays, 2000, 22, (9),
pp. 818 –826
6 Danen, E.H., and Yamada, K.M.: ‘Fibronectin, integrins, and growth
control’, J. Cell Physiol., 2001, 189, (1), pp. 1– 13
7 Renshaw, M.W., Ren, X.D., and Schwartz, M.A.: ‘Growth factor
activation of MAP kinase requires cell adhesion’, EMBO J., 1997,
16, (18), pp. 5592–5599
8 Giancotti, F.G., and Ruoslahti, E.: ‘Integrin signaling’, Science, 1999,
285, (5430), pp. 1028 – 1032
9 Barberis, L., Wary, K.K., Fiucci, G., Liu, F., Hirsch, E., Brancaccio, M.,
Altruda, F., Tarone, G., and Giancotti, F.G.: ‘Distinct roles of the
adaptor protein Shc and focal adhesion kinase in integrin signaling to
ERK’, J. Biol. Chem.,2000,275, (47), pp. 36532 –36540
10 Schlaepfer, D.D., and Hunter, T.: ‘Integrin signalling and tyrosine
phosphorylation: just the FAKs?’, Trends Cell Biol., 1998, 8, (4),
pp. 151 –157
11 Wary, K.K., Mainiero, F., Isakoff, S.J., Marcantonio, E.E., and
Giancotti, F.G.: ‘The adaptor protein Shc couples a class of
integrins to the control of cell cycle progression’, Cell, 1996, 87,
(4), pp. 733–743
12 Asso ian, R.K., and Schwartz, M.A.: ‘Coordinate signaling by
integrins and receptor ty rosi ne kinases in the regulation of G1
phase cell-cycle progre ssion’, Curr. Opin. Genet. Dev., 2001, 11,
(1), pp. 48– 53
13 Giancotti, F.G.: ‘Integrin signaling: specificity and control of cell
survival and cell cycle progression’, Curr. Opin. Cell Biol., 1997, 9,
(5), pp. 691–700
14 Schwartz, M.A., and Ginsberg, M.H.: ‘Networks and crosstalk:
integrin signalling spreads’, Nat. Cell Biol., 2002, 4, (4),
pp. E65 – E68
15 Roovers, K., Davey, G., Zhu, X., Bottazzi, M.E., and Assoian, R.K.:
‘Alpha5beta1 integrin controls cyclin D1 expression by sustaining
mitogen-activated protein kinase activity in growth factor-treated
cells’, Mol. Biol. Cell, 1999, 10, (10), pp. 3197– 3204
16 Kholodenko, B.N., Demin, O.V., Moehren, G., and Hoek, J.B.:
‘Quantification of short term signaling by the epidermal growth
factor receptor’, J. Biol. Chem., 1999, 274, (42), pp. 30169–30181
17 Schoeberl, B., Eichler-Jonsson, C., Gilles, E.D., and Muller, G.:
‘Computational modeling of the dynamics of the MAP kinase
cascade activated by surface and internalized EGF receptors’, Nat.
Biotechnol., 2002, 20, (4), pp. 370 – 375
18 Chapman, S., and Asthagiri, A.R.: ‘Resistance to signal activation
governs design features of the MAP kinase signaling module’,
Biotechnol. Bioeng., 2004, 85, (3), pp. 311–322
19 Hornberg, J.J., Binder, B., Bruggeman, F.J., Schoeberl, B., Heinrich,
R., and Westerhoff, H.V.: ‘Control of MAPK signalling: from
complexity to what really matters’, Oncogene, 2005, 24, (36),
pp. 5533 – 5542
IET Syst. Biol., Vol. 2, No. 1, January 200814
20 Asthagiri, A.R., Nelson, C.M., Horwitz, A.F., and Lauffenburger,
D.A.: ‘Quantitative relationship among integrin-ligand binding,
adhesion, and signaling via focal adhesion kinase and extracellular
signal-regulated kinase 2’, J. Biol. Chem., 1999, 274, (38),
pp. 27119 –27127
21 Morris, M.D.: ‘Factorial sampling plans for preliminary
computational experiments’, Technometrics, 1991, 33, (2),
pp. 161 –174
22 Giannone, G., Dubin-Thaler, B.J., Dobereiner, H.G., Kieffer, N.,
Bresnick, A.R., and Sheetz, M.P.: ‘Periodic lamellipodial
contractions correlate with rearward actin waves’, Cell, 2004, 116,
(3), pp. 431–443
23 Bell, G.I.: ‘Models for the specific adhesion of cells to cells’, Science,
1978, 200, (4342), pp. 618–627
24 Parsons, J.T.: ‘Focal adhesion kinase: the first ten years’, J. Cell Sci.,
2003, 116, (8), pp. 1409–1416
25 Aplin, A.E., Howe, A., Alahari, S.K., and Juliano, R.L.: ‘Signal
transduction and signal modulation by cell adhesion receptors: the
role of integrins, cadherins, immunoglobulin-cell adhesion
molecules, and selectins’, Pharmacol. Rev., 1998, 50, (2),
pp. 197 –263
26 Kaplan, K.B., Bibbins, K.B., Swedlow, J.R., Arnaud, M., Morgan,
D.O., and Varmus, H.E.: ‘Association of the amino-terminal half of
C-Src with focal adhesions alters their properties and is regulated by
phosphorylation of tyrosine-527’, EMBO J., 1994, 13, (20),
pp. 4745 – 4756
27 Bjorge, J.D., Jakymiw, A., and Fujita, D.J.: ‘Selected glimpses into the
activation and function of Src kinase’, Oncogene, 2000, 19, (49),
pp. 5620 – 5635
28 Boerner, R.J., Kassel, D.B., Barker, S.C., Ellis, B., DeLacy, P.,
and Knight, W.B.: ‘Correlation of the phosphorylation states
of pp60c-src with tyrosine kinase activity: the intramolecular
pY530-SH2 complex retains significant activity if Y419 is
phosphorylated’, Biochemistry, 1996, 35, (29), pp. 9519–9525
29 Cole, P., Burn, P., Takacs, B., and Walsh, C.: ‘Evaluation of the
catalytic mechanism of recombinant human Csk (C-terminal Src
kinase) using nucleotide analogs and viscosity effects’, J. Biol.
Chem., 1994, 269, (49), pp. 30880–30887
30 Schlaepfer, D.D., and Hunter, T.: ‘Focal adhesion kinase
overexpression enhances ras-dependent integrin signaling to ERK2/
mitogen-activated protein kinase through interactions with
and activation of c-Src’, J. Biol. Chem., 1997, 272, (20),
pp. 13189 –13195
31 Schlaepfer, D.D., Jones, K.C., and Hunter, T.: ‘Multiple
Grb2-mediated integrin-stimulated signaling pathways to ERK2/
mitogen-activated protein kinase: summation of both c-Src- and
focal adhesion kinase-initiated tyrosine phosphorylation events’,
Mol. Cell Biol., 1998, 18, (5), pp. 2571–2585
32 Downward, J.: ‘The GRB2/Sem-5 adaptor protein’, FEBS Lett., 1994,
338, (2), pp. 113–117
33 Howe, A.K., and Juliano, R.L.: ‘Distinct mechanisms mediate the
initial and sustained phases of integrin-mediated activation of the
Raf/MEK/mitogen-activated protein kinase cascade’, J. Biol.
Chem., 1998, 273, (42), pp. 27268–27274
34 Wary, K.K., Mariotti, A., Zurzolo, C., and Giancotti, F.G.: ‘A
requirement for caveolin-1 and associated kinase Fyn in integrin
signaling and anchorage-dependent cell growth’, Cell, 1998, 94, (5),
pp. 625 –634
35 Ladbury, J.E., Lemmon, M.A., Zhou, M., Green, J., Botfield, M.C.,
and Schlessinger, J.: ‘Measurement of the binding of tyrosyl
phosphopeptides to SH2 domains: a reappraisal’, Proc. Natl Acad.
Sci. USA, 1995, 92, (8), pp. 3199– 3203
36 Payne, G., Shoelson, S.E., Gish, G.D., Pawson, T., and Walsh, C.T.:
‘Kinetics of p56lck and p60src Src homology 2 domain binding to
tyrosine-phosphorylated peptides determined by a competition assay
or surface plasmon resonance’, Proc. Natl Acad. Sci. USA, 1993,
90, (11), pp. 4902–4906
37 Kaneko, T., Kumasaka, T., Ganbe, T., Sato, T., Miyazawa, K.,
Kitamura, N., and Tanaka, N.: ‘Structural insight into modest
binding of a non-PXXP ligand to the signal transducing adaptor
molecule-2 Src homology 3 domain’, J. Biol. Chem., 2003, 278,
(48), pp. 48162–48168 [Epub 16 September 2003]
38 Matsuda, M., Ota, S., Tanimura, R., Nakamura, H., Matuoka, K.,
Takenawa, T., Nagashima, K., and Kurata, T.: ‘Interaction between
the amino-terminal SH3 domain of CRK and its natural target
proteins’, J. Biol. Chem., 1996, 271, (24), pp. 14468–14472
39 Fincham, V.J., Unlu, M., Brunton, V.G., Pitts, J.D., Wyke, J.A., and
Frame, M.C.: ‘Translocation of Src kinase to the cell periphery is
mediated by the actin cytoskeleton under the control of the Rho
family of small G proteins’, J. Cell Biol., 1996, 135, (6, Pt 1),
pp. 1551 – 1564
40 Wang, D., Esselman, W.J., and Cole, P.A.: ‘Substrate conformational
restriction and CD45-catalyzed dephosphorylation of tail
tyrosine-phosphorylated Src protein’, J. Biol. Chem., 2002, 277,
(43), pp. 40428–40433
41 Wang, J., and Walsh, C.T.: ‘Mechanistic studies on full length
and the catalytic domain of the tandem SH2
domain-containing protein tyrosine phosphatase: analysis of
phosphoenzyme levels and Vmax stimulatory effects of glycerol and
of a phosphotyrosyl peptide ligand’, Biochemistry, 1997, 36, (10),
pp. 2993 – 2999
42 Zhang, Z.: ‘Kinetic and mechanistic characterization of a mammalian
protein-tyrosine phosphatase, PTP1’, J. Biol. Chem., 1995, 270, (19),
pp. 11199 – 11204
43 Sydor, J.R., Engelhard, M., Wittinghofer, A., Goody, R.S.,
and Herrmann, C.: ‘Transient kinetic studies on the interaction of
Ras and the Ras-binding domain of c-Raf-1 reveal rapid
equilibration of the complex’, Biochemistry, 1998, 37, (40),
pp. 14292 – 14299
44 Hecker, T.P., Ding, Q., Rege, T.A., Hanks, S.K., and Gladson, C.L.:
‘Overexpression of FAK promotes Ras activity through the
formation of a FAK/p120RasGAP complex in malignant
astrocytoma cells’, Oncogene, 2004, 23, (22), pp. 3962 – 3971
45 Sharma, S.V.: ‘Rapid recruitment of p120RasGAP and its associated
protein, p190RhoGAP, to the cytoskeleton during integrin mediated
cell–substrate interaction’, Oncogene, 1998, 17, (3), pp. 271–281
46 Garcia, A.J., and Boettiger, D.: ‘Integrin–fibronectin interactions at
the cell–material interface: initial integrin binding and signaling’,
Biomaterials, 1999, 20, (23 – 24), pp. 2427–2433
47 Sewing, A., Wiseman, B., Lloyd, A.C., and Land, H.: ‘High-intensity
Raf signal causes cell cycle arrest mediated by p21Cip1’, Mol. Cell.
Biol., 1997, 17, (9), pp. 5588 –5597
48 Woods, D., Parry, D., Cherwinski, H., Bosch, E., Lees, E., and
McMahon, M.: ‘Raf-induced proliferation or cell cycle arrest is
determined by the level of Raf activity with arrest mediated by
p21Cip1’, Mol. Cell. Biol., 1997, 17, (9), pp. 5598–5611
49 Dougherty, M.K., Muller, J., Ritt, D.A., Zhou, M., Zhou, X.Z.,
Copeland, T.D., Conrads, T.P., Veenstra, T.D., Lu, K.P., and
Morrison, D.K.: ‘Regulation of raf-1 by direct feedback
phosphorylation’, Mol. Cell, 2005, 17, (2), pp. 215 – 224
50 Sundberg-Smith, L.J., Doherty, J.T., Mack, C.P., and Taylor, J.M.:
‘Adhesion stimulates direct PAK1/ERK2 association and leads to
ERK-dependent PAK1 Thr(212) phosphorylation’, J. Biol. Chem.,
2005, 280, (3), pp. 2055–2064
51 Brummer, T., Naegele, H., Reth, M., and Misawa, Y.: ‘Identification
of novel ERK-mediated feedback phosphorylation sites at
the C-terminus of B-Raf’, Oncogene, 2003, 22, (55),
pp. 8823–8834
IET Syst. Biol., Vol. 2, No. 1, January 2008 15
