Systems biology analysis of G protein and MAP kinase signaling in yeast
N Hao1, M Behar2, TC Elston1and HG Dohlman1,3
1Department of Pharmacology, University of North Carolina, Chapel Hill, NC, USA;2Department of Physics, University of North
Carolina, Chapel Hill, NC, USA and3Department of Biochemistry and Biophysics, University of North Carolina, Chapel Hill,
Approximately a third of all drugs act by binding directly
to cell surface receptors coupled to G proteins. Other
drugs act indirectly on these same pathways, for example,
by inhibiting neurotransmitter reuptake or by blocking the
inactivation of intracellular second messengers. These
drugs have revolutionized the treatment of human disease.
However, the complexity of G protein signaling mechan-
isms has significantly hampered our ability to identify
additional new drug targets. Moreover, today’s molecular
pharmacologists are accustomed to working on narrowly
focused problems centered on a single protein or
enzymatic process. Here we describe emerging efforts in
yeast aimed at identifying proteins and processes that
modulate the function of receptors, G proteins and MAP
kinase effectors. The scope of these efforts is far more
systematic, comprehensive and quantitative than anything
attempted previously, and includes integrated approaches
in genetics, proteomics and computational biology.
Oncogene (2007) 26, 3254–3266. doi:10.1038/sj.onc.1210416
Keywords: mitogen-activated protein kinases; yeast;
genetics; modeling; systems biology; desensitization
G proteins and MAPK signaling
Mechanism of G protein-mediated signal transduction
The actions of many sensory signals, hormones and
neurotransmitters are mediated by cell surface receptors,
which have in common a 7-transmembrane topology
(7TM receptors or 7TMRs). Receptors of this class are
conserved in organisms as diverse as humans, yeasts and
plants. They represent the largest gene family in the
human genome, and are the target of most pharmaceu-
ticals in use today (Hardman et al., 2001).
Generally speaking, 7TMRs transmit their signals via
G proteins. Upon agonist binding to its receptor, the
G protein a subunit releases guanosine diphosphate
(GDP), binds guanosine triphosphate (GTP) and dis-
sociates from the G protein bg subunit complex (Sprang,
1997). Depending on the system, either Ga or Gbg or
both can then activate multiple downstream effectors
such as adenylyl cyclase, ion channels and phospholi-
pases. In most cases, effector activation occurs at the
plasma membrane, but activation results in the produc-
tion of second messengers (e.g. cAMP or calcium) that
diffuse throughout the cell and trigger global changes
including protein phosphorylation, new gene transcrip-
tion and alterations in cell homeostasis or differentiation
(Neves et al., 2002). Other effectors include protein
kinase cascades that include members of the mitogen-
activated protein kinase (MAPK) family. More recently
in mammalian cells, it has been shown that 7TMRs can
also signal independently of G proteins, and can instead
activate MAPKs through a family of adaptor proteins
called arrestins (Shenoy and Lefkowitz, 2005).
MAPK activation by G proteins can occur through
phospholipase Cb, which in turn activates Ras and/or
protein kinase C. Either protein can then activate Raf
kinase, MAPK kinases, and ultimately a MAPK. Other
routes involve activation of the Ras-family members
Rac, Cdc42 and Rap. Once activated, the MAPK
translocates to the nucleus where it phosphorylates
transcription factors, thereby regulating the expression
of genes that control cell growth and differentiation
(Gutkind, 1998; Marinissen and Gutkind, 2001).
Less studied are non-nuclear substrates for MAPKs.
Only a handful have been identified, but this list includes
G protein signaling components such as the GRK
receptor kinases and b-arrestin, two classes of proteins
that modulate the intensity and specificity of G protein-
coupled receptor signaling (Lin et al., 1999; Pitcher
et al., 1999). Regulators of G protein signaling (RGS
proteins, described below) are likewise phosphorylated
by downstream MAPKs (Garrison et al., 1999; Ogier-
Denis et al., 2000; Parnell et al., 2005). Thus, the activity
of receptors and G proteins may be coordinated in part
by the ability of MAPKs to phosphorylate regulatory
factors acting throughout the signaling pathway, in both
feed-forward and feedback mechanisms.
Ga proteins eventually hydrolyze GTP to GDP, at
which point the G protein subunits re-associate and
signaling stops. GTPase activity is further accelerated by
RGS proteins. So as the specificity of the agonist signal
depends on the ability of G proteins to couple only to
Correspondence: Associate Professor TC Elston, Department of
Pharmacology, University of North Carolina, CB 7365, Chapel Hill,
NC 27599-7365, USA or Professor HG Dohlman, Department of
Biochemistry and Biophysics, University of North Carolina, CB 7260,
Chapel Hill, NC 27599-7260, USA.
E-mail: firstname.lastname@example.org or email@example.com
Oncogene (2007) 26, 3254–3266
& 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00
the appropriate receptor and effector, the intensity
of the signal depends on the rates of GTP binding
and hydrolysis, events catalyzed by agonist-occupied
receptors and RGS proteins, respectively (Neubig and
Siderovski, 2002; Chasse and Dohlman, 2003).
Because of the potential for new pharmacology, an
emerging area of investigation is centered on how signal
intensity is regulated. Below we describe examples of key
regulatory mechanisms acting on the prototypical
7TMR – G protein – MAPK effector cascade. Our
emphasis is on the yeast model, but given the extensive
structural and functional similarities between yeast and
mammalian signaling systems, discoveries made in yeast
are generally applicable as well to humans.
G proteins and MAPKs as drug targets
Just as the majority of all clinically important remedies
are known to act on G protein-coupled receptors at the
cell surface (Hardman et al., 2001), intracellular signal-
ing proteins represent potentially useful drug targets,
particularly in situations where a genetic defect leads to
persistent activation by the receptor. Indeed, the clear
link between genetic alterations in signaling components
and human pathophysiology provides strong evidence
that chemical inhibitors of G protein and MAPK
signaling could be useful and effective as drugs (Roberts
and Der, this issue). Drugs that inhibit MAPK function
could work by interfering with G protein activation,
production, MAPK catalytic function, MAPK phos-
phatase activity or MAPK translocation to the nucleus.
Drug discovery efforts directed at G protein or
MAPK signaling components are still in their infancy,
but early successes have been reported. Academic labs
have identified small molecule inhibitors of RGS
proteins (Roman et al., 2006). Further along in the
drug ‘pipeline’ are small molecule inhibitors of protein
kinases. Striking evidence for the potential of protein
kinases as drug targets comes from the discovery of
imatinib mesylate (Gleevec), an ATP-competitive small-
molecule inhibitor of the Bcr-Abl oncogene product.
Treatment with Gleevec can induce complete remissions
in the early stages of chronic myelogenous leukemia
(Barnes and Melo, 2003). A major issue in any drug
development effort, however, is the identification of
appropriate targets for therapeutic intervention. It is not
sufficient to know that a particular G protein or protein
kinase is activated in a specific disease state, because
dysregulation can be the consequence rather than the
underlying cause of disease pathology. At the very least,
clear genetic or physiological/cell biological data are
needed to implicate a protein kinase as a potential
Another major issue is selecting which isoform(s) of
any given component should be targeted. For example,
there are roughly 40 RGS isoforms, each with broad
tissue distributions (Gold et al., 1997). Among the
MAPK signaling pathways, candidate targets include at
least 11 MAPKs, but there are also another seven
MAPK kinases (MAPKKs) and at least 20 MAPK
kinase kinases (MAPKKKs). The MAPKs fall into five
subfamilies: these are (i) extracellular signal-regulated
kinase (ERK)1/2, (ii) c-Jun N-terminal kinase (JNK)
1/2/3, (iii) p38a, b, g and d, (iv) ERK5 and (v) ERK7.
Splice variants further contributes to the complexity of
selecting appropriate targets. Finally, the broad expres-
sion of RGS and MAPKs, and their function in diverse
biological responses, suggests that undesired effects and
potential toxicity would be observed with inhibitors.
Thus, it remains unclear which, if any, of the specific
isoforms can be inhibited with useful therapeutic
response and lack of adverse effects.
Despite these complications, inhibitors of MAPK-
related processes are in clinical trials or in preclinical
development for a variety of cardiovascular diseases,
including acute coronary syndromes, stroke, hyperten-
sion, diabetes and the metabolic syndrome (reviewed in
(Force et al., 2004; Johnson et al., 2005; Sebolt-Leopold
and English, 2006). Inhibitors of p38 have advanced the
furthest in clinical trials (Kumar et al., 2003). Inhibitors
of the MAPK kinases MAPK/ERK kinase (MEK)1 and
MEK2 are also in clinical trials for the treatment of a
variety of tumors. At least one MAPKKK inhibitor
(CEP-1347) has advanced to Phase II clinical trials
(Roberts and Der, this issue). Although far from
comprehensive, these examples suggest the possibility
of future success.
G protein and MAPK signaling in yeast
Yeast genetics in pharmacology research
Genetic defects in receptors and Ga proteins have been
reported to cause a variety of developmental, metabolic
and endocrine disorders (Gutkind, 1998; Farfel et al.,
1999; Spiegel, 2000; Marinissen and Gutkind, 2001).
Likewise, dysfunction of MAPK signaling is well known
to underlie a broad spectrum of diseases, including
many cancers, cardiac hypertrophy, ischemia/reperfu-
sion injury, angiogenesis and atherogenesis (Cuevas,
Abell and Johnson, this issue). The identification of
genetic defects in human endocrine disease and cancer
has vastly improved our understanding of how G
protein and MAPKs act and how they can fail. They
can also guide efforts towards new pharmacology. One
strategy is to use genetics to identify protein binding
partners of a dysfunctional gene product, as any
interacting proteins represent alternative (nonmutated)
targets for drug therapy (Hartwell et al., 1997). Another
approach has been to screen for second-site mutations
that compensate for the primary defect in a gene. For
example, we exploited the genetic tractability of the
yeast Saccharomyces cerevisiae to identify intragenic
suppressors of a GTPase-deficient Ga mutant, Gpa1Q323L
(Apanovitch et al., 1998a). These second-site suppressor
mutations appear to lock the protein in the inactive
conformation in both the absence and presence of GTP.
Genetic suppression of a disease-causing mutation
provides strong presumptive evidence that pharmacolo-
gical suppression can also be achieved (Gibbs and Oliff,
Systems biology analysis of yeast
N Hao et al
The pheromone/G protein response system in yeast is
arguably the best-characterized signaling pathway of
any eukaryote, and it has long served as a prototype for
hormone, neurotransmitter and sensory response sys-
tems in humans. In yeast G protein signaling is required
for cell–cell communication leading to mating. Haploid
a and a cells secrete type-specific pheromones (a-factor
and a-factor, respectively) that promote cell fusion and
the formation of an a/a diploid. Most components of the
G protein-signaling cascade in yeast were identified
genetically, through isolation of mutants that exhibit a
mating-deficient (or sterile) phenotype. Among these are
genes that encode the pheromone receptor (Ste2), the G
protein b and g subunits (Ste4, Ste18), components of a
MAPK cascade (Ste20, Ste11, Ste7 and Fus3 or Kss1), a
kinase scaffold protein (Ste5) and a transcription factor
(Ste12) (Figure 1). Fus3 acts primarily in the mating
pathway whereas Kss1 acts primarily in the invasive or
‘pseudohyphal’ growth pathway in nutrient-limited
cells. Deletion of both MAPK genes is needed to confer
sterility, however, suggesting that Kss1 can function in
the place of Fus3 (Dohlman and Thorner, 2001).
Dynamic regulation and desensitization
A property of signal-response systems in general, and of
G protein-coupled receptors in particular, is that
prolonged stimulation leads to desensitization. For
mammalian receptors, the mechanisms of desensitiza-
tion are complex, and include rapid alterations such as
phosphorylation by receptor kinases, followed by
uncoupling from the G protein and sequestration within
minutes of agonist activation (mediated in part by
arrestins) (Shenoy and Lefkowitz, 2005). Receptor
desensitization is thought to represent the molecular
basis for habituation to light, odors, chemical stimulants
and narcotics (Dohlman, 2002).
Desensitization typically entails some form of feed-
back inhibition, where a downstream effector alters the
activity of an upstream transducer. For example, many
cell surface receptors are rapidly phosphorylated and
endocytosed following stimulation, and these events
limit their ability to transmit the signal. Such phosphory-
lation-mediated negative feedback mechanisms do not
require new protein synthesis and therefore can act
rapidly, sometimes within seconds of pathway activa-
tion. Other mechanisms can take hours or even days.
For instance, negative regulators such as RGS proteins
or MAPK phosphatases may be transcriptionally
induced. Thus, in any given system there can be multiple
overlapping mechanisms of feedback regulation, each
with distinct temporal characteristics, that collectively
modulate cellular responsiveness to a given stimulus.
Similarly in yeast, prolonged stimulation with mating
pheromones eventually leads to desensitization and a
resumption of normal cell growth (Dohlman and
Thorner, 2001). Genetic methods in yeast have been
used to identify desensitization factors including the first
RGS protein Sst2 (Chan and Otte, 1982a,b; Chasse
et al., 2006). Genetic disruption of the SST2 gene has
no effect on cell viability but allows cells to respond
to concentrations of pheromone at least an order of
magnitude lower than in normal cells, and these cells fail
to recover from pheromone-induced growth arrest even
if the ligand is removed.
With growing appreciation of the similarities between
the mammalian and yeast signaling systems, a detailed
analysis of Sst2 function was initiated in the mid-1990s.
This was done with the expectation – later proven to be
correct – that similar proteins must exist in humans. Sst2
Invasive growthStress adaptation
(Hog1) pathways are shown, as detailed in the text. Note that some pathway components have been omitted for the sake of clarity.
Three MAPK pathways in S. cerevisiae. Signaling components of the mating (Fus3), invasive (Kss1) and osmosensing
Systems biology analysis of yeast
N Hao et al
was shown to act at the level of the G protein and shown
to stimulate Gpa1 GTPase activity, thereby revealing
how Sst2 promotes desensitization in yeast (Dohlman
et al., 1995, 1996; Apanovitch et al., 1998b). Specifically,
Sst2 decreases the lifetime of the active form of the G
protein, thereby dampening the cellular response to
pheromone. Subsequently, more than 40 Sst2 homologs
have been identified in mammals, and are now known
collectively as the RGS protein family (Neubig and
Yeast systems biology
Emerging genomics and proteomics tools in yeast
Yeast is well recognized as an excellent model for genetic
analysis. Their ability to undergo efficient homologous
recombination is particularly useful for studying the
functional role of proteins in vivo, through gene
disruption or gene replacement. Pheromone stimulation
leads to a well-defined series of events required for
mating, including readily-assayed responses such as new
gene transcription, morphological and cytoskeletal
changes, and transient growth arrest in the G1 phase
of the cell cycle. As noted above, the yeast system bears
strong structural and functional similarity to signaling
pathways in mammals. The G-protein and MAPK
components in particular share extensive sequence
similarity with their human counterparts (Dohlman
and Thorner, 2001).
Yeast has also been a pioneering model system for the
development of new proteomic and genomic methodolo-
gies. The yeast genome was the first from any eukaryote
to be fully sequenced (Goffeau et al., 1996). Every gene
has been arrayed for the purpose of transcription
analysis using whole-genome microarrays. Moreover,
most yeast genes have now been fused to green
fluorescent protein (GFP), fused to a variety of affinity
tags (tandem affinity purification (TAP), His6, glu-
tathione-S-transferase (GST)) and deleted (Martzen
et al., 1999; Winzeler et al., 1999; Ghaemmaghami
et al., 2003; Huh et al., 2003; Gelperin et al., 2005).
Indeed, it is now possible to simultaneously track the
expression, localization and activity of nearly every
protein within the cell, from the cell surface to the
nucleus. Notably, most of the available fusion proteins
are expressed at the proper locus and under the control
of the native promoter. These strains are all widely
available from commercial or academic sources.
Case study: systems-level analysis reveals a new signaling
pathway in yeast
In yeast inactivating mutations in the Ga protein Gpa1
do not block pheromone responses but rather produce a
constitutive signal, apparently owing to uncontrolled
pathway activation by free Gbg (reviewed in Dohlman
and Thorner, 2001). These findings provided early
evidence that Gbg is necessary and sufficient for mating
responses, and that Gpa1 functions primarily to
constrain the levels of free Gbg. More recently, however,
GTPase-deficient Gpa1Q323Lwas shown to initiate (in a
dominant manner) many of the same cellular events
triggered by mating pheromone and free Gbg, including
increased MAPK phosphorylation and new gene tran-
scription (Guo et al., 2003). Ga-initiated signaling
requires the phosphoinositide 3-kinase Vps34 and its
regulatory protein Vps15. Gpa1 binds directly to Vps34
and Vps15 in a guanine nucleotide-dependent manner,
promotes an increase in phosphoinositide 3-phosphate
production, and promotes translocation of phospho-
inositide 3-phosphate-binding proteins to the endosome
where Vps34 and Vps15 are active. Deletion or
pharmacological inhibition of Vps34 abrogates all of
these Gpa1 signaling phenotypes (Slessareva et al.,
2006). Thus, genetic, cell biological, biochemical and
pharmacological evidence support the model that Vps34
is a direct effector for Gpa1. Notably, Vps15 and Vps34
represent the first additions to this well-studied signaling
system in recent memory, and illustrate how there
remain many variations of the G protein and MAPK
signaling paradigm yet to be fully appreciated.
The identification of Vps15 and Vps34 represents an
excellent example of how newly available engineered
yeast strains have transformed our approach to signal-
ing research. To identify proteins required for Gpa1-
mediated signaling, the GTPase-deficient mutant was
introduced into each of approximately 5000 gene
deletion strains, along with a transcription-reporter
specific for the mating response pathway. Quantitative
analysis of transcriptional induction revealed several
genes necessary for full Gpa1-mediated signaling,
including the phosphoinositide 3-kinase. Moreover,
proteins that translocate in response to elevated
phosphoinositide 3-phosphate production were identi-
fied by systematic microscopy analysis of GFP-fusion
proteins (Slessareva et al., 2006). These findings
illustrate how signaling research has evolved to become
ever more comprehensive and systematic. If these
findings can be extended to mammalian homologs of
Vps34 and Vps15, these proteins represent potential new
targets for pharmacological modulation of MAPK
Dynamic regulation of signaling pathways
Above we describe examples of how genomics and
proteomics tools can reveal new signaling pathway
components. However, interpreting the results of
genome-scale studies often relies on traditional models
based largely on intuition, often do not consider
signaling dynamics, and are rarely quantitative. Dif-
ferent cellular behaviors are commonly regulated by a
single G protein or MAPK, but the response may
depend on discrete changes in the intensity or duration
of enzyme activity. For example, desensitization is
commonly defined as ‘feedback inhibition after a delay’,
and as such is an excellent example of dynamic behavior
in signaling. Other forms of dynamic behavior are
critically important in cell fate decisions. In one oft-cited
example, epidermal growth factor promotes transient
activation of the ERK MAPK and leads to cell
Systems biology analysis of yeast
N Hao et al
proliferation, whereas nerve growth factor promotes
sustained ERK activation and results in cell differentia-
tion (Marshall, 1995). Thus, a complementary approach
to any systematic or systems-level analysis is the
development of temporally resolved assays, as well as
computational models that consider the temporal
behavior of proteins, including dynamic changes in
protein activity, localization and expression.
Despite the importance of temporal regulation in
signal transduction, the underlying mechanisms leading
to pathway inactivation have been especially difficult to
define. For G proteins, much recent attention has
focused on RGS proteins. With respect to MAPKs,
most attention has focused on feedback inhibition
through phosphorylation, as well as pathway regulation
by protein phosphatases (Morrison and Davis, 2003;
Martin et al., 2005). Phosphorylation is of special
interest because it provides a rapid and reversible means
for the dynamic control of signaling. Thus, the activity
of any MAPK reflects a balance of receptors and RGS
proteins, as well as of activating kinases and phospha-
tases. In the section below, we describe how computa-
tional methods may be applied to the study of signaling
Modeling cell behavior and signaling dynamics
Application of computational methods to biological
problems is not new. Indeed receptor theory, which uses
mathematical modeling to quantify ligand-induced
transitions in the chemical state of the receptor, has
been widely used by pharmacologists for over 25 years
(Kenakin, 2004). The ‘wiring’ of naturally occurring
regulatory networks is more complicated, and certainly
too complex to rely solely on qualitative descriptions
devoid of mathematics. Efforts to model signaling
pathways and regulatory networks has lagged, however,
despite the availability of appropriate mathematical and
computational tools. Early efforts were severely ham-
pered by the lack of suitable experimental data,
including insufficient information about the identity
and activity (in both space and time) of pathway
components. However, new genomics and proteomics
tools, as well as a deeper understanding of the under-
lying biology, have since allowed investigators to
reconsider using computational methods to solve com-
plex biological problems.
With respect to G protein and MAPK signaling, early
efforts have focused on mechanisms of pathway activa-
tion. For example, Yi et al. (2003) modeled the
dynamics of G protein activation in the pheromone
response pathway of yeast. They used fluorescence
resonance energy transfer experiments to measure the
association state of the Ga and Gbg subunits. Mathe-
matical modeling was then used to estimate the in vivo
rates of G-protein activation and deactivation, in the
presence or absence of RGS proteins. Other computa-
tional modeling and experimental investigations led to
the suggestion that RGS proteins act as kinetic scaffolds
(Zhong et al., 2003). In this model the RGS concentra-
tion is uniformly distributed throughout the cell. The
accelerated GTPase activity caused by RGS promotes a
rapid re-coupling of the G-protein to the receptor. This
in turn allows G-proteins to immediately undergo
another round of activation. This has the effect of
increasing the pool of active G-proteins in the vicinity of
receptors, while keeping G-protein activity low else-
where. In another study Bornheimer et al. (2004)
provided a theoretical analysis of the 7TMR-G pro-
tein–RGS cycle. They studied signal propagation under
various relative concentrations of active receptor, G
protein and RGS protein. Their work predicted how
changes in component concentrations would lead to
distinctive signaling behaviors in terms of steady-state G
protein activity and velocity of GTP hydrolysis.
At the level of the effector MAPKs, mathematical
modeling revealed that these kinase cascades can be
intrinsically ‘ultrasensitive’, with Hill coefficients as
large as 5 (Huang and Ferrell, 1996). This prediction
was confirmed in vitro using Xenopus oocyte extracts
(Huang and Ferrell, 1996). Mathematical modeling also
demonstrated that the kinases in MAPK cascades
primarily regulate signal amplitude, whereas the phos-
phatases that mediate pathway deactivation determine
signal duration (Heinrich et al., 2002; Hornberg et al.,
2005). These predictions were confirmed by monitoring
EGF-induced ERK phosphorylation in the presence of
kinase and phosphatase inhibitors (Hornberg et al.,
In the remainder of this review, we focus on temporal
aspects of cell signaling and the mechanisms that
regulate pathway activity. We pay particular attention
to feedback and feed-forward regulation and illustrate
how mathematical modeling has been used to under-
stand the logic of various pathway architectures. We do
not consider mathematical modeling of pathway speci-
ficity (Somsen et al., 2002; Komarova et al., 2005;
Schaber et al., 2006). Although it is an important topic
in cell signaling, the underlying biological mechanisms
that dictate signaling specificity are not well established.
Modeling mechanisms of pathway regulation
In this section, we consider five scenarios of pathway
regulation. The first two are examples of negative
feedback regulation, in which a downstream component
modifies the activity of an upstream component result-
ing in either desensitization or deactivation of the
pathway. The third scenario is an example of positive
feedback regulation, in which a downstream component
promotes further activation of upstream components
leading to signal amplification. Finally, we consider two
scenarios where upstream components modify the
activity of downstream components. Just as the avail-
ability of engineered yeast strains has transformed
experimental approaches to signaling, there are now
several examples of how computational methods have
directed experimental approaches. Thus, we provide
wherever possible specific examples of each scenario
drawn from the yeast model system.
Systems biology analysis of yeast
N Hao et al
Scenario 1: negative regulation by feedback
We begin with the case of feedback desensitization. In
particular, we consider a model in which a downstream
kinase phosphorylates and desensitizes the upstream
receptor. Such phosphorylation-mediated negative feed-
back mechanisms do not require new protein synthesis
and therefore can act rapidly, sometimes within seconds
of pathway activation. Such a mechanism has recently
been shown to underlie short-term attenuation of
the high osmolarity pathway in yeast, mediated by
the osmosensor protein Sho1 and the MAPK Hog1
(Hao et al., 2007; Figure 1). Hog1 is required for cell
adaptation to osmotic stress conditions (Brewster et al.,
1993), and acts by promoting increased production of
glycerol that serves to equalize cellular osmotic pressure
with the external environment. Yeast mutants that
cannot produce or retain glycerol show diminished
viability after an osmotic shock, despite strongly
enhanced Hog1 phosphorylation. On the other hand,
constitutive activation of Hog1 leads to cell death.
Therefore, strict control over the dynamics of Hog1
activation is essential for cell survival (Hohmann, 2002;
Klipp et al., 2005).
Initially, computational models were devised based on
the duration and dose-dependence behavior of Hog1
activity. Given striking differences between pathway
activity in the absence and presence of Hog1 catalytic
activity, all of the models invoked some Hog1-depen-
dent phosphorylation event. A series of simple mathe-
matical models were fit to the data, and these models
suggested the existence of a key phosphorylation event
early in the pathway. Further experimental analysis
revealed that Hog1 phosphorylates Sho1, and also
established that Sho1 exists normally as a homo-
oligomer; moreover, mutation of the phosphorylation
site leads to diminished Sho1 oligomerization and
diminished signaling (Hao et al., 2007). These observa-
tions led to the model of pathway adaptation shown in
Figure 2a. In this model, Sho1 initiates a MAPK
cascade that leads to the activation of Hog1, and Hog1
in turn phosphorylates and inhibits Sho1. It is further
assumed that Sho1 exists as a homodimer and that
feedback phosphorylation causes the dimer to dissoci-
ate. Here, we have allowed for the phosphorylation step
to be reversible, so that the dephosphorylated Sho1 can
reassemble as a dimer and re-enter the signaling pool.
Theoretically speaking, the ability to desensitize the
pathway at the level of the ‘receptor’ has several
advantages for the system. First, if there were compo-
nents shared among multiple signaling pathways (e.g. a
common MAPKKK such as Ste11, Figure 1), desensi-
tization of the receptor would allow any shared
components to remain competent to transmit signals
from other receptors. Second, placing multiple pathway
components between the MAPK Hog1 and its target of
feedback regulation increases the sensitivity of the
system. This effect is illustrated in Figure 2b. The blue
curve in this figure represents the result when Hog1 is
responsible for feedback phosphorylation of the osmo-
sensor Sho1. The red and green curves are the results
when the MAPKKK Ste11 and the MAPKK Pbs2,
respectively, mediate the feedback phosphorylation of
Sho1. This increased sensitivity observed in Figure 2b
results from the delay between the feedback phosphory-
lation event early in the pathway and deactivation of the
MAPK late in the pathway. Such a delay would allow
the cell to adapt to strong signals yet remain sensitive
enough to detect weak signals. Thus, as multicomponent
signaling cascades are well known to confer signal
amplification, we postulate that multiple components
also allow cells to respond appropriately to a wide range
of signal strengths without the need for amplification
(Hao et al., 2007).
Scenario 2: negative regulation by feedback deactivation
Above we considered a scenario in which the regulated
pathway component is transiently removed from the
signaling pool. Next, we consider a scenario in which
negative feedback causes deactivation of a pathway
branch of the high osmolarity pathway in yeast. The Sho1
osmosensor is active as a dimer or higher order oligomer. Signaling
through Sho1 activates a MAPK kinase cascade that consists of
Ste11 (MAPKKK), Pbs2 (MAPKK) and Hog1 (MAPK). Feed-
back phosphorylation of Sho1 leads to desensitization and loss of
oligomerization. The phosphorylated Sho1 monomer can be
dephosphorylated allowing reformation of the oligomer. (b) Time
series of phospho-Hog1. This figure indicates that the system is
capable of near perfect adaptation. It also shows that using Hog1
(blue) as the negative regulator produces a larger response than if
the kinases Ste11 (red) or Pbs2 (green) were responsible for
Feedback desensitization. (a) A diagram of the Sho1
Systems biology analysis of yeast
N Hao et al
component, but the deactivated protein is immediately
available for reactivation. Figure 3a shows a model in
which the stimulus leads to activation of a MAPK. The
MAPK then phosphorylates and activates a phospha-
tase that in turn dephosphorylates the MAPK. At low
stimulus levels feedback dephosphorylation is sufficient
to restore MAPK activity to near prestimulus levels
(Figure 3b, blue curve). This system does not fully adapt
because at steady state a small increase in the level of
active MAPK is required to ensure there is enough
active phosphatase to counteract the stimulus. As the
stimulus level increases the level of active phosphatase
eventually saturates, and the MAPK response becomes
sustained (Figure 3b, green curve). If the activation
kinetics of the phosphatase are ultrasensitive (Goldbeter
and Koshland, 1981) with respect to the active MAPK
concentration, then the transition between a transient
signal and a sustained signal occurs in a switch-like
manner (Figure 3b, inset). For example, a transition
from transient to sustained signaling was proposed to
underlie MAPK signaling specificity in yeast (Sabbagh
et al., 2001).
Feedback desensitization appears to be more common
than feedback deactivation. There may be some inherent
advantages to the desensitization mechanism, as in this
case the sensitivity of the pathway could be increased by
making an upstream component of the pathway the
target of feedback regulation. However, for the case of
feedback deactivation targeting an upstream pathway
component diminishes the ability of the system to adapt
and makes it susceptible to damped or sustained
oscillations (Kholodenko, 2000). Although biochemical
oscillations are well documented in calcium signaling,
circadian clocks and cell cycle regulation, a biological
example of oscillations in MAPK signaling has not been
documented. One possible explanation of this failure to
observe oscillations is that it might require single-cell
measurements of the activation state of pathway
components, which is technically possible but rarely
Scenario 3: positive feedback delays pathway activation
and produces an irreversible switch
We next consider positive feedback mechanisms. Posi-
tive feedback can cause a delay in pathway activation, as
the stimulus needs only to produce a small increase in
pathway activity, and this initial increase is amplified by
the positive feedback eventually leading to full pathway
activation (Figure 4b, shaded region). One example of
such a delay was identified through analysis of the yeast
pheromone pathway (Hao et al., 2003). In this study,
functional effects of genetic perturbations on the path-
way were examined experimentally and compared with
predictions from a model of RGS and G-protein
activity. The initial computational model of the G
protein cycle was devised based on the published
literature. This model failed to capture key features of
phosphorylated MAPK (p-MAPK) activates a phosphatase
(P’ase), which in turn dephosphorylates the phospho-MAPK. (b)
Time series of the phospho-MAPK concentration as a function of
time. This figure illustrates that for low stimulus levels feedback
deactivation can lead to a transient signal (blue curve), whereas at
high stimulus levels the response becomes sustained (green curve).
Inset: A plot of the steady-state phospho-MAPK kinase concen-
tration (red curve) and the maximum amplitude of the response
(dashed magenta curve) as a function of the stimulus strength. The
blue and green dashed horizontal lines correspond to the stimulus
strengths used to generate the time series shown in this figure.
Feedback deactivation. (a) In this example, the
lated MAPK phosphorylates and activates a MAPKK. This system
produces an irreversible switch. (b) Time series of the activated
kinase concentration resulting from a square pulse stimulus
(shaded region). Initially the stimulus and the phosphorylation
levels are low. When the stimulus increases, the phosphorylation
level increases after a delay (see text) and remains elevated even
after the stimulus returns to its original low level.
Positive feedback. (a) In this example, the phosphory-
Systems biology analysis of yeast
N Hao et al
the experimental data, including the delay in activation
kinetics observed when the RGS protein Sst2 was
overexpressed. This suggested the existence of a positive
feedback mechanism that counteracts the negative
effects of Sst2. This cellular behavior was subsequently
found to result from pheromone-promoted ubiquitina-
tion and degradation of Sst2. This represents a striking
example of how model-driven experiments revealed new
mechanisms of cell regulation, which may not have been
found by intuitive means alone.
Another common feature of positive feedback systems
is bistability or ‘switch-like’ behavior. If the positive
feedback is sufficiently strong, it can maintain pathway
activity even after the stimulus is removed creating an
irreversible switch. This property is particularly impor-
tant for cell fate decisions such as cell division or
differentiation, where it is detrimental to abort the
process. Mathematically an irreversible switch occurs
when the model equations possess two stable steady
state solutions for a single stimulus level. This pheno-
menon, known as bistability, is demonstrated in
Figure 4b. At early times both the stimulus and
phospho-MAPK levels are low. When the stimulus level
is increased (shaded area in Figure 4b), the phospho-
MAPK concentration increases to a new steady state.
When the stimulus is removed the positive feedback is
able to maintain phospho-MAPK concentration in an
elevated state. Thus at low stimulus levels, the model
equations possess two stable steady states corresponding
to low and high phospho-MAPK concentrations.
The yeast mating response is one example of such a
cell fate decision process. Pheromone activation con-
verts yeast cells from a vegetatively growing state to a
growth-arrested state. Growth arrest in G1 and the
accompanying morphological changes are pseudo irre-
versible, as the cell will remain in the arrested state long
after the pheromone is removed. Another example of
such a cell fate decision process is Xenopus oocyte
maturation. In response to the steroid hormone
progesterone, oocytes will undergo either maturation
and arrest in metaphase of meiosis II or fail to mature
and remain arrested in a G2 state. They cannot stay in
an intermediate state. Moreover, once an oocyte has
matured, it will remain in the mature state even after the
progesterone is removed (Ferrell and Xiong, 2001,
Mathematical modeling has revealed that the dual
phosphorylation required for MAPK kinase activation
is sufficient to generate bistablility (Markevich et al.,
2004). That is, similar to the results shown in Figure 4,
there is a range of stimulus levels for which there exist
two stable concentrations of the dually phosphorylated
form of the MAPK. Although biologically this appears
to be a case of bistability without positive feedback,
mathematically it can be shown that a distributive
phosphorylation mechanism, which requires two colli-
sions between the MAPK and MAPK kinase, is
equivalent to positive feedback. This mechanism was
suggested to underlie the binary response observed in
the pheromone response pathway of yeast (Wang et al.,
Scenario 4: integral control
Perfect adaptation refers to the situation in which a
pathway component is only transiently activated despite
a persistent stimulus and eventually returns to exactly
basal levels. Figure 5a shows a simple model that
produces perfect adaptation. In this model a pathway
component (e.g. a receptor or protein kinase) is
constitutively expressed. The component can exist in
two states, the inactive state R and the active state R*.
The stimulus increases the rate at which the component
transitions to the active state. Subsequent to activation
the component becomes desensitized and is removed
from the signaling pool. It is important to note that in
this model only the active component can be desensi-
tized, desensitization is irreversible and the rate of
desensitization is independent of the stimulus strength.
Figure 5b shows a time series of the activated
strengths. The stimulus is given at time t¼0 and
remains constant for all later times. Note that for all
stimulus strengths, the active component concentration
returns to its basal levels. This mechanism of adaptation
is known as linear integral feedback control (Alon,
2006), because mathematically it can be shown that the
strength of the ‘feedback’ that restores the time-
dependent concentration of the activated component
R*(t) to its prestimulus level R*(0) is proportional to the
integral over time of the difference between R*(0) and
feedback control is robust in the sense that the system’s
0ðR ? ð0Þ ? R ? ðt0ÞÞdt0). Linear integral
component is produced at a constitutive rate and can exist in one
of two states, R (inactive) and R* (active). The presence of a
stimulus increases the rate at which the receptor transitions to the
active state. The active receptor is then subjected to desensitization
or degradation. (b) Time series of the active receptor concentration
for various agonist concentrations. This figure demonstrates the
system’s ability to perfectly adapt to a persistent stimulus.
Integral control. (a) In this example, a pathway
Systems biology analysis of yeast
N Hao et al
ability to adapt does not depend on the choice of model
parameter values (Alon et al., 1999; Alon, 2006).
This integral control mechanism was first proposed to
underlie models of bacterial chemotaxis (Alon et al.,
1999; Yi et al., 2000) and may contribute as well to
signal attenuation of the pheromone response pathway
of yeast. Following exposure to pheromone, the MAPK
kinase kinase Ste11 rapidly phosphorylates and acti-
vates the MAPK kinase Ste7 (Wang et al., 2003). Ste11
phosphorylation of Ste7 leads to ubiquitination and
increased degradation of the protein (Wang and
Dohlman, 2002; Wang et al., 2003). This mechanism
provides a foolproof mechanism for irreversibly deplet-
ing the cell of activated signaling components. Depletion
of Ste7 might also have detrimental effects, however,
because Ste7 is also a component of a second MAPK
pathway leading to invasive growth (Figure 1). Surpris-
ingly, following pheromone stimulation Ste7 levels are
roughly equal to their prestimulus levels, implying that
yeast have overcome this limitation by concomitant,
pheromone-induced production of Ste7. Thus, the net
levels of Ste7 remain largely constant, and accelerated
degradation is balanced by accelerated synthesis.
Depletion of the signaling pool is one limitation of
linear integral feedback control. Another potential
drawback is that the duration of the response is almost
entirely determined by the rate of desensitization or
degradation. Rapid signal attenuation requires rapid
turnover of the pathway component. This in turn
requires a relatively large constitutive synthesis rate
to produce a strong signal. These limitations can
be overcome by use of a feed-forward architecture
(described below) in which the stimulus activates a
parallel pathway that negatively regulates the signal.
Scenario 5: feed-forward regulation
A feed-forward architecture refers to a signaling net-
work in which the stimulus initiates parallel pathways
that converge on a downstream signaling component.
Feed-forward systems can be devised to behave as
coincidence detectors, relaying a signal only when both
pathways are active, thereby ensuring the pathway is not
activated by a spurious or highly transient stimulus
(Alon, 2006). These systems can also be designed to
relay a signal when only one of the pathways is active,
thereby guarding against a loss of responsiveness if one
pathway becomes temporarily inactive (Alon, 2006).
Additionally, if the two parallel pathways have anta-
gonistic effects, this provides another mechanism of
Feed-forward regulation may underlie the contribu-
tion of the scaffold protein Ste5 to pheromone responses
in yeast. Ste5 binds to multiple MAPK kinase compo-
nents including Ste11, Ste7 and Fus3 (Figure 1) and
binding to each protein is critical for transmitting signals
from membrane-bound G proteins to the cytoplasmic
pool of Fus3. The activation of MAPKs requires
the recruitment of Ste5-bound MAPK cascade to the
plasma membrane through Ste5 binding to Gbg. At the
same time, the p21-activated kinase (PAK) Ste20 is
activated by membrane-bound Cdc24 and Cdc42. Cdc24
acts as a guanine nucleotide exchange factor for Cdc42,
leading to the formation of Cdc42-GTP. Therefore,
coincident activation by Gbg and Cdc42-GTP are
required to fully activate the MAPK Fus3.
A feed-forward mechanism is likewise thought to
underlie transient ERK activation by epidermal growth
factor (EGF) (Sasagawa et al., 2005). Activation of the
EGF receptor results in the Son of sevenless homologue
protein son of sevenless (SOS) being recruited to the
plasma membrane where it acts as a guanine nucleotide
exchange factor for Ras, leading to the formation of
Ras-GTP. Ras-GTP initiates a cascade eventually lead-
ing to MAPK/ERK activation. EGF receptor activation
also results in the recruitment of the negative regulator
Ras-GAP to the plasma membrane through a slower
parallel pathway. In a recent study, Sasagawa et al.
(2005) used experimental analysis combined with
mathematical modeling to show that this feed-forward
mechanism most likely underlies the transient nature
of ERK activation by EGF. The fast activation is
responsible for the initial accumulation of Ras-GTP and
the consequent increase in ERK activation, whereas the
slow build up of Ras-GAP causes the signal to quickly
decay regardless the continuous presence of EGF. A
simplified version of this model is shown in Figure 6.
This system represents the simplest version of perfect
adaptation by a feed-forward mechanism (Tyson et al.,
Signal regulation by multiple control mechanisms
Most signaling pathways contain multiple regulatory
mechanisms and understanding how these mechanisms
are integrated to form a functional signaling system will
likely benefit from mathematical modeling. For exam-
ple, computational analysis of MAPK signaling revealed
that the positive feedback loop formed by the MAPK
and protein kinase C can generate bistability (Bhalla
et al., 2002). It also demonstrated that the negative effect
of a MAPK-induced increase in phosphatase expression
moves the signaling network from a bistable state, in
which a brief stimulus results in sustained MAPK
activation, to a monostable state in which the system
responds in a manner proportional to the stimulus.
These results were confirmed experimentally and high-
light the role of MAPK phosphatases in determining the
timing and duration of MAPK activation (Bhalla et al.,
2002). Induction of protein phosphatase expression has
likewise been proposed to play a critical role in MAPK
signaling in yeast. For example, activation of the
MAPKs Fus3 and Hog1 leads to enhanced expression
of protein phosphatases, such as the dual-specificity
phosphatase Msg5 (acts on Fus3) and the Tyr-specific
phosphatases Ptp2 and Ptp3 (Jacoby et al., 1997;
Wurgler-Murphy et al., 1997; Roberts et al., 2000).
We also note that adding a slow negative feedback
loop to a bistable system is a common mechanism for
generating oscillations. In such a system, when pathway
activity is high the negative regulator slowly builds until
it is able to drive the pathway back to a state of low
Systems biology analysis of yeast
N Hao et al
activity. This causes expression of the negative regulator
to decrease to near basal levels allowing the positive
feedback to reactivate the pathway, where the cycle
repeats. Such oscillations are referred to as hysteresis
oscillations (Tyson et al., 2003). It was recently shown
that when protein degradation is included in a model
for the dual phosphorylation cycle of MAPK kinase
activation hysteresis oscillations are possible (Wang
et al., 2006).
More complex computational models that attempt to
capture the behavior of entire signaling pathways are
also being developed. For example, Klipp et al. (2005)
have constructed a model of the high osmolarity
response pathway in yeast mediated by Hog1 (Hoh-
mann, 2002). Klipp et al. (2005) integrated modeling
and experimental approaches to investigate long-term
changes in response to osmotic shock, including closing
of glycerol channels (preventing the outflow of glycerol
upon stress stimulation) and induction of genes required
for the biosynthesis of glycerol. Their studies indicate
that gene induction is important for maintaining a high
level of glycerol production in adapted cells, and when
combined with rapid closure of glycerol channels is
sufficient to explain an initial glycerol accumulation
after osmotic shock. Similar efforts yielded comprehen-
sive models of the yeast pheromone pathway (Kofahl
and Klipp, 2004; Shao et al., 2006) and mammalian
epidermal growth factor (EGF) receptor pathways
(Schoeberl et al., 2002).
The pharmaceutical industry excels in many endeavors,
but has a mixed record of finding truly novel drugs or
new drug targets. Much effort is focused on developing
second-generation modulators of established targets,
and the improvements to health are often incremental at
best. The 7TMR class of receptors has been invaluable
in drug development. However, the targets for these
drugs have been known for 30 or more years. There is
now a growing effort to focus pharmacological research
on more recently discovered modulators and effectors
of 7TMR-controlled signal pathways. The expectation
is that pharmacological inhibitors of proteins that
modulate 7TMR signaling pathways could be used to
control clinically relevant events. Inhibitors that fail to
reach the clinic could still serve as valuable research
tools, allowing investigators to better define the role of
each target in signaling.
Candidate drug targets include arrestins, G proteins,
RGS proteins, MAPK components, protein phospha-
tases or components of the protein degradation machi-
nery. Such drugs could work in a number of ways, such
as by interfering with receptor or effector degradation,
RGS function, G protein GTP hydrolysis or signal
specificity through arrestins versus G proteins. RGS
proteins and protein kinases in particular hold tremen-
dous promise as drug targets. It is already established
that these proteins play a key role in desensitization, so
selective inhibitors could function to enhance the
responsiveness of 7TMRs (much like serotonin uptake
inhibitors or cGMP phosphodiesterase inhibitors) and
their downstream effectors. Finally, RGS proteins and
protein kinases appear to act through a variety of other
protein–protein (or protein–lipid) interactions mediated
by a diverse collection of interaction domains and
scaffold proteins; many of which remain to be fully
Modeling will be especially useful in situations where
it is impractical to systematically screen for new path-
way inhibitors. By constructing computational models,
anticipated outcomes of experimentally feasible pertur-
bations (e.g. pharmacological inhibition or selective
knockdown of gene expression) can be tested experi-
mentally and evaluated in the context of the model.
With successive rounds of experimentation, modeling
and experimental validation, the best mechanisms can
be investigated in detail. In this way modeling can serve
as a hypothesis generator, helping investigators to best
focus their efforts. Mathematical modeling is already
being used to understand the physiological changes that
occur during oncogenesis. Hornberg et al. (2005) applied
control analysis to a detailed model of EGF-induced
ERK activation. Their analysis revealed that of the 148
reactions considered in the model, only a small subset
receptor activates two parallel pathways. The stimulus causes a
rapid recruitment of the positive regulator SOS (GEF) to the
plasma membrane, which initiates signaling. This is the feed-
forward pathway. At the same time, the stimulus slowly activates
the negative regulator Ras-GAP, which terminates signaling. (b)
Time-series of Ras-GTP as function of time. This figure
demonstrates the system’s ability to perfectly adapt to a persistent
Feed-forward regulation. (a) Signaling through the EGF
Systems biology analysis of yeast
N Hao et al
significantly controlled the amplitude and duration of
ERK activation. In particular, the reactions that were
found to have the greatest effect on signal duration
and amplitude, were MEK phosphorylation by Raf and
Raf dephosphorylation. These results help explain why
Raf is such a potent oncogene.
Insights gained from modeling will also benefit
pharmaceutical research. By quantifying biological
systems and accurately predicting cellular processes, it
should eventually be possible to direct research into
novel pharmaceutical treatments for various diseases.
Thus, the use of computational simulations has the
potential to radically accelerate the development of
modern medicines. Accurate models of feedback regula-
tion in the well-characterized yeast system demonstrates
the validity of the approach, and will facilitate similar
efforts to model more complex biological pathways in
Alon U, Surette MG, Barkai N, Leibler S. (1999). Robustness
in bacterial chemotaxis. Nature 397: 168–171.
Alon U. (2006). An Introduction to Systems Biology. Chapman
& Hall: London ISBN 1-58488-642-0.
Apanovitch DM, Iiri T, Karasawa T, Bourne HR, Dohlman HG.
(1998a). Second site suppressor mutations of a GTPase-
deficient G-protein a- subunit. Selective inhibition of
bg-mediated signaling. J Biol Chem 273: 28597–28602.
Apanovitch DM, Slep KC, Sigler PB, Dohlman HG. (1998b).
Sst2 is a GTPase-activating protein for Gpa1: purification
and characterization of a cognate RGS-Ga protein pair in
yeast. Biochemistry 37: 4815–4822.
Barnes DJ, Melo JV. (2003). Management of chronic myeloid
leukemia: targets for molecular therapy. Semin Hematol 40:
Bhalla US, Ram PT, Iyengar R. (2002). MAP kinase
phosphatase as a locus of flexibility in a mitogen-activated
protein kinase signaling network. Science 297: 1018–1023.
Bornheimer SJ, Maurya MR, Farquhar MG, Subramaniam S.
(2004). Computational modeling reveals how interplay
between components of
regulates signal transduction. Proc Natl Acad Sci USA
Brewster JL, de Valoir T, Dwyer ND, Winter E, Gustin MC.
(1993). An osmosensing signal transduction pathway in
yeast. Science 259: 1760–1763.
Chan RK, Otte CA. (1982a). Isolation and genetic analysis of
Saccharomyces cerevisiae mutants supersensitive to G1
arrest by a factor and a factor pheromones. Mol Cell Biol
Chan RK, Otte CA. (1982b). Physiological characterization of
Saccharomyces cerevisiae mutants supersensitive to G1
arrest by a factor and a factor pheromones. Mol Cell Biol
Chasse SA, Dohlman HG. (2003). RGS proteins: G protein-
coupled receptors meet their match. Assay Drug Dev Technol
Chasse SA, Flanary P, Parnell SC, Hao N, Cha JY, Siderovski
DP et al. (2006). Genome-scale analysis reveals Sst2 as the
principal regulator of mating pheromone signaling in the
yeast Saccharomyces cerevisiae. Eukaryot Cell 5: 330–346.
Dohlman HG, Thorner JW. (2001). Regulation of G protein-
initiated signal transduction in yeast: Paradigms and
principles. Annu Rev Biochem 70: 703–754.
Dohlman HG, Apaniesk D, Chen Y, Song J, Nusskern D.
(1995). Inhibition of G-protein signaling by dominant gain-
of-function mutations in Sst2p, a pheromone desensitization
factor in Saccharomyces cerevisiae. Mol Cell Biol 15:
Dohlman HG, Song J, Ma D, Courchesne WE, Thorner J.
(1996). Sst2, a negative regulator of pheromone signaling in
the yeast Saccharomyces cerevisiae: expression, localization,
and genetic interaction and physical association with Gpa1
(the G-protein a subunit). Mol Cell Biol 16: 5194–5209.
Dohlman HG. (2002). Diminishing returns. Nature 418: 591.
Farfel Z, Bourne HR, Iiri T. (1999). The expanding spectrum
of G protein diseases. N Engl J Med 340: 1012–1020.
Ferrell JE, Xiong W. (2001). Bistability in cell signaling: how
to make continuous processes discontinuous, and reversible
processes irreversible? Chaos 11: 227–236.
Force T, Kuida K, Namchuk M, Parang K, Kyriakis JM.
(2004). Inhibitors of protein kinase signaling pathways:
emerging therapies for cardiovascular disease. Circulation
Garrison TR, Zhang Y, Pausch M, Apanovitch D, Aebersold R,
Dohlman HG. (1999). Feedback phosphorylation of an
RGS protein by MAP kinase in yeast. J Biol Chem 274:
Gelperin DM, White MA, Wilkinson ML, Kon Y, Kung LA,
Wise KJ et al. (2005). Biochemical and genetic analysis of
the yeast proteome with a movable ORF collection. Genes
Dev 19: 2816–2826.
Ghaemmaghami S, Huh WK, Bower K, Howson RW, Belle A,
Dephoure N et al. (2003). Global analysis of protein
expression in yeast. Nature 425: 737–741.
Gibbs JB, Oliff A. (1994). Pharmaceutical research in
molecular oncology. Cell 79: 193–198.
Goffeau A, Barrell BG, Bussey H, Davis RW, Dujon B,
Feldmann H et al. (1996). Life with 6000 genes [see
comments]. Science 274: 546 563-7.
Gold SJ, Ni YG, Dohlman HG, Nestler EJ. (1997). Regulators
of G-protein signaling (RGS) proteins: region-specific
expression of nine subtypes in rat brain. J Neurosci 17:
Goldbeter A, Koshland Jr DE. (1981). An amplified sensitivity
arising from covalent modification in biological systems.
Proc Natl Acad Sci USA 78: 6840–6844.
Guo M, Aston C, Burchett SA, Dyke C, Fields S, Rajarao SJ
et al. (2003). The yeast G protein a subunit Gpa1 transmits a
signal through an RNA binding effector protein Scp160.
Mol Cell 12: 517–524.
Gutkind JS. (1998). The pathways connecting G protein-
coupled receptors to the nucleus through divergent mitogen-
activated protein kinase cascades. J Biol Chem 273:
Hao N, Behar M, Parnell SC, Torres MP, Borchers CH,
Elston TC et al. (2007). Experimental and computational
analysis of feedback inhibition of the Sho1 osmotic stress-
response pathway. Curr Biol (in press).
Hao N, Yildirim N, Wang Y, Elston TC, Dohlman HG.
(2003). Regulators of G protein signaling and transient
activation of signaling: experimental and computational
analysis reveals negative and positive feedback controls on
G protein activity. J Biol Chem 278: 46506–46515.
Systems biology analysis of yeast
N Hao et al
Hardman JG, Limbird LE, Gilman AG. (2001). Goodman &
McGraw-Hill: New York ISBN 0-07-135469-7.
Hartwell LH, Szankasi P, Roberts CJ, Murray AW, Friend SH.
(1997). Integrating genetic approaches into the discovery of
anticancer drugs. Science 278: 1064–1068.
Heinrich R, Neel BG, Rapoport TA. (2002). Mathematical
models of protein kinase signal transduction. Mol Cell 9:
Hohmann S. (2002). Osmotic stress signaling and osmoadap-
tation in yeasts. Microbiol Mol Biol Rev 66: 300–372.
Hornberg JJ, Bruggeman FJ, Binder B, Geest CR, de Vaate AJ,
Lankelma J et al. (2005). Principles behind the multifarious
control of signal transduction. ERK phosphorylation and
kinase/phosphatase control. FEBS J 272: 244–258.
Huang CY, Ferrell Jr JE. (1996). Ultrasensitivity in the
mitogen-activated protein kinase cascade. Proc Natl Acad
Sci USA 93: 10078–10083.
Huh WK, Falvo JV, Gerke LC, Carroll AS, Howson RW,
Weissman JS et al. (2003). Global analysis of protein
localization in budding yeast. Nature 425: 686–691.
Jacoby T, Flanagan H, Faykin A, Seto AG, Mattison C, Ota I.
(1997). Two protein-tyrosine phosphatases inactivate the
osmotic stress response pathway in yeast by targeting the
mitogen-activated protein kinase, Hog1. J Biol Chem 272:
Johnson GL, Dohlman HG, Graves LM. (2005). MAPK
kinase kinases (MKKKs) as a target class for small-molecule
inhibition to modulate signaling networks and gene expres-
sion. Curr Opin Chem Biol 9: 325–331.
Kenakin T. (2004). Principles: receptor theory in pharma-
cology. Trends Pharmacol Sci 25: 186–192.
Kholodenko BN. (2000). Negative feedback and ultrasensiti-
vity can bring about oscillations in the mitogen-activated
protein kinase cascades. Eur J Biochem 267: 1583–1588.
Klipp E, Nordlander B, Kruger R, Gennemark P, Hohmann S.
(2005). Integrative model of the response of yeast to osmotic
shock. Nat Biotechnol 23: 975–982.
Kofahl B, Klipp E. (2004). Modelling the dynamics of the
yeast pheromone pathway. Yeast 21: 831–850.
Komarova NL, Zou X, Nie Q, Bardwell L. (2005). A
theoretical framework for specificity in cell signaling. Mol
Syst Biol 1: 0023.
Kumar S, Boehm J, Lee JC. (2003). p38 MAP kinases: key
signalling molecules as therapeutic targets for inflammatory
diseases. Nat Rev Drug Discov 2: 717–726.
Lin FT, Miller WE, Luttrell LM, Lefkowitz RJ. (1999).
Feedback regulation of beta-arrestin1 function by extra-
cellular signal-regulated kinases [in process citation]. J Biol
Chem 274: 15971–15974.
Marinissen MJ, GutkindJS. (2001). G-protein-coupled
receptors and signaling networks: emerging paradigms.
Trends Pharmacol Sci 22: 368–376.
Markevich NI, Hoek JB, Kholodenko BN. (2004). Signaling
switches and bistability arising from multisite phosphoryla-
tion in protein kinase cascades. J Cell Biol 164: 353–359.
Marshall CJ. (1995). Specificity of receptor tyrosine kinase
signaling: transient versus sustained extracellular signal-
regulated kinase activation. Cell 80: 179–185.
Martin H, Flandez M, Nombela C, Molina M. (2005). Protein
phosphatases in MAPK signalling: we keep learning from
yeast. Mol Microbiol 58: 6–16.
Martzen MR, McCraith SM, Spinelli SL, Torres FM, Fields S,
Grayhack EJ et al. (1999). A biochemical genomics
approach for identifying genes by the activity of their
products. Science 286: 1153–1155.
Morrison DK, Davis RJ. (2003). Regulation of MAP kinase
signaling modules by scaffold proteins in mammals. Annu
Rev Cell Dev Biol 19: 91–118.
Neubig RR, Siderovski DP. (2002). Regulators of G-protein
signalling as new central nervous system drug targets. Nat
Rev Drug Discov 1: 187–197.
Neves SR, Ram PT, Iyengar R. (2002). G protein pathways.
Science 296: 1636–1639.
Ogier-Denis E, Pattingre S, El Benna J, Codogno P. (2000).
Erk1/2-dependent phosphorylation of Galpha-interacting
protein stimulates its GTPase accelerating activity and
autophagy in human colon cancer cells. J Biol Chem 275:
Parnell SC, Marotti Jr LA, Kiang L, Torres MP, Borchers CH,
Dohlman HG. (2005). Phosphorylation of the RGS protein
Sst2 by the MAP kinase Fus3 and use of Sst2 as a model to
analyze determinants of substrate sequence specificity.
Biochemistry 44: 8159–8166.
Pitcher JA, Tesmer JJ, Freeman JL, Capel WD, Stone WC,
protein-coupled receptor kinase 2 (GRK2) activity by
extracellular signal-regulated kinases. J Biol Chem 274:
Roberts CJ, Nelson B, Marton MJ, Stoughton R, Meyer MR,
Bennett HA et al. (2000). Signaling and circuitry of multiple
MAPK pathways revealed by a matrix of global gene
expression profiles. Science 287: 873–880.
Roman DL, Talbot JN, Roof RA, Sunahara RK, Traynor JR,
Neubig RR. (2006). Identification of small molecule
inhibitors of Regulator of G-protein Signaling 4 (RGS4)
using a high throughput flow cytometry protein interaction
assay (FCPIA). Mol Pharmacol 71: 169–175.
Sabbagh Jr W, Flatauer LJ, Bardwell AJ, Bardwell L. (2001).
Specificity of MAP kinase signaling in yeast differentiation
involves transient versus sustained MAPK activation. Mol
Cell 8: 683–691.
Sasagawa S, Ozaki Y, Fujita K, Kuroda S. (2005). Prediction
and validation of the distinct dynamics of transient and
sustained ERK activation. Nat Cell Biol 7: 365–373.
Schaber J, Kofahl B, Kowald A, Klipp E. (2006). A modelling
approach to quantify dynamic crosstalk between the
pheromone and the starvation pathway in baker’s yeast.
FEBS J 273: 3520–3533.
Schoeberl B, Eichler-Jonsson C, Gilles ED, Muller G. (2002).
Computational modeling of the dynamics of the MAP
kinase cascade activated by surface and internalized EGF
receptors. Nat Biotechnol 20: 370–375.
Sebolt-Leopold JS, English JM. (2006). Mechanisms of drug
inhibition of signalling molecules. Nature 441: 457–462.
Shao D, Zheng W, Qiu W, Ouyang Q, Tang C. (2006).
Dynamic studies of scaffold-dependent mating pathway in
yeast. Biophys J 91: 3986–4001.
Shenoy SK, Lefkowitz RJ. (2005). Seven-transmembrane
receptor signaling through beta-arrestin. Sci STKE 2005:
Slessareva JE, Routt SM, Temple B, Bankaitis VA, Dohlman
HG. (2006). Activation of the phosphatidylinositol 3-kinase
Vps34 by a G protein alpha subunit at the endosome. Cell
Somsen OJ, Siderius M, Bauer FF, Snoep JL, Westerhoff HV.
(2002). Selectivity in overlapping MAP kinase cascades.
J Theor Biol 218: 343–354.
Spiegel AM. (2000). G protein defects in signal transduction.
Horm Res 53: 17–22.
Sprang SR. (1997). G protein mechanisms: insights from
structural analysis. Annu Rev Biochem 66: 639–678.
Systems biology analysis of yeast
N Hao et al
Tyson JJ, Chen KC, Novak B. (2003). Sniffers, buzzers, Download full-text
toggles and blinkers: dynamics of regulatory and signaling
pathways in the cell. Curr Opin Cell Biol 15: 221–231.
Wang X, Hao N, Dohlman HG, Elston TC. (2006). Bistability,
stochasticity, and oscillations in the mitogen-activated
protein kinase cascade. Biophys J 90: 1961–1978.
Wang Y, Dohlman HG. (2002). Pheromone-dependent ubi-
quitination of the mitogen-activated protein kinase kinase
Ste7. J Biol Chem 277: 15766–15772.
Wang Y, Ge Q, Houston D, Thorner J, Errede B, Dohlman
HG. (2003). Regulation of Ste7 ubiquitination by Ste11
phosphorylation and the Skp1-Cullin-F-box complex. J Biol
Chem 278: 22284–22289.
Winzeler EA, Shoemaker DD, Astromoff A, Liang H,
Anderson K, Andre B et al. (1999). Functional characteri-
zation of the S. cerevisiae genome by gene deletion and
parallel analysis. Science 285: 901–906.
Wurgler-Murphy SM, Maeda T, Witten EA, Saito H. (1997).
Regulation of the Saccharomyces cerevisiae HOG1 mitogen-
activated protein kinase by the PTP2 and PTP3 protein
tyrosine phosphatases. Mol Cell Biol 17: 1289–1297.
Xiong W, Ferrell Jr JE. (2003). A positive-feedback-based
bistable ‘memory module’ that governs a cell fate decision.
Nature 426: 460–465.
Yi TM, Huang Y, Simon MI, Doyle J. (2000). Robust perfect
adaptation in bacterial chemotaxis through integral feed-
back control. Proc Natl Acad Sci USA 97: 4649–4653.
Yi TM, Kitano H, Simon MI. (2003). A quantitative
characterization of the yeast heterotrimeric G protein cycle.
Proc Natl Acad Sci USA 100: 10764–10769.
Zhong H, Wade SM, Woolf PJ, Linderman JJ, Traynor JR,
Neubig RR. (2003). A spatial focusing model for G protein
signals. Regulator of G protein signaling (RGS) protien-
mediated kinetic scaffolding. J Biol Chem 278: 7278–7284.
Systems biology analysis of yeast
N Hao et al