A System of Counteracting Feedback Loops Regulates Cdc42p Activity during Spontaneous Cell Polarization

Abstract

Cellular polarization is often a response to distinct extracellular or intracellular cues, such as nutrient gradients or cortical landmarks. However, in the absence of such cues, some cells can still select a polarization axis at random. Positive feedback loops promoting localized activation of the GTPase Cdc42p are central to this process in budding yeast. Here, we explore spontaneous polarization during bud site selection in mutant yeast cells that lack functional landmarks. We find that these cells do not select a single random polarization axis, but continuously change this axis during the G1 phase of the cell cycle. This is reflected in traveling waves of activated Cdc42p which randomly explore the cell periphery. Our integrated computational and in vivo analyses of these waves reveal a negative feedback loop that competes with the aforementioned positive feedback loops to regulate Cdc42p activity and confer dynamic responsiveness on the robust initiation of cell polarization.

Full text

Developmental Cell, Vol. 9, 565–571, October, 2005, Copyright ©2005 by Elsevier Inc. DOI 10.1016/j.devcel.2005.08.014
Short ArticleA System of Counteracting
Feedback Loops Regulates Cdc42p
Activity during Spontaneous Cell Polarization
Ertugrul M. Ozbudak,
1
Attila Becskei,
1
and Alexander van Oudenaarden
1,
*
Department of Physics
Massachusetts Institute of Technology
Cambridge, Massachusetts 02139
Summary
Cellular polarization is often a response to distinct
extracellular or intracellular cues, such as nutrient
gradients or cortical landmarks. However, in the ab-
sence of such cues, some cells can still select a po-
larization axis at random. Positive feedback loops
promoting localized activation of the GTPase Cdc42p
are central to this process in budding yeast. Here, we
explore spontaneous polarization during bud site se-
lection in mutant yeast cells that lack functional land-
marks. We find that these cells do not select a single
random polarization axis, but continuously change
this axis during the G1 phase of the cell cycle. This
is reflected in traveling waves of activated Cdc42p
which randomly explore the cell periphery. Our inte-
grated computational and in vivo analyses of these
waves reveal a negative feedback loop that competes
with the aforementioned positive feedback loops to
regulate Cdc42p activity and confer dynamic respon-
siveness on the robust initiation of cell polarization.
Introduction
Establishment of cell polarity is an essential process
both in uni- and multicellular organisms during cell divi-
sion, chemotaxis, differentiation, morphogenesis, and
cell migration (Nelson, 2003). In general, the axis of cell
polarity is selected either by cell extrinsic stimuli or in-
trinsic cues. The role of intrinsic cues is exemplified by
the selection of the bud site during the cell division cy-
cle of the yeast Saccharomyces cerevisiae. In this case,
a new bud forms where landmark proteins were depos-
ited during the previous cell division cycle. The Ras
GTPase Rsr1p (Bud1p) landmark protein recruits the
guanosine-nucleotide exchange factor (GEF) Cdc24p
to the plasma membrane (Toenjes et al., 2004), resulting
in a local activation of Cdc42p (Kozminski et al., 2003;
Park et al., 1997). Subsequently, activated Cdc42p re-
cruits its effectors forming a crescent-shaped polar cap
on the cell periphery, which orchestrates rearrange-
ments of the actin cytoskeleton and polarized secretion
to initiate budding. Cells lacking functional landmarks
initiate budding at a random location (Chant, 1999)by
spontaneous symmetry breaking (Wedlich-Soldner et
al., 2003, 2004; Irazoqui et al., 2003).
This reflects the efficiency of polarization mecha-
nisms to amplify minute asymmetries in the absence of
any spatial cues. In general, this may be accomplished
*Correspondence: avano@mit.edu
1
Lab address: http://web.mit.edu/biophysics
by positive feedback regulation of a polarization activa-
tor that is initially randomly distributed at the plasma
membrane. Theoretical studies show that if activation
requires a limiting diffusible substrate, an initially random
activator distribution has the potential to self-organize
into a highly polarized distribution characterized by a
single, stable, and randomly oriented polarization axis
(Gierer and Meinhardt, 1972). Indeed, recent experi-
mental studies suggest that the regulation of Cdc42p
activity is part of a positive feedback loop (Wedlich-
Soldner et al., 2003; Irazoqui et al., 2003). Potential mo-
lecular candidates for this positive feedback include
actin-mediated transport of secretory vesicles contain-
ing Cdc42p to sites of enhanced activity (Wedlich-
Soldner et al., 2003) and a Cdc42p-induced local poly-
merization of scaffolding proteins at the plasma mem-
brane that enhances the local Cdc42p recruitment rate
(Irazoqui et al., 2003). This symmetry breaking requires
the scaffold protein Bem1p (Irazoqui et al., 2003) and
the cycling of Cdc42p between GTP and GDP-bound
forms (Caviston et al., 2002), which is catalyzed by the
Cdc24p GEF and various Cdc42p-specific GTPase acti-
vating proteins (GAPs) including Rga1p and Bem3p.
Here, we study the dynamics of this symmetry break-
ing in yeast cells with RSR1 deleted. In these cells lack-
ing functional landmarks, we monitor the pool of acti-
vated Cdc42p at the plasma membrane. We find that
cells in the early G1 phase randomly select a polariza-
tion axis. However, this axis is not static but gradually
wanders during the rest of the G1 phase. This is experi-
mentally reflected in traveling activity waves of Cdc42p
that often traverse the circumference of a cell more
than once before budding. This dynamic behavior can-
not be explained by the positive feedback regulation of
Cdc42p activity alone. We propose that negative feed-
back regulation of Cdc42p activity mediated through
the actin cytoskeleton in parallel to a positive feedback
provides the drive of the traveling activity waves. Con-
sistently, the wave mobility is significantly reduced
when the actin cytoskeleton is depolymerized or when
the concentration of Cdc42p activity inhibitors (GAPs)
is reduced.
Results and Discussion
Cdc42 Activity Is Organized into Spatial Waves
In order to monitor the polarization dynamics of acti-
vated Cdc42p, a CRIB
Gic2
-GFP fusion protein was mon-
itored using fluorescence microscopy. The CRIB do-
main of Gic2p, CRIB
Gic2
, binds to the GTP-bound form
of Cdc42p (Burbelo et al., 1995). This probe is expected
to respond rapidly to changes in the Cdc42p-GTP loca-
tion because photobleached areas of polar caps re-
covered their initial fluorescence intensity in less than
one second (Figure S1). This response is considerably
faster than the exchange rate of Cdc42p between the
polar cap and the cytoplasm (Wedlich-Soldner et al.,
2004). Time-lapse experiments showed that, in wild-
type cells, the location of the polar cap did not change

Developmental Cell
566
Figure 1. Dynamics of the Polar Cap Movement
(A) Gic2p
(1–208)
-GFP is used as a reporter for activated Cdc42p. Localized fluorescence signal in wild-type (strain ERT224.1; upper panel) or
rsr1⌬(strain ERT225.1; lower panel) cells. Time is reported in minutes.
(B) Bem1p-CFP localization is shown in an rsr1⌬(strain ERT273.2) cell. Time is reported in minutes.
(C) Time evolution of the polarization angle in a wild-type (open circles) or rsr1⌬(solid red circles) cell. The polarization angle is defined as
the angle difference between the current and incipient polarization site.
after the initial establishment of cellular polarity, al-
though the cap displayed small displacements. Eventu-
ally, budding started adjacent to the previous division
site (Figure 1A, upper panel; Movie S1). Similar beha-
vior was observed in bem1Dcells. This is expected
because in these cells the budding landmark is fully
functional and cells bud in a nonrandom, monopolar
fashion.
If dynamic polarization in the absence of landmark
protein Rsr1p (Bud1p) is solely determined by autoam-
plification of fluctuations in Cdc42p regulators, we ex-
pected that a polar cap would be initiated at a random
location followed by budding at this same location. In-
deed, the polar cap formed at a single site, but contrary
to the above expectations, it started to move around
the cell periphery immediately after symmetry breaking.
The direction of movement changed at random time in-
tervals. Occasionally, polar caps traveled over the en-
tire cell perimeter before settling down at a random
location to initiate budding (Figure 1A, lower panel;
Movies S2 and S3). Similar behavior was observed
when the location of the polar cap components was
monitored by Bem1p-CFP (Figure 1B), Gic2p-GFP, and
CRIB
Cla4
-GFP (data not shown). The interval between
the appearance of the polar cap and the inception of
budding at room temperature is (60 ± 13) min (N= 13).
We find that the wandering stops about 15–20 min be-
fore we can detect the appearance of the bud in the
phase contrast image. The motion of traveling polar
caps was quantified by their angular displacement with
respect to the initial polarization site at the start of the
experiment. Analysis of the polarization angle indicates
that there is a significant difference in mobility of the
polar caps between wild-type and rsr1⌬cells over the
entire G1 stage of the cell cycle (Figure 1C).
The mobility of the traveling activity waves was quan-
tified by plotting the mean square angular displace-
ment of polar caps q
RMS
2
as a function of time, averaged
over multiple cells (Saxton and Jacobson, 1997). The
linear dependence of q
RMS
2
with time for rsr1Dcells re-
veals that traveling waves perform a random uncon-
strained walk at the cell periphery (Figure 2A). Deletion
of other landmark proteins, such as Bud2p and Bud5p,
also resulted in the appearance of traveling waves. The
mobility of the polar caps as measured by the slope of
the q
RMS
2
-time relation showed similar values for rsr1⌬,
bud2⌬, and bud5⌬cells (Figure 3A). This indicates that
cycling between the GTP and GDP-bound state of
Rsr1p is necessary to prevent the polar cap from travel-
ing, because cells expressing constitutively active and
inactive Rsr1p (bud2⌬and bud5⌬cells, respectively)
display highly mobile waves.
Dosage Dependence of Wave Mobility
on Localizing Factors
The above results suggest the existence of pattern-
destabilizing processes, which become dominant in the
absence of localizing factors such as Rsr1p. Next, we
reduced the Rsr1p expression level below the wild-type
level to attain an intermediate level of polarizing strength
and to moderate the effect of pattern destabilization.
For this purpose, the weak P
IXR1
and P
GLN3
promoters
(Holland, 2002) were used to drive to expression of
RSR1. The mobility of the polar cap was reduced in
comparison to rsr1⌬cells as the concentration of Rsr1p
increased (Figure 2A). When RSR1 is driven by the weak
P
GLN3
promoter, the increase of q
RMS
2
slowed down with
time, indicative of confined random motion of the polar
cap (Saxton and Jacobson, 1997). This is consistent
with the model that large fluctuations in pattern desta-
bilization cause the polar cap to escape from the area
of activated Rsr1p localization at low expression level.
However, on average, the polar cap spends more time
in the area of landmark proteins. In order to examine
how the wave mobility affects bud site selection, we
measured the budding angle. The budding angle re-
flects the displacement of the settled, late G1 localiza-
tion of the polar cap with respect to the previous bud-
ding site (see Experimental Procedures). The average
budding angle correlates nonlinearly with the wave mo-

Counteracting Feedback Loops Control Cdc42 Activity
567
Figure 2. The Root Mean Square of the Angular Displacement θ
RMS
of the Polar Cap Quantifies Wave Mobility
θ
2RMS
is plotted as a function of time with respect to the initial polarization angle at the start of the observation. θ
2RMS
was obtained
by averaging over at least ten different cells. The average slope, obtained by a linear least-square fit, of θ
2RMS
-time relation defines the
wave mobility.
(A) Dosage-dependent effect of RSR1 expression on the motility of the polar cap shown for wild-type (strain ERT224.1; black squares), rsr1⌬
(strain ERT225.1; red circles), rsr1⌬/P
GLN3
RSR1 (strain ERT275.2; green triangles), and rsr1⌬/P
IXR1
RSR1 (strain ERT276.3; green triangles)
cells.
(B) Dosage-dependent effect of GAP expression on the motility of the polar cap for wild-type (strain ERT224.1; black squares), rsr1⌬(strain
ERT225.1; red circles), rsr1⌬bem3⌬/P
POL30
RGA1 (strain ERT282.3; blue triangles), rsr1⌬bem3⌬(strain ERT254.1; purple stars), and rsr1⌬
rga1⌬(strain ERT255.1; green triangles) cells.
bility (Figure 3B), revealing that confinement of waves
results in a more deterministic bud site selection.
Regulators of Cdc42p GTPase Activity Determine
the Mobility of Waves
Whereas landmark proteins localize the traveling waves,
activities that enhance the mobility of the waves are
likely to be related to restructuring of the polar cap.
Cycling of Cdc42p between GTP and GDP-bound forms
is crucial for recruitment of polar cap components (Ira-
zoqui et al., 2003, 2004). Inhomogeneous recruitment
of Cdc42p regulators within the polar cap may lead to
a local decrease in the Cdc42p turnover and, conse-
quently, the net dissociation rate of polar cap compo-
nents would increase. As a result, the restructuring and
mobility of the polar cap would be reduced. The spatial
distribution of Cdc42p and its regulators bound to se-
cretory or endocytotic vesicles is in part dependent on
actin-based transport (Wedlich-Soldner et al., 2004; Ira-
zoqui et al., 2005). Therefore, we hypothesized that im-
pairment of actin polymerization or changing the turn-
over of the Cdc42p GTPase cycle would decrease the
wave mobility. First, inhibition of actin polymerization
by latrunculin A abolished wave mobility in rsr1⌬cells
(Figure 3A). While initial pattern formation is indepen-
dent of actin even in the absence of landmark proteins
(Ayscough et al., 1997; Irazoqui et al., 2003;Figure S2),
the mobility and the dynamic nature of polar caps is ac-
tin-dependent. The wave mobility in latrunculin A-treated
rsr1⌬cells is very similar to that observed for wild-type
cells. We do not observe the alternating appearance
and disappearance of the polar cap as was observed
when Cdc42p-GFP was used as a reporter in latrun-
culin A-treated cells (Wedlich-Soldner et al., 2004). This
difference might be attributed to use of a different re-
porter or different expression levels of Cdc42p or Rsr1p
in these strains. Second, the turnover of the Cdc42p
GTPase cycle was impaired by deleting BEM3, a gene
encoding a GAP protein for Cdc42p. This deletion re-
duced the wave mobility substantially in rsr1⌬cells
(Figures 2B and 3A). To verify whether this phenotype
was specific to the deletion of BEM3 or whether it re-
flects the reduction of GAP activity in general, another
GAP gene, RGA1, was deleted. The effect of this dele-
tion was comparable to that observed for the BEM3
deletion. In addition, overexpression of Rga1p in
rsr1⌬bem3⌬cells restored the mobility of the polar cap
(Figures 2B and 3A). When BEM3 is overexpressed in
rsr1⌬cells, the wave mobility does not change appreci-
ably, indicating that the wave mobility in rsr1⌬cells is
close to its maximum saturated value (Figure 3A). This
might indicate that there is another limiting factor that
determines the maximum recruitment rate of the GAPs
to the polarization site. These data suggest that wave
mobility depends on the dosage of GAP activity. Third,
overexpression of a cell membrane-targeted form of
the GEF Cdc24p also resulted in a decreased mobility
(Figure 3A) of a more widely distributed polar cap (Sup-
plemental Data).
A Model of Feedback Regulation of Cdc42p Activity
Explains Wave Formation
The formation of Cdc42p activity waves can be mod-
eled in terms of a feedback control of activators and

Developmental Cell
568
Figure 3. Regulation of Wave Mobility and Effect on the Budding Angle
(A) Wave mobility (units of deg
2
/min) of the polar cap in cells having different genetic backgrounds. The same strains were used as described
in Figure 2. The last two bars were obtained from cells overexpressing myrG2A-CDC24 (strain ERT236.1) and myr-CDC24 (strain ERT237.1).
(B) The wave mobility nonlinearly correlates with the average budding angle of cells expressing varying amounts of landmark proteins. In
wild-type cells, new buds are excluded from the previous budding sites. This results in a minimum budding angle of around 40° for wild-type
cells. An angle of about 120° corresponds to random budding.
inhibitors of Cdc42p (Figure 4A; Supplemental Data).
Interactions between Bem1p, Cdc24p, and Cdc42p are
suggested to form an actin-independent, positive feed-
back loop that enhances the recruitment of activators
to the polar cap (Bose et al., 2001; Butty et al., 2002;
Irazoqui et al., 2003; Wedlich-Soldner et al., 2003, 2004;
Shimada et al., 2004). We hypothesize that this positive
feedback loop accounts for the initial symmetry break-
ing (Figure S2). A second, actin-mediated, positive
feedback could be mediated by the actin-based de-
livery of secretory vesicles containing, for example,
Cdc42p or Cdc24p (Wedlich-Soldner et al., 2003, 2004).
Indeed, we sometimes observe faint Gic2p
1–208
-GFP-
labeled dots moving toward the polar cap (Figure S4).
Positive feedback loops allow an initially homoge-
nous system to polarize in a random static direction
(Gierer and Meinhardt, 1972). Because there is a limited
number of activated Cdc42p molecules distributed
along the membrane, a uniform distribution will still ex-
hibit small local concentration deviations about the
average. Due to the positive feedback regulation, local
concentration maximums will grow at the expense of
the surrounding areas in the membrane. This mecha-
nism can explain the initial symmetry breaking in, for
example, latrunculin-treated cells (Figure 4B, middle
panel), but cannot explain the traveling Cdc42p activ-
ity waves.
We therefore propose that in addition to the positive
regulation, a negative feedback loop must exist. Be-
cause we do not observe waves in latrunculin-treated
cells, it is likely that the negative feedback loop is medi-
ated by the actin cytoskeleton. A potential molecular
implementation of the negative loop might be the deliv-
ery of vesicles containing GAPs along the actin cables.
Alternatively, recent experiments show that actin-
dependent endocytosis might disperse polarizing fac-
tors (Irazoqui et al., 2005). Because actin patch compo-
nents (associated with endocytosis) can be recruited
by Cdc42p-GTP (Lechler et al., 2000), this indicates that
Cdc42p could stimulate dispersal of polarized factors
in an actin-dependent manner, effectively establishing
a negative feedback loop. The negative feedback loop
provides the system with the potential to exhibit travel-
ing waves (Meinhardt, 1999). To observe traveling waves,
it is important that the negative feedback regulation op-
erates at a slower rate than the positive regulation. We
propose that the actin-independent positive feedback
reacts faster to a change in Cdc42p activation state
than the regulation mediated by the actin cytoskeleton.
This is a reasonable assumption because nucleation
and polymerization of new actin cables followed by
vesicle transport might be a slower process than the
formation of the Bem1p scaffolding complex that relies
on relatively fast protein-protein interactions and rapid
diffusion. We propose that the waves observed in rsr1⌬
cells are a result of a competition between a fast, actin-
independent, positive feedback regulation and a slow
and therefore delayed, negative feedback regulation.

Counteracting Feedback Loops Control Cdc42 Activity
569
Figure 4. Hypothetical Model of Wave Formation of Cdc42p Regu-
lators
(A) Three feedback loops regulate the recruitment of Cdc42p regu-
lators: an actin-independent positive feedback loop (left), an actin-
mediated positive feedback loop (middle), and an actin-mediated
negative feedback loop (right).
(B) Initiation of waves (indicated by gray arrow) is expected in cells
for which the actin-mediated negative feedback dominates the ac-
tin-mediated positive feedback (left). This net negative feedback
can destabilize the polar cap and initiate cap motion. No wave mo-
bility is expected when the positive feedback loops dominate. This
can be achieved in the absence of the actin cytoskeleton (+lat A,
middle) or in strains in which the GAP activity is reduced or the
Cdc24p activity is enhanced (gap⌬or GEF+, right). More details on
the model are given in Supplemental Data.
Because the inactivation always lags behind the spread-
ing activation, this induces a wave motion (Meinhardt,
1999). This implies that in rsr1⌬cells, the actin-medi-
ated negative feedback dominates the actin-mediated
positive feedback. Wandering of the polar cap was not
observed in earlier experiments in which the position of
the polar cap was monitored by using a GFP fusion of
wild-type Cdc42p or a constitutively activated version
of Cdc42p (Wedlich-Soldner et al., 2003, 2004). The
strains used in these experiments carried a functional
RSR1 gene and expressed higher levels of activated
Cdc42p compared to wild-type. This elevated Cdc42p
activity leads to a stronger actin-mediated positive
feedback, which might explain the absence of Cdc42p
activity waves in these strains.
In rsr1⌬cells, the effective feedback systems consist
of one actin-independent positive feedback and one
actin-mediated negative feedback (Figure 4B, left panel).
The opposite holds for rsr1⌬cells in which the GAP
activity has been reduced or the GEF activity has been
increased. In these cells, the positive, actin-mediated
loop dominates the negative feedback and therefore
the feedback structure consists effectively of two paral-
lel positive loops that allow symmetry breaking but do
not support wave mobility (Figure 4B, right panel). In
the absence of the cytoskeleton, only one positive feed-
back loop remains, explaining the absence of waves in
latrunculin-treated cells (Figure 4B, middle panel). The
proposed feedback system (Figure 4A) therefore quali-
tatively explains all the experimental data. A numerical
model that explicitly models the dynamics in the dif-
ferent mutants is presented in Supplemental Data. The
above model is likely to account for cellular polarization
in other contexts, for example during cell shape changes
upon exposure of yeast cells to pheromones. Interest-
ingly, wandering polar caps were also observed in pher-
omone-treated mutant strains that lack landmark pro-
teins and are defective in chemotropism (Nern and
Arkowitz, 2000).
Conclusion
Cdc42 is essential for chemotaxis in higher eukaryotes
such as Dictyostelium and neutrophils (Watanabe et al.,
2004; Weiner et al., 2002). In these cells, interlocked
positive and negative feedback loops have been iden-
tified among Cdc42, its regulators, and the actin net-
work (Meili and Firtel, 2003). Similar arguments that ex-
plain the Cdc42p activity waves in budding yeast might
apply to the rotating pseudopod waves observed in
Dictyostelium (Killich et al., 1993, 1994) or the minCDE
oscillatory waves in Escherichia coli (Meinhardt and de
Boer, 2001; Huang et al., 2003). Mechanisms responsi-
ble for symmetry breaking are inherently coupled to
mechanisms that enable restructuring of patterns. A
design of competing feedback regulation loops com-
bines efficient polarization with the ability to dynami-
cally respond to varying intracellular or environmental
conditions.
Experimental Procedures
Construction of Plasmids and Strains
The P
GLN3
,P
IXR1
,P
MYO2
, and P
POL30
promoter sequences corre-
spond to the 600, 858, 677, and 600 base pair regions upstream of
the start codon of the respective genes. P
TET02
corresponds to the
CYC1 TATA region with two upstream rtTA binding sites. The T
RSR1
,
T
CDC24
, and T
CYC1
terminator sequences correspond to the 596,
373, and 204(187) base pair regions downstream of the stop codon
of the respective genes. BamHI-RGA1 (1–60 stop codon)-NotI frag-
ment was cloned into pRS401 backbone downstream of P
POL30
.A
sequence encoding the MGCTVSTQTIGDESDP myristoylation sig-
nal or a mutated signal sequence (G2A) was fused to the N termi-
nus of CDC24 by PCR. The resulting BamHI-myr-CDC24-T
CDC24
-
NotI and BamHI-myr(G2A)-CDC24-T
CDC24
-NotI constructs were
cloned into pRS316 and pRS401 backbones downstream of P
POL30
.
KpnI-P
MYO2
-rtTA-T
CYC1
-NotI was cloned into pRS305 and pRS401
backbones. KpnI-P
GLN3
-RSR1-T
RSR1
-NotI and KpnI-P
IXR1
-RSR1-
T
RSR1
-NotI was cloned into pRS401 backbone. XhoI-P
TETO2
-
(BamHI/BglII)-GIC2
(1–208)
-(BglII/BamHI)-GFP-T
CYC1
-NotI was cloned
into pRS306 backbone. All strains are derivatives of BY4741 (MATa
his3⌬1leu2⌬0met15⌬0ura3⌬0). The strain list is given in Table S1.
Plasmid MJ792 was a gift from M. Peter.
Growth Conditions and Data Acquisition
Yeast cells were grown overnight at 30°C in synthetic selective
dextrose media. Expression of GIC2(1–208)GFP was induced over-
night by doxycycline addition (50 ␮g/l) into the growth media. Cells
were harvested at exponential growth. For time-lapse experiments,
slides were covered with 1% agarose containing synthetic media
supplemented with 2% glucose. Cells were spun down and resus-
pended in the same media and placed on top of a thin agar layer.
Fluorescence signals of single cells were measured using a Nikon

Developmental Cell
570
TE2000 microscope (Microvideo Instruments, Avon, MA) with auto-
mated stage and a cooled back-thinned CCD camera (Micromax;
Roper Scientific, Duluth, GA). Imaging was performed at room tem-
perature (T = 22°C). Images were collected every 60 s to follow
the movement of the polar cap. The site of maximum fluorescence
intensity in the polar cap was marked as the center of the cap. The
angle between the initial and subsequent centers of the moving
polar cap was calculated using Metamorph (Universal Imaging; Re-
search Precision Instruments, Natick, MA). On average, 30 time
series were obtained to calculate the q
RMS
in each genetic back-
ground. The wave mobility was calculated from fitting the q
2RMS
time plot between t= 0 and 10 min using a linear least square fit.
Cells in the G1 stage had polar caps wandering around the cell
periphery which stopped around 15–20 min before the onset of
budding (S phase). In addition, a small fraction of cells in G1 stage
(<20%) had polar caps which either did not move or the movement
paused for a longer time period (around 20–25 min). Some of these
cells did not bud at all, which may reflect that they entered a quies-
cent state. For calculating the mobility coefficients, cells were con-
sidered in which the polar caps did not stop or pause. For measur-
ing the budding angle, cells were stained with ConA (Lew and
Reed, 1993). The angle between a small bud and the birth scar in
daughter and young mother cells was measured. Both rsr1⌬bem3⌬
and rsr1⌬rga1⌬cells continued budding randomly like rsr1⌬cells,
as assayed by budding angle (data not shown), although the sta-
bility of the position of the polar cap in these cells was similar to
that in wild-type cells. For actin depolymerization, cells were
treated with 100 ␮M latrunculin A for 10 min. Cells suspended in
media containing latrunculin A were placed on a thin agar layer and
microscopy was performed as described above. For FRAP experi-
ments, a small region in the middle of the polar cap was bleached
and fluorescence recovery was followed by confocal microscopy.
Supplemental Data
Supplemental data including a computer simulation, discussion, fig-
ures, table, and movies are available at http://www.developmentalcell.
com/cgi/content/full/9/4/565/DC1/.
Acknowledgments
We thank J. Chabot for technical assistance, B. Tam for help with
confocal microscopy, D. Pellman, J. Pedraza, and S. Oliferenko for
helpful suggestions and critical reading of the manuscript, and M.
Peter, D. Lew, E. Bi, and D. Pellman for kindly providing strains and
plasmids. A.B. is a Long-Term Fellow of the Human Frontier Sci-
ence Program. This work was supported by NSF-CAREER (PHY-
0094181) and NIH (GM068957) grants.
Received: February 21, 2005
Revised: August 3, 2005
Accepted: August 29, 2005
Published: October 3, 2005
References
Ayscough, K.R., Stryker, J., Pokala, N., Sanders, M., Crews, P., and
Drubin, D.G. (1997). High rates of actin filament turnover in budding
yeast and roles for actin in establishment and maintenance of cell
polarity revealed using the actin inhibitor latrunculin-A. J. Cell Biol.
137, 399–416.
Bose, I., Irazoqui, J.E., Moskow, J.J., Bardes, E.S., Zyla, T.R., and
Lew, D.J. (2001). Assembly of scaffold-mediated complexes con-
taining Cdc42p, the exchange factor Cdc24p, and the effector
Cla4p required for cell cycle-regulated phosphorylation of Cdc24p.
J. Biol. Chem. 276, 7176–7186.
Burbelo, P.D., Drechsel, D., and Hall, A. (1995). A conserved binding
motif defines numerous candidate target proteins for both Cdc42
and Rac GTPases. J. Biol. Chem. 270, 29071–29074.
Butty, A.C., Perrinjaquet, N., Petit, A., Jaquenoud, M., Segall, J.E.,
Hofmann, K., Zwahlen, C., and Peter, M. (2002). A positive feedback
loop stabilizes the guanine-nucleotide exchange factor Cdc24 at
sites of polarization. EMBO J. 21, 1565–1576.
Caviston, J.P., Tcheperegine, S.E., and Bi, E. (2002). Singularity in
budding: a role for the evolutionarily conserved small GTPase
Cdc42p. Proc. Natl. Acad. Sci. USA 99, 12185–12190.
Chant, J. (1999). Cell polarity in yeast. Annu. Rev. Cell Dev. Biol. 15,
365–391.
Gierer, A., and Meinhardt, H. (1972). A theory of biological pattern
formation. Kybernetik 12, 30–39.
Holland, M.J. (2002). Transcript abundance in yeast varies over six
orders of magnitude. J. Biol. Chem. 277, 14363–14366.
Huang, K.C., Meir, Y., and Wingreen, N.S. (2003). Dynamic struc-
tures in Escherichia coli: spontaneous formation of MinE rings and
MinD polar zones. Proc. Natl. Acad. Sci. USA 100, 12724–12728.
Irazoqui, J.E., Gladfelter, A.S., and Lew, D.J. (2003). Scaffold-medi-
ated symmetry breaking by Cdc42p. Nat. Cell Biol. 5, 1062–1070.
Irazoqui, J.E., Gladfelter, A.S., and Lew, D.J. (2004). Cdc42p, GTP
hydrolysis, and the cell’s sense of direction. Cell Cycle 3, 861–864.
Irazoqui, J.E., Howell, A.S., Theesfeld, C.L., and Lew, D.J. (2005).
Opposing roles for actin in Cdc42p polarization. Mol. Biol. Cell 16,
1296–1304.
Killich, T., Plath, P.J., Xiang, W., Bultmann, H., Rensing, L., and
Vicker, M.G. (1993). The locomotion shape and pseudopodial dy-
namics of unstimulated Dictyostelium cells are not random. J. Cell
Sci. 106, 1005–1013.
Killich, T., Plath, P.J., Hass, E.C., Xiang, W., Bultmann, H., Rensing,
L., and Vicker, M.G. (1994). Cell-movement and shape are nonran-
dom and determined by intracellular, oscillatory rotating waves in
Dictyostelium amebas. Biosystems 33, 75–87.
Kozminski, K.G., Beven, L., Angerman, E., Tong, A.H., Boone, C.,
and Park, H.O. (2003). Interaction between a Ras and a Rho
GTPase couples selection of a growth site to the development of
cell polarity in yeast. Mol. Biol. Cell 14, 4958–4970.
Lechler, T., Shevchenko, A., and Li, R. (2000). Direct involvement of
yeast type I myosins in Cdc42-dependent actin polymerization. J.
Cell Biol. 148, 363–373.
Lew, D.J., and Reed, S.I. (1993). Morphogenesis in the yeast cell
cycle: regulation by Cdc28 and cyclins. J. Cell Biol. 120, 1305–
1320.
Meili, R., and Firtel, R.A. (2003). Two poles and a compass. Cell
114, 153–156.
Meinhardt, H. (1999). Orientation of chemotactic cells and growth
cones: models and mechanisms. J. Cell Sci. 112, 2867–2874.
Meinhardt, H., and de Boer, P.A.J. (2001). Pattern formation in
E. coli: a model for the pole-to-pole oscillations of Min proteins and
the localization of the division site. Proc. Natl. Acad. Sci. USA 98,
14202–14207.
Nelson, W.J. (2003). Adaptation of core mechanisms to generate
cell polarity. Nature 422, 766–774.
Nern, A., and Arkowitz, R.A. (2000). G proteins mediate changes in
cell shape by stabilizing the axis of polarity. Mol. Cell 5, 853–864.
Park, H.O., Bi, E., Pringle, J.R., and Herskowitz, I. (1997). Two active
states of the Ras-related Bud1/Rsr1 protein bind to different effec-
tors to determine yeast cell polarity. Proc. Natl. Acad. Sci. USA 94,
4463–4468.
Saxton, M.J., and Jacobson, K. (1997). Single-particle tracking: ap-
plications to membrane dynamics. Annu. Rev. Biophys. Biomol.
Struct. 26, 373–399.
Shimada, Y., Wiget, P., Gulli, M.P., Bi, E., and Peter, M. (2004). The
nucleotide exchange factor Cdc24p may be regulated by auto-inhi-
bition. EMBO J. 23, 1051–1062.
Toenjes, K.A., Simpson, D., and Johnson, D.I. (2004). Separate
membrane targeting and anchoring domains function in the local-
ization of the S. cerevisiae Cdc24p guanine nucleotide exchange
factor. Curr. Genet. 45, 257–264.
Watanabe, T., Wang, S., Noritake, J., Sato, K., Fukata, M., Takefuji,
M., Nakagawa, M., Izumi, N., Akiyama, T., and Kaibuchi, K. (2004).
Interaction with IQGAP1 links APC to Rac1, Cdc42, and actin fila-
ments during cell polarization and migration. Dev. Cell 7, 871–883.

Counteracting Feedback Loops Control Cdc42 Activity
571
Wedlich-Soldner, R., Altschuler, S., Wu, L., and Li, R. (2003). Spon-
taneous cell polarization through actomyosin-based delivery of the
Cdc42 GTPase. Science 299, 1231–1235.
Wedlich-Soldner, R., Wai, S.C., Schmidt, T., and Li, R. (2004). Ro-
bust cell polarity is a dynamic state established by coupling trans-
port and GTPase signaling. J. Cell Biol. 166, 889–900.
Weiner, O.D., Neilsen, P.O., Prestwich, G.D., Kirschner, M.W., Cant-
ley, L.C., and Bourne, H.R. (2002). A PtdInsP(3)- and Rho GTPase-
mediated positive feedback loop regulates neutrophil polarity. Nat.
Cell Biol. 4, 509–513.