Characterization of Protein Dynamics in Asymmetric Cell Division
by Scanning Fluorescence Correlation Spectroscopy
Zdene ˇk Petra ´s ˇek,* Carsten Hoege,yAlireza Mashaghi,yThomas Ohrt,* Anthony A. Hyman,y
and Petra Schwille*
*Biophysics Group, Biotechnologisches Zentrum, Technische Universita ¨t Dresden, Dresden, Germany; andyMax Planck Institute of
Molecular Cell Biology and Genetics, Dresden, Germany
processes that all rely on the distribution and interaction of proteins. Despite the tight regulation of these processes with respect to
temporal and spatial protein localization, exact quantification of the underlying parameters, such as concentrations and distribution
coefficients, has so far been problematic. Recent experiments suggest that fluorescence correlation spectroscopy on a single
molecule level in living cells has great promise in revealing these parameters with high precision. The optically challenging
situation in multicellular systems such as embryos can be ameliorated by two-photon excitation, where scattering background and
cumulative photobleaching is limited. A more severe problem is posed by the large range of molecular mobilities observed at
the same time, as standard FCS relies strongly on the presence of mobility-induced fluctuations. In this study, we overcame the
limitations of standard FCS. We analyzed in vivo polarity protein PAR-2 from eggs of Caenorhabditis elegans by beam-scanning
FCS in the cytosol and on the cortex of C. elegans before asymmetric cell division. The surprising result is that the distribution of
PAR-2 is largely uncoupled from the movement of cytoskeletal components of the cortex. These results call for a more systematic
future investigation of the different cortical elements, and show that the FCS technique can contribute to answering these
questions, by providing a complementary approach that can reveal insights not obtainable by other techniques.
The development and differentiation of complex organisms from the single fertilized egg is regulated by a variety of
Asymmetric division of early embryonic cells is essential
for future cell diversity and is preceded by the establishment
of cell polarity. A polarization signal leads to distinct con-
centrations of polarity factors in different cellular domains of
A good system for studies on polarization and asymmetric
first cell division is regulated by PAR proteins, among others
(3–5), and requires the motility of the highly dynamic acto-
myosin cortex (contractions and flows) and the asymmetric
localization of PAR proteins (6–8).
The polarization of the fertilized embryo is initiated by an
interaction of the centrosome with the cortex (9) and leads to
a cortical flow of actomyosin from the posterior toward the
anterior part of the cell. This flow is closely connected with
the accumulation of PAR-1 and PAR-2 at the posterior, and
the restriction of PAR-3, PAR-6, and PKC-3 to the anterior
half of the cell.
At a later maintenance phase, the asymmetric redistribu-
tion of the PAR proteins back to the whole cortex is pre-
vented by their antagonistic interactions (10). Non-muscle
myosin NMY-2, a component of the actomyosin cortex, is
required through all phases, and after polarity establishment
asymmetrically distributes with a more contractile cortex on
the anterior side.
Binding of PAR-1 and PAR-2 to the cortex depends on a
functional actomyosin cytoskeleton, as inhibition by cyto-
chalasin B, an actin inhibitor, or downregulation of NMY-2
by RNAi inhibits binding. While direct interaction between
PAR-2 and NMY-2 has not been demonstrated, PAR-1 can
bind directly to NMY-2 in vivo, and PAR-1 localization is
dependent on PAR-2 (7,8).
Although several polarity components of the Caeno-
rhabditis elegans embryo have been discovered, it remains
largely unclear how the asymmetry of the cortex-associated
proteins before andduring the first division is established and
maintained, and how the proteins interact, move, and redis-
tribute within the embryo on the molecular level.
The localization and redistribution of proteins involved
in polarity is usually studied by fluorescence microscopy.
While this imaging technique reveals the protein distribution
and dynamics in living embryo on the second or subsecond
timescale, its low temporal resolution does not allow the
observation of fast dynamics on the scale of microseconds to
Here we apply standard fluorescence correlation spec-
troscopy (FCS) to measure the relatively fast diffusion of
PAR-2, NMY-2, CDC-37, and a membrane-binding PH
domain protein in cytosol. Circular scanning FCS (sFCS) is
employed to investigate the considerably slower motion of
the two cortex-associated proteinsPAR-2and NMY-2, a task
practically impossible with standard FCS.
In FCS, the statistical analysis of the detected fluorescence
signal by means of the autocorrelation function reveals im-
portant information about the dynamics of the diffusing
Submitted April 10, 2008, and accepted for publication August 26, 2008.
Zdene ˇk Petra ´s ˇek and Carsten Hoege contributed equally to this work.
Address reprint requests to Petra Schwille, E-mail: petra.schwille@biotec.
Editor: Alberto Diaspro.
? 2008 by the Biophysical Society
5476 Biophysical JournalVolume 95 December 20085476–5486
species, such as diffusion coefficients and the type of motion
(11–13). Scanning FCS, a modification of the standard FCS
with the measurement volume being scanned in a controlled
fashion across the sample (14), has been introduced to
overcome problems with low statistical accuracy in systems
with slow diffusion (15,16), to minimize the effects of pho-
tobleaching of slowly diffusing molecules, or for other rea-
We show that PAR-2 and non-muscle myosin NMY-2
diffuse freely in cytoplasm, with diffusion coefficients lower
than expected in aqueous environment, most likely due to
collisions with large components of the crowded cytoplasm,
formation of larger complexes, or due to transient binding.
On the cortex, both proteins exhibit highly dynamic non-
uniform distribution. NMY-2 is localized in well-defined
patchesoftheactomyosincortexthat changes itscontractility
throughout the cell cycle (6). Interestingly, PAR-2 assembles
in a less discrete and more dynamic pattern than NMY-2.
Scanning FCS autocorrelations of NMY-2 decay at a longer
timescale with a sharp falloff, characteristic of directed mo-
tion. On contrary, PAR-2 motion is faster and the autocor-
relation decay is more gradual, indicating multicomponent
diffusive or even subdiffusive behavior.
The presented results show that FCS and sFCS are tech-
niques suitable for the study of protein dynamics in an
asymmetric dividing embryo on the temporal scales ranging
from microseconds to seconds, and demonstrate that PAR-2
and NMY-2 show largely independent motion on the cortex.
MATERIALS AND METHODS
The experiments were performed on a homebuilt two-photon laser scanning
microscope (14,20) using an UPLAPO 603 W3/IR objective (Olympus,
Hercules, CA). The excitation was provided by a tunable Ti:Sapphire laser
(Mira 900-F, Coherent, Santa Clara, CA) with the wavelength set to 920 nm,
and an average power of 5 mW. The laser beam is steered by two galva-
nometer scanners fully controllable by software, allowing the system to
operate in two modes: a conventional imaging laser scanning microscope
mode, and an sFCS mode, where the beam is scanned along a circular path
with a user-defined radius and frequency. The fluorescence, collected by the
objective and transmitted through an appropriate emission filter, is detected
by an avalanche photodiode (model No. SPCM-CD2801, PerkinElmer,
Wellesley, MA). For correlation measurements, the stream of detected
photocounts is either directly autocorrelated with a multiple-tau hardware
correlator (ALV-6000, ALV, Langen, Germany), or processed by the SPC-
830 module (21) (Becker & Hickl, Berlin, Germany) to obtain timing of
every photoncount with resolution 13.1 ns, and stored for further analysis.
The time of a point-FCS measurement at one location was 10 3 10 s. The
time of one sFCS measurement was ;100 s, corresponding to 30,000 scan
orbits at the used frequency of 300 Hz. The scan radii used in this work were
in the range 2–9 mm, depending on the size of the flat part of the embryo
C. elegans worms were cultured at 16?C on nematode growth media plates
seeded with OP50 bacteria, and shifted to 25?C a day before imaging (22).
were described before. Briefly, GFP fusion proteins were expressed from
stable transgenic worm lines from the pTH-GFP vector under the control of
the pie-1 promoter. A transgenic worm line expressing GFPTCDC-37 from
the pie-1 promoter (vector pTH-GFP-Gateway, containing the enhanced
GFP variant) was produced by microparticle bombardment of unc-119(ed3)
The fertilized eggs taken from adult worms were released by slicing an-
imals open with a pair of needlesin M9 buffer.Thenthe eggswere placed on
agarose gel spread in a thin layer on a coverslip, and covered with another
coverslip. The specimen was turned upside down and observed on an in-
verted microscope, as described above. The embryos were observed either
with objective focus in the equatorial plane (cytoplasm FCS measurements),
or with the focus near the coverslip, wherethe embryo was partially flattened
and larger part of the embryo cortex was in focus, thus allowing sFCS
The sFCS measurements were performed in the maintenance phase after
pseudocleavage and before the second cell division. During this time period,
the cortex is relaxed after the myosin contractile foci have disassembled, and
does not show strong contractile activity as observed in the polarity estab-
lishment phase (6). No measurements were performed during pseudocleav-
age and cytokinesis because of large-scale cortical movements (cell cortex
rotation) accompanying cell division. However, more precise specification
large-scale coordinated cortical movements (approximately minutes).
The fluorescence autocorrelation curves g(t) obtained with the hardware
correlator according to the definition
gðtÞ ¼ ÆdFðtÞdFðt1tÞæ=ÆFðtÞæ2
were analyzed using the model of free diffusion in three-dimensional
where ÆF(t)æ is the average fluorescence intensity over the whole measure-
ment, dF(t) [ F(t) – ÆF(t)æ, g0is the amplitude of the autocorrelation
function, tD¼ a2/D is the diffusion time, a is the e?1/2width of the
measurement volume approximated by a Gaussian function (a ¼ 0.14 mm in
our case), w is the form factor describing the extension of the measurement
volume in the axial direction, and D is the diffusion coefficient. The
autocorrelation curves measured on the flattened part of the cell cortex
were analyzed by an equivalent model for diffusion in two dimensions,
which can be obtained from Eq. 2 by letting w / N.
The stream of photocounts recorded during the sFCS measurements was
stored and autocorrelated off-line. Since the relevant dynamics occurs on the
timescales only 1–2 orders of magnitude shorter than the permissible time
of the measurement, the correlation analysis is affected by poor statistical
accuracy at long correlation times. For this reason, compensated normali-
zation was chosen in the calculation of the correlations, as it has been shown
to provide higher accuracy estimates of the correlations at long correlation
the following way:
gðtkÞ ¼ ðÆFðtiÞFðti1kÞæi? F0FkÞ=F2
N ? k
N ? k+
Protein Dynamics Measured by sFCS5477
Biophysical Journal 95(11) 5476–5486
Scanning the measurement volume introduces periodic oscillations into
the autocorrelation curve persisting typically as long as the correlation has a
calculated with sufficiently high resolution even at long correlation times.
The common multiple-tau correlation technique does not have sufficient
temporal resolution at larger lag times, causing the oscillations to average
out. The scanning autocorrelation curves were therefore calculated with
linearly spaced channels of constant width (1 ms or 10 ms). To achieve re-
alistic computation times, the correlation was calculated via Fourier trans-
form of the detected photoncount sequence FðFðtÞÞ using the following
The details on the implementation of the calculation of linear autocorrela-
tion of a long data stream using Fourier transform have been published
The long-scale fluorescence fluctuations caused by the global changes of
the fluorescence pattern resulted in fluctuations in the tails of the sFCS au-
tocorrelation functions. These fluctuations do not reflect the local dynamics
of the fluorescence pattern (fluctuations at any given position along the
scanned circle), and were therefore filtered out from the calculated curves by
the following procedure. The fluctuating fluorescence signal F(t) was as-
sumed to consist of the local fluctuating signal f(t) modulated by a slowly
changing global modulation function h(t): F(t) ¼ f(t)h(t). The function h(t)
was estimated from the measured fluorescence F(t) by smoothing F(t) with a
above-mentioned definition that the autocorrelation gF(t) of F(t) is related to
the autocorrelation gf(t) of f(t) and the autocorrelation gh(t) of h(t) in the
FðtÞFðt1tÞdt ¼ F?1ðjFðFðtÞÞj2Þ:
gFðtÞ11 ¼ ðgfðtÞ11ÞðghðtÞ11Þ:
The desired autocorrelation gf(t) was then calculated from Eq. 6 using
knowngF(t) and gh(t). Since the effective number of measurementlocations
along the circle is large, this procedure does not affect the resulting temporal
autocorrelation profile, as would be the case in a measurement with fixed
detection volume. This filtering procedure is demonstrated in Fig. S3 in
Supplementary Material, Data S1.
can produce additional oscillations in the autocorrelation curve, reflecting
this nonuniformity. Again, these fluctuations can be filtered out applying Eq.
6, where h(t) is now the average fluorescence intensity along the scanned
circle over the whole measurement, and gh(t) is the autocorrelation of h(t)
(see Fig. S3 in Data S1). The effect of this procedure is similar to the phase-
normalized autocorrelation introduced by Skinner et al. (18) to separate
fluctuations due to immobilized particles.
Free diffusion on a two-dimensional surface while scanning the beam
with frequency v in a circle of radius R leads to the following model auto-
The preexponential term corresponds to the standard autocorrelation of
diffusion process in two dimensions, and the exponential factor describes
the periodic modulation due to the scanning motion. The scanning auto-
correlation consists of peaks located at correlation times equal to the
integer multiples of the scan period T, that is, at times when the scanning
focus returns to the initial position. The peak maxima coincide with the
autocorrelation value that would be obtained in the absence of scanning (the
preexponential term in Eq. 7). When the scan radius R is much larger than
the size of the measurement volume a, the correlation between the peaks is
practically zero, since the fluorescence at points located far from each other
the autocorrelation increases with the correlation time, as the molecules
diffuse further and further away from their initial position (see Fig. S2 in
Data S1). When the molecular motion is much slower than the scanning
speed, the shape of the peaks in the autocorrelation function is largely
determined by the exponential factor in Eq. 7. Therefore, the exponential
factor with an effective width parameter a92¼ a2(1 1 t/tD)/R2and an
amplitude A as two fitting parameters were used to fit the peaks of the
experimental sFCS autocorrelation functions.
The scanning autocorrelation can be viewed as a part of the full spatio-
temporal correlation g(x, t); that is, the correlation of fluorescence between
two locations spaced by the distance x at two time points delayed by the time
t. This becomes apparent when one realizes that only the values at integral
multiples of the scan period are autocorrelations at the same location, while
the values at all other times are actually spatial cross correlations between
two locations at a distance x:
x ¼ 2Rsinðvt=2Þ:
This equation relates the correlation time t to the spatial correlation
coordinate x, or alternatively, to the phase u of the circular motion: u ¼
vt: Scanning FCS samples the full spatiotemporal correlation at coordinates
(x, t) linked by Eq. 8. This sampling is fine enough when the scanning
motion is much faster than the relevant timescales determined by the
investigated molecular motion. The sFCS autocorrelation can then be
displayed as a two-dimensional plot with the axes formed by the x and t
coordinates, and the correlation value color-coded.
The spatiotemporal correlation g(x, t) contains more information than a
temporal correlation g(t), and possibility to measure it experimentally gives
a better chance to distinguish between different models of transport. For
example, the spatiotemporal correlation of diffusion in two dimensions
exhibits broadening in space with longer correlation times as the molecules
diffuse from their initial position. On the other hand, binding to/detachment
form a surface,
gðx;tÞ ¼ g0e?kte
does not show any spatial broadening, since the molecules do not move
laterally and disappear fast after detachment from the surface. Although
these two models already show different dependence of the autocorrelation
g(t) on the correlation time, the additional differences in spatial correlation,
alternative models (see Fig. S1 in Data S1 for several models of spatiotem-
Diffusion in cytoplasm
The PAR-2 and NMY-2 proteins are known to localize in
cytoplasm and in different parts of the embryo cortex, de-
pending on the phase of the cell cycle (29). Whereas PAR-2
localizes almost exclusively on the posterior cortex after
polarity establishment, NMY-2 localizes predominantly on
the anterior cortex, while being less contractile on the pos-
terior (Fig. 1, A and B). However, posterior localization of
PAR-2 depends on NMY-2 (7). Therefore, an interesting
question is whether PAR-2 can bind to the cortex directly
by association with NMY-2. We tested this hypothesis by
studying the dynamics of both proteins in the cytosol and on
the cortex by FCS.
5478Petra ´s ˇek et al.
Biophysical Journal 95(11) 5476–5486
We have employed two-photon FCS with the measurement
volume positioned in a fixed location within the cytoplasm to
measurement volume was positioned away from the pronu-
clei; the movement of the pronuclei did not influence the
measured protein dynamics because it occurs on a much
longer timescale. The autocorrelation curves were analyzed
using the model of diffusion in three dimensions (Eq. 2). The
distributions of the obtained diffusion coefficients are shown
in Fig. 1 C (GFPTPAR-2) and Fig. 1 D (NMY-2TGFP). The
mean values ofthediffusioncoefficients determinedfrom the
measured distributions are shown in Table 1, with the errors
indicating the widths of the distributions. The differences in
the distributions of the diffusion coefficients of the two
proteins indicate that PAR-2 and NMY-2 in the cytosol do
not diffuse as a part of a common complex. The possibility
to use Eq. 2 to describe the measured autocorrelations indi-
cates predominantly diffusive behavior of the two proteins in
For comparison, we performed analogous FCS measure-
ments on two other proteins in cytosol: GFPTCDC-37 and
GFPTPH. CDC-37, a proteinofsimilarsize toPAR-2 (Table
1), is required for polarity establishment and mutual exclu-
sion of anterior and posterior PAR proteins (30). It is uni-
formly distributed in both anterior and posterior halves of
the embryo and does not localize on the cortex. The PH
domain, a small protein domain derived from mammalian
PLC1d1, is known to bind to PIP2lipid present in the plasma
membrane, and therefore serves as a membrane marker
(24,31). Embryos expressing GFPTPH display very bright
fluorescence on the plasma membrane and weaker fluores-
cence in the cytosol, permitting standard FCS measurements
in both. The measured diffusion coefficients of both proteins
again exhibit broadened distributions. The mean values to-
gether withthestandard deviations indicatingthedistribution
widths are shown in Table 1.
In addition to cytosolic diffusion, it was possible to mea-
sure standard FCS autocorrelations of the GFPTPH on the
membrane, by focusing onto the flattened bottom part of the
embryo as described below, and thus preventing the move-
ment of the membrane out of focus. The diffusion of the
than in the cytosol (Table 1). The relatively fast diffusion of
all the investigated proteins in cytosol and the PH domain on
the membrane (compared to PAR-2 and NMY-2 dynamics
on the cortex; see below) means that no distortions due to
Scanning FCS on the cortex
Attempts to perform standard FCS measurements on PAR-2
and NMY-2 localized on the cortex by positioning the mea-
surement volume on the cortex when focused in the embryo
midplane (Fig. 1, A and B) did not provide satisfactory results.
The main reason was the motion of the embryo caused by its
development, on the timescale of ;100 s needed for the
C) and NMY-2TGFP (B and D) in the cytosol. The
measurement volume was positioned into different parts of
the cytosol, while focused into the midplane of the embryo
(A and B), and the measured autocorrelation curves were
fitted to Eq. 2. The distributions of the obtained diffusion
coefficients D are shown in panels C (GFPTPAR-2) and
D (NMY-2TGFP). (Scale bar, 10 mm. a, anterior; p, pos-
TABLE 1Diffusion coefficients determined with FCS
Protein LocationM/kDa D/mm2s?1
0.9 6 0.2
1.5 6 0.8
4.8 6 1.3
8.1 6 2.0
1.1 6 0.3
The error indicates the width of the distribution of D. The molecular mass
M is inclusive the GFP label.
Protein Dynamics Measured by sFCS5479
Biophysical Journal 95(11) 5476–5486
measurement. Motion of the cortex as a whole out of the
measurement volume cannot be separated from the slow
motion of molecules within the cortex, thereby compromis-
ing the results.
Another problem with measurements with fixed volume is
the low statistical accuracy due to the very slow motion of
the molecules on the cortex: during the maximum realistic
measurement time, limited by the development of the em-
bryo, insufficient number of molecules passes through the
measurement volume, resulting in low accuracy of the av-
eraging procedure. Additionally, photobleaching can occur
during the long residence time in the measurement volume.
To overcome these problems, we have used circle sFCS,
where the measurement volume is moved along a circular
path with a known radius and frequency (16,18). The ob-
jective was focused near the coverslip onto the flattened
bottom part of the embryo. In this way, the largest possible
part of the cortex was simultaneously present in focus. The
radius was chosen as large as possible so that the whole
scanning path lied within the focused part of the cortex (Fig.
2, A and B). Longer scan path means that information from
effectively more independent volumes is averaged, implying
better statistical accuracy. The data obtained in this way
represents an average over the scanned path, and it is there-
fore implicitly assumed that the typical dynamics are the
same at all probed locations.
The fluorescence intensity trace was recorded with the
SPC-830 module and its autocorrelation was calculated off-
line as described in Materials and Methods. Typical auto-
correlation curves of GFPTPAR-2 and NMY-2TGFP are
shown in Fig. 3, A and B, respectively. The peaks on the blue
curves at times nT, where T is the scan period T ¼ 1/300 s,
correspond to the autocorrelation at the same location after n
rotations of the measurement volume. The shapes of the
peaks (not discernible in the figure) were fitted to the model
as expressed by Eq. 7, and the fitted amplitude of every peak
is marked as a black dot in the plots in Fig. 3, A and B. An
example of fits to several selected peaks of a GFPTPAR-2
autocorrelation function is shown in Fig. S2 in Data S1. The
curve formed by the amplitudes of all peaks then represents
the loss of correlation of the fluorescence pattern on the
cortex, and reflects the temporal redistribution of this pattern,
ina similar wayas anordinaryFCScurvereflects thepassage
of molecules through the measurement volume.
The typical decay of autocorrelation of GFPTPAR-2 was
found to differ from that of NMY-2TGFP, as shown in Fig.
4, where several normalized autocorrelation curves (formed
by peak maxima as described above) of each protein are
plotted. The autocorrelation of GFPTPAR-2 decays over a
broad temporal range and cannot be described by simple
suggests multicomponent or anomalous diffusion. The au-
tocorrelation decay of NMY-2TGFP is steeper than the
autocorrelation of a two-dimensional diffusional process,
possibly indicating contribution of translational motion,
which alone leads to a steep exp(?(t/tf)2) temporal depen-
dence (32). The motion of NMY-2TGFP on the cortex is
as characterized by the autocorrelation curves was found
to be similar in both anterior and posterior parts of the cell,
independently of the higher fluorescence intensity in the an-
The distribution of both the proteins on the cortex was
nonuniform, with rapidly changing pattern (Fig. 2, A and B).
The sizes of the features in both cases were near or below the
GFPTPAR-2 (A and C) and NMY-2TGFP (B and D) on
the cortex. The objective was focused near the coverslip
onto the flattened part of the embryo. The circles in panels
A and B indicate the scan path for sFCS measurements.
(C andD) Thefluctuatingfluorescenceintensityrecordedin
the sFCS measurement is displayed in a two-dimensional
plot, where the horizontal axis corresponds to one revolu-
tion (scan period T), and the vertical axis to subsequent
revolutions during the course of measurement (from top to
bottom). The columns in the plot then show the fluores-
cence fluctuations at individual positions along the scanned
circle. (C) GFPTPAR-2, D: NMY-2TGFP. (Scale bar,
10 mm. a, anterior; p, posterior.)
5480 Petra ´s ˇek et al.
Biophysical Journal 95(11) 5476–5486
motion and the dynamics of this pattern rather than only the
dynamics of individual independently moving molecules.
The pattern of NMY-2TGFP consists of bright spots with
size comparable to or below the resolution limit (6), whose
movement appears to be to some extent coordinated and
partially directional (nondiffusive). This can be seen in Fig.
2 D, where the fluorescence intensity from sFCS measure-
ments is plotted in a two-dimensional plot in such a way that
each line corresponds to one revolution, the horizontal axis
corresponds to points along the circle, and the measurement
time increases from the top to the bottom. The inclined ver-
tical traces originate from bright spots crossing the scanned
in the same direction and indicate therefore coordinated mo-
tion. The fluorescence distribution of GFPTPAR-2 on the
cortex is also found to be nonhomogeneous and highly dy-
namic. However, it appears to be much less concentrated into
discrete spots and its motion is less coordinated (Fig. 2 C).
The characterofmotionof fluorescently labeledmolecules
is reflected by spatiotemporal correlation g(t, x). Therefore,
the sFCS correlation data were displayed as functions of the
correlation coordinates (t, x), as described in Materials and
Methods (Figs. 6 and 7). Both PAR-2 and NMY-2 exhibit
spatial broadening of the correlations with increasing corre-
lation time t, as can be better seen in Figs. 6 B and 7 B, where
The broadening appears at shorter times in case of PAR-2
and NMY-2TGFP (thick gray) on the cortex and FCS autocorrelations of
The sFCS autocorrelations are formed by the fitted amplitudes of the peaks
as shown in Fig. 3.
Comparison of sFCS autocorrelations of GFPTPAR-2 (black)
GFPTPAR-2 (A) and NMY-2TGFP (B) with several simple models of
transport: uniform flow (green), binding/dissociation (magenta), normal
diffusion (black), anomalous diffusion (gray), normal two-component dif-
fusion (red), and flow with a Gaussian distribution of speeds, centered at
v ¼ 0 : pðvÞ;expð?v2=ð2s2
istic time constants of the relevant process.
Comparison of a typical sFCS autocorrelation (blue) of
vÞÞ (cyan). The parameters tiare the character-
measurements (shaded). The amplitudes of the peaks obtained from fits to
Eq. 7 are indicated by solid dots. (A) GFPTPAR-2, (B) NMY-2TGFP.
The experimental fluorescence autocorrelations from sFCS
Protein Dynamics Measured by sFCS 5481
Biophysical Journal 95(11) 5476–5486
than NMY-2, which is consistent with the shorter correlation
times, implying faster motion. The normalized correlations
also show that the time t ; 10 s represents a practical ac-
curacy limit in these measurements, beyond which the data
are dominated by noise. The PAR-2 correlation was fitted to
three models: one-component diffusion, two-component
diffusion, and one-component diffusion with one binding/
dissociation component, using an appropriate linear combi-
nation of Eqs. 9 and 10 (Fig. 6, C–H). As expected from
observations made above (Fig. 4), the one-component dif-
fusion model does not describe the data well. Out of the two
models with two components, two-component diffusion
leads to a better fit with the data than one diffusion and one
binding component, as judged by the residual plots and the
values of x2(Fig. 6, G and H). No fitting of NMY-2 spa-
tiotemporal correlation was attempted, because the spatial
features start to appear at correlation times too close to the
accuracy limit (Fig. 7 B).
Scanning FCS on the cortex
The presentedresultsofthe sFCS measurements demonstrate
that this technique is capable of capturing the dynamics,
which is too slow for conventional FCS. By scanning along a
t the temporal correlation coordinate. (A) The spatiotemporal autocorrelation. (B) The spatiotemporal autocorrelation normalized to the maximum at each t
value to emphasize the spatial broadening. (C and F) Fit to a one-component diffusion model and the residuals of the fit (x2¼ 4.61 3 10?6). The white
rectangle denotes the fitting range. (D and G) Fit to a two-component diffusion model and the fit residuals (x2¼ 3.99 3 10?6). (E and H) Fit to a model with
one diffusion and one binding/dissociation components, and the fit residuals (x2¼ 4.37 3 10?6). The correlation coordinate x (corresponding to the distance
along the scanned circle) can be equivalently expressed by the scan phase u; or by the time from the peak maximum t. These three coordinates are related to
each other in the following way: x ¼ 2R sin(vt/2), u ¼ vt:
sFCS autocorrelation of GFPTPAR-2 displayed in spatiotemporal representation, and fits to three different models. The value x is the spatial and
5482 Petra ´s ˇek et al.
Biophysical Journal 95(11) 5476–5486
circle, multiple locations are probed pseudo-simultaneously,
and sufficient averaging can be achieved in a time shorter by
one-to-two orders of magnitude than the time necessary for a
measurement with a fixed detection volume. Furthermore, no
photobleaching of slowly moving molecules was observed
(Fig. 2, C and D), which can be attributed to the fact that each
location along the circular pathis illuminated foronly a small
fraction of the scan period T.
The sFCS correlations represent information averaged
along the scanned circle. To see whether there are any sta-
tistically significant variations in dynamics at different
points, we divided the raw data and calculated autocorrela-
tions at individual locations along the circle (see Fig. S4 in
Data S1). The individual curves are strongly affected by poor
statistics (which is one of the reasons for averaging over a
larger area); however, within the limits of this accuracy, the
dynamics are comparable at all points along the circle.
The comparison of the sFCS autocorrelation curves (Fig.
4) of GFPTPAR-2 and NMY-2TGFP shows that the dy-
namics of bothproteins on the cortex are clearly distinct from
each other. While the motion of GFPTPAR-2 is faster and
has a rather diffusive or subdiffusive character, the motion of
NMY-2TGFP is slower and the sharp falloff of the autocor-
by comparison with several models that the NMY-2TGFP
autocorrelation decay is steeper than that of diffusion, but
more gradual than flow with uniform speed. A reasonable
assumption of distribution of flow velocities results in an
autocorrelation curve with a slope close to the experimental
data, as demonstrated on the example of Gaussian velocity
model matches the data well, the observed lateral motion of
myosin clusters is not compatible with this model.
The partially coordinated flow of myosin might be related
to the deformation of the actin cortex to which the bright
falloff times of the NMY-2TGFP autocorrelation curves are
vÞÞ: Although the binding
likely to be caused by natural variations in cortical motion in
any given experiment. The cortical flow speed during po-
larity establishment has been previously determined as ;4
mm min?1(5) with maximum value of 7.7 mm min?1(6),
which translates into equivalent flow times tfof the corre-
lation curves of 4.1 s or 2.1 s, respectively. This value is
comparable to the characteristic decay times of the NMY-2
sFCS autocorrelations (Fig. 4), suggesting that the myosin
moves with similar velocities at both polarity establishment
and maintenance phases. During the polarity establishment
phase the myosin patches move all in the same direction,
resulting in a net cortex flow, while in the later phases the
motion is directionally uncorrelated, preventing appearance
of any large-scale flows, and being observable only as a re-
organization of the punctuate fluorescence pattern.
The decay of the sFCS autocorrelation of PAR-2 is more
gradual than the loss of autocorrelation due to the following
and normal diffusion, as shown in Fig. 5 A. Combination of
binding with diffusion or flow cannot result in an autocorre-
lation that is flatter than in any of these processes alone. The
PAR-2 autocorrelation decay can be better described by the
model of anomalous diffusion with a time-stretching param-
eter a. Even better match with the experimental data can be
components, each with its own type of transport and time
constant, as demonstrated by two-component diffusion in
favor a two-component diffusion model over a model with
one-component diffusion and one binding/dissociation com-
ponent. The multicomponent diffusion model is consistent
with the presence of slow bright clusters (Fig. 2 C) and faster
smaller complexes or individual molecules. The gradual au-
tocorrelation decay, broadening of sFCS autocorrelation
peaks with increasing lag time, and the dynamics of bright
patches as seen in Fig. 2 C all suggest presence of diffusional
temporal autocorrelation normalized to the maximum at each t.
sFCS autocorrelation of NMY-2TGFP displayed in spatiotemporal representation. (A) The spatiotemporal autocorrelation. (B) The spatio-
Protein Dynamics Measured by sFCS 5483
Biophysical Journal 95(11) 5476–5486
possibility of minor contribution of binding kinetics to one or
more components of motion, with diffusion still being the
dominant transport mechanism.
The differences in the dynamics between PAR-2 and
NMY-2 proteins inferred from the autocorrelation data
show that although there might be minor fractions that bind
each other, the proteins move within the cortex mostly
The faster motion of PAR-2 relative to that of NMY-2,
which is assumed to reflect the cortex dynamics, and the
presence of a clearly detectable pattern (Fig. 2 A), raises the
question oftheroleofthemembrane presentabovethecortex
in the localization of PAR-2. Although the dynamics of
PAR-2 on the cortex are slower than those of the PH domain
on the membrane, the localization of PAR-2 with compo-
nents of the membrane cannot be excluded. The existence of
a PAR-2 pattern implies reduced molecular mobility, and the
autocorrelations measured with sFCS are likely to be domi-
the dynamics of individually diffusing PAR-2 molecules,
therefore leading to a slower-decaying autocorrelation. It is
not clear whether the bright pattern dynamics represents mo-
tion of bright clusters containing many GFP-labeled PAR-2
molecules, or whether the observed dynamics is a net effect of
simultaneous growth and disassembly of otherwise static
clusters (or combination of the two phenomena). Developing
a model of fluorescence autocorrelation and spatiotemporal
correlation resulting from the latter mechanism might help to
discriminate between the two processes.
on the membrane (1.1 mm2s?1) is comparable with that of
fluorescent lipid analogs in plasma membranes of rat baso-
kidney cells (1.4 mm2s?1) (33), suggesting free diffusion of
GFPTPH bound to PIP2in the membrane.
Cheeks et al. (5) used fluorescence recovery after photo-
bleaching (FRAP) to show that photobleached PAR-2 on
cortex completely recovers on the scale of tens of seconds.
Their FRAP results are limited by the time resolution of 2 s,
and do not permit further characterization of the type of the
timescales down to the millisecond range, and identify the
diffusion can be furthermore linked to the observed hetero-
geneous highly dynamic PAR-2 cortex pattern, not reported
previously. Scanning FCS has the advantage of higher tem-
poral resolution over FRAP, thus forming a bridge between
the FRAP technique applicable on the timescales of seconds
and longer, and a standard FCS operating at millisecond
scales and shorter. Representing the sFCS data in terms of
spatiotemporal correlation further increases the information
obtainable with sFCS.
Other fluorescence correlation techniques could in prin-
ciple be applied to study the dynamics of the cortex-localized
proteins. For example, spatiotemporal image correlation
spectroscopy has been used to produce a map of flows from
a sequence of images (34). Our preliminary results with
spatiotemporal image correlation revealed limitations on the
spatial and temporal resolution imposed by the high transport
speeds and the available fluorescence signal in our data (28).
Scanning FCS currently appears to be the optimal approach
for the studied problem on the way between point- and im-
Another technique for studying dynamics is particle
tracking, which, however, requires identification of the fea-
tures to be tracked. While this may be successful for NMY-2,
it would be hardly possible for PAR-2 with our data, where
the brighter fluorescent areas cannot be unambiguously
identified and separated from their surroundings in subse-
quent image frames due to low contrast and imaging rate and
high noise. This is likely to be the case also for some other
PAR proteins. To reliably compare dynamics of two or more
proteins, it would be, however, preferable that the data are
obtained using the same technique. We see a possible alter-
native to tracking in a certain combination of tracking and
correlation techniques, where the image features need not be
identified, but their displacement from one image frame to
another is determined by correlation, in a way similar to
image registration techniques commonly used, for example,
in medical imaging. The feasibility of this approach applied
to a comparatively noisy fluorescence microscopy images
remains to be tested.
Diffusion in cytoplasm
The dynamics of the investigated GFP-labeled proteins in the
cytosol can be reasonably well described by free diffusion.
The observed spread of diffusion times is mainly due to the
presence of long-time fluctuations in the fluorescence inten-
sity, which affect the autocorrelation curves at long times and
measured diffusion coefficients therefore reflect the fastest
component of the motion of the proteins in the cytoplasm.
Although the solvent viscosity of cytoplasm, as measured
by rotational mobility (i.e., microviscosity), is known to be
similar to that of water, the translational diffusion of probes
of various sizes indicates considerably higher and addition-
ally size-dependent apparent viscosity. This effect has been
explained by heterogeneity and crowding of the cytoplasm,
where collisions with large structures and reversible binding
lead to effectively lower diffusion coefficient and stronger
dependence of the diffusion coefficient on the particle size,
The measured diffusion coefficients of the investigated
proteins in cytosol are all significantly smaller than the dif-
fusion coefficient of free GFP in buffer (87 mm2s?1) (37):
the fastest GFPTPH is ;113 slower, and the slowest
were not available to perform direct comparison with their
diffusion in a buffer. The simplifying assumption of a cube
5484Petra ´s ˇek et al.
Biophysical Journal 95(11) 5476–5486
root dependence of the diffusion coefficient on the molecular
mass predicts only 1.5–23 smaller diffusion coefficient of
the individual protein molecules in buffer compared to GFP.
These results indicate the role of cytoplasmatic crowding on
the motion of the investigated proteins.
Although PAR-2 and CDC-37 are of similar size, PAR-2
diffusion is approximately three times slower, indicating
possible localization in a larger complex or self-association.
Similarly, the low diffusion coefficient of NMY-2 may be a
consequence of association with other cytoplasmic compo-
nents and homodimerization. The possible involvement of
PAR-2 and NMY-2 in large protein complexes could be
linked to the mechanism by which they become localized to
We have shown that FCS and sFCS can be used to study the
dynamics of fluorescently labeled molecules on both short
and long timescales even in such a complex and dynamic
system as a polarized embryo that will divide asymmetri-
cally, by overcoming the limitations of low statistical ac-
curacy and photobleaching. Although cortex localization of
PAR-2 depends on the presence of NMY-2 (8), our data in-
dicate that PAR-2 is not recruited to the cortex by binding to
posterior localized NMY-2 patches. By using circular sFCS,
we could show that PAR-2 dynamics are faster than NMY-2,
and further that PAR-2 distribution in the cortex is not uni-
form but heterogeneous, with a highly dynamic pattern dis-
tinct from that of NMY-2. It is therefore more likely that
NMY-2 changes the properties of the cortex in a way that
PAR-2 can associate with it, and presence of PAR-2 on the
cortex might be inhibitory for NMY-2 contractility.
Circular sFCS, with its single-molecule sensitivity and full
utilization of the fluorescence signal, provides information
about the molecular dynamics and the type of motion, which
is too slow for standard FCS, and not resolvable with imag-
ing. Furthermore, sFCS provides information on spatial
correlation in addition to temporal correlation, facilitating
better characterization of transport processes in living orga-
nisms and discrimination between different models on basis
of their spatiotemporal correlation.
Measurement along a perimeter of a relatively large circle
overcomes two significant limitations encountered in FCS
when applied to slowly moving molecules: photobleaching
measurement area, and statistical noise due to the insufficient
the measurement. In comparison to imaging, higher temporal
resolution, determined by the scanning frequency, is achieved
with sFCS. Furthermore, by using two-photon excitation one
additionally benefits from the possibility of long measurement
times without disturbing the embryo development.
Future sFCS studies on other polarity proteins, along with
fluorescence microscopy and RNAi experiments, will con-
tribute to a better understanding of asymmetric cell division
in C. elegans and in other systems.
To view all of the supplemental files associated with this
article, visit www.biophysj.org.
We thank Carrie Cowan and Nathan Goehring for stimulating discussions.
C.H. was supported by an Ernst Schering Foundation Postdoctoral fellow-
ship. Funding was provided by The International Human Frontier Science
Program grant No. RGP 5-2005, and Europa ¨ische Fond fu ¨r Regionale
Entwicklung project No. 4212-06-02.
1. Cowan, C. R., and A. A. Hyman. 2004. Asymmetric cell division in C.
elegans: cortical polarity and spindle positioning. Annu. Rev. Cell Dev.
2. Schneider, S. Q., and B. Bowerman. 2003. Cell polarity and the
cytoskeleton in the Caenorhabditis elegans zygote. Annu. Rev. Genet.
3. Nance, J. 2005. PAR proteins and the establishment of cell polarity
during C. elegans development. Bioessays. 27:126–135.
4. Munro, E. M. 2006. PAR proteins and the cytoskeleton: a marriage of
equals. Curr. Opin. Cell Biol. 18:86–94.
5. Cheeks, R. J., J. C. Canman, W. N. Gabriel, N. Meyer, S. Strome, and
B. Goldstein. 2004. C. elegans PAR proteins function by mobilizing
and stabilizing asymmetrically localized protein complexes. Curr. Biol.
6. Munro, E., J. Nance, and J. R. Priess. 2004. Cortical flows powered by
asymmetrical contraction transport PAR proteins to establish and
maintain anterior-posterior polarity in the early C. elegans embryo.
Dev. Cell. 7:413–424.
7. Guo, S., and K. J. Kemphues. 1996. A non-muscle myosin required for
embryonic polarity in Caenorhabditis elegans. Nature. 382:455–458.
8. Boyd, L., S. Guo, D. Levitan, D. T. Stinchcomb, and K. J. Kemphues.
1996. PAR-2 is asymmetrically distributed and promotes association
of P granules and PAR-1 with the cortex in C. elegans embryos.
9. Cowan, C. R., and A. A. Hyman. 2004. Centrosomes direct cell
polarity independently of microtubule assembly in C. elegans embryos.
10. Cuenca, A. A., A. Schetter, D. Aceto, K. Kemphues, and G. Seydoux.
2003. Polarization of the C. elegans zygote proceeds via distinct
establishment and maintenance phases. Development. 130:1255–1265.
11. Rigler, R., and E. S. Elson, editors. 2001. Fluorescence Correlation
Spectroscopy, 1st Ed. Theory and Application, Chemical Physics
Series. Springer Verlag, Berlin.
12. Bacia, K., and P. Schwille. 2003. A dynamic view of cellular processes
by in vivo fluorescence auto- and cross-correlation spectroscopy.
13. Petrov, E. P., and P. Schwille. 2008. Standardization and Quality
Assurance in Fluorescence Measurements II: Bioanalytical and Bio-
medical Applications, Vol. 6, Springer Series on Fluorescence.
Springer, Berlin, Heidelberg, New York.
14. Petra ´s ˇek, Z., and P. Schwille. 2008. Single Molecules and Nanotech-
nology, Vol. 12, Springer Series in Biophysics. Springer, Berlin,
Heidelberg, New York.
15. Petersen, N. O. 1986. Scanning fluorescence correlation spectroscopy.
1. Theory and simulation of aggregation measurements. Biophys. J.
Protein Dynamics Measured by sFCS5485
Biophysical Journal 95(11) 5476–5486
16. Berland, K. M., P. T. C. So, Y. Chen, W. W. Mantulin, and E. Gratton. Download full-text
1996. Scanning two-photon fluctuation correlation spectroscopy: par-
ticle counting measurements for detection of molecular aggregation.
Biophys. J. 71:410–420.
17. Amediek, A., E. Haustein, D. Scherfeld, and P. Schwille. 2002.
Scanning dual-color cross-correlation analysis for dynamic co-locali-
zation studies of immobile molecules. Single Mol. 3:201–210.
18. Skinner, J. P., Y. Chen, and J. D. Mu ¨ller. 2005. Position-sensitive
scanning fluorescence correlation spectroscopy. Biophys. J. 89:1288–
19. Digman, M. A., C. M. Brown, P. Sengupta, P. W. Wiseman, A. R.
Horwitz, and E. Gratton. 2005. Measuring fast dynamics in solutions
and cells with a laser scanning microscope. Biophys. J. 89:1317–1327.
20. Petra ´s ˇek, Z., M. Krishnan, I. Mo ¨nch, and P. Schwille. 2007. Simulta-
neous two-photon fluorescence correlation spectroscopy and lifetime
imaging of dye molecules in submicrometer fluidic structures. Microsc.
Res. Tech. 70:459–466.
21. Becker, W., A. Bergmann, E. Haustein, Z. Petra ´s ˇek, P. Schwille, C.
Biskup, L. Kelbauskas, K. Benndorf, N. Klo ¨cker, T. Anhut, I.
Riemann, and K. Ko ¨nig. 2006. Fluorescence lifetime images and
correlation spectra obtained by multidimensional time-correlated single
photon counting. Microsc. Res. Tech. 69:186–195.
22. Brenner, S. 1974. Genetics of Caenorhabditis elegans. Genetics. 77:
23. Schonegg, S., A. T. Constantinescu, C. Hoege, and A. A. Hyman.
2007. The rGTPase-activating proteins RGA-3 and RGA-4 are re-
quired to set the initial size of PAR domains in Caenorhabditis elegans
one-cell embryos. Proc. Natl. Acad. Sci. USA. 104:14976–14981.
24. Audhya, A., F. Hyndman, I. X. McLeod, A. S. Maddox, J. R. Yates, A.
Desai, and K. Oegema. 2005. A complex containing the Sm protein
CAR-1 and the RNA helicase CGH-1 is required for embryonic
cytokinesis in Caenorhabditis elegans. J. Cell Biol. 171:267–279.
25. Nance, J., E. M. Munro, and J. R. Priess. 2003. C. elegans PAR-3 and
PAR-6 are required for apicobasal asymmetries associated with cell
adhesion and gastrulation. Development. 130:5339–5350.
26. Praitis, V., E. Casey, D. Collar, and J. Austin. 2001. Creation of low-
copy integrated transgenic lines in Caenorhabditis elegans. Genetics.
27. Kojro, Z. 1991. Normalization and statistical noise level in the nor-
malized autocorrelation function. Compensated normalization. J. Phys.
Math. Gen. 24:L225–L229.
28. Petra ´s ˇek, Z., C. Hoege, A. A. Hyman, and P. Schwille. 2008. Two-
photon fluorescence imaging and correlation analysis applied to protein
dynamics in C. elegans embryo. Proc. SPIE. 6860:68601L.
29. Cowan, C. R., and A. A. Hyman. 2007. Acto-myosin reorganization
and PAR polarity in C. elegans. Development. 134:1035–1043.
30. Beers, M., and K. Kemphues. 2006. Depletion of the co-chaperone
CDC-37 reveals two modes of PAR-6 cortical association in C. elegans
embryos. Development. 133:3745–3754.
31. Hurley, J. H., and T. Meyer. 2001. Subcellular targeting by membrane
lipids. Curr. Opin. Cell Biol. 13:146–152.
32. Palmer, A. G., and N. L. Thompson. 1987. Theory of sample translation
in fluorescence correlation spectroscopy. Biophys. J. 51:339–343.
33. Bacia, K., D. Scherfeld, N. Kahya, and P. Schwille. 2004. Fluorescence
correlation spectroscopy relates rafts in model and native membranes.
Biophys. J. 87:1034–1043.
34. Brown, C. M., B. Hebert, D. L. Kolin, J. Zareno, L. Whitmore, A. R.
Horwitz, and P. W. Wiseman. 2006. Probing the integrin-actin linkage
using high-resolution protein velocity mapping. J. Cell Sci. 119:5204–
35. Luby-Phelps, K. 2000. Cytoarchitecture and physical properties of
cytoplasm: volume, viscosity. diffusion, intracellular surface area. Int.
Rev. Cytol. 192:189–221.
36. Verkman, A. S. 2002. Solute and macromolecule diffusion in cellular
aqueous compartments. Trends Biochem. Sci. 27:27–33.
37. Swaminathan, R., C. P. Hoang, and A. S. Verkman. 1997. Photo-
bleaching recovery and anisotropy decay of green fluorescent protein
GFP-S65T in solution and cells: cytoplasmic viscosity probed by green
fluorescent protein translational and rotational diffusion. Biophys. J.
5486 Petra ´s ˇek et al.
Biophysical Journal 95(11) 5476–5486