You are on page 1of 11

PROTOCOL

Fluorescence correlation spectroscopy in


living cells
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

Sally A Kim1,2, Katrin G Heinze1,2 & Petra Schwille1


1Institute of Biophysics, Dresden University of Technology, Tatzberg 47-51, D-01307 Dresden, Germany. 2Present addresses: Division of

Biology, 114-96, California Institute of Technology, Pasadena, California 91125, USA (S.A.K.) and Research Institute for Molecular Pathology,
Dr. Bohr Gasse 7, 1030 Vienna, Austria (K.G.H.). Correspondence should be addressed to P.S. (petra.schwille@biotec.tu-dresden.de).

Fluorescence correlation spectroscopy (FCS) is an ideal analytical tool for studying concentrations,
propagation, interactions and internal dynamics of molecules at nanomolar concentrations in living
cells1–4. FCS analyzes minute fluorescence-intensity fluctuations about the equilibrium of a small
ensemble (<103) of molecules. These fluctuations act like a ‘fingerprint’ of a molecular species
detected when entering and leaving a femtoliter-sized optically defined observation volume created by
a focused laser beam. In FCS the fluorescence fluctuations are recorded as a function of time and then
statistically analyzed by autocorrelation analysis. The resulting autocorrelation curve yields a measure
of self-similarity of the system after a certain time delay, and its amplitude describes the normalized
variance of the fluorescence fluctuations. By fitting the curves to an appropriate physical model, this
method provides precise information about a multitude of measurement parameters, including diffusion
coefficients, local concentration, states of aggregation and molecular interactions. FCS operates
in real time with diffraction-limited spatial and sub-microsecond temporal resolution. Assessing
diverse molecular dynamics within the living cell is a challenge well met by FCS because of its single-
molecule sensitivity and high dynamic resolution3,4. For these same reasons, however, intracellular FCS
measurements also harbor the large risk of collecting artifacts and thus producing erroneous data. Here
we provide a step-by-step guide to the application of FCS to cellular systems, including methods for
minimizing artifacts, optimizing measurement conditions and obtaining parameter values in the face of
diverse and complex conditions of the living cell. A discussion of advantages and disadvantages of one-
photon versus two-photon excitation for FCS is available in Supplementary Methods online.

MATERIALS
REAGENTS EQUIPMENT
Fluorescently labeled cells to be studied plated on coverslips or Glass coverslips (borosilicate glass; Fisher Scientific) or
glass-bottom chambers chambers with a cover glass bottom (Evotec, MatTek, Nalgene,
Phenol red–free growth medium (Invitrogen) appropriate for Nunc International) that are physically flat, clean and
cell type being used corrosion resistant, free of bubbles, scratches and striations,
Buffered salt solution appropriate for cell type and controlled for high uniformity in size and thickness, and
experimental conditions (for example, HEPES-buffered saline have low or no autofluorescence (expensive quartz glass
solution (HBSS): 116 mM NaCl, 5.4 mM KCl, 0.8 mM MgSO4, coverslips are usually unnecessary; plastic coverslips are not
1.6 mM CaCl2, 20 mM HEPES and 10 mM glucose at pH 7.4) recommended)
Water (HPLC grade; Merck-EMD) tested for low fluorescence FCS rig on vibration-isolated table equipped with a
FCS detection unit and all components described in
Fluorescent dyes for calibration (for example, Rhodamine
Supplementary Table 1 online and illustrated in Figure 1
Green, Alexa Fluor 488, Rhodamine 6G, TMR
(tetramethylrhodamine; TRITC), Alexa Fluor 546, Cy5, Data analysis software for FCS curve fitting: any data analysis
Alexa Fluor 633) software (Origin MicroCal Software, SigmaPlot, MatLab)
allowing for a flexible, user defined nonlinear curve fitting (we
Purified fluorescently labeled proteins when possible
recommend a Marquardt-Levenberg nonlinear least-square
for corresponding in vitro measurements (for example,
fitting routine)
enhanced GFP)

PUBLISHED ONLINE 30 OCTOBER 2007; DOI:10.1038/NMETH1104

NATURE METHODS | VOL.4 NO.11 | NOVEMBER 2007 | 963


PROTOCOL
PROCEDURE
Protein labeling 1| Choose an appropriate fluorophore and labeling method for live-cell experiments on the target
protein. The fluorophore must be bright, photostable and not interfere with the normal function of the
protein under investigation (Box 1). The labeling fluorophore may be used for the setup adjustment and
calibration steps, although a suitable alternative is also acceptable. The purified labeling dye should be
available for diffusion studies to assess potential systematic errors induced by the label itself.

2| Label the target protein by site-specific chemical labeling or by fusion to a fluorescent protein, as
determined in Step 1.
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

Setup adjustment 3| Create a (nearly) diffraction-limited focal spot with a Gaussian point spread function in three
dimensions by epi-illumination of the objective lens with a parallel laser beam. For one-photon excitation,
the best conditions for creating an artifact-free Gaussian observation volume are given by slightly
underfilling the objective back aperture (see ref. 5). For two-photon excitation, a slightly overfilled back
aperture is recommended. No aperture restriction is necessary for two-photon excitation as it inherently
results in a sufficiently ‘clean’ Gaussian shape. Adjust the excitation laser beam diameter using a beam
expander if necessary. Optimize the position and angle of the laser by using at least two adjustable
mirrors with kinematic optical mounts in the beam path to steer the beam into the microscope.

4| Prepare 500–1,000 nM (Solution A) and 10–100 nM (Solution B) solutions of appropriate fluorescent


dye in water or in nonfluorescent buffer.

5| To adjust the excitation pathway, put 20–50 µl of Solution A in a chamber or directly on a coverslip
(for inverted setups) and place it in the sample path of the microscope. If evaporation is a concern, use
larger volumes (200–400 µl) in a closed chamber. If correctly aligned, a small, round fluorescent spot in
the center of the field of view will be visible. If the spot is not visible, appears distorted, is off-center
or becomes dimmer when the spot is moved toward the center (by steering the laser beam), shift the
excitation beam to the correct position where no angle or parallel beam displacement compromises the
point spread function.

BOX 1 CHOICE OF FLUOROPHORES AND LABELING METHOD


Properties of the fluorophores, labeling method as well as position and size of the coupled fluorophore
must be carefully considered when labeling proteins for intracellular experiments. It is critical that
the integrity of the labeled protein is maintained, that the fluorophores emit at a high efficiency, and
that the labeling does not disrupt the biochemical function or cellular localization of the protein. The
following details in particular should be taken into account.

Brightness, photostability and wavelength of excitation of the fluorophore in the context of the
application and the experimental setup. To minimize any bias of intracellular biochemical processes,
the size and inertness of the fluorophore to the intracellular environment should be considered. Red-
shifted fluorophores have the tendency to be more hydrophobic, which can cause aggregation that is
problematic for FCS and molecular labeling in general. The fluorescently labeled protein behavior must be
taken in the context of the target protein to which it will be coupled.

The method of coupling the fluorophore to the proteins. Site-specific labeling of purified proteins
with small organic fluorophores (~1 kDa) offers the benefit of minimizing possible steric hindrance that
can interfere with protein function. But expressing and purifying sufficient quantities of protein can
be laborious and time-consuming, and requires subsequent delivery of the labeled protein to cells. An
alternative is to use genetically expressible intrinsically fluorescent proteins. A major disadvantage is the
relatively large size of the fluorescent protein (~27 kDa in the monomeric form) for labeling purposes.

The position of the labeling. Standard precautions must be taken to ensure that the fluorescent protein
folds correctly to fluoresce and maintain the function of the endogenous target protein. The optimal
location of fusion as well as the length and sequence of the linker region between the fluorescent protein
and target protein must be chosen. Problems can occur if the linker is not sufficiently long (and flexible
enough) or is too long.

964 | VOL.4 NO.11 | NOVEMBER 2007 | NATURE METHODS


PROTOCOL
6| Adjust the detection pathway by placing Solution B in
Glass-bottom chamber
the microscope as in Step 5. Change the optical pathway
Observation in the microscope to send the fluorescence light to the
volume ~0.5 fl
Cell sample
(camera) port where the FCS detection unit (pinhole,
Stage Objective lens detector, correlator and computer) is mounted and start
Correction collar
the correlator. It should show a low fluorescent count rate
without, most likely, an autocorrelation curve. Adjust the
pinhole/fiber position including lens, if applicable, using
micrometer translation until the maximum fluorescence
intensity and a clear, typical FCS curve are achieved.
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

1PE or 2PE lasers ➨ TROUBLESHOOTING

7| If the objective has a correction collar, adjust it to


obtain the maximal brightness parameter η (see Step
10). Rotation of the collar adjusts the height of certain
Dichroic mirror lens elements within the objective to compensate for
Fluorescence
potentially considerable spherical aberrations introduced
by different coverglass thicknesses or refractive index
Lens 1 variations in immersion medium, when using high
numerical aperture objectives. In most cases, the focus
Pinhole
may shift, and the image of the ‘corrected’ FCS spot may
(only required for 1PE) have wandered after adjustment of the correction collar
so that readjustment of the pinhole (particularly in the
Lens 2 z orientation) is necessary after every change of the
correction collar (make only small adjustments at a time)
to ensure optimal FCS performance.
Detector Correlator
8| Replace the dye sample with water or buffer to
determine the amount of static background light. Minimize
any source of significant background.
Figure 1 | FCS setup. A parallel laser beam for one- ➨ TROUBLESHOOTING
photon excitation (1PE; for example, blue laser,
argon-ion, 488 nm) or two-photon excitation (2PE; 9| Obtain a fluorescence autocorrelation curve (10 × Calibration
infrared laser, typically a femtosecond titanium:
10s) using the reference standard for single-component
sapphire) is coupled into the microscope (red
beam path) via a dichroic mirror epi-illuminating three-dimensional Brownian motion (Solution B). Fit
the objective lens. A small, diffraction-limited the curve to a single component model equation of the
focal spot serves as the FCS observation volume in autocorrelation function G(τ):
which fluorescence is generated. Fluorescence light
–1
is collected via the same objective lens, passed G(τ ) = Neff × (1 + τ /τdiff )–1 × (1 + τ /(S 2 τ diff ) ) –1/2 ,
through the dichroic mirror (green fluorescence
path), guided through a pinhole (only for 1PE), and
where Neff is the average number of particles in the
finally projected onto a high-sensitivity detector
(single-photon counting module). Alternatively,effective measurement volume element Veff (see Step 12).
Include an additional term for triplet blinking if necessary
a fiber coupling of the detector can replace the
(see ref. 6). The structure parameter, S, characterizes the
pinhole and second lens (not shown). The digital
photon signal (transistor-transistor logic pulses;
shape of the detection volume. FCS curve fitting of the
TTL) is sent to a hardware correlator. Respective
calibration measurement should give an S value of 3–6
correlation curves are displayed and stored on a
(one-photon excitation) or 2–4 (two-photon excitation).
computer for subsequent analysis.
This S value is then fixed for the subsequent curve fitting
for the experimental session. If necessary, the system
should be readjusted to provide the appropriate S value
before continuing the experiment. The respective diffusion time, τdiff, reflects the average lateral
transit time of a molecule through the focus.
This calibration and adjustment must be repeated when using a different detection channel, changing
the fluorophore or emission filter, or adjusting any excitation and/or emission pathway alignment.
➨ TROUBLESHOOTING
▲ CRITICAL STEP

NATURE METHODS | VOL.4 NO.11 | NOVEMBER 2007 | 965


PROTOCOL

a b
1.4 2.3 0.20
2.2
1.2
2.1
2.0
1.0 0.15
1.9

G(τ )
0.8

G(τ )
0.10
0.6

0.4
0.05
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

0.2

0.0 0.00
0.001 0.01 0.1 1 10 100 1,000 0.001 0.01 0.1 1 10
τ (ms) τ (ms)

Figure 2 | Minimizing artifacts in intracellular FCS measurements. (a) Intracellular power series for determination of optimal
laser power. Sequential FCS measurements were taken at different laser powers in the same place in the same cell from lowest
to highest power. The autocorrelation curves were normalized for visual comparison of shape and τdiff. Photobleaching effects
start to appear at the highest excitation power tested (lowest attenuation, OD = 1.9) indicating the boundary of the ‘safe’
excitation power range for making intracellular measurements without artifacts for this molecule at this wavelength in these
cells. (b) Autocorrelation of cellular autofluorescence in hippocampal neurons. Immature hippocampal neurons (3 d in vitro)
cultured in medium containing phenol red show fibrous-looking autofluorescence that produces a robust autocorrelation curve
(black) when excited at 870 nm. The fit is shown in red.

10| Determine the brightness parameter η (commonly called the counts per molecule per second
or CPM measured in kHz) by dividing the average count-rate by Neff. This can be used to assess the
quality of the adjustment of the system.
Simple subtraction of background counts from the count rate before calculating CPM values is not
recommended. For typical hardware correlator–based FCS setups, both the count rate and correlation
amplitude are affected by background. Therefore, background correction on the hardware correlated
data would bias results as the correlation amplitude (1/Neff) has already been adjusted by background,
thereby indicating a higher-than-‘true’ particle number. If necessary, the setup should be readjusted to
obtain maximal CPM values. Repeat Step 9.

11| Determine the lateral 1/e2 beam dimensions, ro, from τdiff, and known diffusion coefficient, D, of
the standard by ro2 = 4Dτdiff. A classic calibration standard is Rhodamine 6G with D = 2.8 × 10–6 cm2/s
(see ref. 6). Using ro as a calibration value, the diffusion coefficient of an unknown probe of interest
can then be calculated from its measured τdiff.

12| Determine the axial 1/e2 beam dimension zo and the effective volume Veff based on the fit
parameters of the fluorescent standard (Step 11) by zo = S ro and Veff = π3/2 ro2 zo.
For an alternative calculation for two-photon excitation and other details, see Supplementary Methods.

Delivery of 13| For intracellular measurements, deliver the fluorescently labeled molecule to cells (see
fluorescent Supplementary Methods for delivery methods). Transfection of fusion proteins of biological molecules
molecules to cells often produces intracellular concentrations in the nanomolar range, ideal for FCS.

Optimizing 14| Make FCS measurements of the purified labeled component in solution before performing FCS in
intracellular FCS cells. Determine measurement parameters (diffusion, triplet blinking and brightness) of the molecule’s
measurements behavior to guide expectations for intracellular experiments and to troubleshoot potential problems in
the cell.
If necessary, pretreat the glass with a blocking agent (for example, 10% BSA) for 10 min followed by two
short washes in buffer to avoid adsorption and aggregation effects.
➨ TROUBLESHOOTING

966 | VOL.4 NO.11 | NOVEMBER 2007 | NATURE METHODS


PROTOCOL
15| Perform an intracellular power series, 1.25 Rhodamine green in solution
Alexa 488 calmodulin in solution
compare the respective FCS curve fits Alexa 488 calmodulin in HEK293 cells
and select an optimal laser power that 1.00 DiI in HEK293 membrane
is high enough for a sufficient photon
efficiency η to accurately fit the curves 0.75

G (τ)
τ
while staying below intensities that
cause photobleaching or damage the cell 0.50
during the measurement time for FCS data
acquisition. Depending on the FCS setup, 0.25
laser source, power, fluorescent dye and
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

biological application, CPM can vary; 0.00


1,000–5,000 CPM for cellular experiments
0.001 0.01 0.1 1 10 100 1,000
under low power is typical (see ref. 6) and τ (ms)
has been shown to be sufficient for reliable
curve fitting. But even with lower values, Figure 3 | FCS measurements in cells. A representative FCS measurement
of Alexa-labeled calmodulin taken both in solution and in the cytoplasm
as small as 500 CPM, autocorrelation
of a HEK293 cell. For additional comparison a lipid analog (DiI, 1, 1′-
curves may still be clearly resolved in an dioctadecyl-3, 3, 3′,3′-tetramethylinocarbocyanine perchlorate) diffusing
experiment if longer acquisition times in the plasma membrane of a HEK293 cell is shown. These measurements
can be applied, and when bleaching and demonstrate the dynamic range of molecular mobilities accessible by FCS.
temporal resolution are not an issue. The A calibration curve of Rhodamine green is also shown. The autocorrelation
goal is to ascertain a ‘safe’ excitation curves have been normalized for visual comparison. Each curve is an
power range where key parameters (τdiff average of six 10-s acquisitions.
and particle number N) do not depend on
excitation power. To establish the appropriate conditions, take all measurements in the same cell to avoid any
influence of biological heterogeneity (Fig. 2a).
The result of a power series is only valid for the specific excitation wavelength at which it was determined.
➨ TROUBLESHOOTING

16| Check for autofluorescence (typical count rates) in unloaded cells and characterize the behavior
(correlated or uncorrelated, stable or unstable; Fig. 2b). Ideally, a selected cell line will have no or only a very
small amount of autofluorescence that is not correlated (Supplementary Methods).
➨ TROUBLESHOOTING
▲ CRITICAL STEP

BOX 2 DIFFUSION MODELS FOR FCS MEASUREMENTS IN CELLS


Simple diffusion of a single species in a homogeneous medium is adequately described by a single
diffusion coefficient. A small, inert molecule, such as Alexa Fluor 488 hydrazide, which does not interact
with the intracellular environment, can be adequately described by a single diffusion coefficient. But
native biological molecules in complex environments such as the inside of a cell typically cannot be
described by a simple diffusion coefficient. Intracellular data are therefore normally fit to either a
multiple diffusion coefficient (Option A) or anomalous diffusion model (Option B).

Option A. The multiple component diffusion model describes a situation that has two or more
independently diffusing species, each of which can be described by a single diffusion coefficient. In
general fitting should be constrained to the least number of components possible, normally two unless
other components are known and can be fixed for curve fitting. The addition of extra components does
not necessarily provide more insight into the biological meaning of the data, even if it yields a better fit
because of the larger number of free parameters.

Option B. An alternative for fitting complex translational diffusion data is to use the anomalous diffusion
model, which is based on the idea that the mean-square displacement is not linearly proportional to
time but instead follows a power law represented by tα: 〈r 2 〉 = Γ t α , where Γ is the transport coefficient
and α is the time exponent6. The transport coefficient Γ is qualitatively similar to diffusion coefficients,
and α gives a measure of the degree the motion is restricted by the environment. Smaller values of α
correspond to either increased binding or a higher concentration of obstacles (for example, organelles
and cytoskeleton) in the diffusion path. Potential mechanisms of anomalous diffusion are binding and
collisions with mobile as well as immobile obstacles and subresolution confinement.

NATURE METHODS | VOL.4 NO.11 | NOVEMBER 2007 | 967


PROTOCOL
FCS measurements 17| Obtain a fluorescence image of the cells containing the fluorescently labeled protein and select the
in living cells desired measurement position. If no such imaging feature is available, cells should be viewed with bright-field
illumination, and the measurement position visualized by the fluorescence spot.

18| Choose cells manually by moving the x-y stage to select for a particular measurement position.
Avoid cells with very high expression as well as those with subcellular localizations with bright fluorescent
aggregates or large vacuoles.

19| Establish measurement conditions such that both fluorescence intensity and correlation amplitudes
remain constant within a tolerance of 10% throughout the measurement. Multiple runs (5–10) of short
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

measurement times (10 s) have been found to be optimal for intracellular cytoplasmic measurements (Fig. 3).
Ten seconds is typically sufficient time to generate a reasonable FCS curve, and multiple measurements allow
data to be discriminated if photobleaching or sudden perturbations occur during the measurement. Longer
measurement times are necessary for slowly moving populations, such as large proteins or complexes and
lipids in membranes (Fig. 3). Begin acquisition immediately after opening the shutter for FCS illumination.
Close the shutter between delayed FCS runs.
➨ TROUBLESHOOTING

Analysis of 20| Sort FCS curves from single runs individually to determine whether they meet the designated criteria
intracellular FCS (for example, less than 10% photobleaching throughout the acquisition time, no large spikes in the
measurements count rate owing to aggregation, and no differences larger than twice the standard deviation in shape or
amplitude of the curve). Those curves that do not meet the established criteria should not be taken into
account for further analysis (for 10 × 10 s measurement, no more than three runs should be discarded).

21| Fit FCS curves with the appropriate physical model (one-component, two-component, or anomalous
diffusion in either three or two dimensions with blinking dynamics when appropriate) with a Marquardt-
Levenberg nonlinear least-squares fitting routine (see Supplementary Methods for alternative fitting
routines). Various appropriate models should be tried to fit the data. If two different models yield similar fits,
select the simpler model. Intracellular data typically cannot be described by a single diffusion coefficient and
require either a multiple-diffusion coefficient or an anomalous diffusion model (Box 2).
➨ TROUBLESHOOTING
▲ CRITICAL STEP

TROUBLESHOOTING TABLE
PROBLEM SOLUTION
Step 6 No FCS curve is obtained. For in vitro measurements, there could be problems with the setup.
Make sure the laser shutter is open and the laser is operating
properly. Make sure the optical filters and dichroic are compatible
with the dyes. Check the alignment of the detection unit (pinhole or
fiber position and lens).
Step 8 A static high background is This is likely due to photon counts from room-light or excitation-
observed even when not measuring in the light leakage to the detector. Use a dark shroud, or cover sources of
sample. light or leakage spots. For static background, use a correction factor.
Using brighter fluorophores or higher labeling ratios can increase
the signal-to-noise above the background. Make sure there is no
autofluorescence or scattering from the buffer, sample holder or
cover glass.

Correlating fluorescence is measured in For in vitro measurements, this may be due to trace fluorescence
(unlabeled) control samples. impurities in the diluent. Make all samples in water quality
controlled for low fluorescence. Considerable autofluorescence may
also be caused by unlabeled sample targets or reagents. Remove any
reagent that contributes to the autofluorescent signal. If necessary,
use low fluorescence reagents.

968 | VOL.4 NO.11 | NOVEMBER 2007 | NATURE METHODS


PROTOCOL
TROUBLESHOOTING TABLE (CONTINUED)
Step 9 Extraneous fluctuations are Instability of the setup (beam intensity, laser pointing, jitter in
observed in the count rate. mirrors or galvanometers used for beam positioning, electrical
noise, mechanical instability in the stage) can cause oscillations
in the autocorrelation curve. Carefully characterize the setup and
eliminate all sources of instability. To detect unwanted fluctuations
potentially induced by instabilities of the laser, the count rate
provided by the correlator software can be used to monitor the
fluorescence of a highly concentrated (micromolar) dye solution
excited with lower laser intensity. For in vitro measurements, the
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

fluorescently labeled molecules may aggregate in solution. Some


dye molecules and fluorescently labeled proteins need to be made
fresh for every experiment to avoid aggregation, but some dyes
are very stable and can be used repeatedly without problems. Dyes
are best stored at 4 °C to avoid freeze-thaw cycles. Solutions can
be filtered or centrifuged in a microcentrifuge at 14,000 r.p.m. for
30 min at 4 °C to remove aggregates. Aggregation of fluorescently
labeled molecules can occur in living cells and may be due to
overexpression. Cell movement may be another source of extraneous
fluctuations. To alleviate this problem, stop any perfusion during
FCS data acquisition, give cells a sufficient recovery period after
electroporation or microinjection or perform viability tests on the
cells to determine problems of cell health.

Average detector count rate is too high This is likely due to too many fluorescent molecules because
(>1 MHz) throughout the FCS measurement. the concentration of fluorescent probe in the sample is too
high. For in vitro measurements, make serial dilutions of your
sample, and test for optimal count rates and FCS curves. For
intracellular measurements in conjunction with electroporation
or microinjection, dilute the stock concentration before delivery.
Additionally, unlabeled species (nonfluorescent competitors) can be
added. For transfection experiments, simultaneous cotransfection
with the empty vector may help reduce the concentration.

S value for dye solution is incorrect. Either the fitting model used is inappropriate and/or the
excitation-detection light path, particularly the pinhole or the
correction collar of the objective lens, is misaligned. Be sure the
laser epi-illumination is ideal (centered, no angle distortion) and
the correction collar reflects the thickness of the cover glass in
use. Additionally, readjust the pinhole position in x-y direction
so that the CPMs and the amplitude of the autocorrelation
curve are maximal. Re-check fitting parameters; if they are still
unsatisfactory, scan the z position of the pinhole and choose the
position where the CPM and S values are optimal. If not successful,
choose another fitting model and/or retry the adjustment
procedure. If the S value is too low, check whether the excitation
laser beam diameter is too small. If the S value is too high, check
the following: (i) the objective correction collar may be in the
wrong position, not reflecting the thickness of the cover glass in
use, (ii) the x-y-z position of the pinhole, fiber or detector may be
incorrect. Check alignment and scan the position along the optical
axis for optimizing FCS parameters, and (iii) the dye solution is
prepared inappropriately so that aggregation or adsorption effects
have occurred.

Step 14 No FCS curve is obtained. See Troubleshooting Step 6

There are extraneous fluctuations and See Troubleshooting Step 9


excessively high count rates.

Average detector count rate is See Troubleshooting Step 9


too high during the FCS measurement.

NATURE METHODS | VOL.4 NO.11 | NOVEMBER 2007 | 969


PROTOCOL
TROUBLESHOOTING TABLE (CONTINUED)
Step 15 Significant photobleaching of Determine optimal illumination dose (dependent on photostability
the fluorophore is observed during the and mobility of the fluorophores). Decrease laser power, use
measurement. two-photon excitation and/or more photostable fluorophores.
If molecules are too immobile for FCS, particularly in cells,
use alternate techniques, such as fluorescence recovery after
photobleaching (FRAP) or image correlation spectroscopy (ICS) to
assess mobility.

Cell damage is observed due to the Determine optimal illumination dose. Decrease laser power, use two-
stationary parked laser beam. photon excitation (>900 nm) and/or a shorter acquisition time. Close
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

the laser shutter in between delayed FCS runs. Use shortest data
acquisition time needed.

Photobleaching diminishes with repeated For intracellular measurements, this is due to possible higher- and
measurements. lower-mobility populations. Ensure that the extent of the low-
mobility fraction is appreciated by using very low illumination
intensity while choosing the measurement position or, if available,
by positioning the beam based on an image obtained in the scanning
mode. Acquisition should begin immediately with the start of the
FCS illumination. The mobility of the faster diffusing fraction may be
assessed after the ‘prebleaching’ is completed. The mobility of the
slow fraction may need to be assessed by FRAP, if it is inaccessible
by FCS.

Intracellular measurements yield This is most likely background due to autofluorescence or scattering.
significantly lower molecular brightness Check carefully for autofluorescence and scattering signal of unlabeled
values. control cells at the same laser intensity as experimental measurements.
The background tends to lower the amplitude of the autocorrelation
function. If the background cannot be circumvented by different
treatment of the cells (see above), use the following formula to correct
for measured particle numbers (see equations 5 and 6 in ref. 7).

Intracellular measurements yield artificially Photobleaching limits the observation time and causes cumulative
increased τdiff of molecules accompanied dye depletion in restricted subcellular compartments. Decrease laser
by a concurrent increase in the correlation power. Perform an intracellular power series to determine optimal
amplitude. laser power.

Intracellular measurements yield Nonspecific interactions of the fluorescent probe with the
erroneously long τdiff of molecules. intracellular environment (for example, membranes) can slow down
the labeled molecules. Use different fluorophore labels that tend to
be more hydrophilic and therefore are less likely to have nonspecific
intracellular interactions.

Step 16 There is correlating High basal autofluorescence can cause correlating fluorescence in
autofluorescence. intracellular measurements. Use of two-photon excitation with longer
infrared wavelengths (preferably over 900 nm) and/or red-shifted
dyes can help minimize intracellular autofluorescence problems.
Always maintain cells in phenol red–free medium for at least 12 h
before the experiment. Use a cell type with less autofluorescence,
such as HEK293 cells. For the case of primary cell culture, change
culture conditions and check whether autofluorescence is dependent
on days in vitro for primary cell culture. Use brighter fluorophores,
higher concentrations and/or a smaller detection volume.
Step 19 No FCS curve is obtained. For intracellular measurements, this could be due to poor
transfection or electroporation. Check intracellular fluorescence
on a fluorescence imaging system. Fluorescence should be easily
detectable even at low-nanomolar concentrations.
See also Troubleshooting Step 6.

There are extraneous fluctuations and See Troubleshooting Step 9


excessively high count rates.

970 | VOL.4 NO.11 | NOVEMBER 2007 | NATURE METHODS


PROTOCOL
TROUBLESHOOTING TABLE (CONTINUED)
There are photobleaching problems and See Troubleshooting Step 15
incorrect τdiff values.

Step 21 FCS data analysis yields poor fits Sources of this problem may be due to (i) significant unaccounted-
and/or nonsensical parameter values. for photobleaching, (ii) the fit converging to a local minimum with
nonsensical parameter values, (iii) selection of an inappropriate
diffusion model with an unaccounted-for triple fraction (molecular
blinking) or (iv) afterpulsing effects of the detector. First, check for
unaccounted photobleaching (Step 15). For fitting data, provide start
values based on a visual examination of the FCS curve or known values.
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

Fix the S value to the value determined by calibration measurements


and account for any offset from very slow fluctuations. Use a different
or modified diffusion model (including an exponential term for
blinking), if necessary. Last, check for afterpulsing effects from the
detector, which can be clearly identified as a sharp spike at very fast
time scales in the correlation curve (see Supplementary Table 1). If
afterpulsing is present, exclude the first affected data points (typically
2–6) in the FCS curve before curve fitting. If the time scale of
afterpulsing interferes with the time scale of dynamics to be studied,
split the single fluorescent emission into two detection channels and
cross-correlate the signal. This removes afterpulsing effects in the
FCS curve as afterpulsing of two independent detectors will be non-
identical and nonsynchronized (and therefore uncorrelated).

CRITICAL STEPS
Step 9 Calibration is one of the most crucial steps for FCS measurements as quantification of mobility
and binding dynamics depend on the accuracy of these procedures. Note that for data fitting, S and
τdiff are dependent parameters as they both describe the shape of the curve decay in concert. If S is
overestimated or set to higher values, the results for the respective τdiff as a floating parameter tend
to be underestimated by curve fitting and vice versa.

Step 16 Autofluorescence and static background issues can cause large problems for intracellular FCS
measurements. It is important to characterize and optimize the cellular system, including all reagents
and buffers, by performing control FCS measurements before the start of the experiment in cells.

Step 21 FCS data analysis is critical for extracting quantitative parameters from the data. Nonlinear
curve fitting is prone to systematic errors and can generate results of no biological relevance.
Selecting the best mathematical model and reducing the number of free (floating) parameters at
the start of the fit by exploiting parameter values that are known for the system helps optimize this
step. Meaningful start values and boundary conditions are also of crucial importance, particularly
when using a multiple-component model for intracellular measurements. For cellular measurements,
statistical analysis of different FCS curves and fit results for various experiments performed on
different days help to validate the significance of results.

COMMENTS
FCS offers both high spatial and temporal resolution to study biological processes so that mobility and
binding kineticswhich can also be assessed by fluorescence cross-correlation spectroscopy6 (FCCS;
Box 3)can not only be studied between cells but also can be mapped through different compartments of
a single cell. It is unlike many other techniques in which the heterogeneity of a process in a cell population
or within the same cell is lost, because only an average value is determined for a seemingly ‘homogeneous’
ensemble. Investigating heterogeneity, however, is biologically important to understand the extent of
variation in cellular responses. Examples of intracellular applications of FCS include studying directed
transport processes8, counting signaling molecules9, assessing intracellular diffusion in vivo10 and
understanding the dynamics of membrane structure11. Based on such work, FCS holds great promise in
deepening our understanding of molecular dynamics in cells.
The technique of fluctuation analysis underlying the idea of FCS requires two key features of the
molecular system under investigation—low concentrations (in the nanomolar range) and sufficient

NATURE METHODS | VOL.4 NO.11 | NOVEMBER 2007 | 971


PROTOCOL

BOX 3 FLUORESCENCE CROSS-CORRELATION SPECTROSCOPY


Dual-color FCCS, or fluorescence cross-correlation spectroscopy, is a powerful multicolor extension
of FCS that probes the interaction of two differently labeled molecular species with higher precision
than single-color FCS. Instead of correlating the fluorescence signal only with itself (FCS), in FCCS the
fluorescence fluctuations of two spectrally distinct detection channels are also correlated with each other
(‘cross’-correlation). The high specificity of this technique arises from the fact that a cross-correlation
curve is only formed if the differently labeled molecules of interest are bound and therefore co-diffuse
through the FCCS observation volume, and only then will their fluorescence fluctuations correlate in
time. By analyzing the amplitude and decay time of respective curves, FCCS can be used to determine
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

binding constants and mobility of the bound complex as well as the concentrations of all of the species.
A more detailed protocol for making FCCS measurements is available6.

ADVANTAGES OF FCCS
FCCS has been shown to provide the following advantages over FCS for analyzing molecular interactions.

Realization of binding studies for molecular binding partners of similar size and mobility.
Diffusion, as assessed by single-color FCS, is not a very sensitive parameter to monitor binding (τdiff
varies with the hydrodynamic radius and therefore only with the inverse cube root of the molecular mass
for a spherical assembly). Therefore, for detection in a Brownian motion–driven system, the observed
molecular species must be bound to targets of considerably larger size to reliably resolve the binding
kinetics.

Higher dynamic range for determining dissociation constants. In contrast to small changes in τdiff
resulting from binding, the amplitude change in a respective cross-correlation curve is proportional to
the fraction of molecules bound.

Less influence by diffusion driven artifacts. Intracellular diffusive behavior can change for many
reasons, including nonspecific interactions and aggregation, but a change in the cross-correlation
curve is only dependent on the amount of co-diffusing doubly labeled molecules and is therefore highly
specific.

Less influence by the nonlinear fitting routines and model chosen. Tracking binding by measuring
changes in diffusion, particularly for intracellular measurements, normally requires the use of complex
fitting models. Often these models are a crude approximation of the complex reality of intracellular
diffusion, particularly for a molecule with many binding partners. In FCCS, the binding interaction of
interest can be determined directly by the amplitude of the cross-correlation curve, which is a more
robust, stable fitting parameter.

DISADVANTAGES OF FCCS
Disadvantages of FCCS versus FCS are mainly the higher complexity of the molecular assay (dual-color
labeling) as well as of the experimental setup (two emission channels, two laser lines for excitation when
using one-photon excitation). FCCS data, however, must be routinely checked for potential artifacts and
systematic errors resulting from non-ideal experimental conditions, such as dye cross-talk, chromatic
aberrations and autofluorescence.

mobility (D = ~10–6–10–9 cm2/s). High concentrations of fluorophores in the focal volume result in a high
average fluorescence with small relative contributions from each molecule. This makes FCS impossible
because the main parameter of interest is not the fluorescence intensity itself but the deviation from the
mean. Use of FRAP can be used for highly concentrated systems. Molecules must also be mobile for FCS
measurements to induce a diffusion-related decay of the autocorrelation. Slower-diffusing or transported
molecules require a longer acquisition. In practice, photobleaching is the most evident problem. Although
it is possible to deconvolve the immobile fraction from FCS curves that have evidence of photobleaching,
this complicates the analysis. Instead, alternate techniques, such as FRAP and anisotropy, can be used in
conjunction with FCS, particularly for the detection of immobile molecules.
FRAP3 is complementary to FCS because it is applicable to very slowly moving but highly concentrated
fluorescent particles. The lower fluorophore concentrations and laser powers in FCS are advantageous

972 | VOL.4 NO.11 | NOVEMBER 2007 | NATURE METHODS


PROTOCOL
for minimizing probe as well as photochemical and heating artifacts. ICS12, an extension of FCS, is a
conceptually similar technique, which correlates spatially pixel by pixel for a given fluorescent image,
temporally for a stack of imagesor both. In contrast to FCS, ICS allows for determination of molecular
density and concentrations for both living (mobile) and fixed (immobile) samples. A promising recent
development is the spatiotemporal version of ICS (STICS)13 that can also provide directional information
(a velocity vector) for fluorescently labeled proteins. This ‘velocity’ or ‘vector mapping’ of molecules is
particularly powerful for studying intracellular transport of proteins. Along with this technique, there are
robust filtering algorithms available to get rid of the dominant immobile population in living cells, when
looking at the mobile fraction. The spatial resolution of the resulting map mainly depends on the signal to
noise ratio for the given experiment. The time resolution for STICS is lower than for FCS and limited by
© 2007 Nature Publishing Group http://www.nature.com/naturemethods

the speed of image acquisition. This drawback can be overcome by an extension of ICS called raster image
correlation spectroscopy, or RICS, that uses standard laser scanning confocal microscopy in the raster-scan
mode14, but this strategy has not been extended to STICS as of yet. Single-particle tracking15, including
single-dye tracing, is an alternative technique for analyzing diffusion behavior and is mostly used in two-
dimensional (membrane) systems. Since single-particle tracking directly analyzes the behavior of single
molecules, it yields more detailed information about the heterogeneity of the system and higher spatial
resolution. This technique, however, requires more time-consuming acquisitions and offline analysis,
whereas FCS features a more robust signal through its inherent averaging over many single molecule events.
Note: Supplementary information is available on the Nature Methods website.

ACKNOWLEDGMENTS
We thank K. Bacia for critical reading of the manuscript and valuable comments, and E. Haustein, K. Bacia, M.N. Waxham, D.R. Larson
and W.R. Zipfel for helpful and constructive discussions. This work was supported by grants of the Fulbright Kommission to S.A.K., the
German Ministry for Education and Research to P.S. and Human Frontiers to P.S.

Published online at http://www.nature.com/naturemethods/


Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions

1. Rigler, R. & Elson, E.L. (eds.) Fluorescence Correlation large transporting complex is transported via kinesin in
Spectroscopy: Theory and Applications (Springer, Berlin, squid giant axons. Cell 103, 141–155 (2000).
2001). 9. Cluzel, P., Surette, M. & Leibler, S. An ultrasensitive
2. Thompson, N.L. Fluorescence correlation spectroscopy. in bacterial motor revealed by monitoring signaling proteins
Topics in Fluorescence Spectroscopy, Volume 1: Techniques. in single cells. Science 287, 1652–1655 (2000).
(Lakowicz, J.R., ed.) 337–378 (Plenum, New York, 1991). 10. Yao, J., Munson, K.M., Webb, W.W. & Lis, J.T. Dynamics
3. Zipfel, W.R. & Webb, W.W. In vivo diffusion measurements of heat shock factor association with native gene loci in
using multiphoton excitation fluorescence photobleaching living cells. Nature 442, 1050–1053 (2006).
recovery and fluorescence correlation spectroscopy. in 11. Bacia, K., Scherfeld, D., Kahya, N. & Schwille, P.
Methods in Cellular Imaging. (Periasamy, A. ed.) 216–235 Fluorescence correlation spectroscopy relates rafts in
(Oxford, New York, 2001). model and native membranes. Biophys. J. 87, 1034–1043
4. Schwille, P. Fluorescence correlation spectroscopy and (2004).
its potential for intracellular applications. Cell. Biochem. 12. Petersen, N.O., Höddelius, P.L., Wiseman, P.W., Seger, O.
Biophys. 34, 383–408 (2001). & Magnusson, K.-E. Quantitation of membrane receptor
5. Hess, S.T. & Webb. W.W. Focal volume optics and distributions by image correlation spectroscopy: concept
experimental artifacts in confocal fluorescence correlation and application. Biophys. J. 65, 1135–1146 (1993).
spectroscopy. Biophys. J. 83, 2300–2017 (2002). 13. Hebert, B., Costantino, S. & Wiseman, P.W. Spatiotemporal
6. Bacia, K. & Schwille, P. Practical guidelines for dual- image correlation spectroscopy (STICS) theory,
color fluorescence cross-correlation spectroscopy. Nat. verification, and application to protein velocity mapping
Protoc., published online 1 November 2007 (10.1038/ in living CHO cells. Biophys. J. 88, 3601–3014 (2005).
nprot.2007.410). 14. Digman, M.A. et al. Measuring fast dynamics in solutions
7. Schwille, P., Haupts, U., Maiti, S. & Webb, W.W. Molecular and cells with a laser scanning microscope. Biophys. J. 89,
dynamics in living cells observed by fluorescence 1317–1327 (2005).
correlation spectroscopy with one- and two-photon 15. Qian, H., Sheetz, M.P., & Elson, E.L. Single particle
excitation. Biophys. J. 77, 2251–2265 (1999). tracking. Analysis of diffusion and flow in two-
8. Terada, S., Kinjo, M. & Hirokawa, N. Oligomeric tubulin in dimensional systems. Biophys. J. 60, 910–921 (1991).

NATURE METHODS | VOL.4 NO.11 | NOVEMBER 2007 | 973

You might also like