1051
Journal of Cell Science 108, 1051-1062 (1995)
Printed in Great Britain © The Company of Biologists Limited 1995
Mapping of adherens junction components using microscopic resonance
energy transfer imaging
Zvi Kam*, Tova Volberg and Benjamin Geiger
Department of Chemical Immunology, The Weizmann Institute of Science, Rehovot 76100, Israel
*Author for correspondence
SUMMARY
Quantitative microscopic imaging of resonance energy
transfer (RET) was applied for immunological high resolution proximity mapping of several cytoskeletal components of cell adhesions. To conduct this analysis, a microscopic system was developed, consisting of a highly stable
field illuminator, computer-controlled filter wheels for
rapid multiple-color imaging and a sensitive, high resolution CCD camera, enabling quantitative data recording
and processing. Using this system, we have investigated the
spatial inter-relationships and organization of four
adhesion-associated proteins, namely vinculin, talin, αactinin and actin. Cultured chick lens cells were double
labeled for each of the junctional molecules, using fluorescein- and rhodamine-conjugated antibodies or phalloidin.
RET images were acquired with fluorescein excitation and
rhodamine emission filter setting, corrected for fluorescein
and rhodamine fluorescence, and normalized to the fluorescein image. The results pointed to high local densities of
vinculin, talin and F-actin in focal adhesions, manifested by
mean RET values of 15%, 12% and 10%, respectively. On
the other hand, relatively low values (less than 1%) were
observed following double immunofluorescence labeling of
the same cells for α-actinin. Double indirect labeling for
pairs of these four proteins (using fluorophore-conjugated
antibodies or phalloidin) resulted in RET values of 5% or
lower, except for the pair α-actinin and actin, which yielded
significantly higher values (13-15%). These results suggest
that despite their overlapping staining patterns, at the level
of resolution of the light microscope, the plaque proteins
vinculin and talin are not homogeneously interspersed at
the molecular level but form segregated clusters. α-Actinin,
on the other hand, does not appear to form such clusters
but, rather, closely interacts with actin. We discuss here the
conceptual and applicative aspects of RET measurements
and the implications of the results on the subcellular
molecular organization of adherens-type junctions.
INTRODUCTION
actions in adherens junctions. For example, integrin was shown
to bind to both talin (Horwitz et al., 1986) and α-actinin (Otey
et al., 1990), and cadherin was shown to bind to catenins
(Kemler and Ozawa, 1989). Among the plaque proteins,
vinculin was shown to interact with several proteins including
talin (Burridge and Mangeat, 1984), paxillin (Turner et al.,
1990) and α-actinin (Belkin and Koteliansky, 1987), and
possibly with itself (Bendori et al., 1989). α-Actinin, also,
binds to another junctional protein, zyxin (Crawford and
Beckerle, 1991), as well as to actin filaments (Baron et al.,
1987). AJ contain, in addition, actin capping proteins such as
tensin (Butler and Lin, 1989) and a multitude of enzymes
(mostly kinases) which are known to participate in the transmembrane transduction of growth and differentiation signals
(Geiger et al., 1992).
Mainly on the basis of these biochemical data, a general
structural model was offered for adherens junctions (e.g.
Burridge et al., 1988; Geiger et al., 1990), though the actual
molecular topology and substructure of these sites was not
directly demonstrated experimentally. This is mainly due to
Adherens-type junctions (AJ) are a family of cell-cell and cellmatrix adhesions which are characteristically associated with
the actin-based cytoskeleton through a multimolecular submembrane plaque (Geiger and Ginsberg, 1991). The molecular
composition and diversity of AJ have been extensively studied
during the last several years. It was shown that essentially all
AJ share some common components such as actin, α-actinin
and vinculin, while other junctional molecules, notably the
plaque proteins talin, paxillin, plakoglobin and catenins, are
selectively associated with either cell-matrix or cell-cell
adhesions (Geiger et al., 1985, 1990). The transmembrane
receptors of AJ, namely members of the integrin and cadherin
families (mediating contacts with the matrix or with neighboring cells, respectively), show further tissue-specific heterogeneity which reflects their individual binding specificities
(Geiger and Ayalon, 1992; Burridge et al., 1988; Hynes, 1992).
In vitro binding studies with isolated junctional proteins
have provided some insight into the possible molecular inter-
Key words: resonance energy transfer, vinculin, talin, α-actinin,
actin, focal contact, cell adhesion
1052 Z. Kam, T. Volberg and B. Geiger
the fact that the currently available methods fail to resolve
the molecular architecture of these sites. Fluorescence
microscopy, which is the most commonly used technique for
visualizing AJ components, has a rather poor resolution of
0.2 µm at best. Immunoelectron microscopy, on the other
hand, has an excellent spatial resolution, yet the stoichiometric representation of the cellular antigens, using the stateof-the-art methodologies (namely, the number of immunogold particles per antigen molecule), is insufficient to provide
reliable molecular information (see Discussion). Since
cellular ultrastructure is still not amenable for high resolution
analytical methods (like crystallography), novel approaches
for high resolution molecular mapping are essential for
studying the molecular structure and assembly of adhesion
sites in situ.
The approach applied here for the study of AJ organization
is fluorescence resonance energy transfer (RET). This
approach has long been utilized as a spectroscopic ruler for
measurements of nanometer-scale proximities using visible
light (Förster, 1946, 1948, 1959, 1964; Stryer and Haugland,
1967; Stryer, 1978; Stryer et al., 1982; Kuhn, 1982). The phenomenon occurs due to electron orbital dipole-dipole coupling
between an excited donor and adjacent acceptor fluorophores
whose respective emission and absorption spectra overlap. A
characteristic distance between the dyes (r0), typically in the
range of 4-6 nm (Stryer and Haugland 1967; Stryer 1978),
defines the scale factor in the 6th power, which determines the
inverse relationships between RET probability and the
distance (for review see Lakowicz, 1983). RET provides,
therefore, a sensitive measure for inter- or intra-molecular
proximity (Stryer 1978), rendering it a useful tool for examination of the structure of molecular complexes including
membranes (Fernandez and Berlin, 1976; Yguerabide and
Förster, 1981; Szöllösi et al., 1984; Johnson et al., 1984; Jovin
and Arndt-Jovin, 1990; Uster and Pagano, 1986; Kubitscheck
et al., 1991; Wang et al., 1993), cytoskeletal filaments (Taylor
et al., 1981), chromatin (Jovin, 1979; Arndt-Jovin and Jovin,
1989; Cardullo et al., 1988; Ludwig et al., 1992), and multimeric enzymes (Adams et al., 1991). Moreover, RET proved
to be most useful also for the study of the conformation and
folding of proteins in which specific residues were modified
by the acceptor and donor dyes (Beechem and Haas, 1989;
Gettins et al., 1990).
For many years RET measurements were primarily carried
out on molecules in solution using fluorimeters. The application of RET for in situ measurements in microscopic specimens
became practical only recently due to the improved sensitivity
and accuracy of fluorescence digital imaging and processing
(Tsien, 1989; Jovin and Arndt-Jovin, 1989; Herman, 1989;
Kam 1987; Adams et al., 1991). Jovin and Arndt-Jovin
employed an approach for microscopic RET imaging based on
photobleaching decay rates of the donor and the integrated
number of fluorescent photons emitted. RET in single cells was
also used for measurement of kinase subunit interaction as a
function of cAMP levels (Adams et al., 1991).
We present here the background for direct RET measurement based on immunomicroscopy. Using this system we have
explored the intermolecular relationships between several
cytoplasmic constituents of focal contacts, namely vinculin,
talin, α-actinin and actin.
MATERIALS AND METHODS
Immunochemical reagents
Primary antibodies used in this study were as follows: monoclonal
anti-human vinculin (hVin1) was obtained from Sigma Immunochemicals (St Louis, USA); polyclonal rabbit antibodies (R694) were
prepared against pure chicken vinculin. Monoclonal antibodies to αactinin were from Sigma Immunochemicals (clone BM-75.2) and
polyclonal antibodies were prepared in rabbits; both raised against the
chick gizzard protein. Anti-talin antibodies were either monoclonal
(clone 27.2) or rabbit antibodies kindly supplied by Dr K. Burridge
(U. North Carolina, Chapel Hill) and by Dr M. Beckerle (U. Utah,
Salt Lake City). Fluorescent secondary antibodies (goat antibodies to
mouse or rabbit IgG), conjugated to either fluorescein isothiocyanate
(FITC) or tetramethyl rhodamine isothiocyanate (TRITC) were
purchased from Jackson Laboratories (West Grove, USA). Double
conjugated goat anti-mouse IgG, containing comparable levels of fluorescein and rhodamine, were prepared by mixing the commercial
rhodamine-labeled antibodies (3 mg/ml), in 0.2 M carbonate buffer,
pH 9.0, and at room temperature, with a 10-fold molar excess of
dichlorotriazinyl amino fluorescein (DTAF, Research Organics,
Cleveland, USA). The double conjugated molecules were separated
from free dye by gel filtration on Sephadex G-50 (Pharmacia,
Sweden). The molecules selected for immunolabeling contained an
average of 2 moles of each fluorophore per mole of antibodies as
determined spectroscopically from the 280, 495 and 575 nm absorptions. FITC- and TRITC-conjugated phalloidin were purchased from
Molecular Probes (Eugene, Or).
Immunolabeling of cultured cells
Cultured chicken lens cells were prepared as previously described
(Volk and Geiger, 1984) from 7- to 8-day old embryos and immunofluorescently labeled at 30-50% confluence. Cells were simultaneously permeabilized and fixed (for 5 minutes) in 3% paraformaldehyde solution containing 0.5% Triton X-100 in 50 mM MES buffer,
pH 6.0, and then post-fixed for an additional 30 minutes in 3%
paraformaldehyde. This procedure was selected to optimize preservation of antigenicity of junctional proteins and the structure of focal
adhesions.
For immunolabeling, fixed cells on cover slides were inverted on
antibody droplets (cells facing down) and incubated at room temperature for 30 minutes. Following extensive wash with phosphate
buffered saline (PBS) the cells were incubated for 20 minutes with
the secondary, fluorescently conjugated, anti-IgG, at an appropriate
dilution, rinsed again and mounted in 90% glycerol in 0.1 M Tris
buffer, pH 8.0. Antifade agents were not added to the mounting
medium, since they could interfere with RET efficiency. Primary
antibody concentrations used in this study ranged between 20 and 100
µg/ml, depending on the antibody. The secondary antibodies were
usually used at a concentration of 5 µg/ml. Fluorescent phalloidin was
used for labeling at saturating concentrations, as determined directly
by titration, followed by measurement in the digital microscopic
system.
Digital microscopic system
Images were recorded with a Zeiss Axiomat microscope coupled to a
cooled CCD camera (CC/CE 200, Photometrics, Tucson, AZ). The
computerized microscope system (previously described by Hiraoka et
al., 1991; Kam et al., 1993) included fast computer-controlled step
motors driving the excitation and emission filter wheels (Compumotor, Petaluma, CA). The filters used here were bandpass interference
filters DF type (to minimize cross talk) excitation: 488
(bandwidth=30) nm and 580(25) nm; emission: 535(50) nm and
623(40) nm for fluorescein and rhodamine imaging, respectively, and
a matching quadruple-wavelength dichroic mirror (DAPI, FITC,
Texas Red and Cy5; purchased from Omega Co., Brattleboro, VT).
Resonance energy transfer imaging 1053
To reduce the bleed-through between fluorescein and rhodamine
signals, the filters used for rhodamine imaging were the ones originally recommended for Texas Red recording.
For RET analysis three images were consequently recorded. The
first two employed the conventional fluorescein (F) and rhodamine
(R) channels and the third used the fluorescein excitation and the
rhodamine emission filters (denoted here as X channel). As the images
detected through the three filter settings are often different in their
brightness, we adjusted the length of exposure times to keep the
number of measured photons (and thus the noise) roughly constant.
To avoid changes in the profile of light illumination intensity we have
used a fiber optics illumination scrambling system (Kam et al., 1993).
Furthermore, for each channel a homogeneous layer of fluorescent
solution was imaged as a reference for variations in the illuminated
field profile, and all acquired images were pixel-by-pixel normalized,
thus eliminating effects of residual inhomogeneities in illumination as
well as in the CCD response (Hiraoka et al., 1987).
Despite the high optical quality of the filters we found small lateral
shifts between images recorded through the different filters. They
were determined using 0.1 µm Nile Red fluorescent beads (L-5198
FluoSpheres, Molecular Probes, Eugene, OR, USA). To correct these
shifts, images were aligned using the Simplex algorithm (Amoeba
routine taken from Numerical Recipes, Press et al., 1986). Correlations (or r-factor) between two images were maximized following
high-pass filtering and Hanning. Displacements were found reproducible within +0.1 pixel and depended only on the emission filter.
For ×100/1.3 NA Axiomat objective (0.047 µm/pixel) the displacements between images recorded through the fluorescein and
rhodamine emission filters are:
∆x = 2.5 ± 0.1, ∆y = −1.7 ± 0.1 pixels.
Similar correlation analysis for a focal series at 0.2 µm steps did
not point to a measurable shift in focus between fluorescein and
rhodamine images.
Image processing
Due to partial overlap between the excitation and emission spectra of
the donor (fluorescein) and acceptor (rhodamine) fluorophores, the
image recorded through the X channel contains, in addition to RET,
contributions from the donor and acceptor direct fluorescence. These
contributions can be corrected (see Trón et al., 1984) if the fluorescein and rhodamine fluorescence images are measured, and the bleedthrough coefficients for the particular filter sets are known. To experimentally determine these coefficients, chicken lens cells were labeled
for vinculin by antibodies with one fluorophore only (fluorescein or
rhodamine) and recorded through the F, R and X channels. Following
alignment, ratio images were calculated for all pixels above a
threshold level setting at about 30% of the maximum image intensity.
Average background was evaluated at the excluded area below the
threshold, and subtracted. For focal contacts, signal to background
ratio was usually higher than 20:1, and the results were thus largely
insensitive to the precise threshold level. Ratio images were uniform
within the regions above threshold with typical variations of less than
2%. The average ratio between images taken with two filter sets gives
the bleed-through coefficient between the respective channels. The
only significantly large coefficients are the contribution of fluorescein
and rhodamine to the X channel (termed S1 and S2 by Trón et al.,
1984), and the bleed-through of fluorescein sample to the R channel
(termed S3). The values obtained are:
S1 = 0.061 ± 0.005, S2 = 0.01 ± 0.008, S3 = 0.01 ± 0.008.
These values agree with the overlap integrals of the transmission
spectra of the filters used in the system, the fluorophore excitation and
emission spectra, and a typical emission spectrum of the 100 W
mercury arc lamp used.
RET images (IRET) were calculated by applying equation (4) of
Trón et al. (1984) to every pixel in images I1,I2 and I3 (see below),
the intensity of which was higher then the respective signal threshold
levels. The equation can be approximated by:
αE
(I2 − S2I3)
(I2 − S2I3 − S1I1)
IRET = ––––– = –––––––––––– − S1 ≈ –––––––––––––– .
(1 − E) (1 − S2S3/S1)I1
I1
(1)
Thus:
IRET
E = ––––––––– ,
α + IRET
(2)
where E is energy transfer efficiency, I1, I2 and I3 are the images
recorded through the F, X and R channels, respectively, following displacement of I1 (see above) and subtraction of the average background
at regions where pixel intensities were below background threshold
levels (selected 10-20% below the signal threshold levels). The factor
α depends on overlap spectra of the filters and the fluorophores
absorption and emission. The value of α was experimentally determined from the ratio of fluorescence intensities of the same number
of fluorescein and rhodamine dye molecules measured through the X
channel (Trón et al., 1984; see below). This enables to relate the RET
values calculated here to the absolute efficiencies, E, of fluorescence
energy transfer event. Under simplified configurational assumptions,
and if the donor and acceptor transition dipoles orientation factor is
taken as 2/3, the value for random orientation (see Dale and Eisinger,
1975), E can be related to distances. The numerator (I2− S2I3 − S1I1)
in this relation is the net fluorescent intensity due to energy transfer
evaluated from the image of the X channel after subtracting the bleedthrough contributions from fluorescein and rhodamine fluorescence as
evaluated from the F and R channel images. In addition, bleaching of
fluorescein must be taken into account, since exposures in the X
channel may be as long as several seconds. For the excitation intensities used, fluorescein signal bleaches 20% within about 5 seconds.
We therefore approximate the F image by the average between two
recordings; first before and second after the X and R channel
exposures. Rhodamine bleaching during the exposures is less than
2%, and was thus neglected. Analysis was repeated for all pairs of
labeled molecules in 3-5 independent experiments, each consisting of
3-10 images. Data are thus averaged for hundreds of focal adhesions
in more then 10 cells. For each image the averaged values and a
histogram were computed. Processing was also done on small
windows at different regions of the same image.
Direct measurement of α
A 2 µl sample of labeled antibodies (1.4 µg/ml, 3 fluorophores/
antibody molecule) was dried on a slide, the area measured at low
magnification (1× objective and ×0.5 magnification factor for the
microscope-CCD coupling optics) and the average fluorescence intensities measured in 25 fields within each drop using 20× objective,
giving the measured fluorescence intensities per fluorescein and
rhodamine dye molecule (IF, IR in the F and R channels, respectively).
The ratio of fluorescence intensity between the X and F channels, for
fluorescein, and the X and R, for rhodamine-labeled antibodies (RF
and RR) was measured, yielding:
α = (IRRR)/(IFRF) = 0.24 ± 0.04.
The error is dominated by the variations in fluorescence intensity
within the dried drop area.
RESULTS
Microscopic resonance energy transfer (RET) in
immunofluorescently double-labeled cells
To determine the dependence of microscopic RET on the
molecular proximity between antibody-bound fluorophores we
1054 Z. Kam, T. Volberg and B. Geiger
have compared RET values as a function of antibody concentrations using 3 distinct immunolabeling procedures.
(1) Labeling of cells with a monoclonal anti-vinculin
antibody followed by a secondary antibody which is doubleconjugated to both fluorescein and rhodamine. Since both fluorophores are bound to the same molecule, it is expected that
most of the RET in the labeled cells will be contributed by
intramolecular (rather than inter-molecular) energy transfer,
and thus will be largely insensitive to antibody dilution (Fig.
1A).
(2) Labeling of cells with a monoclonal anti-vinculin
antibody, followed by a mixture of fluorescein- and
rhodamine-labeled anti-mouse IgG antibodies. In such samples
only intermolecular interactions contribute to RET, though the
labeled secondary antibodies could be associated either with
the same primary antibody, or with antibodies bound to
different vinculin molecules (Fig. 1B).
(3) Labeling of cells with a mixture of two distinct primary
anti-vinculin antibodies (monoclonal, mouse antibody; and a
polyclonal, rabbit antibody) followed by fluorescein-labeled
anti-mouse IgG and rhodamine-labeled anti-rabbit IgG antibodies (or vice versa). In this experiment RET may occur only
between secondary antibody molecules that are bound to
different primary antibodies. When the primary antibodies are
used at sub-saturation levels it is conceivable that each fluorophore represents a different vinculin molecule (Fig. 1C).
Antibody saturation concentrations were established experimentally by serial dilutions. For example, for vinculin the
intensity of fluorescence labeling reached a plateau at an
antibody concentration of about 30 µg/ml, yielding 4×106
counts/µm2 per second, corresponding to approximately 4×104
fluorophore molecules. This number of fluorophores at saturation per focal contact unit area is comparable to the estimated
number of molecules according to Kreis et al. (1982, 1985).
To reduce the chances of two antibodies bound to the same
antigen molecule (in self-RET experiments) primary antibody
concentration used were selected to yield fluorescence levels
of 20-30% below saturation (defined here as subsaturation).
The saturation is less critical for hetero-RET measurements.
It should be emphasized that RET values, described in this
work, are defined as relative energy transfer intensities, not
absolute efficiencies, E. RET dependence on secondary
antibody dilution is presented in Fig. 1. It is shown that
dilution of the double-conjugated antibodies causes an initial
decrease in the RET calculated according to Eq. (1) from 28%
to 16% (Fig. 1A), probably due to the loss of the intermolecular transfer, and then reaches a plateau, due to RET between
donor and acceptor fluorophores present on the same antibody.
In contrast, RET values in cells double labeled with antibodies, each conjugated to a different fluorophore (Fig. 1B,C),
were highly sensitive to secondary antibody dilution (especially to the dilution of the antibody conjugated to the
acceptor-fluorophore). Although this argument is not rigorous,
due to the non-additive nature of RET, it is noteworthy that
the maximal RET values of about 13% obtained when donor
and acceptor fluorophores were conjugated to different
secondary antibody molecules (linked either to the same or to
different primary antibodies Fig. 1B,C) were comparable to
what we interpret as the intermolecular RET (about 16%)
obtained with the double-conjugated antibody (see plateau in
Fig. 1A). It was thus concluded that RET values indeed
represent the local neighborhood of the fluorophore-conjugated antibodies and hence may provide information on the
average distances between (or density of) adjacent molecules
of the respective antigens.
Molecular organization of vinculin, talin, α-actinin
and actin in AJ and along stress fibers: self-RET
images
To obtain information concerning the local densities of each of
the different junctional molecules (see below) we have labeled
cells for each of these proteins using both mouse and rabbit
primary antibodies, followed by a mixture of the respective fluorophore-conjugate secondary antibodies. As a rule we have
examined the RET values at subsaturating levels of the two
primary antibodies, to avoid binding of two different primary
antibodies to the same antigen molecule, and saturating levels
of the secondary antibodies. Double labeling for actin was
50
RET (%)
Fig. 1. The dependence of image-averaged
A
B
C
RET values on the concentration of
fluorophore-conjugated secondary
antibodies. Cells were incubated for a fixed
10
time with different dilutions of the
fluorescent antibody solution.
(A) Dependence of RET values on the
dilution of the double-conjugated (to both
fluorescein and rhodamine) secondary
antibody (II. F/R α M) applied after labeling
I. M α vin
I. M α vin+ R α vin
I. M α vin
1
of cells with monoclonal anti-vinculin
II. F α M + R α M
II. F α M + R α R
II. F/R α M
primary antibody (I. M α vin).
0.5
0
20
40
60
20
40
60
20
40
60
(B) Dependence of image-averaged RET
values on dilution of the donor- or dilution
ANTIBODY DILUTION FACTOR
of the acceptor-fluorophore (rhodamine or
fluorescein)-conjugated secondary anti-mouse antibodies mixture (II. F α M + R α M) applied following the monoclonal anti-vinculin primary
antibody labeling (I. M α vin). Open squares, dilution of fluorescein-conjugated antibodies, keeping rhodamine-conjugated antibodies
undiluted. Filled squares, dilution of rhodamine, keeping fluorescein undiluted. (C) Dependence of image-averaged RET values on the dilution
of the flurescein-conjugated anti-mouse Ig (open squares) or the dilution of the rhodamine- conjugated anti-rabbit Ig (filled squares) antibodies
(II. F α M + R α R) in the secondary antibody mix. Cells were labeled first with a mixture of monoclonal and rabbit polyclonal anti-vinculin
primary antibodies (I. M α vin + R α vin).
Resonance energy transfer imaging 1055
carried out using a mixture of fluorescein- and rhodamine-conjugated phalloidin.
The patterns of double fluorescent labeling and the calcuR
RET
α-ACTININ
TALIN
VINCLUIN
ACTIN
F
lated RET images in pseudo-color for actin, vinculin, talin and
α-actinin are shown in Fig. 2 and summarized in Table 1. Factin, labeling with fluorescein- and rhodamine-labeled phal-
% RET
24
12
Fig. 2. Double labeling of chick lens cells for one of the four proteins: actin, vinculin, talin and α-actinin (self-RET). Actin double labeling was
carried out by mixture of fluorescein- and rhodamine-conjugated phalloidin. Immunolabeling for the last three proteins (rows designated,
respectively) was carried out with mixture of the mouse and rabbit primary antibodies (see Materials and Methods) followed by fluoresceinlabeled goat anti-mouse Ig (column designated F) and rhodamine-labeled goat anti-rabbit Ig (R). The RET images calculated from Eq. (1) in
Materials and Methods are shown in the column designated RET. The pseudo color scale is linear in RET values (24% full scale). Bar, 10 µm.
1056 Z. Kam, T. Volberg and B. Geiger
Table 1. Summary of the self-RET and hetero-RET values
between double-immunolabeled focal contact components
actin, α-actinin, vinculin and talin
ACTIN-ACTIN
3500
Acceptor
Actin
α-Actinin
Vinculin
Talin
10±5
13±4
3±2
1±1
5±2
1±1
4±2
6±2
3±2
4±2
14±3
5±2
1±1
6±2
5±2
12±4
0
VINCULIN-VINCULIN
The table contains the histogram modal RET values (± half peak width) for
5-10 images.
loidin (Fig. 2 actin/F and R, respectively), yielded essentially
identical and largely uniform patterns both along stress fibers
and at their focal contact-associated termini. RET values, on
the other hand, were quite variable, ranging from 5 to 15%
(Fig. 3). Characteristically, the highest RET values were
observed at or close to the termini of actin cables, where association with the junctional plaque occurs. These regions did
not, necessarily, coincide with those showing highest phalloidin staining. RET between anti-actin antibodies and phalloidin was comparable to RET between fluorescein- and
rhodamine-labeled phalloidin.
Vinculin immunolabeling (Fig 2, vinculin) was most
prominent in focal contacts and intercellular AJ, with high
RET values of up to 25% (average value was 14%, as shown
in Fig. 3). Careful examination indicated that nearly all focal
contacts as well as intercellular AJ (not shown), regardless of
their size or subcellular location, displayed similar RET values.
We did, however, detect non-uniform RET levels within individual focal contacts, manifested, for example, by the fine red
streaks (see Fig. 2) which indicate RET values of 20% or
higher. These streaks were aligned parallel to the direction of
the bundle. It should be mentioned that switching the fluorophores conjugated to the secondary antibodies had no
apparent effect on the results of self-RET.
Talin double labeling gave consistently somewhat lower
RET values compared to vinculin (average of about 12% and
nowhere reaching local values as high as those of vinculin),
despite the fact that the overall labeling intensity for talin was
comparable or higher. It was also noted that central areas of
focal contacts exhibited lower RET (green) compared to the
periphery (yellow).
In contrast to actin, vinculin and talin, self-RET values for
α-actinin were very low, typically below 1%. Again, the low
values could not be attributed to a low labeling intensity, since
the immunostaining was similar to that of the other proteins.
Spatial inter-relationships between different
junctional proteins
Further experiments were conducted to determine the apparent
proximity of pairs of the different AJ-associated molecules
(Figs 4, 5). The relationships of vinculin, talin and α-actinin to
actin were examined using immunolabeling in conjunction
with fluorescent phalloidin. As shown in Fig. 4, and quantitatively by the histograms in Fig. 6, both vinculin and talin
displayed low RET values with actin (below 3%). The actinvinculin values were somewhat higher than those obtained for
NUMBER OF PIXELS
Donor
Actin
α-Actinin
Vinculin
Talin
1000
0
TALIN-TALIN
750
0
α−ACTININ-α−ACTININ
1500
0
0
5
10
15
20
25
RET (%)
Fig. 3. Histograms of self-RET values for actin, vinculin, talin and
α-actinin obtained for the four images of Fig. 2. The histograms
present the distributions of RET values in the analyzed images.
actin-talin, yet the significance of these differences is not clear.
It should be emphasized that these low values were obtained
also when the donor and acceptor labeling was switched,
despite the extensive overlap between the areas labeled with
the two fluorophores, represented by the yellow areas in the
superimposed images (Fig. 4, F/R). In contrast, α-actinin-actin
RET values were very high when the former molecule was
labeled with the donor fluorophore, reaching an average value
of about 13%. RET was significantly lower (about 5%) when
α-actinin was labeled with the acceptor fluorophore, most
likely due to the vast excess of actin in these sites (see Discussion). The highest RET values for these two proteins were
locally detected in association with striations along the stress
fibers.
The analysis carried out to determine the cross inter-relationships between vinculin, talin and α-actinin is shown in Fig.
5 and, quantitatively, in Fig. 6 (see also Table 1). RET values
between these proteins were relatively low, of the order of 5%.
DISCUSSION
There is now growing awareness of the fact that cellular
processes occur in a highly specific structural context (for
review see Ingber, 1993). Thus the assembly of supramolecular structures such as enzyme-substrate or receptor-ligand
complexes, as well as cytoskeletal or karyoskeletal networks,
depends on highly specific and accurate subcellular targeting
and compartmentalization. Elucidation of cellular events at the
molecular level requires, therefore, detailed information not
only on the expression and post-translational processing of the
Resonance energy transfer imaging 1057
relevant components but also on their subcellular distribution
at the highest possible resolution and sensitivity.
The most common approach used for subcellular localization of molecules in, or on, cells is immunofluorescence
microscopy, which has a spatial resolution of approximately
0.2 µm, namely, more than one order of magnitude larger than
the size of most globular proteins. Thus, information about
packing patterns and fine molecular organization is not readily
attainable by light microscopy (LM). The alternative experimental approach, namely electron microscopy (EM), has an
intrinsic resolution comparable to the size of globular proteins
and may be combined with immunochemical labeling for the
identification of specific molecules in cells at an approximate
resolution of about 200 Å (limited by the size of the antibodies
and the gold particles). However, a major limitation of
immuno-EM is the low ratio between the electron-dense
particles resolved and the number of antigen molecules
present. This is usually attributed to EM sample preparation
procedures, including strong fixation and ultrathin sectioning.
For example, in our previous studies (Volberg et al., 1986),
indirect immunolabeling of smooth muscle for vinculin and
talin yielded approximately 500 gold particles per µm2 of the
dense plaques. Similarly, immuno-EM labeling of wet-cleaved
focal contacts (Brands et al., 1990; Feltkamp et al., 1991)
yielded maximal values of 400 particle per µm2. This value is
almost two orders of magnitude lower than the estimated
density of vinculin molecules in focal contacts (Kries et al.,
1985) and therefore the immunolabeling of EM samples cannot
faithfully represent the distribution of vinculin.
Immunofluorescence labeling, on the other hand, appears to
be much more efficient, reaching at saturation fluorophore
densities comparable to those of vinculin. We therefore
propose that the application of RET may yield valuable information on intermolecular proximity and submicroscopic
organization of AJ components. As shown, the regions with
high RET values do not, necessarily, coincide with those displaying highest staining levels, in line with the claim that RET
does not merely report on the average local concentration but
rather on distributions at a higher resolution than that attainable by conventional light microscopy. It should be clarified
that despite the high intrinsic resolution attainable by RET, the
images obtained are still defined within the regular spatial resolution of the light microscope (namely, ≈ 0.2 µm). Nevertheless, the intermolecular proximity scale attainable by this
technique, as represented by the RET value at each pixel, may
be about one order of magnitude better.
The relationships between RET values and actual distance
may be subject to several physical and topological considerations.
(a) The critical distance r0, which falls in the range of 3545 Å for randomly oriented fluorescein-rhodamine pair, defines
the effective range of RET, since the probability of transfer
declines sharply when interfluorophore distance significantly
exceeds r0, due to the (r0/r)6 dependence. While this value has
a marginal effect in indirectly, double-labeled specimens, due
to the large size of the antibodies (see below), it might have a
major effect on self-RET obtained between fluorescein- and
rhodamine-labeled phalloidin, which are small molecules that
bind to actin in a 1:1 molar ratio and place the fluorophores
right at the grooves of the actin molecules (Wulf et al., 1979;
Faulstich et al., 1983). Based on the structural model of actin
(Kabsch et al., 1990) and its binding sites (Pollard and Cooper,
1986), the distance between fluorophores along the filament, at
saturation, is about 50 Å and 1/2 of these are expected to be
fluorescein-rhodamine pairs. Interestingly, we found that RET
values close to the termini of bundles were higher than along
the filaments, suggesting that F-actin in these regions may be
more tightly packed than along the bundle, resulting in interfilament RET.
(b) The lengths of the antibody arms, connecting the fluorophore to the antigen, are considerably longer than r0 and thus
may have an over-riding effect on the dependence of RET on
the distances between the antigen molecules. For example,
indirect immunolabeling may place the fluorophores at a
maximal distances of ≈200 Å from the antigen, namely 5-6
times r0. It should, however, be emphasized that most of the
fluorophores are expected to be located closer to the antigen,
that there are 2-3 fluorophores bound per secondary antibody
and that there might be several labeled antibodies attached to
each of the primary antibodies. The configurational possibilities of secondary antibody-conjugated fluorophores can be
simulated by donors and acceptors each attached at the ends of
two-segment arms with flexible hinges, and anchored at the
other ends, again via flexible rotating hinges. RET dependence
on the separation between the two anchor points can be
estimated by averaging on arm angles. For 100Å arm length,
at a separation of d0=90 Å, RET decays to half of its maximal
value. Compared to a donor-acceptor pair at distance r0, RET
probability is expected to be enhanced by having multiple fluorophores per primary antibody molecule, but smaller due to
possible 3-dimensional distribution, according to: n2(r0/d0)3 ≈
0.5. The rather small compromise on resolution and RET probability is apparently compensated by a significant increase in
flexibility of labeling configuration and labeling intensity. It is
conceivable that the use of Fab fragments and of direct labeling
(rather than indirect labeling with intact antibodies) might
somewhat improve the resolution, though it is still not clear
whether the availability of only about one fluorophore per
antigen at saturation (rather than several, as described in this
study) will yield RET values that will be too low for accurate
density evaluations. Ultimately, one may conjugate fluorophores to the molecules of interest and microinject them into
cells for RET analysis. This approach has the attractive
potential of reporting molecular interactions and probing
dynamic molecular processes in vivo. However, in several preliminary attempts, the RET signals obtained, using such
approach, were very low. The reason for that is not clear yet,
though it could be due either to the dilution of the microinjected labeled molecules in the endogenous unlabeled pool, or
to the difficulty, encountered by rigidly bound fluorophores, of
approaching each other at the right distance and orientation.
(c) Another major factor affecting RET is the saturation
level of antigenic sites, namely the molar ratio between antigen
molecules and antibodies bound to them. Oversaturation of
primary polyclonal antibodies may result in binding of more
then one antibody to the same molecule, which could artificially raise the self-RET values. Under-saturation of labeling
may similarly be misleading, since it may erroneously suggest
large distances between the respective antigenic molecules. In
principle, undersaturation may stem from limited accessibility,
partial loss of antigenicity, or the use of insufficient amounts
of primary and/or secondary antibodies. While the effect of the
1058 Z. Kam, T. Volberg and B. Geiger
first two variables cannot be readily evaluated, the latter factor
can be experimentally controlled. In this study, we have
measured the intensity of labeling versus primary antibody
ACTIN-TALIN
α-ACTININ-ACTIN
F/R
R
F
ACTIN-VINCULIN
concentration, determined the saturation levels and used
antibody concentrations below this level.
The dilution experiment relates to this issue. It was found
RET
% RET
24
12
Fig. 4. Double labeling of chick lens cells for pairs of actin with vinculin, talin or α-actinin. Actin was labeled using fluorescein-conjugated
phalloidin in conjunction with monoclonal antibodies to vinculin and talin (first and second columns), or rhodamine-phalloidin in conjunction
with monoclonal anti-α-actinin (third column). The former two samples were then labeled with rhodamine-conjugated anti-mouse IgG and the
latter with the fluorescein-conjugated secondary antibody (for details, see Materials and Methods). The upper two rows show the fluorescein (F)
and rhodamine (R) channels and the third depicts the superimposed images of the two. The bottom row shows the RET image. The pseudo
color scale is linear in RET values (24% full scale). Bar, 10 µm.
Resonance energy transfer imaging 1059
that RET values declined following dilutions of both donorand acceptor-conjugated antibodies. This is in contrast with
homogeneous solutions of small dyes, where RET efficiency
α-ACTININ-VINCULIN
α-ACTININ-TALIN
RET
F/R
R
F
TALIN-VINCULIN
reflects the distances between acceptor molecules neighboring
each donor fluorophore and is therefore independent of donor
concentration (e.g. Herman, 1989). A possible explanation for
% RET
24
12
Fig. 5. Double labeling and RET analysis of chick lens cells for the other three pairs of the four focal adhesion molecules. The antibodies used
were as follows (from left to right): talin (left column), and α-actinin (middle column)-specific monoclonal antibodies in conjunction with
rabbit anti-vinculin, followed by fluorescein anti-mouse IgG and rhodamine anti-rabbit IgG. The cells shown in the right column were labeled
with rabbit anti-α-actinin and monoclonal anti-talin followed by rhodamine anti-mouse IgG and fluorescein anti-rabbit IgG. The upper two
rows show the fluorescein (F) and rhodamine (R) channels and the third depicts the superimposed images of the two. The bottom row shows
the calculated RET image. The pseudo color scale is linear in RET values (24% full scale). Bar, 10 µm.
1060 Z. Kam, T. Volberg and B. Geiger
ACTIN-VINCULIN
750
0
ACTIN-TALIN
2000
0
NUMBER OF PIXELS
α-ACTININ-ACTIN
2500
0
TALIN-VINCULIN
1500
0
α-ACTININ-VINCULIN
1000
0
α-ACTININ-TALIN
2000
0
0
5
10
15
RET (%)
20
25
Fig. 6. Histograms of RET values presenting the distributions in the
RET images shown in Figs 4-5 for the six possible pair-labeled
samples of the four proteins, α-actinin, talin, vinculin and actin.
this discrepancy is that at high antibody concentrations, close
to saturation levels, most of the antigenic sites are occupied by
antibody molecules. In contrast, at low concentrations, the
probability for binding of two antibodies at nearby sites may
be unfavorable due to steric hindrance, which is reflected by a
decrease in RET with the concentrations of antibodies coupled
to either fluorophore.
In view of the above considerations, and the uncertainty
about antigen-fluorophore ratio and topology, we do not
attempt to propose exact intermolecular distances based on the
measured RET values. It is, however, worth noting that unlike
the situation in solutions (which is a 3-dimensional system) or
in membranes (which is essentially 2-dimensional) fluorophore
average density cannot be estimated for our data, since information on the volume of the labeled structure is limited by the
resolution along the optical axis of the light microscope (≈ 0.5
µm). Nevertheless the relative values of self-RET and RET
between the same antibodies to two distinct antigens can distinguish between random distribution of the target molecule
and the presence of intermolecular clusters. The fact that selfRET values were high for vinculin, talin and actin and very
low for α-actinin, despite their comparable labeling intensity,
is interpreted as indicating the presence of clustered domains
of the first three, and dispersion of the latter. Low RET values
between all pairs of vinculin, talin and actin imply that focal
contacts are composed of segregated clusters of these proteins.
The size of such clusters is not known, yet it could be roughly
estimated based on the ratio between self-RET and heteroRET. This ratio reflects the ratio between the size of the
domains occupied by each protein, and the extent of the
boundary between domains where antibodies against one
protein may closely approach antibodies against the second
protein. Taking the antibody arm length as 20 nm, and the self
to hetero ratio of vinculin and talin as 4, the estimated diameter
of vinculin and talin domains is about 80 nm.
The high RET values between α-actinin and actin support the
notion that the primary interaction of α-actinin is indeed with
actin filaments. An intriguing question is related to the low selfRET values obtained for α-actinin, which is known to be a
dimer, and thus may bind pairs of antibodies at close range. One
possible explanation is that the binding of one antibody
molecule interferes with the binding of the second. However,
even if the efficiency of labeling was high enough to allow most
or even all dimers to be doubly labeled, and half of them bind
a donor and acceptor pair, the expected RET value might still
be low compared to RET from many neighboring molecules.
Apparently, the dimer-to-dimer spacing in situ is too high to
allow for such inter-molecular RET to occur. How do these
findings relate to the various structural models proposed for the
molecular organization of focal contacts? Based mainly on the
binding between isolated junctional molecules, these models
predict that the intrinsic transmembrane adhesion molecules,
integrins, bind to the microfilament system through a complex
chain of molecular interactions. It has been shown that two
cytoplasmic components of focal contacts, talin and α-actinin,
can bind to the cytoplasmic tail of integrin. Talin might further
bind to vinculin, which interacts with α-actinin and through it
with actin (Burridge et al., 1988; Geiger and Ginsberg, 1991).
Alternatively, α-actinin was shown to be capable of directly
interacting with integrin, thus linking it to actin (Turner et al.,
1990; Burridge et al., 1988). These linear models are challenged
by the present findings, which suggest that the submembrane
plaque is dominated by submicroscopic patches, each containing multiple copies of either vinculin or talin. The presence of
vinculin clusters is in line with previous studies, which showed
that vinculin has a tendency to oligomerize through its Cterminal tail (Milam, 1985; Molony and Burridge, 1985), as
well as with the demonstration that binding of vinculin to the
plaque is a highly cooperative process (Bendori et al., 1989). It
has been argued (Geiger et al., 1990) that cell-substratum
adhesion sites are composed of three laminated domains: a
transmembrane region with specific adhesion receptors, mostly
of the integrin family, a vinculin-containing junctional plaque
and an actin-containing cytoskeletal domain. It had been further
shown that actin, together with α-actinin, can be dissociated
from the plaque by severing proteins such as fragmin, without
affecting the organization of vinculin (Avnur et al., 1983), suggesting that the two belong to distinct subcellular domains each
with close intermolecular packing (see Bendori et al., 1989).
This notion is supported by the present results, which show low
RET values between vinculin antibodies and actin-bound phalloidin.
Resonance energy transfer imaging 1061
In summary RET, based on indirect antibody labeling, is not
only simple to apply, but also yields measurable fluorescence
intensities which reflect submicroscopic patterns of assembly
of cellular structural proteins. Future analysis along the lines
set in this work will be used to probe differences in the
assembly of the various junctional components, as well as their
organizational rearrangements during cell spreading, locomotion and growth stimulation.
The authors thank Drs J. W. Sedat and D. A. Agard for a continuous fruitful interaction. We acknowledge with gratitude support for
equipment from the Foundation Mordoch Mijan, the Volkswagen
Foundation through Joseph Cohn Center, and grants by the ICRF, the
Minerva Foundation and the NCRD-DKFZ joint research program.
The paper was written while B.G. was a scholar-in-residence at the
Fogarty International Center for advanced Study in the health science,
National Institutes of Health, Bethesda MD. B.G. is the E. Neter
Professor for Cell and Tumor Biology; Z.K. is the Pollak Professor
of Biophysics.
REFERENCES
Adams, S. R., Harootunian, A. T., Buechler, Y. J., Taylor, S. S. and Tsien,
R. (1991). Fluorescence ratio imaging of cyclic AMP in single cells. Nature
349, 694-697.
Arndt-Jovin, D. J. and Jovin, T. M. (1989). Fluorescence labeling and
microscopy of DNA. In Fluorescence Microscopy of Living Cells in Culture,
part B. Meth. Cell Biol. (ed. D. L. Taylor and Yu-Li Way), vol. 30, pp. 417448. Academic Press, NY.
Avnur, Z., Small, J. V. and Geiger, B. (1983). Actin independent association
of vinculin with the cytoplasmic aspect of the plasma membrane in cellcontact areas. J. Cell Biol. 96, 1622-1630.
Baron, M. D., Davison, M. D., Jones, P. and Critchly, D. R. (1987). The
sequence of chick α-actinin reveals homologies to spectrin and calmodulin.
J. Biol. Chem. 262, 17623-17629.
Beechem, J. M. and Haas, E. (1989). Simultaneous determination of
intramolecular distance distributions and conformational dynamics by global
analysis of energy transfer measurements. Biophys. J. 55, 1225-1236.
Belkin, A. M. and Koteliansky, V. E. (1987). Interaction of iodinated
vinculin, metavinculin and α-actinin with cytoskletal proteins. FEBS Lett.
220, 291-294.
Bendori, R., Salomon, D. and Geiger, B. (1989). Identification of two distinct
functional domain on vinculin, involved in its association with focal
contacts. J. Cell Biol. 108, 2383-2393.
Burridge, K. and Mangeat, P. (1984). An interaction between vinculin and
talin, Nature 308, 744-746.
Burridge, K., Fath, K., Kelly, T., Nuckolls, G. and Turner, C. (1988). Focal
adhesions: transmembrane junctions between the extracellular matrix and the
cytoskeleton. Annu. Rev. Cell Biol. 4, 487-525.
Butler, J. A. and Lin, S. (1989). Actin capping activity of smooth muscle
tensin. J. Cell Biol. 109, 172a.
Brands, R., de Boer, A., Feltkamp, C. A. and Roos, E. (1990). Disintegration
of adhesion plaques in chicken embryo fibroblasts upon rous sarcoma virusinduced transformation: different dissociation rates for talin and vinculin.
Exp. Cell Res. 186, 138-148.
Cardullo, R. A., Agrawal, S., Flores, C., Zamecnik, P. C. and Wolf, D. E.
(1988). Detection of nucleic acid hybridization by nonradiative fluorescence
resonance energy transfer. Proc. Nat. Acad. Sci. USA 85, 8790-8794.
Crawford, A. W. and Beckerle, M. C. (1991). Purification and
characterization of zyxin, an 82,000 dalton component of adherens junctions.
J. Biol. Chem. 266, 5847-5853.
Dale, R. E. and Eisinger, J. (1975). Polarized excitation energy transfer. In
Biochemical Fluorescence: Concepts, vol. 1 (ed. R. F. Chen and H.
Edelhoch), pp. 115-284. Marcel Dekker, NY.
Faulstich, H., Trischmann, H. and Mayer, D. (1983). Preparation of
Tetramethylrhodaminyl-phalloidin and uptake of the toxin into short-term
cultured hepatocytes by endocytosis. Exp. Cell Res. 144, 73-82.
Feltkamp, C. A, Pijnenburg, M. A. P. and Roos, E. (1991). Organization of
talin and vinculin in adhesion plaques of wet-cleaved chicken embryo
fibroblasts. J. Cell Sci. 100, 579-587.
Fernandez, S. M. and Berlin, R. D. (1976). Cell surface distribution of lectin
receptors determined by resonance energy transfer. Nature 264, 411-415.
Förster, T. (1946). Energiewanderung und Fluoreszenz. Naturwissenschaften
33, 166-175.
Förster, T. (1948). Zwischenmolekulare Energiewanderung und Fluoreszenz.
Ann. Physik. (Leipzig) 2, 55-75.
Förster, Th. (1959). Transfer mechanism of electronic excitation. Disc. Farad.
Soc. 27, 7-29.
Förster, T. (1964). Delocalized excitation and excitation transfer. In Modern
Quantum Chemistry, vol. 3 (ed. O. Sinanoglu), pp. 93-137. Academic Press,
NY.
Geiger, B. and Ayalon, O. (1992). Cadherins. Annu. Rev. Cell Biol. 8, 307332.
Geiger, B., Ayalon, O., Ginsberg, D., Volberg, T., Rodriguez Fernandez, J.
L., Yarden Y. and Ben-Zeev, A. (1992). Cytoplasmatic Control of Cell
Adhesion, vol. LVII, pp. 631-642. Cold Spring Harbor Laboratory Press, NY.
Geiger, B., Volk, T. and Volberg, T. (1985). Molecular hetrogeneity of
adherens junctions. J. Cell Biol. 101, 1523-1531.
Geiger, B., Ginsberg, D., Salomon, D. and Volberg, T. (1990). The molecular
basis for the assembly and modulation of adherens-type junctions. Cell Diff.
Dev. 32, 243-354.
Geiger, B. and Ginsberg, D. (1991). The cytoplasmic domain of adherens type
junctions. Cell Motil. Cytoskel. 20, 1-6.
Gettins, P., Beechem, J. M., Crew, B. C. and Cummingham, L. W. (1990).
Separation and localization of the four cysteine-949 residues in human alpha2 macroglobulin using fluorescence energy transfer. Biochemistry 29, 77477753.
Herman, B. (1989). Resonance energy transfer microscopy. Meth. Cell Biol.
30, 219-243.
Hiraoka, Y., Sedat, J. W. and Agard, D. A. (1987). The use of charge-coupled
device for quantitative optical microscopy of biological structures. Science
238, 36-41.
Hiraoka, Y., Swedlow, J. R., Paddy, M. R., Sedat, J. W. and Agard, D. A.
(1991). Three-dimensional multi-wavelength fluorescence microscopy for
the structural analysis of biological phenomena. Semin. Cell Biol. 2, 153-165.
Horwitz, A., Duggan, K., Buck, C., Beckerle, M. C. and Burridge, K.
(1986). Interaction of plasma memmbrane fibronectin receptor with talin - a
transmemmbrane linkage. Nature 320, 531-533.
Hynes, R. O. (1992). Integrins: Versatility, modulation and signaling in cell
adhesion. Cell 69, 11-25.
Ingber, D. (1993). The riddle of morphogenesis: a question of solution
chemistry or molecular cell engineering. Cell 75, 1249-1253.
Johnson, D. A., Voet, J. G. and Taylor, P. J. (1984). Fluorescence energy
transfer between cobra a-toxin molecules bound to the acetylcholine
receptor. J. Biol. Chem. 259, 5717-5725.
Jovin, T. M. (1979). Fluorescence polarization and energy transfer: Theory and
application. In Flow Cytometry and Sorting (ed. M. R. Melamed, P. F.
Mullaney and M. L. Mendelsohn), pp.137-165. John Wiley, N.Y.
Jovin, T. M. and Arndt-Jovin, D. J. (1989). FRET microscopy: digital
imaging of fluorescence resonance energy transfer. Application in cell
biology. In Cell Structure and Function by Microspectrofluorimetry (ed. E.
Kohen, J. S. Ploem and J. G. Hirschberg), pp. 99-117. Academic Press,
Orlando, FL.
Jovin, T. M. and Arndt-Jovin, D. J. (1990). Digital imaging microscopy: the
marriage of spectroscopy and the solid-state CCD camera. SPIE
International Conference on Scientific Optical Imaging, vol. 1439, pp. 109120. SPIE, Bellingham, WA.
Kabsch, W., Mannherz, H. G., Suck, D. S., Pai, E. F. and Holmes, K. C.
(1990). Atomic structure of the actin DNase I complex. Nature 347, 37-44.
Kam, Z. (1987). Microscopic imaging of cells. Q. Rev. Biophys. 20, 201-259.
Kam, Z., Jones, M. O., Chen, H., Agard, D. A. and Sedat, J. W. (1993).
Design and construction of an optimal illumination system for quantitative
wide-field multi-dimensional microscopy. Bioimaging 1, 71-81.
Kemler, R. and Ozawa, M. (1989). Uvomorulin - catenin complex:
Cytoplasmic anchorage of a Ca2+-dependent cell adhesion molecule.
BioEssays 11, 88-91.
Kries, T. E., Geiger, B. and Schlessinger, J. (1982). Mobility of microinjected
rhodamine actin within living chicken gizzard cells determined by
fluorescence photobleaching recovery. Cell 29, 835-845.
Kries, T. E., Avnur, Z., Schlessinger, J. and Geiger, B. (1985). Dynamic
properties of cytoskeletal proteins in focal contacts. Cold Spring Habor
Symp. Molecular Biology of the Cytoskeleton (ed. G. Borisy, D. Cleveland
and D. Murphy), pp. 45-57. Cold Spring Harbor Laboratory Press, NY.
Kubitscheck, U., Kircheis, M., Schweitzer-Stenner, R., Dreybrodt, W.,
1062 Z. Kam, T. Volberg and B. Geiger
Jovin, T. and Pecht, I. (1991). Fluorescence resonance energy transfer on
single living cells: Application to binding of monovalent haptens to cellbound immunoglobulin. E. Biophys. J. 60, 307-318.
Kuhn, H. (1982). Energy transfer mechanism. In Biophysics (ed. W. Hoppe,
W. Lohmann, H. Markl and H. Ziegler), pp. 279-288. Springer Verlag,
Berlin.
Lakowicz, J. R. (1983). Energy transfer. In Principles of Fluorescence
Spectroscopy, pp. 303-339. Plenum Press, NY.
Ludwig, M., Hensel, N. F. and Hartzman, R. J. (1992). Calibration of a
resonance energy transfer imaging system. Biophys. J. 61, 845-857.
Milam, L. M. (1985). Electron microscopy of rotary shaddowed vinculin and
vinculin complexes. J. Mol. Biol. 184, 543-545.
Molony, L. and Burridge, K. (1985). Molecular shape and self association of
vinculin and metavinculin. J. Cell. Biochem. 29, 31-36.
Otey, C. A., Pavalko, F. M. and Burridge,K. (1990). An interaction between
α-actinin and the β1 integrin subunit in vitro. J. Cell Biol. 111, 721-729.
Pollard, T. D. and Cooper, J. A. (1986). Actin and actin binding proteins. A
critical evaluation of mechanisms and functions. Annu. Rev. Biochem. 55,
987-1035.
Press, W. H., Flannery, B. P., Teukolsky, S. A. and Vetterling, W. T. (1986).
Numerical Recipes, pp. 292. Cambridge University Press, NY.
Stryer, L. and Haugland, R. P. (1967). Energy transfer: a spectroscopic ruler.
Proc. Nat. Acad. Sci. USA 58, 719-726.
Stryer, L. (1978). Fluorescence energy transfer as a spectroscopic ruler. Annu.
Rev. Biochem. 47, 819-847.
Stryer, L., Thomas, D. D. and Meares, C. F. (1982). Diffusion-enhanced
fluorescence energy transfer. Annu. Rev. Biophys. Bioeng. 11, 203-222.
Szöllösi, J., Trón, L., Damjanoviek, S., Hellewell, S. H., Arndt-Jovin, D.
and Jovin, T. (1984). Fluorescence energy transfer measurements on cell
surface: A critical comparison of steady state fluorimetric and flow
cytometric Meth. Cytom. 5, 210-216.
Taylor, D. L., Reidler, J., Spudich, J. and Stryer, L. (1981). Detection of
actin assembly by fluorescence energy transfer. J. Cell Biol. 89, 362-367.
Trón, L., Szöllösi, J., Damjanovich, S., Helliwell, S. H., Arndt-Jovin, D. J.
and Jovin, T. M. (1984). Flow cytometric measurement of fluorescence
resonance energy transfer on cell surfaces: Quantitative evaluation of the
transfer efficiency on a cell-by-cell basis. Biophys. J. 45, 939-946.
Tsien, R. Y. (1989). Fluorescent indicators of ion concentrations. Meth. Cell
Biol. 30, 127-156.
Turner, C. E., Glenny, J. R. Jr and Burridge, K. (1990). Paxillin: A new
vinculin-binding protein present in focal adhesions. J. Cell Biol. 111, 10591068.
Uster, P. S. and Pagano, R. E. (1986). Resonance energy transfer microscopy:
Observation of membrane-bound fluorescence probes in model membranes
and in living cells. J. Cell Biol. 103, 1221-1234.
Volberg, T., Sabanay, H. and Geiger, B. (1986). Spatial and temporal
relationships between vinculin and talin in the developing chicken gizzard
smooth muscle. Differentiation 32, 34-43.
Volk, T. and Geiger, B. (1984). A 135 Kd membrane protein of intercellular
adherens junctions. EMBO J. 3, 2249-2260.
Wang, X. F., Lemasters, J. J. and Herman, B. (1993). Plasma membrane
architecture during hypoxic injury in rat hepatocytes measured by
fluorescence quenching and resonance energy transfer imaging. Bioimaging
1, 30-39.
Wulf, E., Deboren, A., Bautz, F. A., Faulstich, H. and Wieland, Th. (1979).
Fluorescent phalloidin, a tool for the visualization if cellular actin. Proc. Nat.
Acad. Sci. USA 76, 4498-4502.
Yguerabide, J. and Förester, M. (1981). Fluorescence spectroscopy of
biological membranes. Mol. Biol. Biochem. Biophys 31, 199-269.
(Received 25 July 1994 - Accepted 10 November 1994)