Analyzing focal adhesion structure by atomic force microscopy

Focal adhesions (FAs) are elongated adhesion structures at the ventral plasma membrane that enable cells to adhere to the extracellular matrix (Burridge et al., 1988). FAs are linked to the termini of contractile actomyosin filaments termed stress fibers and the tension and traction generated through stress fiber contraction allows cells to polarize and to migrate. Furthermore, focal adhesions form important signaling centers where integrins transduce signals from the cellular environment to regulate different cellular functions, such as cell growth, survival and gene expression (Adams, 2002; Cukierman et al., 2002).

Within FAs integrin αβ heterodimers are the main mediators of adhesion to the surrounding substratum. Integrins use their extracellular domain to interact with ECM components, whereas their cytoplasmic tail associates either directly or indirectly with a number of structural, cytoskeletal proteins, such as talin (Burridge and Connell, 1983), paxillin (Turner et al., 1990), vinculin (Geiger, 1979) and α-actinin (Lazarides and Burridge, 1975). These cytoskeletal proteins in turn connect integrins to actin-containing microfilaments, thereby establishing a link between the ECM and the actin cytoskeleton.

FAs have been studied intensively using immunofluorescence and electron-labeling techniques, which showed the localization of more than 50 different proteins to these structures (Zamir and Geiger, 2001). Fusing FA proteins with fluorescent proteins has made it possible to study the dynamics of FA assembly and disassembly (Wehrle-Haller and Imhof, 2002). In addition, biochemical studies have elucidated a large number of protein-protein interactions important for FA function (Zamir and Geiger, 2001). In contrast to these detailed cell biological and biochemical findings, less is known about the FA ultrastructure.

Early studies using interference reflexion microscopy (IRM) and transmission electron microscopy established that FAs occur at the points of closest cell-to-substrate contact where they are associated with the distal ends of stress fibers (Abercrombie et al., 1971; Curtis, 1964; Heath and Dunn, 1978; Izzard and Lochner, 1976). At their proximal end, the microfilaments in FAs appear to be continuous with the linear arrays of microfilaments in stress fibers (Izzard and Lochner, 1976), whereas the distal ends of the microfilament bundles often spread out (Heath and Dunn, 1978). Investigations of platinum replicas of detergent-extracted fibroblasts confirmed that at the distal ends of stress fibers, microfilaments are transformed into flattened plaques closely associated with the substrate (Svitkina et al., 1984). In addition, thin transverse filaments bridging the gap between microfilaments were described and later identified as plectin sidearms connected to vimentin cores (Svitkina et al., 1998).

Immunological electron microscopic (immuno-EM) studies have provided further insight into the vertical and lateral localization of several FA components within the adhesion plaque. Double immunogold labeling performed on ultrathin frozen sections from chick heart fibroblasts showed that vinculin is situated closer to the membrane than α-actinin in FAs (Chen and Singer, 1982) whereas in wet-cleaved chicken embryo fibroblasts, both talin and vinculin are present at high concentrations in a dense network close to the plasma membrane (Feltkamp et al., 1991). Rotary replication in combination with immunogold labeling of sheared Xenopus fibroblasts demonstrated that FA microfilaments are highly bundled and these bundles appear to be linked to the plasma membrane by laterally attached discrete protein aggregates containing vinculin, talin and β1-integrin (Samuelsson et al., 1993). In agreement, β1-integrin was found to localize in linearly arranged discrete clusters by immuno-EM experiments performed on wet-cleaved fibroblasts (Meijne et al., 1994). However, these techniques have so far either only allowed the analysis of a subsection of the entire FA structure or failed to provide a quantitative three-dimensional (3D) model of FA architecture. In addition, the EM techniques require sample preparation protocols that may distort the native array of actin filaments in FAs, such as detergent extraction, sample drying and coating, staining and cutting (Heuser and Kirschner, 1980; Small et al., 1999; Svitkina et al., 1995).

Atomic force microscopy (AFM) produces sample topographies by scanning the surface with a sharp, nanometer scale probe (tip) attached to a flexible cantilever (Binnig et al., 1986). AFM permits imaging in a fluid environment, therefore avoiding artifacts caused by drying and/or coating of samples, and has the advantage of maintaining biological systems and their functionality in physiological conditions (Horber and Miles, 2003; Lesniewska et al., 1998; Muller et al., 2002). Furthermore, cells and biological materials can be imaged directly by AFM with very little sample preparation and a resolution of less than 1 nm. The combination of AFM imaging with fluorescence microscopy is becoming a valuable tool in the investigation of complex cellular structures, in which fluorescence labeling serves to identify protein complexes of interest which can then be imaged at superior resolution by AFM (Kassies et al., 2005; Sharma et al., 2005). We used a combination of fluorescence microscopy and AFM to image FAs in de-roofed cells under physiologically relevant conditions. The resolution of the AFM topographs surpasses that of the light microscope images and provides structural information about the 3D organization of microfilaments in FAs. Furthermore, we show that by correlating the height information of AFM topographs with the fluorescence staining intensities of FA components, differences in their vertical localization can be detected.

REF52 cells (kindly provided by Alexander Bershadsky, Weizmann Institute of Science, Rehovot, Israel) were maintained in DMEM containing 10% FCS, 100 IU/ml penicillin and 100 μg/ml streptomycin. For de-roofing, 5×10⁴ cells were seeded on ethanol-cleaned glass coverslips (22 mm diameter) inserted into 35 mm tissue culture dishes and grown for 72 hours. Cells were washed three times with warm PBS, followed by incubation for 15 seconds with low molecular weight (15,000-30,000) poly-L-lysine. Subsequently, cells were washed three times for 30 seconds with 1/3 strength intracellular buffer (ICB; 20 mM HEPES pH 7.6, 1.5 mM MgCl₂, 5 mM EGTA, 5 mM NaCl, 2 mM CaCl₂, 140 mM potassium glutamate) and then transferred to a 92 mm tissue culture dish containing 10 ml ICB. Using a Hielscher microsonicator, cells were de-roofed with several short (<1 second) ultrasonic bursts at minimum amplitude. After three washes with ICB buffer, cells were fixed for 45 seconds in 2% glutaraldehyde/ICB, followed by 10 minutes in 4% paraformaldehyde/ICB. Using a custom-made metal holder, coverslips were mounted on an inverted Zeiss Axiovert 200 microscope (Carl Zeiss, Göttingen, Germany), on which an AFM (Nanowizard™ AFM, JPK Instruments, Berlin, Germany) was mounted. Coverslips were overlaid with PBS containing 9 mM propyl gallate to minimize photo bleaching and fluorescence images were collected using a Carl Zeiss 63× oil-immersion objective. Using a fluid cell, AFM contact mode images were recorded in liquid using 200 μm long V-shaped cantilevers, with nominal spring constants of 0.01 N/m (MSCT-AUHW, Veeco Instruments, Santa Barbara, CA). The force applied to the cantilever was adjusted manually to ∼50 pN (Muller et al., 1999) and the feedback gains were manually adjusted to obtain the best resolution both on height and deflection channels. Images were collected at a line-scan rate ranging from 0.3 to 0.6 Hz. Light microscopy images were processed using MetaMorph software (Universal Imaging Corporation, Downingtown, PA). AFM images were analyzed using the JPK Image Processing software and FA volumes were estimated using the SPIP software (Image Metrology, Lyngby, Denmark).

REF52 cells develop an extensive contractile apparatus with a multitude of stress fibers terminating in large FAs when grown on glass surfaces (Turner et al., 1989). REF52 cells stably expressing YFP-paxillin as a marker for focal adhesions were grown on glass coverslips for 3 days to enable them to develop mature adhesive contacts (Fig. 1A-C). To make FAs accessible for AFM imaging, it was necessary to remove the apical plasma membrane, the nucleus and the cytoplasm of these cells. Cell de-roofing was achieved by using short ultrasonic bursts based on a method developed by Heuser (Heuser, 2000). The sonication step was performed in a buffer establishing intracellular ionic conditions to preserve the integrity of intracellular structures once cells were disrupted. Cells were usually de-roofed, rinsed and fixed within 1 minute. Sonication caused varying degrees of de-roofing, with some cells remaining intact, some showing partial lysis and some cells being completely de-roofed as judged by phase-contrast microscopy. Alternative de-roofing methods, such as wet cleaving (Brands and Feltkamp, 1988) or low-ionic strength/detergent extraction (Katoh et al., 1998) produced similar results to the sonication protocol but frequently led to less complete removal of cytoplasmic organelles. In these samples subsequent AFM scanning was difficult because the loosely bound organelles frequently contaminated the AFM tip.

In de-roofed REF52 cells, paxillin-YFP continued to localize to oblong structures with dimensions typical for FAs in these cells (usually 0.5-2 μm by 3-8 μm), indicating that FA integrity was preserved during the de-roofing step (Fig. 1B). Paxillin-containing complexes also stained positive for vinculin (data not shown) and F-actin (Fig. 1D), further suggesting that FA structure was resistant to the de-roofing procedure. Stress fibers were rarely retained during the cell de-roofing and usually sheared off precisely at the interface with the FA (Fig. 1D). Addition of protease inhibitors or F-actin stabilizing agents, such as phalloidin or jasplakinolide to the sonication buffer had no detectable influence on the FA structure (data not shown).

Overview AFM images of completely de-roofed cells (Fig. 2A,B) showed retention of different cellular structures vertically protruding between 20 and 500 nm from the substratum. A brief incubation of the cells with poly-L-lysine prior to sonication caused partial preservation of the basal membrane around the cell perimeter (Fig. 2B), whereas the basal membrane was usually completely removed when this step was omitted. In order to identify FAs among the preserved basal protein complexes, the localization of YFP-paxillin was determined by fluorescence microscopy (Fig. 2C). A cluster of eight structures identified as FAs by overlaying the AFM topograph with the paxillin fluorescence image (Fig. 2D) were subsequently imaged at increased resolution (Fig. 2E,F). Filamentous structures were observed that extended throughout the entire lengths of some FAs. Filaments were not always oriented parallel to each other but fanned out towards the cell perimeter at angles ranging from 5 to 8 degrees. Actin-containing microfilaments constitute the predominant filamentous structures in FA and are typically arranged in a fan-shaped pattern (Heath and Dunn, 1978; Svitkina et al., 1984). We consequently assumed that the filamentous structures observed in the AFM topographs were composed of actin-containing microfilaments.

To investigate the organization of microfilaments in FAs in more detail, FAs in de-roofed cells were identified as before by paxillin fluorescence (Fig. 3A) and staining for F-actin using TRITC-phalloidin (Fig. 3B) and then imaged by AFM (Fig. 3D,E). An overlay of the merged fluorescence image (Fig. 3C) with the AFM deflection image (Fig. 3F) demonstrates the good correlation between the two image-generating techniques. A central region within an FA was subsequently imaged with increasing resolution (Fig. 3G,H). Scanning at the highest magnification revealed the predominantly parallel array of filaments in the central part of the FA (Fig. 3H). Occasionally filaments appeared to branch and individual filaments crossed the parallel arrays at different angles (Fig. 3G,H). These crossing filaments could be either actin-containing microfilaments or thinner intermediate filaments (Svitkina et al., 1984; Svitkina et al., 1998). The apparent diameter of the parallel-arranged filaments varied from 90 nm down to 20 nm, which was the limit of the lateral resolution achieved. The variation of the apparent filament width suggests that the observed filaments consist of microfilament bundles in which individual microfilaments could not be resolved because of the limited resolution achieved (see Discussion). However, occasionally several adjacent filaments displaying the minimal apparent width of 20 nm could be resolved (Fig. 3J,K), probably because of slightly looser microfilament bundling in these areas. These smallest filaments are likely to represent single microfilaments whose apparent width in the AFM scans is increased as a result of non-linear tip convolution effects (Schwarz et al., 1994).

Cross sections taken perpendicularly to the microfilament array showed height variations ranging from 10 to 20 nm between adjacent, parallel microfilaments (Fig. 3J) or microfilaments crossing each other (Fig. 3K), suggesting that microfilaments are organized in layers in the FA plaque. The minimal height difference measured between crossing microfilaments of 10 nm is close to the microfilament diameter of 7 nm (Moore et al., 1970), supporting the notion that individual microfilaments could be resolved by our AFM approach.

Along the microfilaments laterally attached globular structures with diameters of around 50 nm could be detected (Fig. 3H), which could represent integrin-containing aggregates observed by immuno-EM (Samuelsson, 1993), although no regularity in the spacing of these globular structures was found. The abundance of these globular structures varied (compare Fig. 3J,K) but was usually highest in the central part of FAs.

The de-roofing method we used maintained the FA structures in physiological conditions during the entire preparation protocol but included an ultimate glutaraldehyde fixation step. Fixation was necessary because higher structures were frequently too pliable for reproducible AFM scanning in unfixed FAs. In order to minimize potential structure artifacts due to glutaraldehyde fixation, such as microfilament fusion (Svitkina et al., 1995), we used a short glutaraldehyde fixation time, followed by longer paraformaldehyde fixation. Nevertheless, at higher magnification, the AFM image quality sometimes became compromised even after fixation because of continued sample pliability and adhesion to the AFM tip. This effect was strongest at the distal end of FAs where microfilaments may not be anchored to the substrate as firmly as in the central part because of active FA assembly or disassembly processes. To reduce the lateral movement of filamentous and globular structures through tip movement, the scan speed had to be kept low (0.3 Hz) and the scanning force had to be continuously minimized. However, comparison of height trace and retrace topographs (Fig. 3H) generated from the central region of an FA demonstrates that optimizing the scan parameters enabled us to image the filament array without any significant structural distortions. Varying the scan angle at such optimized imaging conditions generally had no noticeable effect on the generated topographies.

To illustrate overall FA structure, the AFM topography of a representative FA was presented in a relief (Fig. 4B). The finger-like branching of microfilaments at the end of the FA opposite to the stress fibers could be resolved clearly. A side view of a 3D reconstruction with true (1:1) aspect ratio between the Y and Z axis shows a gradual, constant increase in height from about 50 nm at the front end to about 180 nm at the stress-fiber-associated end of the FA, resulting in an increase of approximately 3°. At the very end of the stress-fiber-associated side, FAs regularly showed a steep increase in height (Fig. 4C). This may not be an intrinsic structural FA feature but rather represent structural damage inflicted on the FA structure as a result of the forceful removal of the stress fiber interface during the preparation protocol. An overlay of five cross sections taken across the FA at intervals of 500 nm shows that FAs are low and wide at the front, but high and narrow at the stress-fiber-associated end (Fig. 4D).

Moving from the distal to the proximal end, FA height increased 2.2-fold, whereas FA width decreased by about the same factor (Fig. 4F). In contrast, the cross section areas varied only by a factor of 1.4 over the length of the FA, reaching a maximum in the central part. The greater constancy of the cross-section area compared to cross-section height and width suggests that the change in FA shape may be the result of a gradual reorganization of structural elements extending throughout the entire FA structure, i.e. microfilaments, rather than being caused by the presence of different structural elements in different parts of the FA. Fig. 4G shows a model of the microfilament organization within FAs. Microfilaments are bundled into a circular array at the stress-fiber-associated end and then gradually flatten out into a planar array towards the opposite end. From the AFM topography, the area of the FA in contact with the underlying substratum (1.9×10⁶ nm²) and the FA volume (1.2×10⁸ nm³) could be estimated. Overall, the FA volumes analyzed from 15 FAs ranged from 5×10⁷ to 5×10⁸ nm³.

In order to quantify the height and spacing of microfilament bundles at the dorsal FA surface, bisecting cross sections perpendicular to the FA long axis were taken as exemplified for a cluster of FAs in Fig. 5A. The corresponding height profiles are shown in Fig. 5B. Because of the high vertical resolution of the AFM, the average height of microfilaments over the surrounding substrate (corresponding to the average value of the local maxima in the cross-section graphs) could be precisely determined. Their average peak height was 118±26 nm (mean±s.d.; 30 FAs analyzed), whereas the average overall height of the cross section was 84±16 nm (mean±s.d.). However, because of spatial constrictions, the AFM tip probably could not fully penetrate between narrowly spaced microfilaments or microfilament bundles, leading to a slight overestimation of the sample height in the interjacent areas. The overall average height is therefore expected to be lower than measured. The distance between neighboring microfilament bundles was determined by measuring the distance between neighboring local maxima in the cross-section graphs and varied from 40 to 240 nm, with an average of 127±50 nm (Fig. 5C).

Occasionally FAs in de-roofed cells contained areas in which the F-actin staining intensity was markedly decreased. Generally, such areas with lower F-actin staining corresponded to regions exhibiting decreased heights in the AFM topographs. These circular indentations could be the result of mechanical disruption of FA structure during de-roofing. An example of an FA complex displaying a local reduction in F-actin staining and height is given in Fig. 6A,D. In contrast to the F-actin staining intensity, the paxillin signal in this area remained unchanged (Fig. 6B). A line scan through the AFM topograph and the corresponding fluorescence images (Fig. 6E) demonstrates a positive correlation between the height and F-actin signal but not the paxillin signal (Fig. 6F). As the removal of the upper (dorsal) part of the FA architecture (∼50 nm) does not decrease the paxillin signal, paxillin localization must be restricted to the lower, membrane-proximal half of the FA. In the opposite situation, because F-actin staining intensity was almost reduced to background levels when the upper FA structure was removed, F-actin appears to be mainly localized in the membrane-distal half of the FA.

We have combined fluorescence microscopy and AFM to generate high-resolution images of FA complexes in de-roofed fibroblasts. Fluorescence microscopy was first used to identify paxillin-containing FAs, which were subsequently imaged by AFM at a resolution surpassing the limits of conventional light microscopy. The AFM topographs yielded structural detail that was not obtainable from light microscopic images and provided insight into the 3D organization of microfilaments within FAs.

AFM topographs of FAs showed the presence of a multitude of filaments spanning the entire length of the FA. These filaments were arranged predominantly in parallel within the central part of the FA but frequently fanned out towards the distal end of the FA, an arrangement consistent with the organization of actin microfilaments in FAs as described in EM studies (Heath and Dunn, 1978; Svitkina et al., 1984). However, whereas transmission EM images of FAs show a tight lateral packing of microfilaments throughout the length of the FA (Heath and Dunn, 1978; Singer, 1979), our AFM topographs suggest that the dorsal FA surface at least, is not formed by a homogeneously dense microfilament layer. Instead, FAs had a corrugated dorsal surface formed by filamentous structures spaced by an average of 127 nm and protruding by 10 to 40 nm over the interjacent areas. The apparently dense packing of microfilaments observed in transmission-EM images may be the result of the projection of the 3D microfilament lattice onto a plane. Such a projection eliminates the information about the vertical separation between microfilaments and consequently reduces the apparent distances between microfilaments. In addition, the superimposition of different horizontal layers of microfilaments may further reduce the apparent spacing between microfilaments in transmission EM images.

The apparent width of filaments observed by AFM varied from 20 to 90 nm. It is well known that the finite size of the AFM tip can cause a non-linear convolution effect in which high, steeply increasing structures appear severely broadened (Schwarz et al., 1994). Considering such tip-induced broadening of the imaged object, the true diameter of filaments should be expected to be much smaller. According to models describing the contribution of AFM tip radius and geometry to the convolution of cylindrical structures in AFM images (Schwarz et al., 1994), the thinnest filaments we observed (apparent width 20 nm) probably correspond to single actin filaments (∼7 nm). Depending on the true radius of the tip used in a particular scan (ranging from 10 to 40 nm, according to the manufacturer's instructions), the tip broadening effect could even account for the apparent widening of single actin filaments into the thicker filaments observed in our AFM images. However, the presence of different filament diameters within the same image (i.e. generated with the same AFM tip, see Fig. 3H-K) suggests that the larger filament diameters were caused by the grouping of different numbers of microfilaments. In these larger filamentous structures individual microfilaments may have been bundled tightly so that they could not be resolved by the AFM tip. Consistent with this idea, EM pictures of rotary replicas show tight bundling of microfilaments within the FA plaque (Samuelsson et al., 1993). Thus, in FAs, microfilaments may be grouped into discrete functional units.

In the membrane-proximal region of the FA, α-actinin functions as a linker between actin filaments and the cytoplasmic tails of integrin subunits (Lazarides and Burridge, 1975; Otey et al., 1990; Rajfur et al., 2002). Nevertheless, α-actinin predominantly localizes to the membrane-distal region of FAs (Chen and Singer, 1982; Geiger et al., 1981) where the bulk of actin filaments is located (Fig. 6A). This points to an additional role of α-actinin as an actin filament crosslinker in FAs. In addition to α-actinin, microfilaments could also be bundled by other actin crosslinkers present in FAs, such as fimbin (Bretscher and Weber, 1980) or filamin (Geiger et al., 1984) and the different dimensions of these linkers could then give rise to filament bundles of varying diameters.

Microfilaments were usually arranged in parallel or at slight angles to each other. Occasionally, however, they were branched and could cross each other (Fig. 3G). We measured height differences ranging from 10 to 20 nm between adjacent microfilaments. This suggests a layered microfilament architecture in which the vertical separation between the layers is in the order of the diameter of a single actin filament. The vertical separation between neighboring microfilaments in FAs may therefore be considerably smaller than the lateral.

Exhibiting a vertical resolution of 1 nm or better, AFM topographs also provided accurate information about FA height and enabled us to determine the FA volume. The average height of bisecting cross sections perpendicular to the FA structure was 84±16 nm, whereas the microfilament bundles protruded 118±27 nm over the substratum (Fig. 5A,B). Given a plasma membrane thickness of 5 to 10 nm and a distance of 10 to 15 nm between the plasma membrane and the substratum at sites of FAs (Heath and Dunn, 1978), the heights we measured have to be reduced by 15 to 25 nm in order to give an accurate account of the thickness of the cytoplasmic part of the adhesion plaque. The FA height we determined is in good agreement with thin section EM studies showing a ∼60-nm-thick density at the site of FAs (Chen and Singer, 1982). FA heights of less than 120 nm also re-emphasize that FAs are ideally suited for imaging by total internal reflection internal fluorescence microscopy (TIRFM) because their height is well within the zone excited by the evanescent wave, which typically penetrates some 150 nm into the cytoplasm of the cell (Adams, 2002; Krylyshkina et al., 2003).

The AFM topographs showed that FA height increases towards the stress-fiber-associated end, whereas the width often decreased. This wedge shape could reflect the functional requirement of FAs to constitute a link between a cylindrical stress fiber and a planar substrate. By flattening towards the stress fiber distal end and the finger-like fanning out of microfilaments, the contact area with the substratum is maximized, whereas the circular cross section at the opposite end assures maximum connectivity with the stress fiber. The relative change in cross-section shape contrasted with much smaller relative changes in cross-section area, indicating that the shape change over the FA length could be caused by a reorganization of microfilaments extending over the entire FA (Fig. 4G).

The ultrastructure of the stress fiber/AF interface is unknown and it remains to be investigated whether there is continuity of actin filaments from stress fibers into FAs. This seems unlikely however, as the stress fiber region immediately next to the FA is a highly dynamic region characterized by intense actin polymerization. Furthermore, stress fiber contraction coincides with twisting of the entire fiber (Katoh et al., 1998). Actin filaments in FAs and stress fibers are functionally distinct. For instance, myosin II, responsible for the force generation in stress fibers is absent from FAs. During our preparation protocol, stress fibers were usually removed from the FAs, although occasionally stress fibers remained attached to FAs (see Fig. 3B). In the future it will be interesting to investigate the structure of the FA/stress fiber interface in more detail.

Disruption of FA structure down to a height of 50 nm above the surrounding substratum left the paxillin-YFP fluorescence signal unaffected (Fig. 6B), indicating that paxillin localizes to the membrane-proximal region of FAs. Allowing a total of 25 nm for the thickness of the plasma membrane plus its distance from the substratum, paxillin localization appears to be further restricted to within 25 nm of the plasma membrane. Membrane-proximal localization of paxillin is consistent with the idea that upon recruitment to FAs paxillin binds to vinculin (Turner et al., 1990), an interaction partner shown to localize to the membrane-proximal part of the FA (Chen and Singer, 1982). In FAs, paxillin may also bind to the cytoplasmic tails of integrin subunits (Chen et al., 2000; Schaller et al., 1995), which would equally confine paxillin to a region immediately adjacent to the plasma membrane. In contrast, F-actin staining was almost completely lost when the membrane-distal half of the FA was disrupted, pointing to a localization of actin filaments primarily in the upper (membrane-distal half) of the FA. Correlating fluorescence intensities with the height information from AFM imaging thus provides information about the vertical localization of FA components within the adhesive plaque.

It is of great interest to understand how changes in the packing density of FA components in the adhesion plaque affect FA function. For `2D' markers (proteins confined to the plasma membrane or a thin adjacent horizontal plane), fluorescence intensities give a measure of how tightly they are packed in the FA. In contrast, fluorescence intensities yield no information about the clustering of `volume' markers (proteins distributed throughout the entire adhesion plaque) (Wehrle-Haller and Imhof, 2002). Because paxillin appears to localize close to the plasma membrane, it can be essentially regarded a 2D marker and consequently, changes in the paxillin fluorescence intensity should correlate with changes in its lateral packing.

In conclusion, we have obtained AFM topographs of FAs using a simple preparation technique that did not require harsh sample treatment and allowed imaging under physiologically relevant conditions. The high vertical resolution of the AFM provided accurate FA height information and offered insight into the organization of microfilaments, such as their layered array and apparent bundling at the dorsal FA surface. By correlating the paxillin-YFP fluorescence intensity with the height information contained in the AFM topographs, we could assign paxillin localization to the membrane-proximal part of the FA. It will now be interesting to investigate the FA ultrastructure with alternative imaging techniques able to provide high-resolution, 3D structural information of native biological samples, such as cryo-electron tomography (Baumeister, 2005; Resch et al., 2002).

This work was supported by the Deutsche Volkswagenstiftung, by the EC and the Free-State of Saxony. We would like to thank Kate Poole, Pierre-Henri Puech and Christian Le Grimellec for helpful discussions and JPK Instruments for their fruitful and collaborative support.