Correlative fluorescence and transmission electron microscopy: an elegant tool to study the actin cytoskeleton of whole-mount (breast) cancer cells

Authors


Kristina A. Jahn. Tel: +61-2-9351-7547; fax: +61-2-9351-7682; e-mail: k.jahn@usyd.edu.au

Summary

Elucidating the structure and dynamics of lamellipodia and filopodia in response to different stimuli is a topic of continuing interest in cancer cells as these structures may be attractive targets for therapeutic purposes. Interestingly, a close functional relationship between these actin-rich protrusions and specialized membrane domains has been recently demonstrated. The aim of this study was therefore to investigate the fine organization of these actin-rich structures and examine how they structurally may relate to detergent-resistant membrane (DRM) domains in the MTLn3 EGF/serum starvation model. For this reason, we designed a straightforward and alternative method to study cytoskeleton arrays and their associated structures by means of correlative fluorescence (/laser)- and electron microscopy (CFEM).

 CFEM on whole mounted breast cancer cells revealed that a lamellipodium is composed of an intricate filamentous actin web organized in various patterns after different treatments. Both actin dots and DRM's were resolved, and were closely interconnected with the surrounding cytoskeleton. Long actin filaments were repeatedly observed extending beyond the leading edge and their density and length varied after different treatments. Furthermore, CFEM also allowed us to demonstrate the close structural association of DRMs with the cytoskeleton in general and the filamentous/dot-like structural complexes in particular, suggesting that they are all functionally linked and consequently may regulate the cell's fingertip dynamics. Finally, electron tomographic modelling on the same CFEM samples confirmed that these extensions are clearly embedded within the cytoskeletal matrix of the lamellipodium.

Introduction

Cancer reaches a critical stage when primary tumour cells invade the surrounding tissue, including nearby blood and lymph vessels, thereby spreading throughout the body. At their destination metastatic cells attach to the blood vessel wall, forming a secondary tumour. This metastatic process seems to be organ-specific (Kieran & Longenecker, 1983). Amongst others, breast cancer causes secondary lymph node, bone, lung and liver metastasis (Jiang et al., 2002). This study uses the MTLn3 breast adenocarcinoma cell line, which maintains its ability to form primary tumours when injected into breast tissue as well as secondary tumours at lymph nodes (Neri et al., 1982).

The actin cytoskeleton generates and maintains cell morphology and polarity, and is required for intracellular transport, contractility, cell division and motility. Mounting evidence suggests that alterations in actin polymerization or reorganization plays a key role in regulating the movement and metastatic behaviour of malignant cells (Rao & Li, 2004). Lamellipodia and filopodia dynamics are crucial to successful cancer cell movement and are regulated by a complex machinery of actin remodelling, activation of oncogenic actin signalling pathways or inactivation of actin-binding proteins that have tumor suppressor functions (Bailly & Condeelis, 2002). Lamellipodia are flat subcellular membrane sheet extensions that are rich in actin especially at a cell's leading edge where a dense band of actin filaments is found. Lamellipodia of MTLn3 cells are visible as early as 50 s after epidermal growth factor (EGF) stimulation and reach maximum activity within 3–5 min. During the first 50 s, a 2.4-fold increase in nucleation activity results in an increase in the number of growing filaments within 100–200 nm of the cell's leading edge. The number of actin filaments within the first micrometer of the leading edge reaches its maximum density after 3 min and is associated with a large increase in filament branching and a concomitant decrease in filament length (Bailly et al., 1999; Mouneimne et al., 2004).

The presence of EGF receptors on MTLn3 cells provides an interesting model for studying the various chemotaxis pathways involved in the metastasis of breast cancer (Bailly et al., 1998, 2000; Yip et al., 2004). Previous studies have shown that these cells actively respond to EGF by generating lamellipodia in the direction of an applied chemoattractant (Bailly et al., 1998). Conversely, multiple and undirected lamellipodia form when EGF is added to cell culture medium without generating a gradient (Segall et al., 1996). MTLn3 cells also possess fine finger-like filopodia extending from the cell rim. Their role in detecting chemical gradients and exploring the surrounding microenvironment is described in detail elsewhere (Kater & Rehder, 1995; Nobes & Hall, 1995). Elucidating the structure and dynamics of lamellipodia and filopodia in response to different stimuli is of considerable interest because they are attractive therapeutic targets. A close functional relationship has been proposed between these actin-rich protrusions and specialized membrane domains (e.g. membrane rafts) (Neumann-Giesen et al., 2004). Membrane rafts lay as discrete patches in the plasma membrane of cells and are commonly studied as detergent-resistant membrane (DRM) structures rich in sphingolipids and cholesterol (Edidin, 2003). These membrane domains have been implicated in a number of physiological and pathological processes such as cell signalling, cellular motility, intracellular trafficking and in function of the immune, digestive and neurological systems (Michel & Bakovic, 2007). Therefore, the aim of this study is to obtain high-resolution structural data of the fine organization of these actin-rich structures and their relationship with membrane rafts in the relevant EGF/serum starvation MTLn3 cancer cell model by applying the methodology of correlative fluorescence (/laser)- and electron microscopy (CFEM).

It is well known, when different imaging techniques are applied on a certain subcellular structure, various outcomes, and sometimes contrary results, can be expected. This all depending on the microscopy method applied. Hence, this illustrates the necessity of correlative imaging approaches to dissect the fine structure of cells across length scales (Giepmans, 2008). Correlative microscopy can be defined as an imaging platform that aims to capture exactly the same cellular structures using two or more different microscopy techniques, preferably with different resolution limits. As such, the correlative microscopy approach allows the investigator to collect additional novel structural information about their sample and this provides a degree of confidence about the structures of interest, as information obtained with one method can be directly compared to that seen with the other methods (Braet & Ratinac, 2007a). One way to image the actin cytoskeleton of mammalian cells using correlative microscopy techniques is through microinjecting fluorescently labelled anti-actin antibodies into living cells. The entire actin cytoskeleton is then visualized using fluorescence and electron microscopy, with data from both correlated using digital software (Svitkina & Borisy, 1998). The strength of this elegant approach is that it allows to combine live cell optical data with static information at the nanometer level. It has to be noted that this technique uses glass cover slips as cell culture substrate thereby excluding whole-mount transmission electron microscopy (TEM) examination. Here, we put forward an alternative approach combining the complementary advances of confocal laser scanning microscopy (CLSM) and TEM to relocate, with the aid of fiducial markers, and examine exactly the same whole-mount cells using CFEM (Peachey et al., 1996; Jahn et al., 2007). We demonstrate the applicability of this combined imaging method by presenting novel complementary information at different resolution limits of the fine structure of the actin cytoskeleton and its associated membrane domains within the same MTLn3 cell before and after exposure to EGF or serum starvation. These samples were used for TEM tomographic imaging without the need for any further preparative steps.

Materials and methods

Cell culture

MTLn3 cells (Welch & Nicolson, 1983) were cultured in high d-glucose Dulbecco's modified eagle medium (DMEM, Gibco-Life Technologies, 31053-028, Australia) containing 5% heat-inactivated foetal calf serum (FCS, Invitrogen, 10438026, Australia), 100  U  mL–1 penicillin and 100  μg  mL–1 streptomycin (Invitrogen, 15140-163, Australia) (Segall et al., 1996). For whole mount studies, 2 × 104 cells per well in a 24-multiwell plate were seeded on Formvar®-coated TEM nickel finder grids (GNi200F2 – 200 Mesh, Proscitech, Australia) pretreated with 20 μg mL–1 collagen type I (Sigma Aldrich, C7661, Australia) as described previously (Braet et al., 1995). The cells were then incubated for 22 h at 37 °C in a humidified incubator with 5% CO2 in air.

FCS starvation and EGF stimulation experiments

After 22 h of culture, MTLn3 cells were starved of FCS by first thoroughly washing them in DMEM containing 1% BSA (Sigma Aldrich, A7030, Australia) and then incubating them in this solution for a further 2 h. For EGF stimulation experiments, MTLn3 cells were first incubated in DMEM as described above and then treated with 5 nM EGF (Invitrogen, 53003018, Australia) diluted in DMEM for 1 or 3 min at room temperature as described previously (Segall et al., 1996). Next, the samples were prepared for correlative imaging (vide infra).

Correlative fluorescence (/laser)- and electron microscopy

MTLn3 cells were prepared for correlative whole mount microscopic examination as described previously (Jahn et al., 2007). Briefly, this was achieved in two main steps: (i) for CLSM investigation, including fluorescent labelling and (ii) for high-resolution TEM (Fig. 1). Fiducial markers of the finder grids were used to find and image areas of interest with both microscopes.

Figure 1.

Schematic drawing illustrating the main steps of the combined fluorescence- and whole-mount transmission electron microscopy (CFEM) method for collecting fluorescence- and electron microscopic information of the same cells. Briefly, the CFEM approach involves three-steps: (A) fluorescent labelling and subsequent confocal laser scanning visualization of cells cultured on electron microscopy finder grids; (B) relocating and imaging the same cells with transmission electron microscopy using fiducial markers (♣); (C) transferring and overlaying the digital images from both microscopies with the aid of image analysis software and calibrated magnification standards.

For CLSM investigation culture medium was removed and the cells were immediately extracted for 5 min at room temperature with cytoskeleton extraction buffer composed of PEM-buffer (100 mM piperazine-N,N′-bis [2-ethanesulfonic acid] (PIPES), 1 mM ethylene glycol bis [2-aminoethylether]-N,N,N′,N′ tetra-acetic acid (EGTA), 1 mM MgCl2) containing 4% polyethylene glycol (PEG) 20 000, 1% Triton X-100 (BDH Chemicals Ltd., Poole, England) and 5 μm phalloidin (Sigma Aldrich, P-2141, Australia) at pH 7.1. The cells were extensively rinsed with PEM-buffer then stained for filamentous actin with 165 nM Alexa Fluor® 594 labelled phalloidin (Invitrogen, A12381, Australia) and with 0.5 μg DAPI per mL for nuclear identification (Sigma Aldrich, D9564, Australia). Both stains were mixed in PEM-buffer and 10 μL of this solution was applied to each grid for 20 min at room temperature under humidified conditions to avoid air-drying. Then the samples were washed twice in PEM-buffer and the grids were recovered using magnetic tweezers to preserve the integrity of the EM supports (Jahn et al., 2007). Grids were mounted on glass slides using an anti-fade mixture composed of 1:1 glycerol-based AF2 Citifluor® (Leica, R1321, Australia) and PEM-buffer containing 3 mg mL–1 ascorbic acid (refractive index 1.398) (Terasaki & Dailey, 1995). The grids were covered with a glass cover slip, sealed with nail polish, and examined immediately (excitation wavelengths: 405, 488 and 594 nm) using a Nikon C1 CLSM equipped with a 100× immersion oil objective (Nikon, Numerical Aperture 1.4, Japan). Each image was averaged four times and the pinhole setting was set on small. Z-stacks were recorded using the EZ-C1 3.0 software at a step size of 0.3 μm, resulting in image data with a calculated optical section thickness of 0.35 μm (Cox, 2007). Regions and cells of interest were imaged and their precise location documented using the fiducial grid marks in transmission light optical mode (Fig. 1A).

For subsequent TEM investigation the grids were recovered and rinsed twice in PEM-buffer. They were then fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer for 20 min at room temperature. After rinsing with PEM-buffer, the samples were incubated in aqueous 0.1% tannic acid solution for 20 min, rinsed twice in distilled water, and incubated for 5 min in distilled water. Next, the samples were treated in 0.16% uranyl acetate for 20 min. The samples were subsequently rinsed twice with distilled water then dehydrated through a graded ethanol series. Before applying 100% ethanol, grids were placed in 100% ethanol containing 0.16% uranyl acetate for 20 min. The grids were rinsed several times in 100% ethanol, stored for 5 min in 100% ethanol and then hexamethyldisilazane (HMDS)-dried (Sigma Aldrich, 999-97-3, Australia) as described previously (Braet et al., 1997). The grids were transferred to a desiccator for at least 30 min before rotary shadowing with platinum (10 s at an angle of 45°) and carbon coating for 20 s at an angle of 45°. Samples were examined with a Philips CM12 (FEI, Eindhoven, The Netherlands) at an accelerating voltage of 120 kV. Cells previously imaged with the CLSM were relocated by using the alphanumerical marks of the finder grids (Fig. 1B). Digital TEM images were captured at the same end magnification using a 11-megapixel CCD camera (Morada, Soft Imaging Systems GmbH) and controlled by the iTEM software package. For high magnification imaging, images of the partly stained whole-mount cells were captured under the scanning transmission electron microscopy (STEM) capturing mode. STEM is more efficient at high magnifications than classical TEM, allowing high-contrast imaging of unstained samples (Jahn et al., 2007).

For image processing and analysis the digital data obtained from both microscopes were transferred to Adobe Photoshop 7 software for colour adjustment and figure assembly (Fig. 1C).

Transmission electron tomography

Areas of interest were examined with a Tecnai TF30 TEM (FEI, Eindhoven, The Netherlands) at 300 kV connected to a CCD camera (Gatan, Ultrascan 895, 4096 × 4096 pixels) and using SerialEM-software (Mastronarde, 2005). Images were binned two times (2048 × 2048 pixels, 0.8 nm per pixel). Semi-automated double-axis tilt series were recorded in 1° incremental steps ranging from −60° to +60°. All subsequent data processing steps were conducted under IMOD 3.9.3 software package (http://bio3d.colorado.edu/imod). In order to align each of the tilt series electron dense particles with a size of approximately 5 nm were used as fiducial markers (Braet et al., 2009). After producing tomograms from each tilt series, the information was merged to produce a higher resolution tomogram and structures were modelled on this tomogram by manual segmenting.

Results

Combined confocal- and electron microscopy imaging of MTLn3 cells cultured in the presence of FCS revealed an extensive filamentous actin cytoskeleton traversing the long axis of each cell (Figs 2A and B). Dot-like actin-rich structures were frequently observed in the cytoplasmic areas. Furthermore, small but distinct lamellipodia containing filamentous actin were observed at the edges of control cells (Figs 2A and B). In contrast, MTLn3 cells starved of FCS for 2 h appeared to retract and the leading edges displayed faint actin staining. Serum starvation also resulted in more filopodia and more dense actin stress fibres (Figs 2C and D).

Figure 2.

Correlative confocal laser scanning- (A, C, E and G) and transmission electron microscopy images (B, D, F and H) of the same control- (A and B), serum starved- (C and D), and EGF stimulated MTLn3 cells, where EGF treatments were for either 1 min (E and F) or 3 min (G and H). Note that the confocal images are three-dimensional projections derived from of the entire cell. (A) MTLn3 cells possessed a filamentous actin cytoskeleton (large arrow), cytoplasmic actin dots (small arrow) and the lamellipodia areas (arrowhead) that sometimes contained filopodia-like protrusions. (C) Serum starved MTLn3 cells displayed the same structural features. However, the cells had more filopodia and the lamellipodial areas (small arrow) are weakly stained for actin when compared to control cells. Note that the actin bundles are more pronounced (large arrow). In contrast, EGF stimulated cells treated for 1 (E) or 3 min (G) show large extended lamellipodial areas containing protrusions. Dense stress fibers were observed (large arrow) as well as an increase in the number of actin dots (small arrow). (B, D, F and H) The corresponding TEM images show the same fine structural details (see, images A, C, E and G for comparative reasons) and reveal the position of the intermediate electron dense DRMs (arrow). Nucleus, N. Scale bars, 20 μm.

Stimulation with EGF for 1 min induced MTLn3 cells to form lamellipodia with very extended leading edges and filopodia (Figs 2E and F). Cytoplasmic stress fibres were prominent around the nucleus and there was an increase in the number of cytoplasmic actin dots. After 3 min of EGF stimulation essentially the same structural changes were observed as described for 1 min EGF-stimulated cells. However, the lamellipodia had reached their maximal size and more actin dots were observed (Figs 2G and H).

Since the actin cytoskeleton of MTLn3 cells responds vigorously to EGF within minutes (vide supra), we investigated the complex organization of these structures in more detail using TEM (Fig. 3) and CLSM (Fig. 4). Under control conditions MTLn3 cells possess lamellipodia as outlined above. However, intermediate magnification TEM imaging revealed that these sheet-like structures are composed of an actin dense region at the cell edge followed by a less intricate and dense cytoskeleton rich area extending towards the nuclear region. Furthermore, DRM remnants of different sizes, that resembling membrane rafts, were found on top and underneath the lamellipodia (Fig. 3A). High-magnification TEM imaging showed single actin filaments extending out beyond the leading edge, sometimes for distances greater than 1.5 μm (Fig. 3B). Some of these finger-like projections were aligned at an angle perpendicular to the leading edge, whereas others ran almost parallel to the leading cell edge.

Figure 3.

High-resolution transmission electron micrographs of the lamellipodium, leading edge and filamentous protrusions of control (A and B) and EGF-stimulated (3 min) MTLn3 cells (C and D). Data were captured under the scanning transmission electron microscopy imaging mode to allow high contrast visualization of the partly stained whole mounts. The dotted lines in figures A and C indicate the border between the lamellipodia area and the leading edge. (A and B) DRMs were consistently found on top (large arrow) and underneath (small arrow) the lamellipodium (asterisk). The boxed area in A (solid line) is enlarged in B, and shows that a number of actin filaments (arrow) extend beyond the cell edge. (C and D) Large and small arrows indicate DRM domains lying on top and underneath the lamellipodial area (asterisk), respectively. The boxed area in C (solid line) is enlarged in D and reveals that long actin filaments extend (arrow) beyond the cell rim. Scale bars A and C, 2 μm; Scale bars B and D, 200 nm.

Figure 4.

High-magnification confocal laser- (A), transmission electron microscopy (B and D) and merged correlative image data (C) of the same cytoplasmic area within a MTLn3 cell exposed to EGF for 3 min. (A) Note the leading edge (large arrow), filopodium (arrowhead) and dot-like structures (small arrow). (B) TEM investigation showed similar structural information, as well as intermediate electron dense structures such as DRMs (arrow). (C) Shows the merged correlative image information of A and B for comparative analysis. The boxed area in C (solid line) is enlarged in D. (D) This high-power TEM micrograph illustrates the complex and intricate organization of these cytoskeletal dot-like structures from which actin filaments conglomerate. Scale bars A–C, 5 μm; Scale bar D, 500 nm.

After 3 min of EGF stimulation, MTLn3 cells contain distinct protruded lamellipodia and the cytoskeleton dense area of the lamellipodium at the cell edge was followed by a less dense filamentous cytoskeleton area (Fig. 3C). However, this cytoskeleton matrix, which was denser and more intricate when compared to the control condition (Fig. 3A). Furthermore, all cells contained discrete membrane islets above and beneath the cytoskeleton matrix of the lamellipodia (Fig. 3C). Some of these detergent-resistant areas were less than 200 nm in diameter and were found to be round, but others had a surface area of up to 8 μm2 and were irregularly shaped. Cytoskeletal filaments could be discerned at the edge of these structures, suggesting that they anchor these membrane islets in place. Finally, an increase in the number of filopodia was observed and these extended beyond the cell edge for 0.5–1 μm (Fig. 3D). These single filamentous structures were again aligned perpendicularly or parallel to the leading edge of the cells.

Throughout this study fluorescently labelled, actin-rich dot structures were repeatedly observed in the lammellipodia regions (Figs 2 and 4A) and were readily correlated to structures observed by TEM (Fig. 4B). Figure 4(C) shows the merged image information of both confocal laser scanning and TEM data of a lamellipodial area of the same cell 3 min after stimulation with EGF. The observed dot-like structures were often located near to or connected to filopodia but others were also found without any filopodia association. Figure 4(D) depicts a filopodium clearly associated with those dots at a nominal magnification of 66 000×. Actin filaments arranged in a network seem to originate from or terminate at the electron dense dot structure. It is likely that cytoskeleton proteins other than actin might be incorporated into this relatively thick region. Membrane remnants were also found in close proximity to these peculiar structural complexes.

In order to dissect the fine structural organization and association of the filamentous extensions with the lamellipodium in further detail (Fig. 5A) transmission electron tomographic analysis was applied to EGF stimulated MTLn3 cells (Fig. 5B). Modelling revealed that the long filamentous actin extension appears to be connected to two shorter filaments, which in turn were anchored further back in the intricate cytoskeleton matrix of the lamellipodium (Fig. 5B). This V-shaped organization was observed for all finger-like complexes studied along the cell edge. Furthermore, three-dimensional modelling and projection of the reconstructions under different angles disclosed the close architectural relationship of these finger-like complexes with the surrounding filamentous cytoskeleton of the lamellipodial area (Fig. 5b).

Figure 5.

High-magnification transmission electron micrograph (A) and corresponding tomographic reconstruction (B) of the leading edge of an MTLn3 cell 3 min after stimulation with EGF. (A) The dotted line box of the low-magnification TEM image in the inset (a) indicates the area from which image A was taken. Note the dense cytoskeleton matrix (*) and filaments protruding from this area (arrows). (B) Filaments in the region of the cell edge indicated by the dotted line in A were modelled based on 120 images per tilt series using dual-axis tilting. From the three-dimensional model constructed of the filamentous extensions (large arrow) and the surrounding cytoskeletal filaments (small arrow), the V-shaped organization (small arrowhead) from which the long filament seems to branch and/or originate became apparent. This cytoskeleton complex is clearly interconnected with the surrounding filamentous cytoskeleton. Discrete knot-like structures (large arrowhead) were observed repeatedly along these cytoskeleton filaments, and the close association of these and their relative height became apparent when the model was tilted (b). Scale bar, 250 nm.

Discussion

In this study, correlative confocal laser scanning and TEM study was performed on the same whole mount breast cancer cells but at different resolutions (Fig. 1). One of the approaches to combine fluorescence and electron microscopic information within a single image is the method of CFEM (Albrecht et al., 1989; Luo & Robinson, 1992; Svitkina et al., 1995; Svitkina & Borisy, 1998). The CFEM method as outlined in this paper can be applied on whole mount cells to be prepared for TEM (Jahn & Braet, 2008) or for STEM (Braet et al., 2007b; Jahn et al., 2007). Light/laser- and electron microscopy each has its own advantages. CFEM, however, allows to take the full advantage of confocal laser imaging with respect to resolution and Z-stack collection. This information can be later combined with the high spatial resolution data obtained from the electron microscope on the same cell. As such, we can expect that having two solutions to the same problem will result in complimentary information (Peachey et al., 1996). Furthermore, cell position or fiducial markers are crucial to the success of this method and serve as beacons to relocate the area, cell and subcellular region of interest over the different types of microscopes. Nowadays, cells relocate cover slips, EM finder grids and the like are on hand from different microscopy suppliers. Even more, computer-controlled correlative sample stages are also available to translate the coordinates of the sample imaged under live cell imaging conditions automatically to the EM goniometer stage (Sartori et al., 2007). The CFEM approach in combination with such a stage opens up new avenues to close the temporal and spatial resolution gap between four-dimensional light- and three-dimensional EM observations (Mironov et al., 2000; Subramaniam, 2005; Verkade, 2005).

To date the structural and functional aspects of the lamellipodia and filopodia of cells as well as the molecular machinery involved in actin filament organization within the leading edge have been largely explored (reviewed in Yamaguchi et al., 2005; Faix & Rottner, 2006). Serum starvation and EGF exposure of MTLn3 cells all alter these membrane sheet-like structures and finger-like protrusions in different ways (Fig. 2) and may reflect different biological roles (Bailly et al., 1999). For example, the lower number of filamentous protrusions observed in resting MTLn3 cells compared to stimulated cells of this study might be due to a difference in motility-associated signalling pathways (Disanza et al., 2005). In addition, previous studies have shown that this subcellular region contains actin filament free barbed ends that are rapidly generated at the leading edge in response to EGF stimulation (Bailly et al., 1999).

The CFEM approach allowed us to collect additional structural insights into how filamentous cytoskeletal proteins lying in the lamellipodium connect to membrane rafts and structurally organize the filopodia (Figs 3 and 4). This complex, made of interlinked subcellular compartments might, to some extent, explain how raft/cytoskeleton-complexes control lamellipodia and filopodia dynamics (see, vide infra). Besides the extensive filamentous actin cytoskeleton, dot-like structures rich in actin were observed scattered throughout the cytoplasm of control and stimulated MTLn3 cells. Since the number seems to increase under stimulating conditions, a direct or indirect involvement of these peculiar actin structures in motility has to be considered.

Transmission electron tomographic analysis on the same whole mount cells prepared for CFEM gave us a three-dimensional glimpse of the filamentous extensions at the cell edge (Fig. 5). This revealed that the long filamentous protrusions was connected to two single lying filaments which were anchored further back in the intricate cytoskeleton matrix of the lamellipodium. Our data on whole-mount detergent-extracted MTln3 breast cancer cells correspond in part with the recent cryo tomography observations on filopodial protrusions of Dictyostelium cells (Medalia et al., 2007).

A literature survey revealed that more is known about the structural organization and dynamic protrusion mechanisms of lamellipodia than of filopodia (Faix & Rottner, 2006). In this study it became clear that DRM domains lie close to filamentous/dot-like structural complexes (Figs 2 and 3), and that these filopodial structures were anchored in the surrounding filamentous cytoskeletal matrix of the lamellipodial area (Figs 4 and 5). It is well known that membrane-binding proteins interact with cytoplasmic protein complexes, which regulate actin polymerization at the cell cortex (Takenawa & Suetsugu, 2007). Based on the correlative imaging data and the three-dimensional data presented in this study we postulate that membrane rafts play a central role in the cell's ability to form protrusions such as lamellipodia and filopodia. Membrane rafts have been shown to play a key role in regulating many cell biological processes (Michel & Bakovic, 2007). In our study we demonstrated that membrane remnants within MTLn3 are closely associated with the cytoskeleton. Membrane rafts lay as discrete patches in the plasma membrane of cells and are known as Triton X-100 resistant structures rich in sphingolipids and cholesterol (Edidin, 2003). These structures are described to be temperature sensitive, able to form large clusters and also to interact with the actin cytoskeleton (Mayor et al., 2006).

We therefore hypothesize that the membrane rafts harbour the family of membrane-binding proteins which trigger the WASP and/or WAVE protein complexes, which in turn activate the ARP2/3 or formin-mediated pathway, and consequently activate actin polymerization and so, regulate lamellipodium and filopodium dynamics (Takenawa & Suetsugu, 2007). It has been shown in previous studies that the leading edge of MTLn3 cells is positive for ARP2/3 (Bailly et al., 1999). The close structural association of the rafts with the cytoskeleton in general and the filamentous/dot-like structural complexes in particular in the lamellipodium suggest that they are all functionally linked together and consequently regulate the cell's fingertips dynamics.

In conclusion, in this study, breast cancer cells were studied as whole mounts using CFEM techniques (Figs 2 and 4). In brief, cells were grown on Formvar-coated finder nickel grids for whole mount CFEM investigation, and preparation steps included prefixation, detergent-extraction, labelling, fixation and drying of the sample (Lindroth et al., 1992). In this study we also illustrated the power of CFEM in combination with labelling techniques to identify and correlate different subcellular compartments at both the optical and electron microscopy levels (Figs 2 and 4). From these findings it is clear that the application of two different, but complementary with respect to sample preparation, high-resolution correlative microscope methods facilitates the collection of structural data that cannot be resolved by classical or standard microscopy methods. Furthermore, the general applicability of this CFEM imaging approach has also recently been proven successful in the study of whole-mounted hepatic endothelial (Braet et al., 2007b), colorectal cancer- (Jahn & Braet, 2008) and plant cells (Barton et al., 2008). Moreover the samples prepared for CFEM investigation were also suitable for subsequent electron tomographic imaging (Fig. 5). Nowadays, transmission electron tomography is the gold standard to obtain high-quality three-dimensional data at 2–4 nm resolution about subcellular structures of interest (Medalia et al., 2002).

In the near future, we expect that electron tomographic investigation of intact vitrified cancer cells in combination with fine structure immunogold labelling studies for actin-binding proteins (e.g. N-WASP, WAVE, etc) and membrane rafts (e.g. lysenin, poly [ethyleneglycol] cholesteryl ether, etc.) will undoubtedly assist us to further dissect the fine structural machinery of lamellipodia and filopodia of cancer cells in three dimensions.

Acknowledgements

The authors are grateful to Prof. J. Condeelis (Albert Einstein College of Medicine, Bronx, NY, USA) for kindly providing us with the MTLn3 cell line. The authors are grateful also to Dr. R. Whan for expert confocal microscopy advice. We want to thank Prof. C. dos Remedios (Muscle Research Unit, The University of Sydney) for reading the manuscript and for expert advice. This work was supported by the ‘NHMRC 402510’ and the ‘Australian Research Council Grant ID DP0881012’.

Ancillary