Introduction

The resolution of standard optical microscopy is limited to ~200 nm because of lights diffraction limit, thus restricting the structural detail that can be achieved when exploring biological samples. To overcome this problem, a range of scanning probe microscopic (SPM) techniques have been developed with considerably higher (sub-nanometer) resolution (Davies et al. 2005; Engel et al. 1997). Atomic force microscopy (AFM) is an SPM, which because of its high-resolution capabilities has proved to be a powerful tool in cell biology, and is being increasingly utilised to image biological material. Figure 1 is a schematic representation of the AFM instrument, which generates an image by systematically scanning a sharp tip mounted on the end of a flexible cantilever over the samples’ surface. As the tip transverses the sample it interacts with the surface causing the cantilever to bend. An optical lever consisting of a laser beam deflected from the end of the cantilever to a position sensitive photo-diode (PSPD) measures the cantilever deflection, thereby generating a topographic map of the surface. The AFM can therefore provide structural detail at nanoscale-resolution that is comparable with scanning electron microscopy (SEM). However, AFM has a distinct advantage over SEM since it can be used in ambient air or under liquid, thereby negating the need for extensive sample preparation and vacuum environments necessary for SEM, and allowing for analysis under near physiological conditions. Furthermore, AFM, by measuring forces within and between biological molecules, can provide additional biophysical information on molecular characteristics by acting as a sensor that quantifies the interaction forces between the tip and sample as they are brought into and out of contact (Butt et al. 2005). AFM-based investigations have therefore revealed nanoscale detail on both fixed and living cells, isolated organelles and individual biomolecules. Some elegant examples of which include imaging the dynamic mechanisms that govern RNA polymerase transcription (Crampton et al. 2006); functional analysis of nuclear (bio)molecular transport through nuclear pore complexes (Kramer et al. 2007); and vascular endothelial growth factor receptor-2 spatial location and ligand binding kinetics in individual cells (Lee et al. 2007).

Fig. 1
figure 1

Schematic representation of the AFM instrument

Full size image

As a stand-alone technique, AFM can provide unique biophysical and ultra-structural information on a sample, but a limitation is the difficulty in correlating structural or mechanical features with functionality (Bowen et al. 2000; Kellermeyer et al. 2006). Hence, several studies have utilised both AFM and optical microscopy to investigate structure-function relationships in biological systems. For example, stand-alone AFM and fluorescence microscopy have been utilised to demonstrate the association of beta-1 integrin with flexible actin-based ridges on melanoma cell surfaces (Poole and Muller 2005) and to detect and measure the mechanical binding properties of HSP60 on stressed and unstressed human endothelial cells (Pfister et al. 2005). However, there are technical difficulties associated with independent microscopic evaluation of the same sample, which can be overcome by coupled AFM and confocal laser scanning microscopy (CLSM).

Integrated AFM–CLSM is a new development that harbours many exciting possibilities for functional interpretation of cellular structures (Owen et al. 2006). It is a powerful combination that enables parallel analysis of the same sample to provide complementary information on several properties simultaneously. The AFM component provides nanoscale structural imaging and quantitative force measurements, while the CLSM contributes depth resolution and fluorescence imaging. The combination of these high-powered imaging systems therefore increases the information that any one technique can generate alone, providing novel mechanistic insights into the relationships between cellular structure and the function of specific molecules.

To harness the versatility of coupled AFM–CLSM, there is a need to standardise the optimal sample preparation requirements to accommodate both microscopic techniques simultaneously. Recently, Kassies et al. (2005) described a custom-designed atomic force fluorescence microscope with which they demonstrated its potential by concurrently imaging membrane fragments from the photosynthetic bacterium Rhodobacter sphaeroids that contain light harvesting complexes. However, imaging intact cells (particularly mammalian cells) that require fluorescence labelling of specific components prior to analysis is more demanding as a delicate balance must be achieved to ensure a strong fluorescence response whilst maintaining cellular architecture. In the present study, we elected to focus upon filopodia as they are notoriously difficult to image because of their dynamic, fragile nature and nanometer dimensions. Consequently, during sample preparation, they are easily damaged and lost; hence, successful imaging has routinely relied on SEM-based approaches (Nakamura 2001). Using filopodia as a model, we therefore describe the development, evaluation and application of an efficient and reproducible sample preparation methodology required to probe the molecular composition of cellular ultra-structural elements by an AFM–CLSM multi-component imaging approach.

Materials and methods

Cell culture

The PNT2 prostate epithelial cell line was obtained from the European Collection of Cell Cultures (Salisbury, UK). The cells were maintained in RPMI 1640 (Gibco, Paisley, UK) supplemented with 10% foetal calf serum and 1% glutamine at 37°C/5% CO2. Growth media was refreshed every other day, and the cells were sub-cultured when they reached 80% confluency.

For imaging, cells were grown on sterile 25 mm circular glass coverslips. Three fixation methods were used, all performed at room temperature: (1) 3% glutaraldehyde in cacodylate buffer (0.1 M sodium cacodylate, 10.8 mM HCl at pH7) for 10 min, followed by a 10 min rinse in distilled water and dehydration through an ethanol series; (2) 4% formaldehyde (Sigma, Dorset, UK) in PBS for 15 min, followed by 3 × 5 min washes in PBS/100 mM glycine; (3) 3.7% ultra-pure methanol-free formaldehyde (Park Scientific Ltd., Northampton, UK) in PBS for 10 min, followed by 3 × 5 min washes in PBS/100 mM glycine. All coverslips were subsequently air-dried prior to imaging.

AFM imaging

Images of glutaraldehyde fixed samples were obtained using the Dimension 3100 atomic force microscope (Veeco Instruments, Cambridge, UK) in tapping mode. Silicon cantilevers (Olympus, Hertfordshire, UK) with resonant frequencies of 288–338 kHz and spring constants of 37.4–61.3 N/m were used. An imaging force of ~10−8 N was maintained, while scan rates were at 0.5 Hz and resolution was 512 × 512 pixels. Appropriate cells to image with the atomic force microscope were located with a light microscope attachment and subsequently multiple scans with increasing magnification were obtained for each region of interest. All ensuing image processing were undertaken with the Nanoscope software (Veeco Instruments, Cambridge, UK).

Multiple tip artefacts can arise from the interaction of soft biological materials with the distal end of the tip, particularly over convoluted cellular regions, where structures may be manipulated on the sample surface (particles may be moved or deposited). This along with adhesion forces between the sample and AFM tip can cause tip contamination. Thus, during imaging special care was taken to avoid damage through abrasion by slow scan rates (0.5 Hz), a relatively light imaging force (10−8 N), and the tips were also frequently replaced.

Immunofluorescence

All stages were performed under ambient conditions unless otherwise stated.

For direct fluorescence labelling of F-actin with Alexa-594, the coverslips were rinsed with PBS, permeabilised with 0.1% Triton-X, blocked in 1% BSA/PBS for 30 min and then incubated with 0.165 μM Alexa Fluor 594-Phalloidin conjugate (Invitrogen, Paisley, UK) for 20 min. To label F-actin and moesin with quantum dots, the coverslips were prepared as above; however, after permeabilisation endogenous biotin was blocked using Biotin Blocking Kit (Invitrogen, Paisley, UK) according to manufacturers’ instructions. Following blocking with 1% BSA/PBS, the coverslips were incubated with either 2 μM biotinylated-phalloidin (Invitrogen, Paisley, UK) for 20 min, or primary moesin antibody (New England Biolabs, Ipswich, UK) diluted 1:100 at 4°C overnight. They were washed with PBS, and then 20 nM Qdot 655-streptavidin conjugate or 20 nM Qdot 525 goat anti-rabbit IgG conjugate (Invitrogen, Paisley, UK) were applied for 30 min or 1 h, respectively. Prior to imaging, the coverslips were rinsed in PBS to remove excess unbound quantum dots, dehydrated through an ethanol series and air-dried.

Coupled AFM–CLSM imaging

The surface morphology of the fixed and fluorescently labelled cells was investigated utilising a PSIA XE-120 atomic force microscope (Park Systems, Suwon, South Korea), equipped with nano-positioning sensors in all three axis and closed-loop feedback for precise, repeatable scanning with sub-nanometer resolution. The AFM laser wavelength was 830 nm. The images were scanned in ‘non-contact’ mode with a resolution of 256 × 256 pixels and scan rates of 0.35–0.5 Hz, using NCHR cantilevers (Park Systems, Suwon, Korea) of 125 ± 5 μm length and a resonant frequency of ~320 kHz. The manufacturers’ specified spring constant was 42 N/m and the typical tip radius was better than 10 nm.

Using a specially machined adapter, the atomic force microscope was integrated with an IX80 inverted microscope (Olympus, Hertfordshire, UK) that was coupled to a Fluoview FV1000 confocal laser-scanning microscope (CLSM; Olympus, Hertforshire, UK). The CLSM scan resolution was typically 512 × 512 pixels with a scan rate of 2.7 s/frame, while the laser excitation wavelength and optical light path and filters were set appropriate for the designated fluorescent label.

All samples were prepared on sterile 25 mm diameter glass cover slips, which located directly within the AFM liquid cell facilitating both wet and dry scanning. The AFM’s built-in optical system, together with the inverted microscope/CLSM allowed for precise positioning of the AFM tip relative to the cellular region of interest and ensured that both the AFM scans and the CLSM images were of identical locations.

Results

Ultra-structural analysis of cell–cell contact sites

Initially, samples fixed with both glutaraldehyde and formaldehyde were imaged by AFM to compare the ultra-structural preservation provided by the fixatives. Glutaraldehyde was found to preserve the cellular architecture to a greater degree than formaldehyde. The highest point of the cells examined was over their nuclear region (Fig. 2), this equated to 1.55 μm (±0.4 μm) on glutaraldehyde fixed cells, while those fixed with formaldehyde had heights of 1.22 μm (±0.3 μm). Thus, cells fixed with formaldehyde demonstrated an approximate 21% reduction in cellular height in addition to increased surface pitting as compared to those fixed with glutaraldehyde.

Fig. 2
figure 2

Light microscopy and topographic AFM images of PNT2 cells. a The boxed region within the light microscopic view of the cells was imaged sequentially at increasing resolution by AFM. L lamellopodia, F filopodia. b The filopodia (boxed region) were further resolved by AFM to demonstrate the quilted appearance of their surface structure and a diameter of approximately 100–150 nm. c Two PNT2 cells in close proximity with their respective filopodia interacting with each other

Full size image

As demonstrated in Fig. 2, large regions such as the nucleus could be clearly defined, whilst high-resolution imaging of the intercellular space between two prostate epithelial cells in close proximity resolved the nanoscale surface detail of peripheral cellular extensions. These included multiple lamellipodia, the width and length of which varied greatly, but typically spanned approximately 100 μm2 and were often observed to have a height of less than 250 nm. Additionally, filopodia with approximate diameters of 100–300 nm and heights of <50 nm were found to extend up to 5 μm from their originating lamellipodium. As shown in Fig. 2b, the ultra-structural analysis of the filopodia by AFM revealed indentations all along their surface resulting in a “quilted” appearance. This structure was consistently observed in multiple images of filopodia on all cells analysed.

When two closely positioned cells were examined, very fine filopodia with widths of 100–150 nm originating from the adjacent cells could be clearly seen entwining with each other and likely represent the early stages of cell–cell contact responsible for promoting adhesion (Fig. 2c). However, at this stage stand-alone AFM can tell us little more of their molecular constituents. Thus, due to the nanometer scale of these very fine structures, they were selected as a model to develop and optimise the appropriate technique required to maintain cellular ultra-structures while simultaneously labelling them with fluorescent tags for further analysis by AFM–CLSM.

Sample preparation for coupled AFM–CLSM

Fluorescence labelling of specific proteins (detected by CLSM) enables functional interpretation of structures identified by AFM. Thus, to further investigate the molecular composition of the filopodia, the necessary experimental conditions required to maximise fluorescence response while maintaining ultra-structure needed to be optimised to facilitate AFM–CLSM. The first complication to be addressed was the most suitable technique for fixation, so the PNT2 cells were fixed with glutaraldehyde, formaldehyde and ultra-pure methanol-free formaldehyde. The F-actin component of the cytoskeleton was then directly fluorescently labelled with an Alexa Fluor 594-phalloidin conjugate.

Glutaraldehyde is the preferred fix for preserving cellular structure and as can be seen in Fig. 2, enables very high topographic resolution. Filopodia are very delicate and often do not survive fixation, yet as Fig. 2 clearly demonstrates, this method preserves the ultra-structure of these cellular ultramicroextensions very well. However, this fixative is unsuitable for fluorescence imaging as it causes excessive autofluorescence—a consequence of the molecules’-free aldehyde groups that will react with amino groups in the labelling antibodies, resulting in a fluorescence response where-ever tissue-bound glutaraldehyde is located (Fig. 3b). This fixative was therefore not suitable for AFM–CLSM.

Fig. 3
figure 3

Concurrent AFM–CLSM imaging of (a, b) glutaraldehyde fixed cells with well-preserved structure, but high background fluorescence; (c, d) ultra-pure formaldehyde fixed cells with slight structural collapse but more specific fluorescence response for F-actin, albeit highly photobleached

Full size image

An alternative is the use of formaldehyde as a fixative as it results in considerably more specific fluorescence signals with reduced background noise; we therefore assessed formaldehyde and ultra-pure methanol-free formaldehyde. We found that fixing the cells in ultra-pure methanol-free formaldehyde resulted in improved fluorescence responses as compared to using regular formaldehyde (data not shown), which usually contains 10–15% methanol to inhibit its oxidation to formic acid. This observation was most likely a consequence of the mechanism by which fixation is achieved with each of these solutions—formaldehyde cross-links proteins, while methanol promotes protein clumping which can affect the quality of the fluorescence signal and structural preservation. When compared to glutaraldehyde fixed cells, formaldehyde fixed cells demonstrated more structural collapse and the fine surface detail of the filopodia was no longer discernable. However, to preserve cellular structure while achieving specific fluorescence signals, high quality methanol-free ultra-pure formaldehyde was the preferred fixative (Fig. 3c, d).

Despite the more specific signal achieved with the ultra-pure formaldehyde fixative, the Alexa Fluor 594 fluorophore proved to be an obstacle primarily because severe photobleaching occurred due to the large number of confocal scans required to zoom in on the appropriate region that corresponded to the AFM scan area. This rapid fading during analysis, combined with the very low fluorescence response that was located in the physically small areas of interest (<150 nm), meant that the fluorescence images were of low resolution, with poor definition (Fig. 3d). Consequently, alternative fluorophores were required, so we elected to assess quantum dots over conventional fluorophores due to both their enhanced quantum yields and photostability. This would likely increase sensitivity and minimise the effects of photobleaching, thereby enabling the detection of smaller fluorescence signals.

The target protein F-actin was indirectly labelled with quantum dots, which were found to substantially improve the quality of images achieved with AFM–CLSM. Quantum dot-conjugated secondary antibodies consistently resulted in strong fluorescence despite the need for continuous scanning to locate the nanometer-scale regions of interest. Furthermore, the size of these areas did not prove to be a limitation in achieving a fluorescence response at those sites (Fig. 4). Thus, we recommend nanocrystal labelling of the molecular targets under study when employing integrated AFM–CLSM.

Fig. 4
figure 4

Concurrent AFM–CLSM imaging of PNT2 cells with quantum dot labelled F-actin (QD655, red) and moesin (QD525, green). AFM-derived amplitude images of cell structure (a, c); CLSM fluorescence images (b, d). Arrows in c and d indicate regions where topographic structures correspond to clusters of fluorescence

Full size image

Molecular make-up of filopodia

To assess the molecular elements that formed the filopodia we utilised our optimised sample preparation technique for integrated AFM–CLSM. PNT2 cells grown on glass coverslips were fixed with ultra-pure methanol-free formaldehyde, and then F-actin and moesin were differentially labelled with quantum dot nanocrystals.

The fine filopodia were well resolved in the AFM images (Fig. 4), and when the fluorescence images of those same regions were examined, both F-actin and moesin were observed within the delicate protrusions, but to varying degrees. F-actin was more concentrated at the base of the filopodia, but did not appear to extend throughout their length (Fig. 4a, b). Additionally, there was little correlation between the topographic structure of the epithelial cells and the location of the F-actin fluorescence signals.

In contrast, moesin was found to concentrate in the central region of the filopodia, extending through their entire length (Fig. 4c, d). Furthermore, high ridges scattered over the apical cell surface resolved by the AFM corresponded very well to punctate clusters of moesin fluorescence, indicating the protein is also likely to be located within these ultra-structural features.

Discussion

With recent advances in the global gene expression and proteomics fields, there are increasing demands on determining the roles of these molecules and how they relate to specific cellular responses, structures and abnormalities associated with disease. Light and fluorescence microscopy have been key tools in trying to decipher such interactions, but the diffraction limit of light stunts the resolution that can be achieved. Consequently, tools such as the AFM have evolved and are now being increasingly applied to investigate biological ultra-structures. However, to obtain functional information it is necessary to combine optical and AFM imaging of cells to correlate high-resolution structural information with functional labelling of specific components. To facilitate a multi-component approach to imaging cells, appropriate sample preparation plays a vital role particularly as the conditions that favour one technique may not complement a different type of analysis. In the current study, we therefore describe the optimised technique required for combined high resolution and fluorescence imaging of human cells using an AFM–CLSM hybrid instrument.

The major obstacle in achieving high quality cellular images with enhanced resolution is appropriate sample preparation, particularly as ultra-structural elements are notoriously delicate. This is exemplified by filopodia, fine protrusions that emanate from the lamellipodium (thin sheets of actin filaments) of motile cells. They are highly dynamic structures that are rapidly assembled and disassembled, with important roles in motility, guidance and adhesion. However, due to the transient and fragile nature of filopodia and the limited resolution achieved by light microscopic techniques, their ultra-structure has remained largely uncharacterised.

Filopodia often do not survive fixation; however, the glutaraldehyde method employed in the current study preserved the ultra-structure of these cellular protrusions very well (Fig. 2). For the first time, we have shown that they have a “quilted” surface structure with widths as narrow as 100 nm and appeared to sense the cells location, reaching out for and becoming interdigitated with those from adjacent cells to initiate and promote cell–cell adhesion. This correlates well with the current data that indicate the formation of adherens junctions, desmosomes, and focal adhesions are initiated when filopodia from adjacent cells engage with each other (Borm et al. 2007; Vasioukhin et al. 2000).

Glutaraldehyde is able to better preserve cellular ultra-structure than formaldehyde, possibly due to the increased speed at which glutaraldehyde is able to penetrate cells and cross-link proteins as compared to formaldehyde, which stabilises biological materials comparatively more slowly (Kiernan 2000). However, despite the superior fixation efficiency of glutaraldehyde for persevering ultra-structure, it is not suitable for fluorescence imaging as its free aldehyde groups react with amino groups in the antibodies required for labelling proteins of interest, resulting in excessive background fluorescence that masks any specific signals. Thus, for combined AFM–CLSM imaging an alternative fixation method is required and in the current study, we found that ultra-pure methanol-free formaldehyde provided the best compromise. Strong and specific fluorescence responses were achieved while the cellular structure was well preserved, but it was noted that some of the fine topographic detail (resolved after glutaraldehyde fixation) was lost when formaldehyde was used. Nonetheless, the advantage of concurrent fluorescence and high-resolution AFM–CLSM imaging offered by the ultra-pure formaldehyde fixation outweighs the slightly compromised surface detail achieved.

The second problem encountered while trying to develop AFM–CLSM imaging was severe photobleaching of traditional organic fluorophores. Multiple scans were often required to zoom into the appropriate regions of interest and this together with the fact that these regions were physically small resulted in very poor fluorescence images. This was overcome by utilising quantum dots in place of the conventional fluorophore labels. These semiconductor nanocrystals offer many advantages including enhanced photostability, high quantum yield (percentage of absorbed photons resulting in emitted photons), and broad excitation with narrow emission spectra (Jaiswal and Simon 2004; Medintz et al. 2005). Quantum dots indeed proved to be substantially more sensitive and resilient labels that enabled concurrent AFM–CLSM imaging, which demonstrated moesin is located throughout the entire length of filopodia, while F-actin is primarily concentrated at their base. The location of moesin and F-actin within filopodia have previously been shown by Amieva and Furthmayr (1995), who demonstrated that F-actin staining was often complementary to moesin statining, with F-actin locating at the base of filopodia while moesin was present within the core of “microextensions” and in “microspikes” on the cell surface. However, harsh, time consuming, expensive and technically demanding methods were required in such studies i.e. immunoelectron microscopy where the moesin protein was labelled with gold particles. Coupled AFM–CLSM holds many advantages over these more traditional techniques as sample preparation is simpler and fluorescence labelling is considerably more versatile.

Furthermore, in the current study, we also observed fluorescence signals corresponding to moesin at locations on the apical cell surface where high ultra-structural ridges had been imaged by the AFM. To the authors’ knowledge, these surface features may therefore represent the first direct visualisation of the sites where moesin is responsible for linking transmembrane proteins to the actin components of the cell cytoskeleton. This anchorage is vital in establishing epithelial cell polarity and promoting the highly ordered cytoskeletal network (Edwards et al. 1997; Speck et al. 2003).

We have demonstrated that with well-optimised sample preparation stages, fluorescence imaging can be achieved concurrently with high-resolution ultra-structural imaging, providing an important platform for further study into the molecular constitution of diseased cells or tissues in relation to their structural characteristics. In this study, we took topographic features that had been visualised by AFM and then investigated their molecular make-up. However, inverse studies can also be performed where a fluorescence region of interest is identified and then the underlying topography at the nanometer scale is focussed upon. Additionally, AFM–CLSM has the potential to be applied to live cells, or coupled to AFM-derived force measurements for further biophysical analysis. This technique will therefore not only provide significant advances in the imaging of cellular structures, but will enable the characterisation of their molecular constituents, biophysical properties and functional aspects. AFM–CLSM is therefore a powerful new tool for the life sciences that has the potential to provide novel insights to further our understanding of cellular structure-function relationships.