TIR-FSM - Wiley Online Library

Journal of Microscopy, Vol. 216, Pt 2 November 2004, pp. 138–152 Received 29 March 2004; accepted 31 July 2004

Signal analysis of total internal reflection fluorescent speckle microscopy (TIR-FSM) and wide-field epi-fluorescence FSM of the actin cytoskeleton and focal adhesions in living cells Blackwell Publishing, Ltd.

M . C . A DA M S , A . M AT OV , D. YA R A R , S . L . G U P T O N , G . DA N U S E R & C . M . WAT E R M A N - S T O R E R Department of Cell Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, U.S.A.

Key words. Cell migration, fluorescent speckle microscopy, total internal reflection FSM.

Received 29 March 2004; accepted 31 July 2004

Summary Fluorescent speckle microscopy (FSM) uses low levels of fluorescent proteins to create fluorescent speckles on cytoskeletal polymers in high-resolution fluorescence images of living cells. The dynamics of speckles over time encode subunit turnover and motion of the cytoskeletal polymers. We sought to improve on current FSM technology by first expanding it to study the dynamics of a non-polymeric macromolecular assembly, using focal adhesions as a test case, and second, to exploit for FSM the high contrast afforded by total internal reflection fluorescence microscopy (TIR-FM). Here, we first demonstrate that low levels of expression of a green fluorescent protein (GFP) conjugate of the focal adhesion protein, vinculin, results in clusters of fluorescent vinculin speckles on the ventral cell surface, which by immunofluorescence labelling of total vinculin correspond to sparse labelling of dense focal adhesion structures. This demonstrates that the FSM principle can be applied to study focal adhesions. We then use both GFP-vinculin expression and microinjected fluorescently labelled purified actin to compare quantitatively the speckle signal in FSM images of focal adhesions and the actin cytoskeleton in living cells by TIR-FM and wide-field epifluorescence microscopy. We use quantitative FSM image analysis software to define two new parameters for analysing FSM signal features that we can extract automatically: speckle modulation and speckle detectability. Our analysis shows that TIR-FSM affords major improvements in these parameters compared with wide-field epifluorescence FSM. Finally, we find that use of a crippled eukaryotic expression promoter for driving low-level GFP-fusion protein expression is a useful tool for FSM imaging. When used in time-lapse mode, TIR-FSM of actin and GFP-conjugated focal adhesion Correspondence to: Dr Clare Waterman-Storer. Tel.: +1 858 784 9764; fax: +1 858 784 9779; e-mail: [email protected]

proteins will allow quantification of molecular dynamics within interesting macromolecular assemblies at the ventral surface of living cells. Introduction Fluorescent speckle microscopy (FSM) is a powerful new method that utilizes high-resolution epi-fluorescence digital light microscopy to analyse the movement, assembly and disassembly of macromolecular structures in vivo and in vitro (Waterman-Storer et al., 1998; reviewed in Danuser & WatermanStorer, 2003). Initially, we devised this method for studying the dynamics of the cytoskeletal polymers, microtubules and actin filaments, in living cells as they locomote or divide. In FSM of the cytoskeleton, polymers are assembled from a very low fraction of fluorescently labelled protein subunits together with unlabelled protein subunits and are imaged with highresolution optics and a sensitive, low-noise, high-dynamic-range digital camera. FSM is achieved in living cells by microinjection or expression of a low amount of fluorophore-conjugated or green fluorescent protein (GFP)-fused polymer subunits that co-assemble with endogenous unlabelled subunits. Stochastic variation in the number of fluorescent subunits per resolution-limited image region assembled into the polymer results in a ‘speckled’ appearance of the polymer structure in high-magnification, high-resolution fluorescence images (Waterman-Storer & Salmon, 1998). In time-lapse FSM of the dynamic cytoskeleton in living cells, movement and changes in speckle intensity act as spatially local reporters for polymer movement, assembly and disassembly throughout the field of view. FSM is fast becoming the method of choice for studies of cytoskeleton dynamics (reviewed in Danuser & Waterman-Storer, 2003), and we have developed automated computer-based image analysis tools (henceforth referred to as ‘qFSM analysis software’) that exploit the wealth of quantitative molecular © 2004 The Royal Microscopical Society

S I G NA L A NA LYS I S I N F S M

dynamic information generated in FSM movies (Ponti et al., 2003; Vallotton et al., 2003). The contrast of the speckle pattern is a critical parameter in FSM imaging. Indeed, the higher the contrast of the speckles, the more readily they can be tracked either by eye in movies or by our qFSM analysis software. Speckle contrast was originally defined for microtubule images as the standard deviation of the number of fluorescent subunits per resolution-limited image region relative to the mean number of fluorescent subunits in all such image regions along the length of the microtubule (Waterman-Storer & Salmon, 1998). Thus, speckle contrast depends indirectly on the size of the resolution-limited image region and the ratio f of fluorescent to non-fluorescent subunits (Danuser & Waterman-Storer, 2003). In three dimensions, the resolution-limited image region is defined by the point-spreadfunction (PSF) of the microscope, whose dimensions in x–y are approximated by the Rayleigh limit r = 0.61λ/NA and whose dimension in z is approximated by the depth of field Z = n λ/ NA2 (where λ is the emission wavelength of the fluorophore, n is the refractive index of the medium between the lens and object, and NA is the numerical aperture of the objective lens). Thus, the best contrast in FSM calls for the highest NA optics available and the lowest level of labelled subunits. In theory, optimal contrast occurs when speckles represent the resolutionlimited image of a single fluorophore (Waterman-Storer & Salmon, 1999; Watanabe & Mitchison, 2002). In practice, however, too low fractions result in a very low number density of speckles, which reduces the spatial sampling of the molecular dynamic events being analysed, and the camera exposure times required reduce the temporal resolution of the analysis. In addition, single fluorophore imaging is quite technically demanding due to the exceedingly high signal-to-noise ratio required, and thus limits the potential of the technique to reach a wider audience. To aid in our biological studies of tissue cell migration, we sought to improve on current FSM technology in two ways. First, by applying the FSM principle to study other macromolecular assemblies besides cytoskeletal polymers that are important in cell motility, and second, by performing FSM using a higher contrast fluorescence microscopy mode. To approach the first issue, we chose to use focal adhesions as a test case. Cells migrating in tissue culture adhere to extracellular matrix (ECM) proteins bound to glass coverslips via structures called focal adhesions (reviewed in Geiger et al., 2001). These are specialized regions of ∼0.25–50 µm2 in the plasma membrane consisting of trans -membrane integrins that bind to the extracellular maxtrix and cluster into two-dimensional supramolecular complexes along with structural proteins including vinculin, talin and α-actinin that link them to actin filaments inside the cell. The actin filaments in cells migrating on ECMcoated glass form a thin, complex, cortical meshwork along the entire ventral cell surface in which forces and tensions are generated by specialized motor proteins. By linking to the cortical actin, focal adhesions thus transmit tension generated © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152

139

inside the cell to the ECM on the glass coverslip to move the cell. Although the motion, appearance and disappearance of whole focal adhesion complexes consisting of thousands of molecules of have been described (reviewed in Webb et al., 2003), little is known about the spatial and temporal dynamics of the molecules making up each complex. Thus, focal adhesion component molecules make intriguing candidates for FSM analysis of a non-polymeric macromolecular assembly. For cells in culture grown on ECM-coated coverslips, focal adhesions and cortical actin are confined to a region of around a hundred nanometres at the ventral cell surface, making them amenable to analysis by fluorescence microscopic methods that have good optical sectioning capabilities. Indeed, we have previously employed a spinning-disc laser confocal microscope to decrease out-of-focus fluorescence in FSM images of actin and microtubules (Gupton et al., 2002; Salmon et al., 2002; Adams et al., 2003). However, as both actin and focal adhesions initiate within 10 nm of the coverslip, they would also benefit from the limited fluorescence excitation field offered by total internal reflection fluorescence microscopy (TIR-FM, reviewed in Axelrod, 2001a). TIR-FM operates on the principle that when light is directed at or above a critical high angle onto an interface between a higher and lower refractive index medium, the entire incident beam is reflected back into the original medium. However, some of the energy of the reflected beam penetrates a small distance beyond the interface as an exponentially decaying evanescent wave. The depth of penetration of the evanescent wave depends on the beam angle, wavelength and the ratio of refractive indices. The thickness of an evanescent field is in the range of 50 nm to several hundred nanometres for a glass/water interface for light in the visible range. This phenomenon provides an opportunity for biological fluorescence microscopy with very high selectivity for exciting fluorophores in an extremely shallow region abutting the interface between the coverslip and an aqueous biological specimen, without exciting fluorescence in the bulk solution or the intracellular milieu (Axelrod, 2001a). Thus, at a given NA, although TIR-FM does not decrease the volume of the PSF as compared with widefield epi-fluorescence, it does decrease the excitation volume, and thus the number of fluorescent subunits per PSF, which should theoretically improve FSM contrast. Therefore, it would be very useful to combine TIR-FM with FSM (TIR-FSM) and the automated qFSM analysis software to allow quantification of molecular dynamics in focal adhesions and actin filaments at ventral surfaces of migrating cells. Materials and methods Reagents X-rhodamine actin was prepared by covalent linkage of succinimidyl ester of X-rhodamine (Molecular Probes) to lysine residues on purified skeletal muscle actin as described in Waterman-Storer (2002). The cDNA encoding an improved

140

M . C . A DA M S E T A L .

GFP (Cormack et al., 1996) fused in-frame to vinculin under the control of the cytomegalovirus immediate early (cmv) promoter in the pcDNA3 vector (Zamir et al., 1999) was the kind gift of Martin Schwartz (University of Virginia). This construct will henceforth be called cmvGFP-vinculin. To produce a construct for a low level of GFP-fusion protein expression, the promoter region of a vector in which 450 bp of the enhancer of the cmv promoter was truncated to reduce insert expression level (kind gift of N. Watanabe; Watanabe & Mitchison, 2002) was subcloned to replace the full-length cmv promoter in the pEGFP-N1 vector (Clontech). The vinculin cDNA was then was inserted in-frame to the enhanced GFP protein. This construct will henceforth be called ∆cmvGFP-vinculin. Rabbit anti-vinculin antibodies (hVin-1) were obtained from Sigma Chemical Co., Texas-red-conjugated anti-rabbit secondary antibodies were obtained from Jackson Immunoresearch and Alexa-488-labelled phalloidin was from Molecular Probes. Cell manipulations Experiments were performed in PtK1 cells (American Type Culture Collection), which are a rat-kangaroo kidney epithelial line. Cells were grown in a humidified 37 °C incubator in an atmosphere of 5%CO2, 95% air in Ham’s F-12 medium containing 10% fetal bovine serum, 25 mm HEPES buffer and antibiotics. Cells were plated on 22 × 22-mm no. 1.5 coverslips (Corning, Zn/Ti borosilicate, n = 1.523), on which they adhere to serum proteins that coat the glass, as well as their own secreted matrix proteins. Cells were microinjected into their nuclei with a mixture of 0.5–1 mg mL−1 X-rhodamine actin and 0.1 mg mL−1 plasmid DNA encoding either cmvGFP-vinculin or ∆cmvGFP-vinculin using an eppendorf Transjector injection system on a Nikon TE-2000S microscope and microneedles pulled on a needle puller (Sutter Instruments). Within 2–8 h following microinjection, X-rhodamine actin was exported from the nucleus and was incorporated into the actin cytoskeleton and GFPvinculin protein was expressed and visible by epi-fluorescence microscopy. For imaging, coverslips of cells were mounted on a slide with two strips of double-sided sticky tape spaced 1 cm apart. The channel between the strips of tape was filled with cell culture media containing 30 u mL−1 Oxyrase (Oxyrase Inc.) to inhibit photodamage and photobleaching of X-rhodamine. The ends of the channel were sealed with a 1 : 1 : 1 mixture of Vaseline, lanolin and paraffin and the slide was transferred to the prewarmed microscope stage for imaging. To compare GFP-vinculin and total vinculin localization, cells were microinjected with ∆cmvGFP-vinculin-encoding plasmid, allowed to express the protein for several hours in the incubator and then fixed with 4% paraformaldehyde in 50 mm PIPES, 50 mm HEPES, 2 mm EGTA, 1 mm MgSO4, pH = 7.0, lysed in 0.1% Triton X-100 in the same buffer, and processed for indirect immunofluorescence localization of vinculin. For localization of total actin filaments in cells injected

with low levels of X-rhodamine actin, cells were fixed in 4% paraformaldehyde in 100 mm MES, 30 mm MgCl2, 1.38 m KCl, 0.02 m EGTA, pH = 7, lysed in 0.5% Triton X-100 in the same buffer, and stained with fluorescent phalloidin. Multimode TIR-FSM ima ging system A detailed description of this microscope system will be published elsewhere. Briefly, the microscope incorporates throughthe-objective multispectral TIR-FM epi-fluorescence (Axelrod, 2001b), wide-field multispectral epi-fluorescence and DIC microscopies. A Nikon TE2000U inverted microscope with a custom-built dual-port epi-illuminator allowed rapid switching between laser illumination for TIR-FM and arc-lamp illumination for wide-field epifluorescence. This was accomplished by introducing a fibre-optically coupled laser beam via one port, and an HBO 100-W Hg arc lamp via the other port. A 50-mW Kr/Ar air-cooled ion laser (Melles Griot) with lines at 488 nm (∼10 mW, for GFP excitation) and 568 nm (∼15 mW, for X-rhodamine excitation) was used for illumination in TIRFM mode. Laser wavelength and intensity were controlled with a polychromatic acousto-optical modulator (PCAOM) (Neos Technologies). Wide-field epi-fluorescence illumination by the arc lamp was controlled with an electronic shutter/ filterwheel device (Sutter Instruments) containing narrowbandpass excitation filters for 488 nm (± 20 nm) and 568 nm (± 20 nm). Both illumination sources were focused at the aperture plane, and were directed to the specimen by a dual dichromatic mirror (Chroma). A 100× 1.45-NA Nikon Plan Apo objective lens allowed the laser beam to be introduced at the outer edge of the objective aperture to achieve a high angle of incidence of the beam on the coverslip surface to achieve TIR and generation of an evanescent excitation field at the interface between the coverslip and the cell. The position of the beam was easily adjusted with a digital readout micrometer that moved the fibre-optic coupler at its input into the illuminator. We calibrated the micrometer readout to allow for accuracy and repeatability in selecting specific beam angles and thus specific evanescent illumination depths. Fluorescence emission from the specimen was collected by the objective lens and selected by a dual bandpass emission filter. Focus was kept constant during through-the-objective TIR-FSM via feedback from the position of the totally internally reflected laser beam (at the coverslip surface) onto a split photodiode (Mashanov et al., 2001), the voltage reading from which was converted to an 8-bit digital signal that was used to control an Applied Scientific Instruments DC servomotor to correct for focus drifts. Images were collected by a low-noise, spectrally sensitive Hamamatsu Orca II-ERG camera, which contains an ER150 interline CCD chip with a 1344 × 1024 array of 6.45-µm2 pixels, each with a full well capacity of 16 000 photoelectrons, and ∼50% quantum efficiency (QE) for green fluorescence and ∼30% QE for red fluorescence. For this study, the camera was operated in 14-bit mode, in which the readout is relatively slow (1.25 MHz) but the © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152


readout noise is exceedingly low (3–5 electrons rms). Control of all electronic hardware and camera acquisition was achieved with MetaMorph software operated on a Pentium PC. Stage temperature was controlled with a custom-modified stage incubator (20–20 Technologies). FSM ima ge acquisition and anal ysis scheme For our quantitative comparison of TIR-FSM and wide-field epi-fluorescence FSM (WF-FSM) images of focal adhesions or actin, a single cell containing GFP-vinculin and X-rhodamine actin was subjected to the following image acquisition scheme, which was automated using macros written in MetaMorph. First, an in-focus WF-FSM image of the GFP-vinculin focal adhesions was acquired using 488-nm light. This was followed immediately by an ‘intensity series’ of four TIR-FSM images using 488-nm light taken with the beam angle set to give a 105-nm evanescent field excitation depth (henceforth called DSmall) calculated according to Axelrod (2001a). The intensity series was performed by using preselected voltages in the PCAOM that corresponded to 0.5, 1.0, 1.5 and 2.0 mW of beam power exiting the objective lens. The purpose of the intensity series was to allow us to compare the speckle signal between WF-FSM and TIR-FSM modes on positionally corresponding image regions of equal average intensity, because different regions of the same cell will have different average intensities depending on the mode of microscopy used. For example, at a given excitation intensity, thick central cell regions will have lower average fluorescent signal in TIR-FSM than in WF-FSM because of the lower out-of-focus background fluorescence, whereas in the periphery of the same cell, the WF-FSM and TIR-FSM fluorescence signal may be similar. The beam angle was then quickly switched to give an evanescent field depth of 145 nm (DLarge), and a second intensity series of five TIR-FSM images using the same beam power settings as the first was acquired. Following the acquisition of such a ‘GFP image set’, a similar nine-image set of WF-FSM, TIR-FSM DSmall and TIR-FSM DLarge images was obtained of the X-rhodamine actin in the same cell. The only differences were that 568-nm epi-fluorescence and laser light were used, DLarge was set at 170 nm and DSmall at 125 nm, and the power settings used for the intensity series were 1.7, 4.0, 4.4 and 4.6 mW. These higher intensities were used because the actin label tended to be dimmer than the GFP and the camera is less sensitive to red emission than green. Each GFP and X-rhodamine nine-image set took ∼10–15 s to obtain, and both sets were acquired for ten different cells, five expressing cmvGFP-vinculin and five expressing ∆cmvGFP-vinculin. To keep camera contributions equal, all images were acquired using 500-ms camera exposure times. To minimize photobleaching, illumination of the specimen was limited to the time of camera exposure by electronic shuttering, and by acquiring the dimmest illumination and thinnest evanescent field images at the beginning of the image set and brighter high-illumination images at the end of the set. © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152

141

We then used custom-modified modules of our qFSM analysis software (Ponti et al., 2003) to compare quantitatively the speckle signal in corresponding regions of the WF-FSM and the TIR-FSM images taken at DLarge and DSmall. First, ten positionally corresponding 3 × 3-µm regions were selected in all images of both GFP and X-rhodamine image sets of a given cell, and their coordinates were stored in memory. The regions were chosen in the GFP ‘channel’ in areas specifically containing a focal adhesion and transferred to the X-rhodamine channel, because focal adhesions are discrete plasma membrane structures, whereas the cortical actin cytoskeleton covers the entire ventral cell surface. In addition, we wished to determine how the speckle signal was affected by location in the cell. To do this, we chose five of these regions to lie within 10 µm of the cell edge where the cell is thin (< 500 nm) and the contribution of diffusible and out-of-focus fluorescent subunits to the image is minimal, and five regions to lie in central cell regions > 10 µm from the cell edge, where the cell is thicker (up to 10 µm) and such fluorescence background is higher. Otherwise, selection was random, i.e. not biased to focal adhesions that appeared either excessively bright or dim, dense or diffuse, long or wide. The qFSM software, programmed in Matlab, was designed to identify speckles as statistically significant image features above image noise, and then track their time of appearance, their intensity fluctuations, time of disappearance, and their trajectory and velocity of motion in order to extract quantitative information on the assembly, disassembly and movement of the macromolecular assembly into which the fluorescent subunits are incorporated. We utilized the first few steps of the complete algorithm as detailed in Ponti et al. (2003) to analyse the speckle signal in all the selected image regions. Briefly, in order to suppress high-frequency image noise, the image was first convolved with a Gaussian kernel whose full-width, half maximal radius (σ) matched the full-width, half maximal radius of the microscope’s PSF according to the formula σ = 0.21λ/NA as derived in Thomann et al. (2002). Then a local maxima detection mask was applied. We then performed a statistical selection scheme to discriminate between local maxima that represent actual speckles vs. maxima due to lowfrequency contributions from shot and dark noise as follows. The noise model parameters were measured and calculated from image sets according to Ponti et al. (2003), including correction for the offset of the Orca II ERG camera. A local minimum detection mask was applied and a triangular mesh was made with the vertices defined by the local minimum positions, and each maximum was associated with the three minima constituting its circumscribing triangle. The average intensity of the three minima was then assigned to the maximum as its local background, and the intensity difference was calculated. Using the noise model, each local maximum was tested as to whether its peak-to-local background intensity difference was greater than the 95% confidence interval in order to be accepted as a significant speckle.

142


The software was modified to return in a text file the following measurements on each 9-µm2 region: number of statistically significant speckles, density of statistically significant speckles, the average modulation of speckles in the region and the mean region intensity. The density of statistically significant speckles expressed in speckles per µm2 will henceforth be called speckle detectability. The speckle modulation ( M) is a normalized measure of peak-to-background intensity difference, as defined by: M = (PS − BL )/(PS + BL), where PS is the peak speckle intensity and BL the average of three local minima. The speckle density was also calculated for the entire area of each cell for both the GFP and the X-rhodamine signals. In addition, the software was modified to output graphics with all statistically significant speckles overlaid by a mark on the image. Expression time-cour se To determine the rate of accumulation of fluorescent vinculin protein driven off two different eukaryotic expression promoters in living cells, several (10–20) cells were microinjected in their nucleus with plasmid DNA, and transferred to a prewarmed microscope for image collection. The microscope (Nikon TE200) was equipped with a robotic stage that was positionally controlled in the x , y and z axes via a closed loop feedback from linear feedback encoders to DC servomotors (Applied Scientific Instruments). The position of each cell was recorded by MetaMorph software, and was visited for image acquisition at 10-min intervals over a period of 4–12 h. Images were captured using a 40× 0.6-NA Plan Fluor ELWD objective lens on an Orca ER cooled CCD camera (Hammamatsu Photonics). Images of a specimen containing GFP fluorescence were acquired on this imaging system and on the multimode TIR-FM system to calibrate a conversion factor for image intensites taken on the two systems so that approximate comparisons could be made. Results and discussion Actin and f ocal adhesion or ganization in PtK1 e pithelial cells and demonstration of the TIR-FSM principle PtK1 epithelial cells cultured on glass coverslips form colonies of 3 –10 cells that adhere tightly to their neighbours, forming a monolayer with each cell spread out to cover ∼1500– 2500 µm2 (Fig. 1A). The cells at the edges of the colonies are specialized for motility, with a flat, thin (< 500 nm) lamellum and lamellipodium (called the ‘leading edge’) extending 10– 20 µm away from the cell centre, while the nucleus resides in the cell central region which is 10 µm thick or greater. We microinjected cells with low levels of X-rhodamine actin or a GFP-vinculin plasmid, incubated the cells to allow incorporation of actin into the cytoskeleton or expression of

GFP-vinculin, then fixed and processed the cells to label total actin with fluorescent phalloidin or total vinculin by immunofluorescence. By wide-field epi-fluorescence microscopy, actin filaments stained with phalloidin were organized into a complex meshwork throughout the ventral cell surface, interspersed with thicker bundles of filaments (Fig. 1B,D). Focal adhesions as visualized with anti-vinculin immunofluorescence were organized into distinct foci that were generally concentrated near the leading edge with fewer in central cell regions (Fig. 1C,F). Particularly for focal adhesion staining, details of structures in central cell regions tended to be obscured by out-of-focus fluorescence (Fig. 1C,F, star). As expected, imaging total actin or total vinculin in the same cells by through-the-objective-type TIR-FM (Axelrod, 2001b) using a 100× 1.45-NA objective lens and an evanescent field depth of ∼150 nm greatly improved the qualitative contrast of both actin filaments and focal adhesions, especially in central cell regions (Fig. 1H,J). We then compared the distribution of total actin to the distribution of the injected fluorescent protein. As previously shown in newt lung epithelial cells (Salmon et al., 2002), the low fraction of X-rhodamine actin incorporated in the actin filament system resulted in a dim speckled distribution of red fluorescence seen by WF-FSM in the thin peripheral cellular regions. However, central cell regions tended to have a more even distribution of fluorescent signal (Fig. 1E, star). Areas with somewhat random distributions of fluorescent speckles in FSM images corresponded to f-actin meshworks or fine factin bundles in the phalloidin image, whereas linear arrays of f-actin speckles corresponded to larger bundles of f-actin (compare Fig. 1D and E). When viewed by TIR-FSM, the contrast of fluorescent speckles in the X-rhodamine actin image qualitatively improved dramatically in the cell periphery, and allowed visualization of speckles in central cell regions where they were obscured by background fluorescence in the WFFSM image (compare Fig. 1E and I, stars). To determine if the FSM principle could be applied to the study of focal adhesions, we compared the distribution of the expressed GFP-vinculin protein (Fig. 1G,K) to total vinculin localized by immunofluorescence (Fig. 1F,J). In WF-FSM images, an extremely dim GFP-vinculin signal (Fig. 1G) co-localized with the vinculin antibody staining (Fig. 1F). At the cell periphery some discontinuity (i.e. ‘speckliness’) in the GFP signal could be seen in focal adhesions that by immunofluorescence had a high density and an even distribution of vinculin staining. In central cell regions, the out-of-focus fluorescent background was quite high, and no GFP speckles could be detected at all in anti-vinculin-stained structures. However, when visualized by TIR-FSM, the low level of expressed GFP-vinculin clearly formed linear arrays and clusters of high-contrast GFP speckles (Fig. 1K), which in the anti-vinculin immunofluorescence image corresponded to focal or fibrillar adhesion structures containing a high, even density of vinculin staining (Fig. 1J). This was true both in central and in peripheral cell regions. Indeed, we have performed time-lapse TIR-FSM of GFP-vinculin in living cells © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152


143

Fig. 1. Comparison of total actin filaments and total vinculin to microinjected or expressed low levels of fluorescent actin and vinculin by wide-field epifluorescence and TIR-FM microscopy in cultured PtK1 epithelial cells. (A– C) Low-magnification (40× objective) views of a group of cells seen by phase contrast (A) with total actin filaments labelled with fluorescent phalloidin (B) and vinculin-containing focal adhesions seen by immunolocalization (C). (D–K) High-magnification (100× 1.45-NA objective) views of the leading edges of cells labelled with fluorescent phalloidin ( D,H), anti-vinculin antibodies (F,J), injected X-rhodamine actin ( E,I ) or expressed ∆cmvGFP-vinculin (G,K) seen by widefield epifluorescence (D – G) or TIR-FM (H–I). Stars show the location of central cell regions in which fluorescent speckles are obscured by out-of-focus fluorescence in the wide-field epi-fluorescence images (E,F) but are of high contrast in the TIR-FM images (I,J).

and observed that the speckle pattern is quite dynamic (data not shown), and should be very interesting for future study. These results demonstrate that FSM can be performed with TIR-FM, and importantly, show for the first time that FSM images can be obtained of a non-polymeric macromolecular assembly, i.e. a focal adhesion. Thus, when used in time-lapse mode, TIR-FSM will allow study of the dynamics of GFP-labelled molecules within focal adhesion structures to be achieved. Signal anal ysis of fluorescent speckles in WF-FSM and TIR-FSM Given the major qualitative differences we observed between WF-FSM and TIR-FSM images in fixed cells, we sought to analyse in living cells the quantitative differences in speckle © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152

signal in the two modes of microscopy and for the different fluorescent probes used. In our initial FSM studies of the linear polymer, microtubules, we described the speckle signal quantitatively with the parameter ‘speckle contrast’, defined as the standard deviation of the fluorescence intensity fluctuations along the microtubule divided by the mean intensity along the microtubule, such that high variability in intensity, i.e. speckliness, corresponded to high contrast values (Waterman-Storer & Salmon, 1998). We found in the present study that for more complex two-dimensional macromolecular arrays, such as the cortical actin polymer meshwork or focal adhesions, this measurement did not reflect the contrast level as observed by eye, especially in more central cell regions where background fluorescence was high. Thus, we define here two new parameters

144


to quantify the speckle signal: speckle detectability and speckle modulation. Detectability is the number of speckles detected per unit area that pass a statistical test to determine if they are significantly different from image noise and significantly brighter than their immediately adjacent local background. We use speckle modulation as a normalized measure of the magnitude of peak-to-background brightness of the identified significant speckles, defined as (PS − BL)/(PS + BL), where PS is the peak speckle intensity and BL the intensity of the immediately adjacent background. To obtain these measurements semi-automatically on large quantities of data, we modified our existing qFSM analysis software (Ponti et al., 2003) to output these parameters for user-defined regions of interest in images, as well as the average region intensity so that we could determine how image brightness related to these parameters.

We performed this analysis on a total of 1800 9-µm2 image regions from living cells containing both X-rhodamine actin and GFP-vinculin (10 different cells × 2 fluorescent probes per cell × 9 images per cell × 10 regions of interest per image) (see Methods for details). Half of the cells had expression of GFPvinculin driven from a robust promoter (cmvGFP-vinculin) and half had low-level expression driven from a crippled promoter (∆cmvGFP-vinculin). Of the nine images taken of each fluorescent probe in each cell, one was wide-field epifluorescence, and eight were TIR-FM images, four each taken at two different evanescent field depths (called DLarge and DSmall, see Methods) and over a range of four illumination intensities. Ten regions of interest were analysed per image, five being within 10 µm from the cell edge and five near the cell centre (see Methods). This image acquisition scheme (outlined in Fig. 2)

Fig. 2. Schematic diagram of the image acquisition and analysis scheme for quantitative comparison of WF-FSM and TIR-FSM images. (A) Image acquisition. The image set includes one WF-FSM image and two series of four TIR-FSM images taken at increasing excitation intensities, one series at a larger evanescent field depth (TIR-FM DLarge) and one at a smaller evanescent field depth (TIR-FM DSmall). Camera exposure times for all images was kept constant at 500 ms. The example shown is for the X-rhodamine actin probe. A similar image set was acquired of GFP-vinculin in the same cell. (B) For image analysis, ten 3 × 3-µm regions were chosen, five within 10 µm of the cell edge and five in central cell areas. Statistically significant speckles were then selected for each region and their modulation was measured. © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152


145

Fig. 3. Examples of automated speckle analysis software: comparison of speckle detection for X-rhodamine actin and GFP-vinculin probes in WF-FSM and in TIR-FSM modes at two different evanescent field depths. Each column of images shows the same living cell containing the fluorescent probe shown (dCMV GFP-Vinculin = ∆cmv-GFP vinculin, X-Rhod actin = X-rhodamine actin) using the different imaging modes shown (WF = WF-FSM, TIRF = TIRFSM). The ten 3 × 3-µm image regions that were analysed in each cell are highlighted. The inset shows one such region with the statistically significant speckles detected by the qFSM analysis software highlighted with a red dot.

allowed us to compare the effects of various imaging parameters on the speckle signal. In Fig. 3, examples of speckle detection by our qFSM software in such regions of interest are shown for images of living cells for X-rhodamine actin, cmvGFPvinculin and ∆cmvGFP-vinculin. TIR-FSM enhances detecta bility and modulation of speckles compared with WF-FSM, especiall y in centr al cell regions Table 1 summarizes the results of our comparison between WF-FSM and TIR-FSM at DLarge (effects of evanescent field depth are compared below), showing the average speckle detectability and modulation for the various probes, imaging modes and cell regions. The values reported are averages (of n = 25 for each GFP-vinculin construct and n = 50 for © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152

X-rhodamine actin) obtained from intensity-matched image region pairs, consisting of two image regions (one from the WF-FSM image, and one from the TIR-FSM Dlarge image) from corresponding positions in the cell, matched in average intensity within ∼10%. What is immediately obvious from these average values is that for both GFP-vinculin and X-rhodamine actin, TIR-FSM greatly improves both detectability and modulation of detected speckles produced with both probes as compared with WF-FSM, particularly in central cell regions. Indeed, speckle detectability in the cell centre of GFP-vinculin focal adhesions (for either promoter; effects of promoter discussed below) is improved 29–30-fold by TIR-FSM compared with WF-FSM, and 4–10-fold in peripheral cell regions, whereas GFP-vinculin speckle modulation is improved by TIRFSM five-fold in the cell centre and 2–3-fold at the cell edge.

146


488 nm

568 nm

∆cmvGFP-vinculin

cmvGFP-vinculin

X-rhodamine actin

WF

WF

TIR (DL )

WF

TIR (DL )

Average speckle detectability (speckles µm−2) Cell edge 0.1144 0.5548 0.0748 Cell body 0.0264 0.76 0.0264

0.764 0.7948

0.1564 0.044

0.8612 0.8472

Average speckle modulation (a.u.) Cell edge 0.0984 0.3032 Cell body 0.0528 0.2696

0. 1844 0.158

0.174 0.045

0.2344 0.1838

TIR (DL )

0.1024 0.0312

For X-rhodamine actin, speckle detectability and modulation in the cell body were improved by TIR-FSM as compared with WF-FSM by 19-fold and five-fold, respectively. Note that our measured speckle detectability values are all less than one significant speckle per µm2. This falls far below the 3.5 speckles µm−2 that we have estimated as an upper limit for speckle density in a mathematical simulation of actin FSM images using 1.4-NA optics (Ponti et al., 2003). This discrepancy is probably the result of several factors, but mainly possible structural differences between our simulated actin meshwork and actin meshworks in vivo, and a lack of accounting in our model for background cellular fluorescence and diffusible fluorescent subunits in the cell. We then sorted the measured speckle signal parameters according to the average intensity of the image region, and reported the average detectability and modulation values for all regions that fell within 25 intensity unit bins (for our camera, this is approximately equal to 6.5 units of dynamic range or 50 –80 photons, depending on wavelength) to allow us to compare the parameters obtained by each mode of microscopy across a range of image brightnesses (Fig. 4). Here, we pooled the results from the ∆cmv- and cmv-GFP-vinculin constructs, and only included TIR-FSM results obtained at DLarge. This analysis revealed for both TIR-FSM and WF-FSM for all cellular regions analysed (Fig. 4) that although speckle detectability is lower at dimmer image intensities (< 50 intensity units (iu)) than at brighter image intensities (> 50 iu), modulation of those speckles that are detected is higher. In addition, brighter average image intensity (> 50 iu) has a much stronger effect of reducing speckle detectability and modulation in WF-FSM than in TIR-FSM. Indeed, often in WF-FSM image regions with intensities above 50 iu there were no detectable speckles at all (or only one, giving insufficient data to generate an average). This was most pronounced for WF-FSM at the cell centre (Fig. 4C) owing to the high level of diffusible GFP in the background. In addition, speckle detectability in TIR-FSM of f-actin in the cell centre (Fig. 4C) was best at very high average image intensity (> 125 iu). However, because fluorophore bleaching is severe at the illumination levels required to produce such bright images, it is not a practically useful finding. The results

Table 1. Comparison of average speckle detectability and modulation in TIR-FSM and WF-FSM images. Averages are from intensity-matched image region pairs (see text for details).

do suggest, however, that for our imaging system, TIR-FSM at DLarge gives the best combination of good speckle detectability and high modulation at illumination intensities that give an average image intensity in the range of 25–100 iu (50 –300 photons). To avoid bias from our selection of analysed image regions, we also determined the total number of detectable X-rhodamine actin or GFP-vinculin speckles per unit area of the whole cell (Fig. 5). This analysis confirmed that TIR-FSM affords 4– 10 times better speckle detectability than WF-FSM throughout the area of the cell. This also shows that total GFP-vinculin speckle density is much lower than that of X-rhodamine actin, as expected given that vinculin resides in discrete sites distributed over the ventral cell surface, whereas the cortical actin cytoskeleton covers the entire ventral cell surface (Figs 2 and 3). The ef fects of evanescent fi eld depth in TIR-FSM on spec kle detectability and modulation We next examined the effect of evanescent field illumination depth D on the TIR-FSM signal in images of GFP-vinculin (DSmall = 105 nm; DLarge = 145 nm) and X-rhodamine actin (DSmall = 125 nm; DLarge = 170 nm). To increase the data pool, the values in Table 2 are not from intensity-matched image region pairs, but are the averages for all image regions analysed in each cell region in each mode (n = 250 regions for GFP-vinculin, and 500 regions for X-rhodamine actin). We expected that both detectability and modulation of fluorescent speckles would be higher for the smaller diffraction-limited illumination volume achieved with a smaller evanescent field. Surprisingly, our analysis revealed that average speckle detectability was 1.3–2.0 times less at TIR-FSM DSmall than at DLarge for both actin and vinculin probes in both central and peripheral cell regions (Table 2). This may reflect the fact that some of the ventral actin or focal adhesion structure, especially at the edges of the evanescent field, may not be excited by DSmall because of their distance from the coverslip surface, thus resulting in a reduced speckle intensity and loss of significant speckle signal. In contrast, the average modulation of detected © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152


147

Fig. 4. Effect of image region brightness on GFP-vinculin and X-rhodamine actin speckle modulation and detectability in WF-FSM and TIR-FSM at DLarge in different cell regions. The measured speckle signal parameters were sorted according to the average intensity of the image region, and plotted are the average detectability and modulation values for image regions in that imaging mode and cell region that fell within 25-iu bins. This allowed comparison of the parameters obtained by each mode of microscopy and cell region across a range of image brightnesses. GFP-vinculin averages include values obtained from cells with expression driven off the cmv or ∆cmv promoter.

© 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152

148


Fig. 5. GFP-vinculin and X-rhodamine actin speckle detectability in individual cells. The total number of statistically significant speckles detected in an entire cell was divided by the cell area.

488 nm

568 nm

∆cmvGFP-vinculin

cmvGFP-vinculin

X-rhodamine actin

TIR (DL)

TIR (DL)

TIR (DS)

TIR (DL)

TIR (DS)

Average speckle detectability (speckles µm−2) Cell edge 0.7591 0.4207 0.9038 Cell body 0.9781 0.5156 1.0433

0.7898 0.8329

0.9453 0.8515

0.5573 0.6983

Average speckle modulation (a.u.) Cell edge 0.2788 0.2526 Cell body 0.2534 0.3036

0.2068 0.165

0.2308 0. 1826

0.2335 0.2139

TIR (DS)

0. 1825 0.1385

significant GFP-vinculin and X-rhodamine actin speckles was not different at the two evanescent field depths in image regions at the cell edge. However, modulation was slightly improved at DSmall compared with DLarge (16 –20%) in image regions from the cell centre. These results show that the loss of

Table 2. Comparison of average speckle detectability and modulation in TIR-FSM images taken at two different evanescent field depths. Averages are from all image regions analysed.

speckle detection at thin evanescent field penetration depths, due possibly to proximity effects, does not outweigh the very small benefit gained in increasing speckle modulation, and for most applications of live cell TIR-FSM, thicker evanescent fields (∼140–170 nm) are preferable. However, if high speckle © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152


modulation is needed for imaging molecular dynamics in central cell regions, the thinner evanescent field may offer a slight advantage. Nevertheless, we can conclude that for throughthe-objective type TIR-FSM of living cells, 1.45-NA objective lenses are sufficient, and more expensive 1.65-NA objective lenses that provide exceedingly thin evanescent field depth are unnecessary. We then sorted the measured speckle signal parameters according to the average intensity of the image region, and reported the average detectability and modulation values for all regions that fell within 25-iu bins to allow us to compare the parameters obtained at each D across a range of image

149

brightnesses (∆cmv- and cmv-GFP-vinculin, and all cellular regions results pooled). This showed that speckle detectability was actually higher in brighter (> 50 iu) images taken at TIR-FSM DSmall than at DLarge, especially for X-rhodamine actin (Fig. 6A,C). The discrepancy with the values reported in Table 2 is likely because the range of average region intensities is heavily weighted towards lower values (< 50 iu) that bias the overall averages. We also compared the total number of detectable X-rhodamine actin or GFP-vinculin speckles per unit area of the whole cell (Fig. 5) at the two evanescent field depths. This analysis confirmed that speckle detectability is slightly better at the larger evanescent field depth.

Fig. 6. Effect of image region brightness on GFP-vinculin and X-rhodamine actin speckle modulation and detectability at two different evanescent field depths. The measured speckle signal parameters were sorted according to the average intensity of the image region, and plotted are the average detectability and modulation values for all cell regions at that evanescent field depth that fell within 25-iu bins. This allowed comparison of the parameters obtained at each D across a range of image brightnesses. GFP-vinculin averages include values obtained from cells with expression driven off the cmv or ∆cmv promoter. © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152

150


Expression of GFP-vinculin fr om a crippled promoter greatly improves speckle modulation in TIR-FSM ima ges We next examined the effects on the focal adhesion speckle signal of using either the ∆cmv or the cmv promoter to drive GFP-vinculin expression in cells. The results are shown in Table 1 (for WF-FSM and TIR-FSM intensity-matched image region pairs) and Table 2 (for all TIR-FSM image regions). For WF-FSM, this analysis shows that using the ∆cmv promoter to drive low-level GFP-vinculin expression curiously increased speckle detectability in focal adhesions at the cell edge by 52%, with no effect on detectability in central cell regions. Modulation in detected WF-FSM GFP-vinculin speckles was not affected by choice of promoter at the cell edge, but was improved by 67% in central cell regions by using the ∆cmv promoter. In contrast, for TIR-FSM (at both DSmall and DLarge), GFP-vinculin speckle detectability in focal adhesions was actually decreased (by 6–87%) by use of the low-level ∆cmv promoter, particularly in regions near the cell edge (19–87%), probably due to insufficient signal. However, detectability of ∆cmvGFP-vinculin speckles was less hampered at the large evanescent field depth than at the smaller depth. These differences of the effects of promoter choice between GFP-vinculin speckle detectability in WF-FSM and TIR-FSM images are likely because the outof-focus diffusible fluorescent background that would be produced by higher levels of expression have more of an effect on the speckle signal in WF-FSM than TIR-FSM. In contrast to its effects on speckle detectability, use of the ∆cmv promoter to drive low-level expression of GFP-vinculin for TIR-FSM improved the average modulation of detected speckles as compared with those produced by the cmv promoter, affording improvements of 22–53% at the cell edge and 83–84% in central cell regions. Thus in general, if the robustness of the speckle signal

(i.e. high modulation) is more important than the density of information, use of the ∆cmv low-level promoter to drive GFP-fusion protein expression is preferable. We then sorted the modulation values obtained for all image regions obtained in all TIR-FSM images and averaged them in 25-iu bins, to allow us to compare the measurement obtained with each promoter independent of differences in image brightness (Fig. 7). This analysis revealed that use of the ∆cmv promoter to drive eGFP-vinculin gave major improvements in speckle modulation over the cmv promoter particularly at mid-range to bright image intensities (50–100 iu). This result shows that for a given number of fluorescent molecules in an image region, more GFP-vinculin proteins produced by the ∆cmv promoter incorporate into focal adhesions than when they are produced faster by the cmv promoter. Although we do not fully understand this result, one possible explanation is that fast expression may produce more non-functional protein, perhaps due to folding problems. Keeping in mind all our analyses, we conclude that the best TIR-FSM speckle signal in focal adhesions is obtained using ∆cmvGFP-vinculin at an average adhesion intensity of ∼50– 100 iu (100–300 photons), where modulation is maximized and fluorescence photobleaching is not excessive. The reduced modulation measured at very low image intensities under all conditions (probe, depth) is probably due to problems with signal to noise, where ‘noise’ consists of both camera noise and non-specific background fluorescence. From the camera, dark noise is the most significant contributor to the difficulty in detecting very low fluorophore numbers, and shot noise is minimal (Ponti et al., 2003). Signal that contributes to a general background haze that makes single or low numbers of fluorophores difficult to detect includes fluorescence of the tissue culture media and reflections within the optics.

Fig. 7. Effect of image region brightness on GFP-vinculin speckle modulation using either the cmv or the ∆cmv promoter to drive expression. The measured speckle signal parameters were sorted according to the average intensity of the image region, and plotted are the average modulation values for all cell regions at DLarge that fell within 25-iu bins. This allowed comparison of the parameters obtained at each D across a range of image brightnesses. © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152


In the majority of biological experiments performed in our laboratory, we use microinjection of plasmid DNA into the nucleus as a means by which to obtain protein expression in living cells. This is our preferred method over bulk transfections because the cells recover rapidly, the level of protein expression is directly related to the time after microinjection and this rate is relatively consistent from cell to cell. Because our above analysis showed that image brightnesses in the range of 50–100 iu produce the best speckle signal, we measured the brightness of the GFP signal in images of cells as a function of time (up to 8 h) after microinjection of the plasmid to allow us to determine the optimal time of GFP-vinculin expression needed to achieve the best speckle signal with either the cmv or the ∆cmv promoter. From this, we found that GFP expression from either promoter was linear (data not shown), with an average rate of increase in fluorescence intensity of 49.2 iu h−1 (n = 6 cells) for the cmvGFP-vinculin, and 14.4 iu h−1 (n = 4 cells) for the ∆cmvGFP-vinculin. Thus, to obtain TIR-FSM images of GFP-vinculin focal adhesions in the range of 50 –100 iu would require ∼1–2 h of expression for cmvGFP-vinculin, and at least 3.5 h for ∆cmvGFP-vinculin. Of course, this time would vary according to the size of the protein being expressed and the cell type used, but could serve as a useful starting point for TIR-FSM studies of other GFPconjugated proteins. Summary and conclusions Here, we first demonstrated that low levels of expression of a GFP conjugate of the focal adhesion protein, vinculin, results in clusters of fluorescent vinculin speckles on the ventral cell surface. By immunofluorescence labelling of total vinculin, we see that these speckles correspond to sparse labelling of dense focal adhesion structures, demonstrating that the FSM principle can be applied to study focal adhesions to the substrate. When used in time-lapse mode, FSM of focal adhesion proteins will probably reveal novel information about the molecular dynamics within these structures during cell motility or metastasis. We then used both GFP-vinculin expression and microinjected fluorescently labelled purified actin to compare quantitatively the fluorescent speckle signal in FSM images of focal adhesions and the actin cytoskeleton by TIR-FM and wide-field epifluorescence. We used our qFSM software for image analysis and define two new parameters for analysing FSM image signal features that we can extract automatically for selected cell regions: speckle modulation as a normalized measure of speckle peak-to-background difference and speckle detectability as a measure of the number of significant speckles per unit area. Our analysis shows that TIR-FSM affords major improvements in these parameters compared with wide-field epifluorescence for imaging macromolecular assemblies at the ventral surface of living cells, both in thin peripheral and in thick central cell regions. Less dramatic, but still significant improvements in TIR-FSM speckle signal modulation can © 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152

151

be achieved by use of a thin evanescent field depth when imaging central cell regions or for GFP-vinculin by driving expression with a crippled promoter. By analysing the speckle signal parameters over a range of image intensities, we found the optimal image intensity for the best compromise between speckle modulation and fluorophore bleaching. The values reported in this study should serve as a benchmark for those who wish to pursue WF-FSM or TIR-FSM imaging, and provide the first demonstration that FSM can be applied to study macromolecular assemblies besides cytoskeletal polymers in living cells. This should open the door to analysis of a wide array of cellular structures in a host of interesting cell-biological applications. Acknowledgements We wish to thank members of the Waterman laboratory for helpful suggestions. We are grateful to John Zemek and Gary Rondeau at Applied Scientific Instruments and Steve Ross and Kunio Toshimitsu at Nikon Instruments for valuable input on the multimode TIR-FSM microscope. This work was supported by NIH GM66050 to C.M.W.S and G.D. C.M.W.S. is a fellow of the Keith R. Porter Endowment for Cell Biology. References Adams, M.C., Salmon, W.C., Gupton, S.L., Cohan, C.S., Wittmann, T., Prigozhina, N. & Waterman-Storer, C.M. (2003) A high-speed multispectral spinning disc confocal microscope system for fluorescent speckle microscopy of living cells. Methods, 29, 29 – 41. Axelrod, D. (2001a) Total internal reflection microscopy in cell biology. Traffic, 2, 764 – 774. Axelrod, D. (2001b) Selective imaging of surface fluorescence with very high aperture microscope objectives. J. Biomed Opt. 6, 6 – 13. Cormack, B.P., Valdivia, R.H. & Falkow, S. (1996) FACS-optimized mutants of the green fluorescent protein (GFP). Gene, 173, 33–38. Danuser, G. & Waterman-Storer, C.M. (2003) Fluorescent speckle microscopy: where it came from and where it is going. J. Microsc. 211, 191 – 207. Geiger, B., Bershadsky, A., Pankov, R. & Yamada, K.M. (2001) Transmembrane extracellular matrix-cytoskeleton crosstalk. Nature Rev. 2, 793 – 805. Gupton, S.L., Salmon, W.C. & Waterman-Storer, C.M. (2002) Converging populations of f-actin promote breakage of associated microtubules to spatially regulate microtubule turnover in migrating cells. Curr. Biol. 12, 1891 – 1899. Mashanov, G.I., Tacon, D., Knight, A.E., Peckham, M. & Molloy, J.E. (2003) Visualizing single molecules inside living cells using total internal reflection fluorescence microscopy. Methods, 29, 142 – 152. Ponti, A., Vallotton, P., Salmon, W.C., Waterman-Storer, C.M. & Danuser, G. (2003) Computational analysis of f-actin turnover in cortical actin meshworks using fluorescent speckle microscopy. Biophys. J. 84, 3336 – 3352. Salmon, W.C., Adams, M.C., & Waterman-Storer, C.M. (2002) Dual-wavelength fluorescent speckle microscopy reveals coupling of microtubule and actin movements in migrating cells. J. Cell Biol. 158, 31 – 37.

152


Thomann, D., Rines, D.R., Sorger, P.K. & Danuser, G. (2002) Automatic fluorescent tag detection in 3D with super-resolution: application to the analysis of chromosome movement. J. Microsc. 208, 49 – 64. Vallotton, P., Ponti, A., Waterman-Storer, C.M., Salmon, E.D. & Danuser, G. (2003) Recovery, visualization, and analysis of actin and tubulin polymer flow in live cells: a fluorescent speckle microscopy study. Biophys. J. 85, 1289–1306. Watanabe, N. & Mitchison, T.J. (2002) Single-molecule speckle analysis of actin filament turnover in lamellipodia. Science, 295, 1083 – 1086. Waterman-Storer, C.M. (2002) Fluorescent speckle microscopy (FSM) of microtubules and actin in living cells. Current Protocols in Cell Biology (ed. by J. S. Bonifacino, M. Dasso, J. B. Harford, J. Lippincott-Schwartz & K. M. Yamada), unit 4.10. John Wiley, New York.

Waterman-Storer, C.M., Desai, A., Bulinski, J.C. & Salmon, E.D. (1998) Fluorescent speckle microscopy: visualizing the movement, assembly, and turnover of macromolecular assemblies in living cells. Curr. Biol. 8, 1227 – 1230. Waterman-Storer, C.M. & Salmon, E.D. (1998) How microtubules get fluorescent speckles. Biophys. J. 75, 2059 – 2069. Waterman-Storer, C.M. & Salmon, E.D. (1999) Fluorescent speckle microscopy of microtubules: how low can you go? FASEB J. 13, S225– S230. Webb, D.J., Brown, C.M. & Horwitz, A.F. (2003) Illuminating adhesion complexes in migrating cells: moving toward a bright future. Curr. Opin. Cell Biol. 15, 614 – 620. Zamir, E., Katz, B.Z., Aota, S., Yamada, K.M., Geiger, B. & Kam, Z. (1999) Molecular diversity of cell-matrix adhesions. J. Cell Sci. 112, 1655–1669.

© 2004 The Royal Microscopical Society, Journal of Microscopy , 216, 138–152