Interactive, computer-assisted tracking of speckle trajectories in fluorescence microscopy: application to actin polymerization and membrane fusion

Abstract

Analysis of particle trajectories in images obtained by fluorescence microscopy reveals biophysical properties such as diffusion coefficient or rates of association and dissociation. Particle tracking and lifetime measurement is often limited by noise, large mobilities, image inhomogeneities, and path crossings. We present Speckle TrackerJ, a tool that addresses some of these challenges using computer-assisted techniques for finding positions and tracking particles in different situations. A dynamic user interface assists in the creation, editing, and refining of particle tracks. The following are results from application of this program: 1), Tracking single molecule diffusion in simulated images. The shape of the diffusing marker on the image changes from speckle to cloud, depending on the relationship of the diffusion coefficient to the camera exposure time. We use these images to illustrate the range of diffusion coefficients that can be measured. 2), We used the program to measure the diffusion coefficient of capping proteins in the lamellipodium. We found values ∼0.5 μm(2)/s, suggesting capping protein association with protein complexes or the membrane. 3), We demonstrate efficient measuring of appearance and disappearance of EGFP-actin speckles within the lamellipodium of motile cells that indicate actin monomer incorporation into the actin filament network. 4), We marked appearance and disappearance events of fluorescently labeled vesicles to supported lipid bilayers and tracked single lipids from the fused vesicle on the bilayer. This is the first time, to our knowledge, that vesicle fusion has been detected with single molecule sensitivity and the program allowed us to perform a quantitative analysis. 5), By discriminating between undocking and fusion events, dwell times for vesicle fusion after vesicle docking to membranes can be measured.

Figure 1

Standard deviation, ε, of the difference between particle position and speckle mark after refining with Gaussian-Fit, versus 1/S/N and ratio of point-spread function width to pixel size, σ/λ. Simulated particles were tracked as described in section S5 of the Supporting Material. The graphs shows data from Fig. 4 of Sbalzarini and Koumoutsakos (26) (σ/λ = 1) and Fig. 6 of Cheezum et al. (33) (σ/λ ≈ 1), who compared Gaussian-Fit and centroid algorithms. We did not include the lowest S/N data in Cheezum et al. (33) because some of these data points fall outside of the graph.

Figure 2

Tracking simulated diffusing particles. (A) Simulated images with increasing diffusion coefficients (left to right). (Bottom row) Same images with increased noise. (B) Marked trajectories. (C) MSD for individual tracks for D = 1 μm²/s (low noise). (D) Averaged MSD plots for different diffusion coefficients in Table 1. Error bars are mean ± 1 SD. (Inset) Enlarged version.

Figure 3

Tracking diffusing CPs at the leading edge of XTC cells. (A) Maximum intensity projection from a time-lapse recording of GFP-labeled CP at the leading edge. (Dashed line) Outline of leading edge. Exposure time was 66 ms and 1 pixel = 80 nm. Diffuse structures are diffusing molecules. (Bright speckles) CP proteins bound to the actin meshwork. (B) Enlarged section of box of panel A, single frame. (Line) Trace of a speckle track. (Middle arrow) Start of track. (Top arrow) Another diffusing speckle. (Bottom arrow) Cloud too mobile to track for enough frames. (C) MSD plots for individual speckle tracks from the time-lapse recording. (D) Distribution of diffusion coefficients found by fitting individual MSD curves with straight lines. Experimental: 22 tracked CPs. Simulated: results of tracking simulated particles for 10 frames with comparable conditions to the experiment: D = 0.6 μm²/s, 66 ms exposure, 1 px = 80 nm. Bin sizes are 0.14 μm²/s. Scale bars, 2 μm.

Figure 4

Speckle lifetime measurements. (A) XTC cell expressing EGFP-actin at high concentrations in which actin filaments in the lamellipodia appear as a continuous field. Scale: 8 μm. (B) Leading edge of lamellipodium with very dilute concentration of EGFP-actin. Single EGFP-actin monomers appear as speckles. (Bottom) Tracked speckles. Scale: 2.65 μm. (C) Intensity profile of speckle marked (arrow) in panel B. (D) Histogram of speckle lifetimes (n = 709). (Squares) Raw data. (Columns) Data normalized for photobleaching. Normalization and half-life estimation as in Watanabe and Mitchison (11).

Figure 5

Single lipid tracking following vesicle fusion on a supported bilayer. (A) Montage of TIRFM images. (Top left) Before docking; (top right) docking; (bottom left) shortly after fusion; (bottom right) more time after fusion. Released lipids diffuse on the membrane. Residual lipids from prior fusion events can be seen in the first frame. (B) Image of tracked lipids. Images taken at 67 frames/s. (C) MSD for individual lipid trajectories. (D) Averaged MSD plots and linear fit from 33 lipids tracked for at least 30 frames. Error bars are 1 SD of the mean. Scale: 2.67 μm (10 pixels).

Figure 6

Detection and analysis of fusion events. (A) Successive frames of image sequence of a vesicle that docked (frame 1246) and fused (frame 1289). (B) Average intensities within an inner circle of 2.5 px radius centered at the position of the vesicle (top curve) and a surrounding ring 2.5 px wide (bottom curve). (C) The y-t projection of an image sequence. Docked vesicle appears as a thin band (left arrow). Fusion results in formation of comet-tail appearance (right arrow). (D) Probability that a vesicle survived without fusion beyond a given delay after docking (178 fusion events from 10 different acquisitions).