We present a software solution for semi-automated tracking of relative protein concentration along the length of dynamic cellular protrusions.
Filopodia are dynamic, finger-like cellular protrusions associated with migration and cell-cell communication. In order to better understand the complex signaling mechanisms underlying filopodial initiation, elongation and subsequent stabilization or retraction, it is crucial to determine the spatio-temporal protein activity in these dynamic structures. To analyze protein function in filopodia, we recently developed a semi-automated tracking algorithm that adapts to filopodial shape-changes, thus allowing parallel analysis of protrusion dynamics and relative protein concentration along the whole filopodial length. Here, we present a detailed step-by-step protocol for optimized cell handling, image acquisition and software analysis. We further provide instructions for the use of optional features during image analysis and data representation, as well as troubleshooting guidelines for all critical steps along the way. Finally, we also include a comparison of the described image analysis software with other programs available for filopodia quantification. Together, the presented protocol provides a framework for accurate analysis of protein dynamics in filopodial protrusions using image analysis software.
Spatio-temporal control of actin regulatory proteins is associated with filopodium dynamics 1,2. Tracking spatially resolved protein concentration along the whole filopodial length through time is thus crucial to advance our understanding of the mechanisms underlying initiation, elongation, stabilization or collapse of these dynamic structures 3,4. Unlike protein analysis in the cytosol, where many cell shape changes occur at a larger scale, filopodia are dynamic micro structures that constantly buckle 5 and bend, thus precluding analysis using a simple approach such as a line-scan.
Different software solutions for tracking filopodial shape are available 6,7,8,9. Likewise, software for ratiometric tracking of protein dynamics within the cell body has been developed 10,11. To combine automated tracking of filopodial shape and spatio-temporal protein analysis, we recently developed an image analysis software based on the convex-hull algorithm 12. This novel analysis method, which is operated via a graphical user interface (GUI), combines for the first time, relative protein concentration along the filopodial length and growth velocity, thus allowing the accurate measurement of spatio-temporal protein distribution independent of movement of these dynamic structures 12.
The idea behind the software (source code is freely available, see below) is that one of the vertices of the convex hull will coincide with the tip of the filopodium (Figure 1A). By looking in the subsequent frame for the nearest vertex of the convex-hull, the moving tip can be tracked throughout the whole movie. Once the tip is detected in each frame, its position is used to draw an axis by joining the tip with a reference point at the base of the filopodium (Figure 1B). Finally, using equidistant nodal points, whose positions are determined by the median pixel with maximum intensity along the line orthogonal to the axis, are used to determine a backbone that follows the filopodial shape. Taking advantage of this adaptive backbone, a kymograph is generated to trace filopodial growth and protein concentrations for up to three channels along the filopodial length (Figure 1C).
Figure 1: Working Principle of the Image Analysis Software. (A) The algorithm behind the software. In Step1 the user specifies the reference (base) and the vertex (the tip) of the filopodium. In Step 2-1 the backbone of the filopodium is obtained using the median pixel with maximum intensity value. In Step 3 the backbone is used for spatial protein intensity profile. In Step 2-2 the software automatically tracks the tip in the subsequent frame. The whole procedure iterates. (B) Snapshot of the algorithm with real filopodium introducing important elements such as the convex hull that is being used for tracking. (C) Overview of parameters that can be measured with the algorithm. This figure has been modified from reference 12. Please click here to view a larger version of this figure.
The image analysis software is operated in Matlab (referred to as programming software) via a graphical user interface. To maximize flexibility and robustness for the particular experimental setting, the user can adjust a series of tracking parameters (e.g. permitted bending angle and inter-frame movement) and also make some corrections to the movies (e.g. cropping, rotation, removal of unwanted objects) (Figure 2A and Table 1).
GUI | No. | Mandatory | Description | Name (in GUI) | ||||
#1 | 1a | Y | Loading stacked .tiff file representing cell body (with box checked in) or create superimposed cell body from channels | CellBody | ||||
#1 | 1b | Y | Loading stacked image file corresponding to protein 1 | Protein 1 | ||||
#1 | 1c | Y | Loading stacked image file corresponding to protein 2 | Protein 2 | ||||
#1 | 1d | Y | Loading stacked image file corresponding to protein 3 | Protein 3 | ||||
#1 | 1e | N | Resets everything to preloaded stacked image files | Reset | ||||
#1 | 2a | Y | Scroll bar to determine the initial frame for analysis in GUI window #2 | NA | ||||
#1 | 2b | Y | Scroll bar to determine the final frame for analysis in GUI window #2 | NA | ||||
#1 | 2c | Y | Scroll bar representing current frame | NA | ||||
#1 | 2d | N | The grey value of the pixels below which all pixels will be set to zero | NA | ||||
#1 | 2e | N | The grey value of the pixels above which all pixels will be set to maximum values | NA | ||||
#1 | 2f | N | Set the intensity values of the pixels specified by <2e> & <2f> | Set | ||||
#1 | 2g | N | Play the intensity-adjusted movie | Play | ||||
#1 | 2h | N | Crop Image | Crop | ||||
#1 | 2i | N | Rotate Image | Rotate | ||||
#1 | 2j | N | Delete regions in the whole stack | Delete regions | ||||
#1 | 3a | Y | Click to open the ‘Analysis Window’ (GUI window #2) | Tracking Window | ||||
#1 | 3b | Y | Enter the size of a pixel in microns | Enter Pixel size | ||||
#2 | 4 | Y | Click to generate the boundary/edge image of the superimposed cell body | Boundary | ||||
#2 | 5 | Y | Click to select the base and tip of the filopodia | Referernce | ||||
#2 | 6a | Y | Enter the number of segments or nodes | No of segments | ||||
#2 | 6b | Y | Enter the scan length (perpendicular to axis) | Scan Width | ||||
#2 | 6c | Y | Enter the length above which filopodia starts bending | Accurate Meas after | ||||
#2 | 6d | Y | Enter the radius of the tip detection circle (i.e. area where the vertex can be localized in the next frame) | Radius of tip detection | ||||
#2 | 6e | Y | Enter the maximum angle the filopodium can bend from the vertical axis | Angle Threshold | ||||
#2 | 6f | N | Add reference points for base and tip for that specific frame | Select reference | ||||
#2 | 6g | N | Enter the length above which filopodia starts bending for that specific frame | Accurate Meas after | ||||
#2 | 6h | N | Enter the radius of the detection circle for that specific frame | Radius of tip detection | ||||
#2 | 6i | N | After entering all the parameters for the specific frame click to store the values to memory and file for further reference | Add | ||||
#2 | 6j | N | Click to delete the set manually parameters for that frame | Delete | ||||
#2 | 6k | N | Click to delete all parameters stored manually using the ‘optional features panel’ for all frames | Reset | ||||
#2 | 6l | N | Check in before tracking to store all tracking results in memory for future reference | History trace | ||||
#2 | 7 | Y | Click to start tracking | Track&Analyze | ||||
#2 | 8a | N | Click to start tracking protein channel intensity | Analyze Protein Intensities | ||||
#2 | 8b | N | Check in to track protein channel intensity along the filopodial length | Whole filopodia | ||||
#2 | 8c | N | Check in for tracking the reference protein or protein A | ProteinA | ||||
#2 | 8d | N | Check in for tracking the protein B | ProteinB | ||||
#2 | 8e | N | Check in for tracking the protein C | ProteinC | ||||
#2 | 8f | N | Check in to track average protein intensity in the tip | Leading tip | ||||
#2 | 8g | N | Enter the length of the tip | Tip Length | ||||
#2 | 8h | N | Enter the minimum distance from base above which the tip starts forming | threshold | ||||
#2 | 8i | N | Click to save the leading tip analysis results to file | Push Button | ||||
#2 | 9a | N | Click to initiate ratiometric protein analysis | Compare | ||||
#2 | 9b | N | Check in to compare protein B with respect to A | log10(B/A) | ||||
#2 | 9c | N | Check in to compare protein C with respect to A | log10(C/A) | ||||
#2 | 9d | N | Check in to compare protein B with respect to A at the tip | log10(B/A) | ||||
#2 | 9e | N | Check in to compare protein C with respect to A at the tip | log10(C/A) | ||||
#2 | 10a | N | Choose other color-map (default: Jetplot) | Color Map | ||||
#2 | 10b | N | Edit the colormap | Edit colormap |
Table 1: Overview of All Functions Present in the GUI Windows #1 and #2.
Once this is accomplished, the program creates a convex hull and automatically tracks the tip throughout the movie. Parameters extracted from the movie, such as a ratiometric kymograph, growth velocity, and filopodial length are displayed and also stored in the work folder as images and as data files. Other parameters such as filopodial lifetime, growth rate and retraction rate can then be extracted and further analyzed from the stored data files (Figure 2B).
Figure 2: Graphical User Interface for using the Image Analysis Software. (A) GUI Window #1 is used for loading and processing images. The program can load up to 3 protein channels, whereby 2 channels are compared pair-wise. The window comes with mandatory (blue) and optional features (green) for pre-processing the images prior to tracking (B) GUI Window #2 is used for tracking the filopodium as well as spatio-temporal and ratio-metric protein analysis. Again, optional features are marked in green. This figure has been modified from reference 12. Please click here to view a larger version of this figure.
Here, we present a detailed protocol for sample preparation and software handling. We start with detailed instructions on culturing cells and acquiring movies optimized for image analysis. This section on data acquisition is followed by a detailed description for operating the image analysis software. Throughout the protocol, we introduce critical steps and optional features that should be considered when collecting and processing data. Finally, we analyze filopodia from different model systems with the image analysis software, before closing with a comparison of the described image analysis software with other programs available for filopodial quantification and a discussion on limitations and future direction.
1. Cell Culture
2. Image Acquisition
NOTE: The length of filopodia varies from 2-10 µm 13. Filopodia grow at an average velocity of 0.05-0.1 µm/s 13,14.
3. Image Pre-processing
NOTE: Use ImageJ or other available software to pre-process images 16,17.
4. Image Analysis – Step 1: Load Images
NOTE: The software described here was written in Matlab (referred to as programming software) and will run only with this program.
5. Image Analysis – Step 2: Generate Trace
6. Image Analysis – Step 3: Spatio-temporal Protein Analysis
7. Image Analysis – Step 4: Ratio-metric Protein Analysis
8. Image Analysis – Step 5: Filopodial Tip Analysis
Using COS cells transfected with a marker for filamentous actin (f-tractin18, red) and a cytosolic reference (green), we found actin-rich filopodial protrusions (Figure 3A, top panel). Time series showed that filopodia rapidly extend and retract (Figure 3A, middle panel). Using the image analysis software, we then traced individual filopodia. Comparison of filopodial length measured by hand vs. the image analysis software showed a Pearson cross-correlation value of 0.947 arguing that the software reliably tracked filopodial extension 12. In addition to growth and retraction rates of an individual filopodia, kymographs at the bottom of Figure 3A also depicts fluorescence intensities of actin (top), the cytosolic reference (middle), and the relative concentration of actin normalized to the cytosolic reference (bottom). Together, these experiments provide evidence that the program reliably measures growth dynamics as well as spatially resolved relative protein concentration throughout the full life-cycle of an individual filopodium in COS cells.
Next, we investigated filopodia dynamics in cultured mouse hippocampal neurons (Figure 3B, top panel). For this, neurons were transfected at day in vitro (DIV) 8 with f-tractin (red) and a cytosolic reference (green) (Figure 3B, middle panel). As in the previous experiment, kymographs depicting fluorescence intensities of actin and cytosolic reference showed that relative enrichment of actin preceded formation of exploratory dendritic filopodia (Figure 3B, bottom panel). These findings are in line with published work, describing formation of actin-rich patches prior to filopodial elongation 3,19,20.
Figure 3: Examples of Filopodia Analyzed by the Software. (A) Overview image (top panel) and time series (middle panel) of COS cells transfected with f-tractin (red) and cytosolic marker (green). Below, plots depicting analysis of filopodial growth dynamics and relative concentration of f-tractin along the filopodium (bottom panels) are shown. (B) Overview image of hippocampal neuron transfected with cytosolic marker (top) and time series of neuron transfected with f-tractin (red) and cytosolic marker (green) (middle panel), as well as plot depicting analysis of dendritic filopodial growth dynamics and relative concentration of f-tractin along the protrusion (bottom panels). (C) SEM image (left panel) and time series of HeLa cells transfected with f-tractin (right panel). Note fluctuations in fluorescence intensity due to filopodial buckling. Scale bars = 20 µm (A, top), 50 µm (B, top), 10 µm (C, top). This figure has been modified from reference 12. Please click here to view a larger version of this figure.
Finally, we aimed to challenge the software by monitoring filopodia in HeLa cells. Filopodia in HeLa cells are long (Figure 3C, left panel) and highly dynamic, often leaving the plane of acquisition. Intriguingly, when focusing on the basal membrane, we observed a sub-fraction of filopodia to undergo helical twisting (Figure 3C, top right panel), whereby parts of the filopodium transiently left the focal plane. Accordingly, kymographs depicting the absolute fluorescence intensity of f-tractin showed wave-like changes in fluorescence intensity through time (Figure 3C, bottom right panel). This behavior is reminiscent of filopodial buckling, a mechanism recently described to be used by filopodia to exert pulling forces 5.
In summary, these experiments provide evidence that the software reliably tracks protein concentration and growth dynamics of filopodial from various origins.
Here we present a detailed protocol for tracking filopodial growth dynamics and analysis of relative protein concentrations in these dynamic structures via the convex-hull algorithm. Using the software, up to 3 channels can be compared pair-wise in a single run, whereby the relative concentrations of two channels (i.e. proteins) is determined throughout the extension/retraction cycle and stored as image and data files in separate folders. In addition to the routine operations, the software also provides a number of parameters that can be modified to optimize analysis for the respective experiment. A detailed description of these modifications as well as troubleshooting can be found in the protocol section and in Table 1.
Image analysis is not limited to a particular microscope type, but rather by the image quality of the time series. Considering its relevance, image quality should therefore be optimized for each channel during acquisition. Our experiments show that the algorithm works best for filopodia with SNR >4 acquired at 1 Hz using an objective with magnification above 40X and no pixel binning. Of course, this applies only as long as the fluorophores are compatible with the filter settings (i.e. no bleed-through; see protocol section 2 for details). Considering that relative intensities are compared against each other, the analysis is not sensitive to small differences in absolute fluorescence signal intensities between individual cells. Therefore, to deduce meaningful data, a well-defined reference channel (e.g. cytosolic marker) is of great importance. However, even if image quality is good there are some biological limitations. For instance, it is not well suited to analyze structures that are branching, tilting more than 45 degree from the axis orthogonal to the cell surface, or where other objects are crossing the structure that is being investigated.
In addition to these experimental limitations, there are critical steps within the protocol that need to be considered when operating the software, as this may otherwise trigger error messages. First, channels corresponding to a particular protein must be of same dimension and represent exactly the same ROI. Second, the stacked files must be in gray scale '.tiff' file format. Third, the stacked image files must be in the work folder. And finally, the names for data files in the work folder should not be changed, unless this is corrected for in the code.
How does the script compare to other existing methods? To probe our image analysis software, we downloaded a number of recently published software solutions that analyze filopodial features 6,10,21,22,23,24,25, and tested all for a number of selected criteria. A detailed comparison is presented in Table 2.
Tsygankov et al. | Barry et al. | Nilufar et al. | Fanti et al. | Constantino et al. | Hendri- cusdottir et al. |
Tarnok et al. | Saha et al. | |
Paper DOI | 10.1083/jcb.201306067 | 10.1083/jcb.201501081 | 10.1186/ 1752-0509-7-66 |
10.1002/dneu.20866 | 10.1016/j.jneumeth. 2008.02.009 |
10.1016/j.jneumeth. 2014.08.016 |
10.1002/cyto.a.22569 | 10.1091/mbc. E16-06-0406 |
software (download link) | http://www.hahnlab. com/tools/ THE softwarepage. html |
http://jcb-dataviewer. rupress.org/jcb/ browse/9059/ |
http://www.perkinslab .ca/pubs/ NMLP20XX. html |
http://www.ifc. unam.mx/ ffm/download.html |
http://fournierlab. mcgill.ca/ styled-7/fTracker.html |
https://fdynamics. wordpress.com/ |
http://cnblab. elte.hu/dfma |
https://campus.uni-muenster.de/index.php?id=13794&L=1 |
program type | Matlab | ImageJ | Matlab | ImageJ | Matlab | Matlab | ImageJ | Matlab |
operation | automated | automated | automated | semi-automated | semi-automated | semi-automated | semi-automated | semi-automated |
single vs. batch analysis | both | both | single | single | single | single | single | single |
whole cell vs. subregion analysis | whole | whole | both | both | both | sub | sub | sub |
multiple vs. individual filopodia | multiple | multiple | multiple | multiple | multiple | ind | ind | ind |
filopodial ID and visual inspection possible | yes | yes | no | yes | yes | yes | yes | yes |
Image processing possible | no | no | no | yes | no | no | no | yes |
filopodial length | yes | yes | yes | yes | yes | yes | yes | yes |
filopodial growth dynamics | yes | yes | no | yes | yes | yes | yes | yes |
filopodial lifetime | yes | yes | no | yes | yes | yes | yes | yes |
ratiometric analysis in filopodia | no | no | no | no | no | no | no | yes |
output / storage format | images / data file | images / data file | images / data file | images / data file | images / data file | images / data file | images / data file | images / data file |
Table 2: Overview of Software-assisted Image Analysis Solutions for Filopodia Quantification.
From the eight tested analysis software solutions, five were operated via Matlab, while the remaining three used ImageJ. Two software solutions (Table 2, to the left) were fully automated and optimized for batch analysis of filopodia length, lifetime and dynamics in whole cells. Of the remaining six software solutions, five required additional manual input (i.e. semi-automated), while one was fully automated. However, only the semi-automated solutions provide detailed information on filopodia growth dynamics, life time and identity. Among these remaining five, three were optimized for subcellular analysis of individual filopodia, while two could process multiple filopodia in parallel. Importantly, from all eight tested software solutions, only one approach allowed to combine information on filopodial growth dynamics with ratiometric protein localization within these finger-like structures (Table 2, to the right).
Although our image analysis software is not suitable for automated batch or whole cell analysis, quantification of an isolated filopodia takes only minutes. Thus, we envision that quantification of protein dynamics in 40-50 filopodia is feasible within a single day with the current software. In contrast, high-throughput screens may require further amendments to the code to streamline input and storage of thousands of individual filopodia in an automated manner. Based on the parameters that are required, other existing software solutions may thus be better suited for such an approach. Keeping these limitations in mind, the software provides a reliable platform for investigating spatio-temporal protein concentration in protrusion like structures in cells. Considering that filopodia are not the only fingerlike cellular structures, it is plausible to envision that the presented image analysis software may also be suitable to study protein dynamics in other biological processes (e.g. neurites, primary cilia).
The authors have nothing to disclose.
The authors acknowledge funding from the DFG (EXC-1003 to MG).
DMEM | Life Technologies | 31966-021 | |
10% Fetal bovine serum | Biochrom AG | L11-044 | |
Lipofectamine 2000 | Life Technologies | 11668-027 | |
1% penicillin/streptomycin | Biochrom AG | 12212 | |
Neurobasal Medium | Life Technologies | 21103-049 | |
B27 | Life Technologies | 17504-044 | |
HEPES (1M stock solution) | Life Technologies | 15630 | |
Citrine-N1 | Addgene | 54593 | |
Labtech | Thermo | 155411 | |
Glutamax-I | Thermo | 35050-061 | |
Hela | Leibniz Institute DSMZ | ACC-57 | |
COS 7 | Leibniz Institute DSMZ | ACC-60 | |
3T3 cells | Leibniz Institute DSMZ | ACC-59 | |
Microscope | Nicon Eclipse | ||
Camera | Andor | DU888 Ultra | |
Confocal Unit | Yokagawa | CSU-X1 | |
Pyruvate | Gibco | 31966-021 |