We present a protocol for the characterization of motility and behavior of a population of hundred micron- to millimeter-sized cells using brightfield microscopy and cell tracking. This assay reveals that Stentor coeruleus transitions through four behaviorally distinct phases when regenerating a lost oral apparatus.
Stentor coeruleus is a well-known model organism for the study of unicellular regeneration. Transcriptomic analysis of individual cells revealed hundreds of genes—many not associated with the oral apparatus (OA)—that are differentially regulated in phases throughout the regeneration process. It was hypothesized that this systemic reorganization and mobilization of cellular resources towards growth of a new OA will lead to observable changes in movement and behavior corresponding in time to the phases of differential gene expression. However, the morphological complexity of S. coeruleus necessitated the development of an assay to capture the statistics and timescale. A custom script was used to track cells in short videos, and statistics were compiled over a large population (N ~100). Upon loss of the OA, S. coeruleus initially loses the ability for directed motion; then starting at ~4 h, it exhibits a significant drop in speed until ~8 h. This assay provides a useful tool for the screening of motility phenotypes and can be adapted for the investigation of other organisms.
Stentor coeruleus (Stentor) is a well-known model organism that has been used to study unicellular regeneration owing to its large size, ability to withstand several microsurgical techniques, and ease of culturing in a laboratory setting1,2,3. Early regeneration studies focused on the largest and most morphologically distinct feature in Stentor—the OA—which is shed completely upon chemical shock4,5,6. De novo replacement of a lost OA begins with the emergence of a new membranellar band—an array of cilia that gradually shift towards the anterior of the cell before forming a functional OA over eight morphological stages3. These stages have been observed sequentially, regardless of temperature, and provide a universal reference point for nearly all studies5.
Mechanistic analysis of Stentor regeneration requires tools for measuring the timing of regeneration that are robust and simple enough to be applied to multiple samples as part of a chemical or molecular screen. The standard method for performing a cell-based assay is imaging, in this case, imaging the formation of new OA during regeneration. However, such imaging-based assays are most effective when the regenerating structure contains distinct molecular components that can be used as markers, so that they would be easily detected in a fluorescence image. In the case of the Stentor OA, the known components (cilia, basal bodies) are also present on the rest of the cell surface; therefore, recognizing the restoration of the OA cannot be achieved simply by looking for the presence or absence of a component.
Rather, some form of shape recognition would be required to detect an OA, and this is potentially very challenging given the fact that Stentor cells often change shape via a rapid contractile process. This paper presents an alternative assay for regeneration that relies on the motile activity of the body and OA cilia. As the OA regenerates, the newly formed cilia undergo reproducible changes in position and activity, which in turn, affects the swimming motility of the cell. By analyzing motility, it is possible to perform an assay for "functional regeneration" that quantifies regeneration by quantifying the function of the regenerated structures. Previous analysis of Stentor ciliary function during regeneration used particle image velocimetry, combined with tracer beads added to the external media, to observe changes in flow pattern at different stages of regeneration7; however, this approach requires laborious imaging of individual cells and their associated flow fields, one at a time.
By using the motion of the cell itself as a proxy for cilia-generated flow, it would be possible to analyze larger numbers of cells in parallel, using low-resolution imaging systems compatible with high-throughput screening platforms. This assay can, in principle, be used to study development and functional regeneration in other swimming organisms in the hundreds of microns to millimeters size scale. Section 1 of the protocol describes the construction of a multiwell sample slide, which allows for high-throughput imaging of a population of cells over up to an entire day. Details are provided for how to adjust for use with other cell types. Section 2 of the protocol covers the acquisition of video data for this assay, which can be accomplished on a dissection microscope with a digital single-lens reflex camera. Section 3 of the protocol provides a walk-through of cell tracking and cell speed calculation using MATLAB code (Supplemental Information). Section 4 of the protocol explains how to turn the numerical results into plots as shown in Figure 1C-F and Figure 2C for easy interpretation of results.
Subscription Required. Please recommend JoVE to your librarian.
NOTE: A population of approximately one hundred S. coeruleus cells were cultured in accordance with a previously published JoVE protocol8.
1. Sample preparation
- Cut a piece of 250 µm-thick silicone spacer sheet (Table of Materials) slightly smaller in both height and width than a microscope slide. Using a 5/16" hole punch, create circular wells. Be mindful of leaving sufficient space between neighboring wells to ensure a good seal.
NOTE: A space of 3 mm between neighboring wells was found to be sufficient. With practice, ten wells can be placed on a single sample slide.
- Initiate regeneration of OA by incubating the cells in 10% sucrose or 2% urea for 2 min (Figure 1A). Then, wash three times with fresh medium8. Gently pipette approximately ten Stentor into each well. Be careful of matching sample volume to the well volume as closely as possible.
NOTE: For the well dimensions used here, 12.5 µL of sample was pipetted into each well.
- Close the wells by gently lowering a piece of coverglass (see Table of Materials) over the wells starting from one edge. Use a narrow and slightly flexible object, such as a 10 µL pipette tip, to press down on the coverglass where it contacts the silicone sheet, to ensure a good seal.
2. Visible light microscopy time-lapse
- Place the sample on the microscope stage, and set magnification to the lowest available such that one well fits in frame in its entirety.
NOTE: A 1.6x camera adaptor and 1x magnification on the objective were used on the microscope for this protocol.
- Begin time-lapse. Acquire a 10 s video at 30 frames per second of each well at each timepoint. If using a microscope setup with a motorized X-Y stage, automate the entire time-lapse. Otherwise, ensure that a user is present at each timepoint to manually translate the sample to record each well.
NOTE: Avoid leaving the sample illuminated when not imaging to avoid heating and evaporation. The sample volumes are small, and evaporation will lead to air bubbles.
- Save movies as TIFF, MOV, or AVI.
NOTE: These are the most common non-proprietary video file types. Depending on the specific microscope software, the videos may save by default to a proprietary file type, but then can be exported to one of the aforementioned file types.
- Use a pixel to millimeter conversion factor for physical scale, and perform a calibration or use known pixel size from the camera used. To calibrate, acquire a clear image of a calibration slide or a clear ruler at the same magnification settings as the videos. Using any image viewing program, count the number of pixels between marks of a known physical distance.
NOTE: For example, if the image of a ruler shows the 1 mm mark and 2 mm mark to be separated by 100 pixels, the conversion factor is 1 mm per 100 pixels. Alternatively, to derive this factor from camera pixel size, simply multiply the camera pixel size by the magnification. For example, if the camera used has 3.45 µm pixels, and the magnification used was 1.6x, then the conversion is 3.45 µm * 1.6 = 5.52 µm per 1 pixel.
3. Cell tracking
- Download the two scripts, TrackCells.m and CleanTraces.m, to an easy-to-remember location on the computer. If the data videos are not already on this computer, transfer them onto this computer.
NOTE: The data videos and the scripts do not need to be in the same folder.
- Organize data videos into folders, one for each timepoint. Use the script TrackCells.m first to perform automated cell tracking in the data videos. Open TrackCells.m and run the script.
- Choose Add to Path if prompted by a pop-up window., which will typically only happen when the script is run for the first time from a given folder. When prompted, Select One or More Data Videos to Initiate Tracking, navigate to the data videos (section 2). Select multiple video files by using shift-click, control-click, or by holding down the left mouse button while moving over the files to highlight them.
- Once satisfied with the list of files in the File name: box at the bottom of the prompt window, click on Open. Perform a test run on one video first to confirm all parameters are set correctly (see discussion below).
NOTE: This script will now create a folder for each video file chosen. It will then write into this folder each frame of the video as a .jpg file and a file named position_estimates.mat, which contains all traces found in the video. Depending on the size of the videos, the number of videos, and the speed of the computer, this script can take hours to run.
4. Trace verification
- Verify that steps described in section 3 were followed correctly by checking that there are no error messages before proceeding. Use the CleanTraces.m script to manually reject anomalous or false traces. Open CleanTraces.m in the MATLAB editor window by double-clicking on the file.
- At the prompt Select data folder output by TrackCells.m. It will have the same name as the video file, navigate to one of the folders created as described in section 3. Choose only one folder.
NOTE: This script will now display the traces one-by-one in a pop-up window. They are overlaid on the frame of the video where the trace starts, in green, and the frame of the video where the trace ends, in magenta. Therefore, a valid trace should link a green cell and a magenta cell.
- When prompted, enter 1 to keep the trace and 0 to reject the trace. Press Enter to move on to the next trace.
NOTE: New traces will continue to display until either there are no more traces, or the first sixty have been shown. When this process is complete, the script will automatically create a folder named CLEAN TRACES for saving the outputs and display all remaining traces on top of the first frame of the video (Figure 1B). This image is automatically saved as LabeledTraces.png for future reference. All traces the user had chosen to keep will be saved in the file clean_traces.mat.
- Complete this step for all videos in one time point before continuing.
NOTE: For the data in this manuscript, one video was acquired for each well at each timepoint, for a total of ten videos per timepoint.
5. Data visualization
NOTE: To visually compare the motility of the entire cell population across different timepoints, all traces from section 4 were translated to begin at the origin and create one radial displacement versus time plot for each timepoint (Figure 1C-F, see Supplemental Figure S1 for all time points).
- Begin by downloading the script titled SpaghettiPlots.m. Open and run the script. When a window file browser window pops up with the prompt Select time folder containing well data (clean_traces.mat) to graph, navigate to the folder of one of the time points. Note that this folder should contain within it a folder for each of the analyzed videos.
- When prompted Calibration: What is the number of pixels per millimeter?, type in the calibration value found in step 2.4, and press Enter.
NOTE: The script will now combine the traces from all the analyzed videos at this time point into a single plot (Figure 1C-F). Faint dotted circles in the plot indicate radial displacements of 1, 2, 3, and 4 mm.
- Adjust axes of the plot as necessary by changing line 55 of the script, which by default, is set to rr = 4 for including a radius of up to 4 mm. Save the plot.
Subscription Required. Please recommend JoVE to your librarian.
The goal of this assay is to quantify the gradual change of movement patterns and phased increase in movement speed from cells within a large (N ~100) regenerating Stentor population. To aid interpretation of results, the custom code included in this protocol generates two types of plots: an overlay of all cell movement traces in a set of video data (Figure 1C-F and Figure S1), and a plot of swim speed by hour since the start of regeneration (Figure 2C). A population of Stentor, which are regenerating normally, should demonstrate a steady transition from directionless swimming to directed swimming (Figure 1G). While each cell undergoes this transition in full during the OA regeneration process spanning ~8 h, variation between individuals necessitates looking at the movement of a large population in aggregate for trends to become clear.
Figure 1C-F show the collective motility of the same population of Stentor cells at 2, 3, 7, and 8 h into the OA regeneration process. These plots show both the expected increase in number of motile individuals (number of visible traces) and the four-fold increase in range of the most motile cells within the population (radial displacement). It is important to note that axes are different between the different panels of Figure 1C-F, with 1 mm (Hour 2), 2 mm (Hours 3 and 7), and 4 mm (Hour 8) chosen as most appropriate to showcase the motion. At 2 h—the earliest time point—no cells move more than 1 mm during the 10 s video, whereas at 9 h, many cells traversed more than 5 mm during the same length of time. Supplemental Figure S1 provides these plots for the full time-lapse, with identical axes used across all plots to give a clearer view of the expected increase of motility in a successful experiment.
There is one caveat to the interpretation of this example series of plots. The fastest swimming cells were able to swim across the well within the video, especially if (by chance) they started close to the edge; thus they are restricted from swimming entirely in a straight line, and their movement range, as visualized in this way, will appear smaller. While this does complicate any attempts to quantify the distance traveled by the cells, it does not alter the qualitative trend that the range of movement increases over at least 8 h. This is consistent with the known morphological regeneration timeline (Figure 2A)3.
To compile population statistics, the tracked cells were classified into three distinct categories: non-motile with no visible holdfast, non-motile with visible holdfast, and motile. These categories of behavior were previously defined by Tartar, but never quantified for their prevalence over time3. Figure 2B shows a stacked histogram summarizing relative cell counts across these categories as a function of hours into regeneration. At all time points, over half of the cell population was not observably motile. Additionally, at later times, a significant number of the cells were found in colonies with visible holdfasts, strongly suggesting they had explored the environment and preferentially attached near others of their kind. An example of this can be seen in the final inset of Figure 2C.
This assay allows for statistical comparison of swim speeds during each phase of motility recovery. The mean and standard deviation for data presented in Figure 2C are summarized below (Table 1). Once these statistical quantities have been extracted by cell tracking, the motion exhibited by the population of cells in the different phases can be rigorously compared using the unpaired t-test. As a concrete example, for the data shown, the t-statistic comparing the swim speeds in Phase 1 and Phase 2 is calculated by:
where x1 and x2 denote the mean speed during Phases 1 and 2, respectively; s1 and s2 denote the standard deviations of the speed during Phases 1 and 2, respectively; n1 and n2 and denote the number of traces extracted during Phases 1 and 2, respectively. The resulting t-statistic t12 can then be translated into a p-value for hypothesis testing. For the data shown in Figure 2C, the transitions from Phase 1 to Phase 2 (p < 0.1%) and from Phase 2 to Phase 3 (p < 0.1%) are statistically significant, whereas the transition from Phase 3 to Phase 4 (p > 20%) is not. It is important to note that the tendency for the cells to settle into small colonies during Phase 4 (Figure 2B, inset), as is apparent in the raw videos, is one example of behavioral difference not well captured by measurement of swim speeds alone. This explains the large standard deviation measured during Phase 4 (0.26 mm/s), which in turn, contributed to a lower t-statistic when comparing it against Phase 3.
Figure 1: Cell tracking reveals gradual recovery of directed swimming. (A) Sucrose shock induces Stentor coeruleus to shed membranellar band. (B) Example tracking result of regenerating cells in a 10 s video with striped circles indicating air bubbles in the well. (C-F) Overlay of all trajectories at 2 h, 3 h, 7 h, and 8 h. Times were rounded to the closest hour during compilation of data. Dotted circles indicate radial displacements of one millimeter. (G) Cell motility evolves from directionless swimming characterized by short circular trajectories to directed swimming characterized by long sinusoidal trajectories. Please click here to view a larger version of this figure.
Figure 2: Motile subpopulation demonstrates four distinct phases of behavior through OA regeneration. (A) Morphology of Stentor regeneration. (B) Normalized counts of cells at each time that were (blue) not motile, with no visible holdfast; (orange) not motile, with visible holdfast; (green) motile. Multiple datasets that span different subsets of hours were combined for this figure, so total cell count per hour ranged from 57 to over 376. Legend shows representative cells under each category as visualized by false color, in magenta and green. (C) Boxplot of swim speed; boxes extend from Q1 to Q3 quartile values at each time with median value indicated in green. The scatter overlays are the average speed of each motile cell. Inset shows representative population behavior during each of the four phases. Abbreviations: OA = oral apparatus; Q1 = first quartile; Q3 = third quartile. Please click here to view a larger version of this figure.
Figure 3: Motility analysis as a phenotype screening tool requires large numbers of cells. (A, B) Tracking results from two wells at 12 h. (C, D) Tracking results from the same two wells at 18 h. Stentor cells display a preference to attach in clusters, with some wells settling into colonies hours sooner than others. Striped areas indicate air bubbles in the well. Please click here to view a larger version of this figure.
|Mean (mm/s)||Standard Deviation (mm/s)|
Table 1: Comparison of swim speeds during each phase of motility recovery.
Supplemental Figure S1: Population swimming patterns through regeneration and beyond. (A-M) Motion of Stentor coeruleus cells 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14, 16, and 18 h, respectively, after the initiation of oral apparatus regeneration. Please click here to download this File.
Supplemental Videos 1-3: Example microscopy videos of S. coeruleus for script debugging. Section 5 of protocol can be performed on this set of three data videos for a quick confirmation that the scripts are running correctly. In addition, these videos provide a concrete example of contrast and resolution requirements for the data videos. Please click here to download this File.
Supplemental Files: Please click here to download this File.
Subscription Required. Please recommend JoVE to your librarian.
Many particle and cell tracking algorithms currently exist, some entirely free. Cost and user-friendliness are often trade-offs requiring compromise. Additionally, many of the existing cell-tracking programs are designed to track slow crawling motion of tissue culture cells, rather than the fast swimming motion of Stentor, which rotates while swimming and can undergo sudden changes of direction. After testing many of these options, the protocol presented here is intended to be a one-stop solution to go all the way from data acquisition to data visualization using only low-cost equipment and a single scientific software package (Table of Materials). This is of particular interest to research labs at teaching-focused institutions or biophysics teaching labs where most of the needed equipment is already available, as it lowers the barrier for students to ask their own questions and arrive at quantitative results within a short timeframe.
The method presented in this manuscript can be adapted to quantitatively characterize the motility of a population of organisms in the hundred microns to millimeter size scale. Application to smaller or larger organisms will require changes to how the video data must be acquired. As applied to a population of regenerating S. coeruleus here, this protocol yielded a timeline to the functional recovery of motility across the cell population, which complements what is currently known about the transcriptional timeline. For example, it is known from RNA sequencing that synthesis of genes encoding ciliary motility proteins does not take place until several hours into the regeneration process, long after the formation of the cilia themselves9, 10. The statistically significant drop in motility from hours 0-3 (Phase 1) to hour 4 (Phase 2) demonstrated above is consistent with this lag in expression of ciliary motility factors compared to the expression of genes involved in ciliary assembly. No transcriptomic study of Stentor regeneration to date had been carried out beyond hour nine, so little is known about the genomic activity. The lack of statistically significant difference between hours 8-10 (Phase 3) and hours 11+ (Phase 4) suggests that the cells have fully regenerated by this time, and few genes relevant to cell motility should be differentially regulated beyond hour 9.
Two major limitations to this method are its failure to track clustered and/or touching cells, and its inability to quantify more complex aspects of motion than swim speeds or trajectories. All particle/cell tracking algorithms have difficulty following cells that are temporarily visually obstructed by another cell, and when multiple cells spend multiple frames in close proximity; the script included here is not an exception. Fortunately, the frequency of the former can be effectively reduced simply by limiting the number of cells placed inside one well as discussed in the next paragraph. Specifically for the S. coeruleus population investigated here, their preference to settle in colonies (Figure 3) caused some of these cells not to be identified by the script. However, as all these cells exhibit little to no motion, statistics on the motile cell population is unaffected. Additionally, videos where multiple cells elude tracking are easy to identify and manually exclude from further analysis. Regarding the quantification of complex aspects of motion, the transition from directionless to directed swimming seen in Supplemental Figure S1 suggests quantitative changes in the time correlation of direction, average acceleration, and potentially other measurables. Speed and range, the only quantities analyzed in this protocol, represent only a small aspect of motion.
Lastly, a few parts of the protocol are of particular importance or consequence in practical use. Assembly of the multiwell sample slides with thin silicone sheets, as described in protocol section 1, requires practice. It is easy to introduce leaks or air bubbles into at least one well while sealing the sample with the coverglass. The wells can be divided over multiple sample slides to allow the use of smaller coverglass, which will make the process easier (and mistakes less costly). Though the assay presented here can be used to characterize the motility patterns of a cell population at a single timepoint, it is demonstrated here as a tool to compare motility patterns of a single population over a long time-lapse (over 10 h). As with any long time-lapse, evaporation of a wet sample is unavoidable. Evaporation can be minimized by a secure seal of the silicone spacer in the imaging chamber to both the glass slide and the coverslip. Failure to do this will result in wells leaking into each other or air bubbles forming inside the wells. For researchers unfamiliar with the use of silicone spacers, pasteurized spring water or other clean medium should be pipetted into each well to test for leakage before use. The dimensions of the sample wells were all chosen to work well for S. coeruleus, which are approximately 1 mm in size. This includes silicone spacer thickness (chosen to constrain the S. coeruleus cells under investigation to a two-dimensional plane), 5/16" diameter, and 10 cells per well. These numbers should be adjusted when adapting this protocol for other cell types.
Several parameters in the TrackCells.m script can be tweaked to optimize automatic cell identification and trace validation. The parameters MinCellArea and MaxCellArea (Lines 15 and 16 of script) allow the user to set the acceptable range of sizes for a cell in square pixels. What values are best depends on many details of the experiment including cell size, camera pixel size, magnification, and whether there are objects which could be misidentified as cells, for example, air bubbles. The default values of 300 and 1500 were optimal for the exact equipment and settings provided in the protocol. After MinCellArea and MaxCellArea have been optimized, tune the parameter Threshold (Line 17) to change image contrast. When incorrectly set, the cell outlines will be poorly identified or missed altogether, thereby hindering correct tracking. If no value of Threshold works well for correct tracking, try with a video with higher contrast or that is more in focus. Line 218 of the script, bad_trks = find(strk_trks > 20); is responsible for discarding traces that deviate from the predicted motion (via Kalman filter) too many times. As is, the script has this threshold for too many set at 20. Decrease this integer value to as far as 1 to discard traces more aggressively. Increase the value to discard less aggressively. Much of TrackCells.m was adopted from the multi-object tracking tutorial freely accessible at http://studentdavestutorials.weebly.com/multi-bugobject-tracking.html, which the reader could refer to for more detail. Supplemental Videos 1-3 are example data videos for a test-run of the scripts.
The ability of S. coeruleus to fully regenerate a lost OA has been first observed over a century ago, yet many open questions remain regarding the recovery of motility and behavioral states. The protocol presented here is intended to demystify the combination of brightfield imaging, automated cell identification and tracking, and data visualization the authors employed to quantitatively characterize the motility of a population of regenerating Stentor. This workflow is broadly applicable to the investigation of motility in general, and the scripts and protocol included can be readily adopted to work with other biological systems of similar size and speed, for example, other ciliates such as Paramecium and Blepharisma. Furthermore, changing the imaging chamber (e.g., using dry wells or different size wells coupled with different imaging magnification) greatly expands the applicability of this workflow. The current tool kit for Stentor includes surgical dissection, osmotic shock, RNAi, DNA/RNA analysis, and small molecule inhibitors6. The high-throughput method of quantifying motility presented here adds a complementary degree of characterization and will be used in future work to investigate motility phenotypes.
Subscription Required. Please recommend JoVE to your librarian.
The authors have nothing to disclose.
This work was supported, in part, by Marine Biological Laboratory Whitman Early Career Fellowship (JYS). We acknowledge Evan Burns, Mit Patel, Melanie Melo, and Skylar Widman for helping with some of the preliminary analysis and code testing. We thank Mark Slabodnick for discussion and suggestions. WFM acknowledges support from NIH grant R35 GM130327.
|0.25 mm-thick silicone sheet||Grace Bio-Labs||CWS-S-0.25|
|24 x 50 mm, #1.5 coverglass||Fisher Scientific||NC1034527||As noted in Discussion, smaller coverglass can be used if fewer sample wells are placed on one slide.|
|CCD camera||We used Nikon D750|
|Chlamydomonas 137c WT strain||Chlamydomonas Resource Center||CC-125|
|MATLAB Image Processing Toolbox||MATHWORKS||needed for TrackCells.m and CleanTraces.m|
- Lillie, F. R. On the smallest parts of stentor capable of regeneration; a contribution on the limits of divisibility of living matter. Journal of Morphology. 12 (1), 239-249 (1896).
- Morgan, T. H. Regeneration of proportionate structures in Stentor. The Biological Bulletin. 2 (6), 311-328 (1901).
- Tartar, V. The Biology of Stentor. Kerkut, G. A. , Pergamon Press. (1961).
- Tartar, V. Reactions of Stentor coeruleus to certain substances added to the medium. Experimental Cell Research. 13 (2), 317-332 (1957).
- Kelleher, J. K. A kinetic model for microtubule polymerization during oral regeneration in Stentor coeruleus. Biosystems. 9 (4), 269-279 (1977).
- Slabodnick, M. M., et al. The kinase regulator Mob1 acts as a patterning protein for Stentor morphogenesis. PLOS Biology. 12 (5), 1001861 (2014).
- Wan, K. Y., et al. Reorganization of complex ciliary flows around regenerating Stentor coeruleus. Philosophical Transactions of the Royal Society B: Biological Sciences. 375 (1792), 20190167 (2020).
- Lin, A., Makushok, T., Diaz, U., Marshall, W. F. Methods for the study of regeneration in Stentor. Journal of Visualized Experiments JoVE. (136), e57759 (2018).
- Sood, P., McGillivary, R., Marshall, W. F. The transcriptional program of regeneration in the giant single cell, Stentor coeruleus. bioRxiv. , 240788 (2017).
- Onsbring, H., Jamy, M., Ettema, T. J. G. RNA sequencing of Stentor cell fragments reveals transcriptional changes during cellular regeneration. Current Biology. 28 (8), 1281-1288 (2018).