High-throughput characterization of cortical microtubule arrays response to anisotropic tensile stress

Background Plants can perceive and respond to mechanical signals. For instance, cortical microtubule (CMT) arrays usually reorganize following the predicted maximal tensile stress orientation at the cell and tissue level. While research in the last few years has started to uncover some of the mechanisms mediating these responses, much remains to be discovered, including in most cases the actual nature of the mechanosensors. Such discovery is hampered by the absence of adequate quantification tools that allow the accurate and sensitive detection of phenotypes, along with high throughput and automated handling of large datasets that can be generated with recent imaging devices. Results Here we describe an image processing workflow specifically designed to quantify CMT arrays response to tensile stress in time-lapse datasets following an ablation in the epidermis — a simple and robust method to change mechanical stress pattern. Our Fiji-based workflow puts together several plugins and algorithms under the form of user-friendly macros that automate the analysis process and remove user bias in the quantification. One of the key aspects is also the implementation of a simple geometry-based proxy to estimate stress patterns around the ablation site and compare it with the actual CMT arrays orientation. Testing our workflow on well-established reporter lines and mutants revealed subtle differences in the response over time, as well as the possibility to uncouple the anisotropic and orientational response. Conclusion This new workflow opens the way to dissect with unprecedented detail the mechanisms controlling microtubule arrays re-organization, and potentially uncover the still largely elusive plant mechanosensors. Supplementary Information The online version contains supplementary material available at 10.1186/s12915-023-01654-7.


Data:
Note regarding image acquisition: any fluorescent microtubule reporter line can be used to acquire Zstacks with a confocal microscope at different time points after applying a mechanical stress. In many cases, the CMT signal is sufficient to also be used as a cell contour signal, but in certain cases it could be necessary to also acquire a second channel specifically for cell contour signal. For a better-quality analysis, it is recommended to use z-intervals of at least 1µm.

Installation
For more details see the installation page on the github repository.
-Fiji: see https://fiji.sc/ -Plugins: To install the required plugins, turn on the corresponding update sites. See https://imagej.net/update-sites/following for explanations if needed.
Click on around the top right corner of the page, and then "download zip". Then, unzip the file in the folder of your choice. Finally, copy the Angle2ablation_Workflow_ToolSet.ijm file and past it in the macros/toolsets folder of your Fiji install folder (on a Mac, access this by right clicking on the Fiji app in a Finder window and selecting "Show Package Contents").
To check if the toolset was loaded properly, open Fiji, and click on at the right end of the Fiji window. You should see Angle2ablation_Workflow_ToolSet in the drop-down menu. Select it and the toolset should appear in your Fiji toolbar.
-A2A_Tmlps_Stats.ipynb: No installation is required as the notebook can be run online simply by clicking on the button on the github page. For local installation of the notebook see the more detailed installation procedure at https://github.com/VergerLab/MT_Angle2Ablation_Workflow.

Workflow:
First, after image acquisition, we advise keeping a separate folder with all the raw untouched data and create a separate folder to convert and save all individual acquired stacks in .tif format (look online, there are many macros that exist to do this task in batch for all your images, and different format of raw images). Each 3D stack should be saved separately, not as a timelapse hyperstack. 1) SurfCut: Cell contour and CMT signal extraction a. Following the information provided in the SurfCut user guide (https://github.com/VergerLab/SurfCut2), extract the cell contours and the outer epidermal CMT signal (This case is specifically covered in the user guide).
b. Make sure to specify suffix for the cell contour images as "_cells" and "_MTs" for the CMTs images when saving them. These suffixes will be recognized by the macros used later on. Note that there is no need to save the Original Projection or the SurfCut Stack, but only the SurfCut Projection.
2) Folder maker: Generate the folder architecture according to your samples and conditions (optional) For this workflow, following SurfCut processing, we advise using a specific folder architecture (Experiment/Genotypes/Samples), that then allows the last steps "A2A" to be run over the whole experiment to generate a single text file containing all the data for further statistical analysis. a. Click on the "FoldrMaker" tool. In the pop-up window, specify the chosen experiment name, location where the folder should be created, name the different genotypes/conditions and the number of samples per genotype/conditions. For genotypes/conditions names, make sure to follow the nomenclature specified in the examples (Names separated by a coma and a space, and of course remove the parenthesis and "e.g."). Then click Ok.
b. After the folders are created, copy the images that have been extracted with SurfCut in step 1, in their corresponding folders.

3) Cell Pre proc: Preprocessing for cell contours improvement (optional)
Before running this tool make sure that "Linear Stack Alignment with SIFT MultiChannel" and "MorpholibJ" plugins are installed.
"Cell Pre proc" is a tool to prepare the segmentation of the cells around the ablation. This tool aligns all images from one time lapse and runs a z-projection using the average intensity. This image will be used to optimize the cell segmentation of the analyzed sample.
a. Click on "Cell Pre proc" in the tool menu. First, we advise running the tool on one sample (time series), to test the parameters, before applying them to all.
b. You will then be prompted to choose a directory. In this case, choose a "sample" folder: a folder containing all the images of a time series for a single sample).
c. In the next step, choose how many images to use for the alignment (starting from the first time point). Note that in some cases it is better to align only the first few images of the time lapse depending on the sample. Over time, the morphology of the sample can change bending/twist of the hypocotyl; this can create artefacts during registration and average projection and render lower quality cell contours. d. To test the quality of preprocessing for further image segmentation, you can then test the morphological segmentation from MorphoLibJ directly from within this macro. This segmentation is simply a test that will not be saved.
e. Finally, you can run the tool on all samples at once. You can then check the result for each time series and go back to processing/make corrections for some of them one by one with the "single" mode if necessary. This will simply overwrite the previous preprocessed image.

4) ROI Maker: Define regions of interest and geometry-based tensile stress pattern prediction.
This macro contains several steps, from cell segmentation, ablation definition, ROI creation, ROI refining, and geometry-based tensile stress prediction. These are put together because some of the same segmentation data are used both for the creation of refined ROIs and the tensile stress pattern prediction.
Note that the geometry based tensile stress proxy is generated form the segmented shape of the ablation which can sometimes be locally concave. In some cases, it can be useful to transform this shape with a convex hull. This is possible by simply changing one parameter in the macro code at line 1044. Open the "Angle2ablation_Workflow_ToolSet.ijm" file from the macros/toolsets folder of your Fiji install by dragging and dropping in Fiji. In the script editor, replace the value of "AblConvexHull" form "false" to "true". Save the file and reload the ToolSet (>>). a. Click on the "ROI Maker" tool. In this case, each samples folder is processed one at a time. In the popup "choose a directory" window, choose a "sample" folder: a folder containing all the images of a time series for a single sample). c. Next, the image obtained with the "Cell preprocessing" tool (or the first cell contour image of the time series) will pop-up with a dialog box to choose a segmentation method. There is a choice between "Morphological" (fully automated) and "Marker watershed" (markers need to be placed manually in each cell) segmentation.
The most straight forward approach is to use the "morphological segmentation" if cell contours are clear (steps d-e. If the result is not satisfying, then switch to marker assisted segmentation (d'-e'). d. The "Morphological" segmentation method opens the segmentation window. The segmentation runs automatically with default parameters. The Watershed segmentation has a tolerance of 10 but this can be modified until the segmentation fits. In the example below the result is not satisfactory. The options are thus either to change the tolerance parameters and re-run the segmentation until the segmentation looks ok, or switch to the marker-based segmentation (see next step).
Note that here on the image below there is clearly some over-segmentation. In some cases, this is acceptable, and can be corrected in a following step by merging some labels.
Below is an example sample for which the automated segmentation worked properly. e. In the dialog box that pops up next, select "Yes" and "OK" to continue the analysis. If not, select "No" and "OK" to return to the segmentation method choice. d'. If you chose "Marker watershed" segmentation, a segmentation window opens. Select manually by using the "Multi-point" tool, the ablation and the cells around it. Note that it is better to select at least two layers of cells around the ablation. e'. Click on "Run". In the result below, there is some under segmentation. One possibility is also to change the position of some markers to see if it improves segmentation or add additional markers and fuse some labels later on.
If none of this fully works for all the cells, there is a possibility in a later step to draw/correct manually some of the ROIs that will be created from these segmentations (see step h.).
If satisfied, click on "OK". The same dialog box as before will appear. If the segmentation is good click on "Yes" to continue the analysis.
f. During the next step you can correct some of the segmentation errors by merging some labels. Follow the instructions in the pop-up window.
In this case, below is the expected result after correction. Not all segmentation errors have been corrected, only those for the ablation and the cell layer around the ablation.
g. In the next step, you are prompted to select the ablation site. Follow the instructions in the pop-up window.
h. Next, the _MTs image appears with the ROIs, including the ablation and a single layer of cells around the ablation. In this step you can further correct the ROIs by deleting and re-drawing some manually if needed. To delete a ROI, select it in the ROI manager and press "Delete". Then select the polygon tool, draw the new corrected ROI, and add it to the ROI manager (Click "add" or hit ctrl+T).
Note that the ROI of the ablation should be the first one on the list of ROI for the next steps of the processing. If you don't modify it will stay as is, but if you delete it and re-draw it, the new ablation ROI will be the last one on the list. Then you need to manually rename the ROI with "0000000001" to make sure that the ablation ROI is first on the list. After renaming the ROI, click on "More" and select "Sort" to refresh the ROI list.
i. Next, the macro will create a simulation of tensile stress pattern around the ablation across the neighboring cells. Here you set a value for a "diameter", which can vary depending on what you expect to measure. In our analysis we selected a value which is roughly the width of the cells in this tissue.
j. Below you see the outer edge of the "simulation" (larger ROI). This perimeter will be used to crop the ROIs of the cells surrounding the ablation. If satisfied click on "Satisfied" and "OK" or if you are not satisfied and then only click on "OK" and you will return to the previous stage where you can change the value until you are satisfied. Make sure to keep a constant value across your experiment and the samples that you want to compare afterwards! k. Next, the ROIs will be cropped according to the defined perimeter. Click on "OK" to continue the analysis.
l. Next, you will define the parameters for the tensile stress "simulation". Define the number of iteration (number of lines drawn) and the spacing between the lines. m. You can then check the simulation image to see if it is equally spread in each ROI. As always, you have the possibility to go back to the previous step if not satisfied.
n. In this step, check the cell ROI positions for each time point of the time lapse. Sometimes there is a slight shift despite the alignment of the images in the time lapse. So, make sure the ROI is perfectly positioned in between the cell contours. Each time you click "OK", the next image in the time series will open. Each time this will save the corresponding ROI files with corrected positions.
o. Finally, the last pop-up window leaves you the choice to stop processing samples for now or to continue directly with the same processing step for another sample.

5) FibrilTool: Microtubule arrays and predicted tensile stress patterns quantification.
This tool quantifies CMT arrays organization in each cell ROI previously defined, and outputs both angle orientation value (average orientation of CMTs (or tensile stress) and an anisotropy value. This version of the tool is adapted from FibrilTool_Batch (https://zenodo.org/record/2528872). a. Click on the "FibrilTool" tool. In this case, each sample folder is processed one at a time. In the pop-up "choose a directory" window, choose a "sample" folder: a folder containing all the images of the time series for a single sample).
b. Next, a FibrilTool window opens to allow you to choose some parameters. These are mostly to change the result display on the output image and do not affect the analysis (lines color, numbering...). Here when you start FibrilTool, run the analysis first by selecting "Real" in the "Target for fibrils". This will perform the quantification on the actual CMT images.
c. Once this is done repeat the same steps, but this time select "Simu" in the "Target for fibrils".
Below are example results images for both analysis with the visual FibrilTool output displayed as red lines. d. All data is automatically saved as txt files. Then repeat these steps for each sample. 6) A2A: Calculation of the "angle to ablation". This macro calculates the "acute absolute angle" (between 0 and 90 degrees) between the actual microtubule array main orientation and the orientation of the predicted stress as measured with FibrilTool. a. Run the "A2A" tool. In the choice for target, select "Simu".
Note that here we leave the possibility to use manually drawn lines for reference of the tensile stress pattern, instead of measurements made on the geometry-based simulation, as was done in some previous studies. See towards the end for some details on how to use the "Line_ROIMaker macro".
b. In the choice for folder, you can either choose "single" and this will run the calculation for a single sample (folder selected in the next step), or "all". To run on "all" folders in the experiment (see file architecture generated at the beginning), all the samples must have already been processed in the previous steps. In this case select the Experiment folder containing all the genotypes/samples/… for this processing step.
Note that the "all" option will also generate a single text file containing all the quantification over the experiment, to be used for data analysis with the python notebook in step 7.
c. Next, it generates an image output and two files for each image from the subfolder containing the measurements (FibrilTool Angle; Drawn Line Angle; Raw angle; Absolute angle; Acute absolute angle to ablation; FibrilTool Anisotropy) and a file containing all the measurements of all the images in the subfolder. The image output (see below) with the angles written (yellow) on the image combined with FibrilTool output for the CMTs (Magenta) and the simulated stress pattern (green). This image is useful to confirm visually that there is no apparent mistake in the angle quantification.