High Throughput Traction Force Microscopy for Multicellular Islands on Combinatorial Microarrays.

The composition and mechanical properties of the cellular microenvironment along with the resulting distribution of cellular devolved forces can affect cellular function and behavior. Traction Force Microscopy (TFM) provides a method to measure the forces applied to a surface by adherent cells. Numerous TFM systems have been described in literature. Broadly, these involve culturing cells on a flexible substrate with embedded fluorescent markers which are imaged before and after relaxion of cell forces. From these images, a displacement field is calculated, and from the displacement field, a traction field. Here we describe a TFM system using polyacrylamide substrates and a microarray spotter to fabricate arrays of multicellular islands on various combinations of extra cellular matrix (ECM) proteins or other biomolecules. A microscope with an automated stage is used to image each of the cellular islands before and after lysing cells with a detergent. These images are analyzed in a semi-automated fashion using a series of MATLAB scripts which produce the displacement and traction fields, and summary data. By combining microarrays with a semi-automated implementation of TFM analysis, this protocol enables evaluation of the impact of substrate stiffness, matrix composition, and tissue geometry on cellular mechanical behavior in high throughput.

traction stress field is the calculated, using the known mechanical characteristics of the substrate.
In the MATLAB code provided with this protocol, the displacement field is calculated using a publicly available digital image correlation (DIC) algorithm (Landauer et al., 2018). Traction forces calculation is performed using an available Fourier transform traction cytometry (FFTC) algorithm (Sabass et al., 2008;Han et al., 2015). These methods were chosen based on their relatively low computational expense and time and relatively few required user inputs. This supports the high throughput nature of the analysis and goal to allow "off the shelf" use for labs who do not focus on TFM as a core competency. However, the code was intended to enable users to substitute alternative displacement or traction field algorithms based on the needs and computational resources of the user and application.
Performing TFM often involves bulk preparations of adhesive substrates which requires relatively large amounts of adhesive proteins, decreasing the practicality of performing TFM on multiple ECM proteins or surface bound ligands. Here we use contact printed microarrays which allows multiple ECM/ligand combinations, with replicates, to be included on a single substrate with low material usage (Flaim et al., 2005). TFM is often implemented either on single cells, or randomly distributed small colonies. In the microarray system, we can assess the mechanical behavior of multicellular islands where forces are transmit to the substrate as well as neighboring cells, resulting in collective behavior (Mertz et al., 2013). Further, these multi cellular islands are of consistent size and shape. This permits analysis of average mechanical behavior of these islands as a function of the environmental conditions. Island diameter can be tuned by using differently sized microarray pins, adding geometry as an additional parameter to investigate. The MATLAB code provided includes techniques to identify island boundaries and align data from replicate islands to enable analysis of these average behaviors. These cellular micro-arrays can be assayed in parallel using immunocytochemistry, which has been described in more detailed elsewhere (Kaylan et al., 2017), to allow correlation of mechanical and phenotypic behavior.
Our application has focused on the relationship between spatial patterns of mechanical behavior and differentiation of liver progenitor cells. This method can be applied to other stem cell systems especially where spatial patterning and collective mechanical behavior is of interest. It can also be applied to investigations of how the mechanical behavior of cancer relates to its microenvironment and further 3 www.bio-protocol.org/e3418 impacts proliferation or response to drugs. Although here we used contact microarray printing, this protocol can easily be applied to other micro patterned systems such as those with non-contact printing (Romanov et al., 2014) and Polydimethylsiloxane (PDMS) stamped substrates (Kane et al., 1999). In this application, we utilize the XY position within the array to know the condition of that island. This protocol could be extended to other systems where spatial position of cells on their substrate is linked to an experimental variable or substrates such as with gradients of stiffness (Hadden et al., 2017) or biomolecules (Dertinger et al., 2002).   j. Invert the dishes, as shown in Figure 1D, and let stand for 15 min to allow beads to migrate towards the surface.

Materials and Reagents
k. Expose dishes to 365 nM UVA for 15 min in the UV crosslinker.
l. Fill dishes with deionized water as shown in Figure 1E n. Dehydrate gels at 50 °C on a hot plate until all water has evaporated from the gel. A dehydrated gel is shown in Figure 1A. Gel dishes can be stored for up to one month in a dark, dry location.

B. Microarray printing
This process prints an array of circular spots of biomolecules onto the polyacrylamide substrates.
These array spots are where cells will adhere forming circular islands.

Note: ECM protein printing (ECMP) buffer is appropriate for most ECM molecules. Growth factor (GF) printing buffer is suitable for other classes of molecules such as growth factors or ligands,
where a low pH could cause issues. For many ECM proteins, 250 µg/m is a suitable concentration. Optimal concentrations will vary depending on the molecule, its retention, and its function. The total volume in each well should be between 5 and 15 µl.
b. In your source plate, you should also include a solution with a fluorescent marker which will be used to convey the orientation of the array to determine the locations of each condition.
We recommend rhodamine-conjugated dextran at a final concentration of 2.5 mg/ml. The source plate configuration will differ based on the arrayer, pin configuration, and desired array layout.
c. Mix each well thoroughly by pipetting. Take care to avoid generating bubbles. Centrifuge the source microplate for 1 min at 1,000 x g. Source plates can be used immediately or 8 www.bio-protocol.org/e3418 e. Prepare the microarrayer and arraying program using the manufacturer's software. The setup and programming will differ based on the arrayer and desired array layout. The program should be devised such that the array orientation is unambiguous, and the locations of each arrayed conditions are known and could be determined by the location relative to the fluorescent marker in any orientation. An example of this is provided in Figure 2.  i. Place dishes into an appropriate adaptor. If the arrayer can fit a standard multiwell plate, the 6-well plate is suitable to hold the dishes.

Note: See the note under Materials and Reagents #5.
j. Begin array fabrication. Check frequently that the humidity has not dropped below 65% RH (non-condensing).
k. When the program is complete, store fabricated arrays covered with aluminum foil at room temperature and 65% RH (non-condensing) overnight. While the array spots are visible, it is helpful to visibly mark the top or bottom of the array so the orientation is known when placing on the microscope. For some hydrogel and pin combinations, it may be necessary to store arrays at ambient temperature and humidity for an additional two days to ensure arrays have dried completely. Arrays can be stored for up to 7 days before use.

C. Seeding cells on microarrays
Here cells are transferred from their normal culture condition on to the microarrayed hydrogel substrates for TFM.
1. To sterilize the gels, add 3 ml of PBS with 1% v/v penicillin/streptomycin. Expose to UV C for 30 min. Exchange penicillin/streptomycin solution for cell culture media.
2. Collect and count cells following the cell appropriate procedure. Resuspend cells in culture media at an appropriate concentration for seeding. This will differ based on cell type but will likely range between 170 x 10 3 and 7 x 10 5 cells/ml. Add 3 ml of cell suspension to each dish.
Incubate dishes at 37 °C and 5% CO2 for 2-24 h, or until confluent cell islands have formed.
Seeding density and time may need to be optimized for your cells and application. Agitation of the dishes every 15-60 min may also aid in forming consistent, confluent islands.
3. Once islands have formed, rinse arrays twice with 3 ml prewarmed media. At this stage, add any experimental treatments such as growth factors or inhibitors. Change media every 1-2 days until time to perform TFM, or as your cell culture protocol suggests, maintaining any treatment concentrations at each exchange. Figure 3 shows an example of an array with cellular islands.  5. Begin the automated imaging of the phase contrast images of the cell islands. Save this file with a suitable name that notes the experiment details as well as that it is the phase contrast image.
6. Switch to the appropriate fluorescent channel for the beads. Individually find and save the Z plane focus of the top surface of the gel under each island. Take care to avoid changing the XY positions. Save this file with a suitable name that notes the experiment details as well as that it is the pre-dissociation image. An example image is shown in Figure 5B. Save this file with the same experiment details as well as that it is the pre-dissociation image.
8. Carefully add 150 µl of the SDS solution to the dish, taking care to not bump or move the dish.
Monitor dissociation of the cell islands using the phase contrast channel. Wait until the islands have completely dissociated from the substrate at which time there the island locations, when viewed in phase contrast channel, should appear mostly blank.

Note: Some cells may require addition of more SDS solution, or higher concentration.
Additionally, Triton-X may be used instead. 1. Ensure the provided MATLAB code has been saved to an appropriate location. Navigate to the folder where this directory has been saved.
2. Make sure all image files have been saved or transferred to a folder available from the computer to be used for analysis.
3. From the command window, run the function run_island_tfm with no inputs.
4. You will be prompted to select the file with the phase contrast image. Navigate to and select the file with the phase contrast image.  Pixel size is pulled from the image metadata but can be set manually. All other fields are set by the user. 7. Once these fields are completed, click "Set Info." This will populate the "Data file name" field using the entered information which will be the file name of the output file. This field can be changed manually. Two additional fields will appear. Enter a name for the first condition in the "Condition 1" field. Enter the numbers of all the islands assigned this condition in the "Islands with Condition 1" field as a list of numbers. Click "Set Condition." If there are additional conditions, repeat these steps for each of the remaining conditions. Once all conditions have been set, the "Done" button will appear. Confirm that all experiment information is correct, then click "Done." All islands assigned a condition in this step will be analyzed. To analyze only some islands, set the number of conditions to the number conditions represented by the islands to be analyzed, and only list those islands in the "Islands with Condition" field. 8. The script will now cycle through the conditions and islands. It will first run a script to correct frame shifts between the before and after dissociation. A GUI will now appear to aid in drawing a boundary around the cell island. This GUI is shown in Figure 7. The software will attempt to draw a boundary around the cell island, which is plotted in red. The slides can be used to adjust the parameters of the trace which will cause the trace to rerun in real time. You can also change the values of these parameters by entering numbers into the text boxes and then hit the "Rerun" button. Alternatively, you can draw the boundary manually by clicking the "draw manual" button. Left click around the island until the boundary is closed. Double click inside the boundary to create the shape. To avoid sharp corners, after the polygon is drawn, a blurring and rounding is applied using the current value in the "Blur" field.
To avoid this rounding, set this field to 1. The manual draw function can be used to trace multiple areas. To reset the manual boundaries, click "Clear boundaries," or click "Rerun" to repeat the automated tracing. When satisfied with the boundary, click "Done." 9. This will bring up the next island. Repeat this process for all islands in the file selected for analysis. 10. When the boundary trace has been completed for each island, the data will be saved. The program will then move to calculate the displacement and traction fields for each island. This process can take up to several hours depending on the number of images, image size, and processors on the computer. Data is saved after each island to limit data loss in the case of issues during analysis. Data is saved in the folder "data out." 11. Most of the computation time is the displacement field estimation. To rerun the traction field calculation with different settings on a file which already has the boundaries identified and displacement field calculated, run the command run_island_tfm('rerun'). You will be prompted to choose a data file which will be reanalyzed. 15 www.bio-protocol.org/e3418 B. View data and generate summaries 1. In the code provided, the output is saved in the folder "data out" with the file name established during analysis. To explore data from a single file, data can be loaded into the workspace by double clicking in the Current Folder explored or using the load function.
2. This loads the cell array "all_cell_data" into the workspace. Each cell corresponds to a single island from the analysis. The output is intended to provide easy access to all relevant data for the user to explore and analyze in MATLAB or export to other programs for analysis as appropriate for their application. Table 1 provides the organization of the data stored within the output file. Data structure elements can be accessed using dot notation of the form structName.fieldName.
Note: See https://www.mathworks.com/help/matlab/matlab_prog/access-data-in-a-structurearray.html for more information on accessing data in structures.  The island boundary is plotted in black. The phase contrast image is also displayed, with the boundary plotted in yellow. The best fit ellipse is shown in red, with the major and minor axes.
4. Using the boundary traced of the island, the script finds a best fit ellipse. This ellipse can be used to align compare many replicates of islands with the same geometry. Use view_one_island_rotandcen to view the data centered and rotated according to the best fit ellipse.
5. The function collect_island_data is provided to collect data from multiple islands across multiple TFM runs. The output of this function is a table with data and information on each of the islands loaded from the files (see Table 1), and a structure holding the displacement and traction data, indexed according to the "summ_ind" field of the summary table This table can be exported to excel or similar. When the function is run without an input, you will be prompted to select data files to load and consolidate. 17 www.bio-protocol.org/e3418    6. The output of the collect_island_data function can further be used to create summary data. The function summarize_islands plots and outputs averaged displacement and traction fields, see Figure 9 for an example. This function uses the aligned the data from each island based on the best fit ellipse. The islands to include in the averaging can be selected by choosing a subset of set of islands of the summary_table by indexing on one or more variables. See data_analysis_examples for an example of using this function.  Figure 10. The function summarize_islands_1D is provided to perform this analysis. Here, the XY position of each data point is converted to a radial coordinate, and the radial position is normalized by the measured radius of the island. The traction data is binned by radial coordinate, and a mean is taken. The data used for this analysis can also be selected using the summary table as discussed previously. The function also outputs the data table with the peak traction of each island amended to the relevant lines. See data_analysis_examples for an example of using this function.