SCRATCH: A programmable, open-hardware, benchtop robot that automatically scratches cultured tissues to investigate cell migration, healing, and tissue sculpting

16 17 Despite the widespread popularity of the ‘ scratch assay’, where a pipette is 18 dragged through cultured tissue to create an injury gap to study cell migration and 19 healing, the manual nature of the assay carries significant drawbacks. So much of the 20 process depends on individual manual technique, which can complicate quantification, 21 reduce throughput, and limit the versatility and reproducibility of the approach. Here, we 22 present a truly open-source, low-cost, accessible, and robotic scratching platform that 23 addresses all of the core issues. Compatible with nearly all standard cell culture dishes 24 and usable directly in a sterile culture hood, our robot makes highly reproducible 25 scratches in a variety of complex cultured tissues with high throughput. Moreover, we 26 demonstrate how scratching can be programmed to precisely remove areas of tissue to 27 sculpt arbitrary tissue and wound shapes, as well as enable truly complex co-culture 28 experiments. This system significantly improves the usefulness of the conventional 29 scratch assay, and opens up new possibilities in complex tissue engineering and cell 30 biological assays for realistic wound healing and migration research. 31 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 2

32 33 The ‘scratch assay’ (Fig. 1A) —dragging a pipette tip or sharp object through a 34 cultured tissue and monitoring the cellular healing response into the resulting gap—is 35 among the most common approaches to study cell migration and healing in vitro (1,2), 36 but also perhaps among the least reproducible and scalable due to the manual nature of 37 the process (2–5). While a popular protocol paper on the manual method has nearly 38 5000 citations at this point (1) and the method is largely free, the traditional scratch 39 assay relies on pressure, tool orientation and brand, speed, and manual stability, and is 40 inherently limited in precision, throughput, and scalability (e.g. it is more difficult in a 96-41 well plate than in 6-well plate). Moreover, there is a missed opportunity to use 42 ‘scratching’ as a form of subtractive manufacturing to produce much more complex 43 tissue geometries and easily prepare unique systems-level co-cultures. Given the 44 ubiquity and importance of scratch assays, new approaches improving the 45 reproducibility, throughput, and versatility can benefit a broad range of research fields. 46 47 While alternative solutions to generate gaps in tissues are well-represented in the 48 literature, none of them address all of the challenges (2). One popular approach is the 49 ‘barrier removal assay’ where cells are seeded on either side of a rubber stencil and 50 then the stencil is removed to generated a ‘gap’ (6–9). While versatile, the approach 51 requires precision pipetting (10), and simply does not scale to small culture vessels. 52 Commercial rubber inserts are available, but are limited in geometry and configuration 53 as well as being costly consumables. Further, there is a concern that barrier removal 54 does may not properly damage the surrounding tissue consistent with actual injury (11). 55 Similarly, DIY parallel scratchers based on machined or molded tips have been 56 effectively used in multi-well plate studies (3,12,13), but the approach relies on 57 sophisticated machine shop CNC capabilities, still requires user applied pressure and 58 speed, can only make straight lines, and is intrinsically limited to a single specific 59 substrate (e.g. 96-well only). While commercial scratch systems exist (14,15), they are 60 also limited to only a few well-plates options (e.g. 24/96-well) and straight lines, and the 61 cost is prohibitively high, relatively speaking (~10k-20k USD at the time of writing). 62 Finally, numerous non-mechanical strategies have been developed that rely on 63 electrical, chemical, and optical patterning allowing improved precision (down to the 64 micron scale), but carry their own limitations to cost, throughput, and versatility (2). 65 Hence, there is an exciting opportunity to redevelop the common, mechanical form of 66 the scratch assay both around flexible, programmable, open-hardware that can be 67 adopted by any laboratory. 68 69 All of the key variables and challenges discussed here are the things that a robot 70 excels at—precision, reproducibility, throughput/repetition, and programmability. 71 Inspired by these advantages, we modified a low-cost robotic platform originally 72 intended for art generation. We call this device SCRATCH—Scalable Cellular Resection 73 Apparatus To Characterize Healing. SCRATCH allows: (1) complete programmability to 74 produce almost any pattern; (2) the use of any scratching tip (e.g. pipette tips, needles, 75 wires, etc.); (3) compatibility with nearly all standard culture vessels (3.5 cm dishes to 76 96-well plates); (4) direct use in a sterile culture hood; and (5) a low net cost of 77 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 3 <500USD at the time of writing (16) . The remainder of this report summarizes how 78 SCRATCH works and demonstrates its capabilities. 79 80

81 82 SCRATCH device working principles and system architecture 83 84 SCRATCH is a fully automated scratch assay system, and its key advantages 85 stem from computer-control of a robotic gantry (Fig.1B). The core of the SCRATCH 86 device is a writing/drawing robot that provides programmable lateral (XY) and vertical (Z) 87 movement of the scratching apparatus (Fig. 1C, color-coded arrows). While SCRATCH 88 can be built using off-the-shelf components from the 3D printing community, for 89 simplicity here we modified a hobby ‘art-bot’ (AxiDraw V3, but many others exist) 90 originally intended to hold pens and markers as this saves considerable time for a 91 minimal cost (~$500). This chassis consists of an XY stepper motor-belt system to 92 position the pipette-tip tool over a tissue culture region, and a servo motor to precisely 93 and gently bring the tool into contact with the tissue in preparation for scratching. 94 Instead of a pen or marker, we 3D-printed a customized pipette tip holder for 10 µL 95 pipette tips (this can be tuned for any pipette tip style) (Fig. 1C). To ensure stability of 96 the tip during scratching, we applied a thin layer of reusable adhesive putty (e.g. FunTak) 97 between the tip and the holder. This tip carrier can then be attached to the XYZ gantry 98 as if it were a pen (see Data Availability for CAD file). At this point, SCRATCH is ready 99 for use (see Video S1 for its operation). 100 101 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 4 102 Figure 1. System mechanism and capability 103 (A) Scratch assay is performed by a pipette tip moving across the cell monolayer, leaving a cell-depleted 104 region. (B) System overview. The lateral movement of the pipette tip is actuated by a stepper motor-105 driven belt sytem. Red and green arrow represent X and Y direction respectively. The vertical movement 106 of the tip is actuted by a servo, indicated with magenta arrow. The 35mm dish is placed on a custom 107 designed fixture. (C) A close-up photo of the device operating on a 60mm dish. (D) Phase-contrast image 108 of dish-scale scratch pattern demonstration, scale bar: 2mm. (E) Cytoplasmic staining of scratch pattern 109 in a 96-well dish, scale bar: 1mm. (F) Close-up photo of device operating on a 96-well plate. Arbitrary 110 pattern and well location can be selected. 111 112 A key design goal was to make SCRATCH as user-friendly and reproducible as 113 possible to enable rapid adoption in cell biology labs, so a key feature of our design is 114 our modular sample-holder directly attached to the frame of SCRATCH that allows most 115 standard culture vessels—from 3.5 cm Petri dishes to 96-well plates (Figs. 1D-F)—to be 116 precisely and reproducibly positioned relative to the pipette tool (see CAD file access 117 instructions in Data Availability; see Fig. S1). This sample holder also incorporates an 118 alignment ring to calibrate the tip position at the beginning of the scratch (see Methods). 119 The use of this fixture allows SCRATCH to be controlled using pre-made template files 120 in open-source drawing software (Inkscape already has plug-in support for many 121 drawing-bots) (Fig S2; see also our shared template files). The user then loads an 122 appropriate template for a given culture vessel, draws their desired patterns in each well, 123 and ‘prints’ the scratch pattern on SCRATCH via a USB connection. 124 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 5 125 We demonstrated the versatility of SCRATCH by creating unique patterns in 126 different types of Petri dishes and culture plates. First, we scratched a large-scale ‘star’ 127 pattern across a layer of primary mouse skin keratinocytes in a 35 mm dish (Figure 1D) 128 to demonstrate the ability to generate complex, precise patterns (see Fig. 1D, right). 129 We then tested SCRATCH on a more challenging culture vessel – a 96-well plate. Here, 130 the small well diameter prevents reproducible or precise manual scratching, and the 131 throughput required to scratch 96-wells is not feasible using the traditional manual 132 approach. However, SCRATCH was able to reliably pattern features (we used a ‘+’ 133 shape) in all 96 wells in <4 minutes. Figure 1F shows a fluorescence image of the 134 resulting patterns. Once calibrated, SCRATCH can automatically and reproducibly 135 scratch arbitrary patterns in most standard culture dishes or plates at high throughput. 136 137 Reproducibility and dynamics characterization 138 139 We first assessed how reproducible SCRATCH patterns were relative to manual 140 patterns using linear scratches made in primary mouse skin keratinocyte layers cultured 141 in 60 mm plates (see Methods); representative results are shown in Fig. 2A. We used 142 the standard deviation of the width of each scratch as the metric for evaluating 143 uniformity. As shown in Fig. 2B, SCRATCH exhibited significantly improved uniformity vs. 144 manual scratching (nearly 4X reduction in standard deviation and on the order of a 145 single cell), while maintaining an average width of ~700 µm (approximate diameter of 146 the 10 µL pipette tip). The observed variations we do see with SCRATCH likely reflect 147 both biological variability in cell orientations and minor vibrations from the motor-belt 148 system (see Fig S3 for high-resolution data on the tip trajectory, and Fig. S4 for a 149 demonstration of the effective resolution limit). 150 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 6 151 Figure 2. Linear scratch quantification and comparison 152 (A) 12 scratches performed by SCRATCH Scale bar: 2mm. (B) Scratch uniformity on keratinocyte 153 monolayer. Edge outline is highlighted in yellow. Device scratch demonstrates lower with variation than 154 manual. Scale bar: 1mm. (C) Wound healing assay on 8 scratches, showing uniform wound closure. 155 Timelapse photos of 0, 12 and 24 hours after scratch are shown. Scale bar: 1mm. (D) Fast and consistent 156 pipette movement from SCRATCH allows low scratch variation on high viscoelasticity tissues. A MDCK 157 monolayer is scratched without calcium chelation. Scale bar: 1mm. 158 159 Therefore, SCRATCH demonstrates superior uniformity to manual scratches in 160 basic tissues, which improves reproducibility of scratch assays and allows higher 161 throughput. As a demonstration these benefits, we rapidly produced an array of 15 162 linear gaps into a primary mouse skin monolayer and quantified the wound closure rate 163 to validate the uniformity (Fig. 2B). Phase-contrast images of 0 hour, 12 hours and 24 164 hours after scratching are shown alongside the quantification (Fig. 2C), and the closure 165 curves indicate relative uniform and tight healing dynamics. 166 167 We next investigated the importance of scratching speed (how quickly the tool is 168 translated through the tissue). This is something impossible to control manually, 169 whereas SCRATCH allows scratch speed to be programmed up to 380 mm/s. Tissues 170 are viscoelastic materials, meaning that their mechanical properties, adhesion to the 171 substrate, and mechanobiological responses depend on the rate at which they are 172 mechanically deformed, not just how much they are deformed, so being able to regulate 173 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 7 the scratching rate should provide unique advantages and a new dimension to consider. 174 In particular, we hypothesized that the high-speed, precise motion of SCRATCH would 175 be particularly useful when working with more challenging tissues possessing strong 176 cell-cell adhesion and relatively weaker cell-substrate adhesion where slow or irregular 177 manual scratching can cause the tissues to delaminate rather than ‘cut’ (17). 178 179 Here, we used the widespread MDCK kidney epithelial model, commonly used in 180 all manner of collective migration experiments and screens and known to exhibit strong 181 cell-cell adhesion and develop collective cell behaviors as a result(9,18–21). We first 182 established a baseline by manually scratching engineered, mature MDCK layers (see 183 Methods) as best we could (Fig. 2D), which resulted in massive, irregular gaps and 184 widespread delamination due to inherent irregularities in the manual process. We 185 observed similar results when set SCRATCH to a slow speed (38 mm/s) and repeated 186 the experiment (Fig. 2D). By contrast, when we repeated the experiment with 187 SCRATCH to the fastest translation speed (380mm/s), we were able to produce highly 188 uniform and more regular scratch patterns in comparison to slower mechanical or 189 manual scratching (Fig. 2D). Overall, SCRATCH was able to deliver more precision, 190 reproducibility, and throughput than manual scratching. 191 192 Subtractive tissue manufacturing: designing complex tissue patterns 193 194 Only laboratory wounds are perfect straight lines, and many studies have 195 emphasized the importance of tissue and wound shape in governing cellular migration 196 and growth (10,22–27). We explored this concept by adapting SCRATCH for subtractive 197 manufacturing of living tissues—gradually removing existing regions of tissue to 198 produce complex patterns (returning to the primary mouse skin monolayer model). 199 SCRATCH enables this by ‘raster cutting’, where it can gradually move the pipette tip 200 tool back and forth while ensuring an overlap in the pattern to fully clear a given region 201 of cells (Figs 3A-B). Here, we chose an approximate overlap of 75%. ‘Positive’ or 202 ‘negative’ patterns can be achieved by selectively scratching the “center” or “edge” of a 203 monolayer, either leaving a solid tissue (‘positive’) or cleared region (‘negative’) (Fig 3C). 204 This subtractive manufacturing method extends the application of SCRATCH beyond 205 pure scratch assays to complex assays evaluating the role of wound size and shape, for 206 example. Moreover, this process is also fully automated within the free software used to 207 control SCRATCH allowing arbitrarily complex patterns as shown in Fig. 3D. 208 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 8 209 Figure 3: Raster mode capabilities and demonstration. 210 (A) Demonstration of area clearance from tip overlap. The programmed path diameters are 3mm, 2mm, 211 1mm and 0.5mm. Scale bar: 2mm. (B) Raster mechanism cartoon and calculation of raster overlap. (C) 212 Demonstration of positive and negative area clearance. Scale bar: 2mm. (D) Complex shape achieved 213 through rastering. Scale bar: 2mm. 214 215 SCRATCH for complex co-cultures 216 217 The “empty space” created by SCRATCH offers new potential for tissue co-218 culture because additional cell types can be back-filled into the newly created empty 219 regions (Fig 4A). As a demonstration of this, we created a complex co-culture using a 220 dermal/epidermal model of fibroblasts (3T3 fibroblasts) and keratinocytes (primary 221 mouse keratinocytes). The resulting spiral pattern is shown in Fig. 4B-C and was 222 produced by first scratching a layer of keratinocytes (pre-stained with a membrane dye), 223 then washing with PBS and backfilling fibroblasts (pre-stained with a different 224 membrane dye) as described in our Methods. The initial population of keratinocytes is 225 shown in cyan and 3T3 in magenta. We also used a nuclear dye (Hoechst 33342) to 226 stain all cell-types. The spiral is clearly visible and the expanded view shows good 227 spatial separation between keratinocytes and fibroblasts. Note that the quality of the 228 backfilling method relies on the confluency of the fist monolayer since the seeded cells 229 will also attach to the area that is outside of intended region. Similar to planar 230 lithography, this process can be repeated multiple times for additional “layers” of cells as 231 long as a co-culture medium exists that can support each cell-type. These data further 232 emphasize the versatility offered by the SCRATCH system to enable not only scratch 233 assays but more complex tissue engineering and cell-cell communication assays. 234 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 9 235 Figure 4. Using SCRATCH for arbitrary geometry tissue co-culture 236 (A) Tissue co-culture through backfilling. A scratch is made on Cell A monolayer. After washing with PBS, 237 desired secondary cell suspension is added. After attachment, the dish is washed with PBS multiple times 238 to remove any unattached cells. Co-culture media is then added and the dish is ready for experiment. (B) 239 Fluorescence image of a spiral scratch on keratinocytes, then backfilled with 3T3 fibroblasts. 240 Keratinocytes are stained with Cellbrite Green and 3T3 fibroblasts are stained with Cellbrite Red, both 241 cells are stained with NucBlue. Scale bar: 2mm. (C) Zoomed in center of the spiral backfill. Scale bar: 242 500um. 243 244

245 246 SCRATCH demonstrates a low cost, fully programmable, and high throughput 247 tool for the popular scratch assay that brings many significant advantages to the method, 248 including improved reproducibility, throughput, and versatility with compatibility for nearly 249 all standard culture plates and dishes. In particular, we showed improved precision, 250 throughput, and reproducibility over manual scratches, as well as the ability to use 251 scratching to produce unique tissue shapes and co-cultures without the need for 252 microfabrication or manual stenciling. 253 254 The open-source and open hardware nature of SCRATCH, combined with its low 255 cost, should substantially aid its adoption, as it can cheaply and easily be incorporated 256 into most cell biological laboratories and used in or out of tissue culture hoods. A key 257 aspect of SCRATCH is that it is easy to modify as a platform, allowing nearly any tip to 258 be incorporated, and allowing for custom programming in Python if unique features are 259 required that the standard graphics software does not allow (for instance, the tip can be 260 programmed to go through a ‘wash’ step where it is agitated in a buffer or ethanol well in 261 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 10 between scratching different wells in a multiwell plate to avoid cross-contamination). 262 Similarly, the SCRATCH style platform can easily be modified with a more precise Z-263 drive to regulate scratching pressure, or enable tip-changes. Moreover, SCRATCH is 264 not dependent on one specific piece of hardware, as any traditional ‘maker’ tool such as 265 a diode laser cutter or 3D printer can be modified to do something similar. This type of 266 versatility can substantially improve the types of applications where scratch-style assays 267 are useful and further aid in its adoption. 268 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 11

269 270 Cell culture 271 272 Primary mouse keratinocytes were provided by the Devenport Laboratory at 273 Princeteon University and cultured in E-medium (Nowak and Fuchs, 2009) 274 supplemented with 15% serum (S11550, Atlanta Biologicals) and 50 μ M calcium. Wild-275 type MDCK-II cells (courtesy of the Nelson Laboratory, Stanford University) were 276 cultured in Dulbecco’s Modified Eagle’s Medium (D5523-10L, Sigma-Aldrich) with 1g/L 277 sodium bicarbonate, 10% fetal bovine serum (S11550, Atlanta Biologicals), and 1% 278 penicillin–streptomycin (15140-122, Gibco). NIH 3T3 fibroblasts were provided by the 279 Schwarzbauer Laboratory at Princeton University. 3T3 cells were cultured in Dulbecco’s 280 Modified Eagle’s Medium with phenol red (D5523-10L, Sigma-Aldrich), 10% fetal bovine 281 serum (S11550, Atlanta Biologicals), and 1% streptomycin/penicillin (15140-122, Gibco). 282 Tissue co-culture media consists of 50% Keratinocyte media and 50% 3T3 fibroblast 283 media (28). All cells were maintained at 37 °C under 5% CO2 and 95% relative 284 humidity. Cells were split before reaching 70% of confluence for maintenance culture, 285 but all the dishes used for scratching had over 90% confluence to ensure even 286 monolayers. 287 288 SCRATCH hardware setup 289 290 Here, we used the Axidraw v3 drawing robot (Evil Mad Scientist, Inc.) to provide 291 XYZ control of our scratching tip. All of the CAD files for the customized attachments 292 and templates we describe here are available at our github repository (See Data 293 Availability section). We designed and 3D printed a custom, modular plate holder that 294 we attached to the Axidraw chassis using two M4 16mm long screws (94500A282, 295 McMaster-Carr) and two M4 nuts (90592A090, McMaster-Carr), and this allows us to 296 mount standard cultureware from 3.5 cm dishes to 96-well plates. We then designed 297 and 3D printed a custom pipette holder with a thin layer of reusable adhesive 298 (10079340647432, Loctite) (FIG S4). The pipette holder assembly was then gently 299 clamped to the vertical stage of the Axidraw using the built-in clamping screw. We 300 calibrated SCRATCH using an alignment ring around the target dish, and press-fit the 301 dish into the modular plate holder. If needed, reusable adhesive can be added to 302 improve stability. With the gantry in pen-up position and powered down (or its motors 303 disengaged), we moved the gantry arm across the dish to ensure vertical clearance 304 through the dish walls, and then aligned the pipette tip with the mark on the alignment 305 ring, this establishes the “origin” of the drawing and the starting point. 306 Upon completion, the pipette tip holder assembly was removed from the vertical 307 stage of Axidraw. Then the dish was removed from the holder and washed with PBS 308 three times to remove cell debris. 309 310 Scratch assay configuration 311 312 The Axidraw V3 is programmed using its official plugin in Inkscape (The Inkscape 313 Team). The “Pen-up” and “Pen-down” range is set to 100% and 0% to ensure vertical 314 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 12 clearance between the wells. Drawing speed is set to 10% (38mm/s) and pen-up 315 movement speed is set to 75% (285mm/s). For contiguous tissues that have high cell-316 cell adhesions, drawing speed is set to 100% (380mm/s). Dialog box “Use constant 317 speed when pen is down” is selected to ensure consistency. Pen raising speed and pen 318 lowering speed is set to “Dead slow” to minimize pipette tip bouncing upon contact. 319 Motor resolution is set to “~2780DPI” for smooth operation and plot optimization set to 320 least to avoid random starting point on a path. For all scratch assays, the programmed 321 path is set to 0.01mm thick and is copied 4 times to the same place for repeated 322 scratches. This ensures good area clearance and avoids uneven scratching due to non-323 conformal contact. 324 For raster mode, we use hatch fills options in Inkscape. Hatch spacing is set to a 325 conservative value 0.1mm, which ensures each region is passed by the pipette tip at 326 least 6 times to avoid any missed scratch zones due to non-conformal contact between 327 the tip and the surface. Hatch angle is set to 45 degrees but can be modified based on 328 the tip. Inset fill from edges option is selected to compensate for the finite tip width, and 329 inset distance is set to 0.187mm (a 75% overlap to ensure path clearance) but should 330 be determined experimentally. 331 332 Tissue co-culturing 333 334 A 35mm dish with confluent keratinocytes was scratched with the steps shown 335 previously. Then the dish was washed with PBS three times and stained with Cellbrite 336 Green (30021, Biotium) at 5µL/mL for 30 minutes. A dish of 3T3 fibroblasts was also 337 stained with Cellbrite Red (30023, Biotium) at 5µL/mL in suspension for 30 minutes. The 338 stained dish is washed with PBS and 2ml co-culture media is added. Stained 3T3 339 suspension is washed with co-culture media 3 times using a centrifuge (5702, 340 Eppendorf). 3T3 suspension is then added to the keratinocyte dish with a density of 341 1000 cells/mm^2. The dish is then incubated for 30 minutes for 3T3 attachment. Then 342 the dish is fixed using 4% paraformaldehyde and stained with Hoechst 33342 (Thermo 343 Fisher) for nucleus. 344 345 Microscopy 346 347 Phase-contrast images were captured with an automated inverted microscope 348 (Leica DMI8) with a 5X objective. For wound healing assays, time-lapse images were 349 captured every 20 minutes. Fluorescence images were captured using an inverted 350 microscope (Zeiss Axio Observer Z1) with a 5x objective, controlled using Slidebook (3I 351 Intelligent Imaging Innovations) with Cy5, FITC, and DAPI filter sets. In both experiment 352 setups, the microscopes were equipped with custom-built incubators maintaining 37 C 353 and 5% CO2. 354 355 356 Image and data analysis 357 358 FIJI (https://imagej.net/software/fiji) is used to process all images, including 359 stitching (29) and wound area calculation for wound healing assay (MRI Wound Healing 360 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 13 Tool, Montpellier Ressources Imagerie). Stitched phase images are processed through 361 FFT bandpass filter (40px max, 2px min) to minimize flat fielding. A custom script is 362 developed to analyze scratch width uniformity. Each scratch is thresholded and 363 segmented to calculate the distance between the edges. Data visualization is performed 364 using GraphPad Prism 10 (GraphPad Software). 365 366 Data Availability 367 368 All CAD files for 3D printing and code necessary to perform the work shown here 369 are available at our laboratory github repository 370 (https://github.com/CohenLabPrinceton/SCRATCH) and we are happy to provide 371 support as needed. 372 Supporting Information 373 Fig. S1: SCRATCH in operation. Close-up image of SCRATCH operating on a 96-374 well plate. The tip is fixed using a thin layer of blue adhesive putty. 375 Fig. S2: SCRATCH programming interface. SCRATCH is programmed through 376 Inkscape software. The dotted line represents scratchable area due to the contact angle 377 between the tip and the edge of the dish. A star pattern is shown here in a 35mm dish 378 template. 379 Fig. S3: Path deviation of SCRATCH. A pen filled with protein-A is fixed on the 380 robot to show vibrations from the X- and Y- motors. The “wobble” deviation is 20um, 381 significantly less than the pipet tip width of 700um. 382 Fig. S4: SCRATCH resolution testing. Scratch resolution using 10L tips. 383 Clearance is lost for scratches less than 1mm apart. Scale bar: 5mm 384 Video S1: SCRATCH operation video. A recording of SCRATCH in operation, 385 accessing arbitrary wells and scratch a “cross” shape in a 96-well plate 386 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 14

387 1. Liang CC, Park AY, Guan JL. In vitro scratch assay: a convenient and inexpensive 388

for analysis of cell migration in vitro. Nat Protoc. 2007 Feb;2(2):329–33. 389 2. Riahi R, Yang Y, Zhang DD, Wong PK. Advances in Wound-Healing Assays for 390 Probing Collective Cell Migration. J Lab Autom. 2012 Feb 1;17(1):59–65. 391 3. Yarrow JC, Perlman ZE, Westwood NJ, Mitchison TJ. A high-throughput cell 392 migration assay using scratch wound healing, a comparison of image-based readout 393 methods. BMC Biotechnol. 2004 Sep 9;4(1):21. 394 4. Cortesi M, Pasini A, Tesei A, Giordano E. AIM: A Computational Tool for the 395 Automatic Quantification of Scratch Wound Healing Assays. Applied Sciences. 2017 396 Dec;7(12):1237. 397 5. Gebäck T, Schulz MMP, Koumoutsakos P, Detmar M. TScratch: a novel and simple 398 software tool for automated analysis of monolayer wound healing assays: Short 399 Technical Reports. BioTechniques. 2009 Apr 1;46(4):265–74. 400 6. Fontanil T, Mohamedi Y, Cal S, Obaya ÁJ. Assessing the Influence of a Protease in 401 Cell Migration Using the Barrier-Migration Assay. In: Cal S, Obaya AJ, editors. 402 Proteases and Cancer: Methods and Protocols [Internet]. New York, NY: Springer; 403 2018 [cited 2024 Apr 7]. p. 133–43. Available from: https://doi.org/10.1007/978-1-404 4939-7595-2_13 405 7. Kroening S, Goppelt-Struebe M. Analysis of Matrix-Dependent Cell Migration with a 406 Barrier Migration Assay. Science Signaling. 2010 Jun 15;3(126):pl1–pl1. 407 8. Das AM, Eggermont AMM, ten Hagen TLM. A ring barrier–based migration assay to 408 assess cell migration in vitro. Nat Protoc. 2015 Jun;10(6):904–15. 409 9. Suh K, Cho YK, Breinyn IB, Cohen DJ. E-cadherin biomaterials reprogram collective 410 cell migration and cell cycling by forcing homeostatic conditions. Cell Reports. 2024 411 Feb 27;43(2):113743. 412 10. Heinrich MA, Alert R, Wolf AE, Košmrlj A, Cohen DJ. Self-assembly of tessellated 413 tissue sheets by expansion and collision. Nat Commun. 2022 Jul 12;13(1):4026. 414 11. Nikolić DL, Boettiger AN, Bar-Sagi D, Carbeck JD, Shvartsman SY. Role of 415 boundary conditions in an experimental model of epithelial wound healing. American 416 Journal of Physiology-Cell Physiology. 2006 Jul;291(1):C68–75. 417 12. Zordan MD, Mill CP, Riese II DJ, Leary JF. A high throughput, interactive imaging, 418 bright-field wound healing assay. Cytometry Part A. 2011;79A(3):227–32. 419 13. Poon PY, Yue PYK, Wong RNS. A Device for Performing Cell Migration/Wound 420 Healing in a 96-Well Plate. J Vis Exp. 2017 Mar 7;(121):55411. 421 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 15 14. Cwycyshyn J, Stansbury C, Meixner W, Hoying JB, Muir LA, Rajapakse I. 422 Automated In Vitro Wound Healing Assay [Internet]. bioRxiv; 2024 [cited 2024 Apr 423 7]. p. 2023.12.23.573213. Available from: 424 https://www.biorxiv.org/content/10.1101/2023.12.23.573213v2 425 15. Yigitbilek F, Conley SM, Tang H, Saadiq IM, Jordan KL, Lerman LO, et al. 426 Comparable in vitro Function of Human Liver-Derived and Adipose Tissue-Derived 427 Mesenchymal Stromal Cells: Implications for Cell-Based Therapy. Front Cell Dev 428 Biol [Internet]. 2021 Mar 26 [cited 2024 Apr 7];9. Available from: 429 https://www.frontiersin.org/articles/10.3389/fcell.2021.641792 430 16. AxiDraw V3 [Internet]. [cited 2024 Apr 7]. Available from: 431 https://shop.evilmadscientist.com/productsmenu/846 432 17. Harris AR, Peter L, Bellis J, Baum B, Kabla AJ, Charras GT. Characterizing the 433 mechanics of cultured cell monolayers. Proceedings of the National Academy of 434 Sciences. 2012 Oct 9;109(41):16449–54. 435 18. Deforet M, Hakim V, Yevick HG, Duclos G, Silberzan P. Emergence of collective 436 modes and tri-dimensional structures from epithelial confinement. Nat Commun. 437 2014 May 6;5(1):3747. 438 19. Overriding native cell coordination enhances external programming of collective cell 439 migration | PNAS [Internet]. [cited 2024 Apr 6]. Available from: 440 https://www.pnas.org/doi/abs/10.1073/pnas.2101352118 441 20. Doxzen K, Vedula SRK, Leong MC, Hirata H, Gov NS, Kabla AJ, et al. Guidance of 442 collective cell migration by substrate geometry. Integrative Biology. 2013 Aug 443 22;5(8):1026–35. 444 21. Poujade M, Grasland-Mongrain E, Hertzog A, Jouanneau J, Chavrier P, Ladoux B, 445 et al. Collective migration of an epithelial monolayer in response to a model wound. 446 Proceedings of the National Academy of Sciences. 2007 Oct 9;104(41):15988–93. 447 22. Heinrich MA, Alert R, LaChance JM, Zajdel TJ, Košmrlj A, Cohen DJ. Size-448 dependent patterns of cell proliferation and migration in freely-expanding epithelia. 449 Rosenblatt J, Stainier DY, Kabla A, editors. eLife. 2020 Aug 19;9:e58945. 450 23. Vedula SRK, Leong MC, Lai TL, Hersen P, Kabla AJ, Lim CT, et al. Emerging 451 modes of collective cell migration induced by geometrical constraints. Proceedings 452 of the National Academy of Sciences. 2012 Aug 7;109(32):12974–9. 453 24. Haeger A, Wolf K, Zegers MM, Friedl P. Collective cell migration: guidance 454 principles and hierarchies. Trends in Cell Biology. 2015 Sep 1;25(9):556–66. 455 25. Xi W, Sonam S, Beng Saw T, Ladoux B, Teck Lim C. Emergent patterns of 456 collective cell migration under tubular confinement. Nat Commun. 2017 Nov 457 15;8(1):1517. 458 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint 16 26. Tarle V, Gauquelin E, Vedula SRK, D’Alessandro J, Lim CT, Ladoux B, et al. 459 Modeling collective cell migration in geometric confinement. Phys Biol. 2017 460 May;14(3):035001. 461 27. Mills RJ, Frith JE, Hudson JE, Cooper-White JJ. Effect of Geometric Challenges on 462 Cell Migration. Tissue Engineering Part C: Methods. 2011 Oct;17(10):999–1010. 463 28. Leal J, Shaner S, Jedrusik N, Savelyeva A, Asplund M. Electrotaxis evokes 464 directional separation of co-cultured keratinocytes and fibroblasts. Sci Rep. 2023 Jul 465 15;13(1):11444. 466 29. Preibisch S, Saalfeld S, Tomancak P. Globally optimal stitching of tiled 3D 467 microscopic image acquisitions. Bioinformatics. 2009 Jun 1;25(11):1463–5. 468 469 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted August 28, 2024. ; https://doi.org/10.1101/2024.08.27.609782doi: bioRxiv preprint