Cyclic uniaxial mechanical stretching of cells using a LEGO^® parts-based mechanical stretcher system

Mechanical signals have been established as integral regulators of cell behavior, tissue physiology and pathophysiology (Dufort et al., 2011; Muncie and Weaver, 2018). For cell biologists, it has therefore become essential to account for mechanical parameters in order to investigate and fully comprehend many biological processes. This necessity comes with an unusual toll for cell biologists – to use biophysical techniques and tools in which they have traditionally not been properly trained. Paradoxically, physicists and engineers that can design and operate such systems are generally not trained in cell biology and biochemistry to address biologically relevant matters. Currently, a rising and expanding generation of biophysicists tends to bridge that gap and, together with multidisciplinary collaborations, contributes to merging these fields and addressing those challenges. Nevertheless, for the everyday biologist, the availability of practical tools on a daily basis, in order to address these challenges, remains critical. We have long explored unconventional means to create new tools and think it is of crucial importance to develop tools that can easily be used by cell biologists on a daily basis without constant supervision by physicists or biophysicists. Here, we present a simple and versatile mechanical cell-stretching system that can be used to generate uniaxial cyclic mechanical stretch on cells in tissue culture. This system is based upon a low-cost battery-powered uniaxial cyclic mechanical stretcher exclusively built out of LEGO^® parts combined with a poly(dimethylsiloxane) (PDMS) tissue culture plate coated with extracellular matrix (ECM) protein to grow and stretch cells. We provide detailed assembly instructions for the LEGO^® mechanical stretcher, as well as a 3D model for the PDMS casting mold and point-by-point instructions to produce the PDMS culture plate. We show that the stiffness of the PDMS culture plate can be adjusted according to the required experimental conditions and that the stimulation is homogeneous across its surface. To illustrate the versatility of our system, we detail a number of applications that we validated and can be performed downstream of mechanical stretch using the current setup of the system, including immunofluorescence, western blotting and biochemical assays, such as determining RhoA activity by a Rho-binding domain (RBD) pulldown assay, or functional assays, such as TGFβ (also known as TGFB1) release from cell-derived ECM. Finally, we illustrate how mechanical stretching using this system can be used to study the effect of mechanical constraints on the behavior of integrin adhesion complexes. In conclusion, we present a self-contained mechanical stretching system that can be used to perform a wide variety of applications downstream of cell stretching, with a simple unique experimental setup that is freely available to the cell biology scientific community and easy to assemble.

The mechanical stretcher is a self-contained system that can perform uniaxial cyclic mechanical stretch on cells in tissue culture (Fig. 1A; Movie 1). The system is compact and measures 50 cm × 40 cm × 10 cm, excluding the battery box (Fig. 1B). It can easily fit on a shelf in a tissue culture incubator, allowing multiplexing by fitting several devices in the same incubator (Movie 2). The system consists of two parts: a battery-powered mechanical stretcher exclusively built out of 326 individual LEGO^® parts and a PDMS culture plate used to grow and stretch cells (Fig. 1B,C). The mechanical stretcher includes two synchronized LEGO^® XL motors powered by an AAA battery box, actioning a displaceable arm that stretches the PDMS culture plate. To control stretch frequency, we included a dual function gearbox between the motors and the stretching arm, which synchronizes the motors and, according to the wheel gear combination, sets the speed of the final 40-tooth wheel, determining stretch frequency (Fig. 1B; assembly booklet, https://doi.org/10.5061/dryad.dbrv15dx8). The use of an AAA battery pack eliminates the necessity to have one or several power outlets in close proximity as well as an electric cord exiting from the incubator, and we have been able to continuously stimulate for 6 h on a single pack (Movie 2; Fig. 5D). The second part of the system is an in-lab-designed PDMS culture plate that can be used to grow cells and stretch them with the LEGO^® mechanical stretcher (Fig. 1C). This elastomer-based tissue culture plate constitutes one example of effective design, but any user-designed stretchable plate could be used with appropriate characterization. Our PDMS culture plate has a three-layer design that provides both stretchability and sufficient structural support to prevent unwanted deformation. It is generated by sequential pouring and curing of a PDMS/curing agent mix (Gutierrez and Groisman, 2011) in a casting mold that can be assembled from Plexiglass^® parts or by 3D printing using the provided model (Fig. 1D; Fig. S1). The surface of the PDMS culture plate is functionalized by silanization and chemically coated with an ECM protein, such as fibronectin (FN), for cell attachment. In this model, the surface area of the PDMS plate is 40 cm², which is roughly half that of a 10 cm culture plate. Additionally, in an eco-responsible and cost-saving attitude, the PDMS plate can be partially recycled for a new experiment by replacing the whole bottom of the plate with a new PDMS layer.

We provide a list of the individual LEGO^® parts required to assemble the mechanical stretcher (Table S1) as well as an assembly booklet with step-by-step detailed assembly instructions (https://doi.org/10.5061/dryad.dbrv15dx8). We estimate the cost of the entire system to be ∼€150-200, split into €100 for the LEGO^® parts mechanical stretcher and ∼€100 for two PDMS plates including the Plexiglass^® molds. Once the mechanical stretcher has been initially assembled, a typical experiment takes at most 2 working days from PDMS plate production to downstream application (Fig. 1E).

In order to characterize the mechanical regimen generated by the system, we recorded the system and tracked the position of several parts of the stretcher: the displaceable arm and an immobile LEGO^® pin as control (Fig. 2A; Movie 3). The movement of the arm was recorded and tracked along x- and y-axes over 60 s, at a sampling rate of 29.70 Hz. The arm executes a very reproducible sine wave movement along the x-axis, which was fitted to a sine curve function as described in the Materials and Methods section (Fig. 2B,F). Its mean amplitude was measured at 0.9301 cm (s.d. 0.0029, n=22) at a frequency of 201.04 mHz, meaning that the elongation generated on the PDMS plate is 12.4% of its initial length. In comparison, along the y-axis, only a very moderate sine wave movement of less than 1 mm of amplitude was recorded (Fig. 2C), which was synced with the uniaxial displacement of the arm along the x-axis, indicating a very limited wobbling of the arm. In contrast, the pin fixed on the immobile structure was almost totally immobile, and we could only detect oscillations of the structure in the range of a tenth of a millimeter along both x- and y-axes (Fig. 2D,E). These seemed to be remotely synced with the movement of the arm but none could be reasonably fitted to a sine wave function (Fig. 2C,E).

This system is widely customizable and, among various modifications, stretching frequency can be adjusted by modifying the gearbox in order to reach frequencies ranging from 200 mHz up to 1 Hz by step increments (Fig. 3A–C). We could even push the system beyond 1 Hz and reach 4 Hz, but only at the cost of structural destabilization of the system (Fig. 3D). We would therefore recommend the use of this system up to 1 Hz but would strongly caution against higher frequencies. Similarly, stretch extent could be changed by modifying the length of the displaceable arm.

Besides the mechanical stretcher, the other major component of the system is the elastomer plate. Several parameters characterize and condition the use of such a plate, including functionalization, stiffness and homogeneity. We routinely functionalize the surface of the plate with FN in order to promote cell attachment and spreading (Fig. 4A), but other ECM proteins, such as collagen I (Fig. 4A), can be used by the end user to comply with their specific needs. The functionalization method is very standard in the field and has been reported numerous times (Gutierrez and Groisman, 2011). Another important parameter that influences the investigation of various physiological or pathophysiological situations is the stiffness of the environment and, by extension, that of the PDMS plate. We designed the PDMS plate and chose its stiffness in order to be as physiological as possible and compatible with the power of the LEGO^® motors. We chose a PDMS formulation of 40:1 (w:w ratio elastomer base to curing agent) that would fit our needs based on Gutierrez and Groisman (2011). We sought to experimentally confirm the stiffness of those surfaces by atomic force microscopy (AFM) and/or tensile test. We indented the surface of the PDMS plate at random positions using an AFM fitted with a 5 µm diameter spherical probe and measured its Young's modulus at 111 kPa, approximately 3-fold over the expected 40 kPa (Fig. S2A). We also calculated E by tensile test in order to determine the stress/strain curve and compute the Young's modulus, measuring E at 26.04 kPa (Fig. S2A). This discrepancy in the measurement of the Young's modulus of PDMS has been observed (Evans et al., 2009; Gutierrez and Groisman, 2011; Livne et al., 2014; Song and Ren, 2014; Wang, 2011), reported and discussed (Megone et al., 2018), and occurs when the percentage of curing agent is below 5%, which is the case here with a 2.5% formulation. We can therefore only characterize the elastomer as having a Young's modulus between 26 kPa and 111 kPa, which, despite this inherent lack of precision, falls into a physiological range.

Besides coating and nominal stiffness, we also sought to characterize the homogeneity of the PDMS plate without stimulation or upon stretching from a material and biological standpoint. Although we cannot determine the absolute E value in this range of elastomer formulation, we can still compare the measurements at different positions of the same gel. Therefore, we measured by AFM the elastic modulus of the gel at three random different positions. We measured E at 108, 103 and 126.5 kPa. The distributions were tested in a one-way ANOVA Kruskal–Wallis test, which indicated that we could not exclude that these distributions are sampled from the same distribution. Therefore, the values of E measured at these three different positions are not statistically different, indicating that although uncertainty remains on the value of E measured by AFM, the values measured at different positions on the gels are undiscernible, indicating its mechanical homogeneity. We also assessed the homogeneity of the surface from a biological standpoint. We grew HeLa cells on 40:1 PDMS culture plates for 24 h and assessed YAP (also known as YAP1) nuclear localization by immunofluorescence at different positions on the PDMS plate as a means of evaluating the homogeneity of the mechanical constraints applied to the cells at different positions (Fig. 4B). We quantified YAP localization based on the intensity of the nuclear YAP compared to cytosolic YAP (see Materials and Methods section). Prior to any mechanical stretch, we observed that YAP had a general tendency to be less nuclear in cells grown in the most central portion of the plate (position 1) compared to those grown in other positions, from quantifications performed (Fig. 4B,C). Although the whole plate is made out of the same material, this seemingly indicates that the PDMS plate harbors subtle local differences that can be sensed by cells. Given the design of the plate, we suspect the structural support provided by the plate edges to be the cause of that slight disturbance. In contrast, upon mechanical stretching, no difference could be observed between cells grown and stretched at different positions on the plate (Fig. 4C). Cell stretching induced nuclear accumulation of YAP in cells at the most central position indistinctly of cells at any other position despite initial differences (Fig. 4C). Altogether, we conclude that, despite slight local mechanical discrepancies, the PDMS culture plate is homogeneous both at rest and during stretch.

From the experimental biologist and purchaser standpoint, one of the shortcomings of some commercially available systems is generally their requirement to acquire additional components for specific downstream applications. Here, we present a simpler versatile system and setup, which allows for most usual cell biology applications. We used the current experimental setup to stretch HeLa cells plated on FN-coated PDMS plates for the indicated time period at ∼0.2 Hz frequency and 12% stretch as previously measured. Subsequently, we assayed a number of applications downstream of uniaxial cyclic stretch. One of the most common and early responses to mechanical force application on cells, using integrin-binding magnetic beads (Guilluy et al., 2011) or uniaxial stretching (Pourati et al., 1998; Smith et al., 2003), is the activation of the GTPase RhoA, which controls cell contractility and stiffening. Therefore, we assessed whether our system could trigger RhoA activation in a similar fashion. We stretched HeLa cells for 5 min prior to cell lysis and performed a GST–RBD pulldown assay to monitor RhoA activation (Ren et al., 1999). We observed that uniaxial stretching induced RhoA activation as early as 5 min (Fig. 5A), as previously reported using commercially available systems (Smith et al., 2003). We therefore show that both western blotting and biochemical assays such as GST–RBD pulldown assay can be performed downstream of uniaxial cell stretching with this system. Alternatively, cells can be fixed and stained for immunofluorescent cell imaging using the very same experimental setup. Practically, actin–GFP-expressing HeLa cells were fixed on the PDMS plate, and PDMS samples were cut out prior to staining for vinculin and mounted on a coverslip for cell imaging. Vinculin staining was observed as punctate staining in focal adhesions (FAs) at the end of actin fibers (Fig. 5C), as classically reported (Geiger, 1979). This demonstrates that immunofluorescent imaging of cells plated on the PDMS culture plate can easily be performed, starting from the very same experimental setup. Also, we performed a functional assay in the form of monitoring TGFβ release from cell-derived matrices (CDM) generated on PDMS plates (Tissot et al., 2018). Fibroblasts were grown on FN-coated PDMS plates for 7 days. Fibroblasts were killed with hygromycin and cell-free CDM were assayed for TGFβ release in the supernatant upon uniaxial mechanical stretching. A 4-fold increase in soluble TGFβ was observed after PDMS plate stretching for 10 min (Fig. 5B).

In addition to short-term stimulations, this system can be used to stretch cells for several hours in order to trigger cell reorientation (Sears and Kaunas, 2016). We stretched subconfluent NIH 3T3 cells for 6 h at 200 mHz, 12% stretch on 40:1 PDMS and quantified cell orientation, i.e. the orientation of the cell long axis relative to the stretch axis. Cells were initially randomly oriented in all directions (Fig. 5D,E). We found that, upon cyclic stretching, cells reoriented along the stretch axis (Fig. 5D,E). Indeed, the distribution of cell orientation in each condition was statistically different as tested with a Kolmogorov–Smirnov test (P=0.0001, approximate value) (Fig. 5E). This reorientation is odd because cells are generally reported to reorient perpendicular to the stretch axis. This nevertheless indicates that this system can perform long-term stimulations and trigger cell reorientation. Altogether, this illustrates the versatility of this system that can be used for a variety of applications and stimulations downstream of uniaxial cyclic stretch.

In order to illustrate how this mechanical stretching system can be used in the context of a biological study, we decided to assess FA and cell behavior upon uniaxial stretch of HeLa cells on 40:1 PDMS plates. Cells were stained for nuclei and vinculin, a mechanoresponsive cell adhesion protein (Grashoff et al., 2010) previously reported as regulated by mechanical stretching (Sen et al., 2011) (Fig. 6A). First, we observed that cell stretching induced an accumulation of vinculin-containing cell matrix adhesions, as the number of adhesions per cell increased from 132.2 to 232 adhesions per cell (n=49 and n=48, respectively). As the number of cell adhesions increased, their distribution was also altered, together with cell shape. After stretching, cells were smaller as quantified by their 15% reduction in cell area and the associated decrease of their form factor score (Table 1 and Fig. 6C). Meanwhile cell eccentricity remained unaltered. FA size was stable at 1.82 µm², indicating that newly formed cell adhesions were the same size as pre-existing adhesions (Table 1). In contrast, FA eccentricity increased upon cell stretching, indicating that FAs were slightly more elongated upon stretching (Table 1). The distribution of FA eccentricity revealed that the FA shape was globally modified, as opposed to a population of FAs of different shapes appearing (Fig. 6D). FA distribution over the cell surface was altered by cell stretching: although cells were generally smaller, the distance between the FA centroid and the cell centroid or the cell edge both increased, indicating that FAs were concentrating in a specific belt-like zone (Table 1 and Fig. 6B,C). More specifically, FAs accumulated at the forefront of cells upon stretching (Fig. 6E).

Overall, this indicates that uniaxial cyclic mechanical stretching modifies cell shape and FA distribution and density, as observed by monitoring a mechanoresponsive FA protein. This also illustrates how our mechanical stretching system can be used to assess the effect of mechanical constraints on a particular aspect or process of cell biology.

Here, we present a very simple and effective experimental system that can be used to apply uniaxial cyclic mechanical stretch to cells in culture. This two-part system consists of a battery-powered LEGO^® bricks-based mechanical stretcher combined with a stretchable PDMS culture plate. The first component of the system, the mechanical stretcher, is exclusively made of LEGO^® parts that can be purchased from various online stores and assembled according to the provided list of parts and assembly instructions. This device is compact and battery-powered, and one or several of these can easily fit in a tissue culture incubator, allowing for multiplexed experiments. This multiplexing possibility is further reinforced by the very modest cost of the device. The present setup allows for a sine wave type of stimulation at a frequency of 0.2 Hz with a 12% stretch (Table 2). By modifying the gearbox, as advised here or, alternatively, according to one's own specifications, the stretch frequency can be adjusted from at least 0.2 Hz up to 1 Hz by step increments. We do not recommend stretching at a higher frequency than 1 Hz since this results in structural destabilization of the device. These frequencies fall into the range generally reported in the literature for various other custom systems, which spans from 0.01 Hz to 5 Hz (Greiner et al., 2013; Livne et al., 2014). Similarly, the stretch elongation of ∼12% achieved by our system is close to what is reported elsewhere for custom systems, generally between 8% and 10% (Greiner et al., 2013; Livne et al., 2014), and commercial systems, up to 12.2% with the Flexcell FX-6000 system. Overall, the most performant custom system could stretch from 0.5% to 30% at frequencies ranging between 0.001 Hz and 15 Hz (Livne et al., 2014). In parallel, one of the most versatile systems in terms of stimulation form and range, the FX-6000 system from FlexCell International (Burlington, NC), can perform uniaxial and equibiaxial cyclic stretch at a wide range of frequencies and stretch elongations. Clearly, this puts our system behind in terms of raw performance and stimulation variety. Nevertheless, this performance has to be put into perspective relative to the versatility of our system, its simple and user-friendly design, and, ultimately, its cost. Obviously, our system is not designed to replace high-end commercially available systems. These systems allow very refined tuning, such as manipulating the waveform of the stimulation, its stretch amplitude or its frequency. They are adapted for very intensive use and provide technical customer support. In contrast, we intended to design our system to be as simple, versatile and cost-effective as can be, while offering a robust sine wave-like stimulation. With a very simple and unique setup, this system offers the possibility to perform a wide variety of downstream applications including western blotting, immunofluorescent staining and imaging. Although, we did not formally test it, we anticipate that it may also be used to purify RNA or DNA. It offers a practical and accessible option for either sporadically performing cell stretching or assessing the relevance of this type of stimulation in the initial steps of a project. Its simplicity also ensures a very low and fast learning curve in order to efficiently use the system compared to commercial systems, which may be more intricate to set up and use. Finally, this system contributes to an expanding trend to generate and provide researchers with open-access platforms based on LEGO^® parts or 3D-printed parts (Almada et al., 2019).

The second component of the system, the stretchable plate, is made of PDMS, an elastomer routinely used for the design of microfluidic devices, which is optically transparent and biologically inactive. We designed the plate such as to fulfill several requirements: a large cell culture area to enable biochemical assays, a multipurpose design and a physiologically relevant stiffness. This plate design is only an example of a design fulfilling our requirements and we encourage each investigator to develop their own design based on individual needs. We promote and used some design principles of our own that do not constitute an inalienable rule. We designed the plate to be a thick surface in order to avoid the use of a spin-coating device, thereby limiting the cost of the device and also limiting the requirement for additional equipment to produce the plate. Although many custom or commercial systems are miniaturized, offering the possibility to be mounted on a microscope, we aimed to design a plate that would, among other possibilities, enable biochemical analysis. Therefore, we designed it as a 40-cm² rectangular-shaped surface, approximately half the size of a 100-mm Petri dish. This is noticeably larger than other systems, which generally report areas in the range of a few cm² (Jungbauer et al., 2008). Incidentally, the caveat associated with such a design is that the stretcher is somewhat too large to be easily mounted on a microscope. Nevertheless, this design, coupled with the optical properties of PDMS, allows for a variety of applications including western blotting, biochemical assays or immunofluorescence. Other devices are generally restricted, by design, to specific applications (Greiner et al., 2013; Livne et al., 2014) or require additional/specific purchase for commercial systems. Finally, we sought to use an elastomer for which stiffness would be relevant to physiological or pathological situations. Other systems generally do not report the stiffness of the PDMS membrane or report values in the range of several MPa (Greiner et al., 2013; Livne et al., 2014). We used a PDMS formulation of 40:1 weight/weight ratio of base to curing agent reported to generate gels of 40 kPa of Young's modulus. We intended to experimentally confirm that elastic modulus using various assays only to face an unexpected oddity of PDMS. Above 5% content of curing agent, any experimental method can be used to determine the elastic modulus of a PDMS gel. In contrast, below that 5% content limit, indentation-based and tensile/compressive methods yield diverging values of the elastic modulus (Megone et al., 2018). It turns out that our PDMS formulation of 40:1 corresponds to a 2.5% content of curing agent and, indeed, AFM and tensile measurements computed E values of 111 kPa and 26 kPa, respectively. Although, this effect has been observed numerous times (Evans et al., 2009; Gutierrez and Groisman, 2011; Livne et al., 2014; Song and Ren, 2014; Wang, 2011), it remains largely confidential and has never been formally addressed. We think this might be due to the widespread use of stiff PDMS for cell culture and for the design of microfluidic devices, which hinders this issue. Therefore, this PDMS formulation prevents us from precisely determining the elastic modulus of the surface, although it does fall into a physiological range.

Finally, we validated the use of this system by reproducing one of the most mainstream experiments performed with uniaxial stretching devices: cell reorientation. We provide evidence that our system can be used for long-term stretching experiments in order to trigger cell reorientation. Surprisingly, our system induced cell reorientation along the stretch axis rather than perpendicular as generally reported. This is most likely due to the fact that we use soft elastomer surfaces in the range of several tenths of kPa as opposed to stiff elastomers in the range of several MPa. Indeed, it has been reported that cells seeded on collagen I-coated silicon membranes orient perpendicular to the stretch axis, whereas cells seeded on collagen I gels reorient along the stretch axis (Walters et al., 2017). The main difference between both conditions is the stiffness of the gel supporting the cells: a 40 shore A durometer silicon sheet has an elastic modulus of ∼2 MPa, whereas an 80 mg/ml collagen I gel has an elastic modulus of ∼40 kPa. This is very similar to the stiffness range of our PDMS plates. As indicated, other systems use PDMS membrane for which stiffness generally falls in the MPa range, whereas our system uses kPa-range PDMS plates. This may explain the discrepancy regarding cell reorientation between our systems and others. Indeed, we tried to stretch MPa-range PDMS plates in order to address more precisely this discrepancy, but the raw power of the motors we use was too weak to actually stretch PDMS plates in this stiffness range.

Altogether, this system constitutes a robust, cheap and accessible uniaxial cyclic stretching system with simple capacities in terms of stimulation, but versatile possibilities in terms of applications, that may allow any cell biologist to address simple mechanobiology-related questions.

Unless mentioned, all chemicals were from Sigma-Aldrich. Unless mentioned, all tissue culture media were from Thermo Fisher Scientific.

The mouse monoclonal anti-RhoA 26C4 (western blotting 1:500), anti-vinculin 7F9 (1:500) and anti-YAP 63.7 (1:100) antibodies were from Santa Cruz Biotechnology. Secondary anti-mouse horseradish peroxidase-coupled IgG (1:10,000) was from Promega (W4021). Secondary goat anti-mouse Alexa Fluor 594 antibody (1:250) was from Thermo Fisher Scientific (11,005). Actin was stained using fluorescein isothiocyanate-conjugated phalloidin (1:1000) from Thermo Fisher Scientific (F432). Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) at 1 µg/ml.

HeLa cells were from American Type Culture Collection (ATCC CCL-2) and were grown in Dulbecco's modified Eagle medium, high glucose containing 10% fetal bovine serum and 2 mM L-glutamine at 37°C and 5% CO₂. MFB-F11 cells were a gift from Dr Tony Wyss-Coray (Stanford University, CA) and were cultivated as described in Tesseur et al. (2006). Cells are routinely tested for mycoplasma infection by PCR.

The LEGO^® bricks-based mechanical stretcher was assembled using LEGO^® parts from the LEGO^® parts list provided in Table S1, according to the detailed assembly instruction booklet provided in https://doi.org/10.5061/dryad.dbrv15dx8. Please note that in assembly steps 89 and 91, the black lift arms are glued to the connecting part.

The three-layered PDMS plate was produced by sequential pouring and curing of PDMS mix. First, the casting mold is assembled from individual Plexiglass^® parts (Fig. S1). Alternatively, the casting mold can be generated by 3D printing using a material that can withstand temperatures of 80°C for several hours. The PDMS plate is generated upside down in the casting mold, meaning that the top of the PDMS plate is casted first. A first 15 ml layer of 1:20 (weight ratio, curing agent to PDMS) PDMS mix is poured in the casting mold and cured for 2 h at 80°C (top dark layer on the schematic plate in Fig. 1C). A second 40 ml layer of PDMS mix is poured and cured for 2 h at 80°C. This layer will constitute the bottom of the culture plate on which cells will grow; therefore, its stiffness needs to be adjusted according to the end-user requirements (gray middle layer in Fig. 1C). A final 5 ml layer of 1:20 PDMS mix is poured and cured for 2 h at 80°C (bottom dark layer in Fig. 1C). The casting mold is disassembled and the PDMS plate carefully removed from the mold.

The mechanical properties of the samples were studied using a BioScope Catalyst atomic force microscope (Bruker) coupled with an optical microscope (Leica DMI6000B). The experiments of nanoindentation were performed using a probe with a Borosilicate Glass spherical tip (5 μm diameter, Novascan) mounted on a V-shaped cantilever with a nominal spring constant of 0.06 N/m. Indentations were performed in relative trigger mode and by setting the trigger threshold to 1 nN or 3 nN. The apparent Young's modulus was calculated using the NanoScope Analysis 1.50 software (Bruker), fitting the force curves to the Hertz indentation model, and using a Poisson's ratio of 0.5. All force-distant curves not having clear base line, a maximum above 1 nN or 3 nN, or a change of slope in the region of the fitting (minimum and maximum force fit boundary 5% and 25%, respectively) were discarded.

The inner bottom of the PDMS plate is sialinized with a mixture of 20% 3-aminopropyltriethoxysilane (APTES) (Thermo Fisher Scientific) in absolute ethanol for 5 min (Gutierrez and Groisman, 2011). The plate is washed three times in absolute ethanol and left to dry. The plate is then chemically coated with ECM proteins. We coated the plates with bovine FN 10 µg/ml in solution in 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (Thermo Fisher Scientific) in PBS for 30 min. Alternatively, PDMS plates were coated similarly using rat tail collagen I at 50 µg/ml (Sigma-Aldrich). The plate is washed once prior to cell seeding.

Cells are seeded 24 h prior to stimulation at the proper cell density. The PDMS plate is carefully equipped with the LEGO^® adaptors and mounted on the cell stretcher. Uniaxial cell stretching is performed for the indicated time, the PDMS plate is retrieved and cells are processed for downstream application.

Whole-cell lysates were prepared using 50 mM HEPES pH 7.4, 250 mM NaCl, 5 mM MgCl₂, 1% Triton X-100, 0.1% SDS, 5 mM dithiothreitol (DTT) and mini EDTA-free protease inhibitors (Roche) at 4°C. Protein lysates were quantified using the bicinchoninic acid (BCA) method (Thermo Scientific Pierce), loaded on gels for SDS-PAGE and analyzed by western blotting.

Activation of RhoA or Rac1, respectively, was measured in a GST–RBD or GST–PBD pulldown assay as described previously. In brief, cells were lysed for 10 min in 25 mM HEPES pH 7.3, 150 mM NaCl, 5 mM MgCl₂, 0.5% Triton X-100, 0.1% SDS, 10 mM NaF, 5 mM DTT and protease inhibitors at 4°C. Triton X-100-insoluble material was removed by centrifugation for 10 min at 9500 g and the lysates were incubated for 40 min with 50 μg immobilized GST–RBD at 4°C.

Dermal fibroblast (5.104 cells/cm²) were grown for 7 days on FN-coated PDMS plates. Cells were killed using 125 μg/ml hygromycin. Plates were then stretched uniaxially at 200 mHz and 12% stretch for 10 min in 1 ml serum-free medium. The supernatant was collected and then used to quantify TGFβ activity. Active TGFβ was quantified using MFB-F11 cells, which produce secreted alkaline phosphatase (SEAP) under the control of SMAD protein-binding promoter in response to TGFβ (provided by Dr Tony Wyss-Coray) (Tesseur et al., 2006). To assess TGFβ levels from serum-free stretched-ECM supernatant, MFB-F11 (4.104 cells/well in a 96-well plate) cells were allowed to adhere for 4 h before being subjected to conditioned medium for 24 h. Twenty-four hours later, SEAP activity was measured using an SEAP Reporter Gene Assay, chemiluminescent (Roche), according to the manufacturer's instructions, using a Centro LB 960 Microplate Luminometer (Berthold).

Cells were fixed in 3.7% paraformaldehyde in PBS, 100 mM sucrose for 30 min, and permeabilized with 0.5% Triton X-100 in PBS for 3 min before incubation with anti-YAP antibody (1:100 in PBS) for 1 h at room temperature. Cells were washed three times and incubated with fluorescently labeled secondary antibody (1:250 in PBS) supplemented with DAPI for 1 h. Hydrogels were mounted in Mowiol^® and imaged on a Zeiss Axiovert 200M inverted wide-field microscope.

YAP localization in cells was quantified as follows: cells were separated into three groups based on the intensity of the nuclear signal of YAP staining relative to the intensity of the cytosolic signal along a line crossing the nucleus and cytosol. Cells were placed in group C (cytosolic) if, on the line scan, the intensity of the staining in the nucleus was equal to or less than that of the cytosol; in group N/C (nuclear and cytosolic) if the signal in the nucleus was more than that of the cytosol but less than 2× the cytosolic intensity; and in group N (nuclear) if the intensity in the nucleus was at least twice that of the cytosol.

CellProfiler (https://cellprofiler.org) was used to identify and quantify FA and cell features. We generated a CellProfiler pipeline in order to identify and characterize FAs and cell shape (available upon request). Briefly, the workflow is, starting with a DAPI and a vinculin staining image, to identify nuclei and cells then identify FAs and restrict analysis to those FAs in cells not touching the sides of the image. The MeasureObjectSizeShape module was used to quantify FAs and cells features. Data were extracted to Microsoft Excel and statistical analysis was performed with GraphPad Prism. Briefly, ‘FA to cell centroid’ measures the distance between the individual FA centroid and the cell centroid. ‘FA eccentricity’ or ‘cell eccentricity’ models the object as an ellipse and measures its shape; when eccentricity equals 0 the object is circular and when eccentricity equals 1 the object is a line. ‘Minimum distance to edge’ measures the distance between the individual FA centroid to the nearest pixel outside the parent object (the cell). The form factor characterizes the form of the object according to the formula ⁠; when the form factor equals 1 the object is circular.

Cells were stained for actin and nuclei with fluorescently labeled phalloidin and DAPI, respectively. Images were analyzed using CellProfiler in order to identify cells and characterize their orientation relative to the stretch axis. The pipeline will be provided upon request.

The mechanical stretcher was video recorded at a sampling rate of 29.70 Hz. The position of the arm and the black LEGO^® pin were tracked using the Tracker software at https://physlets.org/tracker/. Data were collected in Excel and the relative displacement of both objects was calculated. In Fig. 2, the relative displacement of the arm was fitted to a sine wave function of the form a0+a1×cos(x×w)+b1×sin(x×w)+a2×cos(2×x×w)+b2×sin(2×x×w)+a3×cos(3×x×w)+b3×sin(3×x×w)+a4×cos(4×x×w)+b4×sin(4×x×w), using the curve fitting module in MATLAB (MathWorks) with a0=0.5312, a1=−0.427, b1=0.03399, a2=−0.06693, b2=0.0206, a3=−0.02445, b3=−0.00433, a4=0.01553, b4=−0.008956 and w=43.8 (R²=0.9927) for the arm. In Fig. 3, curve fitting provided the following results for each wheel combination: Fig. 3B: a0=0.5017, a1=0.3953, b1=−0.1034, a2=−0.05462, b2=0.03491, a3=0.02167, b3=−0.01518, a4=−0.006378, b4=−0.01592 and w=3.795 (R²=0.9996); Fig. 3C: a0=0.5051, a1=−0.3843, b1=−0.06256, a2=−0.06256, b2=0.02401, a3=−0.01644, b3=−0.01473, a4=0.01287, b4=−0.004125 and w=6.098 (R²=0.997).

The elastic modulus of a collagen I gel can be calculated using the curve fitting equation extracted from Joshi et al. (2018), E=186c^1.226, where E is in Pa and c is the collagen I concentration in mg/ml. For silicon sheets, the relationship between the elastic modulus and the shore durometer scale is E=10^{(0.0235S-0.6403)}, where E is in MPa and S is the type A shore durometer scale value for S values of 20<S<80.

Cell culture experiments were performed at least three times. All quantifications represent mean±s.e.m. Images are representative of experiments that have been repeated at least three times. Group comparison was performed using two-tailed unpaired Student's t-test. Multiple group comparison was performed by two-way ANOVA with Tukey's correction for multiple comparisons. Frequency distribution histogram analysis was performed to assess normal distribution. Variance difference was assessed by F-test. Comparison of the distribution of two populations was performed with a Kolmogorov–Smirnov test.

The assembly booklet has been deposited in Dryad and is available at https://doi.org/10.5061/dryad.dbrv15dx8.

We thank Soline Estrach and Laetitia Seguin for discussions and Laurence Cailleteau for technical assistance.