Control of gene expression in engineered mammalian cells with a programmable shear‐stress inducer

Abstract In humans, cellular mechanoperception serves as the basis of touch sensation and proprioception, contributes to the proper programming of cell fate during embryonic development, and plays a pivotal role in the development of mechanosensitive tissues. Molecular mechanoreceptors can respond to their environment by mediating transient adjustments of ion homeostasis, which subsequently trigger calcium‐dependent alteration of gene expression via specific signaling pathways such as the nuclear factor of the activated T‐cells pathway. Although, mechanoreceptors are potential drug targets for various diseases, current techniques to study mechanically gated processes are often based on custom‐tailored microfluidic systems, which require special setups or have limited throughput. Here, we present a platform to characterize shear‐stress‐triggered, calcium‐mediated gene expression, which employs a programmable, 96‐well‐format, shear‐stress induction device to examine the effects of imposing various mechanical loads on mammalian adherent cell lines. The presented method is suitable for high‐throughput experiments and provides a large tunable parameter space to optimize conditions for different cell types. Our findings indicate that the device is an effective tool to explore conditions in terms of frequency, intensity, intervals as well as extracellular matrix composition alongside the evaluation of different combinations of mechanosensitive proteins for mechanically activated gene expression. We believe our results can serve as a platform for further investigations into shear stress‐controlled gene expression in basic research and drug screening.

Mechanosensitive proteins in the plasma membrane of cells are key mediators of mechanoperception (Geffeney & Goodman, 2012).
Bacterial channels such as the mechanosensitive channel of small or large conductance (MscS or MscL, respectively) are simple exemplars. MscS and MscL each consist of two transmembrane domains and form heptameric (MscS) or pentameric complexes (MscL).
They thereby established that mechanically gated pores are activated by local distortions of the membrane and serve as safety valves to protect bacteria from osmotic stress (Booth & Blount, 2012). These channels have been used as models to study the underlying principles of mechanoperception (Liu et al., 2009). In contrast, two of the major mechanosensors in humans, Piezo-type mechanosensitive ion channel component 1 (Piezo1) and transmembrane channel-like protein 1 (TMC1) comprise much larger and more complex structures. Piezo1 may play a key role in the sensation of touch, proprioception, and potentially blood pressure regulation (Saotome et al., 2018), while TMC1 may be a mediator of hearing (Pan et al., 2018). Mechanosensitive channels respond to mechanical cues by transient modulation of ion homeostasis that can trigger subsequent cellular responses through calciumdependent alteration of gene expression.
As a consequence of their involvement with various essential processes in human health and disease, mechanoreceptors are targets for the development of novel drugs (Xiao, 2020). However, current methods to tap into mechanically induced gene expression are mainly built around sophisticated microfluidic devices or focus on calcium transients rather than alterations of gene expression. Calcium transients alone, however, might not be a suitable means to investigate the full range of effects of the respective receptor.
To enable efficient screening of conditions for mechanically induced gene expression, we used a high-throughput method to assess shear stress-induced gene expression (Xu et al., 2018). We employed a 96-well-format shear stress induction device to explore ways of boosting the intrinsic as well as engineered mechanosensitivity of mammalian cells to trigger gene expression from a calciumresponsive promoter. Parameters such as stimulus intensity, frequency, extracellular calcium concentration, and extracellular matrix composition, as well as the presence of accessory proteins, were assessed for their impact on gene expression. We believe this study provides tools and basic data for future explorations of mechanoactivated gene expression in mammalian cells.

| Plasmid generation
All plasmids were cloned by standard molecular cloning techniques.
See Table 1 for a comprehensive list of plasmids and the genetic elements they carry. Details of the amounts of transfected plasmids for each experiment can be found in Table 2.

Transfection
Polyethyleneimine (PEI, 24765-1, Polysciences Inc.) was used for the transfection of plasmid DNA in a ratio of PEI:DNA of 6:1. At 24 h prior to transfection, cells were seeded as described above.
200 ng or 1000 ng DNA was mixed with 150 µl or 500 µl of serumfree DMEM and 1.2 µg or 6 µg PEI (1.2 µl or 6 µl of a 1 mg/ml stock solution) was added per well of a 96-well plate or 6-well plate, respectively. The mixture was incubated at room temperature for 5-10 min before adding it to the cells. Cells were incubated overnight (>12 h) under standard culture conditions. The medium was replaced by standard culture medium (as described above) the next morning and supplemented with CaCl 2 or KCl as required by the experimental design.

| Generation of stable cell lines
For the generation of stable cell lines, cells were seeded in six-well plates and transfected with donor vectors for the Sleeping Beauty transposase system as described above, using 600 ng of donor vector and 400 ng of pTS395 (P hCMV -SB100X-pA). Selection was done using 1 µg/ml of puromycin (Invivogen, #ant-pr) for at least 10 days.
HEK-293T cells stably expressing mammalian Piezo1 were generated by transfecting pTS391 and pTS395 while HEK-293T and CHO-K1 cells carrying the NFAT-dependent nLuc reporter, NFATc1 expression cassette and constitutive expression of NLuc were generated using pTS2258 and pTS395.

| Assessment of reporter gene expression
Activity of the reporter protein, secreted alkaline placental phosphatase (SEAP), was measured as previously described (Berger et al., 1988). In short, supernatant from cell culture experiments was heattreated for 30 min at 65°C to inactivate endogenous phosphatases.
Approximately 20 µl of heat-treated supernatant was mixed with 60 µl of water and 80 µl of two-times assay buffer (1 M diethanolamine pH 9.8, 0.5 mM MgCl 2 , and 10 mM L-homoarginine). The reaction was started by the addition of 20 µl substrate solution (120 mM p-nitrophenyl phosphate in 2x assay buffer). Reporter activity was assessed by measuring the reaction kinetics at 37°C, by following substrate turnover at 405 nm with an Infinite M1000 microplate reader (Tecan Trading AG). Vector for stable integration of three expression cassettes flanked by insertion and recognition sites of Sleeping Beauty transposase (SB). The first cassette carries a reporter construct for NFAT-induced gene expression of cytosolic nLuc reporter coupled to blue fluorescent protein (mTagBFP2). The second cassette contains infrared-fluorescent reporter protein (iRFP) coupled to a resistance gene for puromycin (puroR) driven by a synthetic RPBSA promoter (P RBSA ) and an ORF encoding nuclear factor and activator of transcription c1 (NFATc1) driven by an internal ribosome entry site (IRES). The third cassette enables expression of cytosolic firefly luciferase (fLuc) reporter for internal normalization of nLuc expression from a constitutive phosphoglycerate kinase promoter (P PGK ).

| Statistical analysis
No statistical analysis was performed. Data are presented as bar graphs representing means ± SD of N = 3 or N = 6 biologically independent samples.   F I G U R E 1 (a) Schematic representation of calcium signaling via the endogenous NFAT signaling pathway. Mechanosensitive channels such as Piezo1 in the plasma membrane trigger elevate cytosolic calcium, activating the trimeric calcineurin A (CnA)-calcineurin B (CnB)-calmodulin (CaM) complex. The activated complex binds to and dephosphorylates phosphorylated NFAT transcription factor through the phosphatase activity of CnA. Dephosphorylated NFAT can enter the nucleus and initiate transcription from NFAT-responsive promoters. Additional channels such as calcium releaseactivated channel (CRAC) or store-operated channels (SOC) that would amplify the calcium signal are not depicted for the sake of clarity. (b) Shear stressinducing turbulent flow was generated by a 3D-printed 96-piston array attached to the membrane of a speaker controlled by a hi-fi amplifier-enhanced signal produced by an Arduino-based signal generator. The up-and-down movement of pistons inside the wells induces turbulent flow, subjecting the cells to shear stress (left insert). Shear stress-induced gene expression is assessed in terms of reporter gene expression (right insert). (c) Various frequencies and timings of shear stress were assessed for their effect on gene expression. To account for unspecific effects, a dual-reporter setup was introduced using NFAT-responsive expression of secreted NanoLuc luciferase (NLuc) and constitutive expression of SEAP from a weak constitutive promoter derived from the simian virus 40 (P SV40 ). By normalizing NLuc activity to SEAP expression, a more robust signal was generated. A clear dosedependent increase of NFAT activity was seen in HEK-293T cells upon stimulation with the piston device. Induction was done at intervals of 5 s every 5 min at the indicated frequencies for 24 h and compared to corresponding uninduced samples. and, hence, NFAT-mediated gene expression. Transfecting cells with Piezo1 yielded higher overall reporter gene expression but lower fold switching compared with wild-type cells (Figure 1d). In addition, Piezo1-transfected cells showed higher sensitivity than wild-type cells, reaching their maximal induction at 30 Hz compared to wildtype cells which were not saturated at 60 Hz.

| RESULTS AND DISCUSSION
As we found the expression of Piezo1 to be slow and heterogeneous within a transiently transfected cell population, we next selected a HEK-293T cell line stably expressing Piezo1 (HEK-293T Pz1 ). We speculated that cell-cell adherence might be a major contributor to improved mechanosensitivity, and we, therefore, used this cell line both to confirm the ability of the prolonged preinduction cultivation used in previous experiments to boost induction fold and to test different induction protocols using induction times of 5 s every 5 min, 2.5 min or 1 min at 30 Hz (corresponding to duty cycles of 1.7%, 3.3%, and 7.7%, respectively). We found the inducibility of the Piezo1-expressing cell line on Day 2 to be superior. Furthermore, higher duty cycles promoted higher fold switching in engineered as well as wild-type cells ( Figure 1e). As the expression of Piezo1 is thought to be constant over time, we speculate that the increased responsiveness on Day 2 points to the possible involvement of cellular adhesion complexes as well as extracellular matrix components that require more time to develop (Jiang et al., 2021).
We next assessed how the expression of key mediators of the NFAT response, NFATc1, and NFATc2, affects shear stress-induced activation of the NFAT pathway. We found that especially overexpression of NFATc1 improves fold switches of HEK-293T cells in response to shear stress induction in a dose-dependent manner in HEK-293T cells (Figure 2a).
To simplify the experimental design, we created a reporter construct that contains the already established NFAT-reporter cassette driving NLuc expression as well as a constitutively expressed firefly luciferase driven by a phosphoglycerate kinase promoter (P PGK ), replacing the SEAP reporter used in previous assays. We also added to the construct a P RPBSA -driven selection cassette that expresses a polycistronic RNA comprising an infrared-fluorescent protein (iRFP) coupled to a puromycin resistance gene (puroR) via a p2a sequence but also contains the CDS for NFATc1, which is driven by an internal ribosome entry site (IRES) (1:P NFAT4 -nLuc-P2A-mTagBFP2-pA-2:P RPBSA -iRFP-P2A-PuroR-IRES-NFATc1-pA-3:P PGK -fLuc-pA). By using cytosolic versions of the luciferase reporters, we aimed to reduce dependency on the secretory pathway to avoid artifacts caused by the expression of multiple membrane proteins.
We used a so generated polyclonal stable HEK-293T cell line to assess the influence of extracellular calcium on the activity of the NFAT pathway. As standard Dulbecco's modified Eagle's medium culture medium contains 1.8 mM calcium, we tested medium containing 1.8, 5, or 10 mM calcium. We found that increased calcium By employing Piezo1 and MscL we could demonstrate that our method can be used to assess mechanoreceptor-mediated activation of gene expression. Additionally, plasma membrane channels involved in calcium signaling such as the L-type calcium channel Ca V 1.2 have already been used to activate gene expression in combination with electrical stimulation or depolarization of cells to treat diabetes and chronic pain (Krawczyk et al., 2020;Wang et al., 2018;Xie et al., 2016). Other potential actuators such as ER-channels of the ryanoidin receptor (RyR) family have yet to be explored in this regard but could be speculated to synergize with that of Ca V . However, because of the sophisticated architecture of Ca V channels and similar ion channels that involve multiple large subunits (Zamponi et al., 2015), studying their responses by means of transient expression might be challenging. Alternative reporters might be used to assess the contributions of alternative signaling pathways in this setup, for example, cytoskeleton-mediated pathways as well (Kirby & Lammerding 2018;Zhang et al., 2015). herence, which in turn leads to better transmission of mechanical forces (Jiang et al., 2021). In accordance with this idea, we found that the density, and hence elasticity, of the plate coating, has a major impact on shear stress responses, especially in the case of CHO-K1 cells. This dependency on ECM might also be indicative of a potential mechanism of activation for MscL and Piezo1 receptors that should be taken into account in future experiments. In this regard, the dependency of (a) (c) F I G U R E 2 (a) NFATc1 and NFATc2 were tested for enhanced shear stress-induced gene expression as single components or in combination (NFAT1/2) using 5 s stimulation each 2.5 min or 5 s stimulation each 0.5 min (3.3% or 14.3% duty cycle, respectively). To improve the robustness of the system and to streamline future experiments, stable cell lines were generated by designing a vector construct comprising NFAT-responsive NLuc expression and constitutive expression of firefly luciferase (FLuc) alongside constitutive expression of NFAT1c. (b) Stable cell lines were used to compare the responses to KCl induction or shear stress-mediated gene expression at different concentrations of extracellular calcium. Stable (c) HEK-293T or d) CHO-K1 cell lines were grown on Matrigel or Geltrex to simulate different extracellular matrix (ECM) compositions and densities as well as to test the effects of additional expression of mechanoreceptors (ii) MscL or (iii) Piezo1. All values are mean ± SD, N = 3 biologically independent samples. NFAT, nuclear factor of activated T-cells; SD, standard deviation STRITTMATTER ET AL.
We believe that the method described here for evaluating shear stress-mediated gene expression has potential applications beyond basic research. For example, the method could be applied for drug screenings targeting mechanosensitive signaling pathways and receptors to provide a more holistic model that includes the underlying signaling cascades on top of calcium transients. Furthermore, mechanosensitive circuits could not only be used in combination with exogenous stimulation but could also be employed to sense physiological parameters such as flow rate inside the vasculature. For example, in next-generation medicine, cells that contain mechanosensitive gene circuits might be employed to detect aberrant levels of blood pressure in the context of cardiovascular diseases, or as a part of stent implants to report on or even to counteract harmful occlusion events.