Compressive Stress Enhances Invasive Phenotype of Breast Cancer Cells via Piezo1 Activation

Uncontrolled growth in solid tumor generates compressive stress that drives cancer cells into invasive phenotypes, but little is known about how such stress affects the invasion and matrix degradation of cancer cells and the underlying mechanisms. Here we show that compressive stress enhanced invasion and matrix degradation of breast cancer cells. We further identified Piezo1 as the putative mechanosensitive cellular component that transmits compressive stress to induce calcium influx, which in turn activate Src/ERK signaling. Interestingly, we observed actin protrusions with matrix degradation ability on the apical side of the cells. Furthermore, we demonstrate that Piezo1 channels were partially localized in caveolae, and reduction of caveolin-1 expression or disruption of caveolae with methyl-β-cyclodextrin led to not only reduced Piezo1 expression but also attenuation of the invasive phenotypes promoted by compressive stress. Taken together, our data indicate that mechanical compressive stress activates Piezo1 channels to mediate enhanced cancer cell invasion and matrix degradation that may be a critical mechanotransduction pathway during, and potentially a novel therapeutic target for, breast cancer metastasis.


Introduction
Cancer invasion is a cumulative result of multiple processes including directed cell migration and extracellular matrix (ECM) degradation.While biochemical factors are well known in mediating cancer invasion, mechanical forces such as compressive stress have also been identified as essential regulators of these processes 1 .For example, an increase of compressive stress inside solid tumor accompanies with cell proliferation and stiffness enhancement 2 .Cancer cells also experience compressive stress when they migrate through capillary and confined tissue microenvironment 3,4 .
Recent in vivo studies show that compressive stress stimulates tumorigenic signaling in colon epithelial cells 5 , and strategies to release compressive stress can indeed enhance the efficiency of anti-tumor treatment 6 .Interestingly, it is demonstrated in vitro that compressive stress directly alters cancer cell proliferation and migration 7,8 and thus drives them to invasive phenotypes 9 .However, whether compressive stress enhances invasive phenotypes by promoting matrix degradation of cancer cells and how compressive stress is sensed and transduced into cellular behaviors is still poorly understood.
Considering that compressive stress stretches cell membrane and thus increases membrane tension, it may alter cancer cells' behaviors through tension-mediated conformational changes of proteins and lipids in the membrane 10 .In particular, the increase of membrane tension can activate several stretch-activated ion channels (SAC) including Piezo and transient receptor potential (TRP) channels [11][12][13][14] .Comparing to TRP channels, Piezo channels are known to respond to membrane tension with more exquisite sensitivity 15,16 , and can be activated via cytoskeleton-mediated forces 17,18   .Moreover, studies in vivo show that Piezo channels mediate diverse physiological activities that are associated with compressive stimulation including touch perception 11 and blood pressure sensing 19 , and are essential in some mechanically related pathological processes such as breast cancer development 20 .In the case of breast cancer development, the role of Piezo1 is substantiated by the shorter survival times of patients with upregulated Piezo1 mRNA expression level 20 .More importantly, it has been shown that the breast cancer cells' response to compression is dependent on Piezo but not TRP channels 21 .Upon the activation of Piezo1 channels, calcium influx evokes several downstream signaling pathways such as Src and ERK which in turn affect dynamics of actin-based structures such as invadopodia/invadosomes 22 that degrade extracellular matrix proteins and thus promote invasion 23 .These data indicate that Piezo1 may be essential for the compression-enhanced cancer invasion.However, whether and how Piezo1 channels mediate compression-enhanced invasive phenotype of cancer cells has not been examined.So far it is generally known that SAC functions at "membrane force foci" such as caveolae 24 .This is because the cholesterol-enriched flask-like membrane invaginations of caveolae may provide proper platforms for harboring and gating SAC [25][26][27] , when caveolae rapidly flatten and disassemble in response to an increase in membrane tension 28,29 .As for Piezo1, structural analysis has shown that there is a pocket sandwiched between Piezo1 repeat B and C, which provides a binding site as a means of interaction with the lipids 14 .Despite such evidence of structure for lipid interaction, the localization of Piezo1 is not well established and thus it remains unclear whether its activity is regulated by caveolae.
In this study, we hypothesized that Piezo1 channels mediate the compressionenhanced invasive phenotype of cancer cells.To test this hypothesis, we examined in vitro cultured human breast cancer cells for their ability to invade and degrade extracellular matrix in the presence or absence of compressive stress, together with corresponding changes in Piezo1 and calcium signaling.We found that compressive stress promoted an invasive phenotype in the breast cancer cells, characterized by enhanced matrix degradation, actin protrusion formation, and calcium signal initiation.
More importantly, the phenotypic changes in these cells appeared to be mediated by compression-induced Piezo1 activation, which in turn was dependent on the integrity of caveolae.

Piezo1
To test whether compressive stress enhances invasion of breast cancer cells, MDA-MB-231 cells were grown on 2D membrane filter (8 μm pore) coated with Matrigel and covered with a thin 1% agarose gel pad, and then exposed to a constant weight (Figure 1a).The compressive stress levels used in this study were 200, 400, and 600 Pa, which were considered pathophysiologically relevant as cells are reported to experience compressive stress at up to about 800 Pa in the core of the solid breast tumor 9,30 .As shown in Figure 1b and 1c, there were more MDA-MB-231 cells that invaded through the Matrigel-coated transwell filters when exposed to compressive stress compared to their un-compressed counterparts (control).The results clearly show that compressive stress enhanced breast cancer cell invasion in a pressuredependent manner.
In order to test whether the compression-enhanced cancer cell invasion was mediated through SAC or more specifically through Piezo1, we first pretreated MDA-MB-231 cells with SAC inhibitors Gd 3+ or more specific inhibitor GsMTx4, separately, followed by exposure to 400 Pa compressive stress.As shown in Figure 1d and 1e, pretreatment with Gd 3+ or GsMTx4 either partially attenuated or completely abrogated compression-enhanced cancer cell invasion.
To further confirm the specificity of Piezo1 in mediating compression-enhanced cancer cell invasion, we first examined the expression of Piezo1 in MDA-MB-231 cells.
We found that Piezo1 was expressed in MDA-MB-231 cells in the form of puncta structures and located not only on the plasma membrane, but also over the intracellular space and nucleus (Figure S1a), consistent with data reported by Gudipaty et al. 31 .We then silenced the protein expression of Piezo1 in MDA-MB-231 cells by using siRNA.
Western blot results confirmed that the efficiency of Piezo1 knockdown (KD) was ~70% (Figure S1b).When MDA-MB-231 cells pretreated with Piezo1 siRNA were exposed to 400 Pa compressive stress, the enhanced cell invasion in response to compression was completely abrogated (Figure 1f and 1g).This supports Piezo1's function in compression-enhanced cancer cell invasion.

Compressive stress enhanced matrix degradation dependent on Piezo1
From the results thus far, we suspect compressive stress may influence cancer cells' capability for matrix degradation.To investigate this, we examined the extent of matrix degradation of MDA-MB-231 cells seeded on FITC-conjugated gelatin-coated glass bottom dish followed by application of compressive stress (Figure 2a).The fluorescence images showed dark puncta areas, corresponding to "holes" formed in the gelatin matrix due to degradation (Figure 2b).Thus, we quantified the extent of matrix degradation, and the results showed that MDA-MB-231 cells exposed to compressive stress from 200 Pa to 600 Pa exhibited a significant and pressuredependent increase of gelatin matrix degradation as compared to their un-compressed counterparts (0 Pa) (Figure 2c).We sometimes found that the holes in the gelatin matrix do not coincide with the cell areas.We suspect this is due to the movement of the cells during the course of compression, since our fluorescent gelatin coating is very uniform in control conditions without cells.Similar as in the case of cell invasion through Matrigel-coated transwell filters, pretreatment of MDA-MB-231 cells with GsMTx4 to inhibit Piezo1 or siRNA probe to silence Piezo1 expression completely abrogated the compression-enhanced gelatin matrix degradation in breast cancer cells (Figure 2c).These data indicate that compressive stress enhanced the matrix degradation capability of breast cancer cells in a Piezo1 dependent manner.
Cancer cells are known to use invadopodia formed on the membrane to promote ECM degradation 32,22,33,34 .Thus, we examined whether compressive stress could promote invadopodia formation in MDA-MB-231 cells.We used immunofluorescencelabeled cortactin, a marker for invadopodia 35 , to visualize and identify invadopodia as cortactin-positive actin puncta on the cell membrane (Figure S2a).The number of invadopodia per cell was manually counted and reported for MDA-MB-231 cells with or without pretreatment with siRNA probe to silence Piezo1, respectively, and with or without exposed compressive stress.The results show that compressive stress increased the number of invadopodia per cell in MDA-MB-231 cells, which was significantly abrogated by silencing Piezo1 (Figure S2b).Although more detailed characterization may be needed to definitively confirm the identity of these protrusions as invadopodia, these results demonstrate that breast cancer cells responded to compressive stress with increased number of actin protrusions that may be responsible for ECM degradation.

Piezo1 mediated compression-induced remodeling of F-actin
Cells under compressive stress may reprogram their intracellular mechanical structures such as F-actin to acquire appropriate mechanical properties.To observe the response of F-actin to compressive stress, we generated a stable cell line expressing RFP-Lifeact and monitored the dynamics of F-actin in MDA-MB-231 cells upon exposure to compressive stress.We observed that compressive stress significantly promoted the formation of stress fibers at the cell cortex region in as early as 10 min in single and collective wild type cells (Figure 3a, top panel).In contrast, the compression-enhanced formation of stress fiber was largely inhibited in the Piezo1 KD cells (Figure 3a, bottom panel).Interestingly, by using 3D spinning disk confocal imaging we discovered actin protrusions at the apical side of the MDA-MB-231 cells when exposed to 600 Pa compressive stress (Figure 3b, top panel).The actin protrusions were reduced in the Piezo1 KD cells exposed to the same amount of compressive stress (Figure 3b, bottom panel).
To quantify the size of these actin protrusions, we first analyzed the size of microspheres of 0.1, 0.2, and 0.5 μm diameter, respectively (Figure S3a) and found that measured pixel values in the vertical direction were in a linear relationship with the sizes of these microspheres (Figure S3b).Using this relationship to calibrate the actin protrusion length, we found that the average length of the apical actin protrusions were about 0.7 μm, respectively (Figure S3c).Such apical actin protrusions appeared to be similar to the above-mentioned invadopodia.To verify whether or not these apical actin protrusions might be invadopodia or not, we examined the agarose gel that covered cells during compression using immunofluorescence microscopy.We found that the apical actin protrusions that were stuck in the agarose gel colocalized with cortactin and Tks5 (Figure S3d).Then, we coated the agarose gel with FITC-conjugated gelatin and used it to cover the cells during compression.We found many dark puncta appeared in the gelatin at the apical side of the cells (Figure S3e).These data together strongly suggest that the apical actin protrusions induced by compressive stress were similar, in both structure and function, to the invadopodia commonly found at the ventral side of the cells.

Piezo1 mediated compressive stress-induced calcium signaling
Recently, internal cell-generated forces have been shown to evoke localized Piezo1-dependent calcium flickers 17 .To de te r m i n e w he t he r an externally imposed force can elicit calcium signaling in a Piezo1-dependent manner, we labeled the cells with canonical calcium dye Fluo-4/AM or transiently transfected a calcium biosensor G-GECO and performed live cell imaging during application of compressive stress to MDA-MB-231 cells.As shown in Figure 4a, calcium signaling, as indicated by the fluorescence intensity of Fluo-4, was activated instantaneously upon exposure to compressive stress (Video 1).The peak magnitude of activation (the relative fluorescence intensity of Fluo-4) increased between ~1.3 to ~2.3 fold as the compressive stress increased from 200 Pa to 600 Pa (Figure 4b).These results were confirmed by using G-GECO (Video 2, Figure 4c and d).The peak magnitude of activation also increased from ~1.5 to ~3.5 fold as the compressive stress increased from 200 Pa to 600 Pa.The kinetics of calcium transients were different between the two indicators, likely due to the different affinity for calcium.
The G-GECO system was used to measure calcium signaling in the following experiments, as it is more convenient than the Fluo-4/AM system.To d et e r m i ne i f calcium signaling was due to extracellular or intracellular calcium, we treated cells transfected with G-GECO with 2 mM EGTA for 15 min to deplete extracellular calcium content before application of compressive stress (400 Pa).Under this condition, we found that compressive stress-induced calcium signaling was abolished, suggesting the signaling was due to influx of extracellular calcium (Figure 4e).Furthermore, calcium influx induced by compressive stress (400 Pa) was also abrogated when cells were pretreated with Gd 3+ or GsMTx4 to block Piezo1, or siRNA probe to silence Piezo1 expression (Figure 4f).Together, these observations support the finding that Piezo1 mediated the cellular response to compressive stress via calcium influx.

Piezo1 mediated compression-enhanced Src/ERK activation
During invadopodia formation and maturation to degrade matrix, several signaling pathways are involved including Src/ERK pathways which promote the formation of invadopodia 23 .To test whether these signaling pathways are activated by compressive stress, we quantified the phosphorylation of ERK and Src in MDA-MB-231 cells following compression.We found that compressive stress significantly activated ERK and Src (Figure 5a).Additionally, Piezo1 KD effectively abolished the compressive stress-promoted signaling of Src, but not for ERK (Figure 5b), suggesting that Src, but not ERK was activated by compressive stress in a Piezo1 dependent manner.

Caveolae regulated the location and function of Piezo1
Previous work suggests that cholesterol content that directly influences the formation of caveolae might regulate Piezo1 functions [36][37][38][39] .Since caveolae assembly and disassembly have been tied to changes in membrane tension 29 , we speculated that compressive stress activation of Piezo1 might also be dependent on caveolae.
We first examined the spatial relationship between Piezo1 and caveolae.We found that both Piezo1 and caveolae formed puncta structures and many of them were colocalized (Figure 6a).The coefficient of colocalization in wild type cells was analyzed with Coloc2 procedure in Fiji software as shown in Figure 6b.The results show that classical Pearson coefficient was 0.76 ± 0.13, which indicates that Piezo1 and Cav-1 were highly colocalized.To test whether caveolae regulate the Piezo1 expression, we quantified Piezo1 protein expression level in wild type, Cav-1 EGFP transiently transfected, and Piezo1 KD MDA-MB-231 cells.We found that Cav-1 EGFP expression increased Piezo1 expression while Cav-1 KD decreased Piezo1 expression (Figure 6c).
To verify the role of caveolae in Piezo localization in the cell membrane, MDA-MB-231 was treated with 5 mM of MβCD that dramatically reduced the number of caveolae (Figure S4a).Consequently, the fluorescence intensity of Piezo1 decreased in the cell membrane, but increased in the nucleus at the 5 min time point (Figure 6d), suggesting caveolae regulated Piezo1 localization in MDA-MB-231 cells.
To test the role of caveolae in regulating the function of Piezo1, cells were pretreated with either 5 mM MβCD or siRNA probe which silenced Cav-1 expression by about 60% (Figure S4b) and then exposed to compressive stress.We found that the compressive stress-induced calcium influx was blocked in both conditions (Figure S4c).Consistent with this results, Cav-1 KD also abrogated the compressive stressenhanced cancer cell invasion (Figure S4d and S4e).Altogether, these data further support the idea that Piezo1 function may be dependent on caveolae.

Discussion
In the present study, we showed that in breast cancer cells, compressive stress enhanced cancer cell invasion, matrix degradation, and the formation of stress fiber and actin protrusion.Additionally, we identified that Piezo1 mediated these processes and the invasive phenotypes also depended on the integrity of caveolae.These findings provide the first demonstration that compressive stress enhances matrix degradation by breast cancer cells and that Piezo1 is an essential mechanosensor and transducer for such stress in breast cancer cells.

Invasion of cells through ECM is a critical activity during cancer metastasis.
Uncontrolled growth-induced compressive stress triggers invasive phenotype by forming leader cells 9 .It is, however, unknown whether such stress would affect the capability of cancer cells to induce ECM degradation.It is also known that the ability of forming actin protrusions such as invadopodia is directly correlated with their invasive potential in breast cancer cells 32 .Consistent with enhanced invasion under compressive stress, we found compressive stress enhanced matrix degradation via promoting actin protrusions in both apical and ventral side of breast cancer cells.Thus, it is plausible that compressive stress in solid tumor might initiate the invasion by enhancing the capability of matrix degradation via actin protrusions.If that is the case in vivo, compressive stress might promote cancer cells to 'dig more holes' in the basement membrane which provide a way for their metastasis.
While it is known that compressive stress affects cancer progression, how cancer cells sense and respond to such stress is not completely understood.Under compressive stress, cell membrane will be stretched which, in turn, increases the membrane tension.Piezo channels are the most notable family of SAC in mammalian cells which are gated by membrane tension 40 .It has been found that Piezo1 channels play essential roles in diverse physiological and pathological processes including cell migration 41,42 and the Piezo1 mRNA expression level is highly correlated with the reduced overall survival of breast cancer patients 20 .Our study also confirmed that Piezo1 channels are essential in mediating the invasion of breast cancer cells.
Emerging evidence indicate that caveolae harbor and modulate ion channels.
TRPV1 channels expression and distribution on the plasma membrane can be disrupted by the removal of caveolae via cholesterol depletion 43 .Similar to TRPV1, we found Piezo1 behaves similarly by shifting its localization to the nucleus upon cholesterol depletion.Interestingly, in stretch-triggered mitosis, Piezo1 was also observed to localize to the nuclear envelope 31 .Thus, it may be a general strategy where a functional relationship between caveolae and Piezo1 regulates force sensing and our data supports the model where caveolae might be the "mechanical force foci" which concentrates Piezo1 to facilitate force sensing and transduction in mammalian cells.
The mechanisms of how Piezo1 channels are gated by mechanical stress is still unclear.Piezo1 channels appear to sense tension in the bilayer and are gated according to "force-from-lipid" principle, an evolutionally conserved gating mechanism 44 .According to this paradigm, the activity and sensitivity of Piezo1 channels can be regulated by lipid membrane.The physical properties of lipid membrane such as thickness, stiffness, and lateral pressure profile found within caveolae may be different from the surrounding membrane.For instance, disruption of caveolae by cholesterol depletion has been demonstrated to change membrane stiffness 36 and results in suppression of epithelial sodium channels and TRP channels 38,45 .In this context, it is plausible that cholesterol-enriched caveolae might affect sensitivity of Piezo1 via controlling the membrane pressure profile.Stomatin-like protein-3 has been reported to tune the sensitivity of Piezo1 channels by controlling the membrane mechanical properties through recruiting cholesterol 36,46 .In this study, we found that the function of Piezo1 in compressive stress sensing is regulated by caveolae.Taken together, it is likely that Piezo1 is located in cholesterol-rich caveolae microdomains and caveolae integrity regulates Piezo1 function.
In addition to membrane structures, cells experiencing mechanical stress can also reorganize their cytoskeletal structures to adapt to a new mechanical microenvironment.Among them, stress fibers are essential cellular structures that control various cellular behaviors.Reports have shown that mechanical tension induces the assembly of stress fibers 47 .In this study, we found that cells under compressive stress quickly assembled new stress fibers in as early as 10 min.This may be to increase their cellular mechanical strength to balance compressive stress.
Our results also demonstrate the regulatory roles of several key pathways of mechanotransduction in the compression-enhanced cancer cell invasive phenotype including Src, ERK and calcium, which are pathways linked to the formation and function of actin protrusion such as invadopodia 48,49 .Our study thus provides a comprehensive understanding from disparate systems in the context of compressionenhanced breast cancer cell invasion (Figure 7).
It is worthy to note that, in this study, we only evaluated the effect of uniaxial compressive stress on breast cancer cells in a 2D culture model.However, cells in vivo grow and live in a 3D microenvironment, which may impact the force direction and change the dynamic response of the cells to compression.Additionally, we only assayed cell invasion in response to compressive stress.Many other features of breast cancer cells such as the loss of acini morphologies in response to compression still need to be explored.For instance, Ricca et al. have shown that brief compression to single malignant breast cancer cell in laminin-rich ECM can stimulate the formation of acinar-like structures, indicating that compressive stress may cause malignancy reversion in breast cancer cells 50 .Furthermore, we just investigated the role of Piezo1 in mediating the breast cancer cell response to compressive stress.However, as another family member, Piezo2 has also been reported to express in these cells 51 , raising the question of its potential implication.Therefore, further studies are required in order to fully elucidate the behaviors and associated underlying mechanisms of breast cancer cells in response to compressive stress during tumor growth and metastasis.
In conclusion, our work may have relevance to human tumors in vivo.As solid tumor experiences high compressive stress due to uncontrolled proliferation and confinement by the stiff extracellular matrix environment, this microenvironment facilitates compression-enhanced cell invasion.The identification of Piezo1's crucial role in this process provides the first demonstration of the dependence of Piezo1 channels on the response of breast cancer cells to physiological compressive stress.
Both of these findings underscore the cardinal role that Piezo1 channels play in regulating cell invasion and may inspire further development targeting Piezo1 as a potential cancer therapeutic target.
After 24 h at 37 °C, the transfection medium was replaced with complete medium, and cells were processed after 24-48 h later.

In vitro compression device
To investigate the effect of compressive stress on cell behaviors, we used a previously described setup 9,53 .Briefly, cells were grown either in a 35 mm culture dish with glass bottom (# 12-565-90, Fisher Scientific) that was coated with/without gelatin, or in a transwell chamber with permeable membrane of 8-μm pores that was coated with Matrigel.Then the cells were covered with a soft agarose disk layer, and subsequently a piston of specific weight was placed on top of the agarose disk to apply a given compressive stress to the cells underneath indirectly.The cross-sectional area of the piston (24 mm diameter) was 4.52 cm 2 but its weight was variable at 9.22 g, 18.45 g, and 27.67 g, corresponding to a stress of 200 Pa, 400 Pa, and 600 Pa, respectively, on the cells.Cells prepared as such but not subjected to piston weight were used as controls.

In vitro transwell invasion assay
To assay the effect of compressive stress on cell invasion, standard transwell invasion assay adapted from Bravo-Cordero 54,55 was performed using 6-well Transwell chambers that were separated as upper and lower chambers by filter membrane with 8 μm pores (# 07-200-169, Corning).For the assay, the transwell filter membrane was coated with 300 µl Matrigel (12 mg/ml, # E1270, Sigma, Burlington, MA) diluted in serum-free DMEM (2 mg/ml final concentration), followed by incubation for 1 h at 37 °C.MDA-MB-231 cells in serum-free medium (5×10 5 cells/well) were placed in the upper chamber, while the lower chamber was filled with 2 ml complete medium.Cells were allowed to grow for 6 h and then compressed for 18 h before being fixed with 4% paraformaldehyde (# 30525-89-4, Electron Microscopy Sciences, Hatfield, PA).The non-invasive cells on the upper chamber were removed with cotton swabs, and the invaded cells in the lower chamber were stained with 0.1% crystal violet (# C6158; Sigma) for 10 min at room temperature, prior to being examined and imaged by light microscopy at 10× magnification (Olympus BX60; Olympus Corporation, Tokyo, Japan).
Then the number of stained cells in the lower chamber was counted using ImageJ software (National Institute of Health, Bethesda, MD) and the enhancement of cellular invasion induced by compressive stress was quantified as percentage (%) of the number of compressed cells over that of the non-compressed cells that had invaded through the filter membrane, i.e. [# of cells in the lower chamber in the presence of a specific weight (experiment group)]/[# of cells in the lower chamber in the absence of a specific weight (control group)].Results are based on the analysis of 10 random fields per transwell in each condition and each experiment was repeated three times.

Evaluation of invadopodia formation and ECM degradation
To determine whether compressive stress enhances cells' ability to degrade ECM, we examined cells cultured on gelatin substrate for their tendency to form invadopodia and associated gelatin degradation, according to a protocol adapted from Artym et al.

56
. Briefly, glass bottom dishes were treated with 20% nitric acid for 1 h, washed with H2O for 4 times, then incubated with 50 μg/ml poly-L-lysine (# P8920, Sigma) in PBS for 15 min and washed with PBS, then further incubated with 0.5% glutaraldehyde in PBS on ice for 15 min followed by thorough washes with PBS.Subsequently, the dishes were coated with 1 ml of gelatin in PBS (1: 9 of 0.1% fluorescein isothiocyanate (FITC)-gelatin (# G13186, Invitrogen, MA): 2% porcine gelatin), then washed in PBS, incubated with 5 mg/ml sodium borohydride (NaBH4) for 3 min, rinsed in PBS, and then incubated in 10% calf serum/DMEM at 37° for 2 h.Afterwards, cells were seeded in each dish at 5x10 5 cells per well and incubated for 8 h, and then subjected to compressive stress of either 200 Pa, 400 Pa, or 600 Pa, respectively, for 8 h as aforementioned.
Upon completion of compression, the cells were imaged with live fluorescence microscopy (at 60x) and the microscopic images were analyzed by using ImageJ to assess the formation of invadopodia and the degradation of gelatin matrix.
Invadopodia were defined as F-actin-positive puncta protruding from the cells into the gelatin matrix underneath the cell in our experiments 57 .For each independent experiment that was performed in triplicates, the number of invadopodia per cell was quantified with cells imaged randomly in >15 microscope view fields, representing a total of ~100 cells per experimental condition.At the same time, the degradation of gelatin matrix was quantified as the percentage of degraded area (dark spots comprised of dense degraded protein products) in the whole area underneath each cell.

Live fluorescence microscopy
To observe the dynamics of actin, Cav-1, and calcium signaling, live cells expressing Lifeact-RFP, Cav-1-EGFP, and G-GECO were imaged with a spinning disk confocal microscope with a 60x or 100x oil immersion objective (Olympus IX73 with Yokogawa CSU-X1).For live fluorescence microscopy, cells were seeded in a 35 mm glass bottom dish that was placed in an environmental chamber mounted on the microscope to maintain constant 37 °C, 5% CO2, and humidity.Cav-1-EGFP was observed at the excitation wavelength of 488 nm.For dynamic tracking of actin in live cells, the cells were consecutively imaged for up to 60 min, and the images were processed using ImageJ.Cells were observed from both top-down and side view for spatial localization of actin, and caveolae by 3D reconstruction of images in Z-stacks (0.4 μm increments for 60x objective and 0.2 μm increments for 100x objective).
To quantify the size of the apical actin protrusion, microspheres of different diameters (0.1, 0.2 and 0.5 μm; cat # F8801, F8810, F8812, ThermoFisher, respectively) were laid flat on the glass cover slip and imaged with the abovementioned methods for imaging actin.We then analyzed the relationship between the measured pixel size in the vertical direction of these microspheres and the actual diameters of these microspheres.We made sure that a linear relationship was established.Then, we used this linear relationship to analyze the length of the upward actin protrusions based on the pixel size of the imaged actin protrusions.

Intracellular Ca 2+ measurement
To evaluate the intracellular calcium concentration ([Ca 2+ ]), we used cells labeled with Fluo-4/AM (# F14201, Thermo Fisher) or transiently expressed with calciumsensitive reporter G-GECO 58 and then evaluated the intensity of intracellular calcium signaling.For the Fluo-4/AM system, cells were incubated with Fluo-4/AM for 1 h at room temperature (25 ± 2 °C) followed by a 0.5 h wash at 37 °C.For the G-GECO system, cells transfected with G-GECO for 48 h were plated into a glass bottom dish, which was further incubated for 24 h.Subsequently, the cells were imaged on a spinning disk confocal microscope (60 x objective), with fluorescence excitation and emission at 488 nm and 533 nm, respectively.For each experimental group, twenty cells were randomly selected and the fluorescence intensity per cell was quantified using ImageJ.

RNA interference
To silence the expression of Piezo1 and Cav-1, Negative Control Medium GC Duplex #2 and siRNA interference for Piezo1 (# AM16708, Assay ID:138387, Thermo Fisher) and Cav-1 (# AM16708, Assay ID: 10297, Thermo Fisher) were used.Briefly, Cells were seeded in 6-well plates at 1 × 10 6 cells/well for 24 h prior to transfection.At 90% confluence the cells were transfected with 30 nmol/L siRNA using Lipofectamine RNAi MAX (# 13778, Invitrogen, MA) in OptiMEM according to the manufacturer's instructions.Transfection mixes were applied to the cells for 24 h, subsequently removed and replaced with 2 ml of growth media.The cells were cultured for 48 h before use in experiments.The protein expression levels of Piezo1 and Cav-1 were ascertained by Western blotting.

Immunofluorescence and colocalization analysis
Cells were fixed with 4% paraformaldehyde for 10 min and permeabilized with 0.1 % TritonX-100 for 10 min at room temperature.Non-specific sites were blocked using 5 % non-fat milk in PBS for 1 h at room temperature.Cells were then incubated in 5 % nonfat milk in PBS containing primary antibodies at 1:100 dilution for 1 h at room temperature.After washing with PBS, cells were incubated with Alexa Fluor 594 or 640 conjugated secondary antibody for 60 min at room temperature.Cells were visualized using the spinning disk confocal microscope with 60x oil immersion objective.For Factin staining, cells were incubated with 1:100 rhodamine phalloidin (# PHDR1, Cytoskeleton Inc.) for 60 min at room temperature.
Colocalization of Piezo1 and Cav-1 was analyzed using Fiji software 59 containing a procedure for colocalization analysis, designated as Coloc2, which are based on pixel-intensity-correlation measurements.Pearson coefficient and 2D intensity histograms were recorded to quantify the degree of the colocalization between Piezo1 and Cav-1.

Western blotting
Western blot assay was used to examine the protein expression and/or activity of Piezo1, Cav-1, Src, and ERK.Cells grown on glass bottom dishes under described assay conditions were lysed using RIPA buffer (# R0278, Sigma) with added cocktail of protease and phosphatase inhibitors (MS-SAFE, Sigma).Protein concentration of cell lysates were determined using the Protein Assay Reagent (#23227, Thermo Fisher Scientific, Waltham, MA).Cell lysis buffer was combined in 4× SDS sample buffer and 2-mercaptoethanol and incubated at 95 °C for 5 min.After loading equal amount of protein per lane, SDS-PAGE was performed.The proteins were transferred onto 0.22 µm nitrocellulose membranes (# 66485, Pall Life Sciences) using Pierce G2 Fast Blotter (Thermo Fisher Scientific, Waltham, MA).Following transfer, the membranes were blocked using 5% nonfat milk in 1x TBST (Tris-buffered saline and 0.1% of Tween-20) for 1 h at RT with gentle agitation and incubated with the primary antibodies overnight at 4 °C under mild shaking condition.After washing three time with 1x TBST, membranes were incubated with goat anti-rabbit secondary antibody (DyLight 800, # SA5-10036, Thermo Fisher) or goat anti-mouse secondary antibody (DyLight 680, # 35518, Thermo Fisher) at RT for 1 h.Signal of immunoblots were detected using the       The opening of Piezo1 mediates the influx of calcium and evokes downstream signaling pathways such as Src.These activated signaling molecules promote actin protrusions both at the apical and ventral sides of cells and the remodeling of actin cytoskeleton, which in turn mediate enhanced matrix degradation and cell invasion.and evokes downstream signaling pathways such as Src.These activated signaling molecules promote actin protrusions both at the apical and ventral sides of cells and the remodeling of actin cytoskeleton, which in turn mediate enhanced matrix degradation and cell invasion.
For matrix degradation and invadopodia experiments, cells were incubated in invadopodia medium containing DMEM supplement with 5% Nu-Serum (# 355104, Corning, NY), 10% FBS, and 20 ng/ml EGF.For labeling actin in live cells, stable cell lines expressing Lifeact-RFP were generated via lentiviral transfection.The lentiviral transfer plasmids pLVX-puro-GFP-Lifeact and pLVX-puro-RFP-Lifeact were cloned from RFP-Lifeact plasmid obtained from Dr. Gaudenz Danuser (UT-Southwestern).Briefly, lentiviruses were produced by transfecting human embryonic kidney (HEK)-293T cells with psPAX2 and pMD2.G (Addgene) and pLVX-puro-GFP-Lifeact viral vectors.Conditioned medium containing viruses was collected after 5 days and then used immediately to infect cells or stored at −80 °C.Transduced target cells were selected with puromycin for 72 h.For optical imaging of dynamic calcium signaling and caveolae localization in live cells, cell lines transiently expressing G-GECO (a green fluorescent genetically encoded calcium indicator) and caveolin-1 (Cav-1)-EGFP respectively were generated via plasmid transfection.The plasmids expressing G-GECO was a generous gift from

Figure captions Figure 1
Figure captions

Figure 2
Figure 2 Compressive stress promoted matrix degradation in MDA-MB-231 cells.a Schematic diagram of the experiment.Cells grown on a glass-bottom dish coated with FITC-conjugated gelatin for 8 h were covered with 1% of agarose gel and compressed with a specific weight.b Representative images (red: actin, green: gelatin) of compression-promoted gelatin degradation at the ventral side of the cell.Gelatin degradation was visualized by confocal microscopy as disappearance of green fluorescence.The arrow in 200 Pa shows an example of the punctate gelatin-degraded region.Insets of 600 Pa images are magnified view of the boxed regions.c The fold change of gelatin degradation area under different treatment conditions as a function of compressive stress normalized to gelatin degradation area at 0 Pa.** p < 0.01 versus control groups.Error bars represent s.e.m..

Figure 3
Figure 3 Piezo1 mediated compressive stress-induced formation of stress fiber and actin protrusions.a Representative images of stress fiber at the ventral side of MDA-

Figure 4
Figure 4 Compressive stress induced calcium signaling in MDA-MB-231 cells.Representative images of intracellular [Ca 2+ ] (a and c) and average time-courses of changing relative mean fluorescence intensity (b and d) of Fluo-4 or G-GECO (normalized to time 0) in MDA-MB-231 cells labeled with Fluo-4/AM or transiently expressing G-GECO before (0 min) and after (1 min) exposure to compressive stress at 200, 400, 600 Pa, respectively.e, f Time-courses of changing relative mean fluorescence intensity of G-GECO in MDA-MB-231 cells pretreated with or without EGTA, Gd 3+ , GsMTx4, and Piezo1 KD in response to 400 Pa compressive stress.Each experiment assayed 10-20 cells and repeated three times.

Figure 5
Figure 5 Compression enhanced the activity of ERK and Src. a The phosphorylation of ERK and Src in MDA-MB-231 cells pretreated with scramble probes in the absence or presence of compressive stress at 200, 400, 600 Pa.b The phosphorylation of ERK and Src in MDA-MB-231 cells pretreated with siRNA for Piezo1 in the absence or presence of compressive stress at 200, 400, 600 Pa.Relative phosphorylation levels were obtained by normalizing to GAPDH expression and 0 Pa value, n = 3. * p < 0.05 versus control groups; ** p < 0.01 versus control groups.Error bars represent s.e.m..

Figure 6
Figure 6 The expression and distribution of Piezo1 in MDA-MB-231 cells were regulated by caveolae.a Representative fluorescence images of Piezo1 (magenta) and caveolae (green) colocalization and 2D intensity histogram output in MDA-MB-231 cells.Insets in both conditions show magnified view of the boxed regions.b Representative image of 2D intensity histogram output of Coloc2 analysis performed

Figure 7
Figure 7 Model of compressive stress-promoted invasive phenotype of MDA-MB-231 cells and associated signaling pathways.Tog e th er, v ertical mechanical compressive stress increases the lateral plasma membrane tension and activates Piezo1 channels.

Figure S1
Figure S1The distribution and the efficiency of siRNA knockdown (KD) of Piezo1 in MDA-MB-

Figure S3 e
Figure S3The characterization of upward actin protrusion.a Representative confocal

Figure S4
Figure S4 Caveolae regulate the function of Piezo1 in responding to compressive stress in MDA-

Figures Figure 1
Figures

Figure 6 The
Figure 6

Figure 7 Model
Figure 7