Piezo2 interacts with E-cadherin in specialized gastrointestinal epithelial mechanoreceptors

Epithelial cells form barriers between outside and inside. In many organs, the epithelial layer is critical to sense luminal mechanical events, such as filling (Marshall et al., 2020) and shear stress (Wang et al., 2016; Zeng et al., 2018). Specialized mechanoreceptors are mechanosensitive cells that interpret forces at the barrier and initiate downstream physiologic responses (Mercado-Perez and Beyder, 2022; Ranade et al., 2014; Woo et al., 2014; Fettiplace, 2017; Fettiplace et al., 2022; Lembrechts et al., 2012; Nonomura et al., 2017). Epithelial mechanoreceptors are tightly embedded into multicellular epithelial layers stitched from single cells using junctional proteins, such as E-cadherin. These junctional complexes can also participate in force sensation and transmission (Lecuit and Yap, 2015), potentially contributing to the function of the specialized mechanoreceptors.

Mechanical cues in the gastrointestinal (GI) tract range from exogenous, such as meals filling the stomach (Umans and Liberles, 2018), to endogenous, like shear stress exerted by moving contents (Beyder, 2019; Mercado-Perez and Beyder, 2022). Mechanical activity in the gut lumen, through stimulation of the mucosal layer, leads to the release of serotonin (5-hydroxytryptamine, 5-HT) (Bulbring and Crema, 1959) via a mechanosensitive subpopulation of 5-HT-releasing cells known as enterochromaffin (EC) cells (Wang et al., 2017; Alcaino et al., 2018). These belong to the wider population of epithelial mechanosensitive enteroendocrine cells (EECs) which release a range of signaling molecules that modulate gut function (Billing et al., 2019; Beumer et al., 2020).

Mechanosensitive EECs are characterized by the expression of Piezo2, which they require to convert mechanical stimuli into signaling molecule release (Alcaino et al., 2018; Billing et al., 2019). Piezo1 and 2 are mechanically gated non-selective cation channels (Coste et al., 2010). Piezo2 specifically may function as a multimodal molecular force sensor. Piezo2 currents are sensitive to lipid membrane tension (Haselwandter et al., 2022; Yang et al., 2022). Additionally, disruption of the actin cytoskeleton with cytochalasin D or latrunculin A decreases force-dependent Piezo2 currents in cultured cell models and dorsal root ganglia neurons (Eijkelkamp et al., 2013; Jia et al., 2016; Verkest et al., 2022). While the molecular details of Piezo2 tethering remain to be elucidated, both intra- and extracellular subdomains likely mediate its cytoskeletal sensitivity. Deletion of the intracellular-facing intrinsically disordered region 5 (IDR5) from Piezo2 eliminates the channel’s sensitivity to cytochalasin-induced actin depolymerization (Verkest et al., 2022). The extracellular Piezo2 cap domain modulates inactivation kinetics (Lewis and Grandl, 2020) and may interact with the ectodomain of the cadherin protein superfamily (Wang et al., 2022). Knockdown of E-cadherin via siRNA decreases mechanically induced ionic currents in an immortalized cell model of Piezo1 currents (Wang et al., 2022). Co-transfection of Piezo2 with E-cadherin in immortalized cell models increases mechanical currents without affecting inactivation kinetics (Wang et al., 2022), but this interaction is not known outside of cultured cell models.

In epithelial cells, cadherins are integral membrane proteins that mediate physical cell–cell adhesion. E-cadherin is expressed by epithelial cells, where it localizes to the lateral plasma membrane and mediates adherens-type junctions between epithelial cells (Lecuit and Yap, 2015). E-cadherin is part of a mechanosensory complex on the epithelial cells’ lateral wall that includes scaffold proteins to anchor intracellularly to the actin cytoskeleton (Lecuit and Yap, 2015). The catenin family of molecules binds to the intracellular tail of E-cadherin (Ozawa et al., 1989; Reynolds et al., 1994), from where it can stabilize E-cadherin (Ireton et al., 2002; Davis et al., 2003), regulate cytoskeletal reorganization (Noren et al., 2000), and facilitate force-dependent tethering to the actin cytoskeleton. The E-cadherin–β/γ-catenin–α-catenin–actin filament chain tethers E-cadherin to the actin cytoskeleton (Sako et al, 1998). Vinculin is a filamentous actin-binding protein that is recruited after tension unfolds the α-catenin molecule (Choi et al., 2012; Yao et al., 2014).

In this study, we hypothesized that Piezo2 interactions with E-cadherin are important for mechanosensation in GI EECs. To test this hypothesis, we set out to visualize Piezo2 co-localization with E-cadherin and assay functional activity in specialized GI epithelial mechanoreceptors, the EECs.

All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the Mayo Clinic.

All materials used for experiments, including antibodies employed in immunofluorescence and co-immunoprecipitation studies, are listed in Table 1.

NeuroD1-Cre mice were a kind gift of Dr. Andrew Leiter (University of Massachusetts, Worcester, MA, USA, since then submitted to The Jackson Laboratories: Jax 028364). Piezo2-Cre and Piezo2^f/f mice were kind gifts of Dr. Ardem Patapoutian (Scripps Research Institute, San Diego, CA, USA, since then submitted to The Jackson Laboratories: Jax 027719, 027720). Mouse lines obtained from The Jackson Laboratories: wild type C57BL/6J (Jax 000664), Polr2aTn(pb-CAG-GCaMP5g,-tdTomato)Tvrd/J (Jax 024477), and B6.Cg-Tg(Vil1-cre)1000Gum/J (Jax 021504). Mice were housed in an institutional facility with ad libitum access to food and water. Mice were sacrificed according to ethical guidelines using a fatal dose of rising levels of carbon dioxide. Cervical dislocation served as a secondary measure of euthanasia.

Primary murine colon cultures were prepared from wild type mice sacrificed at 7 wk based on published culture methods (Alcaino et al., 2018; Treichel et al., 2022). Dissociations were made from colonic epithelia because these are more resilient than their intestinal counterparts and more likely to survive for experimentation. The 10-cm segment of colon was removed and placed in ice-cold phosphate-buffered saline (PBS). A blunt tip syringe was used to flush the colon of fecal matter. Small spring scissors were used to remove any mesentery remaining on the colon. After cleaning, isolation of the submucosa and mucosal layers was done by peeling the outer muscle layers with fine forceps. The remaining mucosa and submucosa were minced until the tissue mass resembled a liquid and subsequently washed three times with fresh ice-cold PBS. Tissue was digested in a prewarmed digestion medium consisting of 500 ml DMEM (Sigma-Aldrich), 500 mg BSA (Sigma-Aldrich), 24 mg collagenase XI (Sigma-Aldrich), and 1 ml of PBS. Tissue was digested in four cycles of varying durations: the first cycle was 5 min, second was 10 min, third was 10 min, and fourth was 15 min. At each digestion, this media-cell solution was shaken vigorously to ensure dissociation into single cells, followed by centrifugation. Supernatants from the first two digestions were discarded. Supernatants from the last two digestions were collected and spun at 100 × g for 5 min. The resulting pellet was resuspended in culture media, which consisted of 500 ml DMEM (Sigma-Aldrich), 25 ml FBS, HI (Sigma-Aldrich), 5 ml L-glutamine (Life Technologies), and 5 ml Pen/Strep (Sigma-Aldrich).

Organoids were generated from four distinct wild type mice. Small intestine organoids were generated, as previously described (Alcaino et al., 2018), to increase our likelihood of consistently identifying EECs making cell–cell contacts with surrounding epithelia. Wild type mice were sacrificed as indicated above. The small bowel was removed from the abdominal cavity. After opening the small bowel longitudinally, it was cut into small sections and washed 15 times with calcium- and magnesium-free Dulbecco’s PBS (DPBS; StemCell Technologies). Washed segments were dissociated with Gentle Cell Dissociation Reagent (GCDR; StemCell Technologies) for 15 min and filtered with a 70 μm cell strainer. The filtered suspension was centrifuged at 500 × g and 4°C for 5–10 min and resuspended in DMEM (ATCC). After another centrifugation step under the same conditions, the pelleted crypts were resuspended in IntestiCult Organoid Growth Medium (IOGM, StemCell Technologies) and mixed at a 1:1 ratio with Matrigel (Corning). 400 μl domes were plated in prewarmed 6-well dishes (Corning) and cultured in 3 ml IOGM per well.

Flat sheets (1 cm × 0.5 cm) of small bowel were used. Tissues were fixed in 4% PFA-PB for 4 h separately, washed in PBS, and moved into 30% sucrose in PBS until they were frozen in an OCT embedding compound (Sakura Finetek). Tissues were cut into 12-µm-thick sections, rinsed with PBS twice for 5 min, and blocked with 200 μl per slide of 1% BSA/PBS/0.3% Triton X/10% normal donkey serum in a humidity chamber. Primary antibodies were added in 200 μl per slide of BSA/PBS/0.3% Triton/10% normal donkey serum and were incubated at 4°C overnight in a humidity chamber. Slides were then rinsed five times for 3 min in PBS. The secondary antibody was incubated for 1 h in the dark. Slides were mounted in SlowFade gold with DAPI (Life Technologies) mounting buffer.

An Elyra PS.1 Super Resolution microscope (Carl Zeiss Microscopy, LLC) with a 63×, 1.4 NA oil-immersion objective and a 1.5× tube multiplier was used. Laser-excitation wavelengths were 405, 488, 561, and 642 nm. The z-sections ranged from 8 to 10 μm, with a slice thickness of 175 nm. A four-image average was used for each z-slice of each of the five rotations of the grid. Channels were acquired sequentially, beginning with the 642-nm excitation, and ending with 405 nm. ZEN deconvolution software for SIM was used with default settings.

Organoids were grown in 50% Matrigel domes for 1-wk after passage. Organoid domes were washed with PBS and fixed in 4% PFA for 30 min at room temperature. Subsequently, they were permeabilized with 0.3% Triton X-100 (Thermo Fisher Scientific), blocked with UltraCruz Blocking Reagent (Santa Cruz), and incubated with primary antibodies overnight at 4°C. On the next day, they were washed with PBS, incubated with secondary antibodies for 1 h at room temperature, and mounted.

An LSM 980 microscope with Airyscan 2 (Carl Zeiss Microscopy, LLC) was used for super-resolution imaging with a 63×, 1.4 NA oil immersion objective. Excitation wavelengths: 353, 493, 577, and 653 nm. Airyscan modality was used during imaging, and default ZEN reconstruction settings were used for creating the super-resolution images.

Fluorescence profiles were generated with Zen blue software (Zeiss) and normalized with respect to maxima and minima in the traces. Manders’ correlation coefficients (MCCs) were generated using the JACoP plug-in in ImageJ (Bolte and Cordelières, 2006; Schneider et al., 2012) from inset images of each of the subcellular domains. Thresholds were applied consistently across analyses. Statistical analyses were performed with Prism 9 (GraphPad).

Total mRNA was isolated from vil^CrexPiezo2^f/f and Piezo2^f/f organoids using the RNeasy micro kit (Qiagen). cDNA was synthesized with SuperScript VILO (Thermo Fisher Scientific). Quantitative real-time polymerase chain reaction (qRT-PCR) was performed using FastStart Essential DNA Green Master (Roche Diagnostics) and analyzed using a LightCycler 96 (Roche Diagnostics). qRT-PCR primers used in this study (Integrated DNA Technologies) are listed in Table 1.

Colon epithelial cell suspensions from six mice were lysed and immunoprecipitated using the Capturem Co-IP kit (Takara Bio) following the manufacturer’s instructions. Samples were centrifuged at 17,000 × g and 4°C for 10 min. The supernatants were divided into five fractions: (1) input fraction, (2) no antibody control fraction, (3) rabbit IgG control fraction, (4) Piezo2 immunoprecipitation fraction, and (5) E-cadherin immunoprecipitation fraction. The number of fractions generated from a single mouse suspension depended on the yield from the primary dissociation. At least three biological replicates were generated for each fraction. Fraction 1 was immediately saved for blotting. Fraction 2 was not incubated with any antibody. Fractions 3–5 were incubated for 20 min at room temperature with a rabbit IgG control antibody or validated antibodies against specified targets at a 1:300 dilution, respectively. The binding of the immunoprecipitate to the column, washing, and elution were carried out for fractions 2–5 exactly to the manufacturer’s instructions using the provided solutions and a tabletop centrifuge. We used a colorimetric Bradford reaction (Bio-Rad) to determine the overall protein concentration in each of the samples using known concentrations of bovine serum albumin (BSA) to build a standard curve. We used the colorimetric reaction to consistently load the same amount of protein per experiment. We loaded 500 ng of protein in all wells for protein detection.

Protein detection was performed via Simple Western in a capillary-based automated instrument that carries out the electrophoretic separation of proteins and their immunodetection in tandem reactions (Jess, ProteinSimple). Samples were diluted with wash buffer and reconstituted with the manufacturer’s 10X sample buffer and 5X master mix for protein detection. We utilized two of the manufacturer’s detection modules for different proteins: 12–230 kDa module for the detection of δ-1 catenin, β-catenin, γ-catenin, α-catenin, vinculin, and actin; 66–440 kDa module for the detection of Piezo2 and E-cadherin. A chemiluminescence assay was selected. We used the manufacturer’s anti-rabbit detection module, which includes a conjugated anti-rabbit secondary antibody, for detection. Assay plates were loaded with 500 ng of protein per sample, biotinylated ladder, primary antibodies, blocking solution, secondary antibody, and luminol-peroxide substrate following the manufacturer’s instructions. Blot-equivalent images, electropherogram profiles, and protein peak areas for quantification were acquired from the Compass for Simple Western software (ProteinSimple).

Calcium imaging protocols are like those previously published (Alcaino et al., 2018; Treichel et al., 2022; Knutson et al., 2022). Dissociated single cells were imaged on an inverted Olympus IX70 epifluorescence microscope equipped with a 16-bit high-speed camera (ORCA-Flash4.0; Hamamatsu) and a CoolLED pE-300Ultra illumination system (CoolLED Limited). pCLAMP10.6 (Molecular Devices) was used to drive a piezotransducer microindenter, LED illumination system, Hamamatsu camera, and Metamorph imaging software (Molecular Devices). The bath solution contained: 127 mM NaCl, 3 mM KCl, 1 mM MgCl₂, 2.5 mM CaCl₂, 10 mM glucose, and 10 mM Hepes, pH 7.3, 320 mmol/kg (adjusted with sucrose). EECs were identified by tdTomato fluorescence (excitation/emission 554/581 nm) and studied functionally using GCaMP5 (excitation/emission 480–505/525 nm). Where noted, dissociated primary cells from three mice were randomly assigned to be transfected with 20 nM of Accell Mouse E-cadherin siRNA SMARTpool (E−041028-00-0005; Dharmacon), or 20 nM of Accell Non-Target control siRNA (D-001910-10-05; Dharmacon). Both cohorts (NT siRNA treated, E-cadherin siRNA-treated) contained cells from the three mice. Experiments were performed 48–72 h after the addition of siRNA-supplemented media.

For mechanical stimulation, a 0.8-µm indentation was delivered for 50 ms with a fire-polished glass microcapillary driven by a piezotransducer P-621.1CD attached to an E-625.CR controller (Physik Instrumente). Raw fluorescence from GcaMP5+ cells was exported to Microsoft Excel and converted to $ΔF/F0 [ΔF/F0=(F – F0)/F0]$⁠, where F₀ is the baseline fluorescence immediately before stimulation. Peak ΔF/F₀ values were extracted from the ΔF/F₀ traces. Areas under the curve (AUCs) were calculated from ΔF/F₀ traces. Statistical analyses were performed with Prism 9 (GraphPad).

Fig. S1 contains additional immunofluorescence imaging of Piezo2 in GI epithelia. Fig. S2 demonstrates our validation of the immunofluorescence approaches, including the Piezo2 antibody used for labeling. Fig. S3 contains additional immunofluorescence imaging of Piezo2 and E-cadherin in intestinal organoids; these images were used to generate average MCCs in Fig. 2. Fig. S4 shows inverse MCCs to those shown in Fig. 2. Fig. S5 contains uncropped versions of the protein immunoassays in Fig. 2. Fig. S6 validates the E-cadherin knockdown approach via immunofluorescence.

Piezo2 is expressed in multiple cellular compartments of mechanosensitive EC cells (Alcaino et al., 2018). We used super-resolution imaging of GI tissues and organoids to examine the subcellular distribution of the ion channel (Fig. 1 A). EC cells were 5-HT immunoreactive using a validated antibody (Beumer et al., 2020). A subpopulation of these 5-HT-positive cells was also positive for Piezo2 in tissues and organoids of all probed mice (three mice used for tissues, four for organoid generation) (Fig. 1, B and E; and Fig. S1). F-actin staining allowed us to discern the cellular borders since F-actin accumulation in the microvilli and the apical plate resulted in a significantly brighter signal toward the apical compartment, and the cortical actin network allowed us to discern the rest of the cell outlines (Fig. 1, B and E; Fig. S1; Fig. S2; and Fig. S3). We corroborated the Piezo2 antibody sensitivity and specificity via immunofluorescent labeling of organoids derived from Piezo2-Cre::GcaMP5/tdTomato and villin-Cre::Piezo2^f/f mice (Fig. S2, A–C). Control immunofluorescence experiments using isotype controls or lacking primary antibodies confirmed the specificity of these results (Fig. S2, D–G).

We noted Piezo2 distributed to three locations in primary EC cells: (1) overlapping with F-actin at the lateral wall (LW) (Fig. 1, B and C denoted by ‡), (2) in basal cytoplasmic (BC) accumulations (Fig. 1, B and C, denoted by ¶), and (3) in the perinuclear (Pn) cytoplasm (Fig. 1, B and C, denoted by §). Surprisingly, we rarely observed the Piezo2 signal in the apical EC cell compartment. We observed the same subcellar distributions in organoid 5-HT+/Piezo2+ double-positive cells (Fig. 1, E and F). Piezo2 and actin fluorescence profiles in both tissue and organoids demonstrated substantial F-actin accumulation with Piezo2 at the LW (Fig. 1, D and G). Profiles did not show as strong a correlation between Piezo2 and actin in the two other domains where the channel localizes (Fig. 1, D and G).

Piezo2 distribution was consistent between organoids and tissues, and since intestinal organoids provide a readily accessible three-dimensional model with spontaneously differentiated EECs, we pursued further imaging in organoids. We sought to establish co-localization between Piezo2 and E-cadherin since the two molecules have been reported to interact in cellular overexpression systems (Wang et al., 2022). We noted the same three pools of subcellular Piezo2 (Fig. 2 A, symbols consistent with Fig. 1 and Fig. S3). As expected, E-cadherin immunofluorescence using a validated antibody (Alvizi et al., 2023) localized mostly to the basolateral aspects of all epithelial cells in the organoid model (Fig. 2, A and C; and Fig. S3). E-cadherin is expected to mediate Piezo2 tethering to the actin cytoskeleton, and since previous fluorescence profiles showed strongest association between Piezo2 and F-actin accumulation at the LW (Fig. 1, D and G), we first fixated on the LW. Fluorescence profiles showed that part of the E-cadherin signal increased at the LW along with Piezo2 (Fig. 2 B and Fig. S3). The correlation between Piezo2 and E-cadherin fluorescence signals was more pronounced at the LW and the basal cytoplasm than in the perinuclear region (Fig. 2, C and D). To quantify these associations, we calculated MCCs of Piezo2 over F-actin and E-cadherin (Fig. 2 E and Fig. S3). Calculating MCCs over the whole cell (WC) revealed that ∼30% of the entire Piezo2 immunofluorescent signal overlapped with F-actin staining. Conversely, two-thirds of the entire Piezo2 immunofluorescent signal overlapped with that of E-cadherin in the WC. When analyzed by subcellular pools, an average of 72% of the Piezo2 signal overlapped with that of F-actin at the LW, whereas the correlation was more modest in the basal cytoplasm (23% on average across all analyzed cells). We consistently found very little overlap between Piezo2 and F-actin in the perinuclear region. The overlap between Piezo2 and E-cadherin was also strongest at the LW (average 78%, very close to the Piezo2/F-actin average). We additionally noted significant Piezo2/E-cadherin overlap in the basal cytoplasm (average 66% across all analyzed cells). As with F-actin, the Piezo2/E-cadherin overlap was consistently nearly absent in the perinuclear region. As expected, the F-actin/Piezo2 and E-cadherin/Piezo2 MCCs were lower for the WC and across all subcellular locations (Fig. S4).

Since the majority of Piezo2 signals at the LW co-localized with that of E-cadherin and actin, we sought to confirm that these molecules interact with each other. We also asked whether the adaptor proteins that link E-cadherin to the actin cytoskeleton were part of the proposed Piezo2–E-cadherin complex in EECs. Mucosal Piezo2 expression is largely restricted to EECs (Alcaino et al., 2018; Billing et al., 2019; Treichel et al., 2022). Hence, we performed co-immunoprecipitation experiments using stripped mucosal segments, which consist mostly of epithelial cells. We immunoprecipitated Piezo2 or E-cadherin and used a high-sensitivity protein separation and immunodetection assay to probe the molecules that co-immunoprecipitated with either protein. E-cadherin co-immunoprecipitated Piezo2, and, as expected, also co-immunoprecipitated actin and the associated proteins catenin δ-1 (p120 catenin), β-catenin, γ-catenin, α-catenin, and vinculin (Fig. 2, F and G). In the Piezo2 precipitates, we co-immunoprecipitated E-cadherin and actin, along with catenin δ-1 (p120 catenin), γ-catenin, α-catenin, and vinculin (Fig. 2, F and G). Protein band quantification confirmed protein accumulation in the precipitates over the column and antibody controls (Fig. 2 G and Fig. S5).

Since Piezo2 interacts with E-cadherin in GI EECs, we next asked whether this interaction has a functional relevance in EEC mechanosensitivity. To answer this, we dissociated epithelia from NeuroD1-Cre::GCaMP5/tdTomato mice, where the genetically encoded calcium indicator GCaMP5 is expressed by cells expressing NeuroD1 (Alcaino et al., 2018; Treichel et al., 2022; Knutson et al., 2022), a late-stage transcription factor defining EEC fate (Haber et al., 2017). Dissociated cells were cultured in the presence of either non-targeting (NT) or E-cadherin siRNA for at least 48 h. Immunofluorescence showed no apparent difference in the expression or distribution of Piezo2 while confirming decreased E-cadherin protein levels (Fig. S6). To test the function, we mechanically stimulated via a single 50-ms blunt indentation (Fig. 3 A). Individual EECs were identified by tdTomato fluorescence (Fig. 3 B), and acute intracellular calcium (Ca²⁺) activity was measured by GCaMP5 fluorescence (Fig. 3, C and D). A proportion of NT siRNA-treated cells (11/16) showed the characteristic mechanosensitive Ca²⁺ response we had previously observed in mechanosensitive EECs and which we had previously proven to be Piezo2-dependent (Alcaino et al., 2018) (Fig. 3 C). E-cadherin knockdown diminished the peak calcium response to mechanical stimulation (Fig. 3, C–E) by 64.9 ± 23.1% and AUCs by 68.5 ± 32.6% (Fig. 3, F and G).

GI EECs are epithelial mechanoreceptors that require Piezo2 to coordinate GI motility and secretion (Alcaino et al., 2018; Treichel et al., 2022). Piezo2 was suggested to interact with E-cadherin in cultured cell models (Wang et al., 2022). Here, we tested whether Piezo2 interacted with E-cadherin in mechanoreceptors of the GI epithelial lining—mechanosensitive EECs. We also demonstrate this interaction could play a functional role in EEC mechanosensitivity. To our knowledge, this is the first demonstration of interactions between Piezo2 and E-cadherin/actin in primary tissues and organoids.

Intracellularly, E-cadherin anchors to the actin cytoskeleton (Lecuit and Yap, 2015). Our co-immunoprecipitation experiments show that E-cadherin, actin, and the linking proteins that span the two co-immunoprecipitate with Piezo2. As others have shown that Piezo2 ionic currents decrease upon actin depolymerization (Eijkelkamp et al., 2013; Jia et al., 2016; Verkest et al., 2022), here, we report that E-cadherin disruption decreases mechanosensitive Ca²⁺ responses. Piezo2 is a multimodal force sensor that detects both membrane deflection (force-from-lipid) and force transmitted along actin tethers (force-from-filament) (Cox et al., 2019; Verkest et al., 2022; Zhou et al., 2023). Several actin-binding proteins that co-immunoprecipitate with Piezo2—specifically α-catenin and vinculin—are mechanosensitive (le Duc et al., 2010; Twiss et al., 2012; Yao et al., 2014; Noordstra et al., 2023). It is possible that force transmission and Piezo2 activation through force-from-filament may be different than through force-from-lipid, resulting in different responses, especially when the Piezo2+ EECs are integrated back into epithelial monolayers. We conclude that force-from-filament transmission requires a particular focus on the EEC LW since this is the cell domain where we found the strongest overlap between Piezo2, E-cadherin, and F-actin. Of note, neither the Piezo2/F-actin nor the Piezo2/E-cadherin MCCs were perfect ones at the LW and displayed some degree of variability. This suggests that while the vast majority (approximately three-quarters) of Piezo2 clusters tether via E-cadherin at the LW, a fraction of Piezo2 molecules in this domain rely on either lipid-based mechanosensing or on an alternative tether.

Alternatively, the interaction between Piezo2 and E-cadherin/actin may localize Piezo2 to areas of sharp force gradients to optimize cellular mechanosensitivity, such as it has been suggested between VE-cadherin and Piezo1 in endothelial cells (Chuntharpursat-Bon et al., 2019, Preprint). Mechanosensitive binding partners such as vinculin might play role in detecting such high-load areas within the cell. Further, the actin cytoskeleton forms “corrals” that compartmentalize regions of the plasma membrane and proteins contained therein (Sadegh et al., 2017), including ion channels (Tamkun et al., 2007). In the EEC, cytoskeletal elements may define mechanical domains for force transduction at the plasma membrane. Channel clustering could extend beyond Piezo2 and include other channels potentially involved in mechanotransduction, similar to how voltage-gated ion channels are clustered in neurons by cytoskeletal domains (Xu et al., 2013; Solé and Tamkun, 2020). Future studies in multicellular cell systems may investigate if E-cadherin and associated structural elements play a role in Piezo2 targeting to, and clustering at, the LW. Our combination of immunofluorescence and co-immunoprecipitation in tissues and organoids support the conclusion that a pool of Piezo2 molecules in GI EECs localizes to the lateral membrane, where they interact with E-cadherin and actin. We report that part of the cytoplasmic pool of Piezo2 at the base of the EEC also co-localizes with E-cadherin, though it does not co-localize as strongly with F-actin. These preliminary clues provide testable hypotheses about the temporal sequence of Piezo2–E-cadherin association and tethering to the actin cytoskeleton. The temporal assembly of E-cadherin junctional complexes has been investigated and suggested to include both early associations and late interactions at the cell junction (Bryant and Stow, 2004; Miranda et al., 2003). Piezo2 and E-cadherin may be stored in intracellular vesicles separate from serotonin-containing vesicles, ready for shuttling to the plasma membrane (Alcaino et al., 2018). If so, the Piezo2–E-cadherin interaction may take place prior to Piezo2 localization to the membrane. This potential interpretation would be consistent with the proposed model of Piezo2–E-cadherin interaction, whereby the junctional protein impales the ion channel so the ectodomain of E-cadherin interacts with the cap domain of Piezo2 (Wang et al., 2022). We did not see interactions between Piezo2 and E-cadherin in the perinuclear pool of Piezo2, perhaps suggesting that actively translated proteins may be too nascent to interact with E-cadherin. However, dedicated experiments that would look at the trafficking of these molecules over time are required to test these hypotheses.

E-cadherin and Piezo2 are co-expressed in several epithelial mechanoreceptors. These include linings in the skin (Woo et al., 2014), lung (Nonomura et al., 2017), and bladder (Marshall et al., 2020), where Piezo2 mechanosensitivity is critical for physiologic function. Yet the interactions between molecular mechanosensors may have implications broader than sensory mechanotransduction. Diseases like cancer display mechanobiological features that impact progression and survival (Rao et al., 2021). While the mechanistic role of Piezo2 in cancer remains to be elucidated, E-cadherin protein expression is still detectable in solid tumors upon Piezo2 overexpression and E-cadherin downregulation (Katsuta et al., 2022). Classically, E-cadherin downregulation is perceived to correlate with more invasive cancer phenotypes as part of the epithelial-to-mesenchymal transition (EMT) (van Roy, 2014). Nevertheless, this is not always the case, as metastases may continue to express E-cadherin, and positive roles for E-cadherin in metastases have been reported (Na et al., 2020). Piezo2 expression correlates with lymph node infiltration in patients with colorectal cancer (Liu et al., 2022). Future experiments will be necessary to discern if E-cadherin similarly persists in colorectal cancer, and if so, whether it is potentiating Piezo2 activity and affecting cancer cell invasiveness. Since the effect of Piezo2 expression in cancer is tissue-dependent (Liu et al., 2022), interactions such as the Piezo2–E-cadherin interaction (and others) may shift the balance between positive and negative outcomes.

In summary, we found that in specialized GI epithelial mechanoreceptors, a fraction of the Piezo2 pool localizes to the lateral membrane, where it interacts with E-cadherin, actin, and other cytoskeletal-binding mechanosensors; and that this interaction may affect Piezo2-mediated cellular mechanosensitivity. It is possible that the Piezo2–E-cadherin interaction is an important anchor for Piezo2 to detect forces transmitted along epithelial tissues, which connect their cytoskeletons into supracellular networks that span across the tissue (Lecuit and Yap, 2015). The connections between Piezo2 and structural proteins, including E-cadherin and actin, have important implications in sensory mechanotransduction and tissue mechanotransduction in health and disease.

The data underlying Fig. 2 are available in the supplemental material. Any further data are available from the corresponding author upon request.

Jeanne M. Nerbonne served as editor.

The authors thank Drs. Deborah Leckband and Gina Razidlo for constructive intellectual discussions and Mr. Eugene Krueger for assistance with super-resolution imaging and image analysis. We thank Mrs. Lyndsay Busby for administrative support.

This work was supported by the National Institutes of Health (NIH) grants: GM065841, DK128913, DP2AT010875, DK123549, DK052766, TR002379, and DK084567 (Mayo Center for Cell Signaling in Gastroenterology).

Author contributions: A. Mercado-Perez: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Visualization, Writing - original draft, Writing - review & editing, J.P. Hernandez: Investigation, Y. Fedyshyn: Data curation, Methodology, Validation, Writing - review & editing, A.J. Treichel: Investigation, Methodology, V. Joshi: Investigation, Writing - review & editing, K. Kossick: Resources, K.R. Betageri: Formal analysis, Investigation, G. Farrugia: Conceptualization, Methodology, Project administration, Resources, B. Druliner: Investigation, Resources, Supervision, Writing - review & editing, A. Beyder: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing - original draft, Writing - review & editing.