Introduction

PIEZO proteins encompass a small family of mechanically-activated and cation-selective ion channels exclusively present in eukaryotic organisms, with only two vertebrate members called PIEZO1 and PIEZO2 (1, 2). PIEZO channels share the same homotrimeric organization, that of a central ion conducting pore surrounded by three curved blade-like domains, each of these blades consisting of a juxtaposition of nine transmembrane helical bundles (called repeats A-I, or transmembrane helical units 9-1) (3–6). The inherent curvature of the PIEZO blades gives these proteins an overall bowl shape that deforms the surrounding membrane beyond the channel-membrane boundary (7–9). Mechanical membrane deformations such as lateral stretch or outward bending promote flattening of the PIEZO blades (10, 11), a conformational change thought to promote opening of the central pore (12–15).

PIEZO1 is expressed in many cell types, whereas PIEZO2 is mainly present in peripheral mechanosensory neurons as well as a discrete number of non-neuronal cells such as muscle and vascular endothelial cells (2, 16, 17). Genetic studies in which Piezo1 and/or Piezo2 genes have been invalidated globally or in specific tissues have revealed numerous biological roles mediated by PIEZO channels, ranging from somatovisceral mechanosensation (18–21) and proprioception (22, 23), to the growth and development of musculo-skeletal (24, 25), cardiac (26), lymphatic (27, 28) and vascular tissues (29, 30). To date, PIEZO channels have been associated with numerous ailments including blood disorders (31–34), mechanical allodynia (35–37), generalized lymphatic dysplasia (38, 39), Prune Belly syndrome (40) and abnormal musculoskeletal development (41–47).

The considerable impact of PIEZO channels on health and diseases has sparked the search for homolog-selective pharmacological modulators, both as molecular probes to dissect homolog-specific biological functions and as potential drugs to treat PIEZO-related pathologies. In this context, Yoda1, a synthetic small molecule proposed to act as a PIEZO1-specific agonist, has been discovered from a high-throughput screening campaign (48). Since its discovery, Yoda1 and its more potent analog Yoda2 (49) have been extensively used as a tool to unravel novel PIEZO1 functions in immune cells (50), bones (51–53), cancer progression (54, 55), and in the vascular system (56–59). In contrast to the common belief that Yoda molecules act as PIEZO1-selective modulators, we show here that they also agonize PIEZO2, prompting careful reinterpretation of prior studies.

Results

Yoda1 exerts distinct measurable effects when used at micromolar concentrations to PIEZO1-expressing cells: 1) it evokes spontaneous (i.e., without external mechanical stimulation) calcium influx; 2) it reduces the minimal force necessary to detect ionic currents (activation threshold) when these currents are evoked by pipette suction (“stretch currents”) or with cellular indentations (“poking currents”); 3) it slows down inactivation and deactivation kinetics of macroscopic currents; and 4) it increases single channel open probability under steady-state membrane stretch conditions (48, 60, 61). The prevailing notion that Yoda1 acts as a PIEZO1 selective modulator mainly arose from Yoda’s inability to elicit spontaneous calcium influx and to slow down inactivation kinetics of poking currents in PIEZO2-expressing cells (48). The effect of Yoda1 on PIEZO2 stretch currents has never been tested to our knowledge, likely because PIEZO2 was thought to be insensitive to membrane stretch until recently (15, 62, 63).

Yoda1 robustly modulates PIEZO2 stretch currents

We first tested the ability of Yoda1 to modulate PIEZO2 stretch currents. Naïve HEK293T^ΔPZ1 cells (which do not produce mechanosensitive currents) (38) were transfected with a mouse PIEZO2 plasmid and measured the next day to capture single channel currents in presence of 20 µM Yoda1 or vehicle control (dimethyl sulfoxide, or DMSO) at a constant patch pressure of -60 mmHg (Figure 1a). Idealization of these currents reveal that Yoda1 increases PIEZO2’s mean open time by ∼3-fold (18.1 ± 2.4 vs. 47.8 ± 8.3 ms) while decreasing PIEZO2’s mean shut time by ∼2-fold (383 ± 62 vs. 161 ± 52 ms). Similar reciprocal microscopic effects of Yoda1 have been reported for PIEZO1 (60). We calculated open probability by dividing the mean open time by the sum of mean open and shut times (64) and found that Yoda1 increases PIEZO2 open probability by about 5-fold in our tested conditions (0.06 ± 0.01 vs. 0.30 ± 0.11, Mann-Whitney U test p-value = 0.0093) (Figure 1b). For comparison, application of 30 µM Yoda1 increases PIEZO1 open probability by 2/3-fold when measured in single channel patches held at -40 mmHg (60). A dwell time analysis of PIEZO2 single channel recordings reveals that the channel populates at least three distinct open and shut states with distinct exponential lifetimes, all of them being reciprocally modulated by Yoda1 (Figure 1c-d), as previously observed for PIEZO1 (60). Next, we measured macroscopic PIEZO2 stretch currents evoked by brief suction pulses of incremental amplitude from +5 to -165 mmHg (Figure 1e). Fitting the relative amplitude of peak currents (I/Imax) using a Boltzmann function indicates that the presence of Yoda1 in the patch pipette correlates with a shift of the mid-point pressure (i.e, pressure needed to reach half of maximal activation, or P_1/2) by about +20 mmHg (-55.4 ± 5.3 vs. -36 ± 4.5 mmHg, Mann-Whitney U test p-value = 0.0117), similar to the reported +15/20 mmHg shift for PIEZO1’s P_1/2 value (48, 61). These results show that Yoda1 overall potentiates PIEZO2 stretch currents to a similar extent than reported for PIEZO1 (48, 60).

Figure 1:

Yoda1 activates PIEZO2 in absence of external mechanical stimulus

Given the strong modulation of PIEZO2 stretch currents by Yoda1, we sought to re-evaluate the ability of this molecule to elicit spontaneous calcium influx in PIEZO2-expressing cells, i.e, in absence of external mechanical stimulation. Two distinct scenarios could explain why this effect has not been previously observed. First, the ability of Yoda1 to modulate mechanically-evoked PIEZO2 currents (measured electrophysiologically) and to spontaneously activate PIEZO2 (measured using calcium imaging) could be mediated by distinct pharmacological mechanisms, as recently suggested for PIEZO1 (65). On the other hand, the activatory effect of Yoda1 on PIEZO2 might be too weak to be detected in absence of external mechanical stimulus, possibly due to PIEZO2’s lower membrane expression, weaker calcium permeability, and/or inherently higher mechanical activation threshold (63). We thus tested if an effect of Yoda1 can be observed by increasing the amplitude of calcium signals. To this aim, we used an isotonic “calcium boosted solution” (CBS) containing 30 mM Ca²⁺ ions, low potassium and no sodium to increase the inward electrochemical driving force for calcium ions. We transfected naïve cells with a control plasmid encoding the calcium sensitive fluorescent reporter GCaMP6m (GC6), or with PIEZO1 and PIEZO2 plasmids co-encoding GC6 through an Internal Ribosome Entry Site (IRES) (66). One minute after Yoda1 perfusion, transfected cells were exposed to a calcium ionophore to normalize maximal calcium signals (maxΔF/F₀) as a function of the maximal GC6 response (to reduce experimental noise due to cell-to-cell variability of plasmid expression). As previously observed, Yoda1 clearly failed to evoke calcium influx through PIEZO2 in the presence of physiological saline (Hank’s Balanced Salt Solution, HBSS) (Figure 2a-b). In contrast, in the presence of CBS, Yoda1 evoked sizeable calcium influx in PIEZO2 transfected cells (Yoda1 vs. DMSO Kruskal-Wallis tests with Dunn’s multiple comparisons p-values <0.0001 for both PIEZO1 and PIEZO2) (Figure 2a-b).

Figure 2:

The PIEZO2 structure was recently shown to adopt the characteristic flattened conformation associated with an open state (15) in response to hypertonic, but not hypotonic shocks (67). This seemed counterintuitive since PIEZO1 flattened under hypotonic, but not hypertonic conditions. This conundrum was elegantly explained by the unique property of PIEZO2 to interact with the cytoskeletal-associated protein filamin-B, altering PIEZO2’s response to distinct mechanical stimuli (67). To test if increasing extracellular tonicity can amplify the effect of Yoda1, we compared the amplitude of Yoda1-induced calcium signals (normalized to DMSO vehicle control) in cells transfected with our PIEZO2-IRES-GC6 plasmid under isotonic (1X HBSS) or hypertonic (2X and 3X HBSS) conditions. Calcium signals were approximately 2- to 3-fold larger in hypertonic vs. isotonic conditions (Kruskal-Wallis test with Dunn’s multiple comparisons p-values = 0.0119 and <0.0001 for 2X and 3X, respectively) (Figure 2c).

Yoda1 modulation of PIEZO2 poking currents

Although Yoda1 was previously found unable to modulate PIEZO2 poking currents kinetics, this conclusion was based on a small sample size (48). We recorded poking currents from a larger number of cells expressing either wild-type mouse PIEZO2 or PIEZO2 variants fused with a green fluorescent protein at the N-terminus (N-GFP) or at the C-terminus (C-mClover3) to better identify cells expressing PIEZO2. As previously observed, PIEZO2 currents display little to no inactivation when elicited by suction pulses (Figure 1e) (63), but exhibit rapid inactivation when elicited by poking stimuli (Figure 3a) (2). Remarkably, PIEZO2 poking currents tend to inactivate about 2-fold slower in the presence of 20 µM Yoda1, an effect that was more pronounced for PIEZO2 C-mClover3 and wild-type (WT) PIEZO2 (unpaired t-test p-value < 10^-5) than PIEZO2 N-GFP (p-value = 0.0180) (Figure 3b). For comparison, PIEZO1 poking currents inactivate 5-10-fold slower in the presence of Yoda1 in similar experimental conditions (60).

Figure 3:

Yoda1 is thought to act as an allosteric positive modulator by binding a hydrophobic pocket sandwiched between PIEZO1 repeats A and B, promoting flattening of the mechanosensory blade domains (61, 65, 68–70). This pharmacological mechanism was initially proposed through engineering a PIEZO1/2 chimera in which a specific amino acid sequence overlapping this binding region (corresponding to mouse PIEZO1 residues 1961-2063) is swapped with its mouse PIEZO2 homologous counterpart (61). This chimera was not modulated by Yoda1 when tested using stretch current protocols, an observation which now seem odd given the present observations that PIEZO2 stretch currents are similarly modulated by Yoda1 (Figure 1e). To complete our electrophysiological characterization, we sought to test an effect of Yoda1 in this chimera with a pairwise poking protocol, using brief poke stimuli to measure deactivation kinetics, the gating parameters being the most modulated by Yoda1 on PIEZO1 (60). Like other tested parameters, deactivation kinetics in this chimera are completely insensitive to 30 µM Yoda1 (p-value = 0.1567), in contrast to wild-type PIEZO1 (paired t-test p-value = 0.0171) (Figure 3c).

Modulation of PIEZO2 poking currents by other small molecules

The replacement of Yoda1’s pyrazine ring for a benzoic acid group led to the creation of Yoda2, an improved analog which agonizes PIEZO1 at nanomolar concentrations (49). We thus sought to test if Yoda2 modulates PIEZO2 currents with similar potency. Although the effect of Yoda1 on PIEZO2 is larger when probed using stretch currents, we used PIEZO2 poking currents instead because these currents are quicker to measure, facilitating the construction of dose-response curves along a wide range of concentrations. Using this approach, we found that Yoda2 slows down PIEZO2 inactivation kinetics at nanomolar concentrations, with a half maximal effective concentration (EC50) at least one order of magnitude lower compared to Yoda1 (∼0.4 vs. ∼10 µM, Figure 4ab). At saturating concentrations, however, Yoda2 exerts a smaller effect than Yoda1 (lower maximal efficacy). A similar efficacy difference between Yoda1 and Yoda2 has been reported for PIEZO1, as Yoda2 was shown to slow down PIEZO1 poking currents inactivation kinetics by about 1.5-2-fold at saturating concentrations (65), while Yoda1 reportedly slows down inactivation of the same currents by up to 5-10 fold (60).

Figure 4:

Next, we tested whether other small molecules can modulate PIEZO2 poking currents, including Dooku1 (a Yoda1 analog that antagonizes the effect of Yoda1 with no known modulatory effect on PIEZO1) and two Yoda1-unrelated PIEZO1 chemical activators (named Cmpd15 and Cmpd64) identified from a structure-guided virtual screening campaign (71, 72). Unexpectedly, Dooku1 slows down inactivation of PIEZO2 poking currents by about 2-3-fold, i.e, similar to Yoda1 (Figure 4c). On the other hand, these currents were not modulated by Cmpds 15 and 64 in our tested conditions, suggesting that these two Yoda-unrelated activators have better PIEZO1-selectivity than Yoda molecules.

A conserved arginine is required for Yoda2’s nanomolar potency

The Yoda binding region is well conserved between the two channel homologs, both at the level of the backbone structure and of amino acid sequence (Figure 5a) (3–6). It thus seems reasonable to speculate that Yoda molecules modulate PIEZO2 by binding the same region. The intracellular side of this binding region possess two conserved arginine residues, R1724 and R2098 in mouse PIEZO1, which could electrostatically interact with Yoda2’s negatively-charged benzoic acid group, potentially contributing to its higher potency relative to Yoda1 (Figure 5b). Although Yoda1 is electrically neutral, virtual molecular docking studies suggest that mouse PIEZO1 R2098 interacts with Yoda1’s pyrazine group (73–75). We tested a pharmacological role for these basic residues in mouse PIEZO1 by substituting them for serine (to conserve their polar character while eliminating their positive charge) and evaluated the impact of these mutations on Yoda1 and Yoda2 dose-response using calcium imaging (Figure 5c, left panels). We fitted these dose-response curves with a standard binding equation and extracted their EC50 values. To ease comparison between mutants and WT channels, we normalized the mutants’ EC50 value to that obtained for WT PIEZO1 and plotted normalized EC50 for both Yoda1 and Yoda2 and for each mutant (Figure 5c, right panel). Clearly, the R2098S mutation has no effect on the potency of either Yoda1 or Yoda2, as the normalized EC50 obtained for both molecules is close to unity, ruling out a pharmacological role for R2098. In contrast, the R1724S mutation reduces Yoda2 potency about 4-fold relative to WT (unpaired Student’s t-test p-value = 0.017), with little effect on Yoda1 potency. The large reduction of Yoda2’s potency in the R1724S mutant suggests that R1724 stabilizes Yoda2 by forming a salt bridge.

Figure 5:

Yoda2 forms a transient salt bridge in the Yoda binding site

The side chain of R1724 displays different orientations in different PIEZO1 experimental structures (Figure 6a). To quantify this orientation, we calculated the pseudo-dihedral torsion angle between the center of the guanidinium group (R1724-Cζ) and three reference atoms (Figure 6b). We next evaluated the temporal distribution of this torsion angle from our recent PIEZO1 simulations (15). Our analysis reveals that this side chain populates two major rotameric orientations, one with torsion angles of 50-100° and the other with torsion angles around -80°. We observed a similar torsion angle distribution across nine PIEZO1 experimental structures (Figure 6c). Critically, rotamers with torsion angles around -80° point the guanidinium group towards the Yoda binding site, while the other rotamers point this group away from it. To explore in more detail how the orientation of R1724 guanidinium group affects Yoda2 binding, we performed all-atom molecular dynamics (MD) simulations of a Yoda2 molecule bound to a truncated PIEZO1 blade (which includes Repeats A-C, the anchor region and the intracellular beam) in the open state, as described in our recent PIEZO1-Yoda1 binding simulations (69) (Figure 6d). Overall, Yoda2 populates binding poses near the intracellular side of the Yoda binding site in the open state, similar to Yoda1 (69). Our analysis reveals a clear correlation between the torsion angle and the formation of a salt-bridge between R1724 and Yoda2: the salt bridge is stable for torsion angle values around -80° and become destabilized when R1724 populates rotamers with positive torsion angles. The dominant Yoda2 binding pose (observed in 82.6%, 49.7%, and 99.8% of all frames in Replica1, 2, and 3, respectively) shows that Yoda2 forms a salt bridge with R1724 and several other non-covalent interactions with I696, F1715, M1719, and Y1692.

Figure 6:

Discussion

Since its discovery, Yoda1 has been used to unravel specific biological functions that depend on PIEZO1 activation. Our discovery that Yoda1 modulates PIEZO2 stretch currents as effectively as PIEZO1’s strongly suggests that some of the reported biological roles attributed to PIEZO1 using this agonist could be (at least partly) transduced by PIEZO2. For instance, our discovery seamlessly reconciles the puzzling fact that reducing PIEZO2 expression in cells that co-express PIEZO1 also reduces the effects of Yoda1 (49, 76, 77). Along the same line, since Yoda1 was shown to increase the expression of the lymphatic valve gene ITGA9 in a PIEZO1-independent manner (27), this genetic effect could be mediated by PIEZO2, which is present in lymphatic endothelial cells (78). Besides, it is also known that some reported effects of Yoda1 are mediated by proteins unrelated to the PIEZO channel family. For example, Yoda1 was recently shown to inhibit Kv1.3 (79), a major potassium channel in immune cells. Since Yoda1 was shown to modulate certain inflammatory responses, these immunological effects could be mediated independently of PIEZO channel activation, for instance through Kv1.3 inhibition (80–82). Lastly, Yoda1 was shown to induce phosphorylation of protein kinase B (Akt) and extracellular-related kinases (ERK) 1 and 2 in vascular endothelial cells (83) in the presence of GsMTx4, a spider toxin which inhibits mechanosensitive channels including PIEZO1 and 2 (84–86).

We previously engineered a Yoda1-insensitive PIEZO1/2 chimera based on the assumption, now obsolete, that PIEZO2 is insensitive to Yoda1 (61). We however show here that this chimeric swap (swapping mouse PIEZO1 residues 1961-2063 for their homologous mouse PIEZO2 counterparts) completely abolishes Yoda1’s ability to modulates channel kinetics, further strengthening our conclusion that this chimeric PIEZO variant is insensitive to this chemical agonist. A closer look at this chimera reveals that only about half of the residues forming the Yoda1 binding site (those residues contained within repeat A) have been swapped. Although what causes the loss of Yoda1-sensitivity in this chimera remains unclear, we can speculate that the chimeric swap might cause Yoda1 to populate its cognate binding poses less effectively (affinity loss) and/or to adopt novel, non-efficacious binding poses (efficacy loss).

Our pressure-clamp recordings show that Yoda1 lowers PIEZO1’s and PIEZO2’s pressure activation threshold to the same extent. Yet, unlike PIEZO1, a spontaneous activation of PIEZO2 by Yoda1 could only be detected by calcium imaging under non-physiological conditions aiming to increase calcium influx. What molecular mechanisms could explain this difference? A possibility may be due to PIEZO2’s inherently higher pressure-threshold (63, 87). Even in the presence of Yoda1, PIEZO2’s activation threshold would remain higher than the resting membrane tension of the cell, limiting the frequency of spontaneous opening (48, 61, 65). Another possibility stems from the notion that Yoda1 was proposed to adopt two distinct binding poses: an upper pose (near mouse PIEZO1 residues A2091, A2094 and F1693) preferentially occupied by Yoda1 when the curved channel is at rest, and a lower, more intracellular pose (near mouse PIEZO1 residues M1719, F1715, and V1714) populated when the channel flattens (69). Remarkably, the A2094W mutation in the upper site specifically abrogates the activatory, but not modulatory, effects of Yoda1 and also prevents Yoda1 from flattening PIEZO1 in absence of external stimulus (65, 70). This suggests that PIEZO2’s weaker response to Yoda1 in calcium imaging assay could result from alteration of this upper binding site. Such a hypothetical mechanism is likely to be allosteric in nature, since most side chains interacting with Yoda1 are conserved between PIEZO1 and PIEZO2. Another remarkable difference noticed in our study is the lower efficacy of the modulatory effects produced by Yoda molecules on PIEZO2 poking currents inactivation kinetics relative to PIEZO1. This difference may be explained by PIEZO1 and PIEZO2 having distinct interactions with cytoskeletal and adhesion proteins known to modulate inactivation kinetics (67, 88, 89).

Last, our study reveals a role for a conserved arginine located in the Yoda binding site (R1724 in mouse PIEZO1) in selectively increasing the binding potency (lower EC50) of Yoda2. The orientation of this arginine fluctuates during molecular dynamics simulations and across several experimental structures. This rotamer dynamics is likely due to the flexibility of the intracellular loop in which this residue resides. Our binding simulations suggests that, while both Yoda1 and Yoda2 bind with similar poses in the Yoda binding site in the open state, Yoda2’s benzoic acid group forms a salt bridge with this arginine when the latter point its side chain towards the binding site. Hence, we propose that the increased potency of Yoda2 arises from the transient formation of a salt bridge with a flexible arginine side chain. This work thus opens new avenues for drug design and optimization.

Materials and Methods

Molecular cloning

We used plasmids encoding mouse PIEZO1 (pCDNA3.1-mPIEZO1) and (PIEZO2Sport6-mPIEZO2) gifted earlier by Ardèm Patapoutian (Scripps Research), as well as a Sport6-mPIEZO2-N-GFP gifted by Tibor Rohacs (Rutgers New Jersey Medical School) and a pMO-mPIEZO2-C-mClover3 plasmid we commissioned Epoch Life Science (TX, USA). The bicistronic plasmids pCDNA3.1-mPIEZO1-IRES-GCaMP6m and pCDNA3.1-mPIEZO2-IRES-GCaMP6m were created in-house earlier (61). mPIEZO1 single point mutations were created and sequence-verified by Epoch Life Science using a pMO-mPIEZO1 plasmid.

Cell culture, transfection and reagents

HEK293T^ΔPZ1 cells (gifted earlier from Ardèm Patapoutian), were cultured at 37°C and 5% CO2 in Dulbecco’s Modified Eagle Medium (GIBCO) supplemented with Penicillin-Streptomycin and 5% fetal calf serum (Life technologies). HEK293T^ΔPZ1 cells were transfected as previously described with slight modifications (90). 3 μg of mPIEZO2 plasmid or 1 μg PIEZO1 plasmid were transfected to 90% confluent HEK293T^ΔPZ1 cells in a one well of a 12 well plate. Yoda1 (Millipore-Sigma), Yoda2 (Tocris), Dooku1 (Tocris), Cmpd15 (Chembridge chemical ID:7410103) and Cmpd64 (Chembridge chemical ID:33734878) were dissolved as stock solutions in DMSO (Millipore-Sigma) at 40 mM, 40 mM and 10 mM, respectively, as described before (69).

Electrophysiology

Transfected cells were re-seeded on glass coverslips coated with Matrigel at a 1/50 dilution (Corning) one day after transfection and used for experiments two days after transfection. Patch pipettes were pulled from G150F borosilicate capillaries (Warner Instruments) to a resistance of 2-3 MΩ using a P-97 puller (Sutter Instrument) and heat-polished using a Microforge-MF2 (Narishige). Single channel and macroscopic PIEZO1 and PIEZO2 stretch currents were elicited using gigaseal pressurization (high-speed pressure clamp, ALA Scientific) or by poking cells with a piezo-electric actuator (Physik Instrumente) in a Hank’s Balanced Saline Solution (HBSS, Thermofisher) as described earlier (60). Single channel dwell times were generated and analyzed as described earlier (64).

Calcium imaging

Cells were washed and incubated for at least 30 min in HBSS prior measurements. The CBS solution contains 1 mM KCl, 30 mM CaCl2, 1 mM MgSO4, 108 mM N-Methyl-D-Glucosamine chloride, and 10 mM HEPES. The pH was adjusted to 7.4 with HCl and has an osmolarity of 288 mOsmol/L (all chemicals were purchased from Millipore-Sigma). For hypertonic treatments, 1X, 2X and 3X HBSS were prepared by dilution of a 10X commercial HBSS solution (Gibco). Fluorescence images were acquired at 1 frame per second either on an inverted Nikon Eclipse microscope through a Nikon CDD camera as described earlier (61), or on a Zeiss LSM880 confocal microscope.

Molecular dynamics simulations

Docking and molecular dynamics (MD) simulations used a truncated all-atom open PIEZO1 model (residues 1131 to 2190) (69). Molecular docking was performed using AutoDock Vina 1.2.3 with the docking grid defined between Repeat B (TM29–30) and Repeat A (TM35–36) using MGLTools-1.5.6 and a grid size of 30 × 30 × 30 Å. The PIEZO1-Yoda2 complex system was assembled using CHARMM36m parameters for the protein, CHARMM36 parameters for POPC lipids, ions, and water molecules, and CHARMM General Force Field (CGenFF) for Yoda2. The system (211,358 atoms in total) was neutralized and maintained at an ionic concentration of 150 mM KCl and solvated using TIP3P water. All-atom MD simulations were performed using GROMACS 2024.2 smp. We followed the input files generated by CHARMM-GUI, including six equilibration steps followed by production simulations. Energy minimization was performed using the steepest descent algorithm for 5,000 steps. Long-range electrostatic interactions were calculated using the Particle Mesh Ewald (PME) method with a real-space cutoff of 1.2 nm. Van der Waals interactions were smoothly switched off between 1.0 and 1.2 nm using a force-switch function. All hydrogen-containing bonds were constrained using the LINCS algorithm, allowing a 2 fs integration timestep. The system was equilibrated at 310.15 K using the V-rescale thermostat with separate coupling groups for the protein-ligand complex, membrane, and solvent. Semi-isotropic pressure coupling was applied using the C-rescale barostat at 1 bar with a compressibility of 4.5 × 10⁻⁵ bar⁻¹ and a relaxation time constant of 5 ps. Positional restraints were applied during equilibration to the protein backbone, lipids, and dihedral angles.

We obtained three production replicas from three different initial structures. The 20 µs replica was run on Anton3 supercomputer by converting a 100 ns GRO file to a Desmond-compatible format using Viparr on Bridges2. The Anton3 simulation was performed using a multigrator integration scheme with a 2.5 fs timestep. Temperature was maintained using a Nose–Hoover thermostat at 310.15 K and pressure was controlled using the Martyna–Tuckerman–Klein (MTK) barostat with semi-isotropic coupling. Two independent 600 ns replicas were carried out using GROMACS. Ligand–protein interaction analyses were performed using MDAnalysis and PLIP 2.1.0, and molecular visualization was carried out using VMD 1.9.3.

Data availability

All experimental data have been deposited into open-access repository Open Science Framework (DOI: 10.17605/OSF.IO/HQG7D).

Acknowledgements

We thank Ardem Patapoutian (Scripps Research) and Tibor Rohacs (New Jersey Medical School, Rutgers University) for the gift of cells and plasmids. This work was supported by NIH grants GM130834 (to JJL and YLL), HL179692 (to JJL and YLL), AT013465 (to JJL), and the Anton3 allocation award MCB190044P (to JJL and YLL), supported by NIH grant GM154042 awarded to the Pittsburgh Supercomputing Center.

Additional information

Author Contributions

TDW, JJL, and YLL designed research; TDW, AC, AB, and JJL performed research and analyzed data; JJL wrote the paper. AC and AB contributed equally to this work.

Funding

HHS | NIH | National Institute of General Medical Sciences (NIGMS) (GM130834)

• Jerome Jacques Lacroix

HHS | NIH | National Heart, Lung, and Blood Institute (NHLBI) (HL179692)

• Jerome Jacques Lacroix

HHS | NIH | National Center for Complementary and Integrative Health (OAM) (AT013465)

• Jerome Jacques Lacroix