Structures of invertebrate PEZO-1 isoforms with a compact architecture and a dispensable pore-distal N-terminal blade

TL;DR

The study reveals that nematode PEZO-1 ion channels have a unique compact structure and can function without a specific part of the protein, suggesting evolutionary differences in how these channels work.

Contribution

The paper presents novel cryo-EM structures of C. elegans PEZO-1 isoforms and shows the N-terminal blade is not essential for mechanosensation.

Findings

PEZO-1G adopts a compact, semi-flattened conformation distinct from mammalian PIEZOs.

The pore-distal N-terminal blade is dispensable for mechanosensation in isoform K.

Different PEZO-1 isoforms impose distinct membrane curvatures, indicating evolutionary divergence.

Abstract

PIEZO channels are mechanosensitive ion channels conserved from plants to humans, yet structures exist for only a few mammalian orthologs. We define the structural and functional diversity of Caenorhabditis elegans PEZO-1, a single gene with extensive alternative splicing, by determining cryo-electron microscopy structures of three representative isoforms: G (full length), K (lacking the pore-distal N-terminal blade), and L (missing most of the blade). PEZO-1G displays mechanically evoked currents yet adopts a compact, semi-flattened conformation that significantly differs from the mammalian domes. The blades exhibit a three-step slope architecture stabilized by inter-blade latching among transmembrane helical units, yielding a circular, steering-wheel-like arrangement. A wider cap enables distinct blade-cap contacts that stabilize a “toggle-down” conformation. Isoform K also exhibits mechanically evoked currents, indicating that the pore-distal N-terminal blade is dispensable for mechanoactivation. Computational membrane-deformation modeling indicates that the isoforms impose distinct curvatures on the bilayer. Our findings indicate an evolutionarily distinct architecture for PEZO-1.

Bell et al. show that nematode PEZO-1 isoforms adopt a compact architecture distinct from mammalian PIEZOs, that the pore-distal N-terminal blade is dispensable for mechanosensation, and that isoforms bend membranes differently, indicating an evolutionarily divergent way to build mechanosensitive channels from a single, highly spliced gene.

Figures

H](#F1)). However, because this intermediate isoform naturally lacks the most N-terminal TM helical units (THU1–THU4, ~38% of the blade), the end of the blades adopts a more radially splayed arrangement reminiscent of mammalian PIEZO1 structures (Figure S1C). Yet the pronounced pivot around THU7 remains. In contrast, the structure of the shortest isoform PEZO-1L, at low resolution, retains only THU8 and THU9; in this structure, the cap appears more mobile and farther from the TM domain, with a longer cap-pore axial dimension (~15 Å) than in the other two isoforms, resembling a balloon tethered

E](#F2)). To quantify this, we measured the signed, perpendicular distance from each THU’s center of mass to the THU9 reference plane (THU9 was excluded from the fitting). A single-line fit to isoform G gives a shallow global slope (*b* = −2.53 Å/THU, *R*^*2*^ = 0.64). In contrast, a three-segment piecewise-linear model with fixed breakpoints after THU2 and THU6 is a better fit (*R*^*2*^ = 0.98) with segment slopes ∣*b*∣ = 6.85 Å/THU (THU1–2), 0.08 Å/THU (THU3–6; essentially flat), and 8.63 Å/THU (THU7–8). On the other hand, PIEZO2 traces a monotonic ramp away from the cap ([Figure 2F](#F2)):

D](#F3)). The broadened PEZO-1G cap engages with the blades via a helix-loop-helix segment (Y2125–K2144) projecting from one subunit to the extracellular loops of the neighboring subunit THU7 (TM27–TM28) and THU8 (TM31–TM32; [Figure 3G](#F3)). This interaction interface includes four hydrogen bonds (R2128–C1278′ backbone, T2127–S1288′, D2123 backbone-K1289′, and N2118–E1689′), a salt bridge (K2144–E1689′), and a cation-π contact (Y2125-K1289′). Conversely, PIEZO1 forms a single salt bridge, E2257–R1761′, between a shorter helix-loop-helix segment (termed Cap^α1-α2^)^[42](#R42)^ and the THU8 ex

D](#F4)). This configuration differs from PIEZO1 in its curved-closed, flattened non-conducting, and intermediate-open conformations, where the cap gate is lined by a mixture of polar and nonpolar residues (D2326, A2328, E2383′, P2382′) and lacks the side chain cis hydrogen bond. PEZO-1’s TM gate measures ~8 Å across V2366 (from each of the three subunits), and its IH packing matches the closed PIEZO1 conformation (~6.8 Å at V2476; [Figure 4D](#F4)). The cytosolic constriction neck, lined by a P2427 ring in PEZO1, has comparable dimensions to PIEZO1’s closed and intermediate-open conformations

D](#F5)).^[39](#R39)^ The mammalian variant PIEZO1.1, which naturally lacks the lateral portal plug, exhibits increased conductance and mechanosensitivity with reduced cation permeability.^[39](#R39)^ In PEZO-1G and 1-K, the portal-lining residue chemistry partially differs from that of the mammalian PIEZOs, yet the plugs resemble the curved-closed PIEZO1 conformation and PIEZO2 ([Figures 5E](#F5) and S12). This arrangement is consistent with the modest cation selectivity we previously reported for PEZO-1G.^[19](#R19)^ Overall, PEZO-1 adopts a configuration with a partially levered beam, ortho

Equations

Full text

INTRODUCTION

Mechanosensitive ion channels transduce mechanical forces into electrochemical signals across all kingdoms of life.^1-4^ Spanning multiple protein superfamilies and varying in sequence, topology, oligomeric state, gating, and reliance on accessory proteins,^3^ they regulate cellular homeostasis,^4-6^ sensory functions,^2,7-17^ organ physiology,^18-22^ organismal regulation,^23-28^ and development.^29^ In 2010, PIEZO1 and PIEZO2 were identified as distinct mechanically activated cation channels in mammalian cells.^30^ Most vertebrate and plant genomes encode two Piezo paralogs, whereas many invertebrates encode a single Piezo ortholog.^31,32^ Accordingly, the model organism Caenorhabditis elegans encodes a single Piezo ortholog, pezo-1. Unlike its mammalian counterparts, pezo-1 is dispensable for embryogenesis and larval viability, and loss-of-function mutants display only subtle phenotypes under standard laboratory conditions.^18,19,27,28^ The in vivo function of pezo-1 remained elusive for over a decade; however, recent work has implicated it in ovulation and pharyngeal pumping—roles without clear mammalian counterparts to date—as well as intestinal function and locomotion, reminiscent of mammalian PIEZO-dependent physiology.^7,15,18-20,22,27,28,33^ In all cases, pezo-1’s physiological roles are consistent with mechanosensation.

The expanding repertoire of pezo-1 functions in C. elegans underscores the value of this model for elucidating family-wide PIEZO principles and provides a platform to dissect conserved vs. divergent mechanisms. Consistent with mammalian PIEZOs, PEZO-1 meets canonical criteria for a mechanosensitive channel: it is necessary for in vivo mechanotransduction and can operate autonomously.^19^ Furthermore, gain-of-function (GOF) mutations that cause hemolytic anemia (PIEZO1) or distal arthrogryposis type V (PIEZO2) in humans cause behavioral defects in worms and confer the same slowed-inactivation gating phenotype when introduced at the equivalent positions in C. elegans PEZO-1.^19,28,34-38^ Mammalian PIEZO2 displays broad, cell-type-diverse splicing, whereas PIEZO1 shows limited, tissue-enriched cassette exon inclusion, with modest functional effects in both mammalian channels.^14,39^ In comparison, C. elegans pezo-1 is extensively alternatively spliced, generating multiple full-length and truncated isoforms (wormbase.org v. WS297).^18^ Previous work has shown that pezo-1 isoforms are not uniformly expressed; longer isoforms are found mainly in pharyngeal muscles, glands, intestine, and reproductive tissues, whereas shorter isoforms are enriched in pharyngeal and mechanosensory neurons. Consistent with this distribution, pezo-1 has been implicated in controlling pharyngeal pumping and particle transfer at the pharyngeal-intestinal valve, as well as in coordinating ovulation, sperm navigation, and the balance between crawling and swimming behaviors.^16,18,22,27^ However, how this extensive alternative splicing fine-tunes PEZO-1 channel function at the molecular level is still not well understood.

Although PIEZO channels are broadly conserved from invertebrates to vertebrates, structural data are currently limited to a small set of mammalian cryo-electron microscopy (cryo-EM) structures: PIEZO1 and PIEZO2 from mouse and human.^39-49^ These structures show a conserved trimer with propeller-like blades consisting of nine transmembrane helical units (THU1 to THU9), an extracellular cap over a central pore, and a beam-anchor-latch linkage that mechanically couples the blades to the conduction pathway. Each of these channels shapes the surrounding membrane into a semi-spherical dome, suggesting a tension-induced flattening mechanism that gates the pore, while lateral portals endow ion entry routes.^41,46^ PIEZO1 and PIEZO2 differ in cap architecture, blade curvature, and pore-neighboring elements, offering a structural basis for distinct inactivation kinetics.^40,41,43,45^ Unlike other ion channel families (e.g., the voltage-gated ion channel superfamily), intersubunit interfaces in PIEZO1 and PIEZO2 are concentrated in the central trimeric module, while the large peripheral blades show little, if any, intersubunit contact. Cryo-EM studies and coarse-grained molecular dynamics (CG-MD) simulations classify PIEZO1 into three blade- and pore-defined conformations: closed (curved), non-conductive (flattened), and intermediate open.^42^ Despite these advances, the gating cycle remains to be determined. With approximately 23% and 21% sequence identity to mouse PIEZO1 and PIEZO2,^50,51^ respectively, C. elegans PEZO-1 and its splice variants provide a tractable vantage point to separate conserved from specialized features across the PIEZO family and to connect structure to gating.

Here, using single-particle cryo-EM, we solved the structures of C. elegans PEZO-1 long (G), intermediate (K), and short (L) isoforms. The long isoform adopts a compact, semi-flattened architecture distinct from the dome-like curved shape reported for mammalian PIEZOs. The blades form a three-step slope stabilized by inter-blade contacts between transmembrane (TM) helical units, closing into a steering-wheel-like ring. Unexpectedly, the intermediate isoform, which lacks the most distal N-terminal blade, retains robust indentation-evoked currents and shows reduced inactivation. Herein, we use “distal” and “proximal” to refer to the positions of N-terminal elements relative to the pore (distal = pore-distal, proximal = pore-adjacent). CG-MD simulations indicate that PEZO-1G and PEZO-1K induce distinct membrane curvatures consistent with their architectures. Together, these data define functional PEZO-1 isoforms and an evolutionarily distinct PIEZO architecture that links blade-cap geometry to the mechanics of activation and inactivation.

RESULTS

Structure-function divergence among C. elegans PEZO-1 isoforms

According to WormBase (wormbase.org v. WS297), pezo-1 encodes 12 isoforms of varying length. Compared with mammalian PIEZOs, eight isoforms are full-length (A–H; ~2,420 amino acids [aa]), three are intermediate (I–K; ~30% shorter), and one is markedly truncated (L; ~60% shorter, Figures 1A and S1A). Although PEZO-1 shares only ~21%–23% aa identity with mammalian PIEZOs, the predicted monomer topology for the longest isoform is conserved (i.e., 38 TM segments, a long cytoplasmic beam helix, three anchor helices, an extracellular cap domain, and cytoplasmic N and C termini; Figure 1B). Each isoform retains the canonical pore module formed by the outer and inner helices. We therefore focused on representative isoforms from each class: G (long), K (intermediate), and L (short) (Figures 1A and 1B). To assess the function of the representative PEZO-1 isoforms, we expressed PEZO-1G, K, and L in N2A^Piezo1−/−^ cells (a mouse neuroblastoma cell line lacking Piezo1)^52^ and recorded whole-cell currents evoked by membrane indentation using a piezoelectric-driven glass probe. PEZO1-G mediates robust indentation-activated currents (Figures 1C and 1D), consistent with our prior characterization of PEZO-1G in insect (Sf9) cells.^19^ Remarkably, splice-mediated loss of 38% of the distal N-terminal blade (isoform K) preserves indentation-evoked gating. In contrast, PEZO-1L did not produce detectable mechanocurrents under these conditions, despite reaching the plasma membrane (Figure S1B). PEZO-1G and -K share similar activation thresholds (Figure 1E). Moreover, PEZO-1G has faster inactivation kinetics than PEZO-1K (Figure 1F). The isoform-specific inactivation and a non-conducting short variant motivated us to frame the central question of how the structural features of PEZO-1 isoforms tune both mechanical sensitivity and inactivation.

Using single-particle cryo-EM, we determined the structures of PEZO-1G, K, and L at overall resolutions of 3.5, 3.1, and >8 Å, respectively (Figures 1G-1I, S1, S2, S3, S4, S5, and S6). Isoforms G and K adopt the canonical trimeric, propeller-shaped architecture of mammalian PIEZOs,^40,45^ whereas isoform L adopts a two-tiered, turret-like assembly. Notably, PEZO-1G displays a conformation in which the blades latch circumferentially across subunits, yielding a compact “steering-wheel” ring that braces the cap (Figure 1G, top view). This closed wheel shape differs from the radially splayed arrangement seen in mouse PIEZO2 (EMD: 9975; Figure S1C).^45^ For PIEZO1, we cannot yet compare this radial arrangement because the N-terminal TM helical units (THU1–THU3) are unresolved in the available structures.^40,41,43,44,48^ Regardless, the closed steering-wheel ring is formed because the pivot around THU7 is more acute than in the mammalian counterparts, facilitating bending, latch formation, and compaction. Interestingly, PEZO-1G is in a conformation where the cap is stabilized in the toggle-down position (Figure 1G, side view), unlike mammalian PIEZOs (EMD: 9975 and EMD: 32592; Figure S1). The intracellular beam, a long helix running parallel to the membrane described in mammalian PIEZO1/2, is also present in PEZO-1G and -K, linking the TM blades to the central pore (Figures 1G and 1H, bottom view).

PEZO-1K maintains the core features described for PEZO-1G (Figure 1H). However, because this intermediate isoform naturally lacks the most N-terminal TM helical units (THU1–THU4, ~38% of the blade), the end of the blades adopts a more radially splayed arrangement reminiscent of mammalian PIEZO1 structures (Figure S1C). Yet the pronounced pivot around THU7 remains. In contrast, the structure of the shortest isoform PEZO-1L, at low resolution, retains only THU8 and THU9; in this structure, the cap appears more mobile and farther from the TM domain, with a longer cap-pore axial dimension (~15 Å) than in the other two isoforms, resembling a balloon tethered to a pole (Figures 1I and S6B). Taken together, PEZO1 isoforms show a distinct architecture with divergent structural features compared to mammalian PIEZO channels.

Structural basis of the “steering-wheel” ring in PEZO-1G

To place the “steering wheel” of PEZO-1G in a structural context, we examined the overall footprint of the longest isoform by quantifying its height (from the base of the cytoplasmic beam to the outermost extracellular point) and its mid-plane dome depth (from the top of the inner helix [IH] to the blade apex). PEZO-1G blades exhibit modest curvature, yielding a semi-flattened configuration with a height of 132 Å and a mid-plane dome depth of 49 Å (Figure 2A). This is in marked contrast to the published dome-shaped curvature of mouse PIEZO2 (PDB: 6KG7),^45^ whose height and mid-plane dome depths are 165 and 98 Å, respectively (Figure 2B). We then asked how the PEZO-1G ring maintains its circular continuity. The steering-wheel-like architecture arises from pivoting at THU7 and inter-blade latching in which THU1–THU2 of one subunit engage THU7–THU8 of the neighboring subunit, stabilizing a circle with a pore-to-THU1 radius of approximately 93 Å (Figures 2A and 2C). A closer inspection shows three discrete contacts: (1) the N-terminal pre-TM1 segment of THU1 in blade A packs against TM27 of THU7 in blade B; (2) TM4 of THU1 in blade A engages the TM27–TM28 interhelical loop of THU7 in blade B; and (3) TM5 of THU2 in blade A contacts TM31 of THU8 in blade B (Figures 2C and S7). Given that PEZO-1K also pivots at THU7 but lacks the latch, we infer that the kink is not induced by these contacts. Consistent with its extreme blade curvature (in the z axis) and limited sequence conservation (Figures S8 and S9), PIEZO2 lacks the THU7–THU8 kink and the inter-blade latches, yielding a broken circle with a larger radius of 139 Å and a distance of ~100 Å between THU1 and THU7 of the neighboring subunits (Figures 2B and 2D).

A distinguishing feature of the PEZO-1G blades is the presence of three discrete positional registers relative to the cap (Figure 2E). To quantify this, we measured the signed, perpendicular distance from each THU’s center of mass to the THU9 reference plane (THU9 was excluded from the fitting). A single-line fit to isoform G gives a shallow global slope (b = −2.53 Å/THU, R^2^ = 0.64). In contrast, a three-segment piecewise-linear model with fixed breakpoints after THU2 and THU6 is a better fit (R^2^ = 0.98) with segment slopes ∣b∣ = 6.85 Å/THU (THU1–2), 0.08 Å/THU (THU3–6; essentially flat), and 8.63 Å/THU (THU7–8). On the other hand, PIEZO2 traces a monotonic ramp away from the cap (Figure 2F): a single-line fit yields b = −10.26 AÅ/THU (R^2^ = 0.94), and a three-segment model offers only a modest improvement (R^2^ = 0.98 with the same breakpoints). Together, these results highlight inter-subunit contacts and a PEZO-1G “steering-wheel” ring arrangement that is semi-flattened.

The PEZO-1 cap-blade interaction stabilizes a toggle-down state

The extracellular cap, the largest soluble domain conserved across PIEZO orthologs, sits atop the TM pore tethered by a flexible linker previously termed the “spring linker” in PIEZO1 (Figures 3A and 3B, arrows).^40,42,43,53^ In PEZO-1G and -1K, this linker is relatively unstructured yet well-resolved. In contrast, the analogous region in mammalian PIEZO1 ranks among the lowest-resolution features outside the pore-distal blade tips, and the corresponding density is not observed in PIEZO2 (PDB: 7WLT and 6KG7; Figures 3B and 3C, respectively).^40,41,43,45^ The PEZO-1 spring linker is approximately 10% shorter than in PIEZO1, a difference that likely contributes to the toggle-down arrangement of the long and intermediate PEZO-1 isoforms.

We next defined the extracellular cap’s footprint and its interfaces with the neighboring TM blades to determine how it shapes PEZO-1 architecture. In PEZO-1G, the cap spans a larger surface area than its mammalian counterparts (321 vs. 247 nm^2^ in PIEZO1 and 265 nm^2^ in PIEZO2; Figures 3D-3F), thereby positioning the cap flanks for direct contact with neighboring subunits (Figure 3D). The broadened PEZO-1G cap engages with the blades via a helix-loop-helix segment (Y2125–K2144) projecting from one subunit to the extracellular loops of the neighboring subunit THU7 (TM27–TM28) and THU8 (TM31–TM32; Figure 3G). This interaction interface includes four hydrogen bonds (R2128–C1278′ backbone, T2127–S1288′, D2123 backbone-K1289′, and N2118–E1689′), a salt bridge (K2144–E1689′), and a cation-π contact (Y2125-K1289′). Conversely, PIEZO1 forms a single salt bridge, E2257–R1761′, between a shorter helix-loop-helix segment (termed Cap^α1-α2^)^42^ and the THU8 extracellular loop, whereas PIEZO2 lacks cap-blade contacts (Figures 3H and 3I). The naturally truncated isoform PEZO-1K, despite lacking THU1–THU4, retains a broadly similar cap footprint and interactions between the cap and blade (Figure S10A and S10B). THU7–THU8 remain positioned to engage the expanded cap, consistent with preserved function (Figures 1C and 1D), supporting the view that this coupling is sufficient to stabilize the cap even when the “steering-wheel” ring is incomplete. Collectively, PEZO-1 shows a larger cap footprint and a shorter spring linker, enabling extensive interactions that stabilize the cap domain in a toggle-down state.

Pore architecture of PEZO-1

To date, the ion permeation pathway of PIEZO1 has been captured in various conformational states, including closed, flattened non-conducting, and intermediate open. Analysis of the HOLE profiles of the worm PEZO-1G and -1K isoforms displays closely matched, likely non-conducting pores (Figures 4A and 4B). Differences are confined to small sections of the extracellular cap and TM domains, where K is wider by ~2.5 Å and ~1 Å at R2082 and V2351, respectively (Figure 4B, asterisks). The hydrophobic gate at V2366 (equivalent to V2476 in mouse PIEZO1, curved-closed state)^42,54,55^ forms one of the tightest constrictions in the permeation pathway with a radius of ~2 and 2.8 Å in PEZO1-K and 1-G, respectively. Moreover, positions N2301 at the top of the extracellular cap domain, and S2381, M2384, and P2427 at the constriction neck (equivalent to M2493 and P2526 in mouse PIEZO1)^40,41,43^ are also narrow, having a pore radius between 0.2 and 1.5 Å. Divergence from mouse PIEZO1 occurs in the central vestibule, extending from the cap gate to the constriction neck. In this region, PEZO-1G and PEZO-1K are more expanded than the PIEZO1 curved-closed (PDB: 7WLT) and flattened non-conducting structures (PDB:7WLU), yet narrower than the intermediate-open state (PDB: 8IXO; Figures 4A-4C). Relative to PIEZO2, PEZO-1G has a narrower cap gate and a wider TM gate (Figure S11). Although PEZO-1 contains expanded regions along the permeation pathway, the overall isoform pore profiles are consistent with non-conducting conformations.

We also directly compared structural features of PEZO-1G that have been deemed relevant in the various observed PIEZO1 conformations. In PEZO-1, the cap gate is formed by interfacial polar residues, Y2196 and N2201 from one subunit and Q2251′ and E2253′ from the neighboring subunit, including a Y2196-N2201 hydrogen bond within the same subunit (Figure 4D). This configuration differs from PIEZO1 in its curved-closed, flattened non-conducting, and intermediate-open conformations, where the cap gate is lined by a mixture of polar and nonpolar residues (D2326, A2328, E2383′, P2382′) and lacks the side chain cis hydrogen bond. PEZO-1’s TM gate measures ~8 Å across V2366 (from each of the three subunits), and its IH packing matches the closed PIEZO1 conformation (~6.8 Å at V2476; Figure 4D). The cytosolic constriction neck, lined by a P2427 ring in PEZO1, has comparable dimensions to PIEZO1’s closed and intermediate-open conformations but is narrower than the flattened non-conducting state (Figure 4D). Overall, PEZO-1 exhibits a closed architecture with a conserved hydrophobic gate and a distinct cap-gate chemistry, distinguishing it from mammalian PIEZOs.

Beam and cytoplasmic latch-plug positions in PEZO-1

The beam is a long intracellular helix, ~88 Å in length, that connects TM28 of THU7 to TM29 of THU8. It runs parallel to the membrane, sitting beneath the pore, and is proposed to participate in the force-transduction pathway that controls pore gating in mammalian PIEZOs (Figure 5A).^39,56^ In C. elegans, the beam length and angular orientation are preserved in isoform K relative to G, despite loss of distal blade segments (Figure 5B), indicating that its geometry is intrinsic to the beam itself and immediately adjacent structures, not to the distal portions of the blades. Across the PIEZO1 and PIEZO2 cryo-EM structures, beam orientation correlates with channel gating: it is levered (i.e., the distal end is tilted relative to the membrane plane, increasing the beam angle) in open-like PIEZO1 conformations and unlevered in the curved, closed PIEZO1 conformation and PIEZO2.^42,45^ In PEZO-1, the beam sits between closed and open-like geometries, exceeding the closed-state angle while remaining less levered at the distal end than open-like PIEZO1 (Figure 5B). Thus, even under the supposed tension-free conditions of a detergent micelle, PEZO-1 favors a partially levered intermediate rather than the unlevered closed state typical of PIEZO1 and PIEZO2 in detergent.

Since the beam is cytoplasmic and lies outside the membrane,^46^ we quantified its electrostatic surface potential. Across nematode and mammalian PIEZOs, positively charged residues are enriched at the proximal beam, whereas acidic residues cluster at the distal tip (Figures 5C and S9). However, the spatial distribution differs between the orthologs. In PEZO-1G and PEZO-1K, the proximal positive potential is more diffuse than in PIEZO1 and PIEZO2, whereas the distal negative potential is more tightly focused, forming an acidic cluster. These patterns suggest that the worm beam’s distal tip may preferentially recruit basic cytoskeletal proteins, while the more localized basic patches in mammalian beams may favor acidic partners, facilitating ortholog-specific electrostatic tuning of cytosolic tethers and force transmission.

The latch adjacent to the cytoplasmic end of the IH and the plug within the lateral portal beneath the pore have been proposed to modulate gating in PIEZO1 in an isoform-specific manner (Figure 5D).^39^ The mammalian variant PIEZO1.1, which naturally lacks the lateral portal plug, exhibits increased conductance and mechanosensitivity with reduced cation permeability.^39^ In PEZO-1G and 1-K, the portal-lining residue chemistry partially differs from that of the mammalian PIEZOs, yet the plugs resemble the curved-closed PIEZO1 conformation and PIEZO2 (Figures 5E and S12). This arrangement is consistent with the modest cation selectivity we previously reported for PEZO-1G.^19^ Overall, PEZO-1 adopts a configuration with a partially levered beam, ortholog-specific electrostatic patterning along the beam, and the latch-plug in a closed-like position.

Isoform-dependent membrane deformation by PEZO-1K and PEZO-1G

To gain molecular insight into how the plasma membrane may accommodate PEZO-1G and PEZO-1K, we embedded both isoforms in a physiological-like lipid bilayer, followed by solvation and neutralization with 140 mM NaCl before performing CG-MD simulations. Each system was equilibrated for 0.5 μs prior to production runs of 10 μs for each isoform, constrained to their structural conformation (Figures 6A, 6B, and S13). Notably, the simulations showed pronounced protein-induced membrane curvature for both isoforms. PEZO-1G induced a localized downward membrane deformation (i.e., toward the cytoplasm) primarily confined within the enclosed area of the “steering-wheel”-like assembly (Figures 6C and 6E; Video S1). In contrast, PEZO-1K induced membrane deformation at the central region bordering the pore, extending outward beyond the protein’s footprint (Figures 6D and 6E; see blue arrows, Video S2). Our CG-MD simulations indicate that PEZO-1G and PEZO-1K induce distinct membrane deformation profiles consistent with variations in their blade length and architecture.

DISCUSSION

Leveraging nature’s comprehensive toolkit, encompassing orthologs and splice variants, has proven a powerful tool for defining ion channel gating. Although most studies focus on mammalian mechanosensitive ion channels, exploring evolutionarily distant orthologs, such as C. elegans PEZO-1, offers naturally occurring perturbations that expose what is conserved, what diverges, and how individual domains contribute to channel function. Our work helps distinguish general principles that are difficult to infer from any single organism and distinguishes features that are fundamental to mechanotransduction from those that are lineage specific.

After PIEZO1 structures became available in detergent and liposomes,^41,43,46,47^ a model emerged in which flattening of the curved blades transduces membrane tension into pore opening and ion conduction. This is consistent with single-molecule experiments showing a correlation between blade expansion and activation.^56^ In this view, the intracellular beam acts as a pivot that relays distal blade conformations to the pore. Our results indicate that this framework needs to be revisited, as the intermediate isoform K, despite lacking THU1–THU4 (16 distal TM segments), preserves indentation-evoked activation, showing that distal blade segments are not required for opening. Interestingly, the large cap footprint in isoforms G and K, together with a shorter spring linker, promotes extensive contacts between the cap and the extracellular regions of the proximal THU7 and THU8, in contrast to mammalian PIEZOs. This arrangement places the cap in a toggle-down position and may contribute to the higher activation thresholds observed for the worm ortholog compared with the published thresholds of mammalian PIEZO channels.^57,58^ Our findings align with Lewis and Grandl’s report that an engineered, DTT-reversible disulfide tether between the cap and the adjacent blade abolishes mechanically evoked currents.^54^ Moreover, a recent structural work on a GOF PIEZO1 mutant challenges a periphery-to-center transduction model and instead supports a center-to-periphery mechanism.^42^ Although we do not resolve the directionality of force transmission to or from the pore, our data place the force sensor within the proximal blades (THU5–THU9) and the cap and pore domains.

The hallmark of PEZO-1G is a compact “steering-wheel” ring, a feature not anticipated by sequence analysis. Recent CG-MD simulations of full-length human PIEZO1 identified a lipid-enabled interblade “handshake” between elements near THU2 and THU8 from neighboring subunits.^59^ Phosphatidylinositol 4,5-bisphosphate stabilizes this contact by engaging clustered basic residues, drawing the blades together, and bending them into a semi-compact shape when compared to the “steering-wheel” shaped worm ortholog. In PEZO-1G, extensive contact between THU1–THU2 and THU7–THU8 (i.e., interblade latching) gives rise to the steering-wheel ring architecture. Our study, along with CG-MD simulations of human PIEZO1, supports critical THU interactions, albeit with different geometrical constraints. Smith and collaborators also determined that disrupting the handshake increases mechanical sensitivity.^59^ In contrast, the naturally truncated isoform K does not show enhanced mechanical sensitivity, as its activation threshold is comparable to that of isoform G and, under matched indentation steps, it inactivates more slowly. These differences may arise from ortholog variations or experimental conditions; nonetheless, the results indicate a clear path to evaluate the significance of this configuration across PIEZOs.

In PEZO-1G, the blades adopt a three-register architecture stabilized by inter-blade latching among transmembrane helical units, with discrete steps that likely mark mechanical pivots. In contrast, published PIEZO1 structures show curved (closed) or flattened (open-like) blades,^42,46^ with the latter displaying an almost flat proximal register. It is tempting to suggest that transitions between curved and flattened states proceed through a stepped geometry rather than a monotonic change. Future work should test whether this arrangement observed in PEZO-1 is also visited by PIEZO1 and PIEZO2 as a three-register state during gating.

Which elements of PEZO-1 determine the rate of inactivation? PEZO-1G inactivates faster than PEZO-1K, as well as all mammalian and Drosophila channels characterized to date.^19,30,52,54^ Three features distinguish the long G isoform from the intermediate K isoform and other orthologs: (1) interblade latching that completes the circumferential ring, (2) two cap-proximal pore regions that are approximately 1–2.5 Å narrower in radius than in isoform K, and (3) a more enclosed membrane deformation. Rather than pointing to a single site, the cumulative functional evidence highlights inactivation as a multi-locus process, with GOF mutations across the blade, beam, cap, and pore of PIEZO1 and PIEZO2 typically converging on the same kinetic outcome (i.e., slower inactivation).^34,36,37,57,60-63^ There is also evidence that the membrane composition can modulate the inactivation rate of PIEZO channels.^38,57,58,64^ Our computational analysis suggests that PEZO-1K induces a membrane deformation that expands radially outward from the pore, extending beyond the blades. It is tantalizing to speculate that this redistribution of membrane deformation observed between G and K isoforms might play a role in the functional differences we determined in inactivation.

Across available PIEZO1 structures, various landmarks have been used as readouts of functional state (Figure 7).^41-43,47^ A primary indicator of a non-conducting pore is the hydrophobic gate centered on V2476 in mouse PIEZO1, which is narrow in closed-like structures. Closed-state features also include constrictions at the cap gate and the intracellular constriction neck. Cytoplasmic geometry can be informative as well; the position of the beam relative to the membrane plane varies across closed, open-like, and inactivated conformations, with a more acute angle commonly assigned to non-conducting states. In addition, it has been proposed that the position of the lateral latch-plug tracks ion channel conductance.^39^ In PEZO-1, several features align with a closed pore: a narrow hydrophobic gate at V2366 and a narrowed intracellular constriction neck. In contrast, the cap vestibule and beam tilt are closer to an intermediate-open geometry than to the closed or non-conducting states. Although PEZO-1 structures retain the lateral latch-plug found in full-length PIEZO1, its conductance matches the plugless isoform PIEZO1.1 (~40 pS).^19,39^ In conclusion, the conformations in our PEZO-1 structures feature a closed pore with cytoplasmic elements occupying an intermediate position between the closed and intermediate-open conformations.

Our findings define ortholog- and isoform-specific architectural solutions for mechanogating within the PIEZO family, identifying cap-blade coupling and the proximal, rather than the distal N-terminal blade, as the principal determinants of mechanical sensitivity. By placing C. elegans PEZO-1 alongside mammalian PIEZO1 and PIEZO2, we leverage natural evolutionary variation to distinguish core from context-dependent features of force transmission and pore control. Our work provides insights into how PIEZO channels capture membrane force and route it to (or from) the pore, with implications for interpreting disease variants, engineering force sensors, and targeting channel mechanics.

Limitations of the study

Our study is primarily focused on the structural characterization of representative C. elegans PEZO-1 isoforms and their comparison with mouse PIEZO1 and PIEZO2, in order to place these architectures in a broader mechanistic and evolutionary context. Although we provide functional evidence that the long and intermediate isoforms can form mechanosensitive channels when activated with a displacement clamp, we did not undertake a systematic structure-function analysis. For example, we did not obtain comprehensive displacement-response relationships for each isoform. However, a recent publication characterized the pressure-response relationships of isoforms G, K, and L.^65^ Furthermore, we did not perform state-dependent kinetic analyses under different lipid conditions or targeted mutagenesis of specific interfaces (e.g., interblade-latch, blade-cap contacts) to test their causal contributions to gating and inactivation. As a result, our data do not yet allow us to assign definitive mechanistic roles to structural features in the activation and inactivation transitions. Rather than providing a structure-function and physiological characterization, the goal of this study is to establish the architectures of representative PEZO-1 isoforms and to highlight domains and interfaces that are strong candidates for controlling mechanoactivation and inactivation. The structures and comparative analyses presented here therefore serve as a framework for future targeted structure-function experiments and in vivo studies aimed at understanding how specific PEZO-1 domains and contacts tune channel function.

Star★Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell culture

Piezo1-knockout Neuro2a (N2A^Piezo1−/−^) cells were used in this study.^52^ Cells were grown on polystyrene cell culture plates (ThermoFisher Scientific) in a 37°C humidified incubator with 5% CO_2_. Growth medium was exchanged 2–3 times per week and was composed of DMEM/F12 (ThermoFisher Scientific) with 10% fetal bovine serum (ThermoFisher Scientific). Cells were passaged once they reached 90% confluency, twice per week, typically at ratios between 1:3 and 1:20. HEK293S GnTI^−^ cells were adapted to suspension culture in FreeStyle293 medium (Invitrogen) supplemented with 2% FBS (Gibco) and maintained at 37°C with 8% CO_2_ in an orbital shaker at 120 rpm. HEK293 cells cultured in 6-well plates were infected at ~70% confluency with 10% P2 baculovirus. Cell lines have not been authenticated. Cell lines tested negative for mycoplasma contamination.

METHOD DETAILS

Channel constructs

For electrophysiology. C. elegans pezo-1 isoforms G, K, and L were codon optimized for mammalian cells and cloned into the pMO vector, a pcDNA3.1-based backbone containing the β-globin 5′ and 3′ untranslated regions, with an IRES-driven GFP cassette to express GFP as a separate cistron (Epoch Life Science). For protein purification experiments, pezo-1 isoforms G and K were subcloned into a pEG BacMam vector. An N-terminal fusion cassette, comprised of a Kozak sequence, MBP (maltose-binding protein), and a TEV protease cleavage site, was inserted under the CMV promoter to facilitate efficient protein expression and purification using amylose resin. For isoform L, the gene, as well as the MBP and TEV cleavage site, was subcloned into a pFastbac vector for insect cell expression.

Transfection

N2A^Piezo1−/−^ cells (^52^) were transfected using Lipofectamine 3000 (Thermo Fisher Scientific), according to the manufacturer’s instructions. Briefly, 25 μL OptiMEM (ThermoFisher Scientific) was combined with 1.5 μL Lipofectamine 3000 per transfection. 26.5 μL of that mixture was added to a separate mixture of 25 μL OptiMEM, 1 μL P3000, and 500 ng of pezo-1-containing plasmid. This transfection mixture was incubated for 15–20 min at RT and then added to one well of a 24-well plate containing pre-plated cells at approximately 70% confluency. Cells were incubated overnight, then checked for green fluorescent reporter expression the following day before being split into assay-appropriate vessels.

Electrophysiology

The extracellular solution consisted of 133 mM NaCl, 3 mM KCl, 2.5 mM CaCl_2_, 1 mM MgCl_2_, 10 mM HEPES, 10 mM glucose, pH adjusted to 7.3 with NaOH, then osmolarity adjusted, if necessary, with sucrose solution to 310 mOsm. The intracellular solution consisted of 133 mM KCl, 5 mM EGTA, 1 mM CaCl_2_, 1 mM MgCl_2_, 10 mM HEPES, 4mM Mg-ATP, 0.4mM Na-GTP, pH adjusted to 7.3 with KOH, then osmolarity adjusted, if necessary, with sucrose solution to 300 mOsm. Indentation-evoked currents were recorded from transiently transfected N2A^Piezo1−/−^ cells with green fluorescence. N2A^Piezo1−/−^ cells that had been incubated with transfection solution for 24 h were split and plated onto 35 mm polystyrene dishes the day prior to whole-cell recording. Immediately before patching, the 35 mm dish was washed with 1 mL of extracellular solution, filled with 1.5–2 mL of extracellular solution, and placed on an inverted Olympus IX73 microscope. Patch pipettes were made from borosilicate glass with an outer diameter of 1.5 mm, inner diameter of 1.1 mm, and length of 7.5 cm using a Sutter P-1000. The patch pipettes were polished with a micro forge MF-830 to a resistance of 2.0–4.5 MΩ. Currents were recorded in whole-cell patch clamp mode at −80mV using a MultiClamp 700B. Signals were digitized using a Digidata 1550 (Axon Instruments) sampled at 100kHz and low-pass filtered at 10 kHz. The indentation stimulus, consisting of 1 μm incremental indentation steps, each lasting 200 ms with a 2-ms ramp in 30-s intervals, was delivered by a blunt glass probe positioned on the cell at an ~60° angle and indentation was driven by a micromanipulator-mounted P841.20 piezoelectric translator.

Electrophysiology analysis

Recordings were analyzed in ClampFit 11.4.1. Recordings with leak currents > 300 pA and an access resistance >20 MΩ were excluded from analyses. Leak currents, before mechanical stimulation, were subtracted offline. Peak PEZO-1 currents were calculated as the maximum current, during the start of the indentation stimulus, among all traces captured for a given cell. The threshold of mechano-activated currents was defined as the indentation step that evoked the first sweep with an amplitude above 30 pA. The tau (τ) of inactivation was calculated by fitting the standard exponential function in ClampFit to the current trace between the maximum current and its return to the baseline.

Immunofluorescence

After 48 h of baculovirus infection, HEK293 cells were transferred onto glass coverslips pre-treated with poly-L-lysine and allowed to attach for 4 h. The samples were then washed three times with PBS, incubated for 5 min at RT with 100 ng/mL FAST DiI (Invitrogen, D7756) dissolved in PBS to label the plasma membrane, and then washed again three time with PBS. Immunofluorescence was subsequently performed as follows. All reagents and wash steps were carried out in PBS. Briefly, cells were fixed with 4% paraformaldehyde for 10 min at 37°C, washed three times, permeabilized with Triton X-100 for 15 min at RT, and washed again. Blocking was performed with 2% BSA for 1 h at RT. Cells were incubated overnight at 4°C with a primary antibody against MBP (sc-13564, Santa Cruz Biotechnology), washed three times with PBS, and subsequently incubated with an Alexa Fluor 488-conjugated anti-mouse secondary antibody (715-547-003, Jackson ImmunoResearch) for 1 h at RT. Both primary and secondary antibody solutions were supplemented with 2% BSA. After three final washes, fluorescence images were obtained. z stack images were acquired using a Revolve microscope (Echo) equipped with FITC and Texas Red filters. Images were processed in ImageJ. Background signal was removed, and a z-projection was generated by summing slice intensities. An FFT bandpass filter (cutoff: 3–40 pixels) was applied to enhance cell contours, followed by contrast adjustment. Finally, the membrane and protein-labeled channels were merged.

PEZO-1 isoforms expression and purification

For PEZO-1G and PEZO-1K, HEK293S GnTI^−^ cells were infected at a density of 2.3 × 10^6^ cells/mL with 2% P2 baculovirus, generated using the Bac-to-Bac Baculovirus Expression System (Invitrogen), according to the manufacturer’s instructions. After 18 h, 10 mM sodium butyrate (ThermoScientific) was added to enhance protein expression, and the incubation temperature was reduced to 30°C without CO_2_. Cells were collected 72 h post-infection by centrifugation at 4,000 g for 15 min at 4°C and resuspended in ice-cold lysis buffer (150 mM NaCl, 2 mM DTT, 50 mM HEPES, pH 7.4, supplemented with 1 mM PMSF, 1 μg/mL pepstatin, 3 μg/mL aprotinin, and 3 μg/mL leupeptin). Cells were homogenized using a high-pressure homogenizer (Avestin) and centrifuged at 3,000 g for 15 min at 4°C to remove inclusion bodies and cell debris. The supernatant was ultracentrifuged at 125,000 g for 55 min at 4°C, and the resulting membrane pellet was resuspended in lysis buffer, flash-frozen in liquid nitrogen, and stored at −80°C until use. Membranes were thawed and solubilized in 1% (w/v) digitonin (GoldBio) for 2 h with gentle rotation. All purification steps were performed at 4°C. Insoluble material was removed by ultracentrifugation at 125,000 g for 55 min, and the supernatant was incubated with amylose resin (New England Biolabs) for 2 h. The resin was washed with 10 column volumes (CV) of buffer containing 300 mM NaCl, 20 mM Tris (pH 8.0), and 0.05% digitonin, and proteins were eluted with buffer containing 150 mM NaCl, 20 mM Tris (pH 8.0), 0.05% digitonin, and 20 mM maltose. Eluted proteins were digested with TEV protease (New England Biolabs) to remove the MBP tag and further purified by size-exclusion chromatography on a Superose 6 column (Cytiva), equilibrated with 150 mM NaCl, 20 mM Tris (pH 8.0), and 0.05% digitonin. Purified proteins were concentrated to 0.1–0.4 mg/mL using Amicon Ultra centrifugal filters (100 kDa MWCO; Millipore). PEZO-1L was expressed in SF9 cells. Cells were infected and harvested after 72 h and processed similarly to PEZO-1G and PEZO-1K, with some modifications. Membranes were solubilized in 2 mM DDM, resin was washed with 150 mM NaCl, 50 mM HEPES (pH 7.4), 2 mM DTT, and 1 mM DDM, and the protein was eluted with wash buffer supplemented with 20 mM maltose.

Cryo-EM sample preparation

Cryo-EM grids were prepared by applying 3 μL of purified PEZO-1 isoform G, K, or L to glow-discharged 200-mesh copper grids with a 2 nm continuous carbon layer (R 2/1, Quantifoil Micro Tools GmbH). Grids were vitrified in liquid ethane using a Vitrobot Mark IV (ThermoFisher Scientific) operated at 4°C and 100% humidity, with a 10 s wait time, blot force of −7, and 3.5 s blot time.

Cryo-EM data collection and image processing

Cryo-EM data were acquired using a Titan Krios G3 TEM microscope (ThermoFisher Scientific) operating at 300 keV and equipped with a BioQuantum energy filter (Gatan Inc.). Images were recorded at a nominal magnification of 130,000× on a K2 Summit direct electron detector (Gatan Inc.), yielding a calibrated pixel size of 1.08 Å. Data were collected over a defocus range of −0.8 to −2.2 μm. Each movie stack was recorded over 7 s, split into 35 frames, with a total accumulating electron dose of ~50 e^−^/Å^2^. Automated data acquisition was carried out using EPU software (ThermoFisher Scientific). Detailed acquisition parameters are listed in Table S1.

Image processing was performed using CryoSPARC (Structura Biotechnology Inc.).^74^ Motion correction and CTF estimation were carried out using Patch Motion and CTF workflows. For all isoforms, particles were picked using the blob picker. Following 2D classification and ab initio reconstruction, particles were subjected to multiple rounds of 2D and 3D classification, heterogeneous refinement, and non-uniform refinement. Particles were also subjected to local CTF refinement and reference-based motion correction. Final map resolutions were estimated using the gold-standard (FSC = 0.143) criterion. Local resolution estimation and orientation diagnostics were also carried out in CryoSPARC.

Model building refinement and validation

Initial model generation and map fitting

A full-length model of the PEZO-1 G isoform was generated with AlphaFold3 (https://alphafoldserver.com/, accessed May 2025).^77^ This model was then fit to the cryo-EM density map for PEZO-1G using UCSF ChimeraX’s (version 1.9)^73^ “Fit in Map” tool, where the cap, transmembrane (TM) region, THU8, and THU9 showed the best initial agreement.

Manual rebuilding and real-space refinement

Local refinement of the cap/TM/THU8/THU9 domain was performed in Coot^75^ to optimize side-chain/backbone placement against the density, followed by real-space refinement in Phenix v1.21 (phenix.real_space_refinement)^72^ with non-crystallographic symmetry (NCS) constraints to maintain the 3-fold protomer relationship. To obtain the full-length G model, we isolated models for the individual domains corresponding to THU1–7 and the beam from the initial AlphaFold model. Individual domains were then manually fit to the density map in ChimeraX using observed α-helices as landmarks. This individual domain registration was then optimized for fit to density with Chimera X’s Fit in Map tool. Again, the entire model was then iteratively optimized and refined to the density map using Phenix and Coot with NCS constraints. Regions lacking interpretable density were conservatively truncated.

To obtain the K isoform, the refined G model was truncated (removing THU1–THU4) and rigidly fit to the K density map in ChimeraX, followed by Coot/Phenix refinement with NCS constraints as above. For the L isoform, a further truncated G model (removing THU1–THU7 and the beam) was fit to the L density map. While the L map was lower resolution (>8Å), the cap density was clearly displaced relative to the TM domain. Accordingly, the cap domain (residues 2074–2342) was fit independently of the rest of the model, which indicated an approximately 15 Å cap shift relative to the TM helices. Final whole-model refinement/validation for each isoform was performed in Phenix.

Reference structures and alignment

For cross-family comparisons, we analyzed our PEZO-1 G/K models together with published PIEZO1 and PIEZO2 structures. Prior to geometric analyses, all models were aligned to a common frame by rigid-body superposition on the TM core to minimize overall rotational/translation differences.

For each domain, the center of mass (COM) was computed from all atoms in that domain. Domain definitions (cap, beam, THU1–THU9, etc.) were derived from the published structures. When a residue range was undefined or truncated for a given isoform, that domain was excluded from statistics for that isoform. Hydrogen atoms, when present, were excluded for consistency across structures. COMs were computed per protomer and then summarized across the 3-fold assembly.

Model comparison

To quantify axial “registers,” we used the THU9 COM as a reference. For each structure, the three THU9 COMs (one per protomer) define a best-fit plane, $[eqn]$ . Let $[eqn]$ be the unit normal to $[eqn]$ oriented from the cap toward the cytosol, and $[eqn]$ point on $[eqn]$ . For any domain COM $[eqn]$ , the signed perpendicular distance to $[eqn]$ is

$$[eqn]$$

By convention, positive values point away from the cap. THU9 itself was excluded from fits using this metric.

For completeness, we also report two angular descriptors relative to $[eqn]$ .

Tilt $[eqn]$ : the polar angle between the vector $[eqn]$ and the plane normal $[eqn]$ :

$$[eqn]$$

Azimuth $[eqn]$ : the in-plane angle of $[eqn]$ projected onto $[eqn]$ , measured from a fixed in-plane axis $[eqn]$ (defined by the projection of the cap-to-TM vector) with right-handed $[eqn]$ :

$$[eqn]$$

To capture axial trends along the blade, we regressed distance vs. THU index using ordinary least squares in a single and three segment model.

Single-line model: $[eqn]$ , where $[eqn]$ is the THU index (e.g., $[eqn]$ , with THU9 excluded when THU9 defines $[eqn]$ ). The slope $[eqn]$ is reported in Å/THU.Three-segment model: fixed breakpoints at THU2 and THU6 (chosen a priori). Independent least-squares fits were performed for THU1–2, THU3–6, and THU7–9, yielding segment slopes $[eqn]$ , $[eqn]$ , $[eqn]$ in Å/THU.

The goodness-of-fit $[eqn]$ for the single and three-segment lines is reported as

$$[eqn]$$

where $[eqn]$ are model predictions and $[eqn]$ is the mean of the observed distances. We report R^2^ for the global single-line model and, for the three-segment model, the overall R^2^ (aggregating residuals across segments). When segment-level R^2^ is shown, it refers to the fit within that segment only.

CG simulations

Structures of PEZO1G and PEZO1K were converted to the Martini3 (CG) representation using Martinize2/Vermouth (vermouth-0.10.1.dev93).^78^ Missing subdomains were not included. Secondary structure information required for the Martini CG mapping was calculated with DSSP integrated within MDTraj.^69^ For both isoform systems, elastic network bonds were implemented (k = 700 kJ·mol^−1^·nm^−2^; distance boundaries: 5–9 Å), and elastic bonds between three monomeric subunits were not removed. The CG systems were constructed using Insane (version 1.2.0),^79^ incorporating a lipid bilayer composition of asymmetric mammalian plasma membrane.^80^ The exact lipid composition of outer leaflet include: 15% POPC, 20.5% PLPC, 2.8% PAPE, 2.8% POPE, 10.3% SSM, 10.3% NSM, 3.7% CMH, and 34.6% cholesterol, while for the inner leaflet the lipid composition is 7% POPC, 11% PLPC, 12% PAPE, 14% POPE, 5% POPI, 11% PAPS, 1% POPA, 5% SSM, 5% NSM, 29% cholesterol. The Insane flags -ring and -fudge 0.2 were used to achieve correct lipid placement near the three-blade domains of both isoforms. Regular-sized water beads (W) were added to achieve box dimensions of approximately 260 × 260 × 220 Å^3^. Monovalent ion beads (Na^+^ and Cl^−^) were included to neutralize the system and achieve a final ionic strength of 140 mM. The vertical (z axis) placement of the protein within the membrane was adjusted according to predictions from the positioning of proteins in the membrane (PPM) 3.0 server.^81^

CG-MD simulations were conducted using GROMACS version 2024.2^(67)^ with the Martini3 New-Reaction-Force (RF) parameter set.^68^ All production simulations employed a 20-fs integration timestep and utilized RF electrostatics. Pressure was regulated at 1 bar semi-isotropically (compressibility: 3 × 10^−4^ bar^−1^) using the Berendsen barostat^82^ for equlibration (0.5 μs) and Parrinello-Rahman barostat^83,84^ (coupling constant: 4 ps^−1^) for the production simulations (10 μs), while the temperature was maintained at 310 K using the velocity-rescale (v-rescale) thermostat.^85^ To prevent simulation artifacts associated with infrequent neighbor list updates,^86^ Verlet-buffer-tolerance was disabled, a conservative neighbor list cutoff of 1.5 nm was adopted, and neighbor searching (nstX variables) was set to occur every 20 simulation steps, in accordance with recommendations for Martini3 New-RF. System equilibration followed the following steps: (1) 500 steps of position-restrained energy minimization in vacuum, (2) 500 steps of position-restrained minimization after solvating the system with water, ions, and lipids, (3) 10 ps of position-restrained NVT equilibration with a 2-fs timestep, (4) subsequent NpT equilibration phases without positional restraints, lasting 25 ps (5 fs timestep), 50 ps (10 fs timestep), 300 ps (20 fs timestep), and 500 ns (20 fs timestep), respectively. Following equilibration (0.5 μs), production simulations were extended to 10 μs (Table S2), with atomic positions recorded every 100 ps for analysis.

Visualization and analysis tools

VMD^70^ was used to analyze trajectories, render molecular images, and create videos. To characterize the spatial organization of membrane components, lipid density maps, and membrane curvature data were generated from production run trajectories. Trajectories were centered and aligned to the protein using only rotations around the membrane normal (z axis). Average volumetric densities of lipid headgroups (PO4) and cholesterol hydroxyl groups (OH) were then computed using the VMD VolMap tool with a grid resolution of 1 Å, resulting in OpenDX-formatted volumetric density files. The resulting volumetric density maps were analyzed and visualized using VMD. A circular $[eqn]$ mask of radius $[eqn]$ centered on the protein was applied. A circular $[eqn]$ mask of radius $[eqn]$ centered on the protein was applied. Mean curvature was computed using the MDAnalysis MembraneCurvature toolkit.^71,87^ All curvature density maps were plotted using a global scale.

QUANTIFICATION AND STATISTICAL ANALYSIS

Data were plotted and statistically analyzed using GraphPad InStat software (version 3.10; GraphPad Software Inc.). Data shown are mean ± SEM. Individual tests, samples size, and exact p values are reported in each of the figure legends. Statistical significance was represented as **p < 0.01, and n.s. not significant. No statistical method was used to predetermine the sample size. Only electrophysiology data that did not meet the criteria above were excluded from the analyses. The experiments were not randomized. The investigators were blind to DNA plasmids whenever possible. Experiments were performed at least three times on different days from different/independent preparations.

Supplementary Material

1

2

3

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.116878.