Membrane potential mediates the cellular response to mechanical pressure

Abstract

Mechanical forces influence cellular decisions to grow, die, or differentiate, through largely mysterious mechanisms. Separately, changes in resting membrane potential have been observed in development, differentiation, regeneration, and cancer. We demonstrate that membrane potential is an important mediator of cellular response to mechanical pressure. We show that mechanical forces acting on the cell change cellular biomass density, which, in turn, alters membrane potential. Membrane potential then regulates cell number density in epithelia by controlling cell growth, proliferation, and cell elimination. Mechanistically, we show that changes in membrane potential control signaling through the Hippo and mitogen-activated protein kinase (MAPK) pathways and potentially other signaling pathways that originate at the cell membrane. While many molecular interactions are known to affect Hippo signaling, the upstream signal that activates the canonical Hippo pathway at the membrane has previously been elusive. Our results establish membrane potential as an important regulator of growth and tissue homeostasis.

Keywords: Hippo; YAP; biomass density; growth control; mapk; mechanotransduction; membrane potential; tissue homeostasis.

Figure 1.. Membrane potential is a sensor and regulator of tissue density

(A) MDCK cells were imaged at 0, 2, and 4 days after reaching confluence with increasing cell number density. Images of cell nuclei and cellular biomass density (in units of g/mL) measured via normalized stimulated Raman spectroscopy (NoRI) are shown. Scale bar: 25 μm. (B) Protein density dependence on cell number density. Cell number density (quantified from Raman: n = 14 regions of interest (ROIs) total from 3 biological replicates from experiment 1 and n = 24 ROIs each from 3 biological replicates from experiment 2; Tomocube: n = 10 ROIs) and protein density (Raman experiment 1: n = 24 cells each from 3 biological replicates and experiment 2, n = 30 cells each from 3 biological replicates, Tomocube: n = 20 cells), measured using NoRI and Tomocube. (C) Biomass density dependence on cell number density from two separate experiments over 5 days, measured using NoRI. Quantified from same ROIs and cells as (B). (D) Images of biomass density from 0, 2, and 4 days of tissue simulation. (E) Tissue simulation recapitulates the experimentally observed relation of biomass density/protein density as a function of number density from 0, 2, and 4 days of confluence (mean ± SD, n = 4 simulations). (F) Images of membrane potential as measured by dye DiOC₂(3), with increasing cell number density. Scale bar: 25 μm. (G) Membrane potential hyperpolarizes with increasing cell number density. Measured by membrane-potential-responsive dye DiOC₂(3) and Voltron biosensor. (Number density of DiOC₂(3) experiment: n = 15 ROIs total from 2 biological replicates; Voltron experiment: number density is binned in intervals of 8.5 cells/100 μm² from a total of 79 ROIs. Membrane potential of DiOC₂(3) experiment: n = 30 cells total from 2 biological replicates; Voltron: n = 8,964 cells total across 79 ROIs, error bars represent SEM from each bin). (H) Images of membrane potential from 0, 2, and 4 days of tissue simulation. (I) Tissue simulation showing hyperpolarization of membrane potential with increasing cell number density. (Mean ± SD, n = 4 simulations.) (J) NoRI images of MCF10A cells at 0, 2, and 4 days after reaching confluence. Scale bar: 25 μm. (K) Biomass density increases with number density in MCF10A cells, measured by NoRI. (Number density quantified from n = 10 ROIs each from 3 biological replicates, and biomass density quantified from n = 24 cells each from 3 biological replicates.) (L) Images of MCF10A nuclei and membrane potential dye DiOC₂(3) from colony edge to colony center showing gradual hyperpolarization in dense regions. Scale bar: 50 μm. (M) NoRI images of EpH4-Ev cells at 0, 2, and 4 days after reaching confluence. Scale bar: 25 μm. (N) Biomass density increases as a function of cellular confluence in EpH4-Ev cells, quantified by NoRI. (Number density quantified from n = 10 ROIs each from 3 biological replicates and biomass density quantified from n = 24 cells, each from 3 biological replicates.) (O) Images of EpH4-Ev nuclei and membrane potential dye DiOC₂(3) from colony edge to colony center, showing gradual hyperpolarization in dense regions. Scale bar: 50 μm. (P) NoRI images of human primary mammary epithelial cells (HMECs) the day before confluence, the day of confluence, and every 2 days after confluence, show linear increase in biomass density. Scale bar: 25 μm. (Q) Quantification of biomass density from 3 individual biological replicates over time (n = 24 cells per biological replicate, error bars indicate SD). (R) Membrane potential dye images of a HMEC colony from edge to center with DiOC₂(3) and DiSBAC₂(3). The two dyes show that cells at the center of the colony are hyperpolarized compared with cells at the edge (by the dye DiOC₂(3), which increases in fluorescence with hyperpolarization), and cells at the edge of the colony are depolarized compared with cells at the center (by the dye DiSBAC₂(3), which increases in fluorescence with depolarization). Scale bar: 100 μm. (S) Quantification of the fluorescence intensity of the two dyes as a function of distance from colony edge, as shown in (R) (n = 8 rectangular ROIs with width 100 μm from 2 biological replicates, error bars indicate SD). (T) Feedback control loop between biomass growth and membrane potential is crucial for epithelial homeostasis. (U) NoRI images of overnight ouabain treatment of MDCK cells. Scale bar: 25 μm. (V) Overnight ouabain treatment (100 nM) showing significant biomass addition in the treated condition compared with vehicle control. (Number density quantified from n = 24 ROIs each from 2 biological replicates, and biomass density quantified from n = 24 cells each from 2 biological replicates). (W–Z) Effects of knocking down the α subunit of the sodium-potassium pump (ATP1A1). (W) Images of membrane potential measured by DiOC₂(3) reveal that ATP1A1 knockdown depolarizes MDCK cells. (X) Quantification of DiOC₂(3) fluorescence intensity after knockdown of ATP1A1 (n = 24 cells each from 3 biological replicates). (Y) NoRI images of control and ATP1A1-KD cells. (Z) Cellular biomass density and cell number density upon knockdown of ATP1A1 compared with wild-type control. (Number density quantified from n = 24 ROIs each from 2 biological replicates, and biomass density quantified from n = 24 cells each from 2 biological replicates). Scale bar: 25 μm. ****p < 0.0001. Data are represented as mean ± SD unless otherwise noted. See also Figures S1 and S2.

Figure 2.. Mechano-electro-osmotic model

Top, the model combines three elementary conservation laws: flux balance across the plasma membrane, mechanical force balance, and charge balance. Left, a cell under low mechanical pressure; right, a cell under high mechanical pressure. Cellular biomass, such as protein and RNA, carries a net negative charge. Cells resist compression of their volume via cytoplasmic osmotic pressure that counteracts and balances external forces (force balance). This osmotic pressure originates from an intracellular concentration of counterions that are attracted to balance negatively charged macromolecules (charge balance). As illustrated on the right-hand side, higher mechanical pressure acting on the cell is thus mostly balanced by a higher concentration of counterions. Higher intracellular ion concentrations result in steeper ion concentration gradients across the membrane, as illustrated in the right inset. Thus, with constant ion channel abundances, a steeper concentration gradient across the membrane results in an augmented diffusive efflux of ions (Fick’s law). With constant active ion transport flux (via sodium-potassium-ATPase), this increased diffusive ion efflux results in a buildup of net negative charge in the cytoplasm and a corresponding membrane potential that counteracts the diffusive efflux of positively charged ions until flux balance is achieved (flux balance for each ion species). Bottom: a one-to-one correspondence between mechanical pressure, cytoplasmic biomass density, and membrane potential emerges from the mechano-electro-osmotic equilibrium illustrated in the top panel, allowing membrane potential to act as the primary mediator of the cellular response to mechanical pressure.

Figure 3.. Cells detect and respond to mechanical forces via changes in membrane potential

(A) Schematic diagram depicting the experimental design, with monolayers grown on stretchable membrane. Left: cells were grown to confluence and a 20% uniaxial stretch was applied. Right: cells were grown to confluence on a pre-stretched membrane, then the stretch was released to induce compression. (B) Images of membrane potential dye FluoVolt on MDCK cells before and after stretching. MDCK cells sometimes form domes that lose basal attachment, and one such dome overlaps with the image. Scale bar: 25 μm. (C) Quantification of (B) (n = 20 cells each from 2 biological replicates, paired t test). (D) Membrane potential predicted by the tissue simulation upon a 20% uniaxial stretch (n = 4 simulations, each point is averaged across all cells in simulation). (E) Images of membrane potential dye DiSBAC₂(3) on MDCK cells before stretching, during stretching, and after de-stretching, showing the reversibility of membrane potential change. Scale bar: 25 μm. (F) Quantification of (E) (n = 8 cells). Increase in fluorescence intensity indicates depolarization and decrease indicates hyperpolarization. Times of stretch and destretch are indicated by cyan and yellow arrows, respectively. (G) Membrane potential predicted by the tissue simulation upon stretch and subsequent de-stretch (n=4 simulations, each point is averaged across all cells in simulation). (H) NoRI images of cellular biomass in response to tissue stretch. Biomass density recovers to the initial biomass density level overnight. Scale bar: 25 μm. (I) Quantification of (H). 0 h represents time of stretch. (n=8 cells each from 2 biological replicates, p values: 0.0018 [initial vs. 0 h], 0.0829 [initial vs. overnight], and one-way ANOVA with Tukey’s multiple comparison test.) (J) Tissue simulation recapitulates the drop in biomass density upon stretch and overnight recovery. (n = 4 simulations). (K) Rescue of de-stretching-induced tissue crowding by ouabain. After 16 h, a significant number of cells had been eliminated in the control condition, consistent with previous studies.^, Scale bar: 50 μm. (L) Quantification of (K) (n = 20 ROIs from 3 biological replicates, unpaired t test, error bars represent SEM). (M) Tissue simulation recapitulates cell elimination after compression and rescue from elimination by depolarizing drugs (n = 4 simulations, error bars represent SEM). ****p < 0.0001. Data are represented as mean ± SD, unless otherwise noted. See also Figure S2.

Figure 4.. A mechanically induced depolarization wave enhances wound healing

(A) Upper panel: DiSBAC₂(3) and fluorescently labeled membrane images of MDCK cells imaged over 3 h after wounding. A depolarization wave was observed from the wound edge up to several layers deep into the tissue. Lower panel: membrane potential from the tissue simulation over the course of wound healing. Scale bar: 50 μm. (B) Quantification of membrane potential as a function of distance from scratch wound at different times (n = 4 rectangular ROIs of 100 μm width). (C) Quantification of mean normalized cell area of cells away from the leading wound edge, as a function of time post wounding, shows expansion of cells during wound healing (n = 20 cells spanning from second to seventh cell row from the wound edge, error bars represent SD). (D) Upper panel: NoRI images over the course of wound healing. Lower panel: biomass density from the tissue simulation over the course of wound healing. Scale bar: 50 μm. (E) Quantification of cellular biomass density as a function of distance from the scratch wound at different times after wounding (n = 4 rectangular ROIs of 50 μm width). (F) Representative images of control, depolarizing drug treatment (ouabain), and hyperpolarizing drug treatment (valinomycin) during wound healing. Scale bar: 500 μm. (G) Quantification of wound area closure as a function of time (n = 2 wells). Depolarization resulted in faster wound healing (ouabain, red circles) compared with the untreated control (cyan circles). Hyperpolarization (valinomycin, yellow circles; 5% PEG, green circles) resulted in slower wound healing. (H) Representative image showing Xenopus embryo tail amputation. (I) Bright-field and membrane potential (DiSBAC₂(3)) images of Xenopus wound edge, 10 min post amputation. Cell layers are progressively depolarized from deep tissue toward the wound edge. Scale bar: 10 μm. (J) NoRI images of Xenopus tail amputation, 0 and 10 min post cut. The closest cell layer becomes dilute in biomass density immediately after the cut, and 5–6 cell layers become dilute within 10 min. Images represent different Xenopus embryos amputated and fixed at the specified times. Scale bar: 25 μm. ****p < 0.0001. Data are represented as mean ± SD unless otherwise noted. See also Figures S2 and S3.

Figure 5.. The Hippo pathway mediates membrane-potential signaling

(A) YAP immunostaining and fluorescent nuclei images in low-density (top) and high-density MDCK monolayers (bottom). Scale bar: 25 μm. (B) Quantification of nuclear/cytoplasmic YAP ratio as a function of cell number density. (n = 20,337 cells, cells binned together by local number density, with a bin size of 1 cell/100 μm².) (C) Schematic representation of YAP localization at different cell number densities. (D–G) Quantification of nuclear/cytoplasmic ratio of YAP in MDCK/MCF10A cells for YAP under sparse (D, F, and H) and dense (E, G, and I) conditions under varied drug perturbations, measured by immunostaining of YAP. In sparse cells, YAP localizes to the nucleus and induction of hyperpolarization significantly promotes nuclear exclusion of YAP. In dense cells, depolarization by ouabain inhibits nuclear exclusion of YAP compared with the control (n = 24 cells total from 2 biological replicates, unpaired t test). Scale bar: 50 μm. (H) YAP immunostaining of sparse MDCK cells with control, 40 μM valinomycin, and 40 μM valinomycin with MST inhibitor (10 μM XMU-MP-1) conditions (n = 24 cells total from 2 biological replicates, one-way ANOVA with Tukey’s multiple comparisons test). Scale bar: 50 μm. (I) YAP immunostaining of sparse MCF10A cells with control, 10 μM valinomycin, and 10 μM valinomycin with MST inhibitor (10 μM XMU-MP-1) conditions (n = 24 cells total from 2 biological replicates, one-way ANOVA with Tukey’s multiple comparisons test, the adjusted p value between sparse and 10 μM Val + MST1 inhibitor is 0.0092). Scale bar: 50 μm. (J) Sparse MCF10A cells, with and without FAT1 knockdown (KD), stained for FAT1 and YAP under control and valinomycin conditions. (n = 20 cells total from 2 biological replicates, unpaired t test, p value between FAT1KD control and valinomycin is 0.558 [ns]). Scale bar: 50 μm. (K and L) MST1 immunostaining of MCF10A/MDCK cells under sparse control, sparse valinomycin, dense control, and dense ouabain conditions. White arrows highlight areas of increased MST1-organized membrane clusters. For quantifications in panels, the localization of MST1 in individual cells in randomized fields of view were analyzed (n = 4 ROIs, unpaired t test. The p value is 0.0012 for MDCK sparse control vs. valinomycin and <0.0001 for other plots). Scale bar: 25 μm. (M) Schematic diagram summarizing signal transduction mechanism mediated by membrane potential. Hyperpolarization of membrane potential causes MST1 to colocalize with the C-terminal tail of FAT1, assembling the Hippo signaling complex. This Hippo “on” state prevents the nuclear translocation of YAP. ****p < 0.0001. Data are represented as mean ± SD unless otherwise noted. See also Figures S3, S4, and S5.