Removal of the mechanoprotective influence of the cytoskeleton reveals PIEZO1 is gated by bilayer tension

Abstract

Mechanosensitive ion channels are force-transducing enzymes that couple mechanical stimuli to ion flux. Understanding the gating mechanism of mechanosensitive channels is challenging because the stimulus seen by the channel reflects forces shared between the membrane, cytoskeleton and extracellular matrix. Here we examine whether the mechanosensitive channel PIEZO1 is activated by force-transmission through the bilayer. To achieve this, we generate HEK293 cell membrane blebs largely free of cytoskeleton. Using the bacterial channel MscL, we calibrate the bilayer tension demonstrating that activation of MscL in blebs is identical to that in reconstituted bilayers. Utilizing a novel PIEZO1-GFP fusion, we then show PIEZO1 is activated by bilayer tension in bleb membranes, gating at lower pressures indicative of removal of the cortical cytoskeleton and the mechanoprotection it provides. Thus, PIEZO1 channels must sense force directly transmitted through the bilayer.

Figure 1. Overview of membrane bleb formation and the efficiency of blebbing solutions on HEK293 cells including the effect on cell viability.

(a) Formation of membrane blebs consists of three phases: initiation, expansion and retraction. Initiation of blebs can be instigated by a variety of stimuli, and certain cell lines continually bleb (for example, M12 cell line68). The initiation phase usually involves a focal weakening or rupture of the cortical cytoskeleton and bleb expansion continues largely devoid of cortical f-actin driven by hydrostatic pressure and actomyosin contractility. Once polymerization of f-actin begins at the bleb membrane, expansion is halted. In the physiological setting, Rho-ROCK signalling then drives bleb retraction again via actomyosin contractility. Illustration of the number of cells blebbing and the corresponding % cells stained with trypan blue in response to treatment with (b) a hypoosmotic NaGluconate solution (∼140 mOsm), (c) a hyperosmotic NaGluconate solution (∼440 mOsm) and (d) a Ca²⁺ free KCl-based solution (first described for use in myocytes3570). (e) Illustration of the dependence of blebbing on the activity of myosin II with almost complete abolition of bleb formation in the presence of 2.5 μM blebbistatin (data points represent mean±s.e.m.; n=4 with each replicate including >250 cells).

Figure 2. Fluorescence imaging of HEK293 cell membrane blebs generated using a hypo-osmotic solution.

(a) Example of pIRES2–EGFP PIEZO1 expression in HEK blebs. The blebs can be clearly seen with free GFP inside the cytoplasm. Blebs sometimes grew to diameters greater than 15 μM (upper row). GFP fusion proteins of PIEZO1 (middle row) and MscL (lower row) can also be seen in the bleb membrane (scale bar, 10 μm). (b) Illustration of staining of cells expressing a pIRES2–EGFP PIEZO1 construct. Arrows represent blebs that contain GFP but are devoid of f-actin staining (red; scale bar, 7 μm). (c) As a control, we also stained cells blebbed in the same manner and stained with phalloidin. Again, the arrow represents a bleb that is not stained and hence deficient in f-actin (scale bar, 20 μm).

Figure 3. Activation of MscL–GFP in bleb-attached patches clearly illustrates the different mechanical environments.

(a) A family of MscL current activity in an excised inside-out patch in response to pressure ramps (300 s to peak) at voltages ranging from +20 to −20 mV. (Inset shows single channel activity elicited by manual pressure application at two opposing voltages). In symmetrical NaCl, 145 mM MscL gives a conductance of 1.57 nS in excised patches. The conductance is somewhat lower than in previous reports owing to a reduction in the bulk conductivity of this recording solution in comparison with the widely used KCl 200 mM and MgCl₂ 40 mM buffer. (b) Illustration of MscL activity in membrane blebs; traces show six sweeps of identical pressures. (c) Pressure thresholds of activation for WT MscL in four configurations; cell-attached, excised, bleb-attached and excised azolectin liposome (data points represent individual experiments with mean±s.e.m. shown for comparison). Pressure thresholds in cell-attached and excised were some three times higher than those seen in bleb-attached patches (*P value <0.01; one-way analysis of variance, Tukey's post hoc test).

Figure 4. Activity of MscL–G22S–cGFP construct in cells provides further evidence for the lack of cortical cytoskeleton in bleb membranes.

(a) Illustration of MscL structure with reference to the hydrophobic lock where the G22S mutation is housed. (b) Comparison of the activation pressures of the G22S mutation with the WT MscL channel in various configurations. (c) Graphic illustration indicating that there is a progressive loss of cortical cytoskeletal involvement from cell-attached to bleb-attached configuration with a corresponding reduction in the ability of the cytoskeleton to redistribute the applied force. (d) Activity of G22S–MscL–cGFP in the three configurations in response to either a ramp (350 ms to peak) or a square wave pressure pulse. A leftward shift in the Boltzmann distribution is obvious, not only between configurations but also between ramps and pulses, both applied via a high-speed pressure servo. (e) Confocal image of MscL–G22S–cGFP expressed in HEK293 cells. The channel is not confined to the plasma membrane and seems to be also incorporated in many organelle membranes (scale bar, 10 μm). (f) Representative trace showing G22S–MscL–cGFP activity in response to a pressure ramp up to a peak of −145 mm Hg (350 ms to peak). (g) Representative trace showing G22S–MscL–cGFP activity in response to a pressure pulse of 350 ms duration. (h) Quantification of leftward shift of P_1/2 between ramp and pulse responses of G22S–MscL–cGFP. (i) Quantification of leftward shift of P_1/2 between ramp and pulse responses of G22S–MscL–cGFP in comparison with the cell-attached patches (c/a, cell-attached; b/a, bleb-attached; i/o, inside-out; data points throughout represent mean±s.e.m.).

Figure 5. Activity of different PIEZO1 fluorescent constructs using GFP and mCherry.

(a) Cryo-EM structure of mouse PIEZO1 with an inset showing the likely position of the 1591 insertion (mCherry or GFP). In addition, we show the putative topology of the transmembrane segments and the positioning of our fluorophore insertions using biochemical data. (b) To create a fluorescent GFP–PIEZO1 protein, we first attached GFP to the N- and C-terminal end of the channel protein. The response of the channel was disrupted as it did not inactivate, and gave rise to small currents (trace labelled C terminus). We then introduced mCherry fluorescent protein into the predicted loop positions indicated in a. We were unable to observe any channel response from channels with mCherry insertions at positions 160, 724 and 855. (b) However, the addition of mCherry into position 1591 (called PIEZO1–1591–mCherry) was similar to WT with an inactivation time constant of 40 ms. The introduction of the fluorescent protein at position 1851 produced low currents and the channel did not inactivate. (c) A representative western blot showing three PIEZO1–GFP fusion constructs (N- and C-linked PIEZO–GFP and PIEZO1–1591–GFP) express full-length protein ∼300 kDa. Despite the lower number of channels encountered in the N- and C-terminal fusion proteins, total expression levels were similar. Lower band (∼100 kDa) shows α-actinin for comparison (n=3).

Figure 6. Whole-cell recording for PIEZO1–1591–mCherry on HEK293 cells produce robust currents.

(a) Whole-cell currents elicited by pressing on the cell with a glass probe to the indicated depths. The membrane voltage was set to −60 mV. Above the current trace is the stimulus waveform. (b) Current plotted as a function of depth, which was incrementally increased at 0.5 micron intervals. Increased depth increases the current. (c) Currents recorded when the probe is set to single depth at varying membrane potentials. (d) Current as a function of voltage (from c) showing that the reversal potential is near 0 mV as in the wild-type PIEZO1 channel (data points represent mean±s.e.m.; n=5).

Figure 7. PIEZO1–1591–GFP fusion protein behaves similar to WT PIEZO1.

(a) PIEZO1–1591–GFP currents elicited from a stepwise increase (−10 to −50 mm Hg) in pressure of 150 ms duration. The peak current generated from each pulse is labelled for clarity. (b) Open probability (P_o) plotted against pressure pulse magnitude for PIEZO1–1591–GFP and WT PIEZO1. Here P_o is estimated as I/I_max from peak currents with increasing pressure pulses. (c) Effect of raising external Ca²⁺ concentration on single-channel conductance of PIEZO1–1591–GFP. Initial conductance in CsCl with ∼1 μM free Ca²⁺ is ∼62 pS (buffered using EGTA and calculated using Ca-EGTA Calculator v1.3, an online EGTA Ca²⁺ chelator calculator), and is reduced by almost half by 1 mM external Ca²⁺ (data points represent mean±s.e.m.; n=4). (d) Example of PIEZO1–1591–GFP voltage-dependent inactivation at three voltages. Inactivation markedly slows with depolarization as seen in WT PIEZO1 (for comparison with previous results: ΔV_patch=−55 mV; τ=55±11 ms, n=6).

Figure 8. Activation of the PIEZO1–1591–GFP fusion protein in bleb-attached patches in comparison with cell-attached.

(a) Mechanosensitivity of PIEZO1–1591–GFP in bleb-attached patches in response to a 150 ms square wave pressure pulse of −55 mm Hg (pipette: CsCl). (b) Single-channel activity at rest as seen post pressure at three different voltages. (c) The increase in P_o at zero pressure is calculated using 5 s records. There is at least a fivefold increase in basal activity (*P<0.01; Student's t-test). (d) Leftward shift in Boltzmann distribution function in bleb-attached patches (cell-attached P_1/2=44±2 mm Hg; n=6: bleb-attached P_1/2=33±3 mm Hg; n=4). (e) Example traces of PIEZO1–1591–GFP activated in blebbed membranes pre-treated with Cytochalasin D (CytoD) 10 μM and colchicine (Colch) 10 μM. Cells were treated for 1 h pre-blebbing, then blebbed in the presence of these agents for 2 h using a hypo-osmotic solution. Final patch-clamp electrophysiology was carried out immediately with the bath solution again containing the same pharmacological agents. (f) Quantification of leftward shift of P_1/2 between cell-attached patches and bleb-attached patches with the identified cytoskeletal interfering agents (c/a, cell-attached; b/a, bleb-attached; data represents mean±s.e.m.; n=4/5).

Figure 9. Estimating the tension required to gate PIEZO1 channels.

(a) Family of MscL–G22S–cGFP channel currents elicited in an excised inside-out patch at +10 mV pipette potential. Inset shows enlargement of a segment of one sweep clearly documenting single MscL channel transitions. (b) Confocal images of the corresponding patch membrane and its deformation over time under the negative pressures. The images corresponding to the pressure steps seen in a allow calculation of membrane tension using Laplace's law (T=Pr/2; where T is tension, P is the applied pressure and r is the radius of patch curvature). Laplace's law provides an upper limit for the tension sensitivity. (c) A Boltzmann distribution is shown for three independent experiments illustrating how membrane tension is linked to channel open probability. The same analysis is shown for PIEZO1 using data accrued from the cell-attached patches as shown in d. The fluorescence is that of GFP being expressed on the same plasmid as used to deliver the WT PIEZO1 channels. The patch-clamp recording below the visualized patch membranes shows the corresponding currents elicited by membrane deformation (blue, 0 mm Hg; black, −45 mm Hg; ΔV_patch=−55 mV; scale bar, 2 μm in all the images).