TREK-1 channels regulate pressure sensitivity and calcium signaling in trabecular meshwork cells

Intraocular pressure (IOP) is the most significant risk factor for glaucoma (Kass et al., 2002; Leske et al., 2003), with current treatment largely limited to IOP-lowering agents that target the secretion of aqueous humor from the ciliary body or its drainage through the pressure-insensitive “uveoscleral” pathway. However, by far the largest outflow component (∼90%) in the primate eye is mediated by the trabecular meshwork (TM), which filters and funnels aqueous humor into the canal of Schlemm (Lütjen-Drecoll and Rohen, 1989). Unlike the ciliary body and muscle, this conventional TM pathway is mechanosensitive and protects the eye from hypertension by autoregulating fluid outflow under different pressure regimens (Brubaker, 1975; Lei et al., 2011). The trabecular outflow mechanism is often compromised in glaucoma, as TM cells adopt the properties of contractile myofibroblasts that chronically augment the tissue resistance to fluid outflow (Flügel et al., 1991; Last et al., 2011). Given that TM has remained intractable to antiglaucoma medications, understanding how its cells sense and transduce pressure is a matter of considerable academic and clinical interest.

We recently identified the nonselective cation channel TRPV4 as a cell volume sensor (Toft-Bertelsen et al., 2017) and likely TM transducer of pressure, swelling, and strain (Ryskamp et al., 2016). In vitro and in vivo experiments revealed that TRPV4 plays a central role in Ca²⁺-dependent cytoskeletal up-regulation, TM resistance to fluid outflow, and regulation of IOP. TRPV4 antagonists lowered IOP in chronically hypertensive eyes but had no effect on healthy eyes (Jo et al., 2016; Ryskamp et al., 2016), suggesting that steady-state tensile homeostasis and mechanoadaptation in TM cells rely on additional mechanosensing mechanisms. In this study, we investigated the background pressure dependence in human TM cells using electrophysiology, pharmacology, RNA silencing, and impedance measurements. We demonstrate that TREK-1, a polymodal mechanosensitive tandem-pore potassium (K_2P) channel (Meadows et al., 2000), represents a crucial molecular link between the membrane potential of primary and immortalized TM cells and their sensitivity to pressure. TREK-1 is activated by multiple types of mechanical force (strain, swelling, shear flow, and stretch) and functions as a regulator of mechanical thresholds, nociception, and stretch-induced contractility in neurons, bladder, colon, and uterine cells (Patel et al., 1999; Talley et al., 2001; Gruss et al., 2004; Heurteaux et al., 2004; Alloui et al., 2006; Nayak et al., 2009; Baker et al., 2010; Monaghan et al., 2011; Noël et al., 2011). Its sensitivity to temperature, pH, long-chain polyunsaturated fatty acids (such as arachidonic acid [AA]), and widely used anesthetics, antidepressants, and neuroleptics (Enyedi and Czirják, 2010; Brohawn et al., 2014; Feliciangeli et al., 2015) allows these channels to integrate the cells’ electrical properties with mechanotransduction to tune a wide spectrum of ambient signals. TRPV4 and TREK-1 channels were reported to regulate the TM cytoskeleton and were implicated in glaucoma (Ryskamp et al., 2016; Carreon et al., 2017), yet it is unclear whether they can be activated by physiologically relevant pressures and how they collaborate in pressure transduction. Here, we characterize the response of TREK-1 to pressure steps and AA and delineate its function as a gatekeeper for Ca²⁺ homeostasis and ECM adhesion. We show that the TM pressure response involves opposing activation of TRPV4 and TREK-1, which cooperate in control of pressure-dependent signaling, calcium homeostasis, and cell–ECM interactions. These findings place TREK-1 in the center of mammalian IOP regulation, and therefore vision, and suggest a novel mechanism for pressure dysregulation in open-angle glaucoma.

Immortalized cells, isolated from the juxtacanalicular region of the human eye (hTM cells), were purchased from ScienCell Research Laboratories. Key physiological features (e.g., TREK-1 dependence of the membrane potential) were also tested in primary TM cells (pTM cells) isolated from corneal rims from three donors (aged between 35 and 60 yr) with no history of eye disease. The rims were acquired and used in concordance with the tenets of the WMA Declaration of Helsinki and the Department of Health and Human Services Belmont Report. The identity of TM cells was validated as described (Ryskamp et al., 2016; Stamer and Clark, 2017), with hTM/pTM exhibiting similar genetic/biochemical/molecular signatures (expression of TM-specific markers, smooth muscle actin, and DEX-induced up-regulation of myocilin), responsiveness to ambient stimuli, and expression of K_2P and TRP channel transcripts. Cell cultures were maintained in the Trabecular Meshwork Cell Medium (ScienCell, 6591) at 37°C and 5% CO₂ and studied at passages 1–10 and 1–3, respectively. For some experiments, cells were transiently transfected with GFP-tagged short hairpin RNA (shRNA) constructs targeting human TREK-1, TASK-1, or scrambled shRNA (shTREK-1, shTASK-1, and scRNA, respectively), using Lipofectamine 3000 according to the manufacturer’s (Thermo Fisher Scientific) instructions. The transfected cells showed no obvious changes in morphology, confluence, and viability and were used for experiments at the second to third day after transfection.

RNA was isolated as described (Ryskamp et al., 2016). Total RNA was extracted with TRI-Reagent (Sigma) with 2 µg used for complementary DNA (XLT cDNA super mix kit; Quanta). Quantitative RT-PCR was performed with Apex-Cyber Green on CFX96 Touch Real-Time PCR (Bio-Rad). PCR conditions were as follows: 95°C for 15 min; 95°C for 15 s, 60°C for 1 min, 40 cycles. The primer sequences and data are shown in Table 1 as mean value of 2^−ΔΔCt ± SEM.

hTM/pTM cells were lysed in a buffer containing the standard protein inhibitor cocktail (Santa Cruz Biotechnology). Total protein was heat inactivated for 3 min at 95°C in Laemmli buffer (Bio-Rad). 30 µg total protein per lane was loaded in 8% mini polyacrylamide gels. Electrophoresis was performed at 90 V for 1 h in the Mini-PROTEAN Tetra apparatus (Bio-Rad; running buffer 25 mM Tris, 195 mM glycine, and 0.1% SDS). Proteins were transferred to a polyvinylidene fluoride membrane (0.2 µM, Bio-Rad) in transfer buffer (25 mM Tris, 195 mM glycine, and 20% methanol) in the Mini-PROTEAN Tetra apparatus at 4°C overnight. The membrane was blocked with 5% skim milk for 30 min, incubated overnight with anti–TREK-1 antibody (1:1,000, Santa Cruz Biotechnology) and anti–TASK-1 antibody (1:1,000, Alomone Labs), and washed with the TBS-T solution (TRIS-base [19 mM], NaCl [150 mM] and Tween-20 [0.1%] diluted in ddH2O, pH 7.5) for 3 × 5 min each. Membranes were incubated with anti-rabbit HRP (1:2,000, Abcam) or GAPDH (1:5,000, Abcam) antibodies for 2 h at room temperature. Protein bands were detected on x-ray film by using enhanced chemiluminescence (Pierce) and developer (AFP Imaging Corp.).

Cells were plated on type I collagen–coated glass coverslips for 24 h and treated with DMSO (control group) or TREK-1 agonists for 1 h at 37°C. After fixation in 4% PFA, the cells were washed in PBS, permeabilized with 0.1% Triton X-100, and exposed to the blocking solution (1% BSA, 0.3% Triton X-100/PBS, and 0.1% NaN₃).

Whole-cell recordings were conducted with borosilicate patch pipettes (tip resistance 6–7 MΩ). To avoid the potentially confounding contributions from intercellular junctions, the recordings were generally conducted in preconfluent cells with spindle-like morphology that is typical of metabolically active cells from the juxtacanalicular TM (Stamer and Clark, 2017). Whole-cell currents and the membrane potential were measured as described (Jo et al., 2016; Molnár et al., 2016) using a Multiclamp 700B amplifier, Digidata 1550 board, and Clampex 10.7 acquisition software (Molecular Devices). Cells were typically held at –30 mV, with current–voltage (I-V) relationships determined from voltage ramps imposed from −100 to 100 mV. Currents were sampled at 10 kHz, filtered at 5 kHz with an eight-pole Bessel filter, leak-subtracted after data acquisition, and analyzed with Clampfit 10.7 (Molecular Devices) and OriginPro 8 (Origin Lab). shRNA-transfected cells were recognized by GFP expression.

The external solution contained (in mM) 126 NaCl, 2.5 KCl, 1.5 MgCl₂, 1.5 CaCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, and 10 d-glucose (pH 7.4 was adjusted with NaOH). In some experiments, 5 mM NaCl was substituted with equimolar TEA-Cl. HEPES replaced NaH₂PO₄ and NaHCO₃ in experiments that used pH changes to test the sensitivity of the resting membrane potential (V_rest). The extracellular saline used to record inward rectifying K⁺ currents contained (in mM) 128 KCl, 1.5 MgCl₂, 1.5 CaCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, and 10 d-glucose. The intracellular (pipette) solution contained (in mM) 135 potassium d-gluconate, 10 KCl, 1 MgCl₂, 10 HEPES, 0.5 EGTA, 0.01 CaCl₂, 0.3 Na-GTP (pH 7.3, adjusted with KOH), and 1 Mg-ATP. All recordings were conducted at room temperature (22–24°C).

Pressure pulses were generated with the high-speed system from ALA Scientific Instruments. Pressure steps were delivered through the patch pipette and controlled through pClamp 10.7. 100 µM 4,4’-Diisothiocyanatostibene-2,2’disulfonic acid disodium salt (DIDS) was added to the extracellular solution to inhibit volume-regulated anion conductances.

Calcium imaging was conducted as described (Ryskamp et al., 2016; Jo et al., 2017; Phuong et al., 2017). Briefly, cells were loaded for 30–40 min with the calcium indicator Fura-2 acetoxymethyl ester (5 µM), placed in recording chambers on an upright epifluorescence microscope (Nikon E600 FN; 20×/0.5 or 40×/0.8 Nikon Fluor objectives), and perfused with isotonic saline containing (in mM) 126 NaCl, 2.5 KCl, 1.5 MgCl₂, 2 CaCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, and 10 d-glucose (300 mOsm; pH 7.4). The high K⁺ external solution contained (in mM): 128 KCl, 1.5 MgCl₂, 1.5 CaCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, and 10 d-glucose. 340/380-nm excitation was delivered from a 150-W Xenon arc source (DG-4; Sutter Instruments). Fluorescence emission was high-pass filtered at 510 nm and captured with cooled EMCCD cameras (Photometrics). Data acquisition, ratio calculation, and background subtraction were accomplished using NIS Elements (Nikon). Results represent averages across cells (three to six slides, each containing 10–30 cells) from at least three separate experiments.

Real-time impedance of hTM monolayers was measured using the electric cell-substrate impedance sensing (ECIS) Zθ system (Applied BioPhysics), as described (Wang et al., 2016). In brief, cells were seeded on 8W10E+ arrays coated with rat collagen before seeding. Once a stable monolayer was established, the cells were exposed to test agents (TREK-1 activators and inhibitors) or control media, and the resistance was measured in each well for 1 h at 10-s intervals in response to a 4-kHz AC frequency (40 electrodes per well). Resistance for each well was normalized to the baseline resistance before the addition of agonist/antagonist to account for baseline differences in the electrodes. Treatment wells were normalized to the control resistance at each time point to determine change in resistance relative to control.

The data are presented as mean ± SEM. Student’s t test and one-way ANOVA followed by Bonferroni post hoc test were used for analyses. Multiple comparisons were done by ANOVA. NS = P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.

We investigated the intrinsic membrane properties, pressure sensitivity, calcium homeostasis, and monolayer impedance in juxtacanalicular hTM cells and validated a subset of the experiments in primary corneoscleral-derived cells prepared from nonglaucomatous human donors (pTM). The phenotype of primary and immortalized cells was validated with PCR analysis of TM-specific markers and steroid-induced up-regulation of myocilin (see Materials and methods).

Although the primate TM is a key regulator of IOP, its capacity for the regulation of fluid outflow is itself pressure sensitive (Brubaker, 1991; Acott et al., 2014), suggesting the potential presence of intrinsic mechanotransducers. To determine whether these include mechanosensitive ion channels, we evoked transmembrane currents with high-speed pressure clamp, a technique developed for the study of stretch- and pressure-activated channels (Sachs, 2010). The average V_rest recorded in the current clamp mode was −33.6 ± 1.2 mV for hTM (n = 53) and −29.4 ± 1.7 mV (n = 16) in pTM cells. The macroscopic current–voltage relationship showed low conductance at negative membrane potentials, conspicuous outward rectification, and reversal at −36.7 ± 2.4 mV (Fig. 1 A). Physiological pressure pulses (15 mm Hg; 3 s) evoked inward and outward currents at holding potentials of ±100 mV (Fig. 1, A–D). Consistent with the recently proposed mechanosensing function of TRPV4 channels (Ryskamp et al., 2016), the amplitude of the pressure-evoked current was reduced by the selective antagonist HC067047 (2 µM; Fig. 1 E). The current–voltage relationship of the HC067047-insensitive pressure-dependent current was dominated by an outwardly rectifying conductance (Fig. 1, F and G). Pressure steps induced a sustained negative shift of the reversal potential by approximately −15 mV, which was preceded by a transient depolarizing current (Fig. 1, H and I).

The implications of these findings are as follows. First, the observation that the TM resting potential (approximately −30 mV) is situated at the pressure-insensitive fulcrum for mechanosensitive channel activation predicts that the TM sensitivity to pressure is determined by the steady-state membrane potential. Second, TRPV4 channels partly mediate the pressure response. Third, pressure stimulation in the presence of TRPV4 inhibitors leaves a large residual outward conductance, indicating that TM mechanotransduction involves additional pressure-sensitive channels.

Given the key role of V_rest in TM pressure sensitivity (Fig. 1), we sought to determine the identity of ion channel subserving the resting membrane conductance. Under current clamp, patching cells with pipettes with K⁺ concentrations equimolar to extrapipette saline abrogated the transmembrane potential gradient (Fig. 2 A). DIDS, an inhibitor of volume-regulated and voltage-operated anion channels, exerted a modest but statistically significant depolarizing shift (3.79 ± 0.63 mV; n = 6; not depicted), indicating that V_rest is principally a function of the K⁺ gradient. Mammalian TM cells express a variety of K⁺ channels (Lepple-Wienhues et al., 1991; Llobet et al., 2001; Gasull et al., 2003; Grant et al., 2013) but it is not clear which, if any, of these maintain or contribute to V_rest. Bath application of TEA (5 mM), a wide-spectrum blocker of voltage-gated (K_v), G protein–coupled inwardly rectifying and Ca²⁺-activated (K_Ca) K⁺ channels, reduced the outward current at nonphysiological (+100 mV) potentials from 494 ± 89 to 309 ± 44 pA (Fig. 2, B and C; n = 6 cells; P < 0.05); however, TEA had no effect on V_rest (Fig. 2 H). Similarly, 4-aminopyridine, a blocker of K_v1.1 (Shaker-related) K⁺ channels (1 mM), had no effect on V_rest (Fig. 2 H). To determine the role of Kir channels, we used hyperpolarizing ramps, which evoked inward currents (n = 7/15) that were suppressed by extracellular Cs⁺ (1 mM; P < 0.05; Fig. 2, D and E). The inwardly rectifying Cs⁺-sensitive current was also unmasked by high (145 mM) extracellular K⁺ (Fig. 2, F and G), yet extracellular supplementation of Cs⁺ had no discernable effects on V_rest (Fig. 2). These results demonstrate that K_v and K_ir channels in human TM cells can be activated at nonphysiological membrane potentials (±100 mV, respectively) but do not contribute to V_rest. In addition, the low [Ca²⁺]_TM (∼50 nM; Ryskamp et al., 2016) and TEA insensitivity argue against the potential involvement of K_Ca channels in V_rest maintenance.

The K⁺ dependence of V_rest and insensitivity to classic K⁺ channel antagonists (Fig. 2 H) pointed at potential roles for the K_2P channel family, which has been associated with leak conductances in a variety of tissues, including smooth muscle cells, which share molecular markers with the TM (Enyedi and Czirják, 2010; Stamer and Clark, 2017). The antimalarial alkaloid quinine (100 µM), which is often used as a pan-K_2P channel blocker (Lesage and Lazdunski, 2000), depolarized hTM cells from −37.7 ± 4.4 to −8.7 ± 2.6 mV (n = 7; P < 0.005; Fig. 2, I and J) and evinced ∼57% suppression of the outward current (Fig. 2, K–M). pTM V_rest and outward currents were similarly sensitive to quinine (P < 0.01 and P < 0.001 for V_rest and whole-cell currents, respectively; Fig. 2, J and M). These data indicate that the hyperpolarizing drive that maintains V_rest in human TM cells under physiological conditions is largely controlled by K_2P channels.

The human K_2P channel family comprises at least 15 members that have been placed into six subfamilies based on sequence identity and functional characteristics. KCNK7 cannot be functionally expressed, TRESK is restricted to the spinal cord and brain, and TALK is confined to islets of Langerhans (Enyedi and Czirják, 2010; Feliciangeli et al., 2015). Analysis of the remaining genes showed broadly consistent expression patterns across hTM/pTM cells, together with the expression of mRNA that encodes TRPV4, a key TM mechanotransducer (Fig. 3 A). The most abundant K_2P transcripts belonged to TASK-1 (TWIK-related acid-sensitive K⁺ channel; KCNK3), TREK-1 (TWIK-related K⁺ channel, KCNK2), TWIK-2 (two-P domain in a weakly inward rectifying K⁺ channel; KCNK6) and THIK-2 (tandem pore domain halothane-inhibited K⁺ channel; KCNK12) subunits. Western blots confirmed TREK-1 protein expression (Fig. 3 C), with the expected bands (de la Peña et al., 2012) at 47 and 58 kD. Moreover, immunostaining showed that TREK-1 and TASK-1 localize to the TM plasma membrane (Fig. 3 D).

K_2P isoforms were functionally characterized using pharmacological tools. In particular, the long-chain polyunsaturated fatty acid AA is an activator of TREK-1, which also inhibits TASK-1 and TWIK-1 (Honoré et al., 2002; Heurteaux et al., 2004). Because AA also activates TRPV4 (Ryskamp et al., 2015) and modulates chloride channels (Meves, 2008), our experiments were generally conducted in the presence of HC-067047 (5 µM), the pan-TRP blocker Ruthenium Red (RuR; 10 µM), and/or DIDS to exclude the confounding contributions from cation/anion channels. Under these conditions, AA (100 µM) evoked sustained hyperpolarizations (arrow in Fig. 4, A and B) that reduced V_rest from −24.2 ± 2.6 mV in unstimulated cells to −57.0 ± 4.6 mV (n = 10 cells; P < 0.01, paired t test). AA-evoked hyperpolarizations were associated with an increase in the outwardly rectifying conductance from 342.4 ± 49.0 pA in unstimulated cells to 2,078 ± 213 pA in the presence of the polyunsaturated fatty acid (Fig. 4, C and D). This effect was antagonized by spadin, a selective inhibitor of TREK-1 (Mazella et al., 2010), which reversibly depolarized the cells to −46.1 ± 4.8 mV (n = 9; Fig. 4, A and B) and partially but significantly (P < 0.005) suppressed the amplitude of AA-evoked outward currents (to 1,518 ± 188 pA; Fig. 4, C and D). The transient depolarization induced by AA in the presence of TRPV4 blockade (arrow in Fig. 4 A) was not investigated further but could involve incomplete TRPV4 inhibition by micromolar HC-067047 and/or activation of arachidonate-regulated Ca²⁺ channels (e.g., Meves, 2008; Thompson and Shuttleworth, 2013).

We next explored whether the cohort of TREK-1 channels that maintain V_rest represents the maximal fraction of activatable channels by exposing the cells to small-molecule TREK-1 activators ML-335 and ML-402 (Lolicato et al., 2017). Both agonists shifted the membrane potential in the hyperpolarizing direction (ΔV_m = −15.5 ± 4.1 and −24.7 ± 6.5 mV by ML335 and ML-402, respectively) while augmenting the whole-cell conductance, with ΔI = 707 ± 98 and 1,458 ± 159 pA for ML-335 and ML-402, respectively (Fig. 5, A–E). Hence, TM cells maintain a substantial fraction of “reserve” TREK-1 channels that are available for activation.

TREK isoforms have been differentiated from other K2P channels with the dihydropyridine analogue amlodipine (Liu et al., 2007). The blocker (10 µM) depolarized hTM cells from −33.4 ± 3.9 to −4.1 ± 1.7 mV and attenuated the amplitude of the outward current (Fig. 5, F–J). Similar inhibitory effects on the membrane potential and outward conductance were observed in pTM cells (Fig. 5, G and I), confirming that V_rest is largely subserved by tonic activation of TREK-1.

The prominent mRNA and protein expression of TREK-1 and TASK-1 in pTM/hTM cells (Fig. 3) prompted us to clarify their relative contributions to V_rest with isoform-specific knockdown. Cells were transfected with lentiviral vectors carrying shRNA or scrambled (Sc) RNAs that were labeled with a fluorescent marker (GFP). Transcript analysis showed 0.57 ± 0.2 knockdown of TREK-1 (P < 0.2) and 0.43 ± 0.02 TASK-1 mRNAs (P < 0.0018) relative to Sc controls, with no apparent effects on cell morphology or survival. Cells transfected with Sc shRNA expressed the outwardly rectifying current (Fig. 6, B and C) typical of untransfected hTM/pTM (Fig. 1 A). Transfection with Sc shRNA had no effect on V_rest (−29.1 ± 2.9 mV), whereas TREK-1 knockdown depolarized the cells to −16.9 ± 2.1 mV (Fig. 6 A). TASK-1 knockdown had no significant effect on V_rest (Fig. 6 D). The transmembrane current in TREK-1 shRNA–transfected cells at V_h = 100 mV was reduced from 480 ± 60 pA in cells overexpressing Sc shRNA to 187 ± 23 pA in TREK-1 shRNA–expressing cells (Fig. 6, A and B). Down-regulation of TASK-1 had a slight but significant depolarizing effect on V_rest in pTM cells (P < 0.05; Fig. 6 J). Similar results were obtained from pTM cells (Fig. 6, G–L). These data confirm the dominant contribution of TREK-1 to the background “leak” conductance in TM cells and suggest a potential auxiliary role for TASK-1 in primary cells.

We investigated the role of K_2P channels in mediating the TM responsiveness to physiological steps of pressure in the presence of HC-067047. Positive pressure steps (15 mm Hg, 3 s) hyperpolarized the V_rest by 3.52 ± 1.49 mV (n = 8 cells, P < 0.05; Fig. 7, A and B), evoking robust outward currents that reached the peak of 2,645 ± 624 pA within 10 s (n = 5; Fig. 7 C). Consistent with K_2P involvement, quinine reduced the pressure-evoked outward current by ∼86% to 369 ± 78 pA (n = 5 and n = 6 cells for control and quinine groups, respectively; P < 0.01; Fig. 7, C and D). Given that TREK-1 is the sole TM-expressed member of the mechanosensitive K_2P group activated by poking, stretching, or pressuring the plasma membrane (Dyachenko et al., 2006; Brohawn et al., 2014), we assessed the pressure response after its knockdown. The pressure-induced current in control Sc-shRNA–transfected cells (3,412 ± 552 pA) was attenuated by ∼80% to 686 ± 63 pA after the transfection with TREK-1 shRNA (n = 11 for Sc shRNA and n = 11 for TREK-1 shRNA groups; respectively, P < 0.01; Fig. 7, E and F). Collectively, these results identify TREK-1 as a regulator of TM mechanosensitivity.

Given the dominant role of the second messenger calcium in TM transmembrane ion flux, volume regulation, contractility, and outflow resistance (Wiederholt et al., 2000; Gasull et al., 2003; Dismuke and Ellis, 2009; Ryskamp et al., 2016), we wondered whether TREK-1 activation coincides with altered Ca²⁺ homeostasis. To test this, cells were loaded with the calcium indicator Fura-2 AM, and the 340/380-nm fluorescence ratio was monitored in the presence of TREK-1 activators and blockers.

Unexpectedly, exposure to the TREK-1 activator ML-402 (40 µM) evoked rapid [Ca²⁺]_i elevations in ∼85% of hTM cells (Fig. 8, A and B). This effect was abolished by RuR (10 µM), suggesting that TREK-1–dependent hyperpolarizations stimulate Ca²⁺ entry via tonically activated TRP-like channels (Fig. 8 C). We tested this possibility in voltage-clamped cells by switching the membrane potential from −30 to 0 mV and −70 mV, respectively. Consistent with the hypothesis that the cells experience constitutive TRP-mediated Ca²⁺ influx driven by the electrochemical gradient, we found that depolarizing steps decreased [Ca²⁺]_i, whereas hyperpolarizations, conversely, increased [Ca²⁺]_i. Reductions (−Δ1.8 mM) and increases (+Δ3 mM) in [Ca²⁺]_o induced corresponding decreases and increases in [Ca²⁺]_i (Fig. 8 D). Another surprising observation was that TREK-1 inhibitors quinine, amlodipine, and spadin also evoke robust increases in [Ca²⁺]_i in resting TM cells (Fig. 8, G–H). The source of Ca²⁺ signals induced by TREK-1 blockers remains to be determined but may include depolarization-activated Ca²⁺ channels and/or release from internal stores (e.g., Wiederholt et al., 2000).

To determine whether channel activators and inhibitors have an impact on the transcellular current flow in TM monolayers, we used ECIS. We have recently used this technique in endothelial monolayers to characterize the permeability of adherens junctions (Phuong et al., 2017), but current flow across nonendothelial monolayers that lack intercellular junctions was proposed to largely represent contractility-dependent adhesivity to the substrate (Qiu et al., 2008; Ramachandran et al., 2011). As illustrated in Fig. 9, 30-min exposure to TREK-1 agonists ML402 and ML67-33 that had been pipetted into the well dose-dependently lowered monolayer impedance. At 100 µM, ML402 and ML67-33 reduced normalized impedance by 9.0 ± 1.7% and 19.0 ± 2.6% (n = 4), respectively (Fig. 9, A–D). The effect of ML402 was transient, whereas ML67-33 demonstrated more sustained reduction of impedance without apparent recovery in the presence of the drug.

Similar to its effect on Ca²⁺ signals, K_2P inhibition likewise decreased the impedance of TM monolayers, with 0.6 ± 0.2% (n = 7) and 10.0 ± 1.4% decreases in peak impedance (n = 8) observed at 10 and 100 µM quinine (Fig. 9, E and F). In contrast to the effect of ML67-33, the effect of quinine was time dependent, as translayer impedance largely recovered in the presence of the drug. The observed TREK-1–dependent reductions in cell-substrate impedance are consistent with increased paracellular spacing between adjacent TM cells, altered cell contractility, and/or changes in cell-matrix adhesion and support a role for TREK-1 in TM–ECM interactions (Goel et al., 2012).

We report that polymodal TREK-1 channels are required to maintain the membrane potential, Ca²⁺ homeostasis, and pressure sensitivity in human TM cells. Our findings include (a) the close relationship between the TM membrane potential and pressure sensitivity; (b) TREK-1 activation underlies the steady-state resting potential, with a reserve pool of nonactivated channels mediating the hyperpolarizing response to pressure and PUFAs; (c) novel links between TREK-1 activation and Ca²⁺ homeostasis; and (d) TREK-1 regulates cell–ECM adhesivity. Together, these results place TREK-1 in the center of mechanosensing and IOP regulation in the primate eye.

The current–voltage relationship and outward rectification of the macroscopic leak current in in physiological K⁺ gradients, together with the abrogation of the membrane potential gradient in symmetric transmembrane K⁺ gradients, indicate that TM V_rest is controlled by the passive behavior of K⁺ leak channels predicted by the Goldman–Hodgkin–Katz equation. Among the variety of K⁺ channels expressed in bovine, mouse, and/or human TM cells are voltage-activated, inwardly rectifying, ATP-sensitive, Ca²⁺-activated large-conductance (BK) and tandem pore K⁺ channels (Llobet et al., 2001; Gasull et al., 2003; Grant et al., 2013), and volume-activated K⁺ channels (Mitchell et al., 2002). Our observation that resting K⁺ fluxes in hTM/pTM cells are insensitive to classic blockers of K_v, K_ir, and K_Ca channels supports the conclusion that the hyperpolarizing component of the V_rest is subserved almost exclusively by K_2P channels. We identified TREK-1 as the key regulator of the background conductance based on the following evidence: (a) hyperpolarizations induced by AA, pressure, and small-molecule activators; (b) obliteration of the resting potential by amlodipine, spadin, and TREK-1:shRNA; (c) outward rectification of the TM current typical of TREK-1 activation in the presence of divalent cations (Maingret et al., 2002); and (d) antibody staining in cultured cells (Fig. 3) and intact tissue (Carreon et al., 2017). We were unable to discern obvious roles for TWIK/THIK-2 subunits, which may have been nonfunctional under our experimental conditions (Bichet et al., 2015), whereas TASK-1 may contribute a small residual outward component in pTM cells.

The extent to which TREK-1 dominates the TM membrane potential seems to be without precedent in eukaryotic cells. Typically, the channel is expressed in cells from organs that experience mechanical deformation (e.g., lung, uterus, stomach, intestine, colon, retinal, and bladder cells), but its contribution to V_rest tends to be auxiliary (Fink et al., 1996; Reyes et al., 1998; Ferroni et al., 2003; Heurteaux et al., 2004; Honoré, 2007; Lembrechts et al., 2011; Cadaveira-Mosquera et al., 2012; Lei et al., 2014). Submaximal TREK-1 activation in resting TM cells may be caused by the low probability of opening at near-zero membrane tension (Fink et al., 1996), whereas TREK-1–mediated hyperpolarizations (∼15–25 mV) induced by pressure, AA, and selective agonists indicate a substantial pool of activatable channels. Activation of this reserve pool might endow the TM with autoregulatory response to mechanical stressors, as documented in heterologously expressing cells (Brohawn et al., 2014). Chloride channels almost certainly contribute to the maintenance of the hyperpolarized state, because blocking Cl⁻-permeable channels (e.g., volume-regulated anion channels, ClC2 and ClC3; Mitchell et al., 2002; Comes et al., 2005) evinced small (∼4–6 mV) depolarizations (O. Yarishkin and D. Križaj, unpublished observations). Studies in bovine TM cells similarly suggested that Cl⁻ fluxes supply ∼10% of the standing outward current (Wiederholt et al., 2000). Another interesting question pertains to the nature of the depolarizing component that balances the resting K⁺ and Cl⁻ fluxes. The identity of this component remains to be determined, with candidates including voltage-operated, store-operated Orai and/or TRP channels (Wiederholt et al., 2000; Abad et al., 2008; Tran et al., 2014).

In addition to maintaining V_rest, the TREK-1 response to pressure appears to modulate calcium influx and cell–ECM interactions. ECIS measurements revealed that channel agonists and antagonists induced significant decreases in monolayer impedance, with a time course that roughly matched that of the [Ca²⁺]_i elevations evoked by TREK-1 activation and inhibition. The precise mechanism remains to be determined but may have involved Ca²⁺-dependent changes in TM contractility, integrin binding (Wiederholt et al., 2000; Campbell and Humphries, 2011), and/or integrin–ECM contacts (Ramachandran et al., 2011), given that Ca²⁺ regulates the interactions between TREK-1 and α-smooth muscle cell actin (Patel et al., 1998), MAPK activation (Bittner et al., 2013), strain-dependent formation of actin stress fibers, and secretion of fibronectin (Ryskamp et al., 2016). Low levels of VE cadherin and lack of adherens and occludens contacts (Bhatt et al., 1995; Pattabiraman et al., 2014) might have contributed to the low steady-state impedance compared with endothelial monolayers (Phuong et al., 2017) that is further decreased through TREK-1 modulation. It remains to be seen whether the increased impedance observed in glaucomatous cells (Torrejon et al., 2016) is mirrored by altered TREK-1 activation in hypertensive eyes.

Among the advantages of the high-speed pressure clamp technique over traditional cell poking assays is the ability to precisely control the timing and amount of force applied to the channel. Using this method, we found hyperpolarizations evoked by physiological pressure steps to be driven mainly by TREK-1. Because the unpressurized gigaseal patch itself imposes nonzero lateral tension on the channel, we did not estimate the threshold for activation by pressure. It is notable that forces that activate TREK-1 (Brohawn et al., 2014) also span the IOPs measured in the healthy eye, ranging from ∼3–7-mm diurnal fluctuations to transient ∼200–300-mm Hg increases evoked by blinking, saccadic eye movements, eye scratching, and sneezing (Downs, 2015). The pressure steps used in our study (15 mm Hg) increased the open probability of human TREK-1 by ∼19-fold (Lei et al., 2014), but it remains to be determined how its open probability is influenced by membrane tensions in hypertensive eyes in which TM cells are stiffer (Last et al., 2011; Pattabiraman et al., 2014) and whether gain/loss-of-function TREK-1 mutations contribute to hypertensive injury in the eye (e.g., Goel et al., 2012; Decher et al., 2017).

One of the conclusions of the present study is that TREK-1 in healthy TM cells functions as a voltage tuner that minimizes pressure responsiveness, Ca²⁺ dysregulation, and contractility through its control of V_rest (approximately −30 mV in human and bovine TM cells; Lepple-Wienhues et al., 1994; Llobet et al., 2001). Tensile steady-state may be so important for TM cells that deviations in depolarizing or hyperpolarizing directions induce a similar compensatory (calcium) response; however, we demonstrate that these compensatory responses involve different intracellular signaling pathways. RuR sensitivity of the hyperpolarizing response suggests that TREK-1 activators facilitate TRP channel–dependent Ca²⁺ influx, possibly via TRPV and/or TRPC channels (Abad et al., 2008; Ryskamp et al., 2016), whereas TREK-1 inhibition presumably depolarizes the cells and induces voltage-dependent Ca²⁺ influx (e.g., Lepple-Wienhues et al., 1991). This ingenious molecular arrangement allows TREK-1 to transduce decreases and increases in pressure into calcium-dependent compensation to modulate TM tensile homeostasis.

We hypothesize that IOP-dependent tensional integration of the TM membrane involves concurrent modulation of at least two intrinsic mechanotransducers, TREK-1 and TRPV4. Compressive, tensile, and/or shear loading prestress the TM membrane by influencing its curvature and fluidity (Anishkin et al., 2014) and thereby facilitating force detection at the TREK-1–lipid interface. In contrast to TREK-1, TRPV4 channels do not regulate the standing conductance, calcium homeostasis, or pressure sensing, as indicated by the lack of effect of HC067047 on the V_rest, [Ca²⁺]_i, and resting IOP (Ryskamp et al., 2016). After the decrease in outflow facility, changes in aqueous shear flow, and/or changes in membrane stiffness that elevate IOP, both TRPV4 (Lakk et al., 2017) and TREK-1 (Brohawn et al., 2014) are activated and cooperate to set a new homeostatic tensile set point. We propose that such set points are likely to be modulated by phospholipid hydrolysis via phospholipase A2 and production of AA (C20:4 ω6), which regulates both TRPV4 and TREK-1. It is noteworthy that exposure to AA recapitulates many aspects of TRPV4- and TREK-1-induced cytoskeletal remodeling, including increases in [Ca²⁺]_i, up-regulation of actin stress fibers, and fibronectin release (Ryskamp et al., 2016; White et al., 2016; Carreon et al., 2017), whereas inhibition of AA production weakens stress fibers and increases aqueous fluid outflow (Pattabiraman et al., 2014). Hence, IOP elevations might regulate the channels directly through force transmission or indirectly through AA and its eicosanoid metabolites (Weinreb et al., 1988; Kirber et al., 1992; Maingret et al., 1999; Enyedi and Czirják, 2010; Ryskamp et al., 2014; Jo et al., 2016). Interestingly, another downstream enzyme, the arachidonate monoxygenase CYP1B1, has been identified as a causative gene in primary congenital glaucoma (Vasiliou and Gonzalez, 2008).

Ocular hypertension is both a cause and consequence of compromised mechanotransduction that underlies the structural and functional changes associated with the decrease in conventional outflow resistance. Prospective randomized multicenter studies demonstrate that IOP reduction delays or even prevents the structural and functional damage of optic nerve axons in glaucoma (Kass et al., 2002; Leske et al., 2003) with an acceptable reduction in IOP by frontline drugs of ∼20%. Given that compromised mechanotransduction may lead to increases in IOP and optic neuropathy by impairing TM filtering and regulation of aqueous outflow, delineation of the intrinsic molecular mechanisms provide new alternatives for treatment. Our findings predict that targeting TRPV4 and TREK-1 could be effective in modulating IOP; however, we also show that antihypertensive treatments will require careful consideration of interlocked signaling pathways that regulate calcium homeostasis, cytoskeletal dynamics, and pressure sensitivity in the TM.

This study was supported by the National Institutes of Health (R01EY022076, R01EY027920, T32EY024234, P30EY014800), Willard L. Eccles Charitable Foundation, Glaucoma Research Foundation, University of Utah Technology Acceleration Grant, and an Unrestricted Grant from Research to Prevent Blindness to the Department of Ophthalmology at the University of Utah.

The authors declare no competing financial interests.

Author contributions: O. Yarishkin and D. Križaj designed research; O. Yarishkin, T.T.T. Phuong, C.A. Bretz, M. Lakk, J.M. Baumann, and K.W. Olsen performed research; O. Yarishkin, T.T.T. Phuong, M.E. Hartnett, and D. Križaj analyzed data; A. Crandall and C. Heurteaux provided tissue and reagents, O. Yarishkin and D. Križaj wrote the paper.

Sharona E. Gordon served as editor.