OSCA/TMEM63 are an evolutionarily conserved family of mechanically activated ion channels

Mechanically activated (MA) ion channels convert physical forces into electrical signals, and are essential for eukaryotic physiology. Despite their importance, few bona-fide MA channels have been described in plants and animals. Here, we show that various members of the OSCA and TMEM63 family of proteins from plants, flies, and mammals confer mechanosensitivity to naïve cells. We conclusively demonstrate that OSCA1.2, one of the Arabidopsis thaliana OSCA proteins, is an inherently mechanosensitive, pore-forming ion channel. Our results suggest that OSCA/TMEM63 proteins are the largest family of MA ion channels identified, and are conserved across eukaryotes. Our findings will enable studies to gain deep insight into molecular mechanisms of MA channel gating, and will facilitate a better understanding of mechanosensory processes in vivo across plants and animals.


Introduction
Mechanotransduction, the conversion of mechanical cues into biochemical signals, is crucial for many biological processes in plants and animals (Arnadó ttir and Chalfie, 2010; Ranade et al., 2015). In mammals, some mechanosensory processes such as touch sensation and vascular development are mediated by the PIEZO family of mechanically activated (MA) ion channels (Coste et al., 2010;Coste et al., 2012;Murthy et al., 2017). For hearing, the mechanotransduction complex in hair cells includes TMC1 as the pore-forming ion channel, but also comprises of TMHS and TMIE (Ballesteros et al., 2018;Pan et al., 2018;Qiu and Müller, 2018). However, other MA ion channels that govern processes such as pain sensation await identification.
In plants, the impact of gravity or soil properties on root development, wind on stem growth, and turgor pressure on plant-cell size and shape are proposed to involve activation of MA ion channels (Hamant, 2013;Hamant and Haswell, 2017). Homologues of the bacterial MA channel MscS-Like (MSLs) exist in plants, and MSL8 is shown to be involved in pollen hydration in Arabidopsis thaliana (Cox et al., 2015;Hamilton et al., 2015). Apart from MSL8, the identity of the MA channels required for most mechanotransduction processes in plants has remained elusive. Beyond MSLs, hyperosmolarity-evoked intracellular calcium increase is shown to be dependent on the genes OSCA1.1 and OSCA1.2 in Arabidopsis thaliana (Hou et al., 2014;Yuan et al., 2014); however, the activation mechanism for these proteins and whether they encode a pore-forming ion channel remains unknown.
We next explored the possibility that OSCA1.1 and OSCA1.2 are mechanosensitive, and that the modest hyperosmolarity-induced currents might be due to osmotic shock causing cell shrinking, and affecting membrane tension (Sachs, 2010). In cells, MA currents are commonly induced by two direct methods: 1) cell-membrane indentation with a glass probe induces macroscopic MA currents in the whole-cell patch clamp mode; 2) cell-membrane stretch induces single-channel or macroscopic MA currents when pressure is applied to a recording pipette in the cell-attached (or excised) patch clamp mode. Surprisingly, MA whole-cell currents recorded from cells transfected with OSCA1.1 or OSCA1.2 were 10-and 100-fold larger than their hyperosmolarity-activated currents, respectively ( Figure 1A,B vs. Figure 1-figure supplement 1), and were comparable to those recorded from cells transfected with mouse PIEZO1, a well-characterized mechanosensitive ion channel ( Figure 1B). Mechanosensitivity of a channel can be estimated by calculating the apparent threshold for activating MA currents that are elicited by membrane indentation. Threshold is measured as the differential of probe distance that first touches the cell and the probe distance that induces the first channel response. Therefore, it is the minimum distance of indentation required to activate the channel. OSCA1.1 and OSCA1.2 whole-cell MA currents had an apparent activation threshold of 8.6 ± 0.9 mm and 6.3 ± 0.7 mm, and inactivated (channel closure in continued presence of stimulus) with a time constant of 10.0 ± 1.3 ms and 10.4 ± 1.7 ms, respectively ( Figure 1B and Table 1). Similarly, robust macroscopic stretch-activated currents were recorded from cells transfected with OSCA1.1 or OSCA1.2 but not from mock-transfected cells ( Figure 1C,D). Stretch-activated currents from OSCA1.1 and OSCA1.2 were reversible and inactivated with a time constant of 24 ± 3.4 ms and 24.6 ± 4.8 ms, respectively ( Figure 1D). The pressure required for half-maximal activation (P 50 ) of OSCA1.1 and OSCA1.2 was -58.5 ± 3.7 mmHg and -54.5 ± 2.2 mmHg, respectively ( Figure 1E). These values are higher than mouse PIEZO1 which has a threshold of -24 ± 3.6 mmHg (Coste et al., 2010;Coste et al., 2015) ( Figure 1E and Table 1), demonstrating that at least in HEK-P1KO cells these proteins evoke high-threshold MA currents. These results suggest that OSCA1.1 and OSCA1.2 are involved in mechanotransduction.
Under physiological ionic conditions, OSCA1.1-and OSCA1.2-dependent stretch-activated currents had single-channel conductance of 184 ± 4 pS and 122 ± 3 pS, respectively ( Figure 1F,G). We note that these values are larger than what was previously reported (Yuan et al., 2014). Unlike our measurements, the single-channel conductance described in Yuan et al., 2014 was measured in the absence of a stimulus, and might not be strictly OSCA-dependent. Interestingly, stretch-activated single-channel current traces from either protein revealed a single sub-conductance state, which was half the amplitude of the full open state (Figure 1-figure supplement 2). Sub-conductance states are a hallmark of many ion channels and are indicative of concerted gating of multiple pores from the same channel, or of changes within a single pore (Dani and Fox, 1991;Hartzell and Whitlock, 2016;Miller, 1982). The presence of a single intermediate state in OSCA1.1 and OSCA1.2 could be suggestive of two cooperative subunits. Indeed, this suggestion is confirmed by the cryo-EM structures of OSCA1.1 and OSCA1.2, which reveal a pore within each subunit of a dimeric channel (Jojoa Cruz et al., 2018;Zhang et al., 2018). Characterization of the stretch-activated currents in asymmetrical NaCl solutions revealed that OSCA1.1 and OSCA1.2 evoked non-selective cation currents with some chloride permeability (OSCA1.1: P Cl /P Na = 0.21 ± 0.06; OSCA1.2: P Cl / P Na = 0.17 ± 0.01) ( Figure 1H). In addition, gadolinium, a generic MA cation channel blocker, inhibited OSCA1.1-and OSCA1.2-induced stretch-activated currents (Figure 1-figure supplement 3). Together, these results demonstrate that OSCA1.1 and OSCA1.2 induce MA non-selective cation currents.

Mm TMEM63C
4.5 ± 0.9 (8) n.d. two types of mechanical stimulation. It is noteworthy that OSCA1.1 and OSCA1.2 which were the only channels activated by membrane indentation, had a high threshold (8 mm) compared to mouse PIEZO1 (4 -5 mm) (Dubin et al., 2017) ( Table 1). Perhaps other OSCA channels have a much higher threshold, technically limiting us from recording reliable MA currents, since HEK-P1KO cells often break near 10 -12 mm of indentation. Alternatively, mechanical indentation and membrane stretch deliver biophysically-distinct forces to the membrane and the different proteins might be tuned to the unique type of force applied. The results described so far indicate that OSCAs evoke stretch-activated currents in a heterologous expression system, meeting at least one of the criteria for MA channels. Indeed, these results by themselves only allow us to conclude that they are a component of an ion channel; it is formally possible that they require endogenously expressed components within HEK-P1KO cells to behave as a functional mechanosensitive ion channel (Arnadó ttir and Chalfie, 2010). To address this possibility, we purified OSCA1.2 fused to a GFP tag, and reconstituted the protein in liposomes to determine whether stretch-activated currents can be recorded from excised proteoliposome patches ( Figure 3A). Purified OSCA1.2-EGFP (117 kDa) appears as a single protein band on a denaturing gel, indicating the absence of other associated proteins (Figure 3-figure supplement 1). Before characterizing OSCA1.2, we first tested whether we could successfully record stretch-activated currents from liposomes reconstituted with the bacterial mechanosensitive ion channel E. coli MscL (Figure 3-figure supplement 1) (Häse et al., 1995). We observed MscL-like stretch-induced channel activity in seven out of eight excised patches ( Figure 3B-D). The stretch-activated single-channel currents had a conductance of 3340 ± 180 pS (N = 6), which is in accordance with previous reports (Kloda and Martinac, 2002) ( Figure 3C). Remarkably, applying negative pipette-pressure in the range of 0 to 50 mmHg to patches excised from liposomes reconstituted with OSCA1.2-EGFP protein also induced robust macroscopic currents ( Figure 3E,F). Applying negative pressure (as high 100 to 120 mmHg) to patches excised from empty liposomes did not change the baseline current. At the single-channel level, the stretch-activated currents exhibited sub-conductance states, were voltage-dependent, and had a conductance of 346 ± 13 pS in 200 mM KCl solution, which matched the single-channel conductance measured from OSCA1.2-expressing cells (304 ± 9 pS) under the same ionic concentration (Figure 3-figure supplement 1 and Figure 3G,H). These results demonstrate that OSCA1.2, like other bona-fide MA ion channels such as MscL, TREK and TRAAK, and PIEZO1, is directly gated by changes in membrane tension, and requires no additional cellular components for activation (Anishkin et al., 2014;Brohawn et al., 2014;Häse et al., 1995;Syeda et al.,    . This represents conclusive evidence that OSCA1.2 encodes a pore-forming subunit and is inherently mechanosensitive (Jojoa Cruz et al., 2018). We next examined if orthologues of OSCAs are also mechanosensitive. Phylogenetic analysis and previous studies have identified the TMEM63 family of proteins as the closest homologues of OSCAs (Hou et al., 2014;Zhao et al., 2016;Yuan et al., 2014) ( Figure 4A and Table 2). To determine whether mechanosensitivity was conserved across different species, we selected homologues in fruit fly, mouse, and human, and tested their ability to induce MA currents in HEK-P1KO cells. In the whole-cell patch clamp mode, mechanical stimulation of cells expressing DmTMEM63, MmTMEM63A, MmTMEM63B, MmTMEM63C, or HsTMEM63A did not elicit MA currents (Figure 4figure supplement 1). However, these clones (with the exception of MmTMEM63C) induced stretch-activated currents when expressed in naïve cells ( Figure 4B,C). The phylogenetic tree illustrates that mammalian TMEM63C is divergent from TMEM63A and TMEM63B ( Figure 4A), which may explain its lack of mechanosensitivity. Alternatively, it is possible that this protein is incorrectly folded and not trafficked to the membrane, or that TMEM63C is not sufficient by itself to induce MA currents.
P 50 values for the TMEM63-induced stretch-activated currents were in the range of -60 to -80 mmHg, suggesting that like OSCAs, these proteins elicit high-threshold MA currents in HEK-P1KO cells ( Figure 4E). The stretch-activated single-channel currents induced by either DmTMEM63 or the mammalian TMEM63s are in the sub-picoampere range and were unresolvable, similar to OSCA2.3 ( Figure 4B vs. Figure 2B). However, compared to plant OSCAs, MA currents induced by the TMEM63 family members had unique gating properties with several fold slower activation and inactivation kinetics ( Figure 4D and Table 1). MmTMEM63A stretch-activated currents are non-selective cationic, and gadolinium inhibits MmTMEM63A-induced MA currents by 75% (Figure 4-figure supplement 2). The stark difference in MA current properties between OSCA2.3/TMEM63s and other OSCA members is interesting. At least MmTMEM63A is expressed at the membrane (Figure 4-figure supplement 3), therefore trafficking issues cannot account for the relatively small stretch-activated macroscopic currents. However, one cannot exclude the possibility that in the native cellular environment, other auxiliary subunits might interact with OSCA2.3 and TMEM63 proteins to alter the channel's permeation and gating properties. Future studies will test this hypothesis. Nonetheless, these results demonstrate that orthologues of OSCAs also induce stretch-activated currents when overexpressed in naïve cells, and that mechanosensitivity is conserved across the various members in the OSCA/TMEM63 family.

Discussion
We provide evidence that members of OSCA family are bona-fide, pore-forming mechanosensitive ion channels: (1) transfection of various members of the family from plants, flies, and mammals give rise to robust MA currents. Importantly, single-channel conductance of individual OSCAs are quite distinct, arguing that the OSCAs contribute to pore properties of these currents. (2) We directly show in vitro (proteoliposomes) that OSCA1.2 is an inherently mechanosensitive ion channel in the absence of other proteins. (3) The accompanying paper describes the high-resolution structure of OSCA1.2 (Jojoa Cruz et al., 2018), and demonstrates that this protein has similar architecture to the TMEM16 family of ion channels (Whitlock and Hartzell, 2017). Furthermore, structure-guided mutagenesis verifies that a residue within the putative pore-forming region contributes to singlechannel conductance (Jojoa Cruz et al., 2018). Therefore, with 15 different members present just in Arabidopsis thaliana (5/6 members we tested were mechanosensitive), OSCA/TMEM63 proteins potentially represent the largest family of mechanosensitive ion channels known to date. Although other eukaryotic ion channels have been previously described, none are conserved from plants to humans. PIEZOs are also present in plants and unicellular organisms; however, evidence that they are mechanosensitive in these species is still lacking.
The existence of 15 members of OSCAs in Arabidopsis thaliana raises the possibility that there might be redundancy, and will make it a challenge to assign function to these proteins individually. However, Yuan et al., 2014reported that under hyperosmotic stress, mutant OSCA1.1 plants had stunted leaf and root growth (Yuan et al., 2014). These phenotypes could indeed be a consequence of impaired mechanotransduction. Expression profile of mouse TMEM63A in public databases (BioGPS.org) suggests expression in tissues that experience mechanical forces such as kidney and stomach, while TMEM63B is high in the nervous system (including dorsal root ganglia) as well as in the heart, skeletal muscle, and stomach. Future studies will investigate the contribution of members of OSCA and TMEM63 families to mechanotransduction in various species.

Materials and methods
Key resources
Cell culture and transient transfection PIEZO1-knockout Human Embryonic Kidney 293T cells (HEK-P1KO, original HEK293T cell RRID: CVCL_0063)) were used for all heterologous expression experiments. HEK-P1KO cells were generated in house using CRISPR-Cas9 nuclease genome editing technique as described previously Lukacs et al., 2015), and were negative for mycoplasma contamination. Cells were grown in Dulbecco's Modified Eagle Medium (DMEM) containing 4.5 mg.ml À1 glucose, 10% fetal bovine serum, 50 units.ml À1 penicillin and 50 mg.ml À1 streptomycin. Cells were plated onto 12 mm round glass poly-D-lysine coated coverslips placed in 24-well plates and transfected using lipofectamine 2000 (Invitrogen) according to the manufacturer's instruction. All plasmids were transfected at a concentration of 600 ng.ml À1 . Cell were recorded from 24 to 48 hr after transfection. Since HsTMEM63A was in a non-fluorescent tagged vector, it was co-transfected with IRES-GFP or pIRES2-mCherry vector.

Electrophysiology
Patch-clamp experiments in cells and in liposomes were performed in standard whole-cell, cellattached, or excised patch (outside-out for cells, inside-out for liposomes) mode using Axopatch 200B amplifier (Axon Instruments). Some whole-cell recordings were done using Axon Multi-clamp700A. Currents were sampled at 20 kHz and filtered at 2 kHz or 10 kHz. Leak currents before mechanical stimulations were subtracted off-line from the current traces. Voltages were not corrected for a liquid junction potential (LJP) except for ion selectivity experiments. LJP was calculated using Clampex 10.6 software. All experiments were done at room temperature. All electrophysiology data was analyzed in Clampfit 10.6 and data was plotted using GraphPad Prism (RRID:SCR_ 002798).
For cell-attached patch clamp recordings, external solution used to zero the membrane potential consisted of (in mM) 140 KCl, 1 MgCl 2 , 10 glucose and 10 HEPES (pH 7.3 with KOH). Recording pipettes were of 1 -3 MW resistance when filled with standard solution composed of (in mM) 130 mM NaCl, 5 KCl, 1 CaCl 2 , 1 MgCl 2 , 10 TEA-Cl and 10 HEPES (pH 7.3 with NaOH). For gadolinium inhibition experiments, 100 mM GdCl 3 stock solution of was diluted in cell-attached pipette solution at a working concentration of 60 mM.

Mechanical stimulation
For whole-cell recordings, mechanical stimulation was achieved using a fire-polished glass pipette (tip diameter 3 -4 mm) positioned at an angle of 80˚relative to the cell being recorded. Downward displacement of the probe towards the cell was driven by Clampex-controlled piezo-electric crystal microstage (E625 LVPZT Controller/Amplifier; Physik Instrumente). The probe had a velocity of 1 mm.ms À1 during the ramp phase of the command for forward movement and the stimulus was applied for 150 ms. To assess the mechanical sensitivity of a cell, a series of mechanical steps in 0.5 or 1 mm increments was applied every 10 -20 s. Threshold was calculated as the differential (y-x) of the probe distance that first touches the cell (x) and the probe distance that induces the first channel response (y).
Macroscopic stretch-activated currents were recorded in the cell-attached or excised, outside-out patch clamp configuration. Membrane patches were stimulated with 500 ms (for OSCA clones) or 1 s or 2 s (for TMEM63 clones) negative or positive pressure pulses through the recording electrode using Clampex controlled pressure clamp HSPC-1 device (ALA-scientific), with inter-sweep duration of 1 min. Since TMEM63 had slower gating kinetics, longer stimulus duration was picked. Negative pressure was applied when patch was in cell-attached and inside-out configuration, positive pressure was applied when patch was in the outside-out configuration. Activation time constant was determined by measuring 10 -90% rise time (between baseline and peak) for currents at saturating pressure stimulus using Clampfit 10.6.
Stretch-activated single-channel currents were recorded in the cell-attached configuration. Since single-channel amplitude is independent of the pressure intensity, the most optimal pressure stimulation was used to elicit responses that allowed single-channel amplitude measurements. These stimulation values were largely dependent on the number of channels in a given patch of the recording cell. Single-channel amplitude at a given potential was measured from trace histograms of 5 to 10 repeated recordings. Histograms were fitted with Gaussian equations using Clampfit 10.6 software. Single-channel slope conductance for each individual cell was calculated from linear regression curve fit to single-channel I-V plots. Single-channel current for MscL in proteoliposomes was measured at one membrane voltage (30 mV) and conductance was calculated at that potential assuming 0 pA at 0 mV. In cells, OSCA1.2 single-channel amplitude in 200 mM KCl, 5 mM MOPS, pH 7.0 (KOH) was also measured at À80 mV (again assuming 0 pA at 0 mV) and used to calculate conductance.

Hyperosmotic stimulation
Hyperosmolarity-induced currents were evoked by two different protocols (Hou et al., 2014;Yuan et al., 2014). 1) Once in the whole-cell-patch clamp configuration in iso-osmotic solution (300 mmol.kg À1 ) currents were recorded continuously in response to voltage ramps from À100 mV to +100 mV applied every 2 -10 s. Once a stable response was achieved, the bath solution was switched to hyperosmotic solution (620 mmol.kg À1 ) and currents were recorded for at least five additional minutes. Maximal response at À100 mV in iso-osmotic solution and hyper-osmotic solution were measured for each cell and plotted. 2) In the whole-cell patch clamp configuration wholecell currents at À80 mV were recorded as cells were first exposed to 1 min of iso-osmotic (300 mmol.kg À1 ) solution followed by 5 mins of hyperosmotic solution (620 mmol.kg À1 ) and back to isoosmotic solution. Maximal current response in the three conditions were plotted for each cell.

Permeability ratio measurements
Reversal potential for each cell in the mentioned solution was determined by interpolation of the respective current-voltage data. Permeability ratios were calculated by using the following Goldman-Hodgkin-Katz (GHK) equations: P Cl /P Na ratios:

OSCA1.2-EGFP expression and purification
Codon optimized OSCA1.2 gene were cloned into vector pcDNA3.1. An EGFP tag was placed at the C terminus and connected to the gene via a PreScission Protease cleavable linker (LEVLFQGP) (OSCA1.2-EGFP). A FLAG tag (DYKDDDDK) was added to the C terminus of EGFP with two intervening alanines as a linker. For protein expression and purification, four liters of HEK293F cells (Thermo Fisher Freestyle 293 F, RRID: CVCL_D603) were grown in Freestyle 293 expression media to a density of 1.2 -1.7 Â 106 cells.mL À1 . All cell lines tested negative for mycoplasma contamination. Each liter was transfected by combining 1 mg.L À1 of the construct with 3 mg.L À1 of PEI in 30 mL of Opti-MEM and then adding the mix to the culture of cells. Transfected cells were grown for 48 hr and then pelleted, washed with ice cold PBS, flash frozen and stored at À80˚C for future use. From this point forward, every step of the purification was carried out at 4˚C unless otherwise stated. Pellets were thawed on ice, resuspended in 200 mL of solubilization buffer (25 mM tris pH 8.0, 150 mM NaCl, 1% Lauryl Maltose Neopentyl Glycol (LMNG), 0.1% cholesteryl hemisuccinate (CHS), 2 mg. mL À1 leupeptin, 2 mg.mL À1 aprotinin, 1 mM PMSF, 2 mM pepstatin, 2 mM DTT) and stirred vigorously for 2 -3 hr. Subsequently, insoluble material was pelleted via ultracentrifugation for 45 min at 30,000 rpm in a Type 70 Ti rotor. Batch binding of the supernatant was performed for 1 hr with 2 mL of FLAG M2 affinity resin (Sigma) previously washed first with 0.1M glycine pH 3.5 and then with wash buffer (25 mM tris pH = 8.0, 150 mM NaCl, 0.01% LMNG, 0.001% CHS, 2 mM DTT). Resin was placed in a gravity flow column and washed with 10CV of wash buffer. Protein was eluted using 2.5 mL of elution buffer (wash buffer and 200 ug.mL À1 3x FLAG peptide). Presence of protein in the elution was confirmed by SDS-PAGE. Sample was concentrated using a 100 kDa MWCO Amicon Ultra centrifugal filter. Concentrated protein was injected onto Shimadzu HPLC and size exclusion chromatography was performed using a Superose 6 Increase column and wash buffer. Fractions corresponding to OSCA1.2-PP-EGFP were concentrated to 2 mg.mL À1 .