A mechanosensitive Ca2+ channel activity is dependent on the developmental regulator DEK1

Responses of cells to mechanical stress are thought to be critical in coordinating growth and development. Consistent with this idea, mechanically activated channels play important roles in animal development. For example, the PIEZO1 channel controls cell division and epithelial-layer integrity and is necessary for vascular development in mammals. In plants, the actual contribution of mechanoperception to development remains questionable because very few putative mechanosensors have been identified and the phenotypes of the corresponding mutants are rather mild. Here, we show that the Arabidopsis Defective Kernel 1 (DEK1) protein, which is essential for development beyond early embryogenesis, is associated with a mechanically activated Ca2+ current in planta, suggesting that perception of mechanical stress plays a critical role in plant development.

M ulticellular development is dependent upon cell−cell communication. Recent research has highlighted the fact that, in addition to chemical signals, the perception of mechanical stress, at both the cell and tissue level, is a key factor underlying growth coordination and morphogenesis (e.g., ref. 1 ). However, while in animal systems mechanosensors with important roles in development have been identified [2][3][4][5][6][7][8][9] , the molecular basis for mechanoperception in plants remains enigmatic. As in animals, plants respond to mechanical stimuli by an elevation in cytoplasmic Ca 2+ 10-12 . Based on the knowledge of plasma-membrane mechanosensing in animal systems, the most probable trigger for Ca 2+ release from internal cellular compartments (Ca 2+ -induced calcium release) is either the opening of plasma membrane-localized mechanosensitive Ca 2+ permeable channels, or the opening of voltage-dependent Ca 2+ channels in response to changes in membrane potential caused by mechanosensitive channels permeable to other ions 12

. Recent research in
Arabidopsis has led to the identification of several plasma membrane-localized mechanosensitive ion channels. These include proteins similar to the bacterial mechanosensitive channel of small conductance (the MSL family [13][14][15][16], the MCA1 protein that rescues the yeast Ca 2+ channel mutant mid1, and its homolog MCA2 [17][18][19][20] . In addition, the membrane protein OSCA1 forms a hyperosmolarity-gated Ca 2+ permeable channel required for osmosensing in Arabidopsis 21 . However, although the MSL8 protein has recently been shown to be required for pollen grains to survive rapid rehydration during fertilization 14 , the very mild developmental phenotypes in single and multiple mutants of genes encoding the channels described above, suggests that they are unlikely to play a major role in mechanosensing during development. Furthermore, these proteins have not conclusively been shown to be responsible for any of the mechanosensitive Ca 2 + currents which have been detected and extensively described by electrophysiologists over the past few decades in planta 20,[22][23][24][25][26] . Since genes encoding voltage-sensitive Ca 2+ channels similar to those identified in animal systems have not been found in plant genomes, it has been proposed that plants may have evolved novel systems for mediating mechanosensitive Ca 2+ fluxes at the plasma membrane to control development (reviewed in ref. 12 ).
The Defective Kernel 1 (DEK1) protein is encoded by a highly conserved uni-gene found in all multicellular plant genomes sequenced so far, and it is absolutely required for both embryonic and post-embryonic development in angiosperms [27][28][29][30][31] . Null dek1 mutant embryos do not develop beyond the early globular stage. When plants with reduced DEK1 activity can be obtained, they show major developmental defects, notably in epidermal differentiation and adhesion 30,32,33 37 ). The epidermis is thought to be under tension during much of plant growth 38 . One plausible function of DEK1 could, therefore, be to perceive and respond to this tension to coordinate epidermal development and maintain epidermal integrity. As the epidermis plays a critical role in organ growth in plants (e.g., refs. [39][40][41][42], this, in turn, has major implications for plant development as a whole.
The activity of the CALPAIN domain has been shown to be Ca 2+ -dependent in vitro 43 . These findings are in line with results in animal systems where the activation of cytoplasmic calpains is associated with the Ca 2+ -dependent autolytic-removal of an Nterminal extension, which in some cases can inhibit calpain activity [44][45][46]  of DEK1 is associated with Ca 2+ dependent mechanoperception at the plasma membrane in plants.

Results
A rapidly mechanically activated plasma membrane Ca 2+ channel. Combining genetic and electrophysiological approaches, the Arabidopsis MSL9 and MSL10 proteins have been shown to mediate a mechanically activated channel activity in protoplasts derived from wild-type plant root cells 13 . Selectivity characterization indicated that anions permeate preferentially through the wild-type MSL channels 13 . Because in this work we used excised membrane patches from callus protoplasts, rather than whole root cell-derived protoplasts, we first investigated whether mechanically activated ion currents due to MSLs and other unidentified channels could also be detected in this material. Reverse transcription quantitative-PCR (RT-qPCR) analysis confirmed the expression of MSL9 and MSL10 in wild-type (Columbia-0) callus tissue ( Supplementary Fig. 1a). When Cl − is provided at the cytosolic face of plasma membrane patches, a mechanically activated channel with a high conductance (46 ± 1.4 pS; ± indicates standard deviation, n = 6) is elicited upon application of positive pressure using a high-speed pressure clamp, as previously described in root protoplasts (Supplementary Fig. 1b) 13 . This activity, which is mainly due to MSL9 and MSL10, was completely abolished in protoplasts obtained from quintuple msl4 msl5 msl6 msl9 msl10 (mslΔ5) mutant callus ( Supplementary Fig. 1c). In addition, to MSL activity, we observed a rapidly activated-inactivated current with a smaller conductance, elicited immediately after the pressure increase ( Supplementary Fig. 1b). Unlike the situation for the MSL current, removing Cl − from the cytosolic face of the membrane did not modify this rapid current (Fig. 1a, Supplementary  Fig. 1b). The rapid current was still present in protoplasts derived from mslΔ5 mutant callus with or without Cl − ions at the cytosolic face (Fig. 1a, Supplementary Fig. 1c). In addition this  mechanically activated channel showed a similar conductance to that seen in Col-0 callus (Fig. 1b), with a reversal potential at positive voltage (E rev WT = 47.5 ± 5.4 mV and E rev mslΔ5 = 60.6 ± 11.5 mV; ± indicates standard deviation, n = 6) suggesting that this rapid current is most likely mediated by Ca 2+ influx (dominant cation gradient, E Ca 2+=128.7 mV). Consistent with this hypothesis, we found that it could be reversibly eliminated by replacing Ca 2+ in the bath solution with the larger TEA + cation (Fig. 1c, d). The pressure required to activate the current was higher than that required to activate MSL channels, suggesting distinct activation thresholds for these channels ( Supplementary  Fig. 1b, c). Thus, we can distinguish at least two independent mechanically activated channel activities in callus protoplasts: one dependent on MSLs activity and the second due to an unidentified Ca 2+ channel which we called the rapid mechanically activated (RMA) channel.
To understand the nature of the RMA-mediated current we tested a range of pipette and bath solutions. The RMA channel conductance is unaffected when Ca 2+ ions are replaced with Ba 2+ ions in the bath solution (Fig. 2a, b) suggesting that the RMA channel is similarly permeant to divalent Ca 2+ and Ba 2+ ions, as is classically found for Ca 2+ channels. To test whether the RMA channel shows characteristics previously described for mechanically activated channels in plants, we investigated the effect of Gd 3+ ions, which are known to affect mechanically activated channel activity in both animal and plant systems 23,[47][48][49][50][51][52] . RMA activity was almost totally blocked by the presence of Gd 3+ ions, and activity could subsequently be restored by washing out Gd 3+ (Fig. 2c). The presence of La 3+ ions, a known blocker of Ca 2+ channels 53-55 strongly reduced the mean open time of the RMA channel (Fig. 2d, e) but did not modify the channel conductance. We conclude that the RMA channel activity fulfils the key criteria previously attributed to mechanically sensitive Ca 2+ channel activities recorded at the plant plasma membrane.
DEK1 trans-membrane domains are required for RMA activation. Loss of DEK1 activity in Arabidopsis leads to early embryo lethality (Supplementary Fig. 2a) 29,30 . By expressing the cytoplasmically localized CALPAIN domain under a constitutive promoter (the RPS5A promoter) in dek1 mutant alleles 32 , we could restore post-embryonic growth and thus investigate the contribution of the DEK1 TM domain to RMA activation. This was done in both the dek1-2 mutant background (which has a T-DNA insertion toward the beginning of the TM encoding regions (Supplementary Fig. 2b, c) 32 ) and in the dek1-3 mutant background (which contains a T-DNA downstream of the TM encoding regions of the DEK1 gene ( Supplementary Fig. 2b, c) 32 ). Calli were generated from both dek1-2 CALPAIN-OE and dek1-3 CALPAIN-OE lines and genotyped ( Supplementary Fig. 3a−c). In callus from CALPAIN complemented dek1-3 mutants (dek1-3 CALPAIN-OE), we detected wild-type levels of the transcript corresponding to the endogenous TM-span encoding region of DEK1 (Supplementary Figs. 2c, 4a, b). This allele therefore has the potential to encode a truncated version of the DEK1 protein, with an intact set of TM spans. However, in callus from complemented dek1-2 mutants (dek1-2 CALPAIN-OE), the TM domainencoding region of the DEK1 gene is physically disrupted (Supplementary Figs. 2c, 4a, c) and an aberrant transcript is produced downstream of the T-DNA insertion (Supplementary Fig. 4b). This transcript is unlikely to be translated to produce a functional truncated protein as the dek1-2 allele shows a null phenotype. Overexpression of the CALPAIN transcript was detected in both backgrounds ( Supplementary Fig. 4b). We also generated an antibody against the CALPAIN domain and verified overexpression at the protein level ( Supplementary Figs. 5a, b, 6).
Because the dek1-2 CALPAIN-OE background cannot produce the intact N-terminal region of DEK1, we next tested the potential effect of DEK1 TM domain disruption on the RMA channel activity. In the majority of responsive patches obtained from dek1-2 CALPAIN-OE calli we observed a complete absence of the RMA current upon membrane stretching (Fig. 3a,  Supplementary Fig. 7). Only a residual current was detectable in most patches of this background. In patches from dek1-3 CALPAIN-OE calli the RMA channel activity was still detectable, but showed 6-8 times more rapid inactivation kinetics (Fig. 3a, b,  Supplementary Fig. 7) compared to those observed in callus from Col-0 plants.
The production of viable dek1 mutant plants, as mentioned before, requires the expression of at least the CALPAIN domain of DEK1. In our lines, the CALPAIN domain is overexpressed in the callus system at the RNA level ( Supplementary Fig. 4). To confirm that the loss of the RMA channel activity in the dek1-2 CALPAIN-OE background is not a consequence of CALPAIN overexpression, we generated calli from Col-0 plants overexpressing the CALPAIN domain (Supplementary Figs. 4d and 5) 18 . This material was submitted to electrophysiological analysis. The RMA current was still present in patches from this material, and behaved similarly to the current in Col-0 patches. (Fig. 3a, b, Supplementary Fig. 7). Therefore, we conclude that the CALPAIN domain overexpression per se is not responsible for the loss of RMA channel activity in the dek1-2 CALPAIN-OE background.
Loss of the DEK1 TM domains does not affect MSL activity. Next, we investigated whether the removal of the DEK1 TM domains could cause a perturbation of other mechanoresponsive channel activities at the plasma membrane by testing whether MSL channels can be activated in the dek1-2 CALPAIN-OE line. We found that the MSL activity in dek1-2 CALPAIN-OE protoplasts was still detectable upon membrane stretching, and its characteristics were comparable to that of Col-0 ( Supplementary  Fig. 8). Thus, loss of the DEK1 TM domains leads to a specific perturbation of the mechanosensitive RMA channel activity in our system.

DEK1 function influences root growth inhibition by Gd 3+ ions.
Based on the mild developmental defects observed in the dek1 CALPAIN-OE lines 32,56 , the CALPAIN domain of DEK1 is likely the "active" domain of the protein. The DEK1 CALPAIN domain is autolytically cleaved from the full length DEK1 protein 32 , and its activity has been shown to be Ca 2+ dependent in vitro 43 . Because the release of the CALPAIN domain upon DEK1 cleavage could be triggered by Ca2+ influx via the RMA channel, blocking the RMA channel might affect development in wild-type plants, but would have a reduced impact in the dek1-2 CALPAIN-OE line, where the expression of a "pre-cleaved" version of the DEK1 CALPAIN would partially bypass the need for the RMA channel activity. Considering the sensitivity of the RMA current to Gd 3+ , and the absence of this current in CALPAIN complemented dek1-2 mutants we tested whether CALPAIN complemented dek1-2 mutants might show such reduced sensitivity to Gd 3+ . Gd 3+ dramatically inhibits Arabidopsis root growth 57 . We generated a dose sensitivity curve of wild-type Col-0 seedlings to concentrations of Gd 3+ between 0 and 200 μM ( Supplementary  Fig. 9). We found strong growth inhibition (40-50%) in wild-type seedlings even at the lowest concentrations tested (30 and 60 μM) ( Supplementary Fig. 9). This inhibition was alleviated in CAL-PAIN complemented dek1-2 plants (Supplementary Fig. 10), formally relating DEK1 function to Gd 3+ sensitivity in planta.
Reintroducing full-length DEK1 restores RMA current. We hypothesize that the loss of RMA current observed in dek1-2 CALPAIN-OE callus is caused by the absence of an intact DEK1 transmembrane region. We propose that in the dek1-3 CALPAIN-OE plants a truncated DEK1 protein, which contains the DEK1 transmembrane spans but which cannot confer normal inactivation kinetics on the RMA current is produced ( Supplementary  Fig. 2c). To test both these hypotheses we investigated the RMA channel activity in both the previously published and characterized dek1-3 line complemented with the full length GFP-tagged DEK1 protein expressed under the RPS5A promoter (dek1-3 DEK1) 30 , and in a dek1-2 mutant background into which the same, complementing full length DEK1-encoding construct had been introduced by crossing (dek1-2 DEK1). In the majority of patches obtained from these lines, we recorded an RMA channel activity with inactivation kinetics and current amplitude resembling those of wild-type plants (Fig. 3a, b; Supplementary Fig. 7). The properties of this restored current were tested in the complemented dek1-2 DEK1 mutant callus and found to be strikingly similar to those of the RMA current ( Supplementary Fig. 11). Our results show that reintroducing the full length DEK1 into dek1 mutant backgrounds leads to the restoration of an RMA current similar to that observed in wild-type callus.

Discussion
In summary, we show that the mechanosensitive activity of a Ca 2+ -permeable channel present in the plasma membrane of Arabidopsis callus-derived protoplast requires the TM domains of the DEK1 protein. It has previously been demonstrated that the CALPAIN domain of DEK1 is released from the plasma membrane by an autolytic cleavage event 32 , and that the CALPAIN activity is enhanced by Ca 2+ ions 43 . We therefore propose a model in which DEK1 activity leads to transient elevation of cytoplasmic Ca 2+ concentration during mechanical stimulation, which is locally transduced by autolytic cleavage (and thus activation) of the CALPAIN domain 32 . Based on the phenotype of lines with reduced DEK1 activity, this mechanotransduction pathway is likely required for the maintenance of cell−cell contacts and epidermis integrity 33 , consistent with embryo lethality in loss of function alleles. In turn, DEK1-dependent epidermis integrity is required for the propagation of mechanical signals between neighboring cells, in a feedback loop. Given the absolute requirement for DEK1 in both coherent embryogenesis and postembryonic growth [29][30][31]33 (summarized in Supplementary  Fig. 12), our work suggests a key contribution of mechanoperception to plant development. The absence of cell migration in plant tissues may explain why the perception of tension between adjacent cells plays such an essential role in development in this kingdom.
This study, together with previously published results from both our work and the work of others 32, 56 supports the idea that the CALPAIN domain of DEK1 is the effector component of the protein in terms of mechanotransduction. Whether the transmembrane portion of DEK1 forms a mechanically activated Ca 2+ channel per se, or it is the mechano-sensor associated with an independent Ca 2+ channel remains to be determined. However, the published Ca 2+ dependence of CALPAIN catalytic activity, combined with our electrophysiological results and the fact that CALPAIN complemented dek1 null mutants develop relatively normally, suggests that once activated by cleavage, the CALPAIN domain of DEK1 can respond to changes in cytoplasmic Ca 2+ levels mediated by other Ca 2+ channels. These could include proteins such as MCA1 and MCA2, or potentially the plant PIEZO protein, which has yet to be functionally characterized. The presence of such channels is consistent with our finding that CALPAIN complemented dek1-2 plants retain some Gd 3+ sensitivity in planta.
Interestingly, intracellularly localized calpains have been proposed to act downstream of mechanosensitive ion channels in animals, to regulate a variety of cellular processes, including cellto-cell adhesion [58][59][60][61] . In this context, the association of a calpain protease with a domain influencing a mechanosensitive Ca 2+ channel activity in the membranes of multicellular plants may reflect evolutive convergence between these two kingdoms.
In animals, the characterization of mechanosensitive channels Piezo1 and Piezo2 has revealed novel roles of mechanoperception in development and physiology. For instance, respiration in lungs relies on Piezo2-expressing sensory neurons, which use mechanical signals to sense airway stretching 62 . Axon growth and trajectory was also shown to depend on Piezo1-dependent perception of the mechanical environment of neurons 63 . The association of DEK1, a key developmental regulator, with a mechanosensitive channel activity paves the way to the identification of new roles of mechanical signals in plant development.

Methods
Plant material. Arabidopsis (Arabidopsis thaliana) Columbia-0 (Col-0) wild-type seeds were obtained from Nottingham Arabidopsis Stock Centre (NASC, School of Biosciences, University of Nottingham, United Kingdom). Plants expressing the pRPS5A:CALPAIN-HIS in a wild-type background (CALPAIN-OE), pRPS5A: CALPAIN-GFP, or pRPS5A:DEK1 in the dek1-3 mutant background and pRPS5A: CALPAIN-GFP in the dek1-2 background have been previously described 18,30 . The pRPS5A:DEK1 construct was transferred from the dek1-3 to the dek1-2 background by crossing, and plants were genotyped as described in Supplementary Fig. 3.
Plant and callus growth conditions. For in vitro cultures, seeds were surfacesterilized with chlorine gas, sown on square plates, and stratified for 2 or 3 days in the dark at 4°C. After stratification, seeds were germinated in a growth chamber under a 16-h light/8-h dark cycle at 21°C.
Electrophysiology. Patch-clamp experiments were performed at room temperature with a patch-clamp amplifier (model 200 A, Axon Instruments, Foster City, CA) and a Digidata 1322 A interface (Axon Instruments). Currents were filtered at 5 kHz, digitized at 20 kHz, and analyzed with pCLAMP8.1 and Clampfit 10 software. During patch-clamp recordings, cells were held at a holding potential (corrected from liquid junction potential) of −16 or −6 mV depending on the composition of the pipette solution and pressure was applied with a high speed pressure-clamp system (ALA Scientific Instrument, NY), allowing the application of precise and controlled pressure pulses in the pipette 15 inhibitor treatments, 0.25 mM LaCl 3 or 0.25 mM GdCl 3 were added to the bath solution, osmolarity was adjusted with mannitol to 450 mOsmol for the bath solution and to 460 mOsmol for the pipette solution using an osmometer (Type 15, Löser Meβtechnik). Gigaohm resistance seals between pipettes (pipette resistance, 0.8-1.5 MΩ) (coated with Sylgard (General Electric) pulled from capillaries (Kimax-51, Kimble Glass)) and protoplast membranes were obtained with gentle suction leading to the whole-cell configuration, then excised to an outside-out configuration. The current inactivation kinetics were fitted with a monoexponential function: F(t) = A × e-t/τ + C, where A is the coefficient, τ is the time constant, and C represents the maximum current intensity.
Genomic DNA extraction and genotyping. Plant callus DNA was extracted using a rapid CTAB isolation technique 65 . dek1-2 and dek1-3 mutants were genotyped using the primers shown in Supplementary Table 1 and Supplementary Fig. 3. The following thermal profile was used: 95°C for 5 min, 35 cycles of 95°C for 45 s, 60°C for 45 s, 72°C for 1 min, and a final extension step of 5 min at 72°C.
RNA extraction and quantitative gene expression analysis. Callus material was collected and snap-frozen in liquid nitrogen for gene expression analysis. For each experiment, at least two independent biological replicates were used. Total RNA was extracted using the Spectrum Plant Total RNA Kit (Sigma-Aldrich). Total RNAs were digested with Turbo DNA-free DNAse I (Ambion) according to the manufacturer's protocol. RNA concentrations were measured with a NanoDrop ND-1000 UV−Vis spectrophotometer (NanoDrop Technologies). One microgram of total RNA was reverse transcribed (RT) using the SuperScript VILO cDNA Synthesis Kit (Invitrogen) according to the manufacturer's instructions. PCR reactions were performed in optical 96-well plates in the StepOne Plus Real Time PCR System (Applied Biosystems). Five microliters of a 1:10 dilution of cDNA was amplified in 20 µL of reaction mix. The following thermal profile was used: 95°C for 10 min, 40 cycles of 95°C for 15 s, 60°C for 30 s. Amplicon melting curves, were recorded after cycle 40 by heating from 60 to 95°C with a ramp speed of 1°C min −1 . Expression levels of each gene, relative to EIF4A, were determined using a modification of the Pfaffl method 33 . Technical triplicates were performed for each biological sample. Primers used for RT-quantitative (q)PCR analysis are shown in Supplementary Table 1.
Antibody production. Anti-IIaIIb rabbit polyclonal antibodies were generated against a peptide from the predicted catalytic domain (RGDKQFTDQEFPPNC) of the Arabidopsis DEK1 protein using the PolyExpress Custom Polyclonal Antibody Production Package Service (GenScript).
Western blot analysis. Callus material for each genotype was snap-frozen in liquid nitrogen, thoroughly ground in pre-cooled porcelain mortars, and proteins were extracted using an extraction buffer containing 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 10% glycerol, 1% deoxycholate, 1 mM EDTA and 1× protease inhibitor cocktail P9599 (Sigma-Aldrich). Samples were incubated on ice for 1 h, centrifuged for 15 min at 14,000×g at 4°C to remove cell debris, and protein concentrations in supernatants were determined using a Bio-Rad protein assay (Bio-Rad). Equal amounts of proteins were loaded and resolved on 7.5% polyacrylamide/0.1% SDS gels. Proteins were transferred onto nylon membranes using an iBlot II dry blotting system (Invitrogen). Membranes were blocked over night at 4°C using 1× PBS/0.2% Tween-20 (Sigma-Aldrich) with 5% non-fat milk (Regilait), incubated for 2 h at room temperature with rabbit polyclonal anti-IIaIIb (Genescript) or mouse monoclonal anti-alpha tubulin clone B-5-1-2 (Sigma-Aldrich, catalogue number T5168) primary antibodies at dilutions of 1:1000 and 1:2000, respectively. Secondary horseradish peroxidase-conjugated anti-rabbit and anti-mouse antibodies (Promega, catalogue numbers S3731 and S3721, respectively) were used at a dilution of 1:45,000. Blots were incubated with Super Signal West Femto reagents (Thermo Scientific) according to the manufacturer's instructions, and then exposed to Super RX film (Fujifilm). Membranes were first incubated with anti-IIaIIb antibodies, and then re-probed with anti-alpha tubulin antibodies.
Gadolinium sensitivity experiments. Seeds were surface-sterilized with chlorine gas, sown on square plates (in a single row in the upper part of the plates), and stratified for 2 days in the dark at 4°C. After stratification, seeds were germinated in a growth chamber under a 16-h light/8-h dark cycle at 21°C, with plates kept in a vertical position. Plates containing 1/5 strength (0.86 gr/L) MS medium (Duchefa) pH = 5.7, 1% sucrose and 0.8% phytoagar, were supplemented with gadolinium (III) chloride hexahydrate (Sigma-Aldrich). Primary root length was scored after 5 or 6 days of growth with ImageJ software. Different genotypes were sown side by side on the same plate and in multiple combinations to buffer possible positional effects.
Data availability. The authors declare that all data supporting the findings of this study are available within the manuscript and its Supplementary Files or are available from the corresponding authors upon request.