Piezo1 channel activation mimics high glucose as a stimulator of insulin release

Glucose and hypotonicity induced cell swelling stimulate insulin release from pancreatic β-cells but the mechanisms are poorly understood. Recently, Piezo1 was identified as a mechanically-activated nonselective Ca2+ permeable cationic channel in a range of mammalian cells. As cell swelling induced insulin release could be through stimulation of Ca2+ permeable stretch activated channels, we hypothesised a role for Piezo1 in cell swelling induced insulin release. Two rat β-cell lines (INS-1 and BRIN-BD11) and freshly-isolated mouse pancreatic islets were studied. Intracellular Ca2+ measurements were performed using the fura-2 Ca2+ indicator dye and ionic current was recorded by whole cell patch-clamp. Piezo1 agonist Yoda1, a competitive antagonist of Yoda1 (Dooku1) and an inactive analogue of Yoda1 (2e) were used as chemical probes. Piezo1 mRNA and insulin secretion were measured by RT-PCR and ELISA respectively. Piezo1 mRNA was detected in both β-cell lines and mouse islets. Yoda1 evoked Ca2+ entry was inhibited by Yoda1 antagonist Dooku1 as well as other Piezo1 inhibitors gadolinium and ruthenium red, and not mimicked by 2e. Yoda1, but not 2e, stimulated Dooku1-sensitive insulin release from β-cells and pancreatic islets. Hypotonicity and high glucose increased intracellular Ca2+ and enhanced Yoda1 Ca2+ influx responses. Yoda1 and hypotonicity induced insulin release were significantly inhibited by Piezo1 specific siRNA. Pancreatic islets from mice with haploinsufficiency of Piezo1 released less insulin upon exposure to Yoda1. The data show that Piezo1 channel agonist induces insulin release from β-cell lines and mouse pancreatic islets suggesting a role for Piezo1 in cell swelling induced insulin release. Hence Piezo1 agonists have the potential to be used as enhancers of insulin release.


INS-1 cells express Piezo1 and respond to Piezo1 agonist.
mRNA, but not Piezo2 mRNA was detected in INS-1 cells by RT-PCR (Fig. 1A). Anticipating that functional Piezo1 channels are also expressed, the activity of Piezo1 chemical agonist Yoda1 on Ca 2+ influx was tested on INS-1 cells. In a dose dependent manner Yoda1 induced increases in intracellular Ca 2+ with an estimated EC 50 of 4 μM (Fig. 1B,C). Poor solubility of Yoda1 above 10 μM prevented an accurate determination of the EC 50 . We performed voltage-clamp recordings to confirm that Yoda1-induced Piezo1 activation leads to ionic influx in INS-1 cells. Yoda1 evoked ionic currents which had a current-voltage relationship (I-V) consistent with activation of Piezo1 channels (Fig. 1D-F). To investigate if the Ca 2+ increase depends on influx of Ca 2+ from the extracellular space, rather than release from intracellular stores, the effect of Yoda1 was tested in the absence of extracellular Ca 2+ . When there was no Ca 2+ outside, Yoda1 did not increase the intracellular Ca 2+ signal, indicating that it induces Ca 2+ influx rather than store release ( Fig. 2A,C). Though no Ca 2+ was added to the extracellular buffer, in the absence of a Ca 2+ chelator there could be some residual Ca 2+ extracellularly whose influx might have resulted in the small increase in fluorescence observed. Most β-cell lines including INS-1 cells express VDCC which are essential for the classical insulin secretion pathway. Therefore, cells were pre-treated with the VDCC inhibitor nicardipine (10 μM for 30 min) before the addition of Yoda1. Nicardipine did not affect the Yoda1 stimulation of INS-1 cells, suggesting that the Yoda1 induced Ca 2+ influx was not through VDCC ( Fig. 2B,C). Though Yoda1 is considered to be specific for Piezo1 channels, the risk of non-specific effects must be considered. Thus we have made additional determinations. The non-specific inhibitors Gd 3+ and ruthenium red (RR), both at 30 μM with 30 min pre-treatment, inhibited Yoda1 induced Ca 2+ influx (Fig. 2D,E). Moreover, the recently developed Yoda1 analogue (Dooku-1), which antagonises Yoda1 activation of Piezo1, significantly reduced Yoda1 induced Ca 2+ influx (Fig. 2F,H) and an inactive Yoda1 analogue (2e) failed to mimic the effect of Yoda1 (Fig. 2G,H). These expression and pharmacological data suggest that INS-1 cells contain functional Piezo1 channels.

Stimulation of insulin release by osmotic or shear stress. An expectation of cells expressing Piezo1
is that they should be sensitive to osmotic stress caused by hypotonic solution and shear stress caused by fluid flow. Hypotonic solution is expected to cause cell swelling because of increased water entry into cells and specifically in β-cells, cell swelling leads to insulin release. Consistent with the presence of functional Piezo1 channels, hypotonicity increased basal and Yoda1 induced intracellular Ca 2+ levels ( Fig. 4A-D). Furthermore, hypotonicity caused insulin secretion which was suppressed by the non-specific Piezo1 channel inhibitor RR (Fig. 4E). Exposing INS-1 cells to the physical force of shear stress (12 dyne/cm 2 for 1 hour) also induced insulin release and this too was inhibited by RR (Fig. 4F). The data provide further evidence of functional Piezo1 channels in β-cells and suggest a link to insulin secretion. However, it should be noted that RR can inhibit VDCC too which could be secondarily activated by Piezo1 stimulation.
Piezo1 perturbation on high glucose induced insulin release. As expected, high glucose (17.8 mM for 1 hour) stimulated insulin release (Fig. 5A). Similarly, Yoda1 (10 μM for 1 hour) stimulated insulin release in low glucose buffer (Fig. 5A). The effects of high glucose and Yoda1 were not additive which may suggest either a common underlying mechanism or a limited pool of available insulin (Fig. 5A). Dooku1 alone did not affect insulin release but pre-treatment with Dooku1 (10 μM for 15 min) significantly reduced the effect of Yoda1 (Fig. 5A). The inactive Yoda1 analogue (2e; 10 μM for 1 hour) did not induce insulin release. As expected, exposure to high glucose (17.8 mM for 30 min) increased basal Ca 2+ level in INS-1 cells (Fig. 5B,C). High glucose also enhanced Yoda1 induced Ca 2+ influx (Fig. 5D,E). To ascertain that Yoda1 stimulated insulin release is through its action on Piezo1, we tested the effect of Yoda1 on INS-1 cells after knocking down Piezo1 expression using siRNA. Piezo1-specific siRNA transfected cells showed a 78% reduction in Piezo1 mRNA expression compared to the scrambled (control) siRNA transfected cells (Fig. 6A). Importantly, Piezo1 knockdown significantly reduced both Yoda1-and hypotonicity-induced insulin release (Fig. 6B). However, it did not significantly inhibit high glucose-induced insulin release (Fig. 6B).

Piezo1 agonist regulates insulin release from primary mouse islets.
To test whether Piezo1 stimulation leads to insulin release from primary β-cells we turned to isolated mouse pancreatic islets. The islets showed expression of both Piezo1 and Piezo2 mRNA (Fig. 7A). Similar to INS-1 cells, high glucose and Yoda1 (10 μM for 1 hour) induced insulin release and the effects were not additive (Fig. 7B). The inactive Yoda1 analogue (2e; 10 μM for 1 hour) had no effect (Fig. 7B). When islets were pre-treated with Dooku1, the effect of Yoda1 was prevented (Fig. 7C). RR produced mild inhibition of hypotonicity induced insulin release from the mouse pancreatic islets as well (Fig. 7D). To further investigate whether Yoda1 induces insulin release by activating Piezo1 channels, we measured Yoda1-stimulated insulin release from pancreatic islets isolated from mice that were haploinsufficient for Piezo1 (Piezo1 +/− ). Yoda1 induced significantly lesser insulin release from Piezo1 +/− islets compared to wildtype islets (Fig. 7E). These data suggest that Piezo1 channel activation is a mechanism for inducing insulin release.

Discussion
Ca 2+ entry and current induced by the Piezo1 agonist Yoda1, together with the observed Piezo1 mRNA expression, suggest the presence of functional Piezo1 channels in two types of β-cell lines -INS-1 and BRIN-BD11. This is further supported by the predicted responses to known chemical inhibitors of Piezo1 and a non-functional Yoda1 analogue (2e) in Ca 2+ measurement experiments. Moreover, two well characterised Piezo1 activators (i.e. shear stress and Yoda1) also induced significant insulin secretion from both the β-cell lines and primary mouse islets. The stimulatory effects of hypotonicity, shear stress and Yoda1 were lost in the presence of nonspecific Piezo1 inhibitors and specific Yoda1 antagonist. Importantly, INS-1 cells transfected with Piezo1-specific siRNA and pancreatic islets isolated from Piezo1 +/− mice showed significantly reduced Yoda1-induced insulin release. Thus, the results support our hypothesis that stimulating mechanosensitive Piezo1 channels by chemical agonist or cell swelling can induce insulin secretion from pancreatic β-cells.
Yoda1 mimics mechanical stimulation and thus facilitates the study of Piezo1 channels without the need for mechanical stimulation, and has no effect on Piezo2 channels 28 . Overexpressed mouse and human Piezo1 channels were originally shown to be activated by Yoda1 with an EC 50 of 17.1 and 26.6 μM respectively 28 . Native Piezo1 channels too have responded to Yoda1 at low micromolar concentrations 23,25,31 . However, EC 50 values were shown to be much lower (2.51 μM for stably overexpressed Piezo1 and 0.23 μM for native channels in human umbilical vein endothelial cells) recently 30 . In β-cells, Yoda1 induced a dose dependent increase in intracellular Ca 2+ with an EC 50 of 4.54 to 9 µM (Figs 1C and 3C). Furthermore, a non-functional Yoda1 analogue (2e) failed to induce any Ca 2+ influx. Both non-selective inhibitors (RR & Gd 3+ ) and Dooku1 inhibited Yoda1 induced Ca 2+ influx significantly. The concentration of Yoda1 required to stimulate Ca 2+ entry and the lack of effect of 2e, the inhibition of Yoda1-mediated effect by removal of extracellular Ca 2+ , the expected effects of Piezo1 inhibitors and the lack of any inhibitory action of the VDCC blocker (nicardipine) indicate that the Ca 2+ influx observed in our study is predominantly through Piezo1 channels. Of note, higher concentration (>∼20 μM) of Yoda1 solutions turn increasingly opaque. Therefore, the apparent EC 50 is likely affected by compound insolubility and may not allow meaningful interpretation.
Ca 2+ elevation upon hypotonic stimulation has already been reported in various β-cell lines and also in both primary mouse and rat pancreatic β-cells 5,11,14 . It was proposed that hypotonic stimulation leads to membrane depolarization because of activation of volume sensitive outwardly rectifying chloride (Cl − ) channels, which in turn activated VDCC resulting in Ca 2+ influx and insulin release 4,10 . In agreement, some studies have shown that hypotonically induced Ca 2+ elevation in rat pancreatic β-cells was nearly abolished by the VDCC blocker nicardipine 5,11 . However, a recent study in HIT clonal cells demonstrated that Cl − channel blockers such as niflumic acid and DIDS failed to inhibit insulin secretion induced by hypotonic stimulation 12  www.nature.com/scientificreports www.nature.com/scientificreports/ shown that hypotonically induced insulin secretion from the HC9 β-cell line is inhibited by the VDCC blockers nitrendipine and calciseptine, but not by the Cl − channel blocker DIDS 7 . Notably, Gd 3+ suppressed hypotonicity induced Ca 2+ elevation in rat pancreatic β-cells and a role for stretch activated cation channels was proposed 14 . Hence, it is clear that there is no consensus from pharmacological studies on the exact mechanism of hypotonicity/cell swelling induced insulin release.
Having neither examined the effect of Piezo1 blockade on the hypotonic response in Ca 2+ entry nor used Cl − channel blockers, we cannot rule either of them out. However, in the light of present data showing potentiation of Yoda1-induced Ca 2+ entry by hypotonicity and high glucose, we propose the following model: hypotonicity induced cell swelling leads to Piezo1 activation and Ca 2+ entry which in turn causes membrane depolarisation and VDCC activation for further Ca 2+ entry (Fig. 7F). Further careful experimentation with VDCC, Cl − , K ATP and Piezo1 channel blockers in different combinations should be performed to clearly elucidate the existence and significance of such a pathway. However, hypotonicity induced insulin release, which could be considered as one of the read-outs for intracellular Ca 2+ elevation is significantly, but only partially, inhibited by Piezo1 blockade (Figs 4E, 5G, 6B and 7D). Hence both pathways (Cl − mediated and Piezo1/Ca 2+ mediated depolarisation upon cell swelling) may be participating in hypotonicity induced insulin release. Interestingly, mouse Piezo1 channels are permeable to Cl − as well with a Cl − to Na + permeability ratio (P Cl /P Na ) of 0.14 32 .
Induction of β-cells to produce/release insulin using secretagogues (e.g. sulfonylureas which are K ATP channel blockers) is a clinically used strategy to manage diabetes mellitus 33 . However, current secretagogue pharmaceuticals are not always favoured due to their adverse cardiovascular reactions and reported induction of β-cell apoptosis 34 . The search for non-pharmacological approaches led to the identification of ultrasound as a Ca 2+ dependent insulin secretion inducer 35,36 . Stimulation of stretch activated channels was proposed as the underlying mechanism. Interestingly, Piezo1 can be activated by ultrasound 37 . Though the significance of Piezo1 in physiological insulin secretion is not evident from our study, using mechanosensitive Piezo1 induction as a strategy to induce insulin secretion has clear clinical potential and warrants further investigation 8 . It is an attractive approach particularly in neonatal diabetes where the disease is caused by gain-of-function mutations in K ATP channels 38 .

Heterozygous Piezo1 knockout mice. All animal use was authorized by the University of Leeds Animal
Ethics Committee and the Home Office, UK. The generation of Piezo1 knockout line was described previously 22 . Mice of age 12-16 weeks were used.
Mouse pancreatic islet isolation. Islets were isolated following the published protocol 39,40 . Briefly, mice were sacrificed and the pancreata harvested in sterile HEPES buffer. In a sterile hood, the pancreas was washed twice with phosphate buffer saline (PBS) supplemented with 1% penicillin (100 U/ml), streptomycin (100 μg/ml). The pancreas was then minced thoroughly in a glass petri dish and digested using 0.1% collagenase-IV (Sigma) prepared in serum-free RPMI1640 media. Digestion was carried out for 5-7 minutes at 37 °C under 5% CO 2 with frequent shaking; 2 ml of fetal bovine serum was added to neutralize the collagenase. The digested tissue was then centrifuged at 800 rpm for 10 minutes. The resulting pellet was resuspended in DMEM medium containing 10% foetal calf serum, penicillin (100 U/ml), streptomycin (100 μg/ml) and seeded in a 60 mm Petri dish and cultured for 48 hours at 37 °C under 5% CO 2 and a humidified atmosphere. C57BL/6 mice aged 2-5 months were used to isolate pancreas, which was conducted in accordance with accepted standards of humane animal care under United Kingdom Home Office Project license No. P606320FB.
Insulin secretion stimulation and assessment. Cells or islets were washed with zero glucose KRBH buffer for 1 hr at 37 °C. Dooku1 pre-treatment was achieved by replacing the wash buffer with fresh buffer containing 10 μM Dooku1 during the last 15 min of the wash period. After washing, fresh low glucose KRBH buffer containing test chemicals (Yoda1 and the analogues) was added onto the cells/islets and incubated for 1 hr at 37 °C. High glucose KRBH and hypotonic SBS were used for measuring high glucose and hypotonicity induced insulin release. Cells were exposed to circular shear stress by placing the culture plates on an orbital rotating platform (Grant Instruments) housed inside the incubator for 1 hr. The radius of culture wells was 10 mm and the rotation rate was set to 210 rpm, which generated swirling motion of medium around the edges of the wells producing tangential shear stress. The shear stress on the cells was calculated as 12 dyne/cm 2 41 . The supernatant was collected and stored at −80 °C until analysis. RR pre-treatment was done by replacing the wash buffer with fresh buffer containing 30 μM RR during last 15 min of the wash period. For islets, buffer change and supernatant collection was performed by centrifuging the islets at 800 rpm for 10 min. Insulin levels were determined using an ELISA kit according to the manufacturer's protocol (from Crystal Chem and RayBio for mouse and rat respectively). For islets, the insulin level was normalised to the total protein of each sample. Protein quantification was carried out by the BCA assay (Pierce) after lysing the islets in the protein lysis buffer containing (in mM), 50 HEPES, 120 NaCl, 1 MgCl 2 , 1 CaCl 2 , 10 NaP 2 O 7 , 20 NaF, 1 EDTA, 10% glycerol, 1% NP40, 2 sodium orthovanadate, 0.5 µg/mL leupeptin, 0.2 PMSF, and 0.5 µg/mL aprotinin (Invitrogen Cell Extraction Buffer). For analysis of the insulin release data, all samples' data were normalised to the control in that particular experiment considering the control as 100%.
Data analysis and statistics. All data were analysed using Origin 2016 software. Results are expressed as mean ± SEM of at least 3 independent repeats. Comparisons within groups were made using paired Students t-tests and between groups using unpaired Students t-test, as appropriate; for multiple comparisons ANOVA with Bonferroni's correction was performed; p < 0.05 was considered statistically significant.

Data availability
The datasets obtained during the current study are available from the corresponding author on reasonable request.