The following compounds were used at the following concentrations: 10 μM AC55541 (PAR2-agonist, Tocris Bioscience, UK), 10 μM AC264613 (PAR2-agonist, Tocris Bioscience, UK), 50 μM FSLLRY-NH₂ (PAR2-antagonist, Sigma-Aldrich, Israel), 2 μM RN1747 (TRPV4-agonist, Tocris Bioscience, UK), 10 μM RN1734 (TRPV4-antagonist, Tocris Bioscience, UK), 10 μM RN9893 (TRPV4-antagonist, Tocris Bioscience, UK), 50 μM D(-)-2-amino-5-phosphonovaleric acid (APV, NMDAR-antagonist, Sigma-Aldrich, Israel), 200 μM (±)-a-Methyl-(4-carboxyphenyl)glycine (MCPG, mGluR-antagonist, Sigma-Aldrich, Israel), KT5720 (protein kinase A inhibitor, Tocris Bioscience, UK), GF109203x (protein kinase C inhibitor, Tocris Bioscience, UK). Pharmaceuticals were added to the perfusion medium with special care to prevent changes in temperature, pH, flow rate, or degree of oxygenation of the artificial CSF (aCSF). Handling and disposal of all drugs carried out in accordance to National and Institutional regulations.

This study and protocol was approved by the Sheba Medical Center Institutional Animal Care and Use Committee (1000/15), which adheres to the national law, and NIH rules. Briefly, 4–5 months old male C57BL/6 mice were rapidly decapitated and 350 μm coronal slices containing the dorsal hippocampus were used. Slices were incubated for 1.5 h in a humidified, carbogenated (5% CO₂ and 95% O₂) gas atmosphere at 33 ± 1°C and were perfused with a CSF [containing (in mM) 124 NaCl, 2 KCl, 26 NaHCO₃, 1.24 KH₂PO₄, 2.5 CaCl₂, 2 MgSO₄, and 10 glucose, pH 7.4] in a standard interface chamber. Recordings were made with a glass pipette containing 0.75 M NaCl (4 MΩ) placed in the stratum radiatum CA1. A cut between CA3 and CA1 was made in order to avoid possible excitability. Stimulation was evoked using a Master 8 pulse stimulator (A.M.P.I., Jerusalem, Israel) and was delivered through two sets of bipolar nichrome electrodes placed on either side of the recording electrode such that two independent stimulation channels were used for each slice. The use of two parallel pathways allowed comparison of the effects of different drug application in the same slice (Maggio and Segal, 2007a,b). Long-term depression (LTD) was induced by low frequency stimulation (LFS) consisting of 1 Hz, 900 pulses, as previously described (Maggio and Segal, 2009). Before applying the protocol, baseline values were recorded at a frequency of 0.033 Hz. Responses were digitized at 5 kHz and stored on a computer. Off-line analysis and data acquisition were performed using Spike 2 software (CED, Cambridge, England). All numerical data are expressed as mean ± SEM, and EPSP slope changes after stimulation were calculated with respect to baseline. There were no systematic differences in the magnitudes of the baseline responses in the different conditions. Unless otherwise indicated, statistical evaluations were performed by applying Student’s t-test for paired and unpaired data, as the case may be (Origin 8.0). p-values of <0.05 were considered a significant difference between means.

The following primary antibodies were used for immunodetection: goat anti-PAR2 (sc-8205, Santa Cruz, 1:25), rabbit anti-TRPV4 (ACC-124, Alomone Labs, 1:50), rabbit anti-PAR2 (APR-032, Alomone Labs 1:500) and mouse anti-GFAP (G3893, Sigma-Aldrich, 1:2000). Hippocampal sections (50 μm) were blocked in 10% normal horse serum in 0.1 M PBS/0.1% Triton for 1 h at room temperature (RT). After 24–48 h incubation at 4°C with the primary antibody (together with 2% normal horse serum), sections were exposed to the appropriate secondary antibody (DyLight^TM 488 conjugated affinity purified donkey anti-goat IgG, 1:800; Alexa Fluor 594 AffiniPure donkey anti-rabbit IgG, 1:2000; Alexa Fluor 488 conjugated AffiniPure donkey anti-mouse IgG, 1:400) for 1 h. The sections were then washed, incubated with Hoechst (b1155, Sigma-Aldrich, 1 μg/ml final concentration) for 10 min (to allow nuclear staining), mounted on dry gelatin-coated slides and finally mounted and cover slipped with Flouromount (F4680, Sigma-Aldrich). Slides were imaged with a Leica SP5 confocal microscope and data were acquired and analyzed using a computer assisted image analysis system.

To test for the role of PAR2 in synaptic transmission and plasticity, we first treated acute hippocampal slices with the selective PAR2-agonist AC55541 (10 μM) while recording evoked field potentials of Schaffer collateral-CA1 synapses. A profound depression of synaptic transmission was observed in these experiments reaching 0.73 ± 0.07 of baseline 30 min after bath-application of AC55541 (p < 0.001; n = 12 slices, Figure 1A). Removal of the PAR2-agonist following induction of LTD did not affect the stability of synaptic depression (Figure 1B). To confirm the specificity of the PAR2-agonist, we repeated experiments in presence of the selective PAR2-antagonist FSLLRY-NH₂ (50 μM; washed in 15 min before 10 μM AC55541). Indeed, in this experimental setting PAR2-LTD was not observed (Figure 1C). Similarly, activation of PAR2 using a different, specific agonist, i.e., AC264613 (10 μM) also resulted in LTD and this effect was blocked by application of the PAR2-antagonist (Supplementary Figures S1A,B, respectively). The effect of the PAR2-agonist (AC55541) was not concentration-dependent, since LTD of similar effects-size was observed when the PAR2-agonist was applied at concentrations of 0.1, 1, and 100 μM (Figure 1D). We conclude from these experiments that PAR2-activation induces robust LTD at Schaffer collateral-CA1 synapses.

We then compared the dynamics of PAR2-LTD with LFS-induced LTD. In a two pathway experimental setting, the delivery of a 1 Hz protocol (900 pulses) resulted in a depression of 0.67 ± 0.06 at 30 min, while the PAR2-agonist induced LTD of similar effect-size at the other pathway (0.69 ± 0.07, p = 0.378, n = 12), without affecting the established LFS-LTD (Figure 1E). Moreover, the PAR2-antagonist did not affect the induction and maintenance of LFS-LTD, while preventing PAR2-LTD at the other pathway (Figure 1F). However, both forms of LTD required the activation of NMDAR, since 50 μM of the NMDAR-antagonist APV blocked LFS-LTD and PAR2-LTD (Figure 1G).

Finally, we tested whether PAR2-mediated LTD is mGluR-dependent by carrying out experiments in presence of the selective mGluR-inhibitor MCPG (200 μM). Consistent with the literature (Maggio and Segal, 2007b; Fitzjohn et al., 2016), LFS-LTD was partially impaired in these experiments (0.82 ± 0.06, p < 0.01, n = 12, Figure 1H). Yet, the induction of PAR2-LTD was not affected by MCPG. Based on these results we conclude that PAR2-LTD requires the activation of NMDAR but not mGluR.

PAR2 is known to mediate its effects, i.e., neuroinflammation and pain in the peripheral nervous system, through the activation of TRPV4 channels (Grant et al., 2007; Chen et al., 2011; Poole et al., 2013). We therefore hypothesized that PAR2 may act on synaptic transmission via TRPV4.

To test this hypothesis, we first examined PAR2 and TRPV4 expression in the hippocampus. Anatomically matched frontal slices containing the dorsal hippocampus were immunostained for PAR2 and TRPV4. Indeed, both PAR2 and TRPV4 were expressed in the hippocampus. A comparable expression pattern was observed: high levels of PAR2 and TRPV4 were detected in CA1 stratum pyramidale. We did not find a prominent colocalization of PAR2 and the astrocytic marker GFAP in these experiments (Figure 2).

We then speculated that TRPV4-activation should also result in LTD, similar to what is observed upon PAR2-activation (c.f. Figure 1). This prediction was tested by exposing acute hippocampal slices to the TRPV4-agonist RN1747 (2 μM). Indeed, a depression in synaptic transmission occurred, reaching 0.63 ± 0.06% of baseline within 30 min (n = 12; Figure 3A; c.f. Figure 1A). This effect was long lasting as it persisted upon the removal of the TRPV4-agonist (Figure 3B). TRPV4-LTD was blocked in the presence of the TRPV4-antagonists RN1734 (10 μM; Figure 3C) or RN9893, respectively (Supplementary Figure S1C). In two pathway experiments TRPV4-LTD reached similar levels of depression as compared to LFS-LTD (0.66 ± 0.07 versus 0.72 ± 0.05 respectively, n = 12, p = 0.19, Figure 3D), while LFS-LTD was not affected by the TRPV4-antagonist (Figure 3E). These experiments disclosed that TRPV4-activation induces robust LTD, similar to what is observed upon PAR2-activation.

In order to test for the interrelation between PAR2- and TRPV4-mediated LTD the following series of experiments was carried out. First, we exposed hippocampal slices to the TRPV4-antagonist before washing in the PAR2-activator (Figure 3F). Indeed, PAR2-activation with AC55541 (10 μM) was not able to induce LTD in presence of the TRPV4-inhibitor RN1734 (10 μM). Conversely, treatment with the TRPV4-agonist in presence of the PAR2-antagonist reliably induced LTD (0.64 ± 0.07% of baseline after 30 min, n = 12; Figure 3G).

We then examined whether sequential PAR2- and TRPV4-activation occlude each other (Figure 3H). Upon induction of PAR2-LTD, the PAR2-agonist was removed before washing in the TRPV4-agonist. In these experiments one of the two pathways was manually reported to the baseline value before TRPV4-activation. Indeed, the TRPV4-agonist failed to induce LTD in this setting, suggesting that the pathway was already saturated, i.e., occluded by the prior application of the PAR2-agonist (Figure 3H).

Finally, we tested whether PAR2- and TRPV4-activation share the same molecular cascade requiring NMDAR but not mGluR activity (c.f. Figures 1G,H). Indeed, we observed that TRPV4-activation did not induce LTD in presence of 50 μM APV (n = 12, Figure 4A), and similar to PAR2-LTD the mGluR-antagonist MCPG (200 μM) was not effective in blocking TRPV4-LTD (0.65 ± 0.04, n = 12, Figure 4B).

To verify that pharmacological activation of PAR2 or TRPV4 induces a genuine LTD at Shaffer collateral-CA1 synapses, in a separate set of experiments we systematically assessed input/output curves and paired-pulse ratios and we did not find any significant effects on these parameters (Supplementary Figure S2). Taken together, we conclude that PAR2 mediates NMDAR-dependent LTD through the activation of TRPV4.

Previous work has indicated that PAR2 acts through PKA to activate TRPV4 in inflammation and pain (Zhao et al., 2015). Hence, to provide further evidence for our major conclusion, we decided to test whether PKA is involved in mediating PAR2-LTD. This hypothesis is relevant also for the known role of PKA as a canonical signaling molecule associated with hippocampal LTD (Collingridge et al., 2010; Hell, 2016; Sanderson et al., 2016). Indeed, in presence of the PKA-inhibitor KT5720 (2 μM) PAR2-activation failed to induce LTD (Figure 5A). This effect was specific, since pharmacological inhibition of protein kinase C (PKC), an additional molecule reported to be involved in LTD (Collingridge et al., 2010), had no apparent effect on PAR2-LTD (Figure 5B).