TRPM4-dependent post-synaptic depolarization is essential for the induction of NMDA receptor-dependent LTP in CA1 hippocampal neurons

TRPM4 is a calcium-activated but calcium-impermeable non-selective cation (CAN) channel. Previous studies have shown that TRPM4 is an important regulator of Ca2+-dependent changes in membrane potential in excitable and non-excitable cell types. However, its physiological significance in neurons of the central nervous system remained unclear. Here, we report that TRPM4 proteins form a CAN channel in CA1 neurons of the hippocampus and we show that TRPM4 is an essential co-activator of N-methyl-d-aspartate (NMDA) receptors (NMDAR) during the induction of long-term potentiation (LTP). Disrupting the Trpm4 gene in mice specifically eliminates NMDAR-dependent LTP, while basal synaptic transmission, short-term plasticity, and NMDAR-dependent long-term depression are unchanged. The induction of LTP in Trpm4−/− neurons was rescued by facilitating NMDA receptor activation or post-synaptic membrane depolarization. Accordingly, we obtained normal LTP in Trpm4−/− neurons in a pairing protocol, where post-synaptic depolarization was applied in parallel to pre-synaptic stimulation. Taken together, our data are consistent with a novel model of LTP induction in CA1 hippocampal neurons, in which TRPM4 is an essential player in a feed-forward loop that generates the post-synaptic membrane depolarization which is necessary to fully activate NMDA receptors during the induction of LTP but which is dispensable for the induction of long-term depression (LTD). These results have important implications for the understanding of the induction process of LTP and the development of nootropic medication.


Introduction
The cellular and molecular mechanisms underlying cognitive brain functions and their deterioration by neurodegenerative and neuropsychiatric disorders are a central theme in contemporary neuroscience. It is widely accepted that during learning, complex sensorial inputs are encoded as changes in the synaptic efficacy of activated neuronal networks. At the level of individual synaptic connections, this is reflected in either long-lasting increases in synaptic efficacy (long-term potentiation (LTP)), long-lasting decreases (LTD), or a reset of previously increased or decreased efficacy to a new level (depotentiation and dedepression, respectively). Of these different forms of synaptic plasticity, LTP was the first that was discovered in the hippocampal formation [2]. Ca 2+ influx through the N-methyl-D-aspartate (NMDA) subtype of glutamate receptors, upon strong post-synaptic depolarization and removal of the Mg 2+ block from NMDA receptors Aurélie Menigoz, Tariq Ahmed, and Victor Sabanov share first authorship.
Rudi Vennekens and Detlef Balschun share last authorship.
(NMDAR), is widely accepted as the central trigger of LTP induction [1,24]. Once the increase in intracellular Ca 2+ exceeds a critical threshold value, biochemical processes necessary for LTP induction and expression are activated by molecular crosstalk within the multiprotein complex of the post-synaptic density (PSD) [20]. Many proteins and molecules have been reported to be important for LTP expression, but only a few have been identified as critical for LTP induction, such as calcium/calmodulin-dependent protein kinase II (CaMKII), cyclic adenosine monophosphate-dependent protein kinase (PKA), protein kinase C (PKC), and the extracellular signal-regulated kinase (Erk)/mitogen-activated protein kinase (MAPK) pathway [7]. In contrast to the increasing complexity of LTP mechanisms downstream of NMDAR activation, the upstream mechanisms of post-synaptic depolarization in response to pre-synaptic glutamate release are fairly established during the last decades, pointing to a dominant contribution of AMPA receptors modulated by dendritic voltage-gated Ca 2+ , Na + , K + , and I h channels [3].
Here, we report a novel critical mediator of LTP induction upstream of NMDA receptor activation. We present different lines of experimental evidence, which support that activation of the transient receptor potential (TRP) channel M4 (TRPM4), a calcium-activated, but calcium-impermeable non-selective cation channel, is mandatory for NMDAR activation and the induction of LTP. The TRPM4 belongs to the melastatin subfamily of the TRP membrane proteins. TRP channels are well described for their role in sensory signaling and can be gated by a large variety of stimuli, from chemical to mechanical and to changes in temperature [11]. Among this family of 28 ion channels, TRPM4 and its closest structural relative TRPM5 exhibit some unique properties (for a review, see [23]). TRPM4 expression has been reported in a large range of tissues including several parts of the cardiovascular system and immune cells such as T cells and mast cells [17,29]. Several studies have also detected TRPM4 messenger RNA (mRNA) and protein in the brain of rodents and humans [31,40]. Recently, excessive TRPM4 activity has been associated with neuronal cell death in experimental autoimmune encephalomyelitis, a mouse model of multiple sclerosis [31].

Animals
Trpm4 −/− transgenic mice were previously described [37]. Female Trpm4 −/− and wild-type (WT) littermates, aged between 8 and 12 weeks, were used for all experiments. All animal experiments were in accordance with the European Community Council Directive (86/609/EC) and approved by the local ethics committee.

In situ staining
Control and Trpm4 −/− mouse brains were dissected out in phosphate-buffered saline (PBS) and fixed in 4 % paraformaldehyde (PFA) for 2 h at RT and cryopreserved in 25 % sucrose overnight at 4°C before embedding in optimal cutting temperature (OCT) compound (Tissue-Tek, Sigma-Aldrich). Sections of 12 μm were cut on a cryostat, collected on ProbeOn Plus microscope slides (Fisher Scientific), and stored at −80°C until used.

Hematoxylin and eosin staining
Brains of control and Trpm4 −/− were dissected out in PBS and immersed in 4 % PFA for 2 h at RT. Samples were then placed in 25 % sucrose overnight at 4°C for cryoprotection, embedded in OCT, and frozen. Coronal 12-μm sections containing the hippocampal area were prepared using a cryostat. Hippocampus sections were stained with hematoxylin and eosin in accordance with the standard procedure.

Golgi staining and morphological analysis
Golgi staining was performed according to the manufacturer's instruction (Rapid Golgi staining kit, FD Neurotechnologies, Inc., Ellicott City, MD, USA). Coronal sections (120 μm) containing identical regions of the hippocampus were selected from WT and Trpm4 −/− mice for analysis. Neurons chosen for analysis had to possess the following characteristics: (i) cell bodies had to be located in the middle third of the slice to avoid analysis of neurons extending into other sections; (ii) dark impregnation through all the dendrites, without breaks; and (iii) the neurons had to be isolated from the neighboring impregnated cells to avoid interferences in analysis. The branching density of the apical dendritic trees of CA1 pyramidal neurons was evaluated using the Sholl method [19], where spine density and length of dendrites arising from the soma are determined in the first-(50 μm), second-(50-100 μm), and third-order segments (100-150 μm) from the center of the soma. The number of intersections between the dendrites and the concentric circles was counted and plotted as a function of the distance from the soma. Tracing and counting were performed with a ×100 oil immersion objective (Zeiss). Curvilinear lengths were measured with Image J software. The spine density was measured on secondary branches situated in the proximal apical area (30-120 μm from the soma). On average, 4-8 neurons were included per animal, and five animals were analyzed per genotype. A MAP2 immunostaining on a separate set of slices was used to assess the neuronal density. Briefly, cryosections containing dorsal region of the hippocampus were probed with anti-MAP2 antibody (#4542; Cell Signaling) for 1 h at room temperature and revealed with anti-rabbit IgG antibody (Alexa fluor 488 conjugate; #4412; Cell Signaling). Images were acquired and analyzed using an Apotome (Zeiss). The number of pyramidal-shaped MAP2positive cells in the CA1 area was determined per mm 2 . Three fields per slices, four slices per animal, and five animals per genotype were analyzed. All these analyses were performed blind to the genotype of subjects.

Multi-electrode array recording
Mice were sacrificed by cervical dislocation, decapitated, and brains transferred to ice-cold oxygenated (95 % O 2 , 5 % CO 2 ) artificial cerebrospinal fluid (aCSF; in mM): NaCl, 124; KCl, 3; NaHCO 3 , 26; NaH 2 PO 4 , 1.25; MgSO 4 , 1; glucose, 10; CaCl 2 . Hippocampi were horizontally sliced (250 μm thick) in a vibratome filled with ice-cold oxygenated aCFS. Hippocampal slices were allowed to recover in an interface chamber filled with oxygenated solution at room temperature for at least 2 h. Then, they were placed in a multi-electrode array (MEA) recording chamber and perfused continuously at a rate of 2 ml/min with aCSF at 32°C. Extracellular stimulations and recordings were performed with a MEA set-up (Multi Channels Systems, Reutlingen, Germany) consisting of a MEA1060-BC preamplifier, a filter amplifier (gain 1100×), and a twochannel stimulus generator (STG2002). MEA biochips with 8×8 titanium nitrite electrodes (30-μm diameter and 200-μm spacing) were used. The bath was grounded with an internal reference electrode. Slice position and contact with the electrodes were secured with a nylon mesh (ALA Scientific Instruments, USA). Stimulation electrodes were positioned at the Schaffer collateral pathway, and evoked field excitatory post-synaptic potentials (fEPSPs) were monitored at proximal stratum radiatum. Data acquisition was performed and controlled by MC_Rack 3.2.1.0 and MC_Stimulus II software (Multichannel Systems, Reutlingen, Germany).
Input/output curves were obtained 20 min after placing the slice on the biochip. Test stimuli were delivered as biphasic pulses (100-μs pulse width) at increasing voltages. For each stimulation, data from three consecutive fEPSPs were collected. Baseline stimulation strength for the rest of the experiments was set to produce 40 % of maximal response.
LTP was induced after 40 min of stable baseline by one of the following protocols: 1 or multiple trains of theta-burst stimulation (TBS, 5 bursts of 4 pulses at 100 Hz repeated at 200-ms interval) or high-frequency stimulation (HFS, 1 s at 100 Hz). LTP was expressed as the percent change in the average slope of the fEPSP taken from 40 to 60 min after LTP induction in relation to the average fEPSP slope during the 10-min baseline recording that preceded the conditioning protocol.
In all experiments, the recording of slices from mutant mice was interleaved by experiments with wild-type controls.
Most recording also included a paired-pulse protocol with following interpulse intervals: 20-50-100-200 ms. Data from three consecutive fEPSPs were collected for each stimulation interval. Paired-pulse facilitation (PPF) was calculated as the ratio of the slope from the second to the first response for each paired stimulation event.
Whole-cell patch-clamp recording Following decapitation, brains were isolated as described above and transverse (400 μm thick) hippocampal slices were prepared with a vibratome (MIKROM HM 650V Microm Instruments GmbH, Wallsdorf, Germany) and stored at room temperature in a holding bath containing oxygenated aCSF. After a recovery period of at least 1 h, an individual slice was transferred to the recording chamber where it was continuously superfused with oxygenated aCSF at a rate of 2.5 ml/min. Wholecell recordings from CA1 pyramidal neurons were performed using a patch-clamp amplifier (MultiClamp 700B, Axon Instruments, Molecular Devices, Inc., Sunnyvale, CA, USA). Neuronal patching was performed under visual control by an infrared differential interference contrast optics system (Axioskop2 FC Plus, Zeiss Instruments, Jena, Germany).
To establish the basic membrane properties and intrinsic excitability of the CA1 pyramidal cells, we measured voltage responses to 500-ms duration −100 pA (hyperpolarizing) and +200 pA (depolarizing) current injections applied to the individual neuron at resting potential. Input resistance was calculated using Ohm's law applied to the peak voltage amplitude in response to a 500-ms injection of 100-pA hyperpolarizing current. Membrane time constants were measured from the same voltage responses through single exponential fitting of the recording trace from the onset of voltage deflection to the peak of the membrane charging curve. Voltage sag (a hallmark of hyperpolarization-activated currents) was quantified as the ratio of the steady-state voltage amplitude over the peak amplitude in response to hyperpolarizing step. Action potential parameters were measured from the first spike in response to depolarizing pulse. Spike after-depolarization (ADP) was measured following a single action potential elicited by a 2-ms depolarizing current pulse of +2 nA. Current-clamp recordings were performed using borosilicate glass microelectrodes (3-5 MΩ) containing (mM) K-gluconate, 130; KCl, 20; K-Hepes, 10; EGTA, 0.2; Na-GTP, 0.3; Mg-ATP, 4; pH 7.3, 287 mOsm. The aCSF was additionally supplemented with CNQX (5 μM), AP5 (5 μM), and picrotoxin (50 μM) to block AMPA/kainate, NMDA, and GABA A receptors, respectively. The liquid junction potential of 13 mV was corrected for arithmetically. Only cells with a corrected junction potential and a resting membrane potential more negative than −60 mV were used for experiments. To keep the pre-stimulus membrane potential at a set level of −60 mV, the automatic slow current injection function of the MultiClamp 700B amplifier (5-s time constant) was used.
In the pairing protocol, CA1 pyramidal neurons were patched with a pipette solution containing 135 mM CsMeSO 3 , 10 mM HEPES, 8 mM NaCl, 0.3 mM EGTA, 4 mM Mg-ATP, 0.3 mM Na-GTP, and 5 mM QX-314. LTP in CA1 pyramidal neurons was induced by pairing stimulation of Schaffer collaterals, for 90 s at 2 Hz, with post-synaptic depolarization at a holding potential of 0 mV in voltage-clamp mode [34]. Recordings were carried out at 25°C. Evoked EPSC amplitudes were normalized to baseline amplitudes prior to LTP induction on a cell-by-cell basis (i.e., % of a base line mean).

Ca 2+ imaging
After recovery, acute hippocampal slices were incubated at room temperature for 60 min in oxygenated aCSF, containing 1 μM Fura2-AM (Biotium). Thereafter, they were placed in the recording chamber of an upright fluorescence microscope (Olympus Deutschland GmbH, Hamburg, Germany) and maintained at 37°C using a heated (40×, water-immersion) objective and a temperature-controlled multichannel perfusion system. A concentric bipolar stimulation electrode (World Precision Instruments, Sarasota, USA) was positioned to stimulate Schaffer collateral afferent fibers. Changes in intracellular calcium were monitored in the proximal stratum radiatum (CA1 region). For excitation (at 350 and 380 nm) of Fura-2, a Polychrome V monochromator (Till Photonics, Munich, Germany) was used. A 409-nm beamsplitter and a 510/590BP emission filter allowed detection of the fluorescent signal by a highspeed EM-CMOS camera (Andor iXON+ DU885KCS-VP, Andor Technology Ltd., Belfast UK). Regions of interest were selected, and analysis was performed with Live Acquisition (LA) Software from Till Photonics (Graefelfing, Germany). Fluorescence data was acquired at 2 Hz (illumination time 200 ms at each wavelength). For offline analysis and statistics, data was exported to Origin 7 software (Originlab, Northhampton, USA). For stimulation of Schaffer collaterals, biphasic pulse trains (duration 500 ms at 100 Hz, 1-ms pulse width) were applied with an isolated pulse stimulator Model 2100 (A-M Systems, Sequim, USA) with increasing stimulation strengths of 50, 100, 200, 300, 400, and 500 μA given at a 75-s interval. Synchronization between electrical stimulation and fluorescence recording was assured via the LA software (Till Photonics, Graefelfing, Germany), which triggered both the Polychrome V monochromator and the pulse stimulator, via Patchmaster software (HEKA, Harvard Bioscience Inc., USA) and an EPC9 amplifier (HEKA, Harvard Bioscience Inc., USA). For data analysis, three active regions of interests (ROI) were selected in the proximal stratum radiatum (CA1 region). The fluorescence intensities of these were averaged. Background correction was performed, by subtracting the fluorescence level, both for 350 and 380 nm, from a nearby non-active region of the same size as the recording region, as measured in parallel. Results are presented as ΔF=F−F 0 . F 0 represents the averaged fluorescence ratio before the start of the stimulation train. Per mouse, at least two hippocampal slices were utilized. Four mice per genotype were used for experiments.

Statistical analysis
Data are expressed as mean±SEM. Data were compared by two-tailed Student's t test and repeated-measures ANOVA with Tukey's post hoc analysis. Statistical analyses were performed using Graphpad Prism 5 (GraphPad Software, Inc., La Jolla, CA, USA) and SPSS 19 (Armonk, NY, USA). Statistical significance was accepted at p<0.05.

TRPM4 is localized in the pyramidal cells of the hippocampal CA1 region
We examined TRPM4 distribution in the mouse hippocampus by in situ hybridization and found TRPM4 mRNA expression in the granule cells of the dentate gyrus and pyramidal cells of the area CA1 and to a lesser extent in the CA3 pyramidal layer (Fig. 1a). Western blotting of plasma membrane-enriched lysates from hippocampus verified the presence of TRPM4 in WT but not in Trpm4 −/− hippocampi (Fig. 1b).
For morphological characterization, we performed Golgi staining on the hippocampus. The apical dendrites of CA1 pyramidal neurons of WT and Trpm4 −/− mice were not significantly different in total dendritic length and the number of branch points (Fig. 1d, e). We observed no significant differences in Sholl analysis or in spine density between the two genotypes ( Fig. 1d-f).

TRPM4 deficiency does not impair basal glutamatergic neurotransmission
To study a role of TRPM4 in synaptic communication and plasticity in the hippocampus, we focused on the Schaffer collateral-CA1 synapse, whose mechanisms are particularly well explored. Given the dominant role of glutamate receptors in these processes, we first tested AMPA and NMDA receptor activity and measured evoked EPSCs (eEPSCs) from CA1 pyramidal neurons using the whole-cell voltage-clamp technique. As shown in Fig. 1i, j, AMPA receptor-mediated eEPSCs recorded at post-synaptic membrane potentials of −70 mV (in the presence of 100 μM picrotoxin) and NMDA receptormediated eEPSCs measured at +40 mV (in the presence of 40 μM CNQX and 100 μM picrotoxin) were identical between genotypes, excluding a shift in the ratio of NMDA to AMPA-mediated currents in Trpm4 −/− . Subsequently, we checked by multi-electrode array recordings from acute hippocampal slices of WT and Trpm4 −/− mice whether the absence of TRPM4 protein affects the efficacy of basal synaptic transmission. As shown in Fig. 2a, the input/ output properties of field excitatory post-synaptic potentials (fEPSPs) were indistinguishable between WT and Trpm4 −/− slices. We also tested paired-pulse facilitation in CA1 neurons as a measure of short-term plasticity and did not detect any difference between WT and Trpm4 −/− at all tested interpulse intervals (Fig. 2b).
Next, we induced long-term potentiation by either 1 or 3 trains of theta-burst stimulation (TBS) and examined the effect on synaptic strength. As depicted in Fig. 2e, f, the initial magnitude of potentiation was significantly lower in Post-synaptically expressed LTP develops in two phases, the induction phase, which is dependent on NMDAR activation, and the expression phase, which is associated with phosphorylation and membrane insertion of AMPAR. In order to decipher whether the lack of LTP in Trpm4 −/− was due to a specific defect in the induction or the expression phase, we investigated the phosphorylation state of Ca 2+ /calmodulin-dependent kinase II (CamKII), which is known to be the first step of the signaling cascade upon Ca 2+ influx through NMDA receptors [21,35]. While the phosphorylation level of the kinase increased after TBS stimulation in WT, no difference was detectable in Trpm4 −/ − (Fig. 2g), supporting a TRPM4-dependent defect during the induction phase, before the activation of CamKII by Ca 2+ influx through NMDA receptors.
It has been reported that HFS induces a compound LTP, which is dependent on the activation of both the NMDARs and voltage-dependent Ca 2+ channels (VDCC) [5,13,15]. To test whether the impaired potentiation in Trpm4 −/− could be due to a difference in activation of either of these pathways, we applied D-APV (100 μM), a specific NMDAR blocker, and nifedipine (100 μM), an L-type VDCC blocker prior to and during HFS stimulation. Application of D-APV reduced the level of potentiation in the WT while it had no effect in Trpm4 −/− (WT control n=8 mice; APV n=4 mice; p<0.05 two-way ANOVA; Trpm4 −/− control n=8 mice; APV n=4 mice; n.s two-way ANOVA) (Fig. 3c). In contrast, nifedipine application totally abolished LTP in Trpm4 −/− and decreased potentiation in the WT group (WT control n=8 mice; nifedipine n=5 mice; p<0.05 two-way ANOVA; Trpm4 −/− control n=8 mice; nifedipine n=5 mice; p<0.05 two-way ANOVA) (Fig. 3d). These results, summarized in Fig. 3e, suggest that in Trpm4 −/− , HFS-LTP is induced only through L-type VDCCmediated Ca 2+ influx, i.e., in an NMDAR-independent manner.

Reduced excitability of Trpm4 −/− hippocampal CA1 neurons
To examine the mechanism underlying the lack of LTP induction in CA1 neurons, we studied the electrophysiological properties of CA1 pyramidal cells in more detail, using whole-cell patch-clamp recordings. First, we analyzed miniature EPSPs (mEPSCs) and miniature IPSPs (mIPSCs) at Schaffer collateral CA1 synapses. We found no difference in the amplitude and frequency of mEPSCs and mIPSCs between WT and Trpm4 −/− mice (mEPSCs amplitude WT These data further support that there is no obvious defect in pre-synaptic release of exocytotic vesicles and basal synaptic transmission to CA1 principal cells. Subsequently, we compared the electrophysiological properties and stimulus-driven firing of CA1 neurons by measuring voltage responses to hyperpolarizing and depolarizing current injections (−100 and +200 pA, respectively, for 500 ms), applied at resting potential. As presented in Table 1, Trpm4 −/− and WT mice did not differ in electrophysiological properties of CA1 neurons, such as resting membrane potential, input resistance, action potential amplitude, and amplitude of spike after-depolarization. Notably, the number of spikes during the depolarizing step was significantly lower in Trpm4 −/− compared to WT mice (WT 12.3±0.82, n=12 mice; Trpm4 −/− 9.9 ±0.79, n=6 mice; p=0.037 Student's t test) (Fig. 4a). Since also the half-width and time constant for relaxation of evoked EPSPs were significantly reduced in Trpm4 −/− CA1 neurons, compared to WT neurons (Fig. 4b), this indicates that a depolarization-induced depolarizing current is reduced in Trpm4 −/− CA1 neurons, which contributes to the postsynaptic EPSP upon stimulation of Schaffer collaterals in WT hippocampal slices.
Trpm4 −/− hippocampal CA1 neurons lack a Ca 2+ -dependent cation current and show diminished Ca 2+ transients To unveil TRPM4 channel activity in CA1 neurons directly, we performed whole-cell voltage-clamp measurements in the presence of a cocktail of blockers of glutamate receptors and voltage-gated Ca 2+ , Na + , and K + channels. To isolate a TRPM4-dependent component from the remaining background currents, we compared currents in response to a voltage ramp from −115 to +35 mV (V h =−85 mV), in the absence or presence of high Ca 2+ (10 μM) in the pipette solution. Such Ca 2+ concentration is reached in spine heads in response to depolarization [30]. In these conditions, we found in Trpm4 +/+ CA1 neurons a Ca 2+ -dependent cation current with properties reminiscent of TRPM4 currents, such as E rev around 0 mVand an outwardly rectifying I-V relation [29] (Fig. 4c). As illustrated in Fig. 4d, this current increased in WT littermates in the high-intracellular Ca 2+ condition by 51.5±19.6 % as compared to the low-intracellular Ca 2+ condition. This Ca 2+ -dependent cation current was absent in Trpm4 −/− CA1 neurons, resulting in current responses in the high-intracellular Ca 2+ condition that were not significantly different from the background currents in the low-intracellular Ca 2+ condition (11.7± Fig. 2 Normal basal transmission, short-term plasticity, and LTD but impaired LTP in Trpm4 −/− hippocampal CA1 neurons. a Input/output curves of fEPSP slopes recorded from CA1 neurons at increasing stimulation intensities (0.5-4 V) did not differ between WT (n=12) and Trpm4 −/− (n=12). Representative analog traces are shown in the inset (calibration 5 ms, 0.35 mV). b Paired-pulse facilitation (PPF) at different stimulus intervals in WT and Trpm4 −/− slices. (WT n=14; Trpm4 −/− n=12). c NMDAR-dependent LTD induced by low-frequency stimulation applied at time B0.^No differences in the level of depression were detected between the two groups (WT n=5; Trpm4 −/− n=5). d mGluRdependent LTD induced by bath application of the group I mGluR agonist DHPG did not differ between genotypes (WT n=5; Trpm4 −/− n=6). e NMDAR-dependent LTP was induced by 1 theta-burst stimulation (TBS 5 bursts of 4 pulses at 100 Hz repeated at 200-ms interval, pulse width 100 μs) in WT (n=10) but not  Fig. 4c, d).
The above data suggest that TRPM4 contributes to postsynaptic depolarization of post-synaptic neurons upon stimulation of Schaffer collaterals, which is a critical step for full activation of NMDA receptors and initiation of LTP induction [20]. Hence, deletion of TRPM4 is expected to diminish membrane depolarization and Ca 2+ influx via NMDA receptors, which should be detectable by Ca 2+ imaging as smaller Ca 2+ transients as compared to WT controls. Therefore, we measured Ca 2+ transients in the CA1 dendritic region in response to high-frequency stimulation applied at various stimulation intensities to Schaffer collaterals. As shown in Fig. 4e, f, at low stimulation intensities, there was no difference between WT and Trpm4 −/− CA1 neurons. However, at high stimulation intensities, as used for LTP induction (see Fig. 3a), Ca 2+ levels are significantly lower in Trpm4 −/− CA1 neurons compared to WT (Fig. 4e, f, RM-ANOVA, p<0.01; F (1,7) =30.4).

Rescue of impaired LTP by facilitating NMDAR activation and membrane depolarization
If TRPM4 activity is necessary for NMDA receptor opening, facilitation of NMDA receptor opening or applying a postsynaptic depolarization should rescue the lack of LTP induction in Trpm4 −/− slices. Such a rescue was achieved by increasing the excitability of Trpm4 −/− CA1 neurons via application of the ß-adrenergic agonist isoprenaline (10 μM;  Mean±SEM. n-Numbers are indicated in brackets and represent the number of cells. No significant differences were detected between WT and Trpm4 −/− Fig. 5a, b) Previous reports described that isoprenaline rescues a lack of TBS-induced LTP in mice with a C-terminally truncated GluN2A subunit [26], likely by inhibiting the A-type K + current in CA1 neurons and thus facilitating depolarization. In our hands, isoprenaline completely restored the induction of 1 TBS-LTP in Trpm4 −/− slices (Fig. 5a, b; Trpm4 −/− untreated n=10 mice; Trpm4 −/− isoprenaline n=7 mice; Bonferroni's multiple comparison t=8.322 p<0.0005). In a second independent approach, we postulated that the LTP deficit should be overcome if the intrinsic depolarization is bypassed by using a pairing protocol for LTP induction under whole-cell patch-clamp conditions (similar as in [34]). This protocol Thus, these data strongly support the hypothesis that the LTP defect is caused by a lack of sufficient post-synaptic depolarization in response to TBS or HFS stimulation.

Discussion
In this study, we took advantage of Trpm4 knockout mice [37] to provide the first direct evidence for a function of TRPM4, and Ca 2+ -activated non-selective cation channels in general, in neurons. Through in situ hybridization and western blotting, we show that TRPM4 is expressed in the CA1 region and the dentate gyrus of the hippocampus and to a lesser extent also in CA3 pyramidal neurons. While Trpm4 −/− mice are normal with regard to morphology, basal synaptic transmission, and paired-pulse facilitation at CA3-CA1 synapses, we found a complete lack of LTP induced by common theta-burst protocols to Schaffer collateral fibers. When high-intensity protocols were applied, such as 5-8 trains of theta-burst stimulation or 2 trains of high-frequency stimulation (HFS) at 100 Hz, Trpm4 −/− CA1 neurons developed LTP but to a lesser level than WT neurons. In contrast to LTP, NMDAR-and mGluR-dependent LTD in CA1 neurons were unchanged in Trpm4 −/− mice. Together, these data indicate that deletion of Trpm4 leads specifically to an increase in the threshold for LTP induction, without affecting the requirements for induction of LTD. Patch-clamp experiments showed that AMPA and NMDA receptor currents and their ratio are indistinguishable between WT and Trpm4 −/− CA1 neurons, excluding changes in the activity or expression of these receptors as a reason for the observed phenotype. Instead, our data indicate an essential function of TRPM4 in the generation of a depolarizing current that is sufficiently strong to fully unblock NMDA receptors as mandatory condition for the induction of LTP. Accordingly, we have identified a Ca 2+ -dependent cation current in CA1 pyramidal neurons, which is absent in Trpm4 −/− mice. The lack of TRPM4 proteins leads to a reduced length of evoked excitatory post-synaptic currents and reduced excitability in CA1 neurons. Moreover, Ca 2+ signals in CA1 neurons upon electrical stimulation of Schaffer collaterals are reduced in Trpm4 −/− neurons compared to WT, which is consistent with a reduction of Ca 2+ influx due to incomplete recruitment of NMDA receptors in Trpm4 −/− mice. Consistent with the idea of insufficient post-synaptic depolarization as underlying cause of the observed LTP deficit, this phenotype can be completely rescued by application of isoprenaline, i.e., an intervention that facilitates the post-synaptic depolarization of CA1 neurons. Most convincingly, when the depolarization of Trpm4 −/− neurons is artificially clamped to the same voltage as WT neurons during a pairing protocol, LTP in both genotypes is indistinguishable with respect to amplitude and time course. Taken together, our data are consistent with a novel mechanistic model of LTP induction at Schaffer collateral-CA1 synapses in hippocampal slices, in which TRPM4 acts as an amplifier of the post-synaptic depolarization, which is indispensable to fully unblock the NMDA receptor and to initiate the induction of LTP in CA1 hippocampal neurons. As summarized in Fig. 6, we propose that in WT hippocampal CA1 neurons, the initial TBS stimulation leads to AMPA receptormediated depolarization of the post-synaptic membrane, which partially relieves the Mg 2+ block of NMDA receptors. The resulting Ca 2+ influx into the post-synaptic neuron activates TRPM4, which mediates a depolarizing current that further depolarizes the post-synaptic membrane, thereby fully unblocking NMDA receptors. Thus, TRPM4 activity facilitates NMDAR gating to a level necessary for activation of crucial plasticity-related downstream signal transduction cascades. In the absence of TRPM4, activation of NMDA receptors is insufficient to properly induce LTP and the downstream signaling mechanisms. In this model, it is not surprising that the reduced post-synaptic depolarization in Trpm4 −/− CA1 neurons is sufficient to induce NMDA receptordependent LTD. At resting membrane potential, the Mg 2+ block of NMDARs is incomplete allowing still a substantial Ca 2+ influx which is sufficient to trigger the downstream signaling pathways that mediate LTD [30]. Moreover, recent work suggested that in young animals, LTD can be induced without any Ca 2+ influx through NMDA receptors [28].
The lack of a substantial NMDAR-mediated component of potentiation during strong stimulation and the apparent reduction to a VDCC-mediated LTP-component in Trpm4 −/− CA1 neurons is likely to have far-reaching consequences on Ca 2+dependent signaling cascades and gene activation. The ability of calcium influx to induce these events varies according to its amplitude, duration, and location [14,30]. Whereas opening

WT
Trpm4 -/- Fig. 6 Putative function of TRPM4 in LTP induction. Pre-synaptic glutamate release causes membrane depolarization (ΔV m ) via activation of AMPA receptors (AMPAR). This depolarization triggers Ca 2+ influx through L-type voltage-dependent calcium channels (L-VDCC) and NMDA receptors (after the Mg 2+ block at the channel pore of NMDA receptors is relieved by depolarization), depending on the conditioning protocol. The initial depolarization and the rise of [Ca 2+ ] i activates TRPM4 which leads to further membrane depolarization which fully unblocks the NMDAR. The resulting large intracellular calcium increase triggers phosphorylation of Ca 2+ /calmodulin-dependent kinase II (CamKII), which in turn leads to phosphorylation of inserted AMPAR and insertion of new AMPAR into the synaptic membrane. Together, this results in the induction of robust LTP. In Trpm4 −/− mice, the lack of TRPM4-mediated depolarization causes only a partial removal of the magnesium block of NMDAR and a reduced calcium influx via this channel, which is insufficient to trigger LTP in Trpm4 −/− mice. Facilitating NMDAR activation (with serine), reducing the inhibitory tone via the A-current (ß-adrenergic stimulation), or additional depolarization of the post-synaptic membrane in a pairing protocol rescues the induction of LTP in Trpm4 −/− mice (see the text for further details) of NMDARs leads primarily to relatively long-lasting Ca 2+ influx into dendritic spines, Ca 2+ influx via VDCC was found to be short-lived and sensed by a multiprotein complex in the vicinity of the VDCCs that contains calmodulin and CaMKII [9,38]. On the other hand, L-type VDCC subtypes (Ca v 1.1. and Ca v 1.2) bind calmodulin at carboxy-terminal recognition sites [14] enabling calmodulin to sense elevated Ca 2+ in microdomains in the immediate vicinity [30] and to mediate the activation of signaling cascades such as the ERK/MAPK and PKA pathways, respectively [9,39]. TRPM4 and TRPM5 are to date the only molecular candidates for the class of Ca 2+ -activated non-selective cation channels [23]. Recently, it was shown that TRPM5, but not TRPM4, contributes to slow after-depolarizations in mouse prefrontal cortex neurons [18]. Others reported that TRPM4 and TRPM5 contribute to, but are not absolutely required for, depolarization-induced slow currents in cerebellar Purkinje cells [16]. While these studies described the involvement of TRPM4 in intrinsic neuronal properties by single-cell recording, the function of TRPM4 in specific neuronal tasks and in synaptic plasticity remained unclear. Previous studies have suggested a physiological role of TRPM4 in close functional relation with glutamate receptors [8,25,27]. The most investigated example is the process of burst firing in the pre-Bötzinger complex neurons [8] where TRPM4-like currents have been proposed to be responsible for amplifying glutamatergic synaptic drive, by transforming the glutamatergic synaptic inputs to long-lasting membrane depolarization [25]. Mrejeru et al. have described a similar mechanism in dopaminergic (DA) neurons of substantia nigra [27]. It was shown that in those neurons, NMDA currents recruit a Ca 2+ -activated non-selective current, which can be blocked by non-specific TRPM4 blockers such as flufenamic acid and 9-phenanthrol. Using Trpm4 −/− mice, Schattling et al. proposed a pathophysiological role for TRPM4 in glutamate stress-induced neurodegeneration [31]. Glutamate stress is an element of the pathophysiology of experimental autoimmune encephalopathy and multiple sclerosis, which is responsible for neuronal cell death and contributes to the progressive loss of motor functions during this condition. Trpm4 −/− mice are partly resistant to the development of EAE, specifically because Trpm4 −/− neurons are resistant to cell death induced by glutamate overstimulation [31]. Our data suggest that the higher resistance of Trpm4 −/− neurons is likely to be caused (at least in part) by reduced Ca 2+ influx through NMDA receptors upon glutamate stress leading to a decrease of Ca 2+ -induced neurotoxicity. Several other TRP channels are highly expressed in the brain, and some TRP channels, most notably TRPV1, have been proposed to participate in mechanisms of synaptic plasticity in the central nervous system (CNS) [36]. However, their roles in synaptic plasticity are a matter of debate (see for instance [6,12,22], but also [4]).
In conclusion, our data highlight TRPM4 as a key mediator for NMDAR-mediated types of LTP while it does not seem to have a function in LTD induction. Our data support a mechanism in which Ca 2+ influx through NMDA receptors and membrane depolarization trigger the activation of TRPM4, which further amplifies membrane depolarization via a TRPM4-mediated Ca 2+ -dependent cation current in order to fully unblock NMDA receptors.