NMDA Receptor-Dependent Synaptic Activation of TRPC Channels in Olfactory Bulb Granule Cells

Abstract

Canonical transient receptor potential (TRPC) channels are widely expressed throughout the nervous system including the olfactory bulb where their function is largely unknown. Here, we describe their contribution to central synaptic processing at the reciprocal mitral and tufted cell–granule cell microcircuit, the most abundant synapse of the mammalian olfactory bulb. Suprathreshold activation of the synapse causes sodium action potentials in mouse granule cells and a subsequent long-lasting depolarization (LLD) linked to a global dendritic postsynaptic calcium signal recorded with two-photon laser-scanning microscopy. These signals are not observed after action potentials evoked by current injection in the same cells. The LLD persists in the presence of group I metabotropic glutamate receptor antagonists but is entirely absent from granule cells deficient for the NMDA receptor subunit NR1. Moreover, both depolarization and Ca²⁺ rise are sensitive to the blockade of NMDA receptors. The LLD and the accompanying Ca²⁺ rise are also absent in granule cells from mice deficient for both TRPC channel subtypes 1 and 4, whereas the deletion of either TRPC1 or TRPC4 results in only a partial reduction of the LLD. Recordings from mitral cells in the absence of both subunits reveal a reduction of asynchronous neurotransmitter release from the granule cells during recurrent inhibition. We conclude that TRPC1 and TRPC4 can be activated downstream of NMDA receptor activation and contribute to slow synaptic transmission in the olfactory bulb, including the calcium dynamics required for asynchronous release from the granule cell spine.

Introduction

The most abundant type of synaptic connection in the vertebrate olfactory bulb (OB) is the reciprocal dendrodendritic synapse between the mitral/tufted cells (MTCs) and granule cells (GCs) where glutamate is released from the MTC dendrite onto the GC spine that in turn can release GABA. Sufficient excitation of GC dendrites results in the spread of inhibition to other MTCs. Thus, the synapse can mediate both recurrent and lateral inhibition of MTCs (for review, see Shepherd and Greer, 2004). Both types of inhibition show a prominent asynchronous component that lasts up to several 100 ms (Jahr and Nicoll, 1980; Isaacson and Strowbridge, 1998) and is probably located within the GC compartment of the reciprocal microcircuit (Chen et al., 2000), contained by large spines also known as gemmules. Possible mechanisms include an I_A current that delays GC spiking (Schoppa and Westbrook, 1999; Kapoor and Urban, 2006), and sluggish extrusion of Ca²⁺ from the gemmule (Egger and Stroh, 2009).

In addition, we have suggested that the nonselective cation current I_CAN first identified in frog GCs (Hall and Delaney, 2002) may also promote asynchronous release (Egger, 2008). In both frog and rat GCs, this plateau current and the associated long-lasting depolarization (LLD) are supported by NMDA receptor (NMDAR) activation and blocked by both flufenamic acid and increased intracellular Ca²⁺ buffering. In the olfactory bulb, LLDs are present in many neuronal classes [e.g., an NMDAR-independent form in mitral cells (Carlson et al., 2000)]. In GCs, a slow global extra Ca²⁺ entry was directly correlated to the duration of the LLD and mediated by the persistent activation of T-type voltage-dependent Ca²⁺ currents (VDCCs) (Egger, 2008).

Several of these properties resemble that of slow synaptic currents mediated by transient receptor potential (canonical) (TRPC) channels that have been described recently elsewhere in the CNS (Amaral and Pozzo-Miller, 2007; Clapham, 2007; Fowler et al., 2007; Hartmann et al., 2008; Becker et al., 2009). The seven TRPC channels, one of six subfamilies of the TRP cation channel superfamily, are expressed in many cell types (for review, see Moran et al., 2004; Birnbaumer, 2009). In the nervous system, TRPCs were initially identified in sensory nerve endings and later found also in axons and dendrites of central neurons. Functionally, TRPCs have been mostly implicated in sensory transduction and neurite outgrowth and development with a role in synaptic transmission emerging only recently. For example, while TRPC2 has been identified to play a crucial role in pheromone detection in the vomeronasal system (for review, see Zufall et al., 2005), there is no functional data on TRPC action in the main olfactory bulb so far.

Here, we tested the hypothesis that a TRPC channel is the molecular correlate of I_CAN in GCs, using electrophysiological recordings and two-photon laser-scanning microscopy of local Ca²⁺ signals in GCs of mice deficient for different subtypes of TRPC proteins. We demonstrate a direct involvement of TRPC channels in synaptic excitation in higher sensory processing and provide evidence for a novel, NMDAR-mediated mechanism of activation of TRPCs that also affects neurotransmitter release from GC dendrites.

Materials and Methods

Animals, slice preparation, and electrophysiology.

TRPC1/TRPC4 double-knock-out (TRPC1/4 DKO) mice were generated by breeding TRPC1^−/− (Dietrich et al., 2007) and TRPC4^−/− mice (Freichel et al., 2001). To generate TRPC1/TRPC4/TRPC5 triple-knock-out mice (used as controls for immunohistochemistry), the DKO mice were bred with mice in which exon 4 of the TRPC5 gene was deleted (Xue et al., 2011). All TRPC-deficient mice were on a mixed 129SvJ genetic background. The mice with the GC-specific NR1-knock-out were the same as the ones used in the study by Abraham et al. (2010) (see their Methods for the generation of the mice, injection of the virus, and the visualization of infected GCs using two-photon imaging). Sagittal olfactory bulb brain slices (thickness, 250 μm) were prepared of WT and mutant mice of either sex (P16–P213, WT Bl6 and 129SvJ strain), in accordance with the European Commission Council directives for animal care. For the influence of age on the physiological readout, see Results. All experiments were performed at room temperature (∼21°C). For all current-clamp experiments in GCs, the intracellular solution contained the following (in mm): 130 K-methylsulfate, 10 HEPES, 4 MgCl₂, 4 Na₂ATP, 0.4 NaGTP, 10 Na-phosphocreatine, 2 ascorbate, at pH 7.2, and in addition 100 μm of the calcium-sensitive dye OGB-1 (Invitrogen) for imaging experiments. The extracellular ACSF was bubbled with carbogen and contained the following (in mm): 125 NaCl, 26 NaHCO₃, 1.25 NaH₂PO₄, 20 glucose, 2.5 KCl, 1 MgCl₂, and 2 CaCl₂. Pharmacological agents used were d-APV (25 μm; Sigma-Aldrich), 6-amino-N-cyclohexyl-N,3-dimethylthiazolo[3,2-a]benzimidazole-2-carboxamide (YM289189), and 2-methyl-6-(phenylethynyl)pyridine (MPEP) (10 and 40 μm; Tocris). For voltage-clamp experiments in mitral cells, the internal solution contained the following (in mm): 125 Cs-methylsulfate, 10 HEPES, 10 TEA, 4 MgCl₂, 2.5 Na₂ATP, 0.4 NaGTP, 10 Na-phosphocreatine, 2 ascorbate, 10 mm QX-314 at pH 7.2. The ACSF was the same as above, except for the omission of MgCl₂. Cells were recorded from in whole-cell mode with an EPC-9 amplifier (HEKA).

Activation of the MTC input was achieved via glomerular stimulation (cf. Egger, 2008) using an STG-1004 stimulator (Multi Channel Systems). To ensure sufficient connectivity for suprathreshold stimulation, most patched GCs were located close to the mitral cell layer (i.e., superficially) and thus were likely to be connected to both mitral and tufted cells (Orona et al., 1983). Synaptic APs in GCs were evoked by glomerular stimulation [glomerular action potentials (glAPs)]; artificial APs [somatic action potentials (sAPs)] were evoked by brief somatic current injections (1 ms, 1000 pA). The average resting potential was at V_m = −79 ± 4 mV (n = 17). GCs were generally held at a potential of −70 to −80 mV; cells that required >50 pA of holding current were not accepted.

For mitral cell recordings in voltage clamp, we used pipettes sized 4–6 MΩ and achieved series resistances R_s of <20 MΩ. R_s was monitored during the experiments. Dendrodendritic recurrent inhibition (DDI) was elicited by depolarizing mitral cells from −70 to 0 mV for 20 ms. To prevent rundown, DDI was measured only every 60–120 s and analyzed after its amplitude stabilized (usually ∼20 min after break in).

A subset of GCs both in WT and TRPC1/4 DKO were also filled with biocytin and processsed and reconstructed using Neurolucida (MicroBrightField), as described previously (Lübke et al., 2000).

Two-photon imaging.

Imaging was performed on an Fluoview 300 system (Olympus), with two-photon excitation at 800 nm provided by a Ti:sapphire solid-state laser system (Mai Tai; Spectra Physics).

Immunohistochemistry.

Young adult WT and TRPC1/4/5 triple knock-out mice were anesthetized and transcardially perfused with 4% PFA in 0.1 m PB. Animals assigned to fine-structure analysis were perfused with a fixative containing 4% PFA and 0.05% glutaraldehyde in 0.1 m PB. The olfactory bulbs were removed and cut coronally (section thickness, 50 μm) on a vibratome, and the sections were stored in PB. For immunofluorescence labeling with rabbit anti-TRPC1, rabbit anti-TRPC4, or rabbit anti-TRPC5, respectively, sections were washed in 0.1 m PB and incubated in a blocking solution (5% normal goat serum, 2% bovine serum albumin, 0.1% Triton X in PB) for unspecific protein binding sites for 60 min. Sections were then incubated with the primary antibodies (final concentration, 0.5 μg/μl in blocking solution; overnight at 4°C). Then, sections were washed in PB and incubated with secondary goat anti-rabbit antibody coupled to Cy3 for 45 min (1:600; Dianova) at room temperature. Finally, sections were washed in PB, counterstained using DAPI (30 s; Hoechst), washed again in PB, and mounted for image analysis.

For electron microscopy, sections of the olfactory bulb sections were washed in PBS and cryoprotected in a solution containing 25% sucrose and 10% glycerol in 50 mm PBS. The sections were freeze-thawed and incubated in blocking solution containing 5% NGS in 0.1 m PB for 1 h, followed by incubation with primary antibodies (final concentration, 2 μg/μl, for 24 h at 4°C). After washing in phosphate buffer, the sections were incubated with biotinylated goat anti-rabbit secondary antibody (1:250; Vector Laboratories) for 4 h at room temperature, washed, and finally incubated for 1 h at room temperature with ABC solution (Vector Elite Kit; Vector Laboratories). Visualization of the antibody binding was performed using peroxidase reaction with DAB (diaminobenzidine) as the substrate. To exclude unspecific staining due to the secondary antibody, we performed negative controls using the same staining procedure except primary antibody incubation. After the sections were treated with OsO₄, they were stained with uranyl acetate, dehydrated, and flat-embedded in epoxy resin (Durcupan ACM; Fluka; Sigma-Aldrich). Ultrathin sections were cut and analyzed with a Philips CM100 electron microscope.

Data analysis.

Two-photon imaging data were analyzed as described previously (Egger, 2008). Changes in calcium were measured in terms of ΔF/F (Egger et al., 2005). Throughout the paper, τ_1/2 denotes half-durations from the peak amplitude of changes in ΔF/F or V_m onward, which were in the case of V_m measured between the onset of the LLD/afterdepolarization (ADP) right after the sodium spike and one-half of its maximum amplitude. Both LLD amplitude and τ_1/2 are often analyzed relative to the ADP following sAPs to eliminate variabilities that arise from the nonsynaptic component of the LLD (i.e., the ADP): the smaller the ratio, the stronger the synaptic contribution. The averaged dendrodendritic inhibition recorded in mitral cells was analyzed with respect to its τ_1/2 and total charge after the stimulus artifact.

To assess statistical significance levels, the nonparametric Wilcoxon matched-pairs signed-ranks test was applied for comparing paired data sets (e.g., parameters of the sAP and glAP in the same cells), while the nonparametric Mann–Whitney test was used to compare unpaired data (e.g., the ADP half-duration in young vs adult animals). The entire set of data from WT, TRPC1 KO, TRPC4 KO, and TRPC1/4 KO animals presented in Figure 3D was compared using the Kruskal–Wallis test; differences between individual data pairs were evaluated with the Mann–Whitney test followed by the appropriate Bonferroni–Holm correction. All averages are given ±SD.

Figure 3.

Presence of TRPC4 and TRPC1 in GCs; TRPC4 deletion reduces the plateau, TRPC1/4 deletion blocks it. A, TRPC4 immunoreactivity is detectable in the olfactory GCL in WT mice (a; red) but not in TRPC1/4/5 triple KO (TKO) mice (b). TRPC1 immunoreactivity is strongly visible in the olfactory GCL in WT mice (c) but not in TRPC1/4/5 TKO mice (d). DAPI counterstain is in blue. Scale bar in d applies to all. B, Examples for weaker or missing LLDs in TRPC4^−/−, TRPC1/4 DKO, and TRPC1^−/− animals. Averaged traces are shown. C, Distributions of the ratios of sAP/glAP ADP τ_1/2 and amplitude for WT and the three different deletions. WT, Black columns; TRPC1^−/−, red line; TRPC4^−/−, blue line; TRPC1/4 DKO, purple columns. Indication of overlap regions of WT and DKO distributions via cityscape. D, Summary of the average ratios for all four conditions. For the TRPC1/4 DKO, the ratio of the sAP-ADP versus the EPSP τ_1/2 just below threshold is also shown. The total significance level for the difference between the most unequal values is p < 0.0001 for both ratios (Kruskal–Wallis test). Significance levels of individual results in KO animals versus WT indicated on the respective bars with asterisks and of single KO animals versus DKO with plus signs; ⁺p < 0.05; ^++,**p < 0.02; ***p < 0.001; no symbol, not significant (pairwise Mann–Whitney tests with Bonferroni–Holm correction). Error bars represent SD.

Results

Synaptic spikes mediate long-lasting depolarization and correlated Ca²⁺ entry in mouse GCs

To verify the presence of the synaptic plateau current I_CAN in mouse GCs, we applied glomerular stimulation of MTC apical dendritic tufts (Schoppa et al., 1998) (Fig. 1A; 500–800 μA pulses, 100 μs duration) in acute brain slices (P17–P132) while recording in whole-cell current-clamp mode from GCs. This stimulation evoked EPSPs in GCs. Upon suprathreshold stimulation strength, the synaptic depolarization resulted in glAPs. Similar to our previous findings in rat, these glAPs were frequently followed by a LLD as shown in Figure 1C (mean amplitude above resting potential, 17.7 ± 6.4 mV; mean half-duration τ_1/2, 176 ± 175 ms; n = 49 GCs). The LLD was significantly larger and longer-lasting than the ADP following current-evoked sAPs in the same cells (by somatic injection of a 1000 pA step for 1 ms), with a mean sAP-ADP amplitude above resting potential of 11.1 ± 4.7 mV and a mean half-duration τ_1/2 of 42 ± 22 ms (p < 0.001 for both; Fig. 1D). On average, the LLD lasted also considerably longer than the glomerular EPSP recorded just below the glomerular AP threshold in the same cell (τ_1/2, 166 ± 174, vs 88 ± 58 ms; n = 39; p < 0.001). Thus, the LLD is not just a barrage of EPSPs due to MTC–GC network activity; its generation most likely requires both synaptic input and a Na⁺ AP.

Figure 1.

Synaptically evoked action potentials cause a plateau current in mouse granule cells. A, Scheme of recording configuration (Egger, 2008). Somatically evoked APs (sAP) are generated by current injection, synaptic APs by glomerular stimulation (glAP). B, Two-photon scan of a wild-type mouse granule cell filled with 100 μm OGB-1. C, Data recorded from the cell in B. Top trace, Somatic voltage recording of an sAP and a glAP. Note the time course of the ADP for the current-evoked signal versus the LLD for the synaptic signal. Bottom traces, Averaged fluorescence transients in response to the above APs recorded in a GC spine and its parent dendrite. D, Scatterplots of LLD versus ADP half-duration τ_1/2 and amplitude (top; n = 49 cells) and of the respective fluorescence signals (ΔF/F)_glAP versus (ΔF/F)_sAP (bottom; n = 43 dendrites and spines).

The occurrence of the LLD did not depend on the age of the animals: upon averaging the LLD and ADP τ_1/2 and amplitude data in three age groups of WT mice—young (P16–P22; n = 32), juvenile (P32–P42; n = 6), and adult (P66–P139; n = 7)—we found that, although both amplitude and duration of the plateau decreased slightly with age, I_CAN was activated in all three groups (all significance levels LLD vs ADP for both amplitudes and τ_1/2: p < 0.005). The average LLD τ_1/2 was always significantly longer than the average EPSP τ_1/2 just below threshold (P66–P139: p < 0.01).

To verify that the LLD is also correlated to global Ca²⁺ entry, we used two-photon laser scan imaging to analyze Ca²⁺ entry into GCs at high spatial resolution. We found that the LLD is accompanied by extra, slow Ca²⁺ entry both in the dendrite and the large reciprocal GC spines similar to our previous findings in rat (Egger, 2008). Figure 1C shows ΔF/F transients in a spine and its parent dendrite of a mouse GC. (ΔF/F)_glAP was found to decay significantly more slowly than (ΔF/F)_sAP in the same locations (τ_1/2, 1200 ± 800 ms, vs 750 ± 460 ms; n = 40 spines in 27 cells; p < 0.001; Fig. 1D). The amplitudes of (ΔF/F)_glAP were as large or larger than of (ΔF/F)_sAP (39 ± 15 vs 33 ± 14%; amplitude ratio glAP/sAP, 1.21 ± 0.28; p < 0.001). For both data sets, there were no significant differences between dendrites and spines, so in Figure 1D all data are pooled (n = 51 locations).

LLD requires GC NMDARs but not mGluRs

In frog and rat GCs, the activation of the plateau current I_CAN was found to require a cytosolic rise in [Ca²⁺]; synaptically evoked plateau currents were sensitive to NMDAR blockade (Hall and Delaney, 2002; Egger, 2008). Like in other species, mouse NMDARs already participate in basal MTC–GC synaptic transmission under near-resting conditions (Laaris et al., 2007; Abraham et al., 2010).

As shown in Figure 2, A and B, application of 25 μm d-APV markedly reduced the LLD half-duration and amplitude (235 ± 236 vs 92 ± 44 ms; 20.8 ± 8.1 vs 15.6 ± 6.8 mV; p < 0.002 for both; n = 15 cells). As to the [Ca²⁺] transients, APV made (ΔF/F)_glAP decay as quickly as (ΔF/F)_sAP and have the same amplitude (τ_1/2 ratio glAP/sAP, control, 2.26 ± 1.10, vs APV, 0.92 ± 0.08; n = 6; p < 0.025; amplitude ratio, control, 1.34 ± 0.29, vs APV, 1.06 ± 0.12; n = 7; p < 0.025). There was no effect of APV on (ΔF/F)_sAP (mean amplitude, control, 28 ± 8%, vs APV, 26 ± 10%).

Figure 2.

The synaptic plateau current is NMDAR dependent but does not require activation of group I mGluRs. A, Individual experiment showing sAP, glAP recorded at the soma and the respective ΔF/F at a dendritic location under control conditions (black traces) and in 25 μm d-APV (gray traces). B, All APV experiments: Effect of APV on LLD τ_1/2 and amplitude (n = 15 cells) and ratios of (ΔF/F)_glAP to (ΔF/F)_sAP for τ_1/2 and amplitude (n = 6 locations). C, Top, Individual recording from a granule cell deficient for NR1 (Abraham et al., 2010) showing sAP and glAP. The glAP trace (black) is shown in overlay of the sAP (gray). Bottom, Histogram plot of average ratios of sAP/glAP ADP τ_1/2 and amplitude of NR1-deficient GCs (n = 5) and WT GCs (n = 49). Error bars represent SD. D, Top, Individual experiment showing sAP, glAP recorded at the soma under control conditions (black traces) and in the presence of 10 μm YM289189 and 40 μm MPEP (gray traces). Calibration is as in A. Bottom, All experiments with blockade of mGluR1/5: effect of blockers on LLD τ_1/2 and amplitude (n = 5 cells).

Given that APV might exert presynaptic effects on MTC NMDA autoreceptors (Isaacson, 1999; Salin et al., 2001) or affect the efficiency of glomerular stimulation (but see Egger et al., 2005; Egger, 2008), we also tested for the existence of the LLD in GCs that were deficient for the essential NMDAR subunit NR1. The GC-specific deletion was achieved via viral expression of Cre recombinase in GCs of conditional GluN1^2lox animals (Abraham et al., 2010). In none of the GCs tested was there any indication of a plateau current. The mean glAP/sAP ratio was 0.97 ± 0.10 for the amplitude and 1.00 ± 0.14 for τ_1/2 of the afterdepolarization (Fig. 2C), while the absolute sAP ADP amplitudes and τ_1/2 were not significantly different from the WT values given above (NR1 KO: ADP amplitude, 13.0 ± 5.6 mV; ADP_τ_1/2, 29 ± 19 ms; n = 5 cells). Because MTCs were not infected, this experiment allows excluding any unspecific effects of NMDAR blockade on synapses other than those of GCs. Thus, the activation of NMDARs located in GCs is essential for the generation of the LLD and the concomitant Ca²⁺ entry.

Slow depolarizing currents such as I_CAN have been associated with activation of G-protein-coupled receptors, in particular group I metabotropic receptors (Congar et al., 1997; Pace et al., 2007). So far, GCs are known to prominently express either mGluR5 or mGluR1, depending on their depth within the GC layer (GCL), and accordingly the application of group I agonists was found to increase GC excitability (for review, see Dong et al., 2009). To test for a role of group I mGluRs, we first recorded control responses, and then coapplied 10 μm of the mGluR1 antagonist YM289189 and 40 μm of the mGluR5 antagonist MPEP and recorded GC responses after at least 10 min of wash-in. This procedure has been reported to block mGluRs in acute brain slices (Fukunaga et al., 2007; Zheng and Raman, 2011). Figure 2D shows that blockade of group I mGluRs did not decrease the LLD τ_1/2 and amplitude (297 ± 287 vs 336 ± 315 ms; 22.2 ± 5.6 vs 21.6 ± 5.3 mV; n = 5 cells; compare with APV results above). Thus, group I mGluRs do not play an essential role in the generation of the LLD.

TRPC channels mediate the LLD

Another possible candidate for the mechanism underlying the plateau current is the activation of TRPC channels, because members of this channel family cause depolarization via their nonselective cation conductance and have been associated previously with plateau-like currents (Birnbaumer, 2009). TRPC4 is the most abundant TRPC in the olfactory bulb (Lein et al., 2007; Zechel et al., 2007). We confirmed the presence of TRPC4 protein in the GCL (Fig. 3Aa,b) and the external plexiform layer (EPL) of the mouse OB, also at the ultrastructural level (Fig. 4A–D) with anti-TRPC4 antibodies. Then we tested for the existence of an LLD in GCs of TRPC4-deficient mice (TRPC4^−/−) (Freichel et al., 2001). To render the graphical depiction more compact, the following data are presented in Figure 3B–D as ratios of sAP to glAP afterdepolarization values. In the absence of TRPC4, the plateau current following the synaptic AP was reduced and, accordingly, the sAP/glAP-ADP τ_1/2 ratio was significantly larger than in the WT (0.64 ± 0.16, n = 14, vs 0.38 ± 0.23, n = 49; p < 0.001 for both), while the ADP ratios, the absolute mean ADP amplitude, and τ_1/2 were similar in TRPC4^−/− and WT (p > 0.05 for all). Thus, NMDAR-dependent plateau currents were significantly reduced but not completely absent in TRPC4^−/− mice (compared with the NR1 deletion described above).

Figure 4.

TRPC4 and TRPC1 are present in granule cell somata and synapses in the external plexiform layer at the ultrastructural level. Ultrastructural analysis using preembedding diaminobenzidine labeling for TRPC4 revealed a strong labeling of the granule cell somata (arrows) in WT mice (A), but not in controls lacking primary antibody incubation (B). In higher magnifications of the EPL, many immunopositive presynaptic (arrowheads) and postsynaptic (arrows) profiles are detectable in WT (C), but not in controls (D). A similar analysis for TRPC1 revealed a strong labeling of the granule cell somata (arrows) in WT mice (E), but not in controls lacking primary antibody incubation (F). In higher magnifications of the EPL, many immunopositive dendrites are detectable in WT (G), but not in controls (H).

TRPC4 belongs to a subgroup of TRPC subunits with two other members—TRPC1 and TRPC5. All three readily associate with each other forming heteromeric channels (Strübing et al., 2001; Goel et al., 2002; Hofmann et al., 2002). Further immunohistochemistry with anti-TRPC1 antibodies confirmed the presence of TRPC1 protein in the GCL (Fig. 3Ac,d) and the EPL of the mouse OB, also at the ultrastructural level (Fig. 4E–H). Expression of TRPC5 in GCs was not observed (data not shown). Thus, we hypothesized that the remaining LLD in the TRPC4^−/− mice is carried by TRPC1. Indeed, in mice deficient for both TRPC1 and TRPC4 (TRPC1/4 DKO), the plateau currents were almost completely absent (Fig. 3B–D). Here, both the amplitude and τ_1/2 ratios were significantly different from the WT (n = 21 vs n = 49; p < 0.001 for both), whereas the absolute sAP-ADP amplitude and τ_1/2 were unaltered in the TRPC1/4 DKO animals compared with WT mice (p > 0.05 for both). The remaining small difference between sAP- and glAP-after depolarizations is explained by the EPSP underlying the glAP: the τ_1/2 values of the EPSP just below threshold and the glAP-afterdepolarization were similar in the TRPC1/4 DKO (71 ± 54 vs 69 ± 38 ms; n = 21); accordingly, the τ_1/2 ratio of sAP-ADP/EPSP is the same as for sAP-ADP/glAP-ADP (Fig. 3D). Moreover, the EPSP τ_1/2 values were not significantly different between TRPC1/4 DKO and WT (p = 0.40), reinforcing the notion that EPSPs alone do not suffice to generate the LLD.

Our finding that both TRPC4 and TRPC1 contribute to the long-lasting synaptic signals in OB GCs was corroborated by results obtained in TRPC1^−/− mice (n = 7 GCs; Fig. 3B–D). In summary, the deletion of either TRPC1 or TRPC4 caused significantly smaller effects on the plateau current τ_1/2 ratios than the deletion of both genes (TRPC1^−/− vs DKO, p < 0.05, and TRPC4 ^−/− vs DKO, p < 0.02; statistical significances for ADP ratio comparisons between different mouse lines evaluated by Kruskal–Wallis test followed by pairwise Mann–Whitney test with appropriate Bonferrroni–Holm correction).

We then tested whether the extra Ca²⁺ entry associated with the plateau was also abolished in the TRPC1/4 DKO animals, using two-photon laser-scanning microscopy. Indeed, the amplitudes and half-durations of (ΔF/F)_glAP and (ΔF/F)_sAP were equal in the DKO GCs (36 ± 13 vs 35 ± 14% and 580 ± 380 vs 590 ± 340 ms; n = 28 locations in 11 cells; both p > 0.1; Fig. 5B,C). Accordingly, both the amplitude and τ_1/2 ratios of (ΔF/F)_sAP to (ΔF/F)_glAP were significantly larger than in the WT (p < 0.001), and not significantly different from 1 (0.97 ± 0.08 and 1.04 ± 0.20, p > 0.25, single-sample t test). Thus, NMDAR-dependent extra Ca²⁺ entry evoked by suprathreshold synaptic input is absent in TRPC1/4 DKO mice.

Figure 5.

The extra Ca²⁺ signal is absent from TRPC1/4 DKO GCs. A, Two-photon scan of a representative TRPC1/4 DKO granule cell. B, Analogous to Figure 1C, but data from TRPC1/4 DKO granule cell. Note the similarity of the responses compared with Figure 1C. C, Scatterplots of glAP-ADP versus sAP-ADP τ_1/2 and amplitudes and (ΔF/F)_glAP versus (ΔF/F)_sAP τ_1/2 and amplitudes, respectively. The plots show both TRPC1/4 DKO data (black inverted triangles) and WT data (open symbols, as in Fig. 1) to facilitate comparison (WT, n = 43 dendrites and spines; TRPC1/4 DKO, n = 28 dendrites and spines).

Deletion of TRPC1 and TRPC4 does not reduce excitability

The general deletion of both TRPC1 and TRPC4 could indirectly impede the LLD generation by detrimental effects on the neuronal excitability of MTCs and GCs (Cvetkovic-Lopes et al., 2010). The aforementioned immunohistochemistry with anti-TRPC1 and anti-TRPC4 antibodies yielded no evidence for the presence of any of the two proteins in olfactory receptor neuron axons. Mitral cells expressed both TRPC1 and TRPC4, while tufted cells showed only expression of TRPC1 (data not shown). Thus, we investigated effects of the TRPC1/4 double deletion on synaptic transmission between MTCs and GCs and also GC excitability per se.

As to the shape of the GC Na⁺ AP in WT and DKO animals, both amplitude and width did not show any significant differences (for absolute data and p levels of this section, see Table 1). The frequency of APs in response to a depolarizing current pulse at 50 pA and 800 ms duration did also not differ significantly between the two populations. The frequency was evaluated for cells responding at this current amplitude; the fraction of silent cells was 0.35 in the WT and 0.39 in the DKO. The input resistance of the two groups was similar. Also, the average extracellular stimulation threshold for the generation of synaptic APs was equal, and as stated above, the EPSP τ_1/2 just below AP threshold was the same for both. We also tested for possible effects of the TRPC1/4 DKO on GC morphology (see Materials and Methods) but did not observe any major changes with respect to the total length of the apical dendrite as well as to the number of its branchpoints.

Table 1.

Excitability and morphology of TRPC1/4 DKO granule cells versus WT granule cells

Parameter	WT		TRPC1/4 DKO		p (Mann–Whitney test)
	Mean value ± SD	n	Mean value ± SD	n
Input resistance	1.07 ± 0.39 GΩ	20	0.89 ± 0.44 GΩ	19	0.14
sAP
Na+ spike amplitude	105 ± 10 mV	35	111 ± 16 mV	20	0.08
Na+ spike half-width	1.9 ± 0.4 ms	18	2.1 ± 0.4 ms	19	0.11
Number of APs at 50 pA, 800 ms	8.8 ± 3.7	30	7.7 ± 3.8	14	0.24
Synaptic input
Extracellular stimulation threshold	620 ± 420 μA	35	620 ± 380 μA	23	0.33
Just subthreshold EPSP τ1/2	88 ± 58 ms	39	71 ± 54 ms	21	0.40
Morphology of apical dendrite		25		9
Length	240 ± 150 μm		215 ± 110 μm		0.67
Number of branchpoints	2.6 ± 1.1		2.0 ± 1.7		0.24

While more subtle effects cannot be excluded, these data speak against a major alteration of GC excitability and MTC-to-GC synaptic transmission in the absence of TRPC1/4.

TRPCs promote slow feedback inhibition of mitral/tufted cells

The MTC–GC microcircuit can provide recurrent inhibition of MTCs via dendritic release of GABA from the reciprocal GC spine. This release occurs in part asynchronously (see Introduction). A role for TRPC4 in dendritic release has been previously demonstrated in thalamic GABAergic neurons (Munsch et al., 2003). Because in GCs TRPC activation via MTC input results in slow extra Ca²⁺ entry, we hypothesized that the slow TRPC-mediated plateau current facilitates recurrent dendrodendritic inhibition of MTCs.

To measure recurrent inhibition, we voltage clamped mitral cells and applied brief depolarizing pulses that are known to activate the microcircuit and cause recurrent GABA release (Jahr and Nicoll, 1980; Isaacson and Strowbridge, 1998). To reduce voltage escape, QX-314 was included in the pipette solution, and NMDAR activation was facilitated by omitting Mg²⁺ from the ACSF (Isaacson and Strowbridge, 1998; Schoppa et al., 1998; Chen et al., 2000; Halabisky et al., 2000). Figure 6 illustrates that the decay kinetics of the resulting barrage of IPSCs was significantly faster in the TRPC1/4 DKO than in the WT mice (τ_1/2 DKO, 70 ± 30 ms, n = 8 cells, vs WT, 280 ± 140 ms, n = 7 cells; p < 0.005). The amplitudes across these experiments were highly variable, probably because of the differing connectivity in acute slices, but not significantly different between the two populations. There was no correlation between the amplitudes and the half-durations (WT, r = −0.27; DKO, r = −0.31). However, the integrals of the responses were larger in the WT (143 ± 115 nA ms) than in the DKO (43 ± 27 nA ms; p < 0.025). Thus, we conclude that TRPC-mediated Ca²⁺ entry contributes to dendritic GABA release from the GC spine and fosters in particular late/asynchronous recurrent inhibition.

Figure 6.

TRPC channels promote recurrent dendrodendritic inhibition of mitral cells. A, Scheme of experiment and recordings of reciprocal DDI in a WT mouse (black) and a TRPC1/4 DKO mouse (gray) mitral cell. B, Averaged DDI recordings for all experiments (n = 7 WT and n = 8 DKO cells). Both the averaged τ_1/2 and integral show a significant decrease for the TRPC1/4 DKO versus WT (p < 0.005 and p < 0.025).

Discussion

TRPC activation cascade

We have demonstrated that TRPC1 and TRPC4 are required for a synaptic, I_CAN-mediated plateau current at the MTC–GC synapse. This conclusion is based on the absence of plateau currents at this synapse in the combined deletion of TRPC1 and TRPC4, while individual deletion of either TRPC1 or TRPC4 is accompanied by partial reductions in the plateau current amplitude and duration. We find that synaptic activation of TRPC1/4 is achieved via NMDA receptor-mediated signaling.

Although TRPC channels are regulated by intracellular Ca²⁺/calmodulin (for review, see Zhu, 2005), so far their activation was found to critically depend on G_q-protein-coupled receptors in other examples of neuronal TRPC action: In cerebellar Purkinje cells, an mGluR1-dependent slow EPSP has been shown to be mediated by TRPC3 (Hartmann et al., 2008). Activation of a TRPC channel downstream of synaptic activation of mGluR1 has been suggested also for hippocampal oriens/alveus interneurons (Topolnik et al., 2006) and CA1 pyramidal cells (El-Hassar et al., 2011). In the amygdala, synaptic responses downstream of mGluR1 and the cholecystokinin-2 receptor are significantly reduced in TRPC5^−/− mice (Riccio et al., 2009). In a heterologous expression system, mGluR1 is able to couple to TRPC1 (Kim et al., 2003).

The dependence of TRPC channel opening on the activation of an ionotropic glutamate receptor has not been described before. Current views hold that TRPC channels are activated by phospholipase C (PLC)-coupled metabotropic receptors (Soboloff et al., 2007). Consistent with that, a group I mGluR-specific agonist-evoked inward current was recorded in OB GCs (Heinbockel et al., 2007). In addition, muscarinic acetylcholine receptors at the MTC–GC synapse were found to be coupled to a flufenamic acid-sensitive I_CAN in GCs (Pressler et al., 2007; Ghatpande and Gelperin, 2009), similarly to neurons in the entorhinal cortex (Zhang et al., 2011). However, the lack of an effect of group I mGluR blockade on the GC LLD, the blockade of the LLD by APV, the triggering of the LLD by single stimulation—versus the repetitive stimulation that is often required to activate mGluRs—and the fact that ryanodine receptors but not IP₃ receptors contribute to basal synaptic Ca²⁺ signaling in GC spines (Egger et al., 2005) rule out an essential involvement of metabotropic receptors in the type of synaptic response that we investigate here.

An mGluR-independent PLC activation following glutamate binding to NMDARs was observed in hippocampal neurons (Codazzi et al., 2006; Horne and Dell'Acqua, 2007). This effect is mediated by PLCδ isoforms (Codazzi et al., 2006) that are not controlled by G_q but directly activated by Ca²⁺ (Allen et al., 1997). We have shown previously that NMDARs contribute a significant fraction to postsynaptic calcium signals at the reciprocal MTC–GC synapse (Egger et al., 2005). Ca²⁺ influx through the NMDAR channel is of crucial importance for the LLD in OB granule cells (Hall and Delaney, 2002; Egger, 2008), and PLCδ isoforms are expressed in the GC layer (Lein et al., 2007). It remains to be elucidated whether NMDAR-mediated Δ[Ca²⁺]_i acts onto the TRPCs via PLCδ or whether the signaling pathway involves other steps. For TRPC3 in cerebellar Purkinje cells, a PLC-independent gating mechanism relies on the small GTP-binding protein Rho and subsequent phospholipase D activation (Glitsch, 2010). TRPC channels are known to be redundantly activated and appear to be located at a point of intersection of diverse signaling pathways (for review, see Hartmann et al., 2011). Moreover, TRPC channels in neurons are part of multiprotein postsynaptic signaling complexes that also contain neurotransmitter receptors, including the NMDAR (for review, see Ambudkar et al., 2006). For example, the cholesterol-binding protein caveolin-1 can scaffold NMDARs and Src protein kinases (Head et al., 2008) and attach TRPC subunits to the plasma membrane (Lockwich et al., 2000; Brazer et al., 2003). Src protein kinases, in turn, have been shown to activate both TRPC1 (Gervásio et al., 2008) and TRPC4 (Odell et al., 2005). At the MTC–GC synapse, the absence of synaptic plateau currents in the NR1-KO GCs argues against an activation of TRPCs via ryanodine receptors.

Since the coincidence of NMDAR activation and a Na⁺ spike is required to activate the TRPCs, we suggest that the relevant signal cascade is triggered by the elevation of [Ca²⁺]_i via NMDARs only if this elevation is enhanced by the membrane depolarization provided by the Na⁺ spike, either via additional relief of the Mg²⁺ block or via activation of VDCCs, just as in classical coincidence detection mechanisms that play a role in synaptic plasticity (Blair et al., 2001).

Finally, it has been debated whether TRPC1 on its own can be a pore-forming subunit (for review, see Wu et al., 2010). We still find a plateau current (albeit reduced) in the absence of TRPC4, which is most likely mediated by TRPC1. Thus, apparently TRPC1 homomultimers can conduct currents in olfactory bulb GCs.

Functional roles of TRPC activation

TRPC activation results in a global Ca²⁺ entry in dendrites and spines of the GC that is due to T-type Ca²⁺ channel activation caused by the depolarizing plateau current (Egger, 2008). This prolonged Ca²⁺ entry persists during the plateau current and thus may contribute to delayed dendritic release of GABA. Indeed, we show that recurrent inhibition is faster and involves less charge transfer in TRPC1/4 DKO animals. We propose that TRPC1 and TRPC4 are key players in the regulation of slow dendritic GABA release from GCs, similar to the role of TRPC4 in thalamic interneurons (Munsch et al., 2003). Obviously, however, TRPCs are not the only effectors of slow output from this system, since asynchronous release also occurs in the presence of TTX (Isaacson and Strowbridge, 1998), is not fully blocked in the DKO, and there are also other mechanisms providing prolonged Ca²⁺ entry such as delayed GC spiking (Schoppa and Westbrook, 1999; Kapoor and Urban, 2006) and sluggish extrusion of Ca²⁺ from the GC dendrite (Egger and Stroh, 2009).

How does TRPC1/4 action then shape odor processing in the olfactory bulb? An implicit answer to this question is provided by the study of Abraham et al. (2010), in which a GC-specific deletion of NR1 was achieved in ∼40% of all GCs of the OB (GluN1^ΔGCL). In these cells, the activation of TRPC1/4 is also blocked, as shown above by the lack of the LLD for glomerularly evoked action potentials. Accordingly, recurrent inhibition following trains of APs is also reduced in GluN1^ΔGCL mice in vitro. Behavioral tests showed that in GluN1^ΔGCL mice learning and memory are not affected, although the discrimination of highly similar odorants takes significantly more time than in WT mice. Thus, TRPC action could also be related to fine odor discrimination; it remains to be elucidated what aspect of the above observations is due to other actions of the NMDAR apart from downstream TRPC-mediated effects.

However, since a major part of TRPC action appears to happen on the slow temporal timescale and odor discrimination has been shown to occur rather fast, we would also expect other roles of TRPC to emerge. For example, TRPC action may couple to the prominent slow, respiration-coupled oscillations intrinsic to the olfactory bulb (Macrides and Chorover, 1972), possibly via their interaction with T-type VDCCs (T-channels). We have already proposed that GC T-channels are periodically inactivated by slow olfactory bulb oscillations and thus GC output is likely to be also periodically modulated (Egger et al., 2003). The TRPC channel-mediated long-lasting depolarizations can activate T-channels and thus produce global Ca²⁺ entry (Egger, 2008) and participate in driving slow bulbar oscillations but may also contribute to the periodic inactivation of T-channels in the course of these persistent depolarizations.

In addition to its importance for recurrent inhibition, the TRPC-mediated mechanism of LLD demonstrated here might play a role in lateral inhibition of neighboring MTCs, again by giving rise to global extra Ca²⁺ entry. Lateral inhibition will usually co-occur with TRPC activation because the latter requires a GC Na⁺ spike. The spike will propagate along the GC dendrite and cause substantial Ca²⁺ entry at other reciprocal spines (Egger et al., 2003; Zelles et al., 2006) that is enhanced by the TRPC-mediated extra global Ca²⁺ entry. Lateral inhibition per se does not necessarily require a Na⁺ spike (Isaacson and Strowbridge, 1998) but may also be triggered by a Ca²⁺ spike in GCs (Pinato and Midtgaard, 2003; Egger et al., 2005) that is probably not capable of activating TRPCs. Nevertheless, the efficiency and duration of lateral inhibition following a GC AP are very likely to be enhanced by TRPC1/4 action. Thus, even though Ca²⁺ entry via NMDARs may not couple directly to recurrent GABA release (Isaacson, 2001), NMDAR activation can contribute to recurrent and lateral inhibition via the TRPC-mediated pathway.

Finally, we have shown that, in GCs, Na⁺ spikes evoked by current injection are not equivalent to Na⁺ spikes evoked by synaptic mitral and tufted cell input because only the latter are capable of TRPC activation and produce a different spatiotemporal pattern of postsynaptic Ca²⁺ entry. This finding may contribute to the recent debate on the validity of paradigms used to determine spike-timing-dependent plasticity (Lisman and Spruston, 2010).