Abstract
Canonical transient receptor potential (TRPC) channels influence various neuronal functions. Using quantitative high‐resolution mass spectrometry, we demonstrate that TRPC1, TRPC4, and TRPC5 assemble into heteromultimers with each other, but not with other TRP family members in the mouse brain and hippocampus. In hippocampal neurons from Trpc1/Trpc4/Trpc5‐triple‐knockout (Trpc1/4/5−/−) mice, lacking any TRPC1‐, TRPC4‐, or TRPC5‐containing channels, action potential‐triggered excitatory postsynaptic currents (EPSCs) were significantly reduced, whereas frequency, amplitude, and kinetics of quantal miniature EPSC signaling remained unchanged. Likewise, evoked postsynaptic responses in hippocampal slice recordings and transient potentiation after tetanic stimulation were decreased. In vivo, Trpc1/4/5−/− mice displayed impaired cross‐frequency coupling in hippocampal networks and deficits in spatial working memory, while spatial reference memory was unaltered. Trpc1/4/5−/− animals also exhibited deficiencies in adapting to a new challenge in a relearning task. Our results indicate the contribution of heteromultimeric channels from TRPC1, TRPC4, and TRPC5 subunits to the regulation of mechanisms underlying spatial working memory and flexible relearning by facilitating proper synaptic transmission in hippocampal neurons.
Synopsis
Heteromers from TRPC1, TRPC4 and TRPC5 define hippocampal synaptic transmission and specific mnemonic behavior.
TRPC1, TRPC4, and TRPC5 assemble into heteromultimers with each other, but not with other TRP family members in the mouse brain and hippocampus.
TRPC1/4/5 deficiency reduced action potential evoked responses in vitro and in situ, but LTP and depotentiation remained unchanged.
In Trpc1/4/5 triple knockout mice cross‐frequency phase‐amplitude coupling is impaired while basal neuronal network oscillations are unchanged.
Trpc1/4/5 triple knock out mice display deficits in spatial working memory and when adapting to a new challenge in a re‐learning task, whereas the acquisition of spatial reference memory remains unaltered.
Introduction
The seven canonical transient receptor potential channels (TRPC1 to TRPC7) form non‐selective cation channels that are activated in response to the stimulation of phospholipase C‐coupled receptors, causing Na+ and Ca2+ entry into the cell, thus leading to the depolarization of the membrane potential and the initiation of Ca2+‐dependent intracellular cascades. Based on structural homology, members of the TRPC subfamily can be divided into three subgroups: TRPC1/TRPC4/TRPC5, TRPC3/TRPC6/TRPC7, and TRPC2 (Montell et al, 2002; Wu et al, 2010). In mice, the Trpc1, Trpc4, and Trpc5 genes are expressed together in several subregions of the hippocampus, as demonstrated by in situ hybridization and immunohistochemistry (Strübing et al, 2001; Freichel et al, 2005; Fowler et al, 2007). They co‐localize in the stratum pyramidale of the hippocampal CA1‐CA3 regions and the granule layer of the dentate gyrus; weaker expression is found in the hilus and the ventral subiculum.
In heterologous co‐expression experiments, TRPC1, TRPC4, and TRPC5 were shown to interact with each other, but not with members of the TRPC3/TRPC6/TRPC7 subgroup (Hofmann et al, 2002). Other reports described heterologous interactions of TRPC1 with all members of the TRPC subfamily (Storch et al, 2012) as well as with TRPV4 and TRPP2 (Ma et al, 2011; Du et al, 2014). Moreover, co‐immunoprecipitations from brain membrane fractions suggested the formation of heteromultimers between TRPC1, TRPC4, and TRPC5 (Goel et al, 2002; Hofmann et al, 2002), as well as with TRPC3 and TRPC6 (Strübing et al, 2003) in embryonic brain. However, these observations are obscured by the lack of subtype‐specific antibodies and rigorous negative controls as provided by the respective target‐knockout mice.
While TRPC channels have been implicated in a variety of neuronal functions, including neuronal excitability (Faber et al, 2006; Stroh et al, 2012; Phelan et al, 2013), excitotoxicity (Phelan et al, 2012, 2013), neurogenesis (Li et al, 2012), and neurite outgrowth (Greka et al, 2003; Li et al, 2005; Hui et al, 2006), the role of TRPC1‐, TRPC4‐, and TRPC5‐containing channels in synaptic transmission and neurotransmitter release remains sparse (Hartmann et al, 2008; Riccio et al, 2009; Shen et al, 2013). TRPC1 was identified as a mGluR1‐evoked slow EPSC channel in cerebellar Purkinje cells based on interference with an anti‐TRPC1 antibody (Kim et al, 2003). However, subsequent experiments with several TRPC‐deficient mouse models revealed TRPC3 as a mGluR1‐evoked slow EPSC channel (Hartmann et al, 2008). In infantile Trpc5−/− mice, synaptic strength was diminished at inputs to the amygdala, but it was not significantly altered in older Trpc5−/− mice (Riccio et al, 2009). In the olfactory bulb, measurements of inhibitory postsynaptic currents (IPSCs) in mitral/tufted cells demonstrated a reduction in GABA release from granule cells in Trpc1/Trpc4−/−‐double‐knockout mice (Stroh et al, 2012). On the organismal level, Trpc4−/− and Trpc5−/− mice exhibit a reduced anxiety‐like phenotype, when presented with stimuli triggering innate fear responses (Riccio et al, 2009, 2014). Both TRPC4 and TRPC5 are causally involved in epileptogenesis, as well as in associated seizures and early death in mice (Phelan et al, 2012, 2013). Additionally, in a mouse model for Huntington's disease (HD), an involvement of TRPC5 activation via S‐glutathionylation of Cys176/178 was reported in the pathogenesis of HD (Hong et al, 2015).
The hippocampus is postulated as storage region for the spatial map, and the formation of the spatial map was shown to depend on the NMDA receptor‐induced plasticity of hippocampal CA3 to CA1 synapses (Morris et al, 1986a,b; Bannerman et al, 1995, 2012; Tsien et al, 1996a,b). This hypothesis was recently experimentally challenged by the finding of Bannerman et al that the NMDA receptors at dorsal CA3 to CA1 synapses are not necessary for the formation of spatial maps but for decision making (Bannerman et al, 2012).
The function of the TRPC1/TRPC4/TRPC5 subfamily for memory formation and synaptic plasticity in the hippocampus has been largely unknown. Here, we demonstrate the specific formation of heteromultimers from TRPC1, TRPC4, and TRPC5 in the mouse brain and hippocampus, using knockout‐controlled affinity purifications together with quantitative high‐resolution mass spectrometry analysis. The combined deletion of Trpc1, Trpc4, and Trpc5 in mice uncovered a role for TRPC1/4/5 channels in neuronal communication and hippocampal function. In the absence of TRPC1/4/5, the action potential‐evoked EPSC amplitude was markedly reduced in autaptic hippocampal neurons, while synapse density and miniature EPSC (mEPSC) frequency in neuronal mass cultures remained unchanged. In the same line, synaptic transmission was also reduced in hippocampal slice recordings.
These cellular peculiarities were accompanied by impaired cross‐frequency phase–amplitude coupling of theta and gamma oscillations in hippocampal networks, as well as by severe deficits in both spatial working memory (SWM) and flexible spatial relearning as demonstrated in independent hippocampus‐specific tasks, while spatial reference memory (SRM) was not affected. These behavioral deficits were not associated with changes in brain morphology.
Overall, our results suggest that heteromultimeric channels composed of TRPC1, TRPC4, and TRPC5 subunits contribute to the regulation of the mechanisms underlying working memory and flexible relearning by facilitating proper synaptic transmission in hippocampal neurons.
Results
TRPC1, TRPC4, and TRPC5 proteins co‐assemble into heterotetrameric channels in the mouse brain and hippocampus
To test the assembly of TRPC1, TRPC4, and TRPC5, we combined quantitative high‐resolution mass spectrometry with affinity purifications (APs) using isoform‐specific antibodies (see Materials and Methods) on membrane fractions prepared from wild‐type and target‐knockout brains. Evaluation of these APs demonstrated the robust co‐assembly of all three TRPC isoforms in both whole brain and hippocampus (Figs 1 and EV1). Thus, TRPC4 and TRPC5 were effectively co‐purified in APs with an anti‐TRPC1 antibody from wild‐type membranes, but not from TRPC1‐knockout material (left part, Fig 1). Similarly, the anti‐TRPC4 antibody specifically and abundantly co‐purified TRPC1 and TRPC5, and the anti‐TRPC5 antibody co‐purified TRPC1 and TRPC4 proteins (middle and right parts, Fig 1). In contrast, mass spectrometry did not detect other members of the TRPC family, nor any other TRP protein in any of the AP eluates analyzed. These results indicate the formation of heteromultimeric channels from TRPC1, TRPC4, and TRPC5 subunits, although the exact stoichiometry of such assemblies cannot be derived.
Abundance ratios (see Materials and Methods) determined for TRPC1, TRPC4, and TRPC5 in affinity purifications with antibodies specifically targeting TRPC1 (anti‐C1), TRPC4 (anti‐C4), and TRPC5 (anti‐C5) proteins, in membrane fractions prepared from brains of wild‐type control, Trpc1−/−, Trpc4−/−, Trpc5−/−, or Trpc1/4/5−/− animals (Trpc1−/−, Trpc4−/−, or Trpc5−/− labeled as C1−/−, C4−/−, or C5−/−, and Trpc1/4/5−/− labeled as C1/4/5−/−). Asterisks denote lack of protein‐specific peptides in the respective affinity purifications. Inset depicts possible subunit assemblies for the respective affinity purifications.
Abundance ratios (see Materials and Methods) determined for TRPC1, TRPC4, and TRPC5 in affinity purifications with antibodies specifically targeting TRPC1 (anti‐C1), TRPC4 (anti‐C4), and TRPC5 (anti‐C5) proteins in membrane fractions prepared from wild‐type control mouse hippocampi.
TRPC1/4/5 channels regulate action potential‐evoked transmitter release
The physiological role of TRPC1/4/5 channels in synaptic transmission was analyzed in autaptic hippocampal neuron cultures. Short stimulation of these neurons in the voltage‐clamp mode evoked single action potentials that triggered a robust synaptic response mediated by AMPA receptors (Guzman et al, 2010). Compared to control cells, TRPC1/4/5‐triple‐deficient neurons displayed a significant reduction in amplitude and charge of the EPSCs (Fig 2A and B). In contrast, TRPC1/4/5 deficiency affected neither mEPSC frequency, amplitude, nor the time course of quantal signaling (Fig 2C and D), indicating that the properties of postsynaptic signal generation and transmitter sensing are unchanged in the absence of TRPC1, TRPC4, and TRPC5 proteins. To investigate whether deficits in synaptic transmission are accompanied by alterations in synapse density, mass cultures of hippocampal neurons were stained with antibodies against the vesicular marker protein synaptophysin. No changes in the number or area of synapses were observed (Fig 2E and F). Taken together, TRPC1/4/5 channels in hippocampal neurons facilitate evoked transmitter release potentially by altering neuronal excitability or presynaptic Ca2+ dynamics.
A. Exemplary recordings of evoked EPSCs from autaptic hippocampal neurons.
B. Summary plots for EPSC parameters. The loss of TRPC1, TRPC4, and TRPC5 reduces the amplitude (**P = 0.0058) and charge of EPSCs (*P = 0.032) (n = 63 for Trpc1/4/5−/−, n = 66 for controls). Statistical significance was determined using two‐tailed unpaired Student's t‐test.
C, D (C) Exemplary recordings of mEPSCs from neurons in mass culture. The cumulative frequency distribution of mEPSC amplitude and charge, as well as the quantitative analyses of both frequency and amplitude (D), shows that TRPC1/4/5 deficiency does not alter the properties of quantal signaling (n = 14 for Trpc1/4/5−/−, n = 20 for controls).
E. Representative epifluorescence images of neurons immunolabeled with synaptophysin. Scale bar (inset), 5 μm.
F. The loss of TRPC channels does not alter the density of synapses determined per 50 μm length of neuronal processes or their respective size (n = 17 for Trpc1/4/5−/−, n = 15 for controls).
Deletion of the Trpc1, Trpc4, and Trpc5 genes does not cause morphological changes in the brain
To test whether the deletion of Trpc1, Trpc4, and Trpc5 affects the cellular integrity of the hippocampus, we compared the hippocampal structures by immunohistological and histochemical stainings of brain slices from adult Trpc1/4/5−/− and control mice. Immunostainings using anti‐GluA1 antibodies (Fig 3A) showed the typical expression pattern of the α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazolepropionic (AMPA) receptor subunit GluA1 (Zamanillo et al, 1999; Jensen et al, 2003). Similar to control mice, strong GluA1 immunostaining was detected in the stratum radiatum, the stratum oriens, and the molecular layer of the dentate gyrus (DG) in the hippocampus of Trpc1/4/5−/− animals. In both control and Trpc1/4/5−/− mice, the GluA1 expression was highest in the CA1 and lowest in the stratum pyramidale (Fig 3A), suggesting a regular dendritic enrichment of AMPA receptors in both CA1, CA2, CA3 pyramidal and DG granule cells. Anti‐GFAP stainings revealed that the manually determined number and the distribution of GFAP‐positive astrocytes in the hippocampal slices were comparable between control and Trpc1/4/5−/− mice (Fig 3B). Similarly, the number and distribution of somatostatin‐positive interneurons, both in the stratum oriens and in the hilus region of the DG, were unchanged (Fig 3C). The histological analysis by Nissl staining of horizontal brain sections showed no obvious differences in the thickness of the CA1, CA3, and the outer DG granule cell layers between the dorsal hippocampus of control and Trpc1/4/5−/− mice, respectively (Fig 3D). In conclusion, the loss of TRPC1, TRPC4, and TRPC5 was not associated with any major alterations in the brain morphology or the thickness of the cortical layer as evaluated by anti‐NeuN staining of coronal sections (Fig 3E).
A–C Immunohistology of coronal brain sections illustrates comparable patterns of GluA1 (A), GFAP (B), and somatostatin (C) expression in Trpc1/4/5−/− and control mice. (B) The number of glial cells (mean ± SEM) is as follows (Trpc1/4/5−/− versus controls): in CA1: subgranular zone: 69.00 ± 6.66 versus 70.67 ± 6.23 (P = 0.8638), Str. moleculare: 110.00 ± 15.63 versus 89.00 ± 6.08 (P = 0.2788), Str. lacunosum: 100.33 ± 9.91 versus 109.67 ± 6.12 (P = 0.4677), Str. radiatum: 121.00 ± 16.29 versus 118.00 ± 12.66 (P = 0.8914), Str. pyramidale: 52.67 ± 3.18 versus 53.00 ± 5.51 (P = 0.9607), Str. oriens: 90.67 ± 2.19 versus 88.00 ± 5.57 (P = 0.6788); in the cortex sector (1.2 × 1.7 mm): 109.00 ± 17.78 versus 128.33 ± 4.26 (P = 0.3498); the right cerebral hemispheres of three mice per genotype were analyzed. (C) The number of somatostatin‐positive interneurons (mean ± SEM) is (Trpc1/4/5−/− versus controls): CA1: 100.50 ± 9.28 versus 78.33 ± 15.21 (P = 0.2419), CA2: 14.33 ± 0.99 versus 15.67 ± 1.71 (P = 0.5143), CA3: 65.17 ± 8.31 versus 52.17 ± 6.94 (P = 0.2577), hilus: 37.00 ± 3.16 versus 34.50 ± 2.63 (P = 0.5569); both cerebral hemispheres of three mice per genotype were analyzed.
D. Nissl staining of horizontal sections does not reveal any obvious changes in neuronal patterns of mutant and control mice. The thickness of the somata containing cell layer (μm; mean ± SEM) is (Trpc1/4/5−/− versus controls): CA1: 29.33 ± 1.63 versus 32.14 ± 2.41 (P = 0.3449), CA3: 49.58 ± 4.99 versus 48.83 ± 3.70 (P = 0.9039), outer DG: 23.79 ± 0.85 versus 21.75 ± 1.21 (P = 0.1819); in two independent tissue slices, both cerebral hemispheres were analyzed in three mice per genotype.
E. The thickness of the cortex layer in anti‐NeuN immunostainings was comparable in control and Trpc1/4/5−/− mice (μm; mean ± SEM): 843.95 ± 17.05 versus 859.20 ± 42.87 (P = 0.7313); both cerebral hemispheres of three mice per genotype were analyzed.
Unchanged basal neuronal network oscillations with impaired cross‐frequency phase–amplitude coupling in Trpc1/4/5−/− mice
Next, we checked whether electrical activity in hippocampal networks of Trpc1/4/5−/− mice was impaired. Freely moving animals were recorded in 5‐h sessions according to the experimental setup depicted in Fig 4A. The frequency distributions displayed typical activity‐dependent features as previously described (Tort et al, 2008; Scheffzuk et al, 2013). In summary, frequency peaks for theta and gamma oscillations during REM sleep were not altered (Fig 4B). Frequency peaks and power for both theta and gamma oscillations during REM sleep were unchanged (Fig 4B, C and F). We further analyzed how gamma amplitude was modulated by the theta phase. General appearance of cross‐frequency couplings was similar to previous findings (Scheffer‐Teixeira et al, 2012) with a modulation of low gamma (50–80 Hz) during REM sleep (Fig 4D). Indeed, gamma oscillations in TRPC1/4/5‐deficient animals were broadly distributed along theta‐phase cycles (Fig 4E), whereas control animals showed the typical “waning” and “waxing” features as described in earlier studies (Chrobak & Buzsaki, 1998). This suggests a desynchronization between gamma oscillations and theta phase. Consistently, the modulation index of cross‐frequency phase–amplitude coupling for low gamma was significantly reduced in Trpc1/4/5−/− animals, compared to the controls (Fig 4G). Taken together, the deletion of TRPC1/4/5 impairs the interaction between low‐ and high‐frequency hippocampal network oscillations.
Scheme of experimental recording setup where video inputs (VIDEO) are synchronized (SYNC) to recorded local field potentials (LFPs) to the computer (PC).
Theta and gamma frequencies are not different between groups. Curves shown as median and 25th and 75th percentiles (n = 5 for Trpc1/4/5−/−, n = 5 for controls).
Peak frequencies for theta and gamma oscillations are not significantly different from each other for both groups (n = 5 for Trpc1/4/5−/−, n = 5 for controls).
Cross‐frequency coupling during REM sleep shows reduced modulation of theta and low gamma for Trpc1/4/5−/− mice compared to controls (white arrows).
Raw LFPs and filtered LFPs for theta and gamma show diffuse distribution of gamma on theta oscillations for Trpc1/4/5−/− mice (red shades) when compared to controls, where gamma is superimposed mainly on the peak of theta cycles (gray shades) with the typical waning/waxing appearance.
Theta and low gamma power is not significantly different between control (n = 5) and Trpc1/4/5−/− mice (n = 5).
Modulation index (MI) for low gamma during REM sleep is reduced in Trpc1/4/5−/− (n = 5) when compared to controls (n = 5; *P = 0.0179). CTR, control; TKO, Trpc1/4/5−/−.
Synaptic transmission and firing output are reduced in hippocampal area CA1 of Trpc1/4/5−/− mice without changing synaptic long‐term potentiation (LTP) or depotentiation
In acute hippocampal slices of adult animals, we analyzed the plasticity of CA3‐to‐CA1 synapses. Upon stimulation of Schaffer collateral CA3 axons (“1” in Fig 5A), comparable axonal spiking of CA3 neurons was obtained (Fig 5B), both in control and in Trpc1/4/5−/− mice. Postsynaptic currents, measured as local field potentials (LFPs) (Fig 5C), in stratum radiatum (“2” in Fig 5A) as well as the postsynaptic firing of CA1 cells, measured in stratum pyramidale (“3” in Fig 5A) as population spikes (Fig 5D), were reduced in slices from Trpc1/4/5−/− mice. Hence, in order to assure comparable baseline LFPs for plasticity experiments below (Fig 5I–K), baseline stimulation intensity was adjusted to higher levels in TRPC1/4/5‐deficient slices (Fig 5E). Equal LFPs elicited comparable firing of the postsynaptic CA1 cells (Fig 5F and G). A left shift (“E‐S‐potentiation”) at the second pulse of a 50‐ms paired pulse was observed in both control (Fig 5F) and Trpc1/4/5−/− slices (Fig 5G), indicating no prominent inhibition on the second pulse under our experimental conditions. When activating the same number of presynaptic fibers (compare Fig 5B), LFP paired‐pulse ratios were increased in Trpc1/4/5−/− mice (Fig 5H, main), pointing to altered short‐term facilitation. Yet, LFP paired‐pulse ratios versus the respective first LFP slopes of the paired pulses (Fig 5H, inset) were found to be similar for Trpc1/4/5−/− mice and controls, suggesting an unchanged synaptic release probability in Trpc1/4/5−/− mice. The transient potentiation after 100‐Hz stimulation was impaired in Trpc1/4/5−/− acute hippocampal slices (Fig 5I), further suggesting altered short‐term plasticity in Trpc1/4/5−/− animals. Since memory function, among others, relies on synaptic plasticity, we studied different aspects of long‐term plasticity similar to Nicholls et al (2008) including a modified NMDAR‐dependent (Fig 5K, arrow 2) and NMDAR‐independent (arrow 3) depotentiation protocol (Kemp et al, 2000). With these protocols, we did not reveal any impairment neither in long‐term potentiation (Fig 5J) nor in successful depotentiation (Fig 5K) in the absence of the three TRPC channel subunits TRPC1, TRPC4, and TRPC5.
(Left) Stimulation of Schaffer collaterals (CA3 axons). Recording electrodes (1) in stratum radiatum (2) and stratum pyramidale (3). (Right) Example traces of recording in stratum radiatum (top) and at the same time in stratum pyramidale (bottom).
Fiber volley in stratum radiatum as a measure for presynaptic axonal spiking versus stimulation intensity (n = 21 (11 mice) for Trpc1/4/5−/−, n = 15 (9 mice) for controls; P = 0.3427, two‐way ANCOVA of type II).
Input–output curve of LFPs versus stimulation intensity. No significant differences between control and Trpc1/4/5−/− outside dashed line (n = 21 for Trpc1/4/5−/−, n = 15 for controls; P < 0.0001, two‐way ANCOVA of type II).
Input–output curve of population spikes versus stimulation intensity (n = 21 for Trpc1/4/5−/−, n = 15 for controls; P = 0.0006, two‐way ANCOVA of type II). No significant differences between control and Trpc1/4/5−/− outside dashed line.
Baseline stimulation intensity used in LTP/depotentiation experiments in control and Trpc1/4/5−/− slices (P = 0.0118, t‐test). Inset, baseline LFPs (n = 21 for Trpc1/4/5−/−, n = 15 for controls; P = 0.9906, t‐test).
“E‐S curve” of LFP (“E”) versus population spike (“S”) of control slices at the first (black) and at the second (blue) pulse of a 50‐ms paired pulse (n = 13 (9 mice); P = 0.0011, two‐way ANCOVA of type II). No significant differences between first and second pulse outside dashed line.
“E‐S curve” of Trpc1/4/5−/− at the first (red) and at the second (purple) pulse of a 50‐ms paired pulse (n = 18 (10 mice); P < 0.0001, two‐way ANCOVA of type II). No significant differences between first and second pulse outside dashed line. Of note, no difference in the first pulse between control (from F) and Trpc1/4/5−/− E‐S curve (P = 0.329, two‐way ANCOVA of type II).
LFP paired‐pulse ratio (50‐ms interstimulus interval) versus stimulation intensity (n = 18 (10 mice) for Trpc1/4/5−/−, n = 13 (9 mice) for controls; P = 0.0007, two‐way ANCOVA of type II) and versus the corresponding first pulse (inset, P = 0.0724, two‐way ANCOVA of type II).
Baseline‐normalized LFP versus time, before, and after 100‐Hz stimulation (arrow 1) (n = 21 (10 mice) for Trpc1/4/5−/−, n = 15 (10 mice) for controls). The first minute (**P = 0.0066, t‐test on absolute LFPs) and the second minute (*P = 0.0417, t‐test) after 100‐Hz stimulation, control and Trpc1/4/5−/− are differently potentiated.
Baseline‐normalized LFP versus time in one subset of (I), before and after 100‐Hz stimulation (arrow 1). Both control (P = 0.0136, t‐test on absolute LFPs) and Trpc1/4/5−/− (P = 0.0033, t‐test on absolute LFPs) exhibit LTP (n = 10 (5 mice) for Trpc1/4/5−/−, n = 8 (6 mice) for controls) with no difference in magnitude (relative LFP) between control and Trpc1/4/5−/− (P = 0.1457, Wilcoxon signed‐rank test).
In the second subset of (I) baseline‐normalized LFP versus time, at baseline, 5 min after 100‐Hz stimulation (arrow 1), 15 min after 1‐Hz paired‐pulse (200 ms) stimulation (arrow 2; Pcontrols = 0.0378, PTKO = 0.0316), and 20 min after 1‐Hz paired‐pulse (50‐ms) stimulation (arrow 3; Pcontrols = 0.2429, PTKO = 0.2888) (n = 10 (5 mice) for Trpc1/4/5−/−, n = 7 (4 mice) for controls). CTR, control; TKO, Trpc1/4/5−/−. Determination of the P‐values is described in Materials and Methods.
Deletion of the Trpc1, Trpc4, and Trpc5 genes impairs spatial working memory and relearning competence
Changes in synaptic transmission are frequently associated with differences in hippocampus‐dependent memory formation or consolidation (Tsien et al, 1996b; Fuchs et al, 2007; Du et al, 2008; Brigman et al, 2010). For characterization of the potential alterations in general behavioral patterns of Trpc1/4/5−/− mice, we tested elementary behavioral skills using a SHIRPA protocol (Filali & Lalonde, 2016; Zhang et al, 2016). No differences in spontaneous behavior and activity, reflexes, visual, or hearing skills were observed. The analysis of a rotarod test revealed no alterations in motor skills. Taken together, these results indicate that there are no major deficits that could impact the animals’ performance in the subsequent learning and memory tasks.
Hippocampus‐dependent behavior was analyzed using well‐established paradigms of the T‐maze, Morris water maze, and radial maze. In the T‐maze test, mice usually prefer to seek a food pellet in a novel arm and therefore need to recall the previously visited test arm. Hence, working memory is assessed in this paradigm (Wenk, 2001; Jang et al, 2013). The time course of error counts, and more clearly the slopes of their log best fits, illustrates that the control mice learned to alternate their choice of visited arms as the T‐maze test progressed. Already from the fifth training day on, they reached an error rate of merely 20%. In contrast, Trpc1/4/5−/− animals consistently performed hardly below the random chance level, indicating impairment in spontaneous alternation and thus in spatial working memory (SWM) (Fig 6A). A comparison of the overall change in performances over time between the two groups confirms the impaired performance of mutant mice observed on individual test days.
Number of errors per day in the delayed non‐match‐to‐place T‐maze task. Trpc1/4/5−/− mice (n = 30) perform consistently worse than the controls (n = 30; days 2–10 P < 0.0067, Mann–Whitney U‐test). The maximum number of errors per day is 6. Analysis of the groups’ individual performance curves confirms that the controls learn to alternate their choices of visited arms (###P < 0.001, Poisson regression with GEE) as the test progresses—unlike the Trpc1/4/5−/− animals. The learning curves of the two groups differ significantly (**P = 0.0015, Poisson regression with GEE).
Number of errors per day in the radial maze test. During the early test phase, the animals lacking TRPC1, TRPC4, and TRPC5 exhibit a higher rate of mistakes (n = 15 for Trpc1/4/5−/−, n = 14 for controls; days 2, 3, 5, 10 *P < 0.05, day 6 **P < 0.01).
To corroborate deficits in SWM for the triple‐deficient animals, we performed a radial maze test, where re‐entries into previously visited (empty) arms are regarded as SWM errors (Schmitt et al, 2005; Bannerman et al, 2008; Penley et al, 2013). Also in this experiment, the number of errors was significantly increased in Trpc1/4/5−/− mice on the majority of days during the early test phase (Fig 6B), emphasizing impaired SWM in TRPC1/4/5‐deficient mice compared to controls.
Spatial reference memory (SRM) was assessed using a standard protocol of the Morris water maze (Fig 7A), in which mice were trained to locate a hidden platform by using distant visual cues (Malleret et al, 1999). Latency and cumulative distance to reach the platform were measured. During the training phase, both parameters declined similarly in control and in Trpc1/4/5−/− mice (Fig 7B and C). Furthermore, after removing the platform on a subsequent probe trial (day 12), both genotypes showed similar preference for the training quadrant during their search, suggesting that all mice retained the location of the platform equally well (Fig 7D). Taken together, these findings indicate unchanged SRM in Trpc1/4/5−/− animals.
A. Schematic representation of the experimental procedure.
B, C The delay (B) and cumulative distance (C) to reach the platform during the acquisition phase are shown. The search performance of all animals is comparable (only day 2 *P < 0.02).
D. Relative time spent in the four quadrants of the swimming pool without hidden platform during the probe trial. Mice of both genotypes spend significantly more time (***P < 0.001) in the training quadrant compared to the other three quadrants. SE quadrant: *P < 0.03.
In the T‐maze task (Fig 6A), the position of the first reward changes between trials, and hence, an adequate performance in this paradigm, requires behavioral flexibility (Nicholls et al, 2008; Kim et al, 2011). To investigate this behavioral quality in more detail, we applied a modified paradigm of the Morris water maze (Garthe et al, 2009) in a naive cohort of Trpc1/4/5−/− animals. To this end, a shortened training phase (days 1–3) with variable starting positions on each day was followed by a reversal part, in which the platform was relocated from the upper left to the lower right quadrant on day 4 to address relearning of a new goal position (Fig 8A). Trpc1/4/5−/− mice responded with a significant initial delay and an increased cumulative distance to the platform on the first training day, but learned on days 2 and 3 to find the hidden platform with similar efficiency as the controls. Strikingly, in the subsequent reversal part (tested on days 4 and 5), when the platform was moved to the opposite pool quadrant (Fig 8B and C), Trpc1/4/5−/− mice showed again significant deficits in platform localization, indicating impaired flexible spatial relearning of a new goal position.
A. Schematic representation of the experimental procedure.
B, C Mutant mice initially display increased delay (B) and cumulative distance (day 1 *P = 0.0129) (C) to the platform (day 1 **P = 0.0073), but comparable performances on the following two days show that both groups successfully learn to navigate to the hidden platform. TRPC1/4/5‐deficient mice perform worse when they are challenged to relearn a new goal position in the reversal part (days 4, 5 P < 0.05) (n = 30 for Trpc1/4/5−/−, n = 30 for controls). Results are shown as mean ± SEM. Statistical significance was determined using two‐tailed unpaired Student's t‐test; ***P < 0.001, **P < 0.01, *P < 0.05.
To investigate whether the impaired performance of Trpc1/4/5−/− mice during relearning can be ascribed more tightly to hippocampus‐dependent learning parameters, a qualitative analysis of defined search patterns was performed (Tsien et al, 1996b; Garthe et al, 2009). To unravel differences in navigational strategies between both genotypes, we defined categories of different swim patterns and applied a classification algorithm that allowed the automated classification in undirected search patterns including thigmotaxis, random search, scanning, and chaining on the one hand, and spatially more precise and thus hippocampus‐dependent strategies on the other hand. Such directed, goal‐oriented search strategies were categorized as “allocentric” search patterns and included directed search, focal search, and straight swimming, essentially as described (Garthe et al, 2009). To be able to apply such search strategies, mice have to form a coherent map of their environment, using the relations of the distant visual cues around the pool.
Allocentric strategies significantly increased as training progressed in controls already from days 1 to 2 and in Trpc1/4/5−/− from days 1 to 3. However, in contrast to the controls, Trpc1/4/5−/− mice did not show improvement in their allocentric performance from days 4 to 5 (Fig 9A and B). Furthermore, mutant mice exhibited a substantially higher proportion of undirected, particularly thigmotactic, search behavior (dark green) throughout the entire test (Fig 9C). In addition, allocentric strategies (orange) were less frequently used by Trpc1/4/5−/− animals, suggesting that they navigate with lower spatial precision. These differences in search behavior became very prominent after relocating the platform during the reversal part of the test. Here, a random search pattern (blue) was used much more frequently by Trpc1/4/5−/− mice. The daily differences in delay to find the hidden platform were correlated with the deficits in the development of new efficient allocentric search strategies of TRPC1/4/5‐deficient mice, indicating that Trpc1/4/5−/− animals stick to previously learned allocentric search strategies and are therefore impaired in relearning and behavioral flexibility (Fig 9D).
A, B Qualitative analysis of search strategies used by controls and Trpc1/4/5−/− mice. Mice of both genotypes show a progression in their search to allocentric (orange) strategies during the training phase (control: day 1 versus 2 *P = 0.02, day 1 versus 3 **P = 0.004; Trpc1/4/5−/− *P = 0.01), but only the control animals change and improve their allocentric search behavior after relocation of the platform in the reversal part. Trpc1/4/5−/− had difficulties to change and adapt new allocentric search behavior.
C. The proportion of individual search strategies of mutant mice was normalized to those of the controls. In Trpc1/4/5−/− mice, the proportion of undirected swimming, especially thigmotaxis (dark green), is increased (days 1–5 ***P < 0.001), and allocentric strategies (orange) are less frequently used (day 3 *P = 0.03, day 5 ***P < 0.001). Notably, they also exhibit more often a random swim pattern (blue) during the reversal phase of the test (*P = 0.04).
D. Mean difference between the groups in delay to the hidden platform correlates with deficits in efficient search modes (n = 30 for Trpc1/4/5−/−, n = 30 for controls). Results are shown as mean ± SEM.
Discussion
In the present study, we demonstrate that TRPC1, TRPC4, and TRPC5 form heteromeric complexes in the brain and particularly in the hippocampus. Using Trpc1/4/5−/− as controls, we show the relevance of TRPC1/4/5 channels for synaptic transmission in cultured hippocampal neurons and hippocampal slices for working memory formation and in relearning tasks. Heteromultimer formation between TRPC1, TRPC4, and TRPC5 was established by co‐immunoprecipitation of all three proteins with antibodies directed against each of them and the crucial isoform specificity of the employed antibodies confirmed by the use of the respective target‐knockout controls. In previous heterologous co‐expression experiments, TRPC1 proteins were shown to interact with all members of the TRPC subfamily, including TRPC3/6/7 subgroup (Storch et al, 2012) as well as with TRPV4 and TRPP2 (Ma et al, 2011; Du et al, 2014). In our experiments using MS‐based protein detection, we neither found members of the TRPC3/6/7 subfamily, TRPV4 nor TRPP2 in affinity purifications of total brain or hippocampi with anti‐TRPC1, anti‐TRPC4, and anti‐TRPC5 antibodies. Our results indicate the formation of heteromultimeric channels from TRPC1, TRPC4, and TRPC5 subunits, although the exact stoichiometry of such assemblies cannot be derived. Thus, TRPC1/4/5 channels may form ternary assemblies (TRPC1‐C4‐C5), as well as heteromers from two distinct subunits (TRPC1‐TRPC4, TRPC1‐TRPC5, TRPC4‐TRPC5). The formation of TRPC homomeric channels cannot be excluded by the results of the MS analysis either.
Functional analyses of synaptic transmission, using autaptic hippocampal neurons, revealed that the loss of TRPC1/4/5 channels significantly impaired action potential‐evoked transmitter release. Furthermore, in neuronal mass cultures, the amplitude, charge, and kinetics of mEPSCs did not differ between Trpc1/4/5−/− neurons and controls, suggesting that TRPC1/4/5 channels are more supportive for action potential‐dependent glutamate release. Moreover, synapse density was found to be unaltered in neuronal cultures of Trpc1/4/5−/−. Overall, these results suggest that TRPC1/4/5 deficiency does not interfere with the postsynaptic glutamate response, but rather regulates excitability or Ca2+ homeostasis of presynaptic terminals.
Previous results have shown that TRPC 5 channels can be regulated/activated by intracellular Ca2+ elevations (Blair et al, 2009) at the millisecond time scale (Gross et al, 2009). Thus, an attractive scenario could be that synaptic Ca2+ influx via voltage‐dependent Ca2+ channels is further enhanced by Ca2+‐dependent activation of synaptic TRPC channels, thereby leading to an overall increase in intracellular Ca2+, which can powerfully regulate the pool size of release‐ready vesicles (Stevens & Wesseling, 1998). Such a mechanism could also influence synaptic short‐term plasticity (Dittman & Regehr, 1998; Wang & Kaczmarek, 1998; Lipstein et al, 2013).
In the same line, the results in acute hippocampal Trpc1/4/5−/− slices indicated impaired synaptic short‐term plasticity (Fig 5H and I) that could contribute to the reduced efficacy of CA3‐to‐CA1 synapses. Both LFPs in the stratum radiatum and the postsynaptic firing of CA1 cells in the stratum pyramidale were reduced in Trpc1/4/5−/− preparations, compared to wild‐type controls. These results point to an impaired postsynaptic firing of the CA1 neurons, due to decreased input by CA3 neurons. Yet, potential alterations, for example, in the number of active synapses cannot be rigorously excluded (Kerchner & Nicoll, 2008). Notably, the similar impact of TRPC1/4/5 deficiency on the evoked response in slice (Fig 5C) and culture experiments (Fig 2A and B) suggests that the deletion of Trpc1, Trpc4, and Trpc5 affects glutamatergic transmission directly, rather than being mediated indirectly by altered GABAergic signaling in acute slices.
Similar findings on excitatory synaptic transmission were described in Trpc5−/− mice in neurons of the lateral amygdala of infantile (P13) mice, where EPSCs were reduced, the magnitude of paired‐pulse facilitation was increased, and the amplitude of mEPSCs was unaltered (Riccio et al, 2009). However, synaptic strength analyzed from input–output curves for AMPA receptor‐mediated EPSCs was unaltered at cortico‐amygdala synapses and thalamo‐amygdala synapses both in adolescent Trpc5−/− (Riccio et al, 2009) and in Trpc4−/− mice (Riccio et al, 2014). In contrast, cortico‐amygdala and thalamo‐amygdala EPSCs, mediated by group I mGluRs, were significantly diminished in slices from TRPC5 (Riccio et al, 2009) and in TRPC4‐deficient animals (Riccio et al, 2014). As we show in this study, long‐term potentiation (LTP) and subsequent depotentiation experiments in acute hippocampal slices did not show any significant differences in Trpc1/4/5−/− mice, supporting the regular postsynaptic function in the absence of TRPC1/4/5. In TRPC5‐deficient mice, LTP was also not affected at cortico‐amygdala synapses (Riccio et al, 2009), but was reduced at Schaffer collaterals, whereas Trpc1−/− and Trpc1/Trpc4−/− mice showed no significant impairments (Phelan et al, 2013). The reasons for these discrepant results remain unknown, but might be due to differences in Trpc5 gene targeting strategies, genetic background of the mice, or experimental setups and design.
A major impairment of neuronal network activity in Trpc1/4/5−/− mice can be excluded by our study. The regular expression patterns of the AMPA receptor subunit GluA1 and the interneuronal key marker protein somatostatin suggest a normal neuronal connectivity in Trpc1/4/5−/− mice. Massive neuronal degradation can be ruled out by Nissl staining, as well as by NeuN and GFAP immunostaining. However, important structural changes might be discovered when stressing Trpc1/4/5−/− animals, subjecting them to disease models, or by more advanced morphologic analyses. For instance, impaired synaptic transmission might also be brought about by a reduction in morphological plasticity. The inactivation of TRPC4 was reported to result in an increase in neurite outgrowth and dendrite branching of hippocampal neurons (Jeon et al, 2013). Yet, similar results were obtained by the expression of a dominant‐negative variant of TRPC5 (Greka et al, 2003), which renders the possibility of morphological alterations, underlying the observed changes in synaptic transmission unlikely, despite the fact that another study suggested that localized Ca2+ influx via TRPC5 channels promotes axon formation via activation of Ca2+/calmodulin kinase kinase (CaMKK) and CaMKIγ (Davare et al, 2009). The integrity of neuronal network function in TRPC1/4/5‐deficient mice finds further support in unchanged basal parameters of typical network patterns as concluded from unaltered theta and gamma oscillations. However, cross‐frequency phase–amplitude coupling (CFC) was impaired in Trpc1/4/5−/− animals. Coordination between slow and fast network oscillations has been suggested to underlie complex mnemonic processes, including working memory tasks (Wulff et al, 2009; Korotkova et al, 2010). The observed impairment of CFC in Trpc1/4/5−/− mice might therefore causally contribute to the mnemonic deficits discussed below.
TRPC1/4/5‐deficient animals are housed as an inbred mouse line. They breed and the number of offspring is normal, without showing any signs of early death. Additionally, the behavioral SHIRPA analysis of Trpc1/4/5−/− mice did not reveal any impairment in spontaneous activity, body position, and tremor, vision, and hearing. The rotarod test showed that the genetic deletion of Trpc1, Trpc4, and Trpc5 did not result in impaired walking behavior, as it has been described for Trpc3−/− mice (Hartmann et al, 2008). Moreover, intact spatial reference learning and memory in two different versions of the Morris water maze show that the ubiquitous and constitutive genetic inactivation of Trpc1, Trpc4, and Trpc5 does not impair spatial reference learning as described for hippocampal lesions (Morris et al, 1982, 1990; Aggleton et al, 1986; Logue et al, 1997; Arns et al, 1999; Deacon et al, 2002; Broadbent et al, 2004).
In contrast to our results, Xing et al have suggested impairments in spatial memory, due to the genetic ablation of Trpc1 (Xing et al, 2016). However, the unconventional Y‐maze protocol applied in the study of Xing et al does neither specifically assess spontaneous alternation, spatial working, nor reference memory, as it has been reliably shown by other groups (Hughes, 2004; Fuchs et al, 2007). Also, the fear conditioning results of Xing et al cannot be related to any (spatial) memory deficits, since Trpc1−/− mice already showed markedly less freezing during the acquisition phase. Moreover, the deficits of Trpc1−/− mice in the step‐down inhibitory avoidance task are rather ambiguous, because the test does not only assess hippocampus‐related long‐term memory formation, but has also been associated with amygdala‐dependent anxiety‐like behavior (Izquierdo & Medina, 1997; Collins et al, 2012).
The performance of Trpc1/4/5−/− mice in the T‐maze and radial maze uncovered distinct SWM deficits in TRPC1/4/5‐deficient animals. Trpc1/4/5−/− mice also differed from controls in their competence to use precise and allocentric search strategies during the acquisition in the reversal learning version of the Morris water maze. Trpc1/4/5−/− animals exhibited a substantially higher proportion of undirected search patterns during the daily trials, demonstrating that Trpc1/4/5−/− mice are impaired in recalling efficiently successful search strategies from previous trials on that day. Nevertheless, Trpc1/4/5−/− mice learned the position of the submerged platform in the Morris water maze, indicating that SRM is operative in the absence of TRPC1/4/5.
Interestingly, Pereira and Wang (2015) demonstrated that the accuracy and robustness of intact working memory is enhanced by the activation of calcium‐dependent nonspecific cationic currents (ICAN), which are reminiscent of TRPC‐mediated currents. These observations are in line with the deficiencies we delineated in synaptic plasticity in acute hippocampal slices of Trpc1/4/5−/− mice and mnemonic behavior.
Mice lacking TRPC1/4/5 also exhibited significant deficiencies to adapt to new challenges in the relearning paradigm of the Morris water maze. Thus, the memory for the submerged platform in the Morris water maze seems to be more stable in Trpc1/4/5−/− mice. A more stable SRM, in the absence of SWM, was also noticed in a classical mouse model with strong impairment in SWM, the GluA1‐knockout mice (Gria1−/−) (Zamanillo et al, 1999). In Gria1−/− mice, the SWM was impaired in the T‐maze, whereas the acquisition of SRM in the Morris water maze and radial maze was still intact (Zamanillo et al, 1999; Reisel et al, 2002). Moreover, GluA1‐deficient mice show more stable SRM for previously visited locations (Sanderson et al, 2009). Thus, the delay in SRM relearning in Trpc1/4/5−/− mice can be taken as an additional indicator that TRPC1/4/5‐deficient animals suffer from short‐term spatial memory deficiencies. A very similar learning impairment was described for mice lacking the receptor subunit GluN2A (Grin2a−/−) of hippocampal NMDA receptors (Grin1ΔDGCA) (Bannerman et al, 2008, 2012). In contrast to previous findings that Grin2a−/− mice suffer from SRM deficits, a subsequent detailed analysis of Grin2a−/− concluded that the NMDA receptor GluN2A is required for rapidly acquired SWM, but not incremental SRM (Bannerman et al, 2008), reminiscent to the learning phenotype of Trpc1/4/5−/− mice. Likewise, Grin1ΔDGCA1 mice that lack NMDA receptors in hippocampal pyramidal CA1 neurons and DG granule cells exhibited no deficit in the acquisition of SRM when analyzed in the classical Morris water maze or the beacon water maze (Bannerman et al, 2012). In these Grin1ΔDGCA1 mice, the relearning of a novel platform position in the Morris water maze was impaired, which was interpreted as a lack of behavioral flexibility (Bannerman et al, 2012; Bannerman et al, 2014). Despite the very similar spatial learning impairments of Trpc1/4/5−/−, Gria1−/−, Grin2a−/−, and Grin1ΔDGCA1 mice, only the Gria1−/−, Grin2a−/−, and Grin1ΔDGCA1 show strong impairments in LTP at hippocampal synapses (Zamanillo et al, 1999; Steigerwald et al, 2000; Kohr et al, 2003; Bannerman et al, 2012). In Tpc1/4/5−/− mice, the hippocampal synaptic plasticity appears to be normal in LTP and in corresponding depotentiation measurements, supporting the view that the LTP analysis in acute hippocampal slices might not represent an ex vivo evaluator for hippocampal function in vivo (Neves et al, 2008).
Our findings render the interaction between TRPC1/4/5 and these ionotropic NMDA or AMPA glutamate receptors to be unlikely, although TRPC1 and TRPC4 are apparently activated downstream of NMDA receptor activation in granule cells of the mouse olfactory bulb (Stroh et al, 2012). Alternatively, it has been proposed that TRPC channels might underlie group I metabotropic glutamate receptor (mGluR)‐dependent conductance in CA3 pyramidal neurons (Kim et al, 2003; Hartmann et al, 2008; Wu et al, 2010). This idea is supported by the observations that the lack of TRPC1 and TRPC4, but not of TRPC5 proteins, abolished the burst firing induced by mGluR activity in lateral septum neurons (Phelan et al, 2012, 2013). Previous studies using group I mGluR agonists, antagonists, or toxins that reduce mGluR expression have assigned a critical role for spatial learning and memory formation to mGluR (Riedel & Reymann, 1996; Balschun et al, 1999; Ayala et al, 2009; Jiang et al, 2014). But a conclusive mechanistic link between the involvement of TRPC proteins in mGluR‐mediated synaptic transmission and their function in hippocampus‐dependent behavior will require further in‐depth studies.
In summary, our data provide novel evidences that TRPC1, TRPC4, and TRPC5 interact in the brain and hippocampus. Based on electrophysiological recordings at hippocampal synapses, these TRPC subunits are crucially involved, likely presynaptically, in the efficiency of synaptic plasticity and neuronal network communication, and might, thereby, participate in spatial working memory and flexible spatial relearning.
Materials and Methods
Ethics statement
All experimental procedures were approved and performed in accordance with the ethic regulations and the animal welfare committees of the Universities of Saarland and Heidelberg. All efforts were made to minimize animal suffering and to reduce the number of animals used.
Animals
A triple‐knockout mouse line Trpc1/4/5−/− was generated by intercrossing mice of the three mouse lines—Trpc1−/− (Dietrich et al, 2007), Trpc4−/− (Freichel et al, 2001), and Trpc5−/− (Xue et al, 2011). Each had been backcrossed to the C57Bl6/N strain (Charles River) for at least seven generations before they were used to generate the Trpc1/4/5−/− line. C57BL6/N control mice were obtained from Charles River and housed in the same animal facility as the Trpc1/4/5−/− mice.
Biochemistry/proteomic analysis
Affinity purification.
Membrane fractions from hippocampi and whole brains of adult wild‐type controls, and membrane fractions from whole brains of adult Trpc1−/−, Trpc4−/−, Trpc5−/−‐single‐knockout, or Trpc1/4/5−/−‐triple‐knockout mice were prepared as previously described (Sailer et al, 2002) and solubilized with ComplexioLyte 47 (CL‐47, Logopharm GmbH) for 30 mins on ice (at a concentration of 1.25 mg/ml). After clearing by ultracentrifugation (10 mins, 125,000 × g, 4°C), the solubilized protein was incubated for 2 h on ice with anti‐TRPC1 (ab4921), anti‐TRPC4 (ab1377), or anti‐TRPC5 (ab777 EB) antibodies, generated in‐house, and cross‐linked to Protein A Dynabeads (LifeTechnologies). Beads were washed twice with CL‐47 dilution buffer (Logopharm GmbH), and bound protein was eluted with non‐reducing Laemmli buffer at 37°C.
Data analysis.
MS/MS analysis was done as detailed in Schwenk et al (2014). Briefly, eluted proteins were subjected to an in‐gel tryptic digest. Nano‐LC‐MS/MS analyses were performed using an UltiMate 3000 HPLC and a LTQ Orbitrap XL mass spectrometer (both Thermo Scientific). Peak lists were extracted with “msconvert.exe” (part of ProteoWizard; http://proteowizard.sourceforge.net/; version 3.0.6906; default Mascot Daemon filter options) and—after pre‐search and linear shift mass recalibration—finally searched against all mouse, rat, and human entries (including P00761|TRYP_PIG, P00766|CTRA_BOVIN, and P02769|ALBU_BOVIN) of UniProtKB/Swiss‐Prot (release 2016_08) with Mascot 2.5.1 (Matrix Science; search parameters as described in Schwenk et al (2014).
Protein abundance ratios in anti‐TRPC affinity purifications (versus IgG controls) were calculated as described in Schwenk et al (2016). Peak volumes (PV) of individual peptides were determined by in‐house written software and are provided in Dataset EV1. Relative protein abundance ratios were calculated by the TopCorr method (Bildl et al, 2012), computing the median of PV ratios for the two to six best correlating protein‐specific peptides.
Electrophysiological recordings in autaptic neurons
Autaptic cultures of hippocampal neurons were prepared at P1‐2 from Trpc1/4/5−/− mice, as described (Bekkers & Stevens, 1991; Schoch et al, 2001; Guzman et al, 2010). Hippocampi were dissected from brain and digested for 20 mins at 37°C with 10 units of papain (Worthington, USA), followed by gentle mechanical trituration. Neurons (density 1,000 cells/ml) were seeded onto a layer of glial microislands, resulting in a co‐culture of glia and nerve cells. Only islands containing single neurons were used for electrophysiology. For mass cultures, neuronal cell suspensions were plated at low density (300 cells/mm2) on 25‐mm cover slips coated with 0.5 mg/ml of poly‐D‐lysine (Sigma). Cultures were maintained at 37°C in an incubator, humidified with 95% air and 5% CO2 in NBA (Invitrogen), supplemented with 2% B‐27 (Sigma), 1% Glutamax (Invitrogen), and 2% penicillin/streptomycin (Invitrogen). Recordings were performed at room temperature on days 14–17 of culturing.
Whole‐cell voltage‐clamp recordings of synaptic currents were obtained from isolated autaptic neurons. All experiments include measurements from more than three different culture preparations and were performed in parallel with age‐matched neurons derived from C57Bl6/N wild‐type mice. Patch pipettes (4–5 MΩ) were filled with intracellular solution containing (in mM): 137.5 K‐gluconate, 11 NaCl, 2 MgATP, 0.2 Na2GTP, 1.1 EGTA, 11 HEPES, 11 D‐glucose, pH 7.3. The standard extracellular solution consisted of (in mM) 130 NaCl, 10 NaHCO3, 2.4 KCl, 4 Ca2+, 4 MgCl2, 10 HEPES, 10 D‐glucose, pH 7.3 with NaOH, osmolarity 300 mOsm. Neurons were voltage‐clamped at −70 mV with an EPC10 amplifier (HEKA Electronik) under the control of the Pulse 8.5 program (HEKA Electronik) and stimulated by membrane depolarizations to +10 mV for 0.7 ms every 5 s (0.2 Hz). Current signals were low‐pass‐filtered at 2.9 kHz. For recording spontaneous miniature EPSCs (mEPSCs), mass cultures of hippocampal neurons were bathed in Ringer's solution, containing 1 μM tetrodotoxin (TTX). To minimize the clamp error, cells with access resistances of < 10 MΩ and with 80–85% resistance compensation were analyzed. To determine the mEPSC properties with reasonable fidelity, spontaneous events with a peak amplitude > 15 pA (~5 times the SD of the background noise and a charge criterion > 25 fC) were analyzed using a commercial software (Mini analysis, Synaptosoft, version 6.0.3).
Electrophysiological recordings in brain slices
Adult mice (14–22 weeks) were anesthetized and transcardially perfused with sucrose saline. Acute transverse slices (280 μm) from the middle hippocampus were cut and stored in a submerged chamber with sucrose saline (in mM: 87 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 7 MgSO4, 0.5 CaCl2, 10 glucose, 65 sucrose, 0.01 sodium pyruvate bubbled with 95% O2/5% CO2, ~300 mOsm). Schaffer collaterals (CA3 axons) were stimulated every 20 s with a 1‐μm‐tip monopolar tungsten electrode (WPI) by biphasic electrical stimulation in stratum radiatum (~300 μm shifted toward CA3). With two glass pipettes filled with 1 M NaCl, local field potentials (LFPs) were recorded in stratum radiatum and the corresponding population spikes were recorded in stratum pyramidale in current clamp (EPC10, HEKA Elektronik, Patchmaster software), as shown in Fig 5A. Acute slices were perfused at 33°C in a submerged recording chamber with saline (in mM: 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 1 MgCl2, 2 CaCl2, 13 glucose, bubbled with 95% O2/5% CO2, ~300 mOsm). Slopes of LFPs were determined by the slope of a 30–70% linear fit; amplitudes of population spikes (negative deflection) were determined relative to baseline. Paired‐pulse LFPs were evoked during input–output curves with an interstimulus interval of 50 ms, but not during plasticity experiments.
Data analysis.
Data were analyzed in Igor Pro 6.37 (Wavemetrics). Results are shown as mean ± SEM. Statistical significances were assessed in GraphPadPrism 5.02 or in R 3.3.2 (package “car”, version 2.1‐3). In plasticity experiments (Fig 5I–K), the absolute LFP slopes of the respective experiments were compared between 10‐min baseline average and the average of the last 5 mins after the respective plasticity induction by a paired t‐test. Statistical differences in the categorical independent factor “genotype” (Fig 5C, D and H) or “first” versus “second paired pulse” (Fig 5F and G) were tested with two‐way ANCOVA of type II, implemented in R. The respective continuous dependent response (from the y‐axis) and the second independent (continuous) factor (from the x‐axis) were taken into account for this analysis.
Electrophysiological recordings in vivo
Animal care and housing conditions.
Animals were housed in groups of up to four, inside a ventilated Scantainer (Scanbur BK) on an inverted 12‐h light/dark cycle (light on at 8:00 P.M.). All animals had free access to food and water. After chronic electrode implantation, animals were housed individually throughout the experiment. After behavioral experiments, animals were killed with an overdose of isoflurane.
Surgery procedure.
A total of 10 male animals, aged between 10 and 14 weeks, weighing ~20–24 g were anesthetized with isoflurane and medical oxygen (4% isoflurane for induction, 1.0–2.0% for maintenance, flow rate: 1 l/min). For analgesia, 0.1 mg/kg of buprenorphine hydrochloride was subcutaneously injected prior and 8 h after surgery. Anesthetized animals were head‐fixed on a stereotaxic apparatus (David Kopf Instruments) with a custom‐made inhalation tube. Body temperature was kept stable at 37–38°C with a heating pad underneath the animal's body (ATC‐2000, World Precision Instruments). After exposure of the skull, holes of a diameter of 0.6 mm were drilled according to stereotaxic coordinates (Paxinos & Franklin, 2001). Two octrodes (eight twisted wires) were implanted into the right and left dorsal CA1 region (AP +/− 1.94 mm, ML +/− 1.25 mm, DV −1.1 mm), and two brass watch stainless steel screws (M1x3) were implanted into the cerebellum to serve as reference and ground electrode, respectively. After surgery, animals were allowed to recover for 7 days.
Electrophysiology.
We recorded LFPs from the dorsal hippocampal CA1 region of freely moving mice. In addition to electrophysiological recordings, three accelerometers registered translational movements in x, y, and z directions. One week after surgery, recordings were performed for 5 h in the animal's home cage (see Fig 4A). Extracellular signals were filtered (1–500 Hz), amplified (RHA2116, Intan Technologies), digitized (2.5 kHz), and stored for offline analyses with custom‐written MATLAB routines (The MathWorks).
Behavioral staging.
We restricted our analyses to selected rapid eye movement sleep phases (REM), where both characteristic and robust theta‐nested gamma oscillations (4–12 Hz and 30–140 Hz) occur. REM sleep was identified by the absence of movement on all accelerometers accompanied by prominent regular theta oscillations.
Spectral analysis.
Power spectral density analysis was calculated by means of the Welch periodogram method using the pwelch.m function from the Signal Processing Toolbox of Matlab (50% overlapping, 4‐s Hamming windows).
Cross‐frequency coupling analysis.
To measure the intensity of coupling between the phase of a slow oscillation and the amplitude of a fast oscillation, we computed the modulation index (MI, for details see Tort et al (2008)). In brief, the MI is a measure to evaluate the divergence of phase–amplitude coupling from a uniform distribution, normalized such that values range between 0 (no coupling) and 1 (maximal coupling). We next selected the MI peaks to evaluate the maximum MI values of low gamma and high gamma to theta coupling, respectively, for further statistical analyses.
Immunostaining in hippocampal neurons
Mass neuronal cultures were fixed for 30 mins at room temperature in PBS containing 4% paraformaldehyde and 4% sucrose. Cultures were quenched for 10 mins with 50 mM NH4Cl in PBS, blocked for 1 h in PBS containing 3% BSA, and 0.1% Triton X‐100. Primary and secondary antibodies were diluted in 3% BSA, 0.1% Triton X‐100 in PBS. Cells were incubated with primary antibodies anti‐synaptophysin (mouse monoclonal, Synaptic Systems, Germany) overnight. After rinsing, they were incubated once with PBS containing 4 M NaCl, then washed briefly in PBS. Alexa‐Fluor 543 (Invitrogen) was applied for 2 h. After washing in PBS, cells were imaged immediately or mounted on 2 μl glycerol. Epifluorescence pictures were digitized with an AxioCam MRm‐CCD camera (Zeiss) at the focal plane of the axonal network with a 25× and 100× objective (Zeiss). Images were acquired with the software Axiovision 4.5 (Zeiss) and analyzed with Metamorph software (Universal Imaging Inc.). Immunopositive spots were determined using a threshold‐based detection routine, with the threshold adjusted to the background signal of the dendrite. Immunosignals were quantified as mean fluorescent intensity. For the analysis of synaptic density, synaptophysin‐positive puncta were counted along 50 μm dendrite length (Guzman et al, 2010).
Histochemistry and immunostainings in brain slices
Experimentally naive 14‐week‐old mice were anesthetized with isoflurane and transcardially perfused with PBS as well as 4% paraformaldehyde (PFA; Roti®‐Histofix 4%, Roth). Prior to and after the postfixation with PFA in PBS (4°C) for 1.5 h, brains were washed in PBS and then embedded in 2% agarose (in PBS). Coronal sections (100 μm) were cut on a vibratome (Leica VT1000S) and kept in PBS. For the immunostaining, they were first pretreated with 0.5% H2O2 (Fluka Analytical) in PBS for 15 mins to quench endogenous peroxidase activity, followed by Solution D1 (1% albumin from bovine serum (BSA; Sigma), 0.3% Triton X‐100 (Sigma) in PBS) supplemented with 2% normal goat serum (NGS; Invitrogen) to block unspecific protein binding sites. Several washing steps with PBS were performed in between. After 1 h, brain slices were incubated with primary antibodies overnight. The dilutions were prepared in D1 and consisted of rabbit antibody targeting either GluA1 (1:1,000; abcam®), glial fibrillary acidic protein (GFAP; 1:500; DakoCytomation), or somatostatin (1:400; abcam®) or mouse antibody targeting NeuN (1:1,000; Millipore®). On the next day, after rinsing with D2 (0.33% BSA (Sigma), 0.1% Triton X‐100 (Sigma) in PBS), the sections were incubated for 1 h with peroxidase‐labeled anti‐rabbit IgG secondary antibody (Vector), diluted 1:600 in D2, followed by washing steps. Slices were then stained with diaminobenzidine solution (0.4 mg/ml DAB in 20 mM Tris (pH 7.6) and 30% H2O2). The reaction was stopped with PBS. Stained sections were mounted onto glass slides using 10 mM Tris (pH 7.6). After drying, slices were embedded in xylene (Merck) with Eukitt® quick‐hardening mounting medium (Fluka Analytical). All staining processes were performed at room temperature. For Nissl staining, brains were taken from experimentally naive 14‐week‐old mice, immediately frozen on dry ice, and stored at −80°C. Transverse sections (12 μm) were cut at −20°C on a cryostat (Leica CM3050 S) and mounted onto microscope slides, previously coated with poly‐L‐lysine. Slides were kept at −20°C. Brain sections were dried at room temperature for 2–3 h before they were incubated for 140 s in thionine solution (0.1% thionine in 0.1 M acetic acid and 0.1 M Na‐acetate, filtered) and rinsed with distilled water. After drying, slices were embedded in xylene with Eukitt®. Bright field images of DAB‐ and Nissl‐stained slices were taken with the Axio Imager M1 (10× magnifying objective, Zeiss) using the Zeiss Axiovision software.
Behavioral assays
Animals were housed in an animal facility with a regular 12‐h light/dark cycle (light on at 7:00 A.M.). Food and water was supplied ad libitum, unless the mice were kept on a food‐restricted diet during an appetite‐motivated learning task (T‐maze, radial maze tests). All behavioral tests were conducted with counterbalanced, gender‐, and age‐matched groups (two‐ to 4‐month‐old male and female adults). We compared C57Bl6/N wild‐type and Trpc1/4/5−/− mice. Different tests were performed on different days.
SHIRPA (SmithKline Beecham, Harwell, Imperial College, Royal London Hospital, phenotype assessment).
Mice were placed in a corner of a clear Plexiglas box and allowed to explore for 30 s. For the evaluation of spontaneous behavior, the body position (lying, sitting, rearing), activity (from immobile to extreme movement), body tremor, twitches, number of urine puddles, and fecal boli were taken into consideration. Vision was tested by lowering the mouse onto a wire grid by the suspension of its tail. The relative distance to the grid at which the animal started to extend its forelimbs was used as a criterion for visual placing responses. Hearing skills were assessed by exposing the mouse directly to a loud sound made by a rubber glove. Sudden twitches were counted as a positive reaction. Tested reflexes included corneal, pinna, or toe pinch reflex. Eye blinking, and active retraction of the ear or the withdrawal of the hindlimbs, respectively, was expected.
Rotarod.
In the rotarod experiment, a rod rotated at a gradually increasing speed. It initially turned at 4 rpm, going up to 40 rpm over a period of 8 mins. The time the animal was able to maintain itself on the rotating rod was measured to judge motor coordination. Mice were subjected to three runs per day (intertrial interval: 15 mins) on 2 days with a 1‐day break in between.
Delayed non‐match‐to‐place T‐maze.
The apparatus was an enclosed T‐shaped maze constructed of gray Plexiglas. Every arm was 7 × 10 × 28 cm. Throughout the entire test period, animals were kept on a food‐restricted diet to motivate them for the food‐rewarded task. Providing ~2 g of pellets per mouse every day, 85–90% of the initial body weight was maintained. At all times, access to water was unlimited.
Over 4 days, the animals were habituated to the maze in four runs per day. In each run, they were allowed to explore one given goal arm. At its end, they were rewarded with a 2‐mg pellet of food, composed of 61.5% carbohydrate, 18.5% protein, 4.9% fat, 4.5% fiber (caloric value: 3.62 cal/mg). The opposing goal arm was closed by inserting a gray Plexiglas divider. During the four runs, the accessible goal arms were randomly alternated. If the mouse had not found the pellet within 120 s, the animal was put back to its home cage. The intertrial interval for each habituation run was ~10 mins. Over the following 10 days, mice performed six trials per day, with a 20‐min intertrial interval. Each trial was composed of two runs in the T‐maze. During the first run (test run), the unrewarded arm was blocked and the animal could only enter the food‐rewarded goal arm. After 10 s, the mouse was placed back to the start arm for the second (free choice) run. In the free choice run, both goal arms of the T‐maze were accessible, but only the previously blocked (alternative) arm contained a food pellet. It was considered an error, when the mouse entered the same arm in the test and free choice run of one trial. The trial was valued as successful if the pellets in both arms were taken. Entering an arm with all four paws was counted as choice. The cutoff time per run was again 2 mins.
Radial maze.
The apparatus consisted of eight equally sized enclosed arms (5.5 × 6 × 27.5 cm) made of gray Plexiglas with transparent lids and walls. These arms were assembled in a radial manner around a circular starting platform. The animals were kept on a food‐restricted diet with unlimited access to water, maintaining 85–90% of their pre‐test body weight. Habituation: On two successive days, all eight arms were open and baited with a couple of oat flakes. Each mouse was given 10 mins to familiarize itself with the environment. Then, the animals were tested on 10 consecutive days with one trial per day with all eight maze arms being open. For each mouse, a specific set of four arms was baited with an oat flake; the other four arms were always unbaited. For every animal, always the same four arms were baited throughout the test. In contrast to the habituation, now only one oat flake was placed behind a low barrier to prevent the mice from seeing whether or not a certain arm was still baited before entering it completely. Some flakes were deposited behind a perforated wall at the very end of each arm. This way, the animals could not distinguish between the presence and absence of the food reward by smelling. Several extra‐maze cues, whose position remained constant throughout the experiment, were provided close to the arms. The maze arms were exchanged after every trial to avoid the use of aromatic cues for orientation. The trial was stopped after all four food rewards were eaten or after 10 mins. Every repeated entry into a previously visited arm was counted as a working memory error.
Morris water maze experiment and pattern analysis.
The apparatus was a circular gray metal pool (diameter: 180 cm; AnyMaze 60235, USA) filled with water (22 ± 1°C) that was made opaque by adding TiO2 and frequent stirring between trials. The circular white escape platform (diameter: 10 cm) was submerged 1 cm below the water surface in the middle of the goal quadrant. Extra‐maze cues were positioned around the pool. Sygnis Tracker software was used to measure swim path length, latency to reach the platform, time spent in each quadrant, and cumulative distance to the platform. If a mouse failed to reach the goal within 90 s, it was guided there manually. After mounting the platform, the animal remained there for 10 s.
For 2 days, the animals were subjected to a visual platform task. Here, a protruding colorful grid made the platform visible. The maze was covered with black curtains to hide the room cues. In each of the four runs per day (intertrial interval: 30 mins), the platform was placed in a different quadrant, while the swimming start position was always diagonal to the platform location.
Spatial reference memory version (Malleret et al, 1999).
On 11 days following the visual platform task, mice were trained to find the hidden platform by using the visual cues surrounding the pool; that is, the colorful grid and the black curtains were removed. Throughout the entire training, the platform remained in the same position (upper left quadrant). The animals performed four runs per day, each starting from a different position along the pool wall (E, SE, S, and SW). The order of the starting positions was chosen randomly but remained the same for all mice throughout a training day. The subsequent probe trial on day 12 consisted of one 120‐s trial in opaque water without any platform. The starting direction was farthest away from the “platform” quadrant (SE).
Reversal learning version (Garthe et al, 2009).
Two days of habituation (visual platform task, see above) were directly followed by 3 days of training and 2 days of reversal learning. Each mouse was given six trials per day. Throughout 1 day, the starting position remained the same, but it was changed every day. During the reversal part, the platform was moved from the upper left (training) to the lower right quadrant.
The analysis of the search strategies allowed a qualitative estimation of the learning process during the Morris water maze task (Garthe et al, 2009). The following categories were defined according to specific criteria, such as time spent or distance traveled in a specific pool region: “Thigmotaxis” describes slow swimming close to the pool wall. The swim path of a “random search” pattern covers a large portion of the pool. During “scanning”, the animal swims mainly in the central area, where it can see the distant visual cues best and where it can “scan” its environment. Circling in a ring zone around the center when the correct distance of the platform to the wall is learned is considered “chaining”. These non‐spatial strategies dominate in the early learning phase. As the training progresses, mice develop spatially more precise and thus hippocampus‐dependent search behavior. “Distracted search” is considered as transition between non‐spatial and spatial strategies. Here, the animals search with preference for a relatively small pool area before they swim straightly to the platform.
In “directed search”, the animal navigates mainly in the “goal corridor”, a direct path between starting point and platform with a given width, still allowing for some uncertainty in the search. In “focal search”, the mouse heads directly to the goal, searches closely to the platform before mounting it. The most direct path between start and goal is used during “straight swimming”. These three swim patterns were summarized in the category “allocentric (other‐centered) search” because here the animal uses the spatial arrangement of the experimental room, rather than its own current location, to find the hidden platform. Lastly, “perseverance” describes a persistent search in the former platform quadrant after moving the goal during the reversal part.
For the automated assignment of the swim paths to the defined categories, the parameters recorded with Sygnis Tracker software were used to derive descriptive criteria for the different search categories. The individual runs were classified using a decision tree built on the parameter space of all obtained data. When the run did not fall into any of the nine categories: (i) straight swimming, (ii) focal search, (iii) directed search, (iv) perseverance, (v) chaining, (vi) thigmotaxis, (vii) scanning, (viii) random search, (ix) distracted search—it was classified as unknown. Throughout our analysis, the rate of unknown strategies was < 8%.
Statistics and data analysis in behavioral tests.
Results are shown as mean ± SEM. Statistical significance was determined using two‐tailed unpaired Student's t‐test, unless otherwise stated. To judge the performance of the animals over time in the T‐maze, Poisson regression with autoregressive GEE (generalized estimating equation) (Ziegler & Vens, 2011) model was applied. In particular, for each genotype, it was tested whether the slope of the log best fit over days 1–10 differed significantly from zero. Similarly, a difference in the overall performances between the two genotypes was statistically tested by examining the interaction between the genotype and time variable, that is, to compare the slopes of the log best fits. Differences with P < 0.05 were considered statistically significant. Significances were depicted as *P < 0.05, **P < 0.01, and ***P < 0.001.
Author contributions
JB‐L planned and performed all behavioral experiments, morphological stainings, and analyzed these data. AK and BF performed affinity purifications and mass spectrometry analysis. VF generated and VF, AK, and BF validated TRPC antibodies. BS, RG, and YS performed electrophysiological analysis and fluorescence microscopy in cultured neurons under supervision of DB. JP and GK performed slice physiology. IM, HS, and RS gave conceptual input in behavioral and morphological studies. AL developed the algorithm for the pattern analysis. VNC, MB, and ADr performed electrophysiological recordings in vivo. PW participated in the generation of mouse lines and mouse breeding. ADi provided a mouse line. The manuscript was initially written by JB and MF. DB, RS, JP, GK, BF, AK, and VF complemented the manuscript and made critical revision. MF and DB conceived, designed, and supervised the study.
Conflict of interest
The authors declare that they have no conflict of interest.
Expanded View
Expanded View Figures PDF [embj201696369-sup-0001-EVFigs.pdf]
Dataset EV1 [embj201696369-sup-0002-DatasetEV1.xlsx]
Acknowledgements
We thank Christin Matka, Tanja Volz, Tom Janke, Annette Herold, and Hans Peter Gensheimer for technical assistance as well as Claudia Pitzer and Barbara Kurpiers (Interdisciplinary Neurobehavioral Core at the Medical Faculty, Heidelberg, University, INBC) for the support during behavioral experiments. This work was supported by HOMFOR (DB) and by the Transregional Collaborative Research Center (TR‐SFB) 152 (MF, DB, BF, ADi, VF), the Collaborative Research Centre (SFB) 1118, FOR 2289, and the DZHK (Deutsches Zentrum für Herz‐Kreislauf‐Forschung—German Centre for Cardiovascular Research) and by the BMBF (German Ministry of Education and Research) (MF). RS, ADr, and GK receive support from the SFB 1134 projects B01, A01, and B05, respectively. RS and ADr are also supported from the SFB 1158 projects A05 and B05.
Funding
Deutsche Forschungsgemeinschaft (DFG)http://dx.doi.org/10.13039/501100001659 TR‐SFB 152SFB 1118SFB 1134SFB 1158Footnotes
‡Correction added on 15 September 2017 after first online publication: Affiliation 10 was added
References
- © 2017 The Authors
