Improved spatial learning is associated with increased hippocampal but not prefrontal long-term potentiation in mGluR4 knockout mice


  • E. Iscru,

  • H. Goddyn,

  • T. Ahmed,

  • Z. Callaerts-Vegh,

  • R. D'Hooge,

  • D. Balschun

    Corresponding author
    • Laboratory of Biological Psychology, Faculty of Psychology and Educational Sciences, Katholieke Universiteit Leuven, Leuven, Belgium
    Search for more papers by this author

Corresponding author: D. Balschun, Laboratory of Biological Psychology, Faculty of Psychology and Educational Sciences, Katholieke Universiteit Leuven, Leuven, Belgium. E-mail:


Although much information about metabotropic glutamate receptors (mGluRs) and their role in normal and pathologic brain function has been accumulated during the last decades, the role of group III mGluRs is still scarcely documented. Here, we examined mGluR4 knockout mice for types of behavior and synaptic plasticity that depend on either the hippocampus or the prefrontal cortex (PFC). We found improved spatial short- and long-term memory in the radial arm maze, which was accompanied by enhanced long-term potentiation (LTP) in hippocampal CA1 region. In contrast, LTP in the PFC was unchanged when compared with wild-type controls. Changes in paired-pulse facilitation that became overt in the presence of the GABAA antagonist picrotoxin indicated a function of mGluR4 in maintaining the excitation/inhibition balance, which is of crucial importance for information processing in the brain and the deterioration of these processes in neuropsychological disorders such as autism, epilepsy and schizophrenia.

Over the last decade, metabotropic glutamate receptors (mGluRs) have been intensively studied as potential targets for treating different diseases of the central nervous system (CNS), such as anxiety disorders, schizophrenia, Parkinson's disease, fragile X syndrome, Lewy body and Alzheimer's disease (Gross et al. 2012; Hopkins et al. 2009; Krystal et al. 2010; Lee et al. 1995; Luscher & Huber 2010; Niswender & Conn 2010; Osterweil et al. 2010; Price et al. 2010; Ribeiro et al. 2010; Sokol et al. 2011). The mGluRs belong to class C heptahelix G-protein-coupled receptors and are subdivided into three groups based on sequence homology, pharmacology and intracellular signaling mechanisms (Niswender & Conn 2010). Group I mGluRs (mGluR1 and mGluR5) are postsynaptically located and coupled to Gq/G11-like G-proteins, resulting in the release of inositol (1,4,5)-trisphosphate (IP3) and diacylglycerol, and the activation of protein kinase C (PKC). Group II mGluRs (mGluR2 and mGluR3) and also group III mGluRs (mGluR4, 6, 7 and 8) are mainly presynaptically expressed and are Gi/o coupled leading to a decrease of cyclic AMP via inhibition of adenylyl cyclase. Once activated these receptors decrease postsynaptic activity by acting as presynaptic autoreceptors and heteroreceptors at glutamatergic and GABAergic terminals, respectively (Bordi & Ugolini 1999; Nakanishi 1994; Niswender & Conn 2010).

The mGluRs are not only involved in a variety of CNS disorders but have also been proven to play an important role in multiple forms of synaptic plasticity [e.g. (Anwyl 2009; Balschun et al. 1999a; Gladding et al. 2009; Huber et al. 2000; Klausnitzer et al. 2004; Luscher & Huber 2010; Moult et al. 2006)] and learning (Altinbilek & Manahan-Vaughan 2007, 2009; Balschun et al. 1999a; Callaerts-Vegh et al. 2006; Lyon et al. 2011; Riedel et al. 2003).

Disturbances in synaptic plasticity and learning, in turn, belong to the hallmarks of many psychiatric, neurodevelopmental, neurodegenerative and cognitive disorders [e.g. (Bear 2005; Goto et al. 2010; Marchetti & Marie 2011; Niswender & Conn 2010; Rowan et al. 2003; Wang et al. 2011)].

Group III mGluRs are the least understood among the mGluRs. Mainly owing to the lack of specific pharmacological tools, only a few studies addressed the function of this mGluR group in synaptic plasticity and learning. Thus, it was reported that intraventricular (i.c.v.) application of group III agonist l-2-amino-4-phosphonobutanoic acid (L-AP4) evoked robust long-term depression in the hippocampal CA1 region and the dentate gyrus (DG) (Manahan-Vaughan 2000; Naie et al. 2006). Long-term potentiation (LTP) in the CA1 region could be blocked by the group III antagonist (RS)-alpha-cyclopropyl-4-phosphonophenylglycine (CPPG) in vitro (Grover & Yan 1999) but not in vivo (Altinbilek & Manahan-Vaughan 2007).

In this study, we examined types of synaptic plasticity and learning in mGluR4 knockout (mGluR4−/−) mice that depend on either the hippocampus or the prefrontal cortex (PFC). The mGluR4−/− mice displayed a markedly enhanced LTP in the hippocampus but not in PFC. Furthermore, increased paired-pulse facilitation (PPF) upon application of GABAA antagonists suggests a function of mGluR4 in the regulation of the excitation/inhibition balance. These specific changes of short- and long-term synaptic plasticity were accompanied by improved spatial working (WM) and reference memory (RM) in mGluR4 mutant mice.

Materials and methods


The mGluR4−/− mice generated as described previously (Pekhletski et al. 1996) were extensively backcrossed to C57BL/6J background. Age-matched wild-type littermates (mGluR4+/+) were used as controls, and all genotypes were confirmed by polymerase chain reaction (Pekhletski & Hampson 1996). The mGluR4−/− mice do not show any gross abnormalities in cytoarchitecture when compared with unaffected wild-type littermates, and they are devoid of any mGluR4 expression (Bradley et al. 1999; Pekhletski et al. 1996; Shigemoto et al. 1997). All mice were bred in the animal facility of Janssen Pharmaceutica and transferred at an age of 12 weeks to Leuven where they were kept as mixed genotype groups in standard animal cages and temperature- and humidity-controlled rooms (12 h light/dark cycle, 22°C) with food and water ad libitum.

Mice were housed in our facilities for at least 2 weeks and habituated daily to handling by an experimenter before subjected to behavioral testing. All experiments were performed with female mice. Experiments were conducted during the light phase of the activity cycle. All protocols have been reviewed and approved by the Ethical Research Committee of the Katholieke Universiteit Leuven and were performed in accordance with the European Community Council Directive (86/609/EEC).


Electrophysiological recordings were performed on slices of the hippocampus and the PFC, prepared from 4- to 6-month-old mice. Animals were killed by cervical dislocation and the brain was rapidly extracted and placed in ice-cold preoxygenated artificial cerebrospinal fluid (ACSF) consisting of (in mM) 124 NaCl, 4.9 KCl, 2.5 CaCl2, 1.3 MgSO4, 1.2 NaH2PO4, 25.6 NaHCO3, 16.6 d-glucose, gassed with 95% O2/5% CO2, pH 7.4. Transverse slices of hippocampus and coronal slices of PFC (400 µm thickness) were prepared by a tissue chopper and incubated for 1 h at room temperature before being placed in a submerged-type four-chamber recording system (Campden Instruments LTD, Loughborough, Leics., UK) and kept there at 32°C and a flow rate of 1.8–2 ml/min/chamber. In all experiments, custom-made monopolar tungsten electrodes were used for stimulation and glass electrodes filled with ACSF (5–7 MΩ resistance) for recording of field excitatory postsynaptic potentials (fEPSPs). The initial slope of the fEPSP served as a measure of this potential. To assess basic properties of synaptic responses I/O curves were established by stimulation with 30–90 µA constant currents (pulse width 0.1 millisecond). The stimulation strength was adjusted to evoke a fEPSP slope of 35% of the maximum and kept constant throughout the experiment. Paired-pulse facilitation was examined by applying two pulses in rapid succession (interpulse intervals of 10, 20, 50, 100, 200 and 500 milliseconds, respectively) at 120-second intervals. In all experiments, the recording of slices from mutant mice was interleaved by experiments with wild-type controls.

Recordings from mPFC

To record LTP in mPFC, the stimulation electrode was lowered into layer II and the recording electrode into layer V, as schematically shown in Fig. 1b. During baseline recording, three single stimuli (0.1-millisecond pulse width) at a 10-second interval were applied every 5 min. Once a stable baseline had been established, LTP was induced by four episodes of high-frequency stimulation (HFS) at 100 Hz for 1 second (0.2-millisecond pulse width), with 5-min interval between consecutive episodes. Recordings were taken 1 min after each HFS train and thereafter every 5 min.

Figure 1.

Basic excitability and LTP in medial PFC of mGluR4+/+ and mGluR4−/− mice. (a) Input/output curves of both genotypes are almost identical. (b) Image of a slice with stimulation and recording electrodes. (c) The mGluR4−/− mice display robust LTP in mPFC, which closely resembles the potentiation observed in wild-type littermates. Note the different kinetics of LTP induction in both genotypes. Because of a significantly faster increase of potentiation, mGluR4−/− mice were able to compensate for an initially lower level of potentiation. Means ± SEM are given. Insets show representative analog traces taken at the times indicated by numbers.

Hippocampal recordings

Recordings in the CA1 region of the hippocampus and the DG were performed as described previously (Balschun et al. 1999b, 2010). The correct location in the medial perforant pathway (mPP) of the DG was verified by obtaining paired-pulse depression at 40-millisecond interpulse intervals. Because of the high level of intrinsic inhibition in this area, 10 µM of the GABAA inhibitor picrotoxin was added to the bath solution.

In DG, LTP was generated by a theta-burst stimulation (TBS) protocol of 15 trains of eight stimuli delivered at 200 Hz, separated by 200 milliseconds, 0.2-millisecond pulse width, applied three times at 10-min intervals (Balschun et al. 1999b). Robust LTP in the CA1 region was induced by applying three trains of TBS (10 bursts of four stimuli at 100 Hz, separated by 200 milliseconds, 0.2-millisecond pulse width) (Larson & Lynch 1986; Larson et al. 1986) at an interval of 5 min. In a subset of CA1 experiments, two tetanization protocols that each induces a weak decremental type of LTP were delivered at an interval of 2 h. First, a single HFS train (100 Hz, 400-millisecond duration, 0.2-millisecond pulse width) was applied (Wilsch et al. 1998) followed 2 h later by a second tetanus, consisting of a single TBS (10 bursts of four stimuli at 100 Hz, separated by 200 milliseconds, 0.2-millisecond pulse width). Synaptic potentiation was monitored for 2 h.


Radial arm maze (RAM)

Mice were tested in an eight-armed RAM (TSE systems, Bad Homburg, Germany). This maze consists of a central platform (20 cm in diameter) and eight radially attached arms (30 × 5 cm). Before starting the experiment, mice were trained to collect the bait at the end of the arm during five trials with one trial per day. Working memory was tested using a delayed-matching-to-sample (DMTS) task, by which four arms randomly chosen by the computer program were baited with a small food sample (piece of almond). The animals got access to the arms through guillotine doors that opened/closed automatically. In each trial, a mouse was placed on the central platform with the guillotine doors initially closed but then given access to all eight arms to collect the bait. Each arm was baited only once and animals needed to use short-term memory and spatial cues to remember which arm they had already entered to retrieve the food bait. Each mouse had a maximum of 5 min to identify the baited arms and to collect the baits. When the animal chose and entered one arm, all the other seven doors closed. When the animal returned to the central platform, the last door closed, and the animal was confined for 5 seconds on the central platform before all doors opened again.

Mice were tested on four trials per day with intertrial intervals (ITIs) of 1 h. The performance of the mice was registered by optical light beams located at the arm entries and by a web camera attached to the ceiling above the RAM (127 cm), as well as build in light beams that registered arm visits and successful bait removal.

Two types of errors were measured: RM errors, quantified as visits into an unbaited arm, and WM errors, quantified as re-entries into a previously visited arm. Entries into a baited arm without collecting rewards were not counted as procedural errors.

Statistical analysis

Data are presented as mean and standard error of the mean (SEM). Within-group differences were evaluated with the Wilcoxon matched-pairs signed-ranks test. Statistical analysis of between-group differences was performed with the t-test or analysis of variance (anova), with Tukey tests for post hoc comparison.

To test for group differences between trial blocks over several days and LTP time series, anova with repeated measures (RM-anova) was used (SPSS 19; IBM, Armonk, NY, USA). As RM-anova is sensitive to any deviation from sphericity, the F-values were corrected in case Mauchly's test of sphericity was significant (Huynh–Feldt correction if epsilon was >0.75, otherwise Greenhouse–Geisser correction).

For intergroup comparisons of paired-pulse ratios, Mann–Whitney U-test was applied. To assess the time course of LTP induction, the fEPSP slopes recorded 1 min after each stimulus train were fitted by linear regression and the results compared by a method equivalent to analysis of covariance (ancova) (GraphPad Prism 5.01; GraphPad Soft Inc., La Jolla, CA, USA).


Deletion of mGluR4 does not affect LTP in mPFC

In the first set of experiments, we examined whether deletion of mGluR4 had any effect on long-term synaptic plasticity in mPFC, a brain area that has been associated with functional changes in cognitive disorders like frontotemporal dementia and Alzheimer's disease. Furthermore, mPFC has a pivotal function in higher order cognitive processes such as WM, decision making and goal-directed behaviors (Barbas 2000a,b; Dalley et al. 2004; Fuster 2001; Goldman-Rakic 1994; Kolb 1984; Petrides 1998). After having tested basic excitability in mPFC that did not differ between knockout and control mice (Fig. 1a), we inspected LTP in this region. The HFS induction protocol evoked a similar robust level of potentiation in both groups, which was stable for at least 2 h (1 min: mGluR4−/−: 132 ± 5%, n = 6, mGluR4+/+: 146 ± 9%, n = 6; 2 h: mGluR4−/−: 127 ± 8%, mGluR4+/+: 129 ± 9%; Fig. 1c; genotype: F1,10 = 0.008, P = 0.929, time: F4.8,48.4 = 4.336, P = 0.003, time × genotype interaction: F4.8,48.4 = 1.23, P = 0.310; RM-anova).

However, a more detailed analysis showed a genotype difference in the kinetics of LTP induction. As evaluated by linear regression, the increase in potentiation of mGluR4−/− caused by every further episode of HFS was faster when compared with control mice (slope of regression: mGluR4+/+: 1.02 ± 0.3; mGluR4−/−: 2.58 ± 0.4; F1,4 = 11.52, P = 0.027, ancova-like analysis). In that way the mGluR4−/− mice were able to compensate for an initially lower level of potentiation.

mGluR4−/− mice show changes in short-term plasticity and tend to have larger dentate LTP

Next, we probed into the role of mGluR4 in synaptic plasticity in the DG. This region of the hippocampal formation was suggested to act as filter for cortical inputs arising mainly from the entorhinal cortex (Coulter & Carlson 2007; Witter et al. 2000) and to play a fundamental role in mediating pattern separation (O'Reilly & McClelland 1994; Rolls 2010).

While there was no difference in basic excitability of mPP-granule cell synapses (Fig. 2a), we detected an interesting functional change in presynaptically mediated short-term plasticity (Anderson 1960) in the absence of mGluR4. Thus, the increased excitability in the presence of 10 µM picrotoxin resulted in paired-pulse responses at 10- and 20-millisecond interstimulus intervals that showed a significantly higher facilitation of the second pulse in mGluR4−/− than in control mice (Fig. 2b). Therefore, while without picrotoxin the paired-pulse ratios at 10 and 20 milliseconds are about the same in the two genotypes (10 milliseconds: mGluR4+/+: 0.58 ± 0.07, n = 7; mGluR4−/−: 0.49 ± 0.06, n = 7; 20 milliseconds: mGluR4+/+: 0.68 ± 0.13; mGluR4−/−: 0.83 ± 0.05), addition of the GABAA antagonist resulted in a much bigger increase in facilitation in mutant mice (10 milliseconds: 349.2%, 20 milliseconds: 211.4%) than in control animals (10 milliseconds: 179.1%, 20 milliseconds 157.1%). This increase was only statistically significant in mGluR4−/− mice (P < 0.001 at 10 milliseconds and P = 0.001 at 20 milliseconds; mGluR4+/+ P = 0.07 at 10 and 20 milliseconds, Mann–Whitney U-test). Noticeably, recordings showed signs of hyperexcitability when picrotoxin was added to the bath as exemplified by representative analog traces in Fig. 2c,d.

Figure 2.

Basic excitability, paired-pulse responses and LTP in the dentate gyrus of mGluR4+/+ and mGluR4−/− mice. (a) Input/output curves of both genotypes do not differ. (b) The mGluR4−/− mice display a marked change of short-term plasticity in the presence of the GABAA antagonist picrotoxin. (c) Representative analog traces at the interpulse intervals of 20 and 100 milliseconds, respectively. (d) The mGluR4−/− mice show consistently higher mean values of potentiation but this does not reach the level of significance. See Fig. 1 for further explanation.

Although the mean LTP values of mGluR4−/− were consistently higher than controls (1 min: mGluR4−/−: 259 ± 18%, n = 7; mGluR4+/+: 236 ± 12%; 2 h: mGluR4−/−: 212 ± 24%; mGluR4+/+: 178 ± 9%, n = 7; Fig. 2d), the genotype difference was not statistically significant (genotype: F1,12 = 1.340, P = 0.269, time: F2.3,27.6 = 18.123, P = <0.001, time × genotype interaction: F2.3,27.6 = 1. 832, P = 0.175, RM-anova).

Synaptic plasticity in CA1 area is significantly enhanced by the absence of mGluR4

As third region for investigating synaptic plasticity, we selected the hippocampal CA1 area, the principal hippocampal output region for spatial information (Gigg 2006) that presumably encodes information arriving from CA3 (Rolls 2010). As in the DG, recordings in the CA1 area did not indicate differences in basic synaptic excitability between mGluR4−/− and mGluR4+/+ mice as clearly evidenced by similar input/output characteristics (Fig. 3a). The paired-pulse protocol, however, was indicative of changes in presynaptically mediated short-term plasticity because the ratio of mGluR4−/− mice at an interpulse interval of 50 milliseconds was significantly increased (P = 0.002; Mann–Whitney U-test) (Fig. 3b). Thereafter, we tested whether deletion of mGluR4 also affected long-term synaptic plasticity in this region. As shown in Fig. 3c, LTP in mGluR4−/− mice was significantly enhanced when compared with control littermates (1 min: mGluR4−/−: 269.57 ± 8%, n = 7; mGluR4+/+: 197.84 ± 3%, n = 6). This difference was maintained for at least 2 h (2 h: mGluR4−/−: 158.32 ± 5%; mGluR4+/+: 121.21 ± 5%; genotype: F1,11 = 22.706, P = 0.001, time: F3.5,38.8 = 42.316, P < 0.001, time × genotype interaction: F3.5,38.8 = 1.777, P = 0.160, RM-anova).

Figure 3.

Basic excitability, paired-pulse responses and LTP in the CA1 region of mGluR4+/+ and mGluR4−/− mice. (a) Input/output curves of both genotypes are not different. (b) The mGluR4−/− mice display an increase in PPF at 50 milliseconds. (c) The LTP in mGluR4−/− mice is markedly enhanced when compared with wild-type littermates. Insets in (b) and (c) depict representative analog traces. (d) The mGluR4−/− mice show a slower fatigue of potentials during tetanization. The slope of the first of four potentials in each of the 10 TBS bursts (see expanded view of a burst) is compared. The slope decay during the first and second trains of TBS is significantly slower in mutant mice when compared with controls. Data are normalized to the slope of the first potential of burst 1. See Fig. 1 for further explanation.

To check whether the higher amplitude of CA1-LTP in mGluR4−/− mice was already triggered during tetanization, we analyzed the dynamics of field responses during the three trains of TBS stimulation in more detail (Fig. 3d). When the slope of first of the four potentials in each of the 10 bursts is compared, there is an overall decay of the slope from burst 1 to burst 10 in both groups. However, during the first and second trains of TBS, the decay is significantly slower in mutant mice when compared with controls, indicating a slower fatigue of potentials in the absence of mGluR4 (Fig. 3d; genotype: slope of first TBS: F1,80 = 8.80, P = 0.04; second TBS: F1,64 = 5.43, P = 0.004; third TBS: F1,64 = 4.34, P = 0.07; time: first TBS: F8,80 = 3.36, P < 0.001; second TBS: F8,64 = 36.99, P < 0.001; third TBS: F8,64 = 53.05, P < 0.001; time × genotype interaction: first TBS: F8,80 = 3.36, P < 0.001; second TBS: F8,64 = 2.96, P < 0.001; third TBS: F8,64 = 0.54, P = 0.8).

Prompted by this marked difference in TBS-induced CA1-LTP, we examined the functional impact of mGluR4 on metaplastic properties of Schaffer collateral-CA1 synapses by applying a combined protocol where two types of weak, decremental LTP were subsequently induced at a time interval of 2 h. First, a weak decremental type of LTP was evoked by single HFS (40 pulses at 100 Hz, 0.2-millisecond pulse width). Such weak types of potentiation have been proven to be particularly vulnerable to any functional disturbance (e.g. Balschun et al. 1999b; Wilsch et al. 1998). The HFS generated a similar decremental LTP in both groups (1 min: mGluR4−/−: 178.71 ± 8%, n = 6; mGluR4+/+: 160 ± 9%, n = 6; Fig. 4) that returned to baseline within 120 min. Although potentiation in mGluR4-deficient mice appeared to be slightly stronger, this difference was not significant (genotype: F1,10 = 2.399, P = 0.152; time: F2.5,24.7 = 44.522, P < 0.001; time × genotype interaction: F2.5,24.7 = 1.016, P = 0.390; RM-anova).

Figure 4.

Metaplasticityis strengthened in mGluR4−/− mice. When two different weak induction protocols are applied, TBS generates a more robust LTP in mGluR4−/− mice when compared with control littermates.

Two hours after the first tetanization, another tetanization paradigm was applied, consisting of a single TBS (10 bursts of four stimuli at 100 Hz, separated by 200 milliseconds, 0.2-millisecond pulse width). In response to TBS, both groups showed a similar initial magnitude of potentiation (1 min: mGluR4−/−: 137.55 ± 17%; mGluR4+/+: 120.04 ± 16%) (Fig. 4). However, the potentiation in controls decayed faster than in mutant mice resulting in a significant genotype effect at the end of recording (last 30 min; genotype: F1,10 = 6.811, P = 0.026; time: F4.6,45.8 = 5.591, P = 0.001; time × genotype interaction: F4.6,45.8 = 1.672, P = 0.166).

mGluR4−/− mice show improved spatial RM and WM in RAM

As PFC and hippocampus are well documented to have a central role in WM and spatial learning, respectively, we examined whether performance of mGluR4 mutant mice is changed in a learning task that critically depends on these two regions. Thus, we employed a DMTS task in a RAM, which allows testing of short-term WM and long-term RM in a spatial task. When examined in this paradigm, mGluR4−/− mice showed significantly better WM performance as assessed by the number of re-entries into a previously visited arm. When compared with their wild-type littermates, mGluR4−/− mice started already with a lower number of WM errors (mGluR4−/−: 3.4 ± 0.7, n = 8; mGluR4+/+: 5.1 ± 0.5, n = 10) and this difference was maintained for the whole 12 days of testing (day 12: mGluR4−/−: 0.9 ± 0.3, n = 8; mGluR4+/+: 1.6 ± 0.2, n = 10; Fig. 5a; genotype: F1,16 = 6.643, P = 0.02; time: F10.1,161.1 = 11.199, P < 0.001; time × genotype interaction: F10.1,161.1 = 0.418, P = 0.937, RM-anova). The better performance of mGluR4−/− mice was not confined to WM because they acquired also a better spatial RM, as evidenced by fewer visits to unbaited arms (e.g. RM error on day 1: mGluR4−/−: 4.6 ± 0.6; mGluR4+/+: 6.4 ± 0.3; day 12: mGluR4−/−: 1.4 ± 0.2; mGluR4+/+: 2.3 ± 0.2; Fig. 5b; genotype: F1,16 = 5.991, P = 0.026; time: F10.3,164.3 = 21.643, P < 0.001; time × genotype interaction: F10.3,164.3 = 0.768, P = 0.663, RM-anova).

Figure 5.

Spatial learning in a DMTS task in an eight-arm radial maze. mGluR4−/− mice committed less WM errors (a) and RM errors (b) than their wild-type littermates.


Groups 2 and 3 mGluRs are both thought to act at glutamatergic and GABAergic terminals as inhibitory autoreceptors and heteroreceptors, respectively. Because of their negative coupling to adenylate cyclase signaling cascades (Niswender & Conn 2010), the primary consequence of their activation is a depression of glutamatergic and GABAergic transmission (Schoepp 2001). In so doing, they make a significant contribution to the tuning of transmission properties of various types of synapses (Ferraguti et al. 2005; Scanziani et al. 1997; Semyanov & Kullmann 2000; Shigemoto et al. 1996).

While the role of group 2 mGluRs in synaptic plasticity is documented by multiple studies, the investigation of group III mGluRs in these processes is only at its beginning. For example, i.c.v. application of the group III agonist L-AP4 evoked robust long-term depression in the hippocampal CA1 region and the DG that could be blocked by the group III antagonist (RS)-CPPG (Manahan-Vaughan 2000; Naie et al. 2006). The same compound had no effect on CA1-LTP (Altinbilek & Manahan-Vaughan 2007). Grover and Yan (1999) reported that NMDA receptor-independent LTP in the CA1 region in vitro could be blocked by the group III antagonist CPPG but not by (RS)-a-methylserine-O-phosphate. However, because the pharmacological effectors used in these studies do not distinguish between the subtypes of group III mGluRs, their genetic deletion is the method of choice to provide further insights into their functional role in synaptic plasticity. The mGluR4-deficient KO mice used in our study displayed a consistent strengthening of LTP in the CA1 region in vitro. This increase was observed irrespective of whether a robust potentiation was generated by repeated trains of tetanization or a decremental potentiation was induced by weak stimulation protocols. This seems to be in contradiction to a recent in vivo study by Altinbilek and Manahan-Vaughan (2007) in rats where application of the broad-band orthosteric group III antagonist CPPG had no effect on CA1-LTP. The difference to our results could be caused by: (1) species differences because rats and mice differ in the laminar distribution of mGluR4 in the hippocampus (Shigemoto et al. 1997) indicating an involvement in different functions. (2) A graded affinity of CPPG for the various group III mGluRs. Low concentrations of the compound are likely to bind to mGluR8 owing to the much higher affinity to this receptor when compared with mGluR4 (Naples & Hampson 2001). The mGluR8 is expressed in a subset of CA1 interneurons but not in pyramidal cells in this region, suggesting a role of mGluR8 in the generation of complex spike bursts (Ferraguti et al. 2005). (3) An incomplete inhibition of mGluR4 by the compound. In a recent study, Niswender et al. (2008) have shown that even at a high concentration as 1 mM, CPPG attained only about 60% inhibition of mGluR4 in vitro. Such high concentrations, however, cannot be applied in physiological experiments because CPPG has been described to inhibit group II mGluRs at higher concentrations.

Detailed analysis of the decay of responses during repeated TBS stimulation showed a slower synaptic fatigue in mGluR4−/− mice. This is indicative of a more sustained glutamate release and depolarization in mGluR4−/− mice during trains of TBS stimulation, thereby facilitating the induction of LTP.

Our findings of improved LTP after genetic abolishment of mGluR4-mediated presynaptic inhibition are reminiscent of a similar strengthening of potentiation after the application of group II antagonists (Behnisch et al. 1998). Interestingly, the improvement of potentiation after weak tetanization in mGluR4 mutant mice was stronger in response to a theta-burst protocol when compared with a weak, HFS of 40 pulses at 100 Hz. Both types of tetanization paradigms were shown to trigger different signal transduction cascades (Hoffman et al. 2002; Selcher et al. 2003).

Although the mean LTP values of mGluR4−/− mice in DG were consistently higher than controls, the genotype difference did not reach the level of significance. In this region, the paired-pulse measurements showed an interesting functional change in mGluR4 mutants, which became only overt when GABAA receptor-mediated inhibition was reduced by adding picrotoxin to the bath solution. This marked increase in paired-pulse response observed in picrotoxin-treated slices of KO mice is most likely owing to a changed interaction of the following mechanisms: presynaptically mediated PPF due to residual calcium that remains in the presynaptic terminal after the conditioning stimulus (Wu & Saggau 1994), a decrease in recurrent GABAA-mediated inhibition caused by picrotoxin (Fisher et al. 1997) and the lack of mGluR4-mediated control of glutamate and GABA release (Niswender & Conn 2010; Schoepp 2001). The latter seems to be particularly important when inhibition is weakened, as indicated by our data.

In further search of specific mechanisms underlying the increased excitability, we have to consider the fact that mGluR4 is not only localized in the presynaptic zones of principal cells but also of GABAergic interneurons (Corti et al. 2002; Kogo et al. 2004). This location is ideal to control GABA release by acting as heteroreceptor on GABAergic nerve terminals (Kogo et al. 2004; Somogyi et al. 2003). Probing further into the mechanisms, Rusakov et al. (2004) identified N-type Ca2+ channels as the principal target of l-(+)-2-amino-4-phosphonobutyric acid (L-AP4)-sensitive group III mGluRs at GABAergic terminals. The resulting depression of presynaptic Ca2+ influx mediates a reduction of GABA release in response to elevations of extracellular glutamate (Rusakov et al. 2004; Semyanov & Kullmann 2000). Thus, deletion of mGluR4 might have led in the DG to a disturbance of inhibition circuits owing to a change of presynaptic GABA release mechanisms and this had become overt at interpulse intervals of 10 and 20 milliseconds, respectively, after bath application of picrotoxin.

The paired-pulse ratio was not only changed in the DG but also in the CA1 region, but there without any use of GABAA antagonists. The mGluR4−/− mice showed increased PPF at 50-millisecond interpulse interval, indicating a lack of mGluR4-mediated tonic inhibition on glutamate release in mutant mice. On the background of different neuronal circuits and proportions of interneuronal subtypes in both regions (Freund & Buzsaki 1996), the increased PPF could be due to similar but not the same mechanisms as suggested above for the DG.

Although mGluR4 has been reported to reside in PFC (Corti et al. 2002; Gupta et al. 2005), we did not find any differences in basic excitability and LTP between mGluR4 mutant mice and their control wild-type littermates in basic excitability and LTP in this region. This does not exclude, of course, that other induction protocols for LTP would have resulted in discernible differences between the two genotypes, or that other types of synaptic plasticity could be more relevant to performance of the DMTS task.

Generally, there is not much known yet about the role of mGluRs in PFC synaptic plasticity. Activation of group II mGluRs was implicated in the induction of long-term depression in PFC (Huang & Hsu 2008; Otani et al. 2002) and mGluR7 was assigned a role in the regulation of NMDA receptors (Gu et al. 2012), which could point to a role of this group III mGluR subtype in plasticity.

The clear enhancement of LTP in the CA1 region of mGluR4−/− mice and the lack of significant effects of mGluR4 deletion on LTP in DG and PFC are difficult to explain because of the premature knowledge about this mGluR subtype (Niswender & Conn 2010; Niswender et al. 2008). Most likely, these differences are caused by different expression profiles of mGluR4 in these regions (Bradley et al. 1999; Shigemoto et al. 1997), which result in a region-specific participation of the receptor in diverse physiological functions such as the fine-tuning of glutamate and GABA release in response to the variety of activation patterns in the intact brain.

More importantly, however, we found a significant role of mGluR4 in LTP in the CA1 area, the principal hippocampal output region for spatial information that presumably further encodes information arriving from CA3 (Gigg 2006; Rolls 2010). Given the essential role of CA1 in NMDAR-dependent spatial learning and plasticity (Tsien et al. 1996), the improved RAM performance of mGluR4−/− mice is well in line with the enhanced LTP in the CA1 region.

One of the advantages of the RAM consists of allowing the distinction of both WM and RM components. Trial-specific WM allows the animal to recall arms entered previously in the current trial and to avoid re-entries. On the other hand, RM refers to stored representations and rules that are useful for all trials. In our study, better performance of mGluR4−/− mice was seen in short-term WM and long-term RM. The RAM learning has been shown to depend not only on the functional integrity of the hippocampus but also of the PFC (Kawabe et al. 1998; Olton & Papas 1979) and changes in WM were described to be positively correlated with LTP changes in PFC and hippocampal subregions (Cui et al. 2011; Dallerac et al. 2011, Holscher et al. 2007; Niewoehner et al. 2007). In contrast to the improved WM and RM, we did not observe changes in prefrontal LTP in vitro. However, there are several examples in the literature (and our own work) where changes in LTP were not accompanied by changes in memory performance and vice versa, or where LTP and memory changes were even negatively correlated (e.g. D'Hooge et al. 2005; Gerlai et al. 1998a; Jolas et al. 2002; Walther et al. 1998). This apparent discrepancy may be due to a number of reasons such as (1) a different sensitivity of LTP and learning performance to the molecular and physiological changes following a gene deletion or the insertion of a transgene, (2) the use of inappropriate or insensitive protocols, (3) the fact that learning represents the sum of putatively different effects of the manipulation in various brain regions, while the LTP effect is the synaptic readout of just one brain region and (4) learning but not LTP (in vitro) can be modulated by non-cognitive factors such as a higher motivational drive.

Interestingly, deletion of mGluR7, another group III mGluR subtype, resulted in partially opposite effects on RAM learning performance. The mGluR7−/− animals committed more WM errors than control mice in a four- and eight-arm RAM but the number of RM errors did not differ between genotypes (Callaerts-Vegh et al. 2006; Holscher et al. 2004). However, these effects were variable because in a configuration where only two of the arms of a four-arm maze were baited, there was no difference in WM errors. In a water maze study with mGluR4−/− mice, Gerlai et al. (1998a,b) found neither functional deficits in acquisition and retention (probe trial) of spatial memory nor in short-term memory (ITI = 90 min) and ‘medium-term’ memory (ITI = 17 h). They did see, however, a faster adaptation, i.e. an increased flexibility of mGluR4−/− mice after platform reversal, which is indicative of the ability of mGluR4−/− mice to show increased performance if demanded by specific circumstances of the task.

Repeated pharmacological blockade of group III mGluRs by the group III antagonist CPPG during RAM learning in rats has been described to result in an impairment of long-term RM while having no effect on short-term WM (Altinbilek & Manahan-Vaughan 2007). However, as outlined already above, because of the strong affinity of this compound for mGluR8 (Naples & Hampson 2001), it is not clear whether at the given dose an action on mGluR4 was involved in the effect.

In conclusion, increased LTP in the CA1 region as well as improved spatial short- and long-term memory after deletion of mGluR4 are in agreement with the presumed function of this mGluR subtype of exerting presynaptic inhibition. Changes in PPF in CA1 and DG point further to a role of mGluR4 in keeping the excitation/inhibition balance, which is one of the most critical variables in the processing of sensory information and of pivotal importance in neuropsychological disorders as autism, epilepsy and schizophrenia (Isaacson & Scanziani 2011; Yizhar et al. 2011).


This work was supported by Katholieke University Leuven (Impulse Fund IMP/04/006, OT/06/23 to D.B. and Fellowship to T.A.), Fonds voor Wetenschappelijk Onderzoek (FWO grants) G.0271.06 and G.0496.07 to D.B. and R.D. and a PhD fellowship to H.G. Z.C.-V. is a postdoctoral fellow financed by an IWT/J&J-PRD Research & Development grant. We would like to acknowledge Robert Duvoisin, Oregon Health & Science University, Department of Physiology and Pharmacology, Portland, OR, USA, for providing breeding pairs of mGluR4−/− mice, Johnson & Johnson Pharmaceutical Research and Development, Beerse, Belgium, for help in maintaining the breeding colony, Ben Vermaercke for his support with the analysis of the responses during TBS stimulation and Eva Staessens for her participation in the experiments. The authors have no conflict of interest to declare.