The effect of mGlu8 deficiency in animal models of psychiatric diseases


M. Fendt, Novartis Institutes for BioMedical Research, WSJ-386.3.28, Forum 1, Novartis Campus, 4056 Basel, Switzerland. E-mail:


The metabotropic glutamate receptor subtype 8 (mGlu8) is presynaptically located and regulates the release of the transmitter. Dysfunctions of this mechanism are involved in the pathophysiology of different psychiatric disorders. mGlu8 deficient mice have been previously investigated in a range of studies, but the results are contradictory and there are still many open questions. Therefore, we tested mGlu8-deficient animals in different behavioral tasks that are commonly used in neuropsychiatric research. Our results show a robust contextual fear deficit in mGlu8-deficient mice. Furthermore, novel object recognition, chlordiazepoxide-facilitated extinction of operant conditioning and the acoustic startle response were attenuated by mGlu8 deficiency. We found no changes in sensory processing, locomotor activity, prepulse inhibition, phencyclidine-induced changes in locomotion or prepulse inhibition, operant conditioning, conditioned fear to a discrete cue or in animal models of innate fear and post-traumatic stress disorder. We conclude that mGlu8 might be a potential target for disorders with pathophysiological changes in brain areas where mGlu8 modulates glutamate and gamma-amino butyric acid (GABA) transmission. Our data especially point to anxiety disorders involving exaggerated contextual fear, such as generalized anxiety disorders, and to conditions with disturbed declarative memory.

The metabotropic glutamate receptor subtype 8 (mGlu8) inhibits transmitter release via Gi/o proteins and associated effector pathways (summarized in Kew & Kemp 2005). Like the closely related subtype 7 (mGlu7), mGlu8 is exclusively located on the presynaptic side of glutamatergic and gamma-amino butyric acid (GABA)ergic terminals (Ferraguti et al. 2005; Shigemoto et al. 1997), but has about a 1000 times higher affinity to glutamate than mGlu7 (Schoepp et al. 1999). Because dysfunctions of the glutamatergic and GABAergic brain system are involved in the pathophysiology of different psychiatric disorders (Lewis & Moghaddam 2006; Nemeroff 2003), it has been suggested that mGlu8 might be a potential target for such diseases.

Currently, little is known about the function of mGlu8. Because specific antibodies for mGlu8 that were validated in mGlu8-deficient mice are not public available, the exact distribution of mGlu8 throughout the whole brain is largely unknown. In situ hybridization of mGlu8 mRNA shows high mRNA levels in the lateral septum, the amygdala and some brainstem nuclei, as well as moderate mRNA levels in the cerebral cortex and the hippocampus [Allen Brain Atlas (; Duvoisin et al. 1995; Messenger et al. 2002; Saugstad et al. 1997]. Some of these brain nuclei are involved in important brain functions such as learning, memory, emotions and locomotor activity (e.g. Eichenbaum 1997; Everitt & Robbins 1997; Fendt & Fanselow 1999) and show pathological changes in different human psychiatric disorders (Dickerson & Sperling 2008; Etkin & Wager 2007; Harrison 1999).

As with antibodies, there is also a lack of specific agonists or antagonists for mGlu8. Therefore, the characterization of mGlu8-deficient mice is currently the only method available to investigate the physiological role of mGlu8. Previous reports state that mGlu8 deficiency is not connected to a very obvious pathological phenotype (e.g. Gerlai et al. 2002). However, when tested in different animal behavioral models, a phenotype was described in innate anxiety models, body weight, pain sensitivity and locomotor activity, but not in animal models of learning, memory, schizophrenia and epilepsy (Duvoisin et al. 2005; Gerlai et al. 2002; Linden et al. 2002; Robbins et al. 2007). It is difficult to compare these studies because different background strains were used, but it is notable that there are contradictory results. For example, mGlu8-deficient animals had an increased body weight in one study (Duvoisin et al. 2005) but not in another (Gerlai et al. 2002). They showed an anxiogenic-like phenotype in several animal models of anxiety (Duvoisin et al. 2005; Linden et al. 2002, 2003; Robbins et al. 2007), but an anxiolytic-like phenotype in a conditioned fear model (Gerlai et al. 2002). Likewise, the phenotype in the open field was described to be hypoactive (Duvoisin et al. 2005), hyperactive (Gerlai et al. 2002) or unchanged (Linden et al. 2002; Robbins et al. 2007). These differences are of critical importance because changes in locomotor activity affect many behavioral tasks.

The aim of the present study was to further analyze the role of mGlu8. Specifically, we were interested in exploring the effects of mGlu8 deficiency in conditioned fear and in learning and memory because the roles of mGlu8 in these phenomena are poorly understood so far. Therefore, we tested mGlu8-deficient mice which are bred on a background (C57BL/6) in animal models of conditioned fear, learning and memory, and also in other models of pivotal importance for neuropsychiatric research.

Material and methods


Originally, mGlu8 knockout mice were made by gene targeting W9.5 embryonic stem cells (129Ola) as described by Duvoisin et al. (2005). Mice that transmitted the mutant allele were mated with C57BL/6J mice (Jackson Laboratory, Bar Harbor, ME, USA) for five generations before animals were transferred to Novartis where they were further backcrossed for five more generations by mating with C57BL/6JNpa mice (Npa refers to Novartis Pharma, Basel, Switzerland). For behavioral experiments, batches of wild-type littermate (mGlu8+/+) and mutant mice (mGlu8–/–) were produced by mating 30 F10-B6 mGlu8+/–males with 60 F10-B6 mGlu8+/–females and the male offspring were used for behavioral analysis. If not stated else, the animals were experimentally naive, and at least 12 weeks old at the start of the experiment.

Genotyping (University of Ulster): Mice were genotyped by multiplexed polymerase chain reaction (PCR), using primers 2 (5′-TAACTACCAGGTGGACGAACTCTC-3′), 3 (5′-CACAAAGGTGGTG GCAATGATTCC-3′) and 4 (above). The wild-type allele generated a 174-bp product using primers 2 and 3, whereas the mutant allele generated a 362-bp product from primers 2 and 4.

Genotyping (Novartis, Basel): For genotyping, tail DNA (DNeasy Tissue Kit; QIAGEN, Hornbrechtikon, Switzerland) was analyzed using either a gel-based method and diagnostic PCR-generated DNA fragments (see above) or using a Taqman procedure. For the Taqman assay, the following primers were used: TGGATTGCACGCAGGTTCT (sense), GTGCCCAGTCATAGCCGAAT (antisense) and CGGCCGCTTGGGTGGAGAGG (5′FAM, 3′BHQ1 labeled probe). For reference, primers detecting the endogenous glucocorticoid receptor gene were used: CGGGACCACCTCCCAAA (sense), CCCCATAATGGCATCCCGAA (antisense) and CTTCATCGGAGCACACCAGGCAGA (5′YY, 3′BHQ1 labeled probe).

Animal holding: The animals were housed in groups of two to four per cage in a humidity and temperature-controlled room under a 12/12-h day/night cycle with lights on at 07:00 h. Water and food were available ad libitum. The animal cages were equipped with plastic nest boxes, wooden chew blocks and nesting material. Experiments were in accordance with international guidelines for the care and use of animals and approved by the local animal welfare councils.


Locomotor activity

A computerized motility measurement system was used (Moti 4.25, TSE Systems, Bad Homburg, Germany). This system automatically measures locomotor activity in transparent boxes (20 × 32 × 17 cm; ca. 150 lux) by frames with infrared beams spaced 5.7–8.4 cm apart. A second frame is able to detect rearing behavior.

Elevated plus maze

The maze was made of gray Perspex and had two open arms (30 × 5 cm) with ledges, 0.3 cm high, and two enclosed arms of the same size with transparent walls 15 cm high, elevated 56 cm above the ground. Illuminance was ca. 240 lux. A camera was mounted vertically above the maze in order to record animals' behavior on the maze for subsequent analysis by the video-tracking system SMART 2.5 (Panlab S.L., Spain). Criterion for arm entry was crossing of the animal's body center.

Elevated zero maze

The maze comprised a gray Perspex annular platform (46 cm diameter, 5.5 cm width) elevated to 41.5 cm above ground level, divided equally into four quadrants. Two opposite quadrants were enclosed by gray Perspex walls (11.5 cm high) on both the inner and outer edges of the platform; the open quadrants had ledges, 0.3 cm high. Illuminance was ca. 550 lux in the open quadrants and 200 lux in the closed quadrants. A video camera was mounted above the maze in order to record animals' behavior on the maze for subsequent analysis by a trained observer.

Light-dark box

The apparatus (45 × 21× 21 cm) consisted of two compartments, one dark (1/3, ca. 35 lux) and one lit (2/3, ca. 915 lux), and separated by a divider with a 12 × 5 cm opening. A camera was mounted vertically above the maze in order to record animals' behavior on the maze for subsequent analysis by the video-tracking system SMART 2.5 (Panlab S.L., Spain).

Conditioned freezing

A computerized fear conditioning system was used (TSE Systems, Bad Homburg, Germany) consisting of four identical transparent Perspex boxes (46 × 46 × 32 cm) placed inside animal detection infrared sensor frames. Each box was located in a sound-attenuating chamber provided with loudspeakers for the acoustic stimuli (background noise of 60 dB SPL and the tone stimuli for fear conditioning), light sources (continuous illumination of 10 lux) and a ventilation fan. The floor of the boxes consisted of removable stainless steel grids (bars: 4 mm diameter, distance: 8.9 mm), which were connected to a shock unit and able to deliver foot shocks of defined duration and intensity. Delivery of all stimuli was controlled by a personal computer. Four additional boxes of the same size as those described above but made of black Perspex (including the floor) served to create a different test context. Movements of the animals were detected by the infrared sensors (distance: 14 mm). The time spent freezing (defined as no infrared beam crosses for more than 1 second) was automatically recorded during all phases of the experiments. Automatically measured freezing in the TSE fear conditioning system is highly correlated with human scoring of freezing (Endres et al. 2007; Misane et al. 2005).

Object recognition

The arena was a white polyester box 60 × 40 cm and with 20-cm high walls. Animal behavior was tracked by a ceiling-mounted camera (Vector tronics; 680 × 800 lines resolution) and a computer-based image analysis system (Biosignals, NY, USA). The room had several visual landmarks such as benches, door and racks for spatial orientation of the mice. Objects used (marbles, dice, glass stopper, pen tops, etc.) were tested against each other to ensure that one object was not explored significantly longer than another object. Objects were of different shape and surface to maximize recognition (e.g. colored glass marble, 1.2 cm diameter; dice, 1.4 cm width, black and white). The tracker program recorded any visits within a 1-cm circle around the object as an exploration event.

Operant behavior

Operant chambers for mice (MED Associates; model no. ENV 307A, 22 × 18 × 13 cm) were enclosed in sound-attenuating boxes with electric fans. Chambers consisted of translucent side panels and aluminum instrument and rear panels. Instrument panels were equipped with two retractable response levers and a house light located at the center top of the panel. Only the left levers were used. Reinforcers (20-mg Noyes food pellets) were delivered by a 28-V DC pellet dispenser to a recessed tray located at the bottom of the panel between the two levers. A computer programmed in MED-PC, which also recorded presses on the retractable levers, controlled all events in the chambers.

Startle system

A SR-LAB startle system (San Diego Instruments, San Diego, USA) was used for the present experiments. It consisted of eight sound-attenuating, ventilated wooden chambers (35 × 35 × 38 cm) equipped with high-frequency loudspeaker for delivering acoustic stimuli and incandescent bulbs for light stimuli. During the startle experiments, the animals were put into acrylic (Plexiglas) cylinders with 4 cm diameter and a length of 10 cm. The cylinders were fixed onto a horizontal acrylic white plate (13 × 20 × 0.5 cm) with a transducer. Animal movements (i.e. startle responses) were detected with the transducer. The output signal of the transducers was digitized for data acquisition (sampling rate 24 bit, 1 kHz) and stored on a computer. The peak-to-peak amplitude of the transducer output within the 100 milliseconds after the startle stimulus was used as the startle magnitude and expressed in arbitrary units.

Experimental procedures

Locomotor activity and phencyclidine-induced hyperactivity

The aim of this experiment was to measure potential effects of mGlu8 deficiency on spontaneous locomotor behavior. Therefore, naive mice were put into the boxes and locomotor activity was measured for 60 min (n = 9/group). The animals were then injected i.p. either with saline or with 2.5 mg/kg phencyclidine (PCP; Sigma, Schnelldorf, Germany; dissolved in saline; injection volume: 10 ml/kg; i.p.), and immediately put back into the boxes, and activity was measured for an additional 60 min. Phencyclidine is known to produce behavioral effects that are characteristic symptoms for schizophrenia, e.g. hyperactivity or prepulse inhibition (PPI) deficits (Geyer et al. 2001; Jentsch & Roth 1999). Therefore, PCP-induced behavioral changes are believed to be animal models of schizophrenia.

Elevated plus maze

This test measures the anxiety induced by open spaces (i.e. absence of thigmotactic cues; Treit et al. 1993). Naive animals were placed in the central platform facing an open arm. The test session lasted 5 min. Two independent experiments with two batches of animals were carried out (n = 10−11/group). Main readouts were open arm ratio for time, entries and distance, i.e. the time spent on the open arm divided by the time of the experiment, as well as the number of entries or distance traveled on the open arm divided by the total number of entries or total traveled distance.

Elevated zero maze

The elevated zero maze is a modification of the elevated plus maze (Shepherd et al. 1994). It has no center which allows uninterrupted exploration. Naive animals (n = 11/group) were placed in the middle of a closed quarter. The test session lasted 5 min.

Light-dark box

The light-dark box test measures the anxiety induced by brightly illuminated areas (Mathis et al. 1994). The animals were placed centrally in the lit compartment facing away from the dark compartment. The test session lasted 10 min. Two different batches of animals (n = 8−10/group) were used in two independent, replicated experiments. The batch used in the first experiment were tested in the elevated plus maze 2 days before; the second batch of animals were experimentally naive.

Conditioned freezing

Conditioned freezing is one of the most recognized animal models for conditioned fear (Fendt & Fanselow 1999; Phelps & LeDoux 2005). In total, we performed four experiments. Experimentally naive mice were used for each experiment. The first experiment was our standard test on the learning, retrieval and extinction of conditioned fear: In adult mice (12–16 weeks, n = 10/group), a fear conditioning session was performed using the transparent Perspex boxes. The animals were individually placed into the boxes. Sixty seconds later, the first of six pairings of a tone stimulus (8 kHz, 80 dB, 30 seconds, pulsed with an interval of 100 milliseconds; intertrial interval: 90 seconds) and a scrambled foot shock (0.6 mA, during the last 2 seconds of the tone stimulus) was presented. Thirty seconds after the last pairing, the mice were returned to the home cage. On the next day, the animals were placed back for 3 min into the conditioning boxes (transparent Perspex boxes) and their freezing response to the conditioning context was quantified. The mice were then returned to the home cage. Two hours later, cued fear was assessed: the animals were put into the black Perspex boxes, and after a habituation period of 2 min, 10 tone stimuli (without foot shocks) were presented with an interstimulus time of 1 min (retention test). These two test sessions were daily repeated on the following 2 days.

In the second freezing experiment, mice of different ages were used. The rationale for this experiment was that we observed in a pilot experiment that the behavioral deficit which is present in 12-week-old mGlu8-deficient mice seems to be absent in younger animals. Therefore, we tested three groups of mice with the following ages: 7 weeks, 8–9 weeks and 12–13 weeks old (n = 9−12/group). The conditioning session, as well as the test on contextual fear, was performed as described earlier; the retention test on cued fear consisted only of five conditioned stimulus (CS) presentations. The third and fourth test day was cancelled.

The aim of the third experiment was to check whether sensitivity to foot shocks is different in mGlu8-deficient mice. We therefore performed conditioned freezing experiments with different intensities of the foot shock unconditioned stimulus (US). The conditioning procedure was identical to those of the previous experiments, but foot shock intensities of 0.4 and 1.0 mA were used (n = 14−16/group). In addition, only the retention test on contextual fear was carried out.

The fourth freezing experiment investigated the long-term behavioral effects of a single traumatic foot shock in wild-type and mGlu8-deficient mice (see Siegmund & Wotjak 2007). The animals (n = 12−15/group) were put into the conditioning boxes (odor of the cleaning agent ‘Thedra’; 50 lux) and after a habituation period of 200 seconds, one single foot shock with a duration of 2 seconds and an intensity of 1.5 mA was administered (‘trauma exposure’). No tone was presented. Sixty seconds later, the animals were put back into the home cages. After 28 days, the animals were then tested for sensitized fear (hyperarousal) by putting the animals into a new context (black Perspex boxes; odor of 70% ethanol; 10 lux) and testing the behavioral response to a 180-second tone (80 dB SPL, 8 kHz; ‘tone test') after a habituation period of 180 seconds. Again, the animals were put back into the home cages after further 60 seconds. On the following day, the freezing response of the animals was tested in two different contexts. The first context shared some stimuli with the conditioning context (grid, transparent Perspex) but others not (odor of 1% acetic acid, triangular outline; 10 lux) and was used to test for context discrimination and generalization of conditioned fear. The second context was the conditioning context, i.e. conditioned fear should be evoked. Both tests last 3 min and were separated by a break of 5–6 h. One day later, the animals were tested for their startle response. After a habituation period of 5 min, 30 blocks with startle stimuli (white noise, 20-millisecond duration) in four different intensities (75, 90, 105 and 115 dB SPL) were presented in a pseudorandomized order, with an interstimulus interval of 13–17 seconds.

Object recognition

Novel object recognition is believed to be an animal model of declarative memory (Winters et al. 2008). In this test, the ability to discriminate between a familiar and a novel object is tested. Animals were handled and exposed to the object recognition box for 2 days before experiments started. In the first session, the animals were exposed to four different objects spaced 10 cm apart for 3 min to evaluate whether some of the objects are preferred by the animals. After 3 h, the same objects were shown again (second run) to evaluate if recognition of familiar objects is similar for the different objects. In a third test session (familiar vs. novel), after 3 h, two familiar objects and two novel objects were shown. The objects were pseudorandomly altered between four defined locations to avoid a spatial bias in the exploration of objects. In each session, total duration of exploration was measured and compared (Dere et al. 2007).

Chlordiazepoxide-induced facilitation of extinction following operant conditioning

In this experiment, we wanted to assess whether mGlu8-deficient mice show a phenotype in operant conditioning. In addition, previous studies have shown that facilitation of extinction following operant conditioning can be induced by a variety of GABAergic agents (Leslie et al. 2004; McCabe et al. 2004). Therefore, this procedure seems to be a good assay for GABAergic activity during extinction training (mGlu8 is localized on GABAergic terminals; Ferraguti et al. 2005). The animals used for this experiment were maintained between 80% and 90% of their free-feeding weight by providing 4–8 g of laboratory chow once daily. In the training phase of each experiment, sessions were only conducted 5 days a week and ad libitum food was available at the weekends. After initial training to lever press for reinforcement, with 20 mg Noyes pellets, mice were placed on a discrete-trial fixed ratio 5 (FR5) schedule with six reinforcers delivered in each session. Training continued until stability, and then 25 daily extinction sessions were carried out (see Shaw et al. 2009 for procedural details). Chlordiazepoxide (CDP; Sigma-Aldrich Company Ltd, Gillingham, UK) was dissolved in 0.9% saline. Injections (via the intraperitoneal route) were made at a dose of 15 mg/kg (i.p., 4 ml/kg). This dose has no effects on locomotor performance, especially lever pressing, in C57BL/6 mice (McCabe et al. 2004). Vehicle injections were given on the final 4 days of training sessions, and vehicle or CDP injections (n = 4−7/groups) were given on each day of the extinction sessions. All injections were given 30 min prior to experimental sessions. For each session, individual inter-response times (IRTs) were recorded to allow for trial-by-trial analysis. The average overall IRTs were also calculated. All IRT data were log transformed to improve homogeneity of variance for statistical analysis.

Prepulse inhibition and PCP-induced prepulse inhibition deficit

Prepulse inhibition is an operational measure of sensorimotor gating. Phencyclidine induces a PPI deficit, which is also observed in schizophrenic patients and which is reversed by narcoleptic compounds (Geyer et al. 2001). Thirty minutes after i.p. injections of either vehicle (saline) or 2.5 mg/kg PCP, the mice (n = 9/group) were put into the startle device. After an acclimatization time of 5 min, the animals received 12 startle stimuli (white noise, 40 milliseconds, 108 dB SPL) with an interstimulus interval of 20 seconds. Then, six blocks of the following stimuli were given in a pseudorandomized order: startle stimulus alone, startle stimuli with prepulses (interpulse interval: 100 milliseconds) of 2, 4, 8, 12 and 16 dB SPL above background noise (60 db SPL).

Statistical analysis

All reported statistical tests were performed using the program systat (SPSS Inc., version 12). For analysis of the behavioral data, analysis of variance (anova; if appropriate with repeated measures) or Student's t-tests were used. In case of a significant anova (P < 0.05), post hoc Tukey tests or least significant difference tests were carried out.


Locomotor activity and PCP-induced hyperactivity

Two different statistical analyses were carried out. First, we checked for potential effects of genotype, as well as for differences between the groups with the same genotype, before treatment. For this analysis, we used the activity means of the whole 60 min before treatment and the means of the first and the last 5-min blocks within this period to detect potential effects on the time–course. Second, we analyzed the effect of PCP treatment within the different groups by using the individual means of the activity in the 60 min before PCP and the 60 min after PCP injection.

Figure 1 depicts the mean activity of the different animal groups measured in traveled distance. In the pretreatment period, there were no main effects of genotype and no differences between the two groups with the same phenotype could be detected. Similarly, no genotype effects and group differences could be detected when the first and the last 5-min blocks of the pretreatment period were analyzed. However, there was a strong effect of the factor time (F1,32 = 383.41,P < 0.001), reflecting habituation of locomotor activity but no interactions between time and genotype or prospective treatment.

Figure 1.

Locomotor activity and PCP-induced hyperactivity. Depicted is the mean traveled distance in meters (±SEM) in an open field during the habituation period (min 0–60), and after i.p. injections of PCP or vehicle (arrow, min 60–120). No effects of mGlu8 deficiency were observed.

Phencyclidine injections clearly increased locomotor activity (treatment: F1,32 = 14.84, P = 0.001; treatment × time: F1,32 = 10.88, P = 0.002). However, this PCP effect was not different in the two tested genotypes.

Elevated plus maze

Figure 2 (upper left panel) depicts the open arm ratio for time, entries and traveled distance (n = 8−10/group). In two independent experiments, these three measures were not significantly affected by the genotype of the animals. Interestingly, the number of entries on the open arm in the first experiment was significantly decreased in mGlu8-deficient animals (t14 = 2.38, P = 0.03; data not shown). However, the total number of arm entries in this experiment also decreased (t14 = 2.57, P = 0.02; data not shown), indicating that the reduced number of open arm entries was not an anxiogenic-like effect. In the second experiment, these two measures were not affected by genotype. In addition, the latency of the first open arm entry, the mean and maximal velocity as well as the total distance traveled were all not significantly affected by genotype in the two experiments (data not shown).

Figure 2.

Elevated plus maze, elevated zero maze and light-dark box. Elevated plus maze (upper left panel): shown is the mean open arm ratio (±SEM) for time, entries and traveled distance of two independent experiments. No effects of mGlu8 deficiency were observed. Elevated zero maze (upper right panel): depicted is the ratio for time spent in the open arm and the line crossings (means ± SEM). mGlu8-deficient animals showed no differences to wild-type littermates. Light-dark box (lower panel): shown are the latency to enter the dark compartment (measured in seconds), the number of transitions between the compartments, the ratios for time spent in the lit compartment and for the distance traveled in the lit compartment (means ± SEM). In two independent experiments, no main effects of mGlu8 deficiency were observed. Only the ratio for traveled distance in the lit compartment was increased in one experiment (**P < 0.01, Student's t-test).

Elevated zero maze

Figure 2 (upper right panel) shows behavioral data for the mice on the elevated zero maze. The ratio of time spent on the open quadrants, as well as crossings between the quarters is shown. Both measures were not affected by the genotype of the animals.

Light-dark box

In Fig. 2 (lower panel), the behavior of the animals in the two different experiments is shown. Three of the standard measures, latency to enter the dark compartment, transitions between the compartments and percentage of time spent in the lit compartment, were not affected by the genotype of the animals in the two experiments. However, the percentage of distance traveled in the lit compartment was significantly increased in the mGlu8-deficient mice in the second experiment (t16 = 2.95, P = 0.009), but not in the first experiment. In parallel, the total distance traveled (data not shown) in the mGlu8-deficient mice was also facilitated in the second experiment (t16 = 25.93,P < 0.001), but not in the first experiment. Other measures of locomotor activity (maximal or mean velocity) were not affected in both experiments (data not shown).

Conditioned freezing

In the first experiment, our standard protocol of one conditioning and three retention sessions was used in adult mice (Fig. 3). During the habituation phase of the fear conditioned session (pre-CS), no freezing behavior was measured for both genotypes. An anova with the freezing response during the CS presentations in the fear conditioning session as the dependent variable, the trial number as within subject factor and the genotype as between subject factor showed successful fear conditioning (trial number: F5,90 = 10.01,P < 0.001), but no differences between the two genotypes. Very similar results were obtained with the analyses of freezing behavior during the three retention sessions on cued fear. Now, freezing behavior could be measured in the pre-CS period and beginning with second retention test, also second-order fear conditioning to this context could be observed (F1,36 = 11.76,P < 0.001). However, the two genotypes were not different in pre-CS freezing. Regarding freezing behavior during CS presentations, there was an effect of the factor trial number beginning with the second retention test (F's > 4.82, P's < 0.001), but neither effects of genotype nor of the interaction genotype × trial number.

Figure 3.

Conditioned fear (freezing). The mean percent time spent freezing (±SEM) in the habituation phases (pre-CS), during the fear conditioning phase (D1, day 1) and during the different retention tests (D2–D4, days 2–4) are depicted. No effects are observed in the freezing responses during the pre-CS periods and during presentations of the conditioned stimulus (cued fear), but mGlu8-deficient mice had a significant decrease in the freezing response to the conditioning context (contextual fear). **P < 0.01, *P < 0.05, post hoc pairwise comparisons after a main effect in anova.

In the tests on contextual fear, significantly attenuated contextual fear was observed in the mGlu8-deficient mice (genotype: F1,15 = 6.23, P = 0.03). Furthermore, a significant effect of test day was found (F2,30 = 10.63,P < 0.001) indicating between-session extinction, and there was a significant interaction between genotype and test day (F2,30 = 3.76, P = 0.04). Post hoc pairwise comparisons showed significant contextual fear deficits in mGlu8-deficient mice on the first two test days (t's > 2.14, P's < 0.05) but not on the last test day.

In the second experiment, we tested if the contextual fear deficit observed in the first experiment is age dependent. Therefore, we compared the conditioned fear behavior in groups of 7, 8–9 and 12–13 weeks old wild-type and mGlu8-deficient mice. Again, we found no influence of the genotype on the freezing behavior during the conditioning session. Similarly, age had no influence. Cued fear in the retention test (Fig. 4) was neither affected by genotype nor by age. In contextual fear, the analysis showed an age-dependent increase of contextual fear (F2,56 = 16.30,P < 0.001) and again a deficit in contextual fear of mGlu8-deficient mice (F1,56 = 9.27, P = 0.004). A significant interaction between genotype and age (F2,56 = 3.34, P = 0.04) indicated that the contextual fear deficit in mGlu8-deficient mice was age dependent. Post hoc pairwise comparisons showed no genotype effects at an age of 7 weeks, a tendency at 8–9 weeks (t19 = 1.98, P = 0.06) and a significant difference at 12–13 weeks (t19 = 3.08, P = 0.006).

Figure 4.

Conditioned fear (effects of animal age and US intensity). In independent experiments, different batches of mice with different ages were tested for conditioned fear. Furthermore, different foot shock intensities were used in different experiments. Depicted is the mean freezing response (±SEM) during exposure to the conditioned stimulus and the conditioning context, as well as the mean activity (distance traveled) during foot shock administration. These experiments confirm the contextual fear deficit of mGlu8-deficient mice and show that this deficit can only be observed in animals older than 7 weeks and with low and moderate foot shock intensities. **P < 0.01, post hoc pairwise comparisons after a main effect in anova.

The third freezing experiment tested whether the contextual freezing deficit observed in the previous experiments is dependent on the intensity of the foot shock US. In Fig. 4, the contextual freezing after conditioning with different US intensities and the direct response of the animals to the foot shock (measured by the mean distance traveled during the foot shock period) were tested. The 12/13 weeks old group from the previous experiment receiving 0.6-mA foot shocks during the training was added to this analysis. As in the previous experiments, the genotype of the animals significantly affected contextual freezing (F1,77 = 17.32,P < 0.001). Furthermore, contextual freezing increased with higher US intensities (F2,77 = 20.77,P < 0.001) but there was no interaction between genotype and US intensity. Analysis of the locomotor response to the foot shocks showed no effects of genotype but of US intensity (F2,77 = 245.15,P < 0.001). There was no interaction between genotype and US intensity.

The fourth freezing experiment investigated the effects of a single traumatic foot shock on the behavior of wild-type and mGlu8-deficient mice. On day 1, the animals were exposed to a single foot shock (Fig. 5). There was a main effect of test phase (pre vs. post shock period: F1,26 = 78.18,P < 0.001), but neither an effect of genotype nor an interaction between genotype and test phase. A very similar result was obtained after analyzing the tone test on day 28 (a measure of sensitization). There was an effect of the tone (pre vs. after tone period: F1,26 = 18.83,P < 0.001) but not of genotype. On day 29, the genotype of the animals had no effect neither on the freezing response in the test environment used for testing context discrimination nor on the freezing response in the conditioning boxes. For the acoustic startle response, a further measure of sensitization, there were no genotype effects but a main effect of stimulus intensity (F3,78 = 25.42,P < 0.001).

Figure 5.

Behavioral changes after exposure to one traumatic foot shock. Depicted is the mean freezing response during the foot shock exposure session (trauma exposure), as well as—after 28 days—to a neutral tone in a novel context (tone test), to a context very similar to the conditioning context (context discrimination) and to the conditioning context (conditioned fear). Furthermore, the mean startle responses to acoustic startle stimuli with different intensities (50–120 dB SPL) are shown. No behavioral changes in this experiment were observed in mGlu8-deficient mice.

Object recognition

The initial test of general attractiveness of the objects used showed that there was no difference between the time spent exploring the objects, as well as no genotype differences (Fig. 6, calibration). When exposing the animals to the same objects 3 h later, exploration times for the objects were reduced for all objects and both genotypes (Fig. 6, second run). No object was found to be preferred in comparison to another. When exposing the animals a further 3 h later to two familiar objects and two novel objects (Fig. 6, familiar vs. novel objects), the novel objects were recognized as novel by both genotypes (factor objects: F's < 5.43, P's < 0.03). However, there was a difference in exploration time between wild-type and mGlu8-deficient mice (factor genotype: F3,36 = 25.52,P < 0.001), and the interaction between time (first vs. third run) and object (familiar vs. novel) was significant in wild-type animals (F3,36 = 5.28, P = 0.004) but not in mGlu8-deficient mice. This is supported by post hoc two-tailed t-tests showing a difference between genotypes for the familiar objects 1 and 2 (t's > 2.18, P's < 0.04), but not for the novel objects 5 and 6.

Figure 6.

Object recognition. Depicted is the mean exploration time (measured in seconds ± SEM) for the different objects in the different test sessions. Objects 1–4 are presented first in the calibration session and are familiar objects in the second or third run, respectively. Objects 5 + 6 are novel objects in the third run (familiar vs. novel). *P < 0.05, post hoc pairwise comparisons after a main effect in anova.

Chlordiazepoxide-induced facilitation of extinction following operant conditioning

Figure 7 shows the final performance on the discrete-trial FR5 schedule of food reinforcement and the effects of extinction and drug administration. The data plotted are mean log IRTs for each group for each session. Note that as extinction progresses, this measure will increase until it reaches a maximum value when no responses occur during a session.

Figure 7.

Operant learning (CDP-induced facilitation of extinction). Shown are the mean scores (log-transformed individual inter-response times) at the end of acquisition and over 25 extinction sessions for wild-type and mGlu8-deficient mice, with and without CDP treatment. The dotted line represents the change from training phase (with reward) to extinction phase (without reward). CDP-induced facilitation was significantly reduced in mGlu8-deficient mice. *P < 0.05, post hoc pairwise comparisons after a main effect in anova.

As evident in Fig. 7, there were no differences between groups during final acquisition sessions, when behavior was stable, or during the first eight extinction sessions, i.e. there is no initial effect of CDP. Thereafter, wild-type mice given CDP before each extinction session slowed down relative to the other three groups; from extinction session 15, knockout mice given CDP slowed down relative to the two saline groups, but not as much as the wild-type CDP group; and from extinction session 21 mice given saline also slowed down. These observations are supported by statistical analyses. In a two-way repeated measures anova for all 25 extinction sessions, there was a main effect of extinction sessions (F24,432 = 19.14,P < 0.001), a main effect of drug treatment (F3,18 = 9.36, P = 0.01) and an interaction between extinction sessions and drug treatment (F72,432 = 1.93,P < 0.05). Pairwise comparisons showed that the mean value for the wild-type group given CDP injections before each extinction session was greater than that for each of the other three groups (P < 0.001 for saline groups, P < 0.05 for mGlu8–/– CDP group, by a least significant difference test). The CDP-treated mGlu8-deficient mice had a higher mean score than the two saline groups (P < 0.05 in each case).

Prepulse inhibition and PCP-induced prepulse inhibition deficit

Figure 8 depicts the mean baseline startle magnitude (first 12 startle trials) and the mean PPI after vehicle and PCP treatment. Analysis of the baseline startle amplitudes showed a significant effect of the genotype (F1,32 = 5.91, P = 0.02) but not of PCP treatment. There was no interaction between treatment and genotype.

Figure 8.

Prepulse inhibition and PCP-induced prepulse inhibition deficit. Shown are the mean startle magnitude and the mean prepulse inhibition with and without PCP treatment (±SEM). mGlu8-deficient animals have a general decrease in startle magnitude, but the percent prepulse inhibition and the prepulse inhibition deficit induced by PCP were not changed. *P < 0.05, group effect in anova.

As expected, PPI was decreased by PCP treatment (F1,32 = 4.06, P = 0.05) but was not affected by the genotype of the animals. Again, there was no interaction of genotype and treatment. In addition, intensity of the prepulse had a significant effect on PPI (F4,128 = 59.60,P < 0.001). All interactions with prepulse intensity failed to reach a significant level.


The present study identifies very specific deficits of mGlu8-deficient mice in contextual fear, object recognition, CDP-induced facilitation of extinction and the acoustic startle response; none of which were reported before. At the same time, the behavior of mGlu8-deficient mice in animal models of anxiety and post-traumatic stress disorder (PTSD) (elevated plus and zero maze, light-dark box, fear conditioning/sensitization after traumatic US), in animal models for schizophrenia (PCP-induced hyperactivity and PPI deficits) and in different models of learning (cued fear conditioning, operant learning) was not affected by mGlu8deficiency. The present experiments also showed normal locomotor activity and normal sensory capabilities in mGlu8-deficient mice. Some of these findings stand in contrast to published data.

The latter findings are of great importance, because changes in general locomotor activity complicate the interpretation of behavioral experiments that are dependent on locomotor activity. However, in several independent experiments with different batches of mice, we found no changes in locomotor activity, neither in an open-field experiment focused on detecting such changes nor in the habituation period of the conditioned fear experiments, or of the startle experiments or in the rate of response during operant conditioning. In line with these data, total activity was not affected in the traditional, unconditioned anxiety models, except for one elevated plus-maze session in which mGlu8-deficient mice had an enhanced total number of arm entries. However, this difference was not observed in subsequent sessions, and the total distance traveled was not changed in this experiment. The observation that locomotor activity is not affected by mGlu8 deficiency stands in contrast to studies reporting changes of locomotor activity in mGlu8-deficient mice in both directions (hypoactive: Duvoisin et al. 2005; hyperactive: Gerlai et al. 2002), but supports other studies with similar observations (Linden et al. 2002; Robbins et al. 2007). These discrepancies may arise from the different genetic backgrounds of the mGlu8-deficient animals used (C57BL/6 vs. ICR) or different ages of animals in the studies (between 12 and 24 weeks). Whatever the reason, we can safely conclude that in the experiments in the present study the effects observed are not because of changes in locomotor activity.

In addition, there was no evidence of sensory deficits in mGlu8-deficient mice in the present experiments. The acoustic cue for fear conditioning, as well as the weak acoustic prepulse used in the PPI experiment, had very similar effects in wild-type and mGlu8-deficient mice. Furthermore, different contexts were readily discriminated, and the response to foot shocks was not affected. These findings are in line with previous studies where sensory impairment was only described in the mouse strains used for backcrossing but not in the mGlu8-deficient mice (Duvoisin et al. 2005; Gerlai et al. 2002; Linden et al. 2002; Robbins et al. 2007).

A surprising result of the present study was that mGlu8-deficient mice expressed no anxiogenic phenotype in unconditioned behavioral models of anxiety, including elevated plus maze, elevated zero maze and light-dark box, in contrast to reports in several publications (Duvoisin et al. 2005; Linden et al. 2002, 2003; Robbins et al. 2007), but see also Gerlai et al. 2002. We carried out five independent experiments, all of which showed no effect of genotype. The effect of CDP on operant extinction was partially reversed, but we have shown elsewhere that this effect of CDP is not because of its anxiolytic action (McCabe et al. 2004). It is difficult to explain the discrepancy from the literature with traditional, unconditioned anxiety models. However, one possible explanation is that we always used naive animals in our experiments (except for practical reasons in one light-dark box experiment). This means that we can exclude effects of prior experience in our studies, whereas this is not the case in some of the published experiments (e.g. Duvoisin et al. 2005). It is known that prior test experience affects unconditioned anxiety-like behavior (Carobrez & Bertoglio 2005; Holmes et al. 2001) and this effect could be modulated by mGlu8. Other factors that can influence anxiety behavior are the housing conditions or the laboratory environment (Crabbe et al. 1999; Whitaker et al. 2009). Our animals were housed in environmentally enriched cages which can decrease anxiety behavior (Friske & Gammie 2005). Furthermore, in some earlier studies animals were used which had been backcrossed with ICR mice (Gerlai et al. 2002; Linden et al. 2002, 2003). ICR mice suffer from deficits in vision (Gerlai et al. 2002) and mGlu8 deficiency could potentially increase this deficit. Last, Duvoisin et al. (2005) tested relatively old animals (ca. 6 months) in the published study, whereas we used mice in an age of 2–3 months.

Perhaps more surprising than the lack of anxiogenic-like phenotype in unconditioned tests of anxiety was to find a contextual fear deficit in mGlu8-deficient animals. Conditioned fear tests are believed to be better predictors of clinical efficacy than the tests of innate or unconditioned anxiety (Garakani et al. 2006; Grillon 2002). In several experiments, we found a deficit in the expression of contextual conditioned fear but not of cued fear. The freezing response in the fear conditioning phase was not affected. In a previous study, Gerlai et al. (2002) described a delayed fear response in mGlu8-deficient mice in the fear conditioning phase. However, the maximal freezing responses reported in that study were about 5–15%, indicating a severe deficit in freezing behavior in the ICR strain which was used for backcrossing. C57BL/6 mice (the genetic background of our animals) show a normal freezing response (50–70%).

Our data also showed that the observed contextual fear deficit is a very specific deficit and not indirectly induced. First, a normal freezing response was observed during presentations of the CS showing (1) that the ability to express a freezing response is not affected by mGlu8 deficiency and (2) that there are no general (fear) learning deficits. Second, the ability to discriminate two very similar contexts was not affected in mGlu8-deficient animals, showing that these animals have normal processing of contextual information. However, the contextual fear deficit observed in the present study was not present in animals younger than 2 months or with foot shocks of very high and traumatic intensity. This can be explained by the hypothesis that for the processing of high intensity and traumatic foot shocks other and/or additional neural structures are used (e.g. Walker & Davis 1997), which are not affected by mGlu8 deficiency. In addition, a recently published electrophysiological study indicated that mGlu8 in the hippocampus, a brain region of high importance for contextual fear (Maren 2001), might play different roles during development. It should be noted that mGlu8 is not only densely localized in the hippocampus but also in other brain regions believed to be involved in contextual fear conditioning, such as the amygdala and the septum (Calandreau et al. 2007; Maren 2001).

The present study also detected a mild deficit in novel object recognition, which is considered to be a measure of declarative memory. The mGlu8-deficient mice were able to differentiate the novel objects from the familiar ones, but this effect was much less pronounced than in wild-type animals. This finding is in contrast to Duvoisin et al. (2005) showing no deficits in novel object recognition. However, we used a more complex experimental protocol with four objects, which makes higher demands on memory systems and is probably more sensitive in detecting deficits.

Overall, mGlu8-deficient mice did not express a general learning deficit. The phenotype of mGlu8-deficient animals in fear conditioning to a discrete cue and operant conditioning was normal. However, in that experiment, the facilitation of extinction by the benzodiazepine agonist CDP was also investigated. Our present findings replicate and extend those of previous studies of the effect of CDP (Leslie et al. 2004, 2005; McCabe et al. 2004; Shaw et al. 2004). They show that the effect of CDP in facilitating extinction is further delayed in mGlu8-deficient mice, suggesting that GABAergic processes are diminished in these mice. Metabotropic glutamate receptor subtype 8 is localized on GABAergic terminals (Ferraguti et al. 2005; Shigemoto et al. 1997) and the lack of mGlu8 may lead to an alteration in inhibitory neurotransmission function that is manifest in behavioral changes seen here.

Lastly, we also found a slight reduction in the acoustic startle response, which was not because of a hearing deficit or to lower body weight in mGlu8-deficient animals. However, the startle response is also an indicator of anxiety state (Koch 1999; Koch & Fendt 2003; Lang et al. 1990; Ray et al. 2009; Yilmazer-Hanke et al. 2004). This finding would thus be consistent with an anxiolytic-like phenotype in contextual fear.

Taken together, the present series of experiments detected specific behavioral changes in mGlu8-deficient animals, which point to an important role of mGlu8 in the processing of contextual fear and object recognition. Furthermore, our experiments indicate a role of mGlu8 in the modulation of GABA release. Metabotropic glutamate receptor subtype 8 is located in brain areas processing contextual fear, extinction and object recognition, including the amygdala, hippocampus, the prefrontal and the perirhinal cortex. Both glutamate and GABA as well as the metabotropic glutamate receptors play important roles within these brain sites (Ferraguti et al. 2005; Palazzo et al. 2008; Schmid & Fendt 2006; Winters et al. 2008). Therefore, mGlu8 might be a potential target for disorders with pathophysiological changes in brain areas where mGlu8 modulates glutamate and GABA transmission. The present data especially point to anxiety disorders with extreme contextual fear, such as generalized anxiety disorders (Grillon 2002), and to diseases with disturbed declarative memory.


We are very grateful to Robert Duvoisin, Oregon Health & Science University, USA, for providing mGlu8-deficient mice (NIH EY09534), to Carsten Wotjak and Yulia Golub, Max Planck Institute of Psychiatry in Munich, Germany, for helpful tips and discussions and to Laura Jacobson, Novartis Institutes for BioMedical Reasearch, for very helpful comments to the manuscript. Potential conflict of interest: some authors are employees of Novartis Pharma AG, which is generally interested in developing compounds for CNS disorders.