C. T. Wotjak, Max Planck Institute of Psychiatry, Kraepelinstrasse 2, D-80804 Munich, Germany. E-mail:firstname.lastname@example.org
To investigate the impact of averseness, controllability and familiarity of a test situation on the involvement of the endocannabinoid system in the regulation of exploratory behaviour, we tested conventional and conditional cannabinoid receptor type 1 (CB1)-deficient mice in behavioural paradigms with different emotional load, which depended on the strength of illumination and the ability of the animals to avoid the light stimulus. Complete CB1 null-mutant mice (Total-CB1-KO) showed an anxiogenic-like phenotype under circumstances where they were able to avoid the bright light such as the elevated plus-maze and the light/dark avoidance task. Conditional mutant mice lacking CB1 expression specifically in cortical glutamatergic neurons (Glu-CB1-KO), in contrast, failed to show a similar phenotype under the same experimental conditions. However, both mutant lines showed increased avoidance of open arm exploration during a second exposure to the elevated plus-maze. If tested in situations where the fear eliciting light could not be avoided, Total-CB1-KO mice showed increased thigmotaxis in an open field, decreased social investigation and decreased novel object exploration under aversive light conditions, but not under non-aversive low light. This time, Glu-CB1-KO also showed decreased exploratory behaviour towards objects and conspecific juveniles and increased thigmotaxis in the open field. Taking into consideration that the behavioural performance of wild-type mice was only marginally affected by changes in light intensities, these data indicate that the endocannabinoid system renders exploratory behaviour largely independent of the test averseness. This process differentially involves endocannabinoid-controlled glutamatergic transmission, depending on the controllability of the test situation.
The behavioural performance of rodents in a novel environment is determined by two opposing motivations, the fear of predators and/or physical harm on one hand and the innate drive to explore novel situations on the other hand. The ability of anxiolytic compounds to shift the balance of these two motivations towards the side of exploration led to the development of a series of anxiety tests, in which exploration of open and illuminated areas (in case of the elevated plus-maze test, open field test and light/dark emergence tests; Hascoet et al. 2001; Milner & Crabbe 2008; Rodgers & Dalvi 1997) or the investigation of conspecifics (in case of social investigation tasks; File & Seth 2003; Haller et al. 2004; Kalueff et al. 2007) or novel objects (in case of object investigation; Kalueff et al. 2007; Wolfer et al. 2004) are assessed as measures of anxiety. The aversiveness of the test situation is typically modified by changing the light conditions because rats and mice, as nocturnal animals, avoid brightly lit compartments. However, depending on the task, this aversive stimulus can be avoided by the animals (e.g. in case of the elevated plus-maze or the light/dark task) or not (e.g. in case of the open field, social investigation or novel object investigation tasks). An unavoidable light stimulus may lead to a series of fear reactions (including freezing, changes in autonomous nervous system and hormonal stress responses) that likely interfere with approach/avoidance behaviour (i.e. anxiety-related behaviour) shown in that particular situation (McNaughton & Corr 2004). Consequently, perception of a given test situation (i.e. its emotional load) depends on its aversiveness, controllability and familiarity.
This conceptual frame work has been supported by a recent study, which showed a higher activity of the hormonal stress response upon exposure to a novel cage (where the animals cannot escape from this situation) compared with exposure to an elevated plus-maze [where the animals can escape from this aversive situation by hiding in the closed arms; (Thoeringer et al. 2007)]. Accordingly, DBA/2 mice, showing twice as much anxiety-related behaviour on the elevated plus-maze compared with C57BL/6J mice, failed to show higher corticosterone levels (Thoeringer et al. 2007). Therefore, the ability to control a particular situation vs. the inability to do so may have important implications in terms of emotional load and brain structures/neurotransmitter systems involved in anxiety-related behaviour (McNaughton & Corr 2004).
We recently showed that, after conditioning or sensitization with inescapable electric footshocks, endocannabinoids control fear adaptation to tones via the cannabinoid receptor type 1 (CB1) (Kamprath et al. 2006, 2009). The fear-alleviating effects of CB1 became evident in highly aversive situations only and seemed to primarily involve the control of cortical glutamatergic transmission (Kamprath et al. 2009). One could assume that similar processes account for behavioural adaptation to bright light, in particular, if the aversive light cannot be avoided by the animals. Accordingly, a number of studies reported increased anxiety in CB1-deficient mice in social investigation (Haller et al. 2004; Uriguen et al. 2004), exploration of the open arms of an elevated plus-maze (Haller et al. 2002, 2004; Uriguen et al. 2004) or of the lit compartment of a light/dark box (Martin et al. 2002; Uriguen et al. 2004). However, there are also reports that failed to obtain any effect of impaired CB1 signalling on anxiety-related behaviour on the elevated plus-maze in particular during low light conditions (Marsicano et al. 2002; Thiemann et al. 2009). Others even showed anxiolytic-like effects in the shock-probe burying test (Degroot & Nomikos 2004), which, in contrast to the previously mentioned exploration-based tasks, measures active coping strategies. Taken together, evidence is accumulating that the endocannabinoid system controls fear and anxiety primarily under highly aversive conditions (Haller et al. 2004; Haller et al. 2009; Naidu et al. 2007; Kamprath et al. 2009; Thiemann et al. 2009).
The current study was designed to systematically investigate the role of CB1 in exploration-based tests with low vs. high emotional load modified by the light intensity, the controllability of light exposure and familiarity with the test procedure. Experiments were performed with mice lacking CB1 expression in the entire body (Total-CB1-KO; Marsicano et al. 2002) or specifically in glutamatergic cortical neurons (CB 1f/f;Nex−Cre = Glu-CB1-KO; Monory et al. 2006). Because the endocannabinoid system was shown to mediate habituation to homotypic stressors (Patel et al. 2005; Patel & Hillard 2008), including adaptation to aversive tones (Kamprath et al. 2006), and to play a particular role during re-exposure to exploration-based anxiety tests, such as the elevated plus-maze test (Rodgers et al. 2005) and the open field test (Thiemann et al. 2007), we additionally studied the consequences of repeated testing (i.e. familiarity) for test situations with avoidable (elevated plus-maze) or unavoidable (open field) light exposure.
Material and methods
All experimental procedures had been approved by the Committee on Animal Health and Care of the State of Bavaria (Regierung von Oberbayern, Munich, Germany) and were performed in strict compliance with the European Economic Community (EEC) recommendations for the care and use of laboratory animals [European Communities Council Directive of 24th November 1986 (86/609/EEC)].
Male mice at the age of 8–16 weeks were used throughout the experiments. Mutant mice (KO) and their respective wild-type littermate controls (WT) were generated and genotyped as described previously (Total-CB1: Marsicano et al. 2002; Glu-CB1: Monory et al. 2006). Mice of the two lines were backcrossed to C57BL/6NCrl for six generations in-house. Total-CB1-KO and Total-CB1-WT were derived from heterozygous breeding pairs. Conditional mutants were maintained by breeding pairs with the genotypes Cre (−);CB1fl/fl (i.e. wild-type) for the mother and Cre(+);CB1fl/fl (i.e. knockout) for the father to avoid effects of the genetic modification on maternal care. Glu-CB1-KO lack expression of CB1 specifically in cortical glutamatergic neurons including neocortex, hippocampus and basolateral amygdala complex (Monory et al. 2006). For the experiments, the offspring of interest was taken from different breeding pairs with respect to closely matching birth dates (the maximum variation among birth dates was 6 weeks). Equal numbers of male mutant and wild-type offspring occurred rarely in one litter, thus breeding pairs usually did not contribute equal numbers of mutant and wild-type animals to the experimental groups. However, care was taken that at least one wild-type littermate was tested together with each mutant mouse and vice versa.
Mice were maintained under standard conditions with food (Altromin Standard Diet 1310, Altromin GmbH, Germany) and tap water ad libitum in Makrolon type II cages (27 × 16 × 12 cm3) with sawdust bedding (Altromin Faser Einstreu, Altromin GmbH, Germany), at 22 ± 2°C room temperature and 55 ± 5% humidity. Mice were kept in an inverse light/dark cycle with lights off at 0900 h and lights on at 2100 h. To avoid confounding influences of social status, subjects were kept individually in cages for 2 weeks before starting with the experiments. Only in case of social interaction experiments, mice were kept in groups of three to five mice per cage to reduce aggressive and sexual-oriented behaviour in the subsequent confrontation with the juveniles (male C57BL/6NCrl mice purchased from Charles River at an age of 19–23 days and kept in groups of five per cage). Mice were transported to the experimental room 24 h before testing and tested during the dark phase of the light/dark cycle between 1000 and 1400 h. A total of 83 Total-CB1-KO, 83 Total-CB1-WT, 65 Glu-CB1-KO and 68 Glu-CB1-WT were used in this study. The sample sizes used for the different experiments are indicated in the respective figures.
Setup and procedure were essentially the same as described before (Kromer et al. 2005; Touma et al. 2008). The elevated plus-maze was made of dark-grey plastic and consisted of two open arms (30 × 5 cm), and two enclosed arms (30 × 5 × 15 cm). A neutral zone of 5 × 5 cm interconnected all four arms, making up the shape of a plus sign. The open arms had a rim with a height of 0.5 cm to facilitate the grip on the open arms. The elevated plus-maze was located 120 cm above the floor. The setup was illuminated by two halogen lights (white light) spotting onto the end of the open arms, thus resulting in a light intensity of 300 lux (lx) at the end of the open arms and of 70 lx at the starting point of the open arms. The light intensity in the closed arms was 50 lx. The plus-maze was surrounded by a black curtain. A CCD camera (Conrad Electronics, Munich, Germany) was mounted over the plus-maze to observe and videotape the behaviour of the mouse without disturbing the animals. At the beginning of the experiment, each mouse was placed on the central platform facing a closed arm and was then allowed to freely explore the maze for 5 min. Mice were exposed to the maze for a second trial 24 h later. After each test, the plus-maze was cleaned with water containing detergent and dried with tissue. Behaviours scored from videotape included conventional and ethological plus-maze variables. Conventional spatio-temporal measures were as follows: number of arm entries (open arms, closed arms and open + closed = total arms, with the criterion for an arm entry defined as all four paws in an arm of the maze), percent open arm entries (expressed as number of open arm entries/total arm entries × 100) and percent time spent on the open arms (expressed as time spent on open arms/time spent in open and closed arms × 100). Ethological variables comprised the frequency of closed arm returns (i.e. exiting the closed arm with the forepaws only and then returning into the same arm), number of head dipping (i.e. scanning over the sides of the maze towards the floor) and number of stretch-attend postures (SAP; i.e. forward elongation of head and shoulders, followed by retraction to the original position without any forward locomotion). The latter two behaviours were further differentiated as protected (i.e. occurring from the relative security of a closed arm exit or the centre platform) or unprotected (i.e. occurring on an open arm). Furthermore, total duration scores for grooming (i.e. licking, scratching and washing of the head and body) and immobility (lack of movement) as well as the number of mice which failed to explore the open arms at all (i.e. non-explorers) were assessed. Behavioural scoring was carried out using the customized computer software program Messprogramm Fuer Plusmaze Version 3.0 (K. Scheidemann, Munich, Germany).
The plus-maze procedure was validated pharmacologically in C57BL/6N mice (i.e. the background strain of the mutant mice used in the current study) by systemic treatment with either a benzodiazepine (diazepam) or a GABA uptake inhibitor (tiagabin). If applied 30 min before testing, both diazepam (1 mg/kg) and tiagabine (2.5 mg/kg) caused a significant increase in percent open arm entries and percent open arm time compared with vehicle-treated controls without significantly changing locomotor activity (Micale, Marsch & Wotjak, unpublished observations).
Setup and test procedure were essentially the same as described before (Muller et al. 2003; Timpl et al. 1998). The light/dark box was divided into two compartments, one dark compartment (L 15 × W 20 × H 38 cm3) with black walls and one lit compartment (L 30 × W 20 × H 38 cm3) with white walls made of plastic. The two compartments were connected by a 4-cm long tunnel. Light intensity was 700 lx in the lit compartment and 15 lx in the dark compartment, measured at floor level. A session started by placing the animal in the centre of the dark compartment facing a wall and lasted for 5 min. After each test, the light/dark box was cleaned with water containing detergent and dried with tissue. The animals' behaviour was videotaped. Entries made into the lit compartment and times spent in the lit compartment were assessed by analysing the videotapes by a trained observer blind to the genotype. These two variables were expressed as a percentage of the total observation period and the total number of light/dark transitions, respectively.
Open field test
Animals were placed in an open field box (L 26 × W 26 × H 38 cm3, Coulbourn Instruments, Allentown, PA, USA) for 30 min on two consecutive days. Total-CB1 mice were randomly assigned to one of three experimental groups that performed the task with 0, 30 or 700 lx (measured at floor level). Glu-CB1 mice were randomly assigned to two groups performing the task either with 0 lx or with 700 lx. Mice were re-exposed to the open field under identical experimental conditions 24 h later. Horizontal and vertical activities were automatically recorded by two infrared sensor rings (2 cm and 5 cm above floor level; the infrared beams of the sensor rings were spaced apart by 1.52 cm). Both sensor rings were connected via an interface to a computer equipped with the Tru Scan Software Version 1.1 (Coulbourn Instruments). The sampling rate was 4 Hz. Each box, including the sensor rings, was surrounded by an additional box made of opaque Plexiglas side walls (L 47 × W 47 × H 38 cm3) without roof and floor. At the end of the trial, mice were returned to their home cages, and test boxes were carefully cleaned with water containing detergent. The open field was subdivided in a centre and margin compartment with the margin clasping an area 4.5 cm away from the walls. Behaviours assessed included horizontal locomotion, vertical exploration (i.e. the number of rearings), immobility time and the distance moved along the walls normalized to the total distance moved (i.e. thigmotaxis).
Novel object investigation
The object investigation test was performed at 30 lx (which still allowed the assessment of exploration of the objects) or 500 lx (i.e. room light measured at level of the test cages). Experimental subjects were habituated to the test arena (a Makrolon Type III cage, 36 × 22 × 14 cm3, with sawdust bedding material and transparent walls) on 2 days for 10 min (the cage remained the same for each animal without cleaning and changing of bedding). On the third day, mice were transferred into the same test cages, and after 10 min two identical objects were placed in a symmetrical position at the short walls of the cages. Between animals, objects were thoroughly cleaned with water containing detergent to eliminate olfactory cues. Four different objects were used: (1) a cone made of aluminium, (2) a light bulb on a plastic socket, (3) a flask made from yellow plastic filled with sand and (4) a black bottle made of glass filled with water. Objects were heavy enough that a mouse could not displace them. Objects were equally distributed among knockout and wild-type mice of the two lines. The duration of the session was 10 min. A small CCD camera (Conrad Electronics) was mounted above the Makrolon cage to enable behavioural observation and leaving the animal undisturbed. The time spent in exploring each object was recorded online manually by means of a personal computer. Investigation was defined as follows: directing the nose towards the object at a distance of not more than 2 cm and/or touching the object with the nose and paws. Sitting on the object was not considered as exploratory behaviour.
Group-housed experimental subjects were separated by transferring them to a new Makrolon Type II cage (27 × 16 × 12 cm3) with fresh bedding material 2 h before starting the experiment. Light intensity in the testing room was 0 lx (i.e. red light) or 500 lx (i.e. room light measured at level of the test cages). A social interaction session comprised a 4-min exposure of a juvenile conspecific in the adult's cage. During the exposure, the duration of active investigatory behaviour (mainly sniffing and licking of the anogenital region, mouth, ears, trunk and tail of the juvenile and close following of the juvenile) was recorded online manually by means of a personal computer. Aggressive and sexual behaviour (biting, riding on back) was rarely observed and, therefore, not quantified.
Behavioural experiments were conducted in four screens to reduce the number of animals used for the study with separate cohorts of animals for every screen. Care was taken to carry out the most stressful tests at the end of a screen. If not stated otherwise, the different screens were accomplished with 2–3 days in between two consecutive tests. The order of the tests was as follows: Screen 1, object investigation (30 lx) and open field test first exposure (0 lx) followed by open field test second exposure (0 lx) 24 h later; screen 2, social investigation (500 lx) and open field test first exposure (700 lx) followed by open field test second exposure (700 lx) 24 h later; screen 3, social investigation (0 lx) and object investigation (500 lx); screen 4, light/dark avoidance (not always) and elevated plus-maze first exposure followed by elevated plus-maze second exposure 24 h later. Note that all different test paradigms were performed in different laboratories.
The experimenter was blind to the genotype of the animals in all experiments carried out in this study. Data are shown as mean ± standard error of the mean (SEM). For each mouse line, the effects of genotype (Total-CB1-KO or Glu-CB1-KO vs. corresponding wild-type controls) on the behavioural variables collected by the various experimental conditions were tested for significance with either analyses of variance (anova; univariate or multivariate) or non-parametric univariate tests (χ2 tests) according to the data nature of the variables (metrical or non-metrical variables, respectively). The anova designs were generally one-factorial, but sometimes two- and more-factorial designs have been used.
The behavioural performance on the elevated plus-maze was analysed in a more detailed manner. For each mouse line, differences in the behavioural variables between the two genotypes and the two consecutive days were tested by multivariant analysis of variance (manova) with repeated measures design (exposure). The number of non-explorers at each exposure was compared between the genotypes by χ2 tests.
To summarize the genotype differences observed in the various locomotion- and anxiety-related parameters, we normalized the behavioural performance of every mutant mouse to the mean behavioural performance of the respective wild-type controls according to the formula y = [KO − mean (WT )]/mean (WT ). Statistically significant deviations from WT were confirmed by one-sample t-tests with a theoretical mean of 0.
We accepted α = 0.05 as nominal level of significance. All post hoc tests by the parametric or non-parametric inferential statistics were performed at a Bonferroni-corrected level of significance to keep the type I error ≤0.05.
Total-CB1-KO, but not Glu-CB1-KO, showed increased risk assessment during the first exposure to the elevated plus-maze
Although Total-CB1-KO spent less time on the open arms and made fewer entries into the open arms of the elevated plus-maze, compared with Total-CB1-WT (Fig. 1a), the differences between the two genotypes in these behavioural characteristics did not reach statistical significance by the performed analysis of variance. At the same time, Total-CB1-KO showed significantly more stretch-attend postures [F(1,40) = 8.64,P = 0.006; Fig. 1b] but almost equal number of closed arm entries [F(1,40) = 1.44,P = 0.238; Fig. 1c], thus indicating an increase in anxiety-related behaviour in Total-CB1-KO in the absence of alterations in locomotion. Glu-CB1-KO, in contrast, failed to significantly differ from their wild-type controls in any of these variables (Fig. 1d–f).
Both Total-CB1-KO and Glu-CB1-KO showed increased avoidance of open arm exploration during the second exposure to the elevated plus-maze
In order to assess the influence of the endocannabinoid system on repeated exposure to test situations, where the aversive stimuli can be avoided, Total-CB1 and Glu-CB1 mice were tested in a second trial on the elevated plus-maze. manovas for all the conventional and ethological parameters showed significant genotype effects [Total-CB1: F(1,40) = 4.12,P = 0.049; Glu-CB1: F(1,46) = 8.69,P = 0.005] as well as significant changes from the first to the second exposure (P < 0.0001) for both Total-CB1 (Table 1) and Glu-CB1 mice (Table 2). In no case was there a significant Genotype × Exposure interaction. Tables 1 and 2 display the significant genotype differences of the individual variables. In general, Total-CB1-KO and Glu-CB1-KO showed decreased exploration of the open arms compared with the respective wild-type controls, as indicated by decreased time spent on open arms and reduced number of open arm entries and head dips. The number of non-explorers was dramatically changed from the first to the second exposure (Tables 1 and 2). During the first exposure, about the same number of Total-CB1-KO and Total-CB1-WT (χ2 = 0.67,P = 0.41), and Glu-CB1-KO and Glu-CB1-WT (χ2 = 0.69,P = 0.41), were non-explorers. During the second exposure, however, significantly more Total-CB1-KO (χ2 = 4.8,P < 0.05) and Glu-CB1-KO (χ2 = 6.1,P < 0.01) avoided the open arms completely, as compared with their respective wild-type controls (Tables 1 and 2).
Table 1. Statistical analysis of the behavioural performance of plus-maze-naï ve and plus-maze-experienced Total-CB1 mice
NS, not significant; SAP, stretch-attend postures (i.e. risk assessment).
Variables were analysed in two sets by manovas with repeated measures designs. manovas showed in each case significant effects for genotype (between-subjects-factor; P < 0.05) and exposure (within-subjects-factor; P < 0.001) but no significant Genotype × Exposure interactions. Here, we show the results of the subsequently performed univariate F-tests. Part of the data is shown in Fig. 1a–c.
*P < 0.05 compared with WT second exposure, χ2 test.
Table 2. Statistical analysis of the behavioural performance of plus-maze-naï ve and plus-maze-experienced Glu-CB1 mice
NS, not significant; SAP, stretch-attend postures (i.e. risk assessment).
Variables were analysed in two sets by manovas with repeated measures designs. manovas showed in each case significant effects for genotype (between-subjects-factor; P < 0.005) and exposure (within-subjects-factor; P < 0.001) but no significant Genotype × Exposure interactions. Here, we report the results of the subsequently performed univariate F-tests. Part of the data is shown in Fig. 1d–f.
*P < 0.05 compared with WT second exposure, χ2 test.
Total-CB1-KO, but not Glu-CB1-KO, showed increased anxiety-related behaviour in the light/dark test
Total-CB1-KO spent significantly less time in the lit compartment [F(1,22) = 9.24,P = 0.006] and entered the lit compartment less often [F(1,22) = 11.08,P = 0.003] than their wild-type controls (Fig. 2a). In contrast, Glu-CB1-KO showed no significant differences in the time spent in the lit compartment or in the entries made into the lit compartment compared with their wild-type controls (Fig. 2b). One Glu-CB1-WT and one Glu-CB1-KO mouse had been identified as outliers (in each case the mouse failed to enter the lit compartment at all) and were excluded from further statistical analyses.
Total-CB1-KO rather than Glu-CB1-KO showed differences in exploratory behaviour in an open field under low and high light intensity
Different cohorts of Total-CB1-mice and Glu-CB1-mice were exposed to an open field under low (0 lx) and high light conditions (700 lx). Exploratory behaviour was analysed in terms of horizontal locomotion (i.e. distance moved), immobility time and vertical exploration (i.e. number of rearings).
Analysis of variance with genotype and time as influential factors showed for the Total-CB1-KO and their corresponding control mice a significant interaction effect of the factors under the low light condition [Genotype × Time : F(9,252) = 5.81,P < 0.0001; Fig. 3a, left panel]. The subsequent analysis of the simple effects pointed out that Total-CB1-KO increased locomotion in the beginning of the open field exposure, compared with wild-type controls, but rapidly habituated to the environment, as shown by similar locomotion from the second observation interval on. Under 700 lx, where the influential factors showed also a significant interaction [Genotype × Time : F(9,225) = 2.99,P = 0.002; Fig. 3b, left panel], Total-CB1-KO showed the same locomotion as controls in the beginning of the exposure, but a more pronounced habituation (univariate F-tests of simple effects in manova, P < 0.05).
Glu-CB1-KO showed slightly more locomotion at the beginning and a slightly more pronounced habituation over the course of the exposure at 0 lx [Genotype × Time : F(9,198) = 2.32,P = 0.017; Fig. 3a, right panel]. In contrast, there were no significant differences in locomotion between Glu-CB1-KO and respective wild-type controls at 700 lx (Fig. 3b, right panel).
Immobility time largely mirrored the data obtained regarding locomotion. Comparisons of Total-CB1-KO and Total-CB1-WT showed significant interaction effects in the manovas under both 0 lx [Genotype × Time : F(9,252) = 3.27,P = 0.001] and 700 lx [Genotype × Time : F(9,225) = 4.03,P < 0.001]. Total-CB1-KO showed less immobility than Total-CB1-WT under 0 lx, in particular, at the beginning of the exposure (Fig. 3c, left panel). Under 700 lx, in contrast, Total-CB1-KO showed more immobility, in particular, towards the end of the exposure (Fig. 3d, right panel).
Glu-CB1-KO showed no statistically significant differences in immobility time as compared with Glu-CB1-WT either under 0 lx (Fig. 3c, right panel) or under 700 lx (Fig. 3d, right panel).
Even though the Genotype × Time interaction has been found to be significant in the low light condition [Genotype × Time : F(9,252) = 2.76,P = 0.004; Fig. 3e, left panel], Total-CB1-KO did not differ significantly at the single 3-min bins from their wild-type controls in the development of the number of rearings at 0 lx. At 700 lx, however, Total-CB1-KO showed significantly less rearing than wild-type controls towards the end of the exposure (univariate F-tests of simple effects in manova, P < 0.05). In the low light condition, Total-CB1-WT, but not Total-CB1-KO, showed an increase in the number of rearings over the course of the exposure [Time: F(9,117) = 6.78,P < 0.001]. Under high light conditions, Total-CB1-KO but not Total-CB1-WT showed a rapid decline [Time: F(9,81) = 3.88,P < 0.001] in the number of rearings that corresponded to the decrease in horizontal locomotion and the increase in immobility time shown by the same animals.
Glu-CB1-KO showed significantly less rearing than wild-type controls over the course of the exposure at 0 lx [Genotype × Time : F(9,198) = 3.15,P = 0.001; univariate F-tests of simple effects in manova, P < 0.05; Fig. 3e, right panel]. These differences likely related to the significant increase in the number of rearings in wild types over the course of the exposure [Time : F(9,90) = 2.51,P = 0.013] that was not evident in Glu-CB1-KO. At 700 lx, however, there were no significant genotype differences detectable (Fig. 3f, right panel).
Both Total-CB1-KO and Glu-CB1-KO showed increased thigmotaxis under high light conditions
In addition to exploratory behaviour, we analysed thigmotaxis as a measure of fear-/anxiety-related behaviour in the open field. To avoid that genotype differences in locomotion confounded the interpretation of this variable, we expressed the distance moved along the walls of the open field as a percentage of the total distance moved. Compared with wild-type controls by means of univariate F-tests of simple effects in manova, Total-CB1-KO showed less thigmotaxis at 0 lx [F(1,28) = 10.56,P = 0.003], but increased thigmotaxis at 700 lx [F(1,22) = 4.79,P = 0.039; Fig. 4a]. No genotype differences were obtained from an additional batch of mice, which had been tested at 30 lx (Fig. 4a; data of exploratory behaviour are not reported for this group for the sake of clarity, brevity and comparability with Glu-CB1 mice). If the development of thigmotaxis over the course of the exposure was considered, two-factorial ANOVAs for repeated measurements showed a significant genotype [F(1,28) = 11.61,P = 0.002] but not a significant Time or Genotype × Time interaction effects at 0 lx, indicating that Total-CB1-KO consistently showed less thigmotaxis than wild-type controls over the exposure time (Fig. 4b). At 700 lx, there was a significant interaction effect [F(9,225) = 3.60,P < 0.001]. Subsequent analyses of the simple effects showed that Total-CB1-KO showed more thigmotaxis than wild-type controls in particular at the end of the exposure (univariate F-tests of simple effects in manova, P < 0.05; Fig. 4b). These differences were caused by the permanent increase in thigmotaxis of the knockout mice over the exposure time [Time: F(9,90) = 2.51,P = 0.013] that could not be observed in the wild-type mice.
Glu-CB1-KO showed a similar phenotype as Total-CB1-KO, although less pronounced in magnitude. If tested at 0 lx, Glu-CB1-KO showed less thigmotaxis than Glu-CB1-WT at the beginning of the exposure [Genotype × Time : F(9,198) = 2.41,P = 0.013; univariate F-tests of simple effects in manova, P < 0.05; Fig. 4c,d]. At 700 lx, in contrast, they showed increased thigmotaxis [Genotype: F(1,20) = 5.83,P = 0.025; Fig. 4c,d].
Genotype differences persisted or became more pronounced during a second exposure to the open field
In order to assess the influence of the endocannabinoid system on repeated exposure to test situations, where the aversive stimuli cannot be avoided, Total-CB1 and Glu-CB1 mice were tested for a second time in the open field test at 0 or 700 lx. In general, mice showed a decrease in locomotion and in the number of rearings, which coincided with an increase in immobility and thigmotaxis, in particular, if tested repeatedly at 700 lx (supporting information, Fig. S1). Genotype differences observed during the first exposure largely persisted during the second exposure. In case of locomotion, immobility and rearings, Glu-CB1-KO, but not Total-CB1-KO, showed more pronounced effects of repeated testing at 700 lx than Glu-CB1-WT, which could be statistically secured by significant Genotype × Day interactions (Fig. S1; statistics not shown).
Both Total-CB1-KO and Glu-CB1-KO show decreased object and social investigation
Univariate F-tests in manova showed that Total-CB1-KO spent less time exploring the objects compared with their wild-type controls under 500 lx [F(1,28) = 4.36,P = 0.046], but not under 30 lx (Fig. 5). Evaluation of Glu-CB1 showed a similar phenotype with Glu-CB1-KO showing reduced investigation duration at 500 lx [F(1,19) = 12.81,P = 0.002] and this time also at 30 lx [F(1,20) = 9.02,P = 0.007] compared with the respective wild-type controls (Fig. 5).
Similar to object investigation, Total-CB1-KO spent also less time investigating a novel juvenile compared with wild-type controls under 500 lx [F(1,22) = 7.84,P = 0.010], but not under 0 lx (Fig. 5). Analysis of Glu-CB1 showed essentially the same phenotype with Glu-CB1-KO showing reduced investigation duration at 500 lx [F(1,21) = 10.04,P = 0.005], but no significant changes in investigation at 0 lx compared with the respective wild-type controls (Fig. 5).
Overview over genotype differences in exploratory behaviour
To summarize the genotype differences observed in the different exploration-based behavioural paradigms and to better illustrate the biological significance of the results, we calculated for each mutant line the deviation from respective wild-type controls as described in the Material and methods section. As shown in Fig. 6a, Total-CB1-KO consistently showed altered exploratory behaviour reminiscent of an anxiety-like phenotype if tested under high light intensities, but not in darkness, irrespective of whether or not the animals could avoid the light. Glu-CB1-KO, in contrast, showed a specific deficit in exploration of juveniles and objects, irrespective of the light conditions (Fig. 6b). In contrast, the effect sizes of deviations from WT mice in other parameters were rather small (statistics not shown).
The current study was designed to investigate the consequences of genetic deletion of CB1 either throughout the entire body (Total-CB1-KO) or specifically in glutamatergic cortical neurons (Glu-CB1-KO) in exploration-based behavioural paradigms. In particular, potential interactions with the aversiveness of the test situation (modified by changes in the light intensity), controllability of the test situation (avoidable vs. unavoidable exposure to bright light) and familiarity with the test procedure (repeated exposure to the elevated plus-maze and the open field) were considered. We could show that Total-CB1-KO showed changes in exploratory behaviour indicative of an anxiety-like phenotype in a variety of behavioural test paradigms irrespective of the controllability of the aversive situation, however, only if the animals were tested under highly aversive conditions (cf. Fig. 6a). A decrease in exploratory behaviour associated with an increase in thigmotaxis in the open field and a decrease of open arm exploration on the elevated plus-maze became evident in test-experienced Total-CB1-KO mice. In contrast, test-inexperienced Glu-CB1-KO failed to show behavioural changes akin to increased anxiety in test situations, where the mice could avoid the lit compartments (i.e. elevated plus-maze, light/dark test; cf. Fig. 6b). However, a weak-to-moderate anxiety-like phenotype, measured as thigmotaxis, became evident if the animals were tested under highly aversive conditions in an unavoidable situation (i.e. in the open field). Furthermore, exploration of novel objects or novel juvenile conspecifics was strongly attenuated under both moderate and highly aversive conditions. Similar to Total-CB1-KO also test-experienced Glu-CB1-KO showed a decrease in exploratory behaviour associated with an increase in thigmotaxis in the open field and a decrease in open arm exploration on the elevated plus-maze. Strikingly, the behavioural performance of Total-CB1-WT and Glu-CB1-WT appeared to be largely unaffected by changes in light conditions. Together, these data indicate that the endocannabinoid system serves as a kind of buffering system, which controls –among others –glutamatergic transmission and, thus, ensures constant levels of exploratory behaviour by rendering it independently of the aversiveness of the test situation. This keeps the balance in conflicting approach-avoidance situations with high emotional load on the side of exploratory behaviour. Apparently, this does only apply to situations where the aversive light stimulus cannot be avoided by the test-naï ve animals. Otherwise, CB1 on other neuronal populations, such as serotonergic and noradrenergic neurons of the brainstem (Bambico et al. 2007; Gobbi et al. 2005; Haring et al. 2007) or on GABAergic neurons (Haller et al. 2007), is likely to mediate the pro-exploratory effects of endocannabinoids.
On the elevated plus-maze, Total-CB1-KO differed from their wild-type controls primarily in the number of stretch-attend postures shown (Fig. 1b). This ethological parameter can be interpreted in terms of ’risk assessment', i.e. information gathering behaviours displayed in potentially threatening situations (Blanchard & Blanchard 1989). Risk assessment is often neglected in analyses of plus-maze behaviour, although it is particularly sensitive to benzodiazepines (e.g. Blanchard et al. 2001; Rodgers et al. 1999, 2003 and references therein). Because risk assessment shows a high correlation with corticosterone responses to the plus-maze (Rodgers et al. 1999), the fact that Total-CB1-KO, but not Glu-CB1-KO, displayed an increased number of stretch-attend postures may relate to the differential glucocorticoid responses observed in the two knockout lines (Steiner et al. 2008).
Reports on the effects of acute pharmacological blockade of CB1 receptors on spatio-temporal parameters on the plus-maze showed contradictory findings with anxiolytic-like effects (Haller et al. 2002), anxiogenic-like effects (Patel & Hillard 2006; Rodgers et al. 2005) and the failure to obtain significant effects at all (Moreira et al. 2008; Rodgers et al. 2003). Interestingly, in the few studies considering ethological parameters, none of the CB1 receptor antagonists tested caused a change in the number of stretch-attend postures (Rodgers et al. 2003). This renders it likely that the phenotype observed in Total-CB1-KO can be attributed –at least in part –to developmental effects of the life-long absence of CB1. However, a recent publication reported an increase in risk assessment in mice treated with the CB1 receptor antagonist SR141716 (rimonabant) before testing on the plus-maze at 0 lx. Strikingly enough –and because of yet unknown reasons –neither AM251 (i.e. another CB1 receptor antagonist) nor genetic deletion of CB1 showed the same phenotype (Thiemann et al. 2009).
Cannabinoid receptor type 1 on glutamatergic neurons prominently protects neuronal circuits from excessive excitatory neuronal activity (Marsicano et al. 2003; Monory et al. 2006) and is causally related to some of the effects of Δ9-tetrahydrocannabinol intoxication (Monory et al. 2007). In this study, we show for the first time that CB1-mediated regulatory constraints of glutamatergic transmission enable constant levels of exploratory behaviour over a broad range of changes in aversiveness of the test situation, thus substantially extending our recent observations concerning sensitized fear (Kamprath et al. 2009). The fact that Glu-CB1-KO, but not Total-CB1-KO, show reduced novel object investigation already at low light conditions may indicate a higher light sensitivity of these animals. Alternatively, CB1 on different neuronal populations (e.g. glutamatergic vs. GABAergic) may exert opposite effects on exploratory behaviour (Haller et al. 2007), with the consequence that population-specific deletion of CB1, rather than ubiquitous deletion, causes a disequilibrium of endocannabinoid-controlled transmitter systems and, thus, a more pronounced phenotype in the respective behavioural task. Future studies will have to use mutants with specific deletion of CB1 in GABAergic neurons (Monory et al. 2006) to substantiate this hypothesis.
Our findings support the notion that the endocannabinoid system interacts with glutamatergic transmission to dampen fear/anxiety responses to highly aversive and inescapable stressors. Therefore, we propose a scenario according to which unavoidable and highly aversive situations lead to a pronounced activation of glutamatergic transmission, which not only promotes fear responses (Millan 2003) but also triggers endocannabinoid synthesis and release from postsynaptic neurons (Varma et al. 2001). Once released, endocannabinoids inhibit glutamatergic transmission via presynaptically localized CB1 receptors (Katona et al. 2006; Kawamura et al. 2006; Monory et al. 2006; Takahashi & Castillo 2006). This ultrashort feedback action of endocannabinoids might keep the activation of a fear matrix (McNaughton & Corr 2004; Singewald 2007) in a physiological range, similarly to the feedback regulation of stress hormone secretion proposed at the level of the hypothalamus (Di et al. 2003) and thereby facilitate exploratory behaviour in approach-avoidance conflicts.
In the open field test, we observed bimodal effects of the endocannabinoid system on regulation of thigmotactic behaviour: Total-CB1-KO and, to a lesser extent, Glu-CB1-KO showed an increase in thigmotaxis if tested under high light conditions. In contrast, both Total-CB1-KO and Glu-CB1-KO showed less thigmotaxis than their wild-type controls at 0 lx because of yet unknown reasons. These findings clearly illustrate the necessity of testing mutant mice in situations with different levels of aversiveness.
It is well known that re-exposure to the plus-maze elicits a fundamentally different emotional state during trial 2 compared with trial 1 in a process called one-trial sensitization, which is characterized by increased avoidance of the open arms (for review see Carobrez & Bertoglio 2005). As another characteristic of one-trial sensitization, test-experienced animals do not longer respond to the anxiolytic effects of benzodiazepines (Holmes & Rodgers 1999). Interestingly, systemic treatment of mice with the selective CB1 antagonist SR141716 (Rodgers et al. 2003), but not AM251 (Rodgers et al. 2005), exerted anxiolytic-like effects during trial 2, but not trial 1, on the elevated plus-maze. Our data obtained from CB1-deficient mice show the opposite effect, namely a further increase in avoidance of open arm exploration during trial 2 in both Total-CB1-KO and Glu-CB1-KO (as evident from the significantly increased number of non-explorers), suggesting that endocannabinoid-controlled glutamatergic transmission plays a role in one-trial sensitization. The same seems to apply to repeated exposure to an open field at high light intensities. It is remarkable that endocannabinoid-controlled glutamatergic transmission seems to be responsible for the altered state of fear/anxiety during the second exposures rather than the ‘usually suspected transmitter systems' such as serotonin or noradrenaline (Gonzalez & File 1997; Griebel 1995; Ressler & Nemeroff 2000). However, taken into account, for instance, that (1) the dorsal raphe nucleus receives glutamatergic afferences from the prefrontal cortex (Levine & Jacobs 1992; Tao & Auerbach 2003) and (2) these pyramidal projecting neurons likely express CB1 (Marsicano & Lutz 1999), it is conceivable that enhanced glutamatergic input in Total-CB1-KO and Glu-CB1-KO mice might lead to increased activity of neurons in the raphe nuclei and eventually increased release of serotonin from efferences in target sites, such as the hippocampus. Here, 5-HT1A agonists mediate an anxiety-like effect, and, interestingly, this is specifically the case for trial 2 on the elevated plus-maze (File et al. 1996).
During the first exposure to the test apparatus, rodents apparently acquire some kind of memory related to exploration of the potentially dangerous areas of the maze (Carobrez & Bertoglio 2005; File et al. 1993; File & Zangrossi 1993). Consequently, there is the option that endocannabinoids modulate one-trial sensitization via their influence on memory processes (Riedel & Davies 2005). Attenuated CB1 signalling seems to coincide, among others, with increased short-term (Degroot et al. 2005) and long-term habituation (Thiemann et al. 2007) to an open field. Our data partially support these observations: Total-CB1-KO showed pronounced within-exposure habituation in the open field under bright light, but no differences to wild-type controls in between-exposure habituation. An increased between-exposure habituation was detectable only in the case of Glu-CB1-KO. It remains to be shown in future studies whether increased familiarity with the test environment indeed accounts for the decreased exploration in test-experienced CB1-deficient mice and, thus, shifts the motivational balance from novelty seeking towards avoidance-like behaviours.
Taken together, our data show a prominent role of CB1 in keeping exploratory behaviour at a constant level despite changes in the aversiveness of the test situation. The controllability of the test situation determines whether endocannabinoid-controlled glutamatergic transmission or other neurotransmitter systems are involved in this process.
We would like to thank Martina Reents and Rita Murau for excellent technical assistance, Drs V. Micale and R. Marsch for pharmacological validation of the plus-maze and Prof. Dr M. Engelmann for his critical comments on an earlier version of this manuscript. This study was supported in part by a grant from the Hübner-Foundation to C.T.W.