Deletion of Metabotropic Glutamate Receptors 2 and 3 (mGlu2 & mGlu3) in Mice Disrupts Sleep and Wheel-Running Activity, and Increases the Sensitivity of the Circadian System to Light

Sleep and/or circadian rhythm disruption (SCRD) is seen in up to 80% of schizophrenia patients. The co-morbidity of schizophrenia and SCRD may in part stem from dysfunction in common brain mechanisms, which include the glutamate system, and in particular, the group II metabotropic glutamate receptors mGlu2 and mGlu3 (encoded by the genes Grm2 and Grm3). These receptors are relevant to the pathophysiology and potential treatment of schizophrenia, and have also been implicated in sleep and circadian function. In the present study, we characterised the sleep and circadian rhythms of Grm2/3 double knockout (Grm2/3 -/-) mice, to provide further evidence for the involvement of group II metabotropic glutamate receptors in the regulation of sleep and circadian rhythms. We report several novel findings. Firstly, Grm2/3 -/- mice demonstrated a decrease in immobility-determined sleep time and an increase in immobility-determined sleep fragmentation. Secondly, Grm2/3 -/- mice showed heightened sensitivity to the circadian effects of light, manifested as increased period lengthening in constant light, and greater phase delays in response to nocturnal light pulses. Greater light-induced phase delays were also exhibited by wildtype C57Bl/6J mice following administration of the mGlu2/3 negative allosteric modulator RO4432717. These results confirm the involvement of group II metabotropic glutamate receptors in photic entrainment and sleep regulation pathways. Finally, the diurnal wheel-running rhythms of Grm2/3 -/- mice were perturbed under a standard light/dark cycle, but their diurnal rest-activity rhythms were unaltered in cages lacking running wheels, as determined with passive infrared motion detectors. Hence, when assessing the diurnal rest-activity rhythms of mice, the choice of assay can have a major bearing on the results obtained.


Introduction
Schizophrenia is a debilitating psychiatric disorder that affects 0.3-0.7% of the world's population [1], while sleep and/or circadian rhythm disruption (SCRD) is observed in 30-80% of schizophrenia patients [2]. Sleep disturbances in schizophrenia include increases in sleep latency and sleep fragmentation, and reductions in total sleep time, sleep efficiency, rapid eye movement (REM) sleep latency, REM sleep density and slow-wave sleep duration [2][3][4]. Typical circadian abnormalities in the disorder include the abnormal phasing, instability and fragmentation of rest-activity rhythms [5][6][7][8][9]. Some of these abnormalities may be secondary to the disorder and its drug treatment, but the co-morbidity of schizophrenia and SCRD may also stem from dysfunction in common brain mechanisms (e.g. specific neurotransmitter systems) [10,11].
The glutamate system is one such candidate mechanism. There is strong evidence of glutamatergic dysfunction in schizophrenia [12,13], while glutamate is the principal neurotransmitter within the retinohypothalamic tract (RHT), which conveys photic entrainment cues from the retina to the circadian timing system in the suprachiasmatic nucleus (SCN) [14][15][16]. Glutamatergic signalling is also important for bilateral communication between the left and right SCN [17], while the SCN stimulates the paraventricular and ventrolateral preoptic nuclei via glutamatergic efferents [18][19][20][21][22]. Glutamate release in the prefrontal cortex displays rhythmic fluctuations during the sleep-wake cycle; it increases during wakefulness and REM sleep episodes, but decreases during non-REM sleep [23][24][25].
At the receptor level, group II metabotropic glutamate receptors may be particularly relevant to the pathophysiology of schizophrenia. mGlu2 and mGlu3, encoded by the genes Grm2 and Grm3, are presynaptic inhibitory autoreceptors that are widely expressed in both rodent and human brain [26][27][28][29][30][31]. Grm2 expression is predominantly neuronal, whereas Grm3 expression is primarily glial [28]. In a recent meta-analysis of genome-wide association studies, a strong association was reported between the Grm3 locus and schizophrenia [32], extending prior evidence that Grm3 is involved in the genetic predisposition to psychosis [33][34][35][36]. Polymorphic variation in Grm3 has also been associated with cognitive performance, prefrontal glutamatergic neurotransmission and hippocampal pathophysiology [33,37]. In addition, there is clinical and preclinical evidence that mGlu2 & 3 may be therapeutic targets for the disorder, although no drug in this category has yet been licensed for use [38].
In the present study, we characterised the sleep and circadian rhythms of Grm2/3 double knockout (Grm2/3 -/-) mice. Although expression of the dimeric form of mGlu3 is reduced in the brains of schizophrenia patients [53], mGlu2 and mGlu3 are not completely absent, so the Grm2/3 -/mouse cannot be considered a 'disease model'. Instead, our primary aim was to obtain additional evidence for the involvement of group II metabotropic glutamate receptors in the regulation of sleep and circadian rhythms. Based on the evidence described above, we predicted that Grm2/3 -/mice would demonstrate a reduction in total sleep duration, increased sleep fragmentation, and heightened sensitivity to the circadian effects of light (e.g. lightinduced phase shifts of greater magnitude).
Age-matched males were used in all experiments. Mice were at least 2 months old at the onset of behavioural screening and were no older than 9 months upon completion of testing. Only males were tested due to the potentially confounding influence of the oestrus cycle on wheel-running activity [59]. 2 transgenic cohorts were used; cohort 1 consisted of 11 Grm2/ 3 +/+ mice and 12 Grm2/3 -/mice, while cohort 2 was comprised of 12 mice of each genotype. For the experiment involving the administration of the drug RO4432717 (F.Hoffmann-LaRoche, Basel, Switzerland), a separate cohort of 22 male wildtype C57Bl/6J mice was used (Harlan, UK). Mice were singly-housed and had access to food and water ad libitum. All behavioural procedures were performed in accordance with the United Kingdom Animals (Scientific Procedures) Act of 1986 and the University of Oxford Policy on the Use of Animals in Scientific Research. All experiments were approved by the University of Oxford Animal Welfare and Ethical Review Board, and were conducted under the PPL 30/2812 by PILs 30/9339 and 30/8553.

Order of tests
Cohorts 1 and 2 completed an extensive wheel-running screen consisting of several standard protocols (full details below). Video-tracking, passive-infrared (PIR) motion detection and object interaction testing were only performed on cohort 2. The order of these tests was counterbalanced across genotype groups. Half of cohort 2 underwent video-tracking followed by PIR motion detection, then wheel-running, and finally object interaction testing. The other half of cohort 2 were subjected to PIR motion detection followed by video-tracking, then object interaction testing, and finally wheel-running. contained 6 cages, with 3 mice of each genotype housed in alternating positions. A standard 12:12 h light/dark (12:12 LD) cycle was employed and illuminance was 100 lux at the base of each cage. Three 24 h videos (i.e. 3 separate 12:12 LD cycles) were recorded at weekly intervals, using a near-infrared CCTV camera (Maplin Electronics, Rotherham, UK). All videos were recorded at 3 frames per second, in the AVI file format. An acrylic block was placed under the food hopper to keep the mouse in the recording field at all times.
Video files were stored on a digital hard drive recorder (Samsung, Suwon, South Korea) prior to analysis with ANY-maze 4.5 (Stoelting, Wood Dale, Illinois). Multiple immobilitydetermined sleep parameters were extracted from the video footage: total sleep time, light and dark phase sleep time, light and dark phase sleep bouts, and light and dark phase sleep bout duration. Sleep was defined as a period of immobility of at least 40 s, a previously established proxy measure of sleep [60]. This measure has an extremely high concordance (>95%) with EEG-based sleep determination [60]. Immobility sensitivity was set at 95% to prevent the detection of movements caused by breathing during sleep.
In addition to the sleep parameters outlined above, four activity parameters were extracted from the video footage using ANY-maze 4.5: total activity, light and dark phase activity (distance travelled in metres), and relative light phase activity (the percentage of total activity occurring during the light phase). Note that one Grm2/3 +/+ mouse was excluded from all dark phase sleep and activity analyses as it escaped from its cage during dark phase recording.
3-5 days after the third 24 h recording, mice were exposed to a 1 h 100 lux light pulse from zeitgeber time (ZT) 16 to 17. (ZT0 refers to the onset of the light phase, while ZT12 denotes the onset of the dark phase). Negative masking-percent activity suppression induced by the light pulse-was computed to provide an indication of photosensitivity [61]. This was calculated by comparing activity levels during the light pulse with average activity levels during the same time window (ZT16-17) across the previous 3 days.
Passive-infrared (PIR) motion detection. Due to the prohibitive size of the video files, it was impractical to analyse more than 72 h of video data. PIR motion detection provides a less data-heavy alternative, enabling the estimation of sleep over a longer period; in this case, 14 consecutive days of 12:12 LD at 100 lux. It also enables the computation of specific parameters that cannot be extracted from isolated 24 h video recordings, such as period length (see below). Full details of this novel technique, developed in our laboratory, should be published in the near future [Brown et al., in review]. Housing conditions for the PIR protocol were identical to that of the video-tracking protocol outlined above. Again, an acrylic block was placed under the food hopper to keep the mouse in the motion detection field at all times. Raw data from the PIR motion sensors took the form of % time active per 10 s epoch, with sensors activated both by gross locomotion and small movements such as turning of the head. Three immobilitydetermined sleep parameters were extracted from the raw PIR data using Microsoft Excel 2013 (Microsoft, Redmond, USA): total sleep time, light phase sleep time and dark phase sleep time. Note that methods do not yet exist to automatically compute parameters pertaining to the frequency and duration of sleep bouts from PIR data. As with the previously described videobased method of sleep determination [60], sleep was defined as a period of immobility of at least 40 s. We have recently confirmed that this measure has an extremely high concordance (>95%) with both video-based sleep determination and EEG-based sleep determination [Brown et al., in review].
In addition to the sleep parameters outlined above, four activity parameters were extracted from the raw PIR data using the ClockLab toolbox (Actimetrics, Illinois, USA) for MatLab (MathWorks, Massachusetts, USA): total activity, light and dark phase activity (expressed in arbitrary units), and relative light phase activity (as above).
Standard diurnal rest-activity rhythm parameters were also computed from the raw PIR data using ClockLab; these comprised two measures of diurnal rest-activity rhythm fragmentation (daily activity bouts and daily activity bout duration), two measures of diurnal rest-activity rhythm consistency (chi-square periodogram amplitude and onset tau error), and period length. Activity bouts were defined using pre-established criteria for ClockLab (maximum gap: 18 min; threshold: 5%). A sixth diurnal rest-activity rhythm parameter-interdaily stabilitywas computed using the software package Actiwatch Activity & Sleep Analysis 7 (Cambridge Neurotechnology, Cambridge, UK). Interdaily stability is a non-parametric measure that quantifies the similarity of activity patterns across multiple days [62].

Diurnal rest-activity rhythm and circadian screening
Wheel-running. This screen was based on an established protocol [63][64][65]. Housing conditions were identical to that of the video-tracking and PIR protocols described above, except that cages were fitted with running wheels (18 cm diameter). As previously, illuminance was 100 lux at the base of each cage. Mice in both cohorts were exposed to 14 days of 12:12 LD and 11 days of constant dark (DD). In addition, cohort 1 was subjected to 13 days of constant light (LL), and a 6 h phase advance during 12:12 LD.
During 12:12 LD, four activity parameters were extracted from the raw wheel-running data using ClockLab: total activity, light and dark phase activity (wheel rotations), and relative light phase activity (as above). The six diurnal rest-activity rhythm parameters computed from the raw PIR data (see above) were also computed from the raw wheel-running data, using Clock-Lab and Actiwatch Activity & Sleep Analysis 7. Again, activity bouts were defined using preestablished criteria (maximum gap: 18 min; threshold: 5 counts/min). Note that two Grm2/3 -/mice were omitted from these activity bout analyses as their activity levels were too low to derive bout data. ClockLab was also used to quantify period length under free-running conditions (i.e. DD and LL), while re-entrainment following the 6 h phase advance was scored manually from individual actograms as the number of days taken to re-entrain.
Both cohorts were subjected to type I and type II phase-shifting light pulses [66] (using previously described protocols [64,67]) to induce a phase delay in their wheel-running rhythms, providing an indication of circadian photosensitivity. For the type I pulse, mice were exposed to 100 lux light for 15 minutes at circadian time (CT) 16 during DD. The magnitude of the phase delay was quantified by fitting one regression line through 10 consecutive activity onsets preceding the light pulse, and another through 6 consecutive activity onsets following the pulse. The first 2 activity onsets following the light pulse were disregarded because of possible transition effects [64]. For the type II pulse, mice were exposed to 100 lux light for one hour at ZT16 during a standard 12:12 LD cycle. Following the pulse, mice were released into DD. The magnitude of the phase delay was quantified by fitting one regression line through 6 consecutive activity onsets preceding the light pulse, and another through 6 consecutive activity onsets following the pulse. This time, only one activity onset following the light pulse was disregarded [64]. For both light pulses, the magnitude of the phase delay was calculated as the time difference between the two regression lines on the first day after the light pulse. Negative masking was also computed for the type II pulse, using the same method employed in the videotracking screen.
Administration of an mGlu2/3 negative allosteric modulator to wildtype mice prior to a type I phase-shifting light pulse To further investigate the hypothesis that group II metabotropic glutamate receptors modulate photic entrainment, the mGlu2/3 negative allosteric modulator RO4432717 was administered to wildtype mice in a standard phase-shifting paradigm. The pharmacology of RO4432717 (previously known as compound 7i) is described elsewhere [68]. 22 male wildtype C57Bl/6J mice (aged 80 days) were individually housed in large cages (44 cm long × 26 cm wide × 12 cm high), fitted with running wheels (18 cm diameter), under a 12:12 LD cycle at 400 lux. All mice were kept in DD for 1 day prior to the administration of a 30 min type I phase-shifting light pulse (400 lux) at CT16. 3 mice underwent no pre-treatment, while 6 mice received an intraperitoneal injection of saline (0.3 ml) at CT12 on the day of the light pulse. The remaining 13 mice received an intraperitoneal injection of RO4432717 at CT12. Three dosages were used: 3 mg (4 mice), 10 mg (5 mice) and 30 mg (4 mice). Injection volumes were kept constant at 0.3 ml. The magnitude of the phase delay was calculated using the method described above.

Home-cage object interaction test
To evaluate whether the altered wheel-running behaviour of Grm2/3 -/mice was a consequence of increased object neophobia (toward their running wheels), a simple object interaction task was conducted within the home-cage. The results of this test are presented in S1 Fig.

Statistical Analyses
All statistical analyses were performed with SPSS 22.0 (IBM, Armonk, New York). Unless otherwise stated, all reported statistics are the result of analyses of variance, with genotype included as the principal independent variable. Differences were considered to be statistically significant at p-values < 0.05. Greenhouse-Geisser corrections were applied where appropriate, but uncorrected degrees of freedom are reported in order to preserve the transparency of the statistical design. For analyses involving both experimental cohorts, further analyses of variance were conducted with cohort included as an extra independent variable. However, no additional effects or interactions were observed. In all figures, Ã indicates a p-value 0.05, ÃÃ indicates a p-value 0.01, and ÃÃÃ indicates a p-value 0.001. Error bars depict the standard error of the mean. M = mean.
Wheel-running and video-tracking analyses: Negative masking during nocturnal light pulses. Negative masking was unaffected by genotype, regardless of the method used to assay locomotor activity. Genotype had no effect on the degree of activity suppression induced by a type II light pulse in either the wheel-running screen (F 1,45 = 0.994, P = 0.324; Fig 2E) or the video-tracking screen (F 1,22 = 0.111, P = 0.743; Fig 2F).
Wildtype mice are more sensitive to the circadian effects of light following the administration of an mGlu2/3 negative allosteric modulator To discount the possibility that the increased light sensitivity of Grm2/3 -/mice is a consequence of altered neurodevelopment, we evaluated the impact of the mGlu2/3 negative allosteric modulator RO4432717 on the phase-shifting responses of wildtype C57Bl/6J mice. There was no difference in the magnitude of phase delays (induced by a type I light pulse at CT16) between untreated wildtype mice and wildtype mice injected with saline at CT12 (F 1,7 = 0.096, P = 0.766; Fig 2G). Relative to saline-treated mice, phase delays were greater in mice injected with RO4432717 at CT12. This difference was statistically significant at  (Fig 2C) and type II (Fig 2D) phase-shifting light pulses are larger in Grm2/3 -/than Grm2/3 +/+ mice. (E & F) Grm2/3 +/+ and Grm2/3 -/mice show similar levels of negative masking during a type II light pulse, as assayed by wheel-running ( Fig 2E) and video-tracking (Fig 2F). (G) In wildtype C57Bl/6J mice, phase delays induced by a type I light pulse are enhanced following the administration of the mGlu2/3 negative allosteric modulator (NAM) RO4432717. The diurnal rest-activity rhythms of Grm2/3 -/mice are dependent on the assay used to measure locomotor behaviour Wheel-running activity analyses. Over 14 days of continuous recording under 12:12 LD, total wheel-running activity was markedly reduced in Grm2/3 -/mice relative to Grm2/3 +/+ mice (F 1,45 = 8.165, P = 0.006; Fig 3A). This reflects the fact that Grm2/3 -/mice were significantly less active than Grm2/3 +/+ mice during the dark phase (F 1,45 = 8.503, P = 0.006; Fig 3C). By contrast, genotype had no effect on activity during the light phase (F 1,45 = 1.257, P = 0.268; Fig 3B); indeed, numerically speaking, Grm2/3 -/mice were more active than Grm2/3 +/+ mice during the light phase. Consequently, the percentage of total daily activity which occurred during the light phase was greater in Grm2/3 -/mice than Grm2/3 +/+ mice (F 1,45 = 7.585, P = 0.008). (Note that mean values for all wheel-running analyses are included in Table 1, and representative actograms of 12:12 LD are depicted in Fig 4A).
Video-tracking activity analyses. When mice were housed in cages lacking running wheels, and locomotor activity was determined with near-infrared cameras, the rest-activity profiles of Grm2/3 -/mice were very different from those derived using running wheels (see Fig 3D). As with running wheels, light phase activity was greater in Grm2/3 -/mice than Grm2/ 3 +/+ mice (F 1,22 = 6.166, P = 0.021; Fig 3E), although the percentage of total daily activity which occurred during the light phase did not vary according to genotype (F 1,21 = 0.625, P = 0.438). Crucially, however, genotype had no effect on dark phase activity (F 1,21 = 0.217, P = 0.646; Fig 3F) or total activity levels (F 1,21 = 0.395, P = 0.536). There were no effects or interactions involving recording day for any of these analyses, reflecting the fact that the activity profiles of Grm2/3 +/+ and Grm2/3 -/mice were relatively consistent across the three recording sessions. Note that specific parameters such as period length cannot be extracted from isolated 24 h video recordings.
PIR activity analyses. When mice were housed in cages lacking running wheels, and locomotor activity was determined with PIR motion detectors, the rest-activity profiles of Grm2/ Sleep & Circadian Rhythms in Grm2/3 Double Knockout Mice 3 -/mice were again very different from those derived using running wheels (see Fig 3G). As with running wheels, light phase activity was greater in Grm2/3 -/mice than Grm2/3 +/+ mice (F 1,22 = 2.551, P = 0.124; Fig 3H), and the percentage of total daily activity which occurred during the light phase was higher in Grm2/3 -/mice (F 1,22 = 3.991, P = 0.058), although neither of these effects reached statistical significance. More importantly, however, genotype had no effect on dark phase activity (F 1,22 = 0.606, P = 0.445; Fig 3I) or total activity levels (F 1,22 = 0.122, P = 0.730). (Note that mean values for all PIR analyses are included in Table 2, and representative actograms of 12:12 LD are depicted in Fig 4B).

Discussion
The present study investigated sleep and circadian rhythms in Grm2/3 double knockout (Grm2/3 -/-) mice and wildtype (Grm2/3 +/+ ) controls. We report three novel findings. Firstly, Grm2/3 -/mice displayed a decrease in immobility-determined sleep time and an increase in immobility-determined sleep fragmentation. Secondly, Grm2/3 -/mice demonstrated heightened sensitivity to the circadian effects of light, as determined using running wheels. More specifically, they showed increased period lengthening in LL, and greater phase delays in response to type I and type II phase-shifting light pulses. Thirdly, the wheel-running activity of Grm2/ 3 -/mice was significantly perturbed under a standard 12:12 LD cycle. Table 2. Descriptive statistics for selected rest-activity parameters derived from 14 consecutive days of passive-infrared (PIR) data.
In which brain region(s) do group II metabotropic glutamate receptors exert their influence on REM sleep? The amygdala is a promising candidate, since it is known to influence sleep regulation [69][70][71][72][73][74], and Grm2 is highly expressed in this region [26,28,30]. Indeed, Dong and colleagues observed reduced REM sleep in wildtype rats following microinjection of the mGlu2/3 agonist LY379268 directly into the basal amygdala [46]. Another region of interest is the thalamic reticular nucleus, which forms part of the ascending arousal system; mGlu2/3 antagonists are known to modulate the inhibitory output of this structure [51,52].
In addition to reduced sleep time, Grm2/3 -/mice exhibited an increase in sleep fragmentation. Consistent with this observation, both the mGlu2/3 antagonist LY341495 and the mGlu2 negative allosteric modulator RO4491533 increase the number of transitions between sleep and wakefulness in wildtype rats [44]. Hence, group II metabotropic glutamate receptors may be involved in the putative 'flip-flop' circuit involving the ventrolateral preoptic nuclei that is hypothesised to control sleep onset and offset [75]. The inability of Grm2/3 -/mice to remain asleep for sustained periods might also explain the overall reduction in their sleep time.
mGlu2 and/or mGlu3 modulate the sensitivity of the circadian system to light Period length in LL was significantly greater in Grm2/3 -/mice than Grm2/3 +/+ mice, despite the fact that period length was unaffected by genotype in DD. Moreover, following both type I and type II phase-shifting light pulses, phase delays were greater in Grm2/3 -/mice than Grm2/ 3 +/+ mice. Negative masking during the type II pulse was unaffected by genotype, although this may reflect a ceiling effect, since the suppression of wheel-running activity was very high in mice of both genotypes. Consistent with our observations in Grm2/3 -/mice, light-induced phase delays in wildtype C57Bl/6J mice were enhanced by the administration of the mGlu2/3 negative allosteric modulator RO4432717. While it has previously been reported that mGlu2/ 3-modulating drugs can influence the magnitude of light-induced phase delays in wildtype hamsters [41,42], our own data represent the first demonstration of such an effect in wildtype mice. It is also significant that we obtained analogous results in a transgenic model and an acute pharmacological model; this suggests that the heightened light sensitivity of the Grm2/ 3 -/mouse is not a consequence of altered neurodevelopment.
Collectively, these results demonstrate that mGlu2 and/or mGlu3 can modulate the sensitivity of the circadian system to light, which raises the possibility that mGlu2/3-targeting compounds may be useful for the treatment of circadian rhythm disorders. The relevance of these findings to schizophrenia is unclear, however. In humans, the suppression of nocturnal melatonin secretion by light is the standard measure of circadian photosensitivity; there is some evidence that this response is upregulated in bipolar disorder and seasonal affective disorder [76], but such studies have not been undertaken in individuals with schizophrenia.
It is possible that the enhanced light-induced phase delays shown by Grm2/3 -/mice reflect the absence of mGlu2 and/or mGlu3 from RHT-SCN synapses. Photic entrainment cues are transmitted from the eye to the SCN via the axons of photosensitive retinal ganglion cells, which together make up the RHT [77,78]. While ionotropic glutamate receptors mediate the transduction of light information from the retina to the SCN via the RHT [79][80][81][82][83][84], metabotropic glutamate receptors may play a modulatory role at RHT-SCN synapses [40][41][42]. Consistent with the fact that group II metabotropic glutamate receptors are presynaptic inhibitory autoreceptors, mGlu2/3 agonists serve to inhibit NMDA-evoked calcium influx in SCN neurons [40], presumably via a reduction in presynaptic glutamate release. By the same logic, the antagonism or deletion of group II metabotropic glutamate receptors might be expected to increase presynaptic glutamate release at RHT-SCN synapses, explaining the enhanced light-induced phase delays witnessed in Grm2/3 -/mice and in wildtype C57Bl/6J mice following the administration of RO4432717. It should also be noted that Grm2 is expressed in the retina, in cholinergic amacrine cells [85][86][87], although its role here is not well understood. Grm3 is not expressed in the mammalian retina [85][86][87].
Origins and implications of the perturbed wheel-running activity of Grm2/ 3 -/mice Under 12:12 LD, the diurnal wheel-running rhythms of Grm2/3 -/mice were characterised by increased activity fragmentation and reduced consistency, while total activity levels were substantially reduced. By contrast, general home-cage activity levels-assessed with both nearinfrared cameras and PIR motion detectors-were unchanged in Grm2/3 -/mice. The consistency of home-cage diurnal rest-activity rhythms, as measured with PIR motion detectors, was also unaffected by genotype. This implies that the perturbed diurnal wheel-running rhythms of Grm2/3 -/mice are in some way related to the use of the wheels themselves.
There are several possible reasons for the significantly reduced wheel-running activity of Grm2/3 -/mice, which are not mutually exclusive. It could reflect heightened object neophobia--a manifestation of increased anxiety-since all mice were housed in empty home-cages prior to the introduction of running wheels. This seems unlikely, however, given that Grm2/3 +/+ and Grm2/3 -/mice showed equal exploration of another novel object (a transparent play-tube) immediately after its introduction to the home-cage (see S1 Fig). Moreover, a previous study found no evidence of altered anxiety in Grm2/3 -/mice in either the elevated plus maze, open field, black and white alley, or novelty-suppressed feeding tasks [54]. A more straightforward explanation is a deficit in motor function. In support of this hypothesis, Grm2/3 -/mice demonstrate mild but significant impairments in standard tests of motor coordination; they are impaired at both the accelerating rotarod task and the multiple static rods task [54]. A final possibility is that Grm2/3 -/mice are less motivated to engage in wheel-running activity; perhaps the physical act of wheel-running is inherently less rewarding for Grm2/3 -/than Grm2/ 3 +/+ mice. Long-term voluntary wheel-running is known to be rewarding for rodents, and produces plastic changes in mesolimbic reward neurocircuity [88]. Significantly, striatal dopamine is reduced in Grm2/3 -/mice, particularly in the nucleus accumbens [89], a brain region involved in reward and motivation [90,91].
It is somewhat more difficult to explain why the PIR-derived diurnal rest-activity rhythms of Grm2/3 -/mice were less fragmented than those of Grm2/3 +/+ mice, when their diurnal wheel-running rhythms were more fragmented. By contrast, Grm2/3 -/mice demonstrated greater light phase activity than Grm2/3 +/+ mice in all three assays (i.e. wheel-running, PIR and video-tracking). This is likely a direct consequence of their reduced light phase sleep time.
Interestingly, our results deviate slightly from those of previous experiments that have studied diurnal rest-activity rhythms in Grm2/3 -/mice. Lyon and colleagues observed that Grm2/ 3 -/mice were less active than Grm2/3 +/+ mice in cages lacking running wheels, particularly in the dark phase [55]. These results should be treated with caution, however, as activity was assessed over a relatively short period (less than 3 light/dark cycles), and recording began immediately after transferring the mice from the familiar home-cage environment to a novel activity-monitoring cage [55]. Furthermore, mice in this experiment were group-housed in the home-cage prior to single-housing in the activity-monitoring cages [55]. By contrast, mice were singly-housed throughout the present study.
Relative merits of wheel-running, video-tracking and PIR assays The Grm2/3 -/mouse clearly demonstrates that the rest-activity phenotype of a mouse can vary markedly depending on the method used to assay locomotor behaviour. This is perhaps unsurprising given that wheel-running is a measure of voluntary exercise rather than general homecage activity. Indeed, wheel-running is a complex behaviour that may be subject to multiple influences including arousal, motivation, anxiety, motor coordination, and sensitivity to reward [92][93][94]. Hence, wheel-running might not be the most appropriate assay for disease-relevant mouse models that show alterations in one or more of these parameters.
The present study is the first to use PIR motion detectors to simultaneously assess sleep and diurnal rest-activity rhythms in a transgenic mouse. The use of PIR motion detectors for sleep estimation was recently validated with simultaneous EEG-based sleep determination in wildtype mice [Brown et al., in review]. The similarity of the PIR-derived and video-derived sleep data presented here serves as further validation of this novel technique. PIR assays offer two main advantages over video-tracking assays: firstly, they enable the estimation of sleep over a longer period of time, providing a more representative view of an animal's sleep profile; and secondly, they enable the computation of specific parameters that cannot be extracted from isolated 24 h video recordings, such as period length.

Sleep disruption in neuropsychiatric disorders
As outlined previously, up to 80% of schizophrenia patients are afflicted by SCRD [2]. The prevalence of SCRD in schizophrenia has been attributed to the use of antipsychotic medication and the absence of social routine [10,11], but, contrary to these explanations, SCRD is seen in both medication-naïve patients [11] and patients that follow a fixed daily routine [7]. Instead, the co-morbidity of schizophrenia and SCRD may reflect dysfunction in common brain mechanisms, such as the glutamate system, which is clearly relevant to sleep [23][24][25], photic entrainment [14][15][16] and schizophrenia [12,13]. At the receptor level, multiple lines of evidence implicate group II metabotropic glutamate receptors in sleep regulation and photic entrainment [14,[39][40][41][42][43][44][45][46][47][48][49][50][51][52], while there is also a robust association between the Grm3 locus and schizophrenia [32][33][34][35][36]. The observation of reduced and fragmented sleep in Grm2/3 -/mice implies-but does not prove-that abnormal group II metabotropic glutamate receptor signalling may be relevant to sleep disruption in schizophrenia. Having said this, it is important to acknowledge that the Grm2/3 -/mouse is not a 'disease model' in itself; although there is evidence for the altered dimerisation of mGlu3 in schizophrenia [53], neither mGlu2 or mGlu3 are completely absent in the brains of schizophrenia patients.
Group II metabotropic glutamate receptors could be relevant to SCRD in neuropsychiatric disorders besides schizophrenia. Grm3 is also associated with bipolar disorder [95,96], and, in contrast to schizophrenia, reduced REM sleep is very common in this condition [97]. This is significant given that REM sleep is reduced in wildtype rats after the administration of various mGlu2/3-modulating drugs [43][44][45][46][47][48][49][50]. Moreover, there is evidence that circadian photosensitivity is elevated in bipolar disorder [76], just as it is in Grm2/3 -/mice.

Conclusions
The results of the present study add to a growing body of evidence implicating group II metabotropic glutamate receptors in both photic entrainment and sleep regulation pathways. Consistent with the effects of a wide range of mGlu2/3-modulating drugs, we provide the first evidence that the genetic ablation of mGlu2 & 3 increases sleep fragmentation and reduces sleep time in mice. Our findings are also the first to demonstrate that the genetic ablation of mGlu2 & 3 heightens the sensitivity of the circadian system to light. These observations are compatible with our hypothesis that SCRD and specific neuropsychiatric disorders share common mechanistic origins [10,11], and tentatively suggest that altered group II metabotropic glutamate receptor signalling could be relevant to sleep disruption in schizophrenia, and to sleep and circadian disruption in bipolar disorder. In addition, we report that the rest-activity phenotype of Grm2/3 -/mice is critically dependent on the method used to assay locomotor behaviour. We argue that wheel-running might not be the most appropriate assay for diseaserelevant mouse models with altered arousal, motivation, anxiety, motor coordination and/or sensitivity to reward, since each of these parameters has the potential to influence wheelrunning activity.
Supporting Information S1 Fig. Home-cage object interaction is unaltered in Grm2/3 -/mice. Genotype has no effect on interaction with a novel object during the first 10 min (A) or first hour (B) after its introduction to the home-cage (at ZT16), or during the entirety of the subsequent dark phase (C). The object was a transparent acrylic play-tube. Object interaction was defined as the amount of time spent in contact with the play-tube (including time spent within it). (TIF) S1

Author Contributions
Conceived and designed the experiments: DP AJ LAB SNP. Performed the experiments: DP AJ. Analyzed the data: DP AJ. Contributed reagents/materials/analysis tools: SG LAB. Wrote the paper: DP AJ LAB SKET SH SG PJH DMB RGF SNP.