Deletion of the clock gene Period2 (Per2) in glial cells alters mood-related behavior in mice

The circadian clock regulates many biochemical and physiological pathways, and lack of clock genes, such as Period (Per) 2, affects not only circadian activity rhythms, but can also modulate feeding and mood-related behaviors. However, it is not known how cell-type specific expression of Per2 contributes to these behaviors. In this study, we find that Per2 in glial cells is important for balancing mood-related behaviors, without affecting circadian activity parameters. Genetic and adeno-associated virus-mediated deletion of Per2 in glial cells of mice leads to reduced despair and anxiety. This is paralleled by an increase of the GABA transporter 2 (Gat2/Slc6a13) and Dopamine receptor D3 (Drd3) mRNA, and a reduction of glutamate levels in the nucleus accumbens (NAc). Interestingly, neuronal Per2 knock-out also reduces despair, but does not influence anxiety. The change in mood-related behavior is not a result of a defective molecular clock, as glial Bmal1 deletion has no effect on neither despair nor anxiety. Exclusive deletion of Per2 in glia of the NAc reduced despair, but had no influence on anxiety. Our data provide strong evidence for an important role of glial Per2 in regulating mood-related behavior.


Results
Deletion of Per2 in glial cells by cross-breeding of mice. In order to study the importance of the Per2 gene in glial cells, we crossed our floxed Per2 (Per2 fl/fl ) animals 13 with mice expressing the Cre-recombinase under transcriptional control of the human Gfap promoter (GCre) 53 . In families where the Cre was inherited from the maternal side, 58% of Cre-negative offspring (out of 80 progeny) showed germline recombination leading to total body Per2 heterozygous deletion. In comparison, 0% of paternally inherited Cre (out of 260 progeny; Supplemental table 1) showed any germline recombination, as previously described 30 . Therefore, we used only male Cre-positive mice for matings with female Per2 fl/fl animals in order to obtain glial-specific deletion of Per2 (GCre + Per2 fl/fl termed GPer2).
Next, we verified glial deletion of PER2 in brain tissue collected at zeitgeber time (ZT) 12 (where ZT0 is lights on and ZT12 is lights off) of GPer2 mice using immunohistochemistry. Sections of the dorsal part of the suprachiasmatic nuclei (SCN) containing mainly arginine-vasopressin (AVP) neurons displayed immunoreactivity with a PER2 antibody (green) (Fig. 1A). In control animals co-staining with a GFAP antibody (red) showed partially overlapping signal with PER2 leading to yellow and orange staining, indicating PER2 expression in glial cells (white arrows, Fig. 1A, left panel). In GPer2 mice, the red signal obtained with antibodies against the neuronal marker NeuN almost entirely overlapped with the green PER2 signal, resulting in the yellow/orange color (Fig. 1A, middle panel). This indicated that in the GPer2 animals PER2 is still present in neurons. The PER2 (green) and GFAP (red) signal did not overlap in the GPer2 animals and no yellow color was observed (Fig. 1A, right panel), strongly suggesting that PER2 was absent in glial cells. Hence, the green signal is due to neuronal PER2 expression, suggesting specificity of our approach.
To corroborate these observations, we performed immunohistochemistry in the striatum at ZT12 (Fig. 1B). The signals for PER2 (green) and GFAP (red) overlapped in the cytoplasm (yellow) and not in the nucleus (blue) of control animals (Fig. 1B, left panel). In GPer2 animals, NeuN signal (magenta) overlapped with the PER2 signal (green) resulting in the yellow color (Fig. 1B, middle). Note that there was no green signal observed that could stem from glial PER2 expression. In the GPer2 mice the PER2 signal (green) did not overlap with the GFAP signal (red) and PER2 appeared to be mainly present around the nuclei (blue) of neurons. Overall, our immunohistochemistry data indicate that PER2 was absent from glial cells of GPer2 mice.  (left panel) and GPer2 mice (middle and right panels) collected at ZT12. Sections were incubated with antibodies against PER2 (green) and glial fibrillary acidic protein (GFAP), a glial marker (red, left and right panels), or NeuN (red, middle panel), a neuronal marker. White arrows in the control panel (left) indicate overlapping PER2 and GFAP signal (yellow) in glial cells, green indicates neuronal PER2 signal and red astrocytic GFAP. In GPer2 sections, PER2 is mostly detected in neurons that are NeuN-positive, giving rise to the yellow color. In contrast, PER2 (green) is not seen in glial cells (red) of GPer2 mice, indicating a glial specific deletion of PER2. Scale bar: 100 µm. (B) Immunohistochemistry of sections from the dorsal striatum (DS) of control (left panel) and GPer2 mice (middle and right panels) collected at ZT12. Sections were stained with DAPI (blue) to reveal cell nuclei and were incubated with antibodies against PER2 (green) and GFAP (red, left and right panels) or NeuN (magenta, middle panel). The yellow color indicates overlapping signal of PER2 and GFAP in control animals (left panel) or PER2 and NeuN in neurons (middle panel). No yellow color in GPer2 sections stained for PER2 and GFAP is observed. Scale bar: 10 µm. (C) Immobility time in the forced swim test (FST) of GPer2 (GCre + Per2 fl/fl , red) and control (GCre -Per2 fl/fl , blue) animals are shown (n = 13 or 11, respectively, two-tailed t-test, **P < 0.01). (D) FST of the Credriver line (GCre + Per2 WT) and the Per2 control line (GCre -Per2 WT) are shown (n = 6, two-tailed t-test, no significant difference is observed). (E) FST of Per2 mutant (Per2 Brdm1 , purple) mice and their littermate controls (WT, green) show a significant difference as previously observed 23 (n = 13 and 8, respectively, two-tailed t-test, ***P < 0.001). (F) Representative swimograms of GPer2 (GCre + Per2 fl/fl , red) and control (GCre -Per2 fl/fl , blue) animals. Note the longer stretches of immobility in the control animals (right panel, blue). (G) Representative swimograms of Per2 mutant (Per2 Brdm1 , purple) mice and their littermate controls (WT, green). Note the longer stretches of immobility in the control animals (right panel, green). www.nature.com/scientificreports/ GPer2 mice display reduced despair. The glial Per2 knock-out animals (GPer2) were tested for despairbased behavior, which is one of the important manifestations of depression. To this end, we used the forced swim test (FST) to assess time of immobility at ZT6 23 . We observed that GPer2 mice (GCre + Per2 fl/fl ) were significantly less immobile compared to control animals (GCre -Per2 fl/fl ) (Fig. 1C). The Cre-driver animals used for crossing (GCre + Per2 WT and GCre -Per2 WT ) showed an immobility time comparable to the control animals (GCre -Per2 fl/fl ) (Fig. 1D). This illustrated that the lower immobility observed in the GPer2 mice was specific to the lack of Per2 in glial cells. As a positive control, we reproduced the lower immobility phenotype of Per2 Brdm1 mutant mice (Fig. 1E) as described previously in male mice 23 , but this time using female animals to demonstrate that the phenotype was not gender specific. Figure 1F shows examples of primary data depicted as mobility over time (swimograms) with the corresponding color-coding. The data were obtained using a self-developed movement analysis software to determine mobility in an automated unbiased manner (see methods). It was evident that the control animals displayed long stretches of continuous baseline signal, which corresponded to long time stretches of immobility ( Fig. 1F,G, right panels). These long stretches of immobility were very short or absent in both the GPer2 (GCre + Per2 fl/fl ) and Per2 Brdm1 mutant mice (Fig. 1F,G, left panels). Taken together, our data show that lack of Per2 in glial cells was sufficient to reproduce the manic-like phenotype observed previously in wholebody Per2 mutant mice 23 . This suggests that Per2 in glial cells may play an important role in the development of despair-based behavior contributing to depression.

Deletion of Per2 in glial cells using adeno-associated virus (AAV)-delivered
Cre recombinase. The Gfap-Cre driver mouse line used above has been reported to potentially express Cre in some neuronal progenitor cells 24,40 . Furthermore, the Cre recombinase is most often inserted randomly into the genome and its correct position and number of copies are unknown. Additionally, Cre leakage was observed in some cases leading to unwanted recombination events 29,46 . Due to these potential problems and to exclude developmental effects in our deletion approach above, we delivered into adult mice a construct expressing the codonimproved Cre-recombinase (iCre) using engineered adeno-associated viruses (AAVs). An engineered AAV for efficient non-invasive gene delivery that can cross the blood-brain barrier (BBB) was used 11 . This AAV-PHP.eB contained a vector expressing iCre under control of the human Gfap promoter with an enhanced green fluorescent protein (eGfp) as reporter. After intravenous (i.v.) injection via the lateral tail vein, fluorescence in the brain was detectable after 3 weeks in both the synthetic CAG-driven positive control 35 , as well as in the Gfap-driven constructs ( Fig. 2A, red and yellow signals). No fluorescence and hence no BBB permeability was detected when using the natural variant AAV9 to deliver the CAG-driven control. Similarly, the non-injected control brain showed only baseline signal ( Fig. 2A, brown). Intraperitoneal (i.p.) injection of the AAV-PHP.eB Gfap-driven Cre-construct also showed only baseline signal after 3 weeks. After 2 months, however, even i.p. delivery of the AAV-PHP.eB Gfap-driven construct resulted in detectable fluorescence, comparable to that of the i.v. injection (Fig. 2B, right), which is quite remarkable and is shown here to work for the first time. We termed the AAV mediated deletion of Per2 in glial cells vGPer2.
Next, we wanted to check if the approach was specific for glial cells (astrocytes and microglia). We used flow cytometry and fluorescence-activated cell sorting to separate microglia (CD11b + ), astrocytes (GLT1 + ) and neurons (CD90.2 + ). In a first step, the cells were depleted from debris (Fig. 2C, left panel), then the CD11b + microglia were removed (Fig. 2C, middle panel) and subsequently the GLT1 + astrocytes were separated from the CD90.2 + neurons (Fig. 2C, right panel). The presence or absence of iCre was then determined by PCR on the different Figure 2. Adeno-associated virus (AAV)-mediated deletion of Per2 in glial cells of adult mice leads to a depression resistant behavior. (A) Fluorescent imaging of whole brains 3 weeks after no injection (left), intravenous (i.v.) injection of the engineered AAV-PHP.eB, which can pass the blood-brain barrier (BBB), containing the general CAG driver (second from left) or the glial Gfap driver (middle). The second to last brain is from an animal with i.v. injected AAV9, which does not pass the BBB, containing the general CAG driver. The last brain (right) is from an animal injected intraperitoneally (i.p.) with the AAV-PHP.eB Gfap-driven construct. Note that only the brains of animals that received the AAV-PHP.eB i.v. display significant fluorescent signal after 3 weeks (orange and yellow color). (B) Fluorescent imaging of whole brains 2 months after injection of the AAV-PHP.eB. Note that the fluorescence is still maintained after 2 months post injection and that even the i.p. injected AAV-PHP.eB Gfap is showing signal in the brain now. (C) Sorting of neurons and astrocytes by flow cytometry from brain tissue including the nucleus accumbens (NAc). The left panel shows the removal of debris from a single cell suspension, showing the distribution of debris in the forward as well as in the side scatter (FSC and SSC, respectively). The middle panel shows the removal of CD11b + cells (microglia) from the cell suspension. The CD11bcells (bottom half from middle panel) were then sorted into two distinct cell populations corresponding to astrocytes (GLT1 + /CD90.2 -) and neurons (CD90.2 + /GLT1 -) (right panel). (D) PCR analysis of astrocytes and neurons from the cell sorting. Microglia (CD11b + ) as well as astrocytes (GLT1 + /CD90.2 -), but not neurons (CD90.2 + /GLT1 -) from PHP.eB Gfap-iCre infected animals express iCre, indicating that only glia and not neurons could express iCre in order to delete Per2 in the Per2 fl/fl mice. All probes and markers were loaded on the same gel and the resulting photograph was not separated and recomposed. (E) Immunohistochemistry of vGPer2 brain tissue from nucleus accumbens (NAc) isolated at ZT6. The signal for PER2 (green) mainly overlaps with neuronal NeuN signal (red) giving rise to the yellow color. Scale bar: 100 µm. (F) Immobility time in the forced swim test (FST) of vGPer2 (PHP.eB Gfap-iCre, green) and control (PHP.eB control, blue) animals are shown (n = 7, two-tailed t-test, **P < 0.01). (G) Representative swimograms of vGPer2 (AAV-PHP.eB Gfap-iCre, green) and control (AAV-PHP.eB control, blue) animals. Note the longer stretches of immobility in the control animals (right panel, blue). www.nature.com/scientificreports/ cell populations. We observed that the PHP.eB Gfap-iCre infected microglia (CD11b + ) and astrocytes (GLT1 + ) expressed iCre. In contrast, neurons (CD90.2 + ) and non-infected controls did not express iCre (Fig. 2D, full size gel in Suppl. Figure 1A). Unfortunately, the sensitivity was not sufficient to reliably detect Per2 expression in WT non-infected controls using this approach. However, immunohistochemistry on nucleus accumbens tissue isolated at ZT6 confirmed that PER2 protein could still be detected in neurons, while outside of neurons PER2 was barely or not detected (Fig. 2E).
vGPer2 mice display reduced despair. The animals with intravenously (i.v.) applied AAV-PHP.eB, which mediated the deletion of Per2 in glial cells (vGPer2), were subjected to the FST in order to assess their immobility in this despair-based behavioral test. We observed that vGPer2 (PHP.eB Gfap-iCre) mice showed significantly lower immobility times compared to control animals, which were injected with a comparable virus, but lacking iCre (PHP.eB control) (Fig. 2F). Examples of swimograms illustrating swimming behavior between minute 2 to 6 are shown (Fig. 2G, Suppl. Figure 2). Longer resting bouts were only detected in control animals ( Fig. 2G, right panel, Suppl. Figure 2). This was consistent with the result we obtained by cross-breeding Per2 floxed with Gfap-Cre mice (Fig. 1C, F, G), indicating that the phenotype can be reproduced by deleting Per2 in glial cells of the adult animal. Hence, developmental processes are very unlikely to be responsible for this reduced despair phenotype. GPer2 mice display normal circadian parameters. Since GPer2 mice display less immobility in the FST and swim more, we tested these animals for the circadian parameters of total activity, circadian period and body temperature fluctuation. We measured running-wheel revolutions in the cages of GPer2 and control animals and observed identical activity profiles with low activity during the 12 h light phase and high activity during the 12 h dark phase (Fig. 4A). General activity was measured via an intraperitoneally implanted transmitter that was traced by a detector plate under the floor of the cage. Comparable to the wheel-running activity ( Fig. 4A), no difference between the genotypes was observed (Fig. 4B). However, the activity in the second half of the dark phase (ZT16-22) was for both genotypes higher in the wheel-running assessment compared to the general activity pattern. These results show that the reduced immobility time of GPer2 animals in the FST was not due to a higher activity level compared to controls and, hence, the higher activity in the FST is related to despair rather than to general activity. Using the wheel-running activity data, we also determined the circadian period (τ) of the GPer2 and total Per2 knock-out (KO) animals under constant darkness conditions. The circadian period of GPer2 mice was normal and not significantly different from control animals (Fig. 4C, D, E), indicating that loss of Per2 in glial cells does not affect circadian period. In contrast, whole-body Per2 KO mice displayed a period shorter than 23 h (Fig. 4F), consistent with previous findings 13,51 .
Next, we assessed body temperature over 24 h under a 12 h light : 12 h dark cycle (LD 12 :12). No significant differences in the body temperature profiles were observed between GPer2 and control animals, in both males (Fig. 4G) as well as females (Fig. 4H). This is consistent with the phenotype in Per2 Brdm1 whole body mutant mice at 21 °C 12 and liver, neuronal and total body Per2 knock-out animals 13 .
Taken together, our data illustrate that loss of Per2 in glial cells does not phenocopy all the characteristics of Per2 whole-body KO animals, such as a shortening of the free-running period. This highlights a specific role of glial Per2 in mood-related behaviors.

Molecular changes in GPer2 mice.
Mood-related behaviors including depression are regulated by a number of different brain nuclei and regions 4,5,7,19,20,54 . We investigated three of those brain regions, the nucleus accumbens (NAc), the medial prefrontal cortex (mPFC) and the amygdala (AMY), as well as the hypothalamus (HYP) as a control brain region. The aforementioned brain regions also showed altered activity in response to swim stress 19 . We focused our attention on expression of genes involved in the synthesis, reuptake and degradation of monoamine neurotransmitters, as well as on genes involved in the clearance of glutamate and GABA from the synaptic cleft, because some of these processes were observed to be altered in Per2 mutant mice 23,47 . Special attention was given to astrocyte-specific genes (Supplemental table 2).
In the hypothalamus (HYP), GABA transporter 1 (Gat1/Slc6a1) mRNA was significantly decreased in GPer2 (GCre + Per2 fl/fl ) compared to control (GCre -Per2 fl/fl ) animals (Fig. 5A). In the nucleus accumbens (NAc), mRNA expression of GABA transporter 2 (Gat2/Slc6a13) and dopamine receptor D3 (Drd3) was increased in GPer2 (GCre + Per2 fl/fl ) compared to control (GCre -Per2 fl/fl ) mice (Fig. 5B, C). In contrast, Gat1/Slc6a1 and Gat3/Slc6a11 were unaltered (Supplemental table 1). Interestingly, genes coding for enzymes involved in monoamine synthesis, such as tyrosine hydroxylase (Th) and monoamine degradation, such as monoamine oxidases A (Maoa) and B (Maob), as well as catechol-O-methyltrasferase (Comt) were similar between the two genotypes in the NAc (Fig. 5D-G), while Per2 mRNA was significantly reduced in GPer2 mice (Fig. 5H). These results suggest that Per2 in glial cells is involved in the regulation of GABA signaling, rather than the regulation of monoaminergic signaling. This was further underlined by our findings on the analysis of neurotransmitters. We found that glutamate www.nature.com/scientificreports/ (Glu) levels were significantly decreased in the NAc of GPer2 (GCre + Per2 fl/fl ) mice (Fig. 5I). This difference, however, was not significant in the other brain areas investigated (dorsal striatum (DS), medial prefrontal cortex (mPFC), Supplemental table 3). We did also not observe changes in GABA and glutamine in the NAc, DS and mPFC (Supplemental table 3). Taken together, our molecular studies examining 4 regions of interest suggest that lack of Per2 in glia affects mostly the NAc, and there the signaling pathways involving GABA and glutamate, as well as signaling via DRD3. NAc is therefore likely one of the regions involved in the observed behaviour.
Deletion of Per2 in glial cells of the NAc is sufficient to elicit a reduced despair phenotype. Since we observed most of the molecular changes in the NAc of GPer2 animals, we wondered whether deletion of Per2 in glial cells of the NAc alone could elicit the phenotypes described above. Therefore, we injected our viral vectors directly into the NAc of Per2 fl/fl mice (Fig. 6A) using a stereotactic injection apparatus and termed the animals GNAcPer2. We saw that both AAV vectors, the PHP.eB Gfap-iCre as well as the AAV9 Gfap-iCre significantly reduced immobility time in the FST (Fig. 6B, C). The results were comparable to the observations in GPer2 and the vGPer2 animals, in which Per2 was deleted in glial cells throughout the brain ( Fig. 1 and  2). In contrast, we did not observe an effect on the time spent in the open section of the O-maze when AAV9 Gfap-iCre was injected into the NAc (Fig. 6D), suggesting that glial Per2 in the NAc was not involved in the regulation of anxiety-related behavior. www.nature.com/scientificreports/ Taken together, our data provide evidence that Per2 expression in glial cells of the NAc is responsible for the regulation of despair-based behavior.
The phenotype of GPer2 mice is not the result of a defective glial molecular clock. To answer the question whether the observed phenotype of GPer2 mice comes from a defect of the molecular clock in glia, we used the aforementioned viral approach to generate mice lacking glial Bmal1 (vGBmal1), an essential molecular clock gene (Fig. 7A, full size gel Suppl. Figure 1B). These mice showed no changes of immobility time in the FST (Fig. 7B) and no difference in time spent in the open section of the O-maze (Fig. 7C). Gat2/Slc6a13 was not significantly decreased, while Drd3 expression was significantly lower in the NAc (Fig. 7D-E). The behaviour of vGBmal1 animals shows that the phenotype observed in GPer2 mice is not a direct result of a defect of the molecular clock, but rather a change in the output of the clock involving the Per2 gene. This is probably related to the function of the PER2 protein as a nuclear receptor co-regulator 44 .
To test whether the observed reduced despair and anxiety are specific to glial Per2 function, we produced mice lacking Per2 in neurons by utilizing the PHP.eB vector for genetic manipulation in adult animals. We systemically  Figure 1C). To our surprise, these mice also exhibited a lowered immobility time in the FST (Fig. 7G), albeit to a smaller degree than vGPer2 mice (Fig. 2). However, we did not observe any change in the O-maze test (Fig. 7H). Interestingly, these mice also showed an increase of Gat2/Slc6a13, but normal expression of Drd3 in the NAc (Fig. 7I-J). Syn1-driven neuronal Per2 knock-out in adult animals was also associated with low activity levels under LD conditions (Fig. 7K). Interestingly, some mice showed difficulties in entrainment to the LD light cycle (Fig. 7L) compared to normal activity patterns of mice injected with an equivalent virus lacking iCre (Fig. 7M). This was an observation that we did not expect, as cross-bred neuronal Per2 KO mice (Nestin-driven Cre) show normal activity patterns under LD conditions 13 .

Discussion
Mood-related behaviors including depression are determined by several and sometimes synergistic factors illustrating the multifactorial nature of these neurological disturbances. In addition to the molecular complexity of these processes, cellular heterogeneity and topological organization complicate the understanding of the functioning of the brain even further. Previous studies have shown that at the molecular level genes involved in the regulation of the circadian clock can affect mood-related behaviors in mice 15,23,34 and most likely also in humans 28,38 . Anatomical and genetic tracing studies have also revealed particular brain regions to be involved in various aspects of mood regulation.
Our previous studies have implicated a role of the clock gene Per2 in mood-related behavior involving the mesolimbic dopaminergic system 23 . However, the contribution of glial cells in this process is poorly understood 52 . Therefore, we were interested to investigate the role of glial Per2. We used three approaches to investigate this question. First, we used genetic tools to delete Per2 specifically in glial cells, including astrocytes (Fig. 1). Second, we deleted Per2 in glial cells of the adult animal via systemic application of an engineered AAV that could pass the BBB (Fig. 2). This experiment was performed to exclude potential developmental effects that could contribute to behavioral alterations in the genetic approach. Third, we deleted Per2 specifically in glial cells of the NAc by injection of AAVs expressing Cre recombinase. All three approaches, the genetic as well as the AAV approaches revealed that absence of Per2 in glial cells, in particular the NAc, reduced the immobility time of mice in the FST, a test that assesses despair-based behavior, one of the aspects of depression. Glial deletion of Per2 resulted in more active swimming in the FST, correlating with a manic behavior as we have previously observed in wholebody Per2 mutant mice 23 . Interestingly, pharmacologic glial ablation in the prefrontal cortex has been reported to increase immobility in the FST and as a consequence made the animals more depressive 5 . This phenotype is the opposite from what we observed in our GPer2 and vGPer2 mice and indicates that in our animal models glial cells are still functional and not eliminated. Hence, our observations are specific to glial Per2 function and are very unlikely related to processes leading to glial cell death.
The processes affected by lack of Per2 in glial cells are most likely not related to the circadian clock mechanism, because GPer2 animals display no abnormalities in clock parameters such as period or activity distribution and body temperature fluctuations (Fig. 4). If the disturbance of the clock mechanism in astrocytes would play a major role, a desynchronization between astrocytes and neurons would be expected, which would ultimately lead to arrhythmic activity of mice under constant darkness conditions. This is, however, not what we observed (Fig. 4). Furthermore, deletion of Bmal1 in glial cells did not affect immobility time in the FST (Fig. 7), suggesting that the clock itself does not cause the phenotype we observed in glial Per2 knock-out animals. Therefore, we www.nature.com/scientificreports/ hypothesize that the phenotypes we describe here are probably related to PER2 output functions, e.g. as nuclear receptor co-regulator 44 . PER2 may potentially act as a nuclear receptor co-regulator, which would correlate with the altered levels of Gat2/Slc6a13 and Drd3 mRNA that we observed in the NAc (Fig. 5B). GABA transporters as well as Drd3 have been implicated in mood-related behavior. Drd3 has been shown to be regulated by the transactivator RORα and repressor REV-ERBα 25 , both of which bind to the same promoter elements (ROREs) and of which REV-ERBα is modulated by PER2 44 . However, the regulation of Gat2/Slc6a13 remains elusive. Since depression is not only related to despair, we also tested our animals in the elevated O-maze. This test addresses anxiety-related behavior, which is an important component of depression. We observed that both GPer2 as well as vGPer2 mice were less anxious, because they spent more time in the open area of the O-maze (Fig. 3). Interestingly, however, the animals with Per2 deleted in the glia of the NAc only (GNAcPer2) did not show this phenotype (Fig. 6D). From this we conclude that Per2 in glia of the NAc may not be important for anxiety-related behavior and specifically affects despair. Hence, glial Per2 in other brain regions, such as the amygdala 17 , the bed nucleus of the stria terminalis (BNST) 41 or the subgenual anterior cingulate cortex (sgACC) 3 , part of the medial prefrontal cortex (mPFC), may be involved in the modulation of anxiety. Interestingly, vNPer2 mice showed no difference in anxiety perception (Fig. 7H) and it seems that the reduced anxiety is mediated mainly through glial Per2. This suggests that distinct brain regions and also different underlying mechanisms lead to reduced despair and anxiety in GPer2 and vGPer2 mice. Whereas the vNPer2 mice also showed increased Gat2 levels, as observed in GPer2 mice, they had no alterations in Drd3 expression. . All probes and markers were loaded on the same gel and the resulting photograph was not separated and recomposed. In comparison to vGBmal1, vNPer2 mice have (G) a reduced immobility time in the FST (n = 11 and 12, respectively, two-tailed t-test, *P < 0.05), but (H) show no difference to controls injected with an equivalent virus, but lacking iCre, in the O-maze (n = 11 and 12, respectively). (I) Their Gat2/Slc6a13 expression is upregulated (n = 3, two-tailed t-test, **P < 0.01), while (J) the Drd3 levels are unaltered in the NAc (n = 3). (K) A representative vNPer2 mouse shows reduced locomotor activity in a wheel-running cage, as represented with an actogram, where each line represents a day of activity, the latter illustrated with the height and frequency of vertical lines. (L) Some KO mice also showed difficulties entraining to the light cycle. (M) The mice injected with the control virus, which lacked iCre, showed normal locomotor activity. www.nature.com/scientificreports/ We did not test anhedonia, another aspect of depression, because whole-body Per2 mutant mice did not show any changes in the sucrose preference test 47 . Therefore, we assumed that glial Per2 plays a subordinate or no role in the anhedonia aspect of depression.
Our previous studies describing whole-body Per2 mutant mice highlighted a functional involvement of Per2 in the regulation of monoamine degradation 23 . In this study, we find that glial Per2 appears to play a role in the regulation of glutamate, GABA and dopaminergic signaling, but not monoamine metabolism (Fig. 5), at least not at the observed time-point. There is ample evidence that both, monoaminergic and GABAergic signaling regulate depressive disorders 31 , and Per2 might be involved in both. If this hypothesis was correct, then we would expect that deletion of Per2 in neurons would still affect mood-related behaviors. Our results are consistent with this view, because vNPer2 mice show reduced immobility in the FST (Fig. 7F).
Alterations of glial function are considered to modify glutamate reuptake 14 , which may be the reason why we observed reduced glutamate levels in the NAc of our GPer2 knock-out animals (Fig. 5I). Interestingly, however, we did not observe changes in expression of glutamate transporters Eaat1 and Eaat2 (Supplemental table 2). This is in contrast with our previous study investigating whole-body Per2 mutant mice, which show decreased Eaat1 levels 47 . Hence, it appears that lack of Per2 function in neurons and astrocytes leads to changes in Eaat1 expression, illustrating the complex interplay between neurons and astrocytes in the regulation of glutamate signaling and mood-related behavior.
Overall, our results provide evidence for a specific role of glial Per2 in mood-related behavior, accompanied by dysregulation of components of the glutamatergic, GABAergic and dopaminergic signaling pathways.

Materials and methods
All experiments were approved by the Bundesamt für Umwelt BAFU and the State Veterinarian of the Canton of Fribourg.
Animals and housing. All mice were housed with food and water ad libitum in transparent plastic cages (267 mm long × 207 mm wide × 140 mm high; Techniplast Makrolon type 2 1264C001) with a stainless-steel wire lid (Techniplast 1264C116), kept in light-and soundproof ventilated chambers. All mice were entrained to a LD 12:12 cycle, and the time of day was expressed as zeitgeber time (ZT; ZT0 lights on, ZT12 lights off). Twoto four-month-old males and females were used for the experiments unless otherwise stated. Males were used for the FST, while females were used for the O-maze and mixed genders to monitor activity. Housing as well as experimental procedures were performed in accordance with the guidelines of the Schweizer Tierschutzgesetz and the Declaration of Helsinki. The state veterinarian of the Canton of Fribourg approved the protocols. The study was carried out in compliance with the ARRIVE guidelines.
For intravenous delivery of viruses, mice were sedated and chemically restrained with an intraperitoneal injection of 40 mg per kg body mass ketamine and 0.15 mg per kg medetomidine in saline and placed on a heating pad, while hydrogel was applied to their eyes. We injected 10 11 viral genomes per mouse in a total volume of 200 μl PBS via the lateral tail vein. The sedation was reversed with atipamezole in 5 times the dose of medetomidine.
Whole-brain fluorescent imaging was performed using the IVIS Lumina II (Caliper LifeSciences) and the accompanying Living Image software (version 4.2.0.14335). Settings: excitation 465 nm, filter GFP, exposure time 1 s, light level low, binning low/small. Forced swim test (FST). The Porsolt's forced swim test has been described in detail elsewhere 23,39 . Briefly, the mice were placed into a cylinder filled with water, where they were left for 6 min. The first 2 min were discarded as adaptation period, while the following 4 min were scored. The test was performed for four consecutive days. The first day is regarded as adaptation and was discarded. The following three days were manually scored for immobility time -the amount of time a mouse passively floated during the 4 min test window. An average immobility time for each mouse was calculated and then these were pooled according to the assigned experimental group. In the case of animals injected with the PHP.eB virus, the forced swim test was performed 4 weeks post injections 10 . All forced swim tests were performed at ZT 6, 6 h after lights were switched on in a 12 h light, 12 h dark environment 23 . The water temperature was always adjusted to 26 + /− 1 °C.
The FSTs were recorded from the side view. The automated forced swim test analysis was performed with a custom written program that evaluated the relative pixel difference in a rectangular area spanning the width of the swim tank and having the height equal to one sixth of the width, measured from the water surface towards the bottom of the tank. Each pixel in the test area was evaluated for changes between frames in red, green and Surgical procedures. The surgical procedures were performed as previously described 33 . Briefly, mice were anesthetized with 80 mg per kg ketamine and 0.30 mg per kg medetomidine, while the anesthesia was reversed by atipamezole in 5 times the dose of medetomidine. Before surgical procedures, the depth of anesthesia was checked by an absence of a reflex when pinching the skin between the toes of the mouse 5-10 min after application of the anesthetic. In case of a reflex, a quarter of the initial dose of the anesthetic was additionally administered. For measurements of internal body temperature and general activity, a wireless biochip (Vital-View system) was implanted into the abdominal cavity (Starrlife Sciences, VitalView Data Acquisition System, Instruction Manual, Software Version 5.1). Stereotactic injections were performed with a pulled glass pipette, which allowed injections of 2 × 200 nl of the virus bilaterally into the nucleus accumbens (NAc) with stereotactic coordinates of 1.60 anterior-posterior, + /-1.00 medio-lateral and 4.50 dorso-ventral. Surgical procedures were followed by administration of carprofen at 10 mg per kg.
Immunohistochemistry. Brains were harvested after mice were cardiovascularly perfused with saline and 4% PFA, and the tissue was left in PFA over night, and then transferred to 30% sucrose in PBS for two days for cryoprotection. 40 µm sections were cut using a cryostat and kept at -20 °C in an antigen preservation solution (1% m/m polyvinyl pyrrolidone in a 1 : 1 mixture of PBS and ethylene glycol) until subsequent steps of washing, blocking with permeabilization, antigen retrieval and antibody staining (anti-PER2, Alpha Diagnostic, cat. PER21-A, 1 : 200 dilution; anti-GFAP, Abcam, cat. ab53554, 1 : 500 dilution; anti-NeuN, Merck Millipore, cat. mab377, 1 : 250 dilution). The procedures have been described by our laboratory in detail elsewhere 9 . Images were acquired using a Leica SP5 confocal microscope.
Wheel-running experiments, general activity and internal body temperature measurements. Wheel-running experiments were performed using custom built cages, according to local legislation on animal experimentation, which had a stainless-steel wire running wheel with a diameter of 11.5 cm. On the axis, a system with a magnet closed a switch for each wheel revolution. The data was digitalized using an interface from Actimetrics and the activity was recorded and processed using the ClockLab software version 6.0.54. This allowed acquisition of activity patterns in LD conditions, as well as monitoring of the subjective day length under constant conditions. For this, the mice were entrained to LD conditions and then released into total darkness for a minimum of 10 days. The first three days were discarded as assimilation, while the subsequent 7 days were evaluated for a shift of activity onset on each consecutive day in order to give a measure of the length of the subjective day or free-running period, as previously described 27,33 .
Telemetrics was performed with the VitalView system, which is used for wireless measurements of general mouse activity, as well as internal body temperature, allowing monitoring of behavioural and physiological changes in free-running animals. For the biochip implantation, please see 'Surgical procedures' . All the procedures have been described in detail elsewhere 33 . Flow cytometry. At the desired time-point, brain tissue was harvested, dissected, stored in a test tube and submerged into cooled isopentane. The samples, still in the isopentane bath, were then stored at − 80 °C 43 . For flow cytometry, the samples were diced with razor blades and digested under heat and agitation with the Neural Tissue Dissociation Kit (T) (130-093-231) from Miltenyi Biotech according to the manufacturer's protocol. Single-cell suspensions were obtained by straining through a 70 µm mesh filter, after which the strained cells were washed twice in FACS buffer (PBS with 5% fetal calf serum and 5 mM EDTA). Then, myelin debris were removed using the Myelin Removal Beads II kit (130-096-731) from Miltenyi Biotech. Myelin-depleted cell suspensions were incubated with an anti-CD16/CD32 Fc-blocking antibody (BD Biosciences) for 30 min at 4 °C to avoid Fc receptor binding and then washed once with FACS buffer. The cells were stained with fluorophoreconjugated antibodies and analyzed on a MACSQuant flow cytometer (Miltenyi Biotech) or sorted using the FACSAria Fusion cell sorter (BD Biosciences). Flow cytometry analysis was based on viable and single-cell gaiting strategies, as previously described 45