PER2 Mediates CREB-Dependent Light Induction of Per1


 Light affects many physiological processes in mammals such as entrainment of the circadian clock, regulation of mood, and relaxation of blood vessels. At the molecular level, a stimulus such as light initiates a cascade of kinases that phosphorylate CREB at various sites, including serine 133 (S133). This modification leads CREB to recruit the co-factor CRCT1 and the histone acetyltransferase CBP to stimulate the transcription of genes containing a CRE element in their promoters, such as Period 1 (Per1). However, the details of this pathway are poorly understood. Here we provide evidence that PER2 acts as a co-factor of CREB to facilitate the formation of a transactivation complex on the CRE element of the Per1 gene regulatory region in response to light. Using in vitro and in vivo approaches, we show that PER2 modulates the interaction between CREB and its co-regulator CRTC1 to support complex formation only after a light or forskolin stimulus. Furthermore, the absence of PER2 abolished the interaction between the histone acetyltransferase CBP and CREB. This process was accompanied by a reduction of histone H3 acetylation and decreased recruitment of RNA Pol II to the Per1 gene. Collectively, our data show that PER2 supports the stimulus-dependent induction of the Per1 gene via modulation of the CREB/CRTC1/CBP complex. Remarkably, our results indicate that the molecular mechanism that transduces the light signal to the clock is similar to the one in the filamentous fungus Neurospora crassa to induce frequency (Frq). This suggests an evolutionarily conserved mechanism of this process despite the divergent sequences of the individual components.


Introduction
Light perception is one of the most important mechanisms that provoke biological responses in organisms, from resetting the circadian clock to cell division, metabolism, and redox state regulation 1-5 . In mammals, light can entrain the circadian clock, regulate mood behaviors, and blood vessel relaxation 6-8 . Light is perceived by the retina by intrinsically photosensitive retinal ganglion cells (ipRGCs) and transduced via the retinohypothalamic tract (RHT) to many different brain regions. For instance, in the suprachiasmatic nuclei (SCN) located above the optic chiasm in the ventral part of the hypothalamus, the signal is responsible for the release of neurotransmitters such as glutamate and PACAP 9 . Light-evoked signals provoke Ca 2+ in ux in SCN neurons and activate a phosphorylation cascade, including protein kinase A (PKA), protein kinase G (PKG), Ca 2+ /calmodulin-dependent protein kinase (CaMK), and mitogenactivated protein kinases (MAPK) also known as extracellular-signal-regulated kinases (ERK). This cascade nally promotes phosphorylation of cAMP response element-binding protein (CREB) at serine 133 and 142 [10][11][12] . CREB dimers recognize a speci c motif of an 8-base pair palindromic sequence (TGACGTCA) called cAMP response element (CRE) present on regulatory regions of target genes such as Per1 13,14 . In neurons, cAMP-regulated transcriptional co-activator 1 (CRTC1) is required for e cient induction of CREB target genes during neuronal activity 15,16 . Upon stimulation of L-type voltage-gated calcium channels, CRTC1 is dephosphorylated by calcineurin. Subsequently, it translocates into the nucleus, where it interacts via its N-terminal CREB-binding domain (CBD) with CREB 17 . Finally, the histone acetyltransferase CREB binding protein (CBP) binds to phospho-Ser133 CREB (pS133-CREB) 18 .
A light pulse (LP) applied to mammals in the dark phase of a 12 h light: 12 h dark cycle, where zeitgeber time (ZT) 0 is lights on and ZT12 is light off, elicits phase-shifts in locomotor activity 10,38 paralleled by induction of Per1 gene expression 6, 39 . Here we measured the expression of pre-mRNA of Per1 and Per2 genes in the SCN of wild type (wt) mice in the SCN before and after an LP of 15 minutes duration at ZT14 (Fig. 1A). The pre-mRNA of Per1 was strongly induced after the LP, whereas the pre-mRNA of Per2 was signi cantly less responsive (Fig. 1A). To corroborate this observation in vitro, we stimulated NIH3T3 cells with forskolin, which mimics the effect of phase-shifting via activation of the Protein kinase A (PKA) signaling pathway 40 . We observed that Per1, but not Per2, pre-mRNA was induced after 25 and 40 minutes of forskolin treatment (Fig. 1B). These results suggested that the Per2 gene is less inducible by light or foskolin than Per1. Because Per2 KO mice do not phase delay after a light pulse at ZT14 41 , our general question was whether PER2 was an upstream factor rather than a downstream target of stimulus-mediated clock resetting.
To challenge this hypothesis, we analyzed Per1 pre-mRNA expression in SCN samples obtained from wt and Per2 knock-out (KO) mice 42 collected either in the dark or 15 min after LP at ZT14. We observed that, in Per2 KO mice, the light inducibility of Per1 is dampened compared to wt (Fig. 1C, S1A upper panels). Although Per2 pre-mRNA was less light-inducible compared to Per1, it was not inducible in the Per2 KO mice (Fig. 1C, S1A lower panels). Subsequently, we analyzed Per1 and Per2 pre-mRNA expression in broblast cells derived from wt or Per2 KO animals. After forskolin treatment of the cells we observed induction of Per1 but not Per2 after 25 min. in wild-type cells (Fig. 1D, S1B, upper panels). In contrast, Per1 was not induced in Per2 KO cells, but Per2 was slightly increased after 40 min. (Fig. 1D, S1B, lower panels). These results are in line with the observation that in the SCN Per1 induction is affected by the loss of Per2. Taken together, our observations suggest that Per2 is important for the regulation of the light-and forskolin-dependent induction of Per1 expression. Of note, other clock genes did not respond in the same way as Per1 after both light and forskolin stimuli, in vivo as well as in vitro (Fig. S1 C-D).
Because cFos is an immediate-early gene (IEG) and can be induced by light in the SCN 23,24 , we wanted to test whether cFos induction may also be affected by the lack of Per2. In the SCN, we observed an induction of c-Fos after an LP in wt animals and surprisingly an even stronger induction in Per2 KO mice (Fig. 1E, S1E). Similarly, forskolin treatment caused stronger cFos induction in broblasts after forskolin Per2 KO broblasts than in wt cells (Fig. 1F, S1F). These results suggest that Per2 plays an important role in regulating light/forskolin responsive genes, which is different from its role as a circadian regulator. Furthermore, Per2 can modulate light/forskolin responses in a positive (e.g., Per1) or negative (e.g., cFos) fashion.
PER2 physically interacts with CREB and modulates its binding to a CRE element in the Per1 regulatory region Light and forskolin activate signaling pathways that lead to CREB phosphorylation at serine 133 (Ser-133) and subsequently evoke target gene expression such as cFos and Per1 10-12 . The Per1 gene contains a CREB response element (CRE) 14 (Fig. S2A) and is regulated by CREB 14,43 . Since induction of Per1 gene expression by light or forskolin is affected by the presence or absence of the Per2 gene (Fig. 1C, D), we set out to test whether the PER2 protein affects CREB binding to the Per1 promoter. For this purpose, we studied the CRE-element present in the Per1 regulatory region, including intron 1 (Fig.   S2A) that was shown to be functional 14 . We applied forskolin to wt and Per2 KO broblast cell lines and collected samples at various time points. Then, we performed chromatin immunoprecipitation (ChIP) using an antibody against CREB, followed by RT-qPCR to measure CREB occupancy at the CRE-element of Per1 and an unrelated element in the Per1 promoter as control (Fig. S2A). We found that forskolin modulated CREB occupancy at the intronic CRE-element of the Per1 gene in a time-dependent fashion with a peak at 25 min after the stimulus. On the other hand, this time-dependent change was absent in the control region ( Fig. 2A, black columns). This transient CREB occupancy increase mirrored the Per1 pre-mRNA pro le observed in Fig. 1B. Hence, it appeared that CREB was not recruited in a constant fashion to the Per1 promoter but bound the chromatin upon speci c stimulation. Surprisingly, lack of Per2 led to an increase of CREB binding to the Per1 promoter already after 10 min ( Fig. 2A, grey columns), but did not further increase after 25 min. This indicated that the two pro les of the occupancy of the regulatory region was different between the two genotypes (two-way ANOVA, p < 0.01). Taken together, these results indicate that PER2 modulates CREB recruitment onto the chromatin of Per1 in a stimulusdependent manner.
Next, we asked whether PER2 could affect CREB recruitment to the Per1 gene regulatory region via modulation of CREB phosphorylation at serine 133 (Ser-133), which is associated with light or forskolininduced phase shifts 44 . Therefore, we performed a ChIP assay using the pSer133-CREB speci c antibody as a bait. We observed that this phosphorylated form of CREB was recruited to chromatin in a similar manner as we observed when using the available CREB antibody for both wt (Fig. 2B, black columns) and Per2 KO (Fig. 2B, grey columns) derived chromatin. This result suggests that lack of Per2 did not alter the CREB binding pro le to chromatin via changes in phosphorylation of Ser-133.
Above, we described that lack of the PER2 protein affected the recruitment dynamics of CREB to the CREelement of the Per1 gene regulatory region. This effect could be of direct or indirect nature. In order to test this, we performed a ChIP assay on the same genomic region using a PER2 speci c antibody. Our experiments revealed that the PER2 pro le of recruitment to the chromatin (Fig. 2C) was similar to the one for CREB ( Fig. 2A) with a maximum at 25 min after forskolin. Interestingly, the circadian clock component BMAL1 was not recruited to the Per1 CRE element in response to forskolin (Fig. S2B). This observation indicated that the recruitment of PER2 was not due to a general effect on the circadian clock.
Since CREB and PER2 show the same chromatin binding pro le upon forskolin stimulation, we wondered whether they might physically interact. We stimulated broblast cells with forskolin and collected samples at various time points. Subsequently, we performed a Western blot (WB) and adjusted the protein loading to CREB accumulation. A proper forskolin induction was con rmed by monitoring CREB phosphorylation at serine 133, which was detected with a peak about 25 min after the forskolin stimulus ( Fig. 2D left panel). An immunoprecipitation assay (IP) followed by WB revealed that PER2 binds CREB independently from the forskolin induction and that the interaction is speci c (Fig. 2D, right panel, Fig.  S2C). Next, we investigated whether we could observe this interaction in vivo in SCN tissue of mice. We immunoprecipitated CREB from SCN protein extracts collected in the dark or 15 min after LP, followed by immunodetection. We observed that PER2 speci cally binds CREB (Fig. S2D) and that this was independent of the LP and of CREB phosphorylation (Fig. 2E), as evidenced by the PER2 signal in the absence of the pSer133 signal in the dark. Our in vitro and in vivo data suggest that PER2 physically interacts with CREB and it does so independently from the stimulus (forskolin or light). Therefore, we hypothesize that this interaction in uences CREB occupancy at the CRE element in the Per1 regulatory region.
PER2 modulates CREB: CRTC1 interaction upon light or forskolin stimulation CREB occupancy of CRE elements in various promoters has been described to involve a family of factors called CRTCs. It has been demonstrated that in particular CRTC1 localized to the nucleus and dimerized with CREB after exposure to cAMP or calcium 45 . Interestingly, CRTC1 is the main CREB co-activator regulating light responses in the SCN. Upon an LP, CRTC1 was observed to translocate into the nucleus to support CREB activity 46 . These observations prompted us to investigate whether the light or forskolin mediated effect on the induction of the Per1 gene involved CREB: CRTC1 dimerization and whether PER2 modulated this complex.
We applied forskolin to wt and Per2 KO broblast cell lines and collected samples at various time points. Then, we performed chromatin immunoprecipitation (ChIP) using an antibody against CRTC1, followed by RT-qPCR to measure CRTC1 occupancy at the CRE-element of Per1 and an unrelated element in the Per1 promoter as control (Fig. 3A). We observed that in wt cells, CRTC1 recruitment to the Per1 intronic chromatin after forskolin treatment displayed a pro le comparable to CREB ( Fig. 2A, black columns) with a peak at 25 min after forskolin treatment (Fig. 3A, black columns). Conversely, in Per2 KO cells, the recruitment of CRTC1 to the Per1 promoter after forskolin application was reduced (Fig. 3A, grey column) and comparable to the CREB pro le ( Fig. 2A, grey columns). Thus, PER2 appeared to in uence CREB and CRTC1 recruitment to the Per1 CRE element in a similar way.
Next, we investigated whether CRTC1 abundance was affected by the lack of PER2. The WB with samples obtained from wt and Per2 KO cell lines collected before and 25 min after forskolin stimulation did not show any difference in CRTC1 levels between the two genotypes. The induction mediated by forskolin increased CRTC1 amounts in both genotypes in a similar manner (Fig. 3B). Interestingly, the interaction between CREB and CRTC1 appeared to be increased by the forskolin stimulus, as shown by immunoprecipitation (IP) using an anti-CREB antibody as bait (Fig S3A). We wondered whether PER2 might affect CREB: CRTC1 interaction. IP with an anti-CRTC1 antibody as bait revealed that PER2 interacted with CRTC1 independently from the forskolin stimulus (Fig. 3C). Hence, PER2 binds to both CREB (Fig. 2D, E) and CRTC1 (Fig. 3C), although CREB and CRTC1 interact only after a stimulus 45 (Fig.  S3A). To test whether PER2 could modulate the CREB: CRTC1 interaction, we immunoprecipitated CREB in both wt and Per2 KO cell lines, without and 25 min after forskolin treatment. Surprisingly, CREB and CRTC1 interacted after this treatment only in wt cells. In contrast, in the Per2 KO cells (Per2 −/− ), the interaction was stimulus-independent ( Fig. 3D), as evidenced by the co-IP of both components before forskolin treatment.
Next, we investigated whether our observations in cells were also manifested in the SCN of mice. Similar to what we observed in the broblast cell line, CREB interacted with CRTC1 only after the light stimulus in SCN samples (Fig. S3B). Immunoprecipitation experiments showed that PER2 interacted with CRTC1 in a speci c manner (Fig. S3C). This interaction was independent of an LP, as revealed when sample loading between the two conditions was adjusted to equal amounts of CRTC1 (Fig. 3E). Furthermore, lightinduced CRTC1 accumulation was not affected by the lack of PER2 (Fig. 3F), and CREB was still phosphorylated at Ser-133 after the LP, as revealed by WB (Fig. 3F). Although CREB phosphorylation and CRTC1 accumulation were unaffected by PER2, the CREB: CRTC1 interaction was modulated by the clock factor (Fig. 3G) in a comparable manner as observed in cells (Fig. 3D) when a stimulus (light or forskolin) was applied. Additionally, CREB and CRTC1 interacted before the LP in Per2 KO mice (Fig. 3G) which was also observed in cells before the forskolin stimulus (Fig. 3D). Altogether these results de ne PER2 as a modulator of CREB: CRTC1 interaction. This modulation depends on speci c triggers, such as forskolin for cell cultures or light for the SCN cellls in animals.
PER2 mediates CREB: CBP interaction upon light or forskolin stimulation CREB binding protein (CBP) is a histone acetyltransferase that can heterodimerize with CREB when CREB is phosphorylated at Ser-133. This CREB: CBP complex then promotes gene expression 18 . To determine the CBP recruitment pro le to the CRE-element of the Per1 regulatory region we performed a ChIP assay using an antibody against CBP, followed by RT-qPCR ( to interaction with BMAL1 47 , since this region contained a BMAL1 binding site (Fig. S2B). In contrast, in Per2 KO cells CBP did not bind to the CRE-element of Per1 and to a reduced extent to the control element after forskolin treatment compared to background signal at time point 0 ( Fig. 4A, grey columns). These results suggested that PER2 affected CBP binding to both Per1 promoter regions.
Since we were interested in the CREB mediated regulation of the Per1 gene, we wondered how PER2 could affect CBP recruitment to the CRE element. We investigated whether PER2 could favor CBP nuclear localization. We tested the cells for CBP expression before and after forskolin treatment (Fig. 4B). We saw that CBP accumulation and localization in the nuclei were increased after forskolin treatment, but no distinct difference between the two genotypes was observed (Fig. 4B). We performed immuno uorescence on wt and Per2 KO broblasts to evaluate the speci city of our PER2 antibody (Fig.  S4A). Next, we tested whether forskolin treatment induced Ser133 phosphorylation of CREB in our assay using an anti-pSer133 CREB antibody. The phosphorylation of CREB was induced as expected in both wt and Per2 KO cells 25 min after but not before forskolin treatment (Fig. S4B). Since, CBP can still accumulate in the nuclei after forskolin induction in the absence Per2, we investigated whether PER2 could affect the interaction between CBP and CREB. This interaction is essential for the CBP acetyltransferase activity on chromatin. Therefore, we immunoprecipitated CREB at 0 and 25 min after forskolin treatment of cells and performed IP using an antibody against CREB followed by WB (Fig. 4C). We observed that CBP was co-precipitated with CREB 25 min after forskolin treatment.
In contrast, this co-precipitation was not observed in Per2 KO cells (Fig. 4C). Subsequently, we wanted to see whether a Per2-dependent CREB: CBP interaction could also be observed in the SCN of mice after the LP stimulus. Immunohistochemistry on SCN tissue showed that the amount of CBP was increased after LP in both wt and Per2 KO tissue, indicating that lack of Per2 did not affect the increase of CBP levels in response to LP (Fig. 4D, control Fig. S4C). Next, we immunoprecipitated CREB from SCN extracts collected in the dark and after LP at ZT14 of both genotypes (Fig. 4E). Similar to the observation in cells (Fig. 4C), CBP was co-precipitated with CREB in wt extracts but not in extracts of Per2 KO mice. Taken together, our ndings indicate that PER2 is necessary for the formation of the CREB: CBP complex after a stimulus such as forskolin or light pulse.

FRET analysis indicates that PER2 supports the interaction of CREB and CBP domains
The results presented above lend support to the notion that PER2 serves as a scaffold for CREB: CBP interaction. To further challenge this idea, we performed Förster resonance energy transfer (FRET) experiments, a widely used method to investigate molecular interactions between proteins such as CREB: CBP in living cells 48 . We used a sensor called ICAP (an indicator of CREB activation due to phosphorylation). The sensor is composed of three different elements: 1) the KID domain of CREB containing the Ser-133, which is phosphorylated upon forskolin induction, 2) the KIX domain of CBP, which is responsible for the dimerization with phospho-CREB and 3) a short linker that separates the KID from the KIX domain. KID is anked by a cyan uorescent protein (CFP), while KIX is anked by a yellow uorescent protein (YFP). When KID is not phosphorylated at the resting phase, the ICAP conformation allows CFP to transfer energy to YFP, producing FRET resulting in yellow light emission. After a stimulus (forskolin), the serine in KID is phosphorylated and binds to KIX. The dimerization separates CFP from YFP, leading to decreased FRET resulting in blue light emission (Fig. 5A upper panel).
We hypothesized that depletion of the endogenous PER2 could affect the KID: KIX dimerization (Fig. 5A bottom panel). To test this hypothesis, we co-transfected ICAP and either scrambled (scr), or Per2 directed shRNA (shPer2) into NIH3T3 cells. Subsequently, we acquired the FRET signal. 3 min after forskolin treatment, which is the standard latency time for FRET signal detection, we noted that only about 10% of shPer2 transfected cells (Per2 knock-down (KD)) were responsive to the stimulus compared to about 90% of scr transfected cells (Fig. 5B). As evidenced by single-cell traces (Fig. S5A), most shPer2 transfected cells were not responsive, but some with a substantial delay, compared to the scr controls. These results suggested that the knock-down of Per2 affected the KID: KIX interaction. Quanti cation of the FRET signal acquired over 30 min from 50 cells showed a difference in the pro les of scr and shPer2 transfected cells (Fig. 5C, S5B). 50% of the scr control cells responded to the forskolin treatment within about 3 min, while shPer2 cells needed over 10 min (Fig. 5D). Additionally, the relative KID: KIX dimerization was signi cantly higher in scr control cells compared to shPer2 cells up to around 20 min after the stimulus (Fig. 5E). Altogether these results reinforce our notion that PER2 may support the CREB: CBP dimerization via their respective KID: KIX domains.
PER2 modulates CBP-mediated chromatin acetylation at lysine 27 of histone H3 after forskolin or light stimulation CBP binds to chromatin by dimerizing with pSer133-CREB to exert its histone acetyltransferase (HAT) activity 49 . Lysine 27 (K27) of histone H3 is the main target of CBP HAT activity to acetylate (Ac) K27 of histone H3 50 . We wondered whether this epigenetic modi cation (AcH3K27) could be observed on chromatin containing the Per1 Cre-element in cells after forskolin treatment. A ChIP assay, using an anti-AcH3K27 antibody for IP, followed by RT-qPCR revealed that chromatin remodeling at that speci c amino acid of histone H3 is forskolin-dependent (Fig. 6A, black columns). We observed highest acetylation of histone H3 in the Per1 CRE-element containing chromatin between 25 and 40 min after stimulus application (Fig. 6A, black columns). Interestingly, acetylation was dramatically reduced in the Per2 KO cells (Fig. 6A, grey columns). This suggested that CBP needed PER2 as a co-factor for assembling a functional complex allowing acetylation of histone H3 on the Per1 CRE-element. This observation is the rst evidence indicating that acetylation at K27H3 is a result of PER2-dependent forskolin stimulation.
Acetylation of histone H3 leads to the assembly of the RNA polymerase II (RNA Pol II) complex to promote gene transcription 20 . Therefore, we tested whether the formation of such an RNA Pol II containing complex on the Per1 CRE-element was promoted by forskolin treatment of cells and whether PER2 may be involved in this process. We performed a ChIP assay using an antibody against RNA Pol II as bait. We noted that RNA Pol II recruitment to chromatin of wt cells (Fig. 6B, black columns) paralleled the acetylation dynamics of histone H3 on K27 (Fig. 6A, black columns). In contrast, the recruitment of RNA Pol II was lower in cells lacking Per2 (Fig. 6B, grey columns), mirroring the absence of acetylation of H3K27 (Fig. 6A, grey columns). These results provide additional evidence for the importance of PER2 as a facilitating component for assembling the stimulus-dependent transcriptional complex on the intronic CRE-element of the Per1 gene.
A previous study that described light-dependent histone phosphorylation in the SCN of mice 51 prompted us to ask whether acetylation of histone H3 at K27 in the SCN was light-dependent. We applied an LP at ZT14 to mice as described before and subsequently collected SCN tissue. We performed an immuno uorescence assay using an anti AcH3K27 antibody. At ZT14 H3 in the SCN of wt mice displayed a basal level of chromatin acetylation, which was strongly increased after application of an LP (Fig. 6C upper row). In contrast the immuno uorescence signal was partially dampened in the SCN of Per2 KO mice (Fig. 6C lower row). To con rm this observation, we performed WB on pooled SCN tissue collected in the dark or 15 min after LP from both genotypes (Fig. S6A). Quanti cation con rmed that AcH3K27 was increased right after the light stimulus in wt SCN tissue, whereas the induction was blunted in SCN extracts obtained from Per2 KO mice (Fig. 6D). Together, these data suggested that the light signaling cascade did regulate chromatin acetylation and that PER2 modulated this process in the SCN.

Discussion
In the present study, we report a role of the clock protein PER2 as a modulator of the light-dependent CREB signaling in the early night. In particular, we observed that PER2 acted as a positive factor in stimulus dependent Per1 gene expression, while it functioned as a negative regulator in cFos gene induction (Fig. 1C-F). This functional dichotomy indicated that the transcriptional complex involving PER2 is not identical on the Per1 and cFos promoters. That PER2 can act as a positive or negative regulator is consistent with previous observations that this protein has a modulatory in uence on transcriptional regulation in both directions in mammals 52,53 . Because deletion of cFos in mice attenuated behavioral responses to light only marginally 54 , we focused on the role of PER2 in the stimulus-dependent activation of the Per1 promoter. Since light activates several signaling pathways in mammals culminating in the induction of the Per1 gene 39, 6 , we investigated the role of PER2 as cofactor in CREB mediated transcription, a common downstream target of protein kinase cascades 18 .
ChIP analysis on the intronic CRE-element of the Per1 gene (Fig. S2A) revealed that CREB and its known co-factor CRTC1 are recruited in a time-dependent manner after a forskolin stimulus with a peak around 25 min. However, although CREB and CRTC1 bound to the same region in the Per2 KO cell lines, their binding was less pronounced and occurred earlier, with a peak around 10 min ( Fig. 2A, 3A). On the other hand, the histone acetyltransferase CBP and the RNA Pol II were recruited to the same element only in wt, but not in Per2 KO cells (4A, 6B). Our observations are in agreement with previous ndings that describe CREB as an activator of Per transcription 14,55 . Interestingly, PER2 could be recruited to the CRE-element of the Per1 promoter as well, with the same temporal pro le as CREB ( Fig. 2A, 2C). These results suggested that PER2 could be part of the CREB-containing transcriptional complex. The hypothesis was further corroborated by the observation of a discrete reduction and temporal pro le of CREB and CRTC1 binding to the Per1 promoter in Per2 KO cells.
Immunoprecipitation experiments revealed that PER2 co-precipitated with CREB (Fig. 2D, E) and with CRTC1 (Fig. 3C, E) in a stimulus-independent manner, suggesting that PER2 could interact with both CREB and CRTC1 before the stimulus occurred (Fig. 7, left). Interestingly, CREB bound to CRTC1 independent of a stimulus when PER2 was absent, however in the presence of PER2 this interaction became stimulus-dependent (Fig. 3D, G). In contrast, interaction between CREB and CBP did not occur in absence of PER2, but appeared in presence of it in a stimulus-dependent fashion (Fig. 4C, E). Hence, a stimulus led to rearrangement of the components CREB, CRTC1 and CBP with PER2 most likely acting as a scaffold to facilitate this rearrangement (Fig. 7, right). Thus, we conclude that PER2 is the factor that mediates the stimulus-dependent build-up of the complex facilitating CRE element-dependent transcription. The limitation of our observation is, that we do not know whether this is a general mechanism or whether this mechanism is restricted to a subset of CRE elements that are anked by speci c uncharacterized regulatory sequences that are present in the Per1 gene. However, the conclusion that PER2 appears to be the stimulus-dependent factor that facilitates gene activation in a subset of CREB dependent target genes seems reasonable. A number of previous observations relate Per gene activation to stimulus-dependent behavioral responses such as entrainment by light 41 , response to drugs 56-59 , and adaptation to temperature and humidity 60 . The role of PER2 in the CREB complex appears not to be essential (Fig. 1C). However, PER2 increases the transcriptional activation potential of the CREB complex. Since PER2 protein is expressed in a circadian fashion the transcriptional potential of the CREB complex is guided to a particular time window and boosted. Hence, PER2 seems to be important to increase environmental signal transduction to particular times of the day.
Dynamic chromatin remodeling in the SCN has been suggested to occur in response to a physiolgical stimulus such as light 51 . These ndings are in line with our observation that stimulus-dependent activation of Per1 transcription involved the acetylation of histone H3 at lysine 27 and that this acetylation was depending on the presence of Per2 (Fig. 6A). The acetylation pro le on the Per1 promoter was also mirrored by the recruitment of RNA Pol II and suggested a functional importance of this Per2 dependent H3K27 modi cation (Fig. 6B). The same mechanism is likely to be present in the SCN (Fig. 6C, D). Of note is the strong decrease of the magnitude of H3K27 acetylation in Per2 KO SCN (Fig. 6D). This can not be accounted for by an event happening only at the Per1 genomic locus. It is likely, that PER2 has a more widespread function to promote CBP action in order to acetylate additional genomic loci. Interestingly, light has been described to affect other epigenetic changes in the genome. For example, changes in day length affected the methylation pattern on chromatin in the SCN in a reversible manner 61 . Taken together, it appears that stimuli such as light can affect gene expression via acetylation, methylation, or phosphorylation to modulate transcriptional responses in order to adapt the clock to the current environmental parameters (Fig. 7).
The molecular mechanism that regulates light-dependent responses described here appears to be different from drosophila, but similar to the one described in Neurospora crassa. In this organism, the transcriptional complex transducing the light signal is composed of WHITE COLLAR 1 (WC-1), WHITE COLLAR-2, and Neurospora GCN5 like-1 (NGF-1) 32,33 . After a light pulse, the large White Collar Complex (WCC) can be rearranged in a similar way as shown in the model we propose here for the mouse (Fig. 7).
The WCC can transiently bind the promoter of target genes in a time-dependent fashion with a peak around 15 min after stimulation 62 . This dynamic is similar to the one observed here for the PER2: CREB complex (Fig. 2). In our model, the promoter occupancy of forskolin-dependent genes is transient with a peak around 25 min after the stimulus. Similarly, the WCC can be stabilized on the chromatin by the factor Submerged protoperithecia-1 (SUB1) 63 . This parallels the role of CRTC1 in our model (Fig. 7) in stabilizing CREB recruitment to the chromatin. WC-2 can be considered as a functional homolog of CREB. Both are light-associated transcription factors that require a platform for stabilization in order to bind to chromatin. WC-2 undergoes light-dependent phosphorylation steps in a time-dependent fashion 64 , similar to CREB. However, the phosphorylation of WC-2 is more stable over time than the one of CREB.
Our results highlight many functional similarities between WC-1 and PER2 in the regulation of stimulusmediated responses. The absence of a functional WC-1 affects the light-mediated gene expression 33 in the same way as PER2 does for Per1 (Fig. 1). WC-1 is necessary for WC-2 to be functional 65 . Our results show that PER2 regulates the promoter occupancy of CREB. Without PER2, CREB can recognize the CRE elements to a lesser extent ( Fig. 2A). Finally, PER2 modulates the interaction between CREB and CRTC1 to happen only after a stimulus. Interestingly, a similar role for WC-1 has been suggested for light-dependent responses. Lack of the Zinc Finger domain in WC-1 led to the light-independent activation of normally light-dependent pathways 66 .
In many organisms, chromatin modi cations can be stimulated after a signal. This usually involves speci c co-activators such as histone acetyltransferases (HATs). CBP/P300, PCAF, and GCN5 appear to be the mediators in this process 67 . In Neurospora crassa, the histone H3 acetylation at the lysine 14 was described as the rst epigenetic marker associated with light-dependent responses 32 . We observed a similar process in mice in this study. After forskolin/light stimulation, histone H3 was acetylated at lysine 27 (Fig. 6), and this acetylation depended in part on PER2. CBP/p300 targeted this speci c lysine in a stimulus dependent manner. Hence, our results suggest that in mammals a similar epigenetic acetylation event occurs as in Neurospora. This view is further supported by our observation that PER2 mediated the physical interaction between CREB and the HAT CBP, which is necessary for CBP to acetylate the target histone (Fig, 4-6). Our observations suggest that even if CREB phosphorylation at Ser-133 is a "condition sine qua non" for CBP recruitment, it seems not to be su cient if PER2 is not there. Lack of PER2 abolished CREB: CBP dimerization and was associated with less H3 Lys-27 acetylation in vivo (Fig. 6C,  D), which mirrors the mechanism observed in Neurospora. Like CBP in mammals, the histone acetyltransferase NGF-1, a homolog of the mammalian GCN5 68 , is involved in the light-dependent chromatin remodeling in Neurospora. After the stimulus, WC-1 starts to be phosphorylated and undergoes a conformational change. This event is responsible for NGF-1 recruitment onto the chromatin of target genes followed by histone H3 acetylation 33 . Lack of WC-1 impairs NGF-1 recruitment to the chromatin, and as a consequence histone H3 acetylation is abolished after a light pulse. Taken together, our data suggest that light-mediated responses are conserved from the invertebrate Neurospora crassa, to vertebrates such as mice.

Light pulse and tissue isolation
Light pulse (LP., circa 500 lux) was given at ZT14, and mice were subsequently sacri ced within 15 min.
As a control experiment, mice were sacri ced in the dark a ZT14. Brains were collected and SCN tissue isolated. For immuno uorescence experiments, mice were perfused with 4% PFA.

Cell culture
Cell lines described in the paper were maintained in Dulbecco's modi ed Eagle's medium (DMEM), containing 10% fetal calf serum (FCS) and 100 U/mL penicillin-streptomycin at 37°C in a humidi ed atmosphere containing 5% CO 2 . Forskolin stimulation (10-100M) was used to mime in vitro the molecular pathway activated by light in mice. Samples were collected at speci c time points mentioned in the text.
The Per2 KO cell line was ampli ed starting from mouse embryonic broblasts (MEFs) and split every three days until immortalization was reached.

Plasmids and transfection
The following plasmids were used for the project:

RNA extraction from cells
Cells were grown to con uency on 6 cm Petri dishes and induced with 10 µM forskolin (50 mM stock in dimethyl sulfoxide) for the indicated time. Total RNA was extracted using the Nucleospin RNA II kit (Machery & Nagel) and adjusted to 1 µg/ µl with water. An amount of 1 µg was reverse-transcribed using Superscript II with random hexamer primers (Thermo Fisher). Real-time PCR was performed using the KAPA probe fast universal master mix and the indicated primers on a Rotorgene 6000 machine. The relative expression was calculated compared to the geometric mean of expression of the inert genes Nono, SirT2, Atp5h, and Gsk3β 69 . For a complete list of primers used in the paper, please see Table 1.
RNA extraction from the SCN RNA from SCN samples was isolated using the Macherey-Nagel RNA Plus kit. Subsequently, 500 ng of puri ed RNA was used for producing cDNA by reverse transcription (Invitrogen SuperScript II). Real-time PCR was performed using the KAPA probe fast universal master mix and the indicated primers on a Rotorgene 6000 machine.

Chromatin Immunoprecipitation
Chromatin immunoprecipitation from cells was performed as described before 70 Table 2) and the immune complexes were captured with protein A agarose fast-ow beads (Sigma-Aldrich) for 1 h at RT.
Protein extraction from cells.
Total con uent cells plated in 10 cm dishes were washed two times with 1x PBS (137 mM NaCl, 7.97 mM Na 2 HPO 4 × 12 H 2 O, 2.68 mM KCl, 1.47 mM KH 2 PO 4 ). Then PBS was added again, and plates were kept for 5 min at 37°C. Cells were detached and collected in tubes and frozen in liquid N 2 . They were subsequently resuspended in Ripa buffer (50 mM Tris-HCl pH7.4, 1% NP-40, 0.5% Na-deoxycholate, 0.1% SDS, 150 mM NaCl, 2 mM EDTA, 50 mM NaF) with freshly added protease or phosphatase inhibitors, and homogenized by using a pellet pestle. Homogenates were kept in ice for 15 min followed by sonication (10 s, 15% amplitude). Right after, the samples were centrifuged for 15 min at 16,100 g at 4°C. The supernatant was collected in new tubes and pellet discarded.
Protein extraction from SCN tissue Isolated SCNs obtained from 5 different mice were pooled together according to the speci c condition (either dark or 15 min after the light pulse). The pooled tissues were frozen in liquid N 2 and resuspended in a brain-speci c lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.25% SDS, 0.25% sodium deoxycholate, 1 mM EDTA). They were subsequently homogenized using a pellet pestle kept on ice for 30 min and vortexed ve times for 30 s each time. The samples were sonicated (10 s, 15% amplitude) and subsequently centrifuged for 20 min at 16,000 g at 4°C. The supernatant was collected in new tubes and the pellet discarded.

Immunoprecipitation
Circa 400 µg of protein extract was diluted with the appropriate protein lysis buffer in a nal volume of 250 µl and immunoprecipitated using the indicated antibody (ratio 1:50) at 4°C overnight on a rotary shaker. The day after, samples were captured with 50 µl of 50% (w/v) protein-A agarose beads (Roche) for 3 hr at 4°C on a rotary shaker. Before use, beads were washed three times with the appropriate protein buffer and resuspended in the same buffer (50% w/v). The beads were collected by centrifugation and washed three times with NP-40 buffer (100 mM Tris-HCl pH7.5, 150 mM NaCl, 2 mM EDTA, 0.1% NP-40).
After the nal wash, beads were resuspended in 2% SDS 10%, glycerol, 63 mM Trish-HCL pH 6.8, and proteins were eluted for 15 min at RT. Laemmli buffer was nally added, samples were boiled 5 min at 95° C and stored at -20° C.

Western blot
The indicated amount of protein was loaded onto 10% SDS-PAGE gel and run at 100 Volt for two hours.
Once the migration was completed, we performed a semidry transfer (40 mA, 1 hour 30 s) using Hybond® ECL™ nitrocellulose membranes followed by red ponceau staining (0,1 % of Ponceau S dye and 5% acetic acid) to validate the success of the transfer. The membrane was subsequently washed with TBS 1x/Tween 0.1% and blocked with TBS 1x/Milk 5%/Tween 0.1% for 1 hour. After washing, the membrane was stained with the appropriate primary antibodies ( were added to the cells. To avoid buffer perturbations due to the addition of the drug stimulus, the latter was added between frames, so it had approximately 20 s to diffuse and activate a response in the cells.

FRET imaging analysis
The time-lapse recordings were analyzed using LAS X software (Leica), adapting a previously described method 72      two genotypes, *** p < 0.001. (C) Immunohistochemistry (IHC) on wt and Per2 KO SCN before and after a light pulse (LP) at ZT14. AcH3K27 (red signal) is stronger after LP in wt compared to Per2 KO SCN. Scale bar: 121 µm. (D) Quanti cation of 3 independent Western blot experiments of wt and Per2 KO SCN before and after an LP at ZT14. Acetylation of H3K27 is increased after an LP with a tendentially lower increase in Per2 KO. Two-way ANOVA, p = 0.059.

Figure 7
Model of PER2 modulating the assembly of the CREB/CRTC1/CBP transcriptional complex Left: In the early dark phase at ZT14, PER2 interacts with CREB and CRTC1. On the other hand, CREB and CRTC1 do not dimerize at this stage, and all the elements do not bind to the CRE element in the Per1 promoter. CBP is not recruited to the regulatory region of Per1 and the histone H3K27 is not acetylated. Per1 is not transcribed through this pathway. Right: In the SCN or in cells a light pulse or forskolin treatment initiates a kinase cascade that leads to CREB phosphorylation and assembly of the CREB: CRTC1: CBP complex which is facilitated by PER2. As a consequence of this assembly, histone H3K27 is acetylated (grey star) and the Per1 gene is transcribed. The hatched oval with question mark indicates that most likely additional factors contribute to the initiation of transcritpion by the RNA polymerase II (POLII).