Non-coding cis-element of Period2 is essential for maintaining organismal circadian behaviour and body temperature rhythmicity

Non-coding cis-regulatory elements are essential determinants of development, but their exact impacts on behavior and physiology in adults remain elusive. Cis-element-based transcriptional regulation is believed to be crucial for generating circadian rhythms in behavior and physiology. However, genetic evidence supporting this model is based on mutations in the protein-coding sequences of clock genes. Here, we report generation of mutant mice carrying a mutation only at the E′-box cis-element in the promoter region of the core clock gene Per2. The Per2 E′-box mutation abolishes sustainable molecular clock oscillations and renders circadian locomotor activity and body temperature rhythms unstable. Without the E′-box, Per2 messenger RNA and protein expression remain at mid-to-high levels. Our work delineates the Per2 E′-box as a critical nodal element for keeping sustainable cell-autonomous circadian oscillation and reveals the extent of the impact of the non-coding cis-element in daily maintenance of animal locomotor activity and body temperature rhythmicity.

vidence shows that non-coding cis-regulatory elements are as important as protein-coding sequences for determining cell identity and morphological development [1][2][3] . The degree of the effects of non-coding cis-regulatory elements on animal behavior and physiology after development, however, remains elusive. Circadian clocks generate~24 h rhythms in behavior and physiology, which allow organisms to anticipate and adjust to daily environmental changes. The rhythm generating mechanism of the circadian clock involves clock genes, which regulate their own transcription in a negative transcription-translation feedback loop [4][5][6] . In this canonical feedback model, non-coding cis-elements are fundamental to both initiating and closing the loop [7][8][9] . Genetic evidence supporting this model, however, is still exclusively based on the effects of mutations in the protein-coding sequences of the clock genes. Therefore, while it is tempting to speculate that the circadian clock-related non-coding cis-elements are operational for daily dynamic regulation of behavior and physiology, there is currently no direct evidence to corroborate this notion.
Notwithstanding the recent intriguing discoveries of circadian oscillations in peroxiredoxin superoxidation in transcriptionally incompetent anucleate erythrocytes 10 and the expression of such cycles in the neurons in the suprachiasmatic nucleus (SCN) 11 , the consensus conjecture regarding the mammalian clockwork supports the transcriptional feedback model. However, the literature addressing mutations or deletions in the protein-coding sequences of the clock genes cannot negate the possible contribution of post-translational (or non-genomic) clock mechanisms. Moreover, the extent of the impact of the cis-regulated transcriptional feedback cycle on establishing circadian rhythmicity in vivo remains the subject of considerable debate 4,12 , due in large part to the lack of genetic research on non-coding cis-element mutations.
In the current study, we develop mutant mice carrying a mutation in a circadian cis-acting element of the core clock gene Per2. Although Per2 was cloned as a secondary period gene in mammals, gene knockout studies revealed that Per2 mutant mice displayed a loss of circadian rhythmicity, revealing its prominent role in the mammalian molecular clockwork [13][14][15] . Moreover, familial advanced sleep phase syndrome in humans is attributed to a missense mutation in the Per2 gene 16 .
The Per2 E′-box sequence (5′-CACGTT-3′) located near the putative transcription initiation site 17 (-20 to -15) has been demonstrated to be the principal circadian cis-element that is sufficient to induce oscillating levels of reporter transcription via the mouse Per2 minimal promoter 17,18 . The importance of this particular cis-element is further implied by the high degree of DNA sequence conservation in its flanking region between humans and mice 17 and by extremely enriched clock protein binding to this element 19 . Publicly available genome-wide ChIPseq data highlight the predominant peak of clock protein binding activity around this E′-box (see Supplementary Fig. 1).
The present study was designed to investigate the role of this unique E′-box sequence of the Per2 promoter (hereafter, Per2 E′-box) as a potential nodal cis-element in the mammalian clockwork. By introducing a site-specific mutation at this element in vivo, we provide empirical evidence to show that the Per2 E′-box is essential for maintaining cell-autonomous circadian oscillations. The cells without the Per2 E′-box cannot maintain circadian molecular oscillations in culture conditions. At the organismal level, mice lacking the Per2 promoter E′-box show destabilized locomotor activity and body temperature rhythms under altered light conditions, including constant light (LL) and experimental jet-lag conditions. Due to compensatory mechanisms in vivo, the mutant mice kept under constant dark (DD) conditions remain rhythmic but exhibit considerably shorter circadian periods than WT mice. Our data therefore define the degree of the impact of the deletion of the Per2 E′-box on the organismal clock: The Per2 E′-box is essential for the period determination of behavioral rhythms in DD conditions and for sustaining stable rhythms under LL and jet-lag conditions. These data underscore the roles of the non-coding cis-element in the regulation of daily behavior and physiology in adulthood.

Results
Generation of Per2 E′-box mutant mice. We developed mutant mice harboring a targeted mutation in only the Per2 E′-box. To do this, we used the piggyBac (PB) transposase tool for genome engineering because it allows for seamless removal of the PBflanked marker cassette (PB-Neo) from the host genome after the mutation is introduced (Fig. 1a). Conventional methods that rely on the Cre/loxP or Flp/FRT system leave behind a single loxP-or FRT-derived ectopic sequence after marker excision. Such a footprint sequence could affect promoter architecture and/or enhancer communication of the target gene 20 . The PB transposase (PBase)/PB system circumvents this problem 21 (see Fig. 1b-d). Southern blot analysis confirmed that the PB-neomycin cassette was deleted without re-integration into the host genome (Fig. 1c). The footprint-free disappearance of the marker cassette was further confirmed by DNA sequencing (Fig. 1d). To reduce the confounding effects of a mixed genetic background, the mutant mice were backcrossed with C57BL/6J mice until microsatellite markers covering all individual chromosomes were congenic to the C57BL/6J strain (see Supplementary Fig. 2). The targeted mutation of the E′-box sequence was finally verified by DNA sequencing using homozygous mutant mice (Fig. 1d). Chromatin immunoprecipitation (ChIP) assays using homozygous mutant mouse liver, sampled at 4 h intervals for 24 h under DD conditions, revealed that the clock proteins PER1, CRY1, CRY2, and CLOCK had consistently low binding levels to the mutated Per2 promoter (Fig. 1e). In a control experiment with WT (+/+) samples, the peak binding levels of PER1, CRY1, CRY2, and CLOCK to the Per2 promoter were observed at CT12, CT4, CT20, and CT8, respectively, consistent with previous reports 19,22 . These results confirmed that Per2 E′-box function was abolished in the mutant mice.
Per2 E′-box is essential to maintain cellular circadian oscillations. To investigate the role of the Per2 E′-box in maintaining cellular oscillation, we generated primary fibroblast cultures from WT (Per2E′ +/+ ) and mutant (Per2E′ m/m ) mouse lung tissues and monitored circadian fluctuations in the PER2 protein levels over 80 h (Fig. 2a). As expected, the endogenous PER2 protein in synchronized WT cells displayed characteristic circadian oscillations in both abundance and electrophoretic mobility, which continued over multiple cycles under constant culture conditions 23 (Fig. 2a). In contrast, Per2E′ m/m cells failed to maintain normal PER2 protein oscillations (Fig. 2a). It is important to note that the mutant cells could competently form the first surge of PER2 expression after synchronization (4-12 h). However, following a subsequent increase (28 h), PER2 expression in the mutant cells remained at mid-tohigh levels and eventually lost apparent circadian variation after 60 h (see also Supplementary Fig. 3). Electrophoretic migration rhythms also disappeared from the mutant cells (Fig. 2a). The relative intensities of the three major bands of the PER2 protein remained unchanged in the Per2E′ m/m cells. The post-translational rhythms were thus affected either directly or indirectly by the mutation of the E′-box. We experimentally confirmed that the protein-coding sequences of PER2 and known clock protein kinases and phosphatases were unaffected in our mutant mice. These results indicate that the E′-box is indispensable for maintaining normal PER2 protein oscillations.
In the Per2E′ m/m cells, we observed that both Per2 mRNA and pre-mRNA (intronic RNA) levels remained constitutively high, with values exceeding those of the circadian peak in the WT cells at 72 h (Fig. 2b). Thus, Per2 transcription remains active even without this particular E′-box sequence in the promoter. Moreover, extensive mRNA profiling revealed that the effects of the mutation of the Per2 E′-box were not limited to Per2 transcription. The dysfunctional Per2 E′-box also abolished the circadian expression of other clock genes and output genes, including Per1, Per3, Cry1, Cry2, Bmal1, Nr1d1, Dbp, and E4bp4 (Fig. 2b). These pervasive effects of the mutation provide evidence that the Per2 E′-box is a fundamental cis-element that maintains normal molecular clock oscillations.  Per2 E′-box is essential to maintain molecular oscillations in the SCN. The SCN in the hypothalamus is the primary regulator of daily rhythms of behavior and physiology in mammals 24,25 . We next examined the effect of the Per2 E′-box mutation on the SCN clock. To study the molecular rhythms in the SCN, we used a Bmal1 promoter-driven luciferase reporter (Bmal1-Eluc) 26 , as it allowed the assessment of molecular rhythms that are not a simple reflection of the Per2 loop. In agreement with previous reports 27 , organotypic SCN slices prepared from control mice (Per2E′ +/+ ; Bmal1-Eluc) displayed persistent circadian rhythms of luminescence, which continued for over a week in culture (Fig. 3a, see also Supplementary Fig. 4). In contrast, all tested SCN slices from the Per2 E′-box mutant mice (Per2E′ m/m ;Bmal1-Eluc) displayed attenuated rhythms of luminescence, which were damped within 2-3 cycles (Fig. 3a). Fast Fourier transform (FFT) analysis of the de-trended waveforms of days 1.5−7.5 (Fig. 3b) confirmed the reduced rhythmicity in the Per2E′ m/m SCN. This finding indicated the relevance of the Per2 E′-box in maintaining normal molecular circadian oscillations in the SCN. Nevertheless, detectable rhythms were noted in the attenuated Per2E′ m/m SCN slices over the initial 2-3 cycles (Fig. 3a). These remaining rhythms were likely dependent on Per1, because SCN explants from Per2E′ m/m ; Per1 heterozygous null mice (Per2E′ m/m ; Per1 +/− ;Bmal1-Eluc) exhibited even fewer persistent rhythms, which were damped within 2 cycles (Fig. 3a-c). Importantly, similar attenuation of Bmal1-Eluc rhythmicity by Per2E′ m/m and further by Per1 +/− could be reproduced in cultures of lung ( Fig. 3d-f) and adrenal explants ( Supplementary Fig. 4), Destabilized organismal rhythms of animal lacking Per2 E′-box. Finally, to assess the effects of the mutation at the behavioral level, C57BL/6J-backcrossed WT and Per2E′ m/m mice were housed in a 12-h light:12-h dark (LD) cycle and then subjected to DD, LL, or experimental jet-lag conditions. Actograms of the animals in the respective conditions were acquired (see Supplementary Figs. 5-9). Interestingly, mutant mice kept in DD showed overt circadian locomotor activity rhythms, although with a significantly shortened free-running circadian period relative to WT mice (WT vs. Per2E′ m/m : 23.71 ± 0.02 h vs. 23.50 ± 0.02 h; see Supplementary Fig. 5). This contrast with the in vitro data suggests that a compensatory mechanism functions in vivo. We thus examined the circadian behavior of Per1 −/− ; Per2E′ m/m double-deficient mice (see Supplementary Fig. 6). Although their period length was unstable and changed gradually over extended exposure to constant darkness, Per1 −/− ; Per2E′ m/m mice displayed behavioral rhythms over 40 days. With analogy, Drosophila strains expressing period or timeless under a constitutive promoter were reported to show behavioral rhythms with an altered period length 12,28 . These data suggest that organisms might have a compensatory mechanism to mitigate defective transcriptional feedback regulation 4,24,29 .
Compared to the modest phenotype in DD, we observed unstable rhythmicity of the mutant mice in LL (Fig. 4a). When placed in LL for 3 weeks, WT subjects (n = 7) stayed rhythmic with a period longer than 24 h, which is consistent with data from previous reports 30,31 . However, none of the Per2E′ m/m mice (n = 9) remained rhythmic (Fig. 4a, see also Supplementary  Fig. 7); all tested mutant mice displayed behaviors characterized by a gradual decrease in the power of rhythmicity (FFT spectrogram, Fig. 4a). Since light has a negative masking effect on locomotor activity, we measured core body temperature fluctuations and confirmed that the body temperature recapitulated the reduced rhythmicity of the mutant mice in LL conditions (Fig. 4b, Rayleigh vector length of WT vs. Per2E′ m/m : 0.64 ± 0.06 vs. 0.42 ± 0.02).
The Per2 E′-box mutant mice were also distinct under experimental jet-lag conditions (Fig. 4c, see also Supplementary  Fig. 8). When the ambient LD cycle was advanced by 8 h, WT mice re-entrained progressively over 8 to 9 days 32 . In contrast, the Per2E′ m/m mice adapted to the new cycle within 2-3 days. The 50% phase-shift value (PS 50 ) 32 , measured from activity onset, indicated a rapidity of 5.33 ± 0.28 days for WT mice and 1.26 ± 0.17 days for Per2E′ m/m mice (Fig. 4c). Notably, the difference between the two genotypes was even pronounced when the animals were exposed to a single 8-h light advance, followed by DD conditions (Fig. 4d, see also Supplementary Fig. 9). Under this light regime, all tested mutant mice shifted forward by approximately 12 h, which is nearly 180°out of phase in a 24-h . The graph shows changes in FFT power in LL. **P < 0.01, two-way repeated-measures ANOVA, Bonferroni post hoc test (Per2E′ +/+ , n = 7; Per2E′ m/m , n = 9). b Representative body temperature records of Per2E′ +/+ and Per2E′ m/m mice in LL. The mean temperature of the entire data series was calculated, and the data points above the mean are plotted. Rayleigh plots show phase distribution of elevated body temperature on days 7-21 (n = 3, each genotype). Data from independent animals are color-coded. Arrow length reflects the r-value of each distribution. *P < 0.05, two-tailed unpaired t-test. c Representative records of locomotor activity (left) and plots of activity onset (middle) of Per2E′ +/+ and Per2E′ m/m mice before and after an 8-h phase advance in LD cycles. The PS 50 values represent the time required for 50% phase-shift (right). ***P < 0.0001, two-tailed unpaired t-test (Per2E′ +/+ , n = 7; Per2E′ m/m , n = 9). d Representative locomotor activity records (left), activity onset (middle), and magnitude of phase-shift (right) of Per2E′ +/+ and Per2E′ m/m mice subjected to an 8-h phase advance on day 1 and released to DD. ***P < 0.0001, two-tailed unpaired t-test (Per2E′ +/+ , n = 7; Per2E′ m/m , n = 11). The data are the mean ± s.e.m. Source data are provided as a Source Data file NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10532-2 ARTICLE cycle, relative to WT mice (Fig. 4d). The underlying mechanism of this extremely large shift is unknown. While this could be due to an indirect effect of the weak clock, a limit cycle oscillator model 33 predicts that a reduced-amplitude pacemaker in the mutant mice (reduced radius of the limit cycle) could have this effect 34 . We also noticed that after a light pulse exposure, the mutant mice show a slightly enhanced Per2 mRNA induction in the SCN (see Supplementary Fig. 10). This perhaps also partly contributes to the enhanced resetting of the mutant mice.

Discussion
Non-coding cis-regulatory elements are known to play a critical role in development, but their precise effects on daily behavior and physiology in adulthood remain elusive. The present study was designed to provide genetic evidence for the roles of the circadian cis-element in vivo. Despite the presence of compensatory mechanisms in vivo, our work shows that the Per2 E′-box is essential for maintaining optimal locomotor activity rhythms under LL conditions and enabling the phase of the clock to resist abrupt shifts in LD cycling. In general, adapting to new LD cycles is rarely instantaneous and requires repeated 24 h cycles. However, the Per2 E′-box mutant mice adapted to a new phase immediately. These data elucidate the impact of the deletion of the non-coding cis-element in daily maintenance of behavioral activity in adulthood.
In addition, our cell culture data reinforce the concept that circadian clock phenotypes are more drastic at the cellular level than the organismic level 4,24 . Per2 E′-box mutation leads to unsustainable gene expression rhythms in organotypic SCN slices and cultured peripheral tissues and fibroblasts. These observations substantiate the general conjecture that cis-regulatory element-based gene transcription is essential for sustaining cellular clock oscillations.
Not surprisingly, the overall behavioral phenotypes of the Per2 E′-box mutant mice reveal that the Per2 E′-box is not an absolute requirement for behavioral rhythm generation. Under DD conditions, its deletion affects only the circadian period (Supplementary Figs. 5 and 6), a phenotype analogous to that observed in transgenic Drosophila strains expressing period or timeless via a constitutive promoter 12,28 . At the gene expression level, Per2 expression in the liver and SCN in the DD-kept mutant mice in vivo was still rhythmic, albeit with an increased baseline (see Supplementary Fig. 11). These observations suggest that organisms have a compensatory mechanism to mitigate defective transcriptional feedback regulation. Correspondingly, previous studies demonstrate that behavioral rhythms do not necessarily reflect cellular clock phenotypes 29,[35][36][37] ; superior circadian performance of behavioral rhythmicity has been observed in several clock gene mutant mice, compared to tissue cultures obtained from the same animal strains 29 . In vivo multicellular and/or inter-organ systemic circuitry might compensate for poor core clock function within individual cells 4 . In this regard, it is worth noting that systemic extracellular signals are known to affect the Per2 promoter via non-E′-box cis-regulatory elements, such as Ca 2+ /cAMP response element (CRE) 38 and glucocorticoid response element (GRE) 39 . Extracellular circadian feedback pathways through these non-E′-box elements might contribute to the compensation mechanisms in vivo 40,41 .
Our work differs from transgenic studies in which non-native promoters are used 12,17 . Transgenic studies and rescue experiments can be affected by shorter promoter regulatory sequences and position effects. In comparison, native promoters are characterized by endogenous enhancer elements and normal chromatin structure, allowing for the preservation of regulation by epigenetic factors, which are known to be crucial for controlling transcription 5 . Under these near-native conditions, we noticed that the deletion of the Per2 E′-box leads to accumulation of Per2 mRNA and protein in cultured fibroblasts. Constitutively un-suppressed transcription of Per2 likely underlies the constant accumulation of PER2 protein and resultant compromised gene expression rhythms in the mutant cells.
Cis-element provides a place where both active and repressive transcription complexes are recruited. The net effect of its absence thus seems context-dependent. The question of this context specificity is still unresolved in the field of transcription and chromatin remodeling. Particularly, in our case, the mechanism(s) that allows the continued transcription of Per2 without relying on the E′-box is still unknown. It is possible that native chromatin structure of the Per2 might permit its transcription. It is also conceivable that other circadian cis-elements on the Per2 promoter, such as D-box and CRE, might contribute to the basal transcription of Per2 9,18,42 . In this regard, the continued expression of Per2 might reflect a confounding effect of being chronically deficient in the functional E′-box. Given that Per2 is under regulation of interlocked feedback cycles 6,9 , remaining Per2 transcription might be a homeostatic consequence of a nonfunctional clock in the mutant cells. A complete understanding of circadian regulation of Per2 in vivo under native conditions remains a challenge for future study.
Genome-wide association studies recently identified many noncoding variants that account for human chronotypes [43][44][45] . Given the physiologic relevance of the Per2 cis-element, a targeted point mutation strategy would facilitate hypothesis-driven approaches to understand the extent of the impact of the non-coding elements on the daily physiology and pathophysiology of the organism.

Methods
Generation of the Per2 E′-box mutant mice. The CACGTT E′-box sequence located in the Per2 promoter (-20 to -15; +1, the putative transcription start site 17 ) was targeted. Note that numbering of the relative position of the E′-box sequence will change according to the definition of the transcription initiation site 17,18 (see also RefSeq database in the UCSC genome browser: https://genome.ucsc.edu/). CACGTT was mutated to GCTAGT, because this change is reported to abrogate CLOCK:BMAL1-mediated activation of Per2 promoter activity in vitro 17 . We employed piggyBac (PB) transposon-based gene engineering system 46,47 , because conventional methods relying on Cre/loxP or Flp/FRT systems leave behind a single loxP or FRT-derived sequence after excision, which causes the creation of 'footprint' (residual ectopic sequence) that might affect promoter architecture and/ or enhancer communication of target gene 48 . The targeting vector to generate PB mutant allele of Per2 was constructed by using a Red/ET recombination system (Gene Bridges). The bacterial artificial chromosome (BAC) containing Per2 was obtained from BACPAC Resources at the Children's Hospital and Research Center at Oakland (RP23-343F13). With a Red/ET cloning method (Gene Bridges), a 10kb genomic region of Per2 (-5899 to +4,103; +1, the putative transcription start site 17 ) that corresponds to the region of 6.5 kb upstream (long arm) and 3.5 kb downstream (short arm) of the Per2 E′-box (-20/-15) was cloned into a vector containing a diphtheria toxin A gene (pDTA vector). Then, a DNA amplicon of Per2 (−90 to +650) that contains a mutated E′-box (GCTAGT) at −20/−15 was obtained by fusion PCR, and this was further modified by inserting a PB terminal repeat sequence (5′TR and 3′TR)-flanked neomycin cassette (PB-NEO) into the TTAA quadruplet sequence of Per2 (+600/+603). With a second-round Red/ET cloning, this fragment was recombined into the 10-kb Per2 BAC/pDTA vector. The resultant Per2 targeting vector was verified by DNA sequencing.
Cell culture and immunoblotting. WT and Per2E′ m/m primary fibroblasts were isolated from the lung tissue of adult male mice according to a protocol established by Seluanov et al. 56 . Cells used for time course assay were between passages 4 and 6. Dispersed cells were uniformly plated in 24-well plates at a density of 1 × 10 5 cells per well and cultured for 2 days prior to synchronization, for which cells were treated with dexamethasone (DEX, final concentration, 200 nM) for 3 h, followed by medium refreshment at Time0. Cells were harvested every 4 h in either TRIzol reagent (Invitrogen) for RNA analysis or 2 × Laemmli buffer for Western blot analysis. Immunoblotting was performed using an affinity-purified anti-mPER2 rabbit polyclonal antibody 23 (final concentration, 2 μg ml −1 ) and β-Actin antibody (A5441, Sigma, 1:1000).
Real-time bioluminescence recording and data analysis. Tissues from mice carrying a Bmal1 promoter-driven luciferase reporter (Bmal1-ELuc) 26 were harvested from adult (2-3 months-old) male mice between ZT10-12 and cultured according to a published method 58 . In brief, brains were sliced coronally into 400μm sections, and the paired SCN, which contains minimal volumes of the non-SCN, were cultured 59 . The lungs were cut into a small piece (~2 × 2 × 0.3 mm 3 ) 59 . The adrenal glands were sliced into a section of 0.3 mm thickness. All tissues were cultured separately on a Millicell membrane (PICMORG50, Millipore) with DMEM medium containing10 mM HEPES (pH 7.2), 2% B27 (Invitrogen) and 1 mM luciferin, in 35-mm dish, and air sealed. Bioluminescence recording was started immediately upon placement in culture with a dish-type luminometer (Kronos Dio, ATTO) maintained at 37°C. Luminescence was measured for 3 min for each dish at 30-min intervals. The data were smoothed using a 2 h moving average and further detrended by subtracting a 24 h running average for FFT analysis. A fourth-order Blackman-Harris window was applied before the power spectrum calculation. The spectrum was normalized to an integral of one by dividing each of its elements by the sum of all elements. Circadian rhythmicity was defined as relative spectral power density at the peak in the circadian range (20-30 h). The peak values represent the power within a frequency band of 0.009.
Locomotor activity recording and data analysis. C57BL/6J-backcrossed Per2E′ m/+ mice were intercrossed to produce homozygous mutant (Per2E′ m/m ) and WT (Per2E′ +/+ ) progenies using in vitro fertilization. We used adult male mice (8-to 10-week old). The animals were housed individually in light-tight, ventilated closets under indicated lighting conditions with ad libitum access to food and water. Locomotor activity was recorded via passive infrared sensors (PIRs, FA-05F5B; Omron) with 1-min resolution and analyzed with CLOCKLAB software (Actimetrics) 60 . Free-running period in DD was determined with χ 2 periodogram, based on animal behaviors in a 40-day interval taken 3 days after the start of DD condition. Rhythmicity in LL was evaluated using FFT-Relative Power (CLOCKLAB, Actimetrics). A fourth-order Blackman-Harris window was applied before the power spectrum calculation. The spectrum was normalized to an integral of one by dividing each of its elements by the sum of all elements. Circadian rhythmicity was defined as relative spectral power density at the peak in the circadian range (20-36 h). The peak values represent the power within a frequency band of 0.006. Speed of behavioral re-entrainment was evaluated using 50% phase-shift value (PS 50 ) 32,61 .
Body temperature recording and data analysis. Precalibrated temperature data loggers (Thermochron iButtons, DS1921H, Maxim) were surgically implanted to the peritoneal cavity of mice under general anaesthesia. After a week of recovery in LD, mice were transferred into LL. The iButtons were programmed for collecting temperature data every 20 min. For data analysis, the mean temperature of the entire data series during days 7 to 21 in LL was calculated, and the data points above the mean were displayed in double-plotted format using CLOCKLAB. For circular plot, data were detrended by subtracting a 24 h running average, and periods of top 20% temperature were deployed in Rayleigh format using Oriana 4 software (Kovacs Computer Services, UK). CT12 was extrapolated from the onsets of locomotor activity during days 3-6. Phase distribution was assessed by mean vector length (Oriana 4).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The source data underlying Figs. 1c, e, 2a, b, 3c, f, 4a-d, Supplementary Figs. 3, 4e, 6b, c, 10a, b, 11a, b are provided as a Source Data file. All the other data supporting the findings of this study are available within the article and its Supplementary Information files and from the corresponding authors upon reasonable request. A reporting summary for this article is available as a Supplementary Information file.