All experiments were approved by the Committee for Animal Experimentation of Hiroshima University (approval no. A23-115-2). The experiments were performed in accordance with the law (No. 105) and notification (No. 6) of the Japanese government.

Male ICR mice (8 weeks old; Charles River Laboratories Japan, Inc.) were housed individually under controlled environmental conditions: room temperature 23 ± 2°C, humidity 40% ± 10%. Following a one-week acclimation period under a 12:12 light–dark cycle (lights on at 08:00 and off at 20:00), the light schedule was shifted to 22:00–10:00 to accommodate the experimental interventions. The mice were acclimated to new light cycles for 3 weeks until their activity patterns were fully adapted. All mice were provided ad libitum access to food and water. n = 4–5 mice per group at each time point.

In our mouse model, we observed reduced activity and body temperature during natural siesta (approximately ZT20–23). Mice without any intervention served as the control group. For siesta disruption model establishment, beddings were replaced twice daily at ZT20 and ZT22 using beddings previously utilized by mice within the same group. To prevent the mice from reusing the original bedding, fresh bedding was introduced every 5 days. A single bedding change typically increased activity for about 1 hour; however, a change at ZT20 alone was insufficient to fully suppress the siesta, as rest behavior resumed around ZT22–23 (Supplementary Figure S1).

Compared to other commonly used sleep disruption methods—such as gentle handling or auditory stimulation—bedding change offers a more practical alternative for siesta-specific interventions. Manual stimulation (e.g., tapping or brushing) is labor-intensive and subject to variability between experimenters, while auditory stimuli may not reflect the natural context of siesta interruption. Bedding change, by contrast, passively increases spontaneous activity may mimic real-world behaviors such as post-meal walking in humans to counteract drowsiness. Moreover, this method is well suited for repeated interventions over extended periods.

Body weight and food intake were monitored weekly throughout the study. On Day 14, immediately following the final siesta disruption, 5 mice in each group were sacrificed at ZT0 to capture the immediate molecular responses. The remaining mice were sacrificed at ZT12 on Day 15, 12 hours after the final siesta disruption. This timing followed a full light-phase sleep opportunity, enabling the assessment of delayed effects after partial recovery from siesta disruption. Serum and frontal cortex samples were collected for subsequent analyses. A schematic of the experimental timeline is shown in Figure 1.

Blood was collected from mice and allowed to clot at room temperature for over 30 min. The samples were then centrifuged at 3,000 × g for 20 min at 25°C. The serum was carefully collected and stored at −80°C until further analyses. Serum corticosterone (501320) (Cayman Chemical, Ann Arbor, MI), superoxide dismutase activity (SOD, 706002) (Cayman Chemical, Ann Arbor, MI), and Malondialdehyde (MDA, M496) (Dojindo, Kumamoto, Japan) were measured using ELISA kit. Total antioxidant capacity (TAC) was assessed using a ready to use kit (AC01D) (Metallogenics, Chiba, Japan).

Spontaneous locomotor activity was individually monitored using an infrared radiation sensor (SE-10; Akizuki Denshi Tsusho Co., Ltd., Tokyo, Japan) in accordance with our previously described method (Tahara et al., 2024). Movements were detected and recorded in 20-min epochs. Data collected for 2 weeks were averaged and analyzed using CLOCKLAB software (Actimetrics, IL, USA).

Core body temperature was measured using a small temperature sensor (Thermochron SL; KN Laboratories Inc., Osaka, Japan) designed to resemble a button battery. The mice were anesthetized with isoflurane, and the sensor was implanted with a sensing surface positioned near the liver. After a recovery and light shift period of 3 weeks, temperature recordings were automatically taken at 20-min intervals for 10 days. Data were analyzed using RhManager software (Version 2.27; KN Laboratories Inc., Osaka, Japan) as reported previously (Ohnishi et al., 2014).

Surgical implantation of a sleep-monitoring head mount (Pinnacle Technology, Inc., USA) was performed under isoflurane anesthesia. Four stainless-steel EEG electrodes (#8209 0.10″EEG Screws, Pinnacle Technology, Inc., USA) were implanted into the skull over the cortex to monitor EEG signals. Silver conductive epoxy adhesive (MG Chemicals, Canada) was applied to the screws and contact surfaces of the head mount to ensure electrical conductivity. The head mount was securely affixed to the skull using resin cement (Super-Bond, Sun Medical, Japan). Two pre-soldered stainless-steel wires for EMG recording were inserted into the neck muscles. After securing the head mount, an ultraviolet-cured adhesive (Transbond, 3 M Japan Innovation, Japan) was used to stabilize the mount and seal the surgical incision.

Following a one-week recovery period, the animals underwent at least 2 days of habituation in the recording cages before the sleep recordings commenced. The measurements were conducted for two consecutive days under two conditions: control and siesta disruption. EEG and EMG signals were amplified and filtered (0.75–30 Hz) using a Data Conditioning and Acquisition System connected to preamplifiers and digitized through SIRENIA® software (Pinnacle Technology, Inc., USA). The digitized data were analyzed semi-automatically using SleepSign® software (Kissei Comtec, Japan) and visually inspected for accuracy. Sleep stages were classified as follows: wakefulness (Wake), high EMG activity; non-rapid eye movement (NREM) sleep, low EMG activity and a prominent delta band power (0.5–4 Hz); and rapid eye movement (REM) sleep, minimal EMG activity and a prominent theta band power (7–8 Hz). For additional analysis, EEG data were subjected to fast Fourier transformation within the 0–24.2 Hz range using SleepSign®. The hourly theta/delta power density (%) during REM and NREM sleep was calculated as the ratio of the theta/delta band power to the total EEG power within the same frequency range for each hour of light-phase sleep.

All behavioral tests were conducted at the end of the light phase (ZT10–12) on day 11 or day 12. All sessions were recorded using a video camera, and the videos were analyzed by a trained observer blinded to the group assignments. Initially, each group consisted of 10 animals used for general experimentation; however, not all behavioral tasks were performed in this primary cohort. To assess additional behavioral outcomes, including the open field test, object location test, and novel object recognition test, a follow-up cohort of 5 animals per group was used. Consequently, sample sizes varied (n = 5–10 per group) depending on the specific behavioral assay.

Total RNA was extracted from the frontal cortex using the TRIzol reagent (Life Technologies, Carlsbad, CA, USA), followed by purification using DNase I (Thermo Fisher Scientific, USA). RNA integrity and concentration were assessed using a NanoDrop spectrophotometer (Thermo Fisher Scientific, USA). RNA sequencing was performed by Novogene Co. Ltd. (Tokyo, Japan). Sequencing libraries were prepared using the NEBNext Ultra II RNA Library Prep Kit (Illumina) in accordance with the manufacturer’s protocol. Clean reads were aligned to the Mus musculus reference genome (GRCm39, Ensembl release 111) using STAR 2.7.0a. Alignment metrics indicated mapping rates exceeding 85% for all samples.

Gene expression levels were quantified as fragments per kilobase of transcript per million mapped reads using RSEM 1.3.3. Differentially expressed genes (DEGs) were identified using DESeq2 package 1.42.0 in R 4.3.2. The DESeq2 workflow included normalization and statistical testing using Wald tests for pairwise comparisons. DEGs were filtered based on fold change > 1 and adjusted p-value < 0.05 for Wald tests. One outlier sample was identified during the analysis. Functional enrichment analyses of the DEGs were conducted using Metascape (v3.5.20240901) with a cut-off criterion of adjusted p-value < 0.05 for all groups. Venn diagrams and bar plots for enrichment analysis were generated using OECloud tools¹ provided by OE Biotech Co., Ltd. To validate the reliability of the RNA-seq results, we performed RT-PCR on a subset of representative genes from key enriched pathways, using the same sample set. The expression trends observed by RT-qPCR were consistent with the RNA-seq data, supporting the robustness of the transcriptomic findings (Supplementary Figure S2). Previous studies have shown that high-quality RNA-seq datasets, when processed under stringent protocols, provide reliable and reproducible gene expression profiles without requiring validation of every individual gene (Everaert et al., 2017).

Body temperature recordings, EEG/EMG measurements, and a subset of behavioral tests (open field test, object location test, and novel object recognition test) were performed using separate cohorts of mice. Although these cohorts were not used for molecular or transcriptomic analyses, all animals were subjected to the same siesta disruption protocols and housed under identical environmental conditions. This approach was used to avoid repeated invasive procedures in the same animals and to ensure data quality across different experimental modalities.

Statistical analyses were performed using SPSS version 29.0.0.0 (IBM Corp., Armonk, NY) and graphical representations were created using GraphPad Prism version 10.1.0 (GraphPad Software, San Diego, CA, USA). Normality of the data was assessed using the Shapiro-Wilk test, and homogeneity of variance was evaluated with Levene’s Test or an F-test, as appropriate. For parametric data, two-way ANOVA followed by Bonferroni’s post-hoc test was used for multiple comparisons, and unpaired two-tailed t-tests were applied for pairwise comparisons. Welch’s correction was applied when the assumption of equal variances was violated, as automatically determined by SPSS. Non-parametric data were analyzed using the Kruskal-Wallis test with Dunn’s post-hoc test. Data are presented as mean ± standard error of the mean.

For the statistical analysis of 24 h body temperature, activity, and sleep-related time-series data, which did not pass the equal variance tests, differences between two groups at specific time points were assessed using unpaired two-tailed t-tests. Statistical significance was set at p < 0.05, with significance denoted as follows: *p < 0.05.

Figures 2A,B presents the results for body weight and dietary intake, which remained stable across experimental groups. No significant differences in serum corticosterone level (Figure 2C), MDA level (a marker of lipid peroxidation, Figure 2D), SOD activity (a key antioxidant enzyme (Michiels et al., 1994), Figure 2E), and TAC level (Figure 2F) were found between control and siesta disruption. In the control group, corticosterone level (Z = −2.67, p = 0.045) and TAC level (F (interaction) = 0.02, p = 0.041) were significantly higher at ZT12 than ZT0, indicating a time-dependent change, while, siesta disruption became flat and did not show any time-dependent change.

Both core body temperature and locomotor activity exhibited clear circadian rhythms, with lower levels during the light phase and higher levels during the dark phase. During the second half of the dark phase, both parameters declined, corresponding to the siesta period. To disrupt this rest phase, we introduced siesta disruption interventions at ZT20 and ZT22. A single intervention at ZT20 induced a brief increase in arousal, but its effect diminished quickly, allowing a rebound siesta around ZT22–23 (Supplementary Figure S1). A second intervention at ZT22 reactivated both physiological measures. These observations demonstrate that dual interventions are required to sustain arousal throughout the siesta window, as confirmed by increased body temperature (F = 0.02, p = 0.008) and locomotor activity (F = 7.82, p = 0.018) during ZT20-24 (Figures 3A,B), with notable elevations at specific time points detailed in Supplementary Table S1. During the light phase sleep (ZT0–12) and early dark phase (ZT12–20), core body temperature did not differ significantly between groups. However, locomotor activity was significantly elevated in the siesta disruption group at some special time points during the light phase: ZT1.67 (F = 26.22, p = 0.046), ZT2.67 (F = 16.42, p = 0.046), ZT3 (F = 28.14, p = 0.049), ZT7.33 (F = 23.33, p = 0.048), ZT8.67 (F = 8.32, p = 0.037), suggesting that the arousal effects of siesta disruption persist into the subsequent rest period.

Interventions at ZT20 and ZT22 significantly increased wakefulness during ZT20–24 (F = 0.07, p = 0.001), particularly at ZT20 (F = 1.17, p = 0.012), ZT21 (F = 0.02, p = 0.000), and ZT22 (F = 18.54, p = 0.031) (Figure 4A), accompanied by a corresponding reduction in NREM sleep during the same period (F = 0.01, p = 0.001), especially at ZT20 (F = 1.83, p = 0.016), ZT21 (F = 1.25, p = 0.000), and ZT22 (F = 7.60, p = 0.029) (Figure 4B). REM sleep also decreased significantly across ZT20–24 (F = 0.66, p = 0.001), particularly at ZT20 (F = 1.54, p = 0.011) (Figure 4C). Overall, this protocol resulted in an approximate 70% reduction in total sleep during the intervention window, confirming the effectiveness of the siesta disruption model. Moreover, a significant rebound increase in REM sleep was observed in the siesta disruption group at ZT8 (F = 0.09, p = 0.015). In contrast, no significant differences were observed in the total durations of wakefulness, NREM sleep, or REM sleep during ZT0–12 and ZT12–20 among the groups (all p > 0.05). Furthermore, the power density of NREM and REM sleep during ZT0–12 was not affected by siesta disruption (Figure 4D). Although NREM delta power is a well-established marker of sleep pressure, no group differences were found, potentially due to rapid homeostatic compensation or the relatively limited contribution of siesta sleep to overall sleep drive. These results indicate that siesta disruption primarily alters sleep architecture during the intervention window (ZT20–24), while leaving the overall sleep patterns during the subsequent light phase (ZT0–12) largely unchanged.

Behavioral responses to siesta disruption were assessed at the end of light phase using multiple paradigms. Anxiety-like behavior was evaluated using the open field test, which revealed no significant differences in the number of center entries or time spent in the center zone across the groups (Figure 5A). This result indicated that siesta disruption did not alter anxiety-like behavior. Cognitive performance was assessed using the object recognition test, which included the recognition of moved and novel objects. Both measures remained unaffected by siesta disruption (Figure 5B), indicating that cognitive performance remained intact. Hedonic behavior was assessed by measuring the preference for saccharin over water, and no differences were observed between the groups (Figure 5C), suggesting that hedonic behavior was not affected by siesta disruption. Finally, depression-like behavior was evaluated using the forced swim test, in which immobility, struggling, and swimming duration were similar between the groups (Figure 5D). While these results suggest minimal behavioral consequences of siesta disruption, caution is warranted in interpretation. The relatively small sample sizes may have limited the statistical power to detect subtle effects, and the paradigms used may not fully capture the range of possible functional alterations. These limitations should be considered when evaluating the absence of group differences. However, as our data showed no apparent trend toward behavioral differences between groups, we consider the lack of significance to reflect a true absence of effect, rather than a false negative.

The Venn diagram (Figure 6A) reveals both overlapping and unique sets of DEGs in the siesta disruption groups compared to controls. At ZT0, 90 genes were upregulated and 46 were downregulated, while at ZT12, 81 genes were upregulated and 10 were downregulated. The complete lists of DEGs are provided in Supplementary Table S2. Notably, Pseudopodium-enriched atypical kinase 1 (Peak1) was the only gene consistently upregulated at both ZT0 and ZT12 (Supplementary Figure S3).