Melatonin regulates the circadian rhythm to ameliorate postoperative sleep disorder and neurobehavioral abnormalities in aged mice

In total, 18‐month‐old specific pathogen‐free C57BL/6J mice (weighing 20–25 g, 50% males) were housed in an insulated and sound‐proof room under an automatically controlled 12 h/12 h light/dark cycle and given food and water ad libitum. Lights were turned on at 8:00 a.m.—Zeitgeber time (ZT) 0— and turned off at 8:00 p.m. (ZT12). All the experimental procedures met NIH guidelines for appropriate care and use of animals in research.

Intraperitoneal (i.p.) melatonin injections (Sigma Aldrich) were administered daily (10 mg/kg body weight) at ZT12 for 7 consecutive days. Melatonin was dissolved in 0.5% absolute ethanol and saline. Luzindole (1 mg/kg i.p., a competitive MT₁/MT₂ melatonin receptor antagonist; Tocris) was administered at ZT12 for 7 consecutive days before laparotomy under sevoflurane (Abbvie) anesthesia.

Mice that were administered the vehicle (0.5% absolute ethanol) or melatonin prior to sevoflurane anesthesia and/or laparotomy were randomly assigned to the following experiments.

Sevoflurane exposure was performed according to a previously described protocol.30 Briefly, sevoflurane exposure and/or laparotomy was started at ZT0. Mice were exposed to 2% sevoflurane in 22% oxygen for 2 h in an anesthesia chamber; the concentration was monitored with a gas outlet. The heart rate, blood‐oxygen saturation, and rectal temperature were monitored. The mice breathed spontaneously, and sevoflurane was well tolerated, with all monitored variables in the physiological range.

Laparotomy was aseptically performed with a method previously used in mice.31 Mice were anesthetized for 2 h with 2% sevoflurane and intracutaneously injected with 0.2% ropivacaine along the planned incision line. A 2‐cm vertical incision was made in the middle of the abdomen, the gastrointestinal tract was exteriorized and vigorously rubbed for 30 s, and the organs (liver, spleen, kidneys, and bowel) were gently probed with cotton for 30 min. The intestines were then placed back into the peritoneal cavity, and the skin was sutured with surgical staples. EMLA cream (2.5% lidocaine and 2.5% prilocaine, AstraZeneca, Sweden) was applied to the incision wound at the end of surgery and then every 8 h for 2 days for surgical pain relief. Body temperature was maintained with a heating pad during anesthesia/surgery.

A battery of behavioral tests described in a previous study was performed to assess changes in both natural and learned behaviors following surgery and anesthesia to examine POD.32 All the mice underwent the tests in the following order: buried food, open field, and Y maze test at 24 h before anesthesia/surgery (baseline) and at 6, 9, 24, and 48 h after anesthesia/surgery. The behavior tests were completed within 50 min to simulate clinical evaluation of delirium in patients. The latency to eating food (buried food test); time spent in the center, latency to the center, and total distance traveled (open field test); and duration and entries in the novel arm and total distance traveled (Y maze test) were recorded. All the behavioral data were analyzed with an animal tracking system (Smart 3.0, RWD Life Science Co., Ltd).

Polygraphic recordings and vigilance state analysis were performed according to a previous study.33 In brief, mice were implanted with electrodes for polysomnographic EEG and EMG recordings under 2% sevoflurane anesthesia. The implant comprised two stainless‐steel screws (1 mm diameter) serving as EEG electrodes, one of which was placed epidurally over the right frontal cortex (1.0 mm anterior and 1.5 mm lateral to bregma) and the other over the right parietal cortex (1.0 mm anterior and 1.5 mm lateral to lambda). Two insulated Teflon‐coated, silver wires (0.2 mm in diameter), which were placed bilaterally into the trapezius muscles, served as EMG electrodes. All the electrodes were attached to a micro‐connector and fixed onto the skull with dental cement. The mice were given intracutaneous injections of 0.2% ropivacaine along the wounds once per day for 2 days. After the surgical procedure, mice were maintained undisturbed in the housing room for 14 days. The mice were habituated to the recording cable for 2 days before polysomnographic recording. Baseline recordings of wakefulness and sleep were conducted for 2 days prior to treatment. Wakefulness and sleep recordings were started immediately after 2 h of anesthesia/laparotomy and lasts for 2 days.

The EEG/EMG signals were amplified and filtered (EEG, 0.5–30 Hz; EMG, 20–200 Hz), digitized at a sampling rate of 128 Hz, and recorded using SLEEPSIGN software (KISSEI COMTEC CO., LTD.). The vigilance states were automatically classified offline, in 4‐s epochs, into rapid eye movement (REM) sleep, nonrem (NREM) sleep, and wakefulness by SLEEPSIGN according to the standard criteria.34 As a final step, defined sleep–wake stages were examined visually and corrected if necessary.

Western blotting was performed to determine the CREB, phosphorylated (p)‐CREB, ERK, and p‐ERK protein levels in the hippocampus and prefrontal cortex and the MT₁ and MT₂ receptor levels in the SCN. Briefly, the primary antibodies used for western blotting were as follows: anti‐MT₁ receptor antibody (1:1000; Abcam, Cambridge, UK, ab203038), anti‐MT₂ receptor antibody (1:100; Abcam, ab203346), anti‐ERK1/2 antibody (1:200; Abcam, ab17942), anti‐p‐ERK1/2 antibody (1:50; Santa Cruz Biotechnology, sc‐81492), anti‐CREB antibody (1:1000; Abcam, ab32515), anti‐p‐CREB antibody (1:1000; Abcam, ab32096), and β‐actin (1:5000; Servicebio, GB12001). Band intensities were quantified by infrared scanning densitometry (Odyssey Imaging Systems; LI‐COR Biosciences).

Immunofluorescence was performed to determine the p‐CREB, ERK1/2, and p‐ERK1/2 expression and distribution in the hippocampus. Briefly, the primary antibodies included p‐CREB (1:100; Abcam, ab32096), ERK1/2 (1:200; Abcam, ab17942), and anti‐p‐ERK1/2 (1:50; Santa Cruz Biotechnology, sc‐81492), and the secondary antibodies included Goat anti‐Rabbit Alexa‐Fluor 488‐ (1:200; Abcam, ab150077) and Goat anti‐Mouse Alexa‐fluor 594‐conjugated antibodies (1:200; Abcam, ab150116). Nuclei were counterstained with DAPI (1:5000; Roche, 236276). Images were captured with a confocal fluorescence microscope (Nikon DS‐U3) for analysis of the hippocampal CA3 region.

Total RNA was extracted from the SCN using an Eastep Universal RNA Extraction Kit (Promega) according to the standard protocol. The RNA concentrations were determined using a Nanodrop spectrophotometer (Thermo Scientific). Total RNA was reverse‐transcribed using the GoScript Reverse Transcription System (Promega). The cDNA solution was subjected to quantitative PCR in a Bio‐Rad iCycler iQ system using GoTaq® qPCR Master Mix (Promega). Quantitative PCR consisted of 40 cycles of 15 s at 95°C and 60 s at 60°C. The primer sequences were as follows:

Per1 forward primer 5′‐CTCTTCTGGCAATGGCAAGGACTC‐3′,

reverse primer 5′‐CTCAGGAGGCTGTAGGCAATGGA‐3′;

Clock forward primer 5′‐GACGGCGAGAACTTGGCATTGA‐3′,

reverse primer 5′‐TGAGACTGCGGTGTGAGATGACT‐3′;

Bmal1 forward primer 5′‐ATAAGGACTTCGCCTCTACCTGTTCA‐3′,

reverse primer 5′‐CCTCGTTGTCTGGCTCATTGTCTT‐3′;

Cry1 forward primer 5′‐GCCAGCAGACACCATCACATCAG‐3′,

reverse primer 5′‐GGGAAGGAACGCCATATTTCTCATCA‐3′;

GAPDH forward primer 5′‐AGAAGGTGGTGAAGCAGGCATCT‐3′,

reverse primer 5′‐CGGCATCGAAGGTGGAAGAGTG‐3′.

Temperature controlled melting curve analysis revealed a single peak corresponding to the specific amplification product. mRNA expression levels of clock genes in the SCN which related to the circadian rhythm, were calculated using the 2^−ΔΔCT method and analyzed using cosine software (Chronos‐Fit program).

The Shapiro–Wilk test were used to assess data distribution. Parametric results in normal distribution are presented as mean ± SEM. Statistical analyses were performed with SPSS 25.0 (SPSS Inc.). The statistical chart is drawn with GraphPad Prism 8.0. Differences were considered statistically significant at p < 0.05.

We analyzed the changes in time spent in REM and NREM sleep every 2 h after 2% sevoflurane anesthesia and laparotomy. The time course of changes revealed that the REM sleep percentage was decreased at ZT2–ZT4, ZT4–ZT6, ZT8–ZT10, ZT10–ZT12, and ZT0–ZT2(D2) after sevoflurane anesthesia and increased at ZT14–ZT16, ZT16–ZT18, and ZT22–ZT0 in mice (all p < 0.05, Figure 2A). The A/S group showed a decreased REM sleep percentage at ZT4–ZT6, ZT6–ZT8, ZT8–ZT10, and ZT10–ZT12 on D1 and ZT0–ZT2, ZT2–ZT4, ZT4–ZT6, and ZT6–ZT8 on D2 and an increase in the REM percentage at ZT22–ZT0 after laparotomy↗ (all p < 0.05, Figure 2B).

As shown in Figure 2D, time course changes revealed that the NREM sleep percentage in mice increased at ZT18–ZT20 after sevoflurane anesthesia and decreased at ZT4–ZT6 on D2 (all p < 0.05). The A/S group showed a decreased NREM sleep percentage at ZT2–ZT4, ZT4–ZT6, ZT6–ZT8, ZT10–ZT12(D1) and ZT0–ZT2, ZT4–ZT6, and ZT6–ZT8(D2) and an increased percentage at ZT12–ZT14, ZT14–ZT16, ZT18–ZT20, ZT20–ZT22, and ZT22–ZT0(D1) and ZT12–ZT14 and ZT14–ZT16(D2) (all p < 0.05, Figure 2E).

It is noteworthy that during the ZT10–12 period, the decrease in REM and NREM sleep was greater than that in the preceding periods in the A/S group. Figure 2C,F show the inflection points of the percentages of time spent in REM and NREM sleep for each mouse. The inflection points of time spent in REM and NREM sleep were both shifted forward by approximately 2 h following the transition from the light period to the dark period (L₁ → D₁) and the dark period to the light period (D₁ → L₂) after laparotomy↗, but the shift disappeared on D2.

To assess the effects of anesthesia/laparotomy on the natural habits of mice, we first performed the buried food test (Figure 2G). Compared with the C group, the latency to finding food in the A/S group was longer at both 6 h (91.12 ± 4.26% vs. 122.93 ± 7.22%) and 9 h (86.48 ± 6.32% vs. 129.21 ± 9.25%) (all p < 0.05) but not at 24 or 48 h.

The open field test was used to evaluate whether anesthesia/laparotomy affected the fear and novelty seeking state of mice (Figure 2H,I). Laparotomy decreased the time spent in the center region at 6, 9 and 24 h (111.60 ± 8.50% vs. 54.28 ± 7.57%, 102.54 ± 8.98% vs. 59.22 ± 9.36%, 87.35 ± 8.89% vs. 60.61 ± 5.83%, all p < 0.05). Therefore, laparotomy under anesthesia had a time‐dependent adverse effect on the emotional state of mice.

The spontaneous Y‐maze test was performed to evaluate spatial learning and memory in the mice. Compared with the C group, both the A and A/S groups showed decreased time spent in the novel arm at 6 h (144.68 ± 15.20% vs. 96.43 ± 8.87%, 144.68 ± 15.20% vs. 74.01 ± 8.79%, all p < 0.05, Figure 2J). Furthermore, the A/S group showed significantly fewer entries into the novel arm than the C group at both 6 and 9 h (136.71 ± 10.09% vs. 66.10 ± 6.74%, 127.53 ± 10.45% vs. 72.05 ± 7.71%, all p < 0.05, Figure 2K). In addition, the total distance traveled was not significantly different at any time point (Figure 2L). The above results demonstrate that laparotomy undermines the natural ability of mice to find food, has adverse effect on the emotional state and impairs spatial learning and memory in a time‐dependent and motor‐independent manner.

Upon further analysis of the REM and NREM sleep inflection points for all mouse groups (Figure 3A,B), we found that melatonin pretreatment reversed the forward shift of the inflection point, however, no shift was found in the M group. Because the inflection points for REM and NREM sleep were also shifted forward by approximately 2 h on D1 after laparotomy↗ in the M + L + A/S group, we speculate that laparotomy↗ under anesthesia can cause circadian fluctuations and melatonin may reverse these changes through melatonin receptors.

We then investigated the total time spent in REM and NREM sleep of each group from ZT4 to ZT8. The total time spent in REM sleep decreased 14.3%, 23.9%, and 18.1% in the A, A/S, and M + L + A/S groups, respectively (all p < 0.05, Figure 3C), which was consistent with the decreases of 18.2% and 17.8% in NREM sleep in the A/S and M + L + A/S groups, respectively (all p < 0.05, Figure 3D). Meanwhile, the times spent in REM and NREM sleep were unaltered in the M group. These results suggest that laparotomy under sevoflurane anesthesia had a greater impact on sleep times than did sevoflurane alone from ZT4 to ZT8, and pretreatment with melatonin can normalize the sleep disturbance.

To better understand the postoperative changes in sleep architecture, the distribution of the number of periods in each stage from ZT4 to ZT8 was determined as a function of duration (Figure 3E,G,I). Laparotomy under sevoflurane anesthesia decreased the number of periods of REM sleep with durations of 0–16, 16–32, 32–64, and 64–128 s (all p < 0.05). Simultaneously, the numbers of periods of NREM sleep with durations of 0–16, 16–32, and 32–64 s (all p < 0.05) were significantly decreased. The number of periods of wakefulness with durations of 0–16, 16–32, and 32–64 s (all p < 0.05) were decreased; only the number of periods of wakefulness with a duration of 64–128 s increased (p < 0.05). After melatonin pretreatment, only the number of periods of REM with durations of 0–16 s and wakefulness with durations of 512–1024 s were decreased (all p < 0.05, Figure 3F,H,J). These results suggest that the sleep architecture changed during ZT4–ZT8 after the surgery, and this change was alleviated by melatonin pretreatment.

Compared with the A/S group, the latency to finding food in the M + A/S group was shorter at both 6 and 9 h (122.93 ± 7.22% vs. 86.53 ± 4.29%, 129.21 ± 9.25% vs. 84.50 ± 6.33%, all p < 0.05, Figure 4A,B) in the buried food test. In addition, compared with the M + A/S group, the latency to finding food in the M + L + A/S group was longer at 6 h (86.53 ± 4.29% vs. 125.67 ± 8.35%, p < 0.05).

In the open field test, compared with the A/S group, M + A/S group increased the time spent in the center region at 9 h (59.22 ± 9.36% vs. 105.54 ± 7.48%, p < 0.05) but not at 24 h (Figure 4C,D). Luzindole pretreatment did not significantly antagonize melatonin pretreatment at either 9 or 24 h. Furthermore, the total distance traveled was not significantly different among the five groups at any time point (Figure 4E), suggesting that different treatments did not cause motor dysfunction.

Compared with the A/S group, M + A/S group increased the time spent in the novel arm at 6 h (74.01 ± 8.79% vs. 139.42 ± 12.29%) in the Y maze test (Figure 4F) and increased entries into the novel arm at both 6 h (66.10 ± 6.74% vs. 135.75 ± 7.80%, Figure 4G) and 9 h (72.05 ± 7.71% vs. 134.05 ± 8.57%, all p < 0.05, Figure 4H). Furthermore, luzindole pretreatment showed significant antagonism in the M + L + A/S group, as the time spent in the novel arm decreased (139.42 ± 12.29% vs. 91.76 ± 5.59%) at 6 h, and the entries into the novel arm decreased (134.05 ± 8.57% vs. 94.54 ± 7.02%) at 9 h (all p < 0.05). These results demonstrate that melatonin can alleviate spatial learning and memory impairments caused by laparotomy under anesthesia, and melatonin receptor antagonists can partially prevent this improvement.

A 2 h laparotomy under sevoflurane anesthesia significantly altered clock gene expression in the SCN for 2 days after surgery, especially on D1 (Figure 5A,E,I,M). Compared with the C group, Bmal1, Clock, and Cry1 mRNA expression showed a delayed peak on D1 after surgery (5.48 ± 1.30, 4.30 ± 0.52, 10.49 ± 0.53 h, all p < 0.05), with no phase shift on D2. Per1 expression showed no phase shift within 2 days after surgery. It is noteworthy that melatonin restored clock mRNA expression to baseline levels after surgery (Figure 5B,F,J,N).

Laparotomy did not change the amplitude of clock gene expression; however, a change was observed in the M + A/S group (Figure 5C,G,K,O). Melatonin pretreatment increased the amplitude of Bmal1 mRNA expression in mice on D1 and D2 compared with the C group (216.71 ± 49.50% and 180.96 ± 19.67% of baseline, all p < 0.05). Melatonin also increased the amplitude of Per1 mRNA expression on D1 (128.45 ± 10.93% of baseline, p < 0.05).

Similar to the change in amplitude, melatonin pretreatment affected the MESOR of clock gene expression, but laparotomy did not cause a significant change in the MESOR compared with the C group (Figure 5D,H,L,P). Melatonin pretreatment increased the MESOR of Bmal1 mRNA expression in mice on D1 and D2 compared with the C group (229.09 ± 24.42% and 248.18 ± 36.42% of baseline, all p < 0.05). Melatonin also increased the MESOR of Cry1 and Per1 mRNA expression on D2 (200.61 ± 40.60% and 143.36 ± 12.66% of baseline, all p < 0.05).

To assess the effect of laparotomy under anesthesia on melatonin receptors in the SCN of aged mice, MT₁ and MT₂ receptor expression levels were analyzed by western blotting analysis and showed that laparotomy under anesthesia significantly increased MT₁ receptor expression levels at 6 and 12 h (all p < 0.05) but not at 2 days after laparotomy. Pretreatment with melatonin could normalize the laparotomy‐induced changes in MT₁ receptor levels in the SCN, and the corrective action of melatonin was antagonized by luzindole (both 6 and 12 h; all p < 0.05). MT₂ receptor expression was not detected by western blots within 2 days after treatments, which suggests that MT₂ receptor labeling was sparse in the SCN (Figure 6A–D).

We then detected MT₁ expression in the SCN 6 h after treatment by immunofluorescence (Figure 6E,F). The results also suggested that laparotomy under anesthesia increased MT₁ receptor expression levels in the SCN, melatonin can normalize the laparotomy‐induced change in MT₁ receptor levels, and the corrective action of melatonin could be antagonized by luzindole (all p < 0.05).

We next assessed whether the ERK/CREB signaling pathway participates in postoperative depression‐like behavior.

In the hippocampus of aged mice, p‐ERK/ERK decreased on D1 after laparotomy (p < 0.05) compared with the control condition and returned to normal on D2 (Figure 7A,C,E). When melatonin pretreatment was applied, the decrease in p‐ERK/ERK 12 h after laparotomy was alleviated, however, this improvement was antagonized by luzindole (all p < 0.05, Figure 7B,D). Similar trends were observed for p‐CREB/CREB, which were decreased on D1 after laparotomy (p < 0.05) and returned to normal on D2 (Figure 7F,H). When melatonin pretreatment was applied, the decrease in p‐CREB/CREB 12 h after laparotomy was alleviated (all p < 0.05). In contrast to p‐ERK/ERK, the decrease in p‐CREB/CREB in the M + L + A/S group was not statistically significant compared with that in the M + A/S group (Figure 7G,I,J), which may indicate that luzindole (1 mg/kg) has an insufficient antagonistic effect on p‐CREB/CREB in the hippocampus 12 h after laparotomy. In addition, melatonin alone did not trigger ERK and CREB phosphorylation in the hippocampus.

In the prefrontal cortex, p‐ERK/ERK was decreased on D1 after laparotomy (p < 0.05) and returned to normal on D2 (Figure 8A,C). When melatonin pretreatment was applied, the decrease in p‐ERK/ERK 12 h after laparotomy was alleviated, however, this improvement was antagonized by luzindole (all p < 0.05, Figure 8B,D). Compared with the control group, no significant difference in p‐CREB/CREB was observed on D1 and D2 after laparotomy (Figure 8E,G). When melatonin pretreatment was applied, the decrease in p‐CREB/CREB 12 h after laparotomy was alleviated, and this improvement was antagonized by luzindole (all p < 0.05). Likewise, melatonin alone did not trigger ERK and CREB phosphorylation in the prefrontal cortex (Figure 8F,H).

The datasets generated and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.