Disrupting daily rhythms early in life changes adult brain and behavior in mice

Beginning on the day of birth, female mice and their litters were exposed to either a regular LD cycle or a disrupting LD cycle where dark onset was advanced by 8 h every second day. To estimate the disrupting effects of such an LD cycle, adult males housed in the same room and exposed to these same LD cycles had their wheel running rhythms recorded (Fig. 1). The strength of the locomotor rhythms was assessed using an F-periodogram analysis. The power at the 24 h period was significantly less (t₍₂₎ = 8.93, p = 0.012) in the disrupted animals (0.31 ± 0.01) than in animals from a regular LD cycle (0.63 ± 0.05). The period with the greatest power in the disrupted animals was at 20 h (0.46 ± 0), and the power at this period was also significantly lower (t₍₂₎ = 5.00, p = 0.038) than that of the peak power observed in control animals at 24 h. This demonstrates that the LD was highly disrupting to the locomotor rhythms in these sample animals with activity frequently occurring during the light phase, and occurring in a fragmented fashion.

To examine anxiety-like behavior, mice (control n = 12, disrupted n = 12) were assessed using the Elevated Plus Maze (EPM). The disrupted animals had significantly less entries into the open arms (1.5 ± 1.5 entries) than did the undisrupted control group (2.75 ± 1.1 entries, Mann–Whitney U = 32.5, n1 = n2 = 12, p = 0.02 two-tailed, Fig. 2A). Such diminished exploratory behavior is consistent with higher anxiety-like behavior in the disrupted animals. In addition, the disrupted animals spent significantly less time in the exposed arms (8.1 ± 11.0 s) compared to the controls (18.5 ± 7.3 s Mann–Whitney U = 24.0, n₁ = n₂ = 12, p = 0.005 two-tailed, Fig. 2B) consistent with higher anxiety-like behavior.

To examine spatial memory, the same mice were then assessed using Morris Water Maze (MWM). Animals from both groups showed significant improvements in the time they took to locate the platform across the four training days (ANOVA on transformed data, main effect of days, Greenhouse–Geisser, F_{(2.11, 46.38)} = 16.01, p < 0.001, µ² = 0.42, Fig. 3A). However, the disrupted animals took significantly more time to reach the platform compared to controls over the course of the learning days (ANOVA on transformed data, main effect of groups, F_(1,22) = 10.20, p = 0.004, µ² = 0.31). While there was no significant interaction between day and group in latency to find the platform (F_(3,88) = 0.127, p = 0.127), planed comparisons revealed a significant pairwise difference on the final day of training. Disrupted animals took significantly longer (13.1 ± 9.1 s) to get to the target quadrant compared to the controls (6.3 ± 4.1 s) on testing day 5 (Independent t-test on transformed data, t₍₂₂₎ = 2.35, p = 0.028, Fig. 3B). The disrupted animals passed through the target quadrant less often than the controls, but this difference was not statistically significant (Independent t-test on transformed data, t₍₂₂₎ = 1.74, p = 0.095, β = 0.39, Fig. 3C). Moreover, there was no significant difference in the time spent in the target quadrant between disrupted and controls (t₍₂₁₎ = 0.93, p = 0.365, β = 0.39, Fig. 3D). There was a significant interaction between day and group in swim speed (F_(4,381) = 16.9, p < 0.001), with the disrupted animals swimming faster (0.15 ± 0.04 m/s) than controls (0.11 ± 0.06 m/s) on day 1 of training and slower (0.12 ± 0.05 m/s) than controls (0.21 ± 0.04 m/s) on day 4 of training. No pairwise differences were observed in swim speed on other training days or on testing day 5.

There was significant improvement across the days and trials of the test, as indicated by decreased number of wrong entries over time, both within and across days (ANOVA on transformed data, main effect of days F_(7,154) = 3.33, p = 0.009, µ² = 0.13, main effect of trials F_(7,154) = 7.91, p < 0.001, µ² = 0.26, Figs. 4A,B), respectively. Control animals committed significantly fewer errors compared to disrupted animals on Delayed Match-to-Sample (DMS) task (ANOVA on transformed data, between groups effect F_(1,22) = 17.75, p < 0.001, µ² = 0.45).

Performance on the DMS task can also be assessed with latency to find the platform. Animals from control group significantly improved in performance across the days and the trials of the DMS task, in which they became faster in reaching the platform than the disrupted animals over time (ANOVA on transformed data with Greenhouse–Geisser correction, main effect of days F_(4.90,07.83) = 7.06, p < 0.001, µ² = 0.24, main effect of trials F_{(3.30, 72.53)} = 2.76, p < 0.001, µ² = 0.11, Fig. 4C,D). Disrupted animals took longer to reach the platform across testing days compared to the controls, but their performance showed improvement towards the end of the testing days (ANOVA on transformed data with Greenhouse–Geisser correction, days × groups interaction F_{(4.90, 107.83)} = 8.04, p < 0.001, µ² = 0.27, trials × groups interaction F_{(3.3, 72.53)} = 2.76, p = 0.043, µ² = 0.11). In particular, disrupted animals took longer to get to the platform on days 1, 3, 5, and 6 in comparison with the controls (Simple main effect with Holm’s correction, day 1, F_(1,23) = 12.96, p = 0.002; day 3, F_(1,23) = 5.57, p = 0.028; day 5, F_(1,23) = 77.69, p < 0.001; and day 6, F_(1,23) = 6.23, p = 0.021, Fig. 4C). Results from latency to platform across trials showed that disrupted animals and controls did not show differences in performance during the first few trials collapsed across days, but subsequently the disrupted animals performed significantly worse than controls (ANOVA on transformed data with Greenhouse–Geisser correction, trials x groups interaction F_{(3.3, 72.53)} = 2.76, p = 0.043, µ² = 0.11, Fig. 4D). In particular, disrupted animals took significantly longer time to find the platform on trials 2, 4, 5, and 6 compared to controls (Simple main effect with Holm’s correction, Trial 2 F_(1,23) = 4.68, p = 0.042, Trial 4 F_(1,23) = 15.37, p = 0.001, Trial 5 F_(1,23) = 15.39, p = 0.001, Trial 6 F_(1,23) = 26.81, p < 0.001).

When examining neuronal complexity of pyramidal cells, the branching order of the disrupted animals was significantly lower compared to controls in the medial prefrontal cortex (mPFC, Welch’s F_(1,10) = 20.6, p = 0.001, Fig. 5) and in the dorsal hippocampus (dHC, Welch’s F_(1,10) = 7.6, p = 0.02), but was not significantly different from controls in basolateral amygdala (BLA, Welch’s F_(1,10) = 0.168, p = 0.690, β = 0.087). When examining dendritic length, animals that experienced early life circadian rhythm disruptions exhibited a significant reduction in dendritic length of the pyramidal neurons in the mPFC (Welch’s F_(1,10) = 29.1 p < 0.001), in the dHC (Welch’s F_(1,10) = 35.3 p < 0.001), and in BLA (Welch’s F_(1,10) = 47.5, p < 0.001, Fig. 5) in comparison with the controls.

As adults, the disrupted animals were significantly more active than the control animals when total daily activity was considered (Control 9173 ± 3245 revs/day, Disrupted 11,386 ± 1337 revs/day, t₍₁₄₎ = 1.934, one tailed p = 0.037). This difference did not carry over to activity levels in constant darkness (DD, Control 7718 ± 3836 revs/day, Disrupted 9255 ± 2853 revs/day, t₍₁₄₎ = 0.918, one tailed p = 0.18). When activity levels were examined more closely at different circadian phases, there were no significant differences between disrupted and control groups in DD (F_(1,70) = 0.84, p = 0.374, Fig. 6). In LD, while activity at the various night phases was moderately higher in the disrupted animals, this difference was not significant (F_(1,70) = 3.74, p = 0.074) and there was no significant interaction between phase and group (F_(5,70) = 1.34, p = 0.256). In addition, there was no significant difference in the phase angle of entrainment between the control and disrupted groups (t₍₁₄₎ = 0.59, p = 0.560). There were no differences in the duration of the active phase (alpha) between controls and disrupted animals in either LD (t₍₁₄₎ = 0.22, p = 0.415) or DD (t₍₁₄₎ = 0.38, p = 0.384). Free-running period did not differ between groups when assessed with either a χ² periodogram (Controls 23.83 ± 0.18 h, Disrupted 23.78 ± 0.08 h, t₍₁₄₎ = 0.767, p = 0.456) or when assessed with regression lines fit to activity onsets (Controls 23.83 ± 0.12 h, Disrupted 23.81 ± 0.12 h, t₍₁₄₎ = 0.345, p = 0.735). In addition, no significant effects were observed between groups for onset variability, mesor, amplitude, acrophase, robustness of the rhythm, or χ² strength, in either LD or DD (Table 1). In most cases, the power for these tests were below β = 0.8, so negative results should be interpreted cautiously.

All experimental protocols were approved by the Life and Environmental Sciences Animal Care Committee at the University of Calgary. All methods are reported in accordance with ARRIVE guidelines for the reporting of animal experiments. All methods were carried out in accordance with the ethical guidelines and regulations of the Canadian Council on Animal Care. To generate the animals used in this study, we bred adult female C57Bl/6 mice (n = 12), obtained from the University of Calgary Life and Environmental Science Animal Resource Centre breeding colony. Pregnant mothers were randomly assigned to be housed with their litters in a regular 12:12 light–dark (LD) cycle (n = 6) or a disrupting LD cycle (n = 6). Mothers in the regular LD cycle produced 47 pups (n = 20 females, n = 27 males), while mothers in the disrupting LD cycle produced 55 pups (n = 32 females, n = 23 males). A maximum of 2 pups from each litter were used for each separate experiment. Sample sizes used in each experiment were based on sample sizes shown previously to be sufficient to detect meaningful differences in circadian and learning experiments in our lab^43,53,54. Because circadian rhythms of infant mice in a litter are not easily assessed, additional adult C57Bl/6 male mice (n = 4) were used to assess the acute effects of the disrupting LD cycle (n = 2) on circadian locomotor rhythms relative to that observed in a normal LD cycle (n = 2).

On the day of birth (postnatal (PN) day 0), cages with the mothers and their litters assigned to the disrupting LD cycle (6 cages, n = 55 pups) were transferred to the circadian disruption colony room. The 12:12 LD cycle was advanced by 8 h every 2nd day. Litters were exposed to the disrupting LD cycle from birth until weaning (PN0-21). Control litters (6 cages, n = 47 pups) were similarly transported to a room where they were exposed to a stable 12:12 LD cycle, with lights on at 6 a.m. On the day of weaning all animals were transferred back to the main mouse colony room with a regular 12:12 LD cycle (lights on at 6am) and housed with same sex littermates (n = 2–4 animals per cage). Food and water were available ad libitum, and room temperature and humidity were maintained at 21 ± 1 °C and 50% ± 10%, respectively. At the age of 3 months (PN90), animals were assigned to one of three experiments (Behavioral Assessment, Circadian Rhythm Assessment, Golgi-Cox Neuroanatomy), with a maximum of 2 animals per litter contributing to each.

To assess the level of circadian disruptions experienced by animals under the disrupting LD cycle during early life period, adult male mice were placed in cages equipped with wheels and exposed to the disrupted LD schedule (n = 2) or the regular LD cycle (n = 2) in parallel to the mothers and their litters.

Animals were between PN90-110 days of age at the start of behavioral testing. Two animals, a male and a female, were randomly selected from each litter for the behavioral assessment (n = 12 from each light cycle). On the first day of testing, the animals from both groups were weighed (average weight: control 30.9 ± 6.6 g, disrupted 31.6 ± 6.6 g). The animals were exposed to three different tasks to assess different behaviors. Tasks were presented in the following order over different days: Elevated Plus Maze (EPM, day 1), Morris Water Maze (MWM, days 4–8), Delayed Match-to-Sample (DMS, days 12–19). This order was selected so that the tasks progressed from least to most complex and allowed for anxiety to be assessed by the EPM before the mice had habituated to extensive handling. Animals had a 2–3 day break between each test to minimize any effects that the previous task would have on the subsequent tasks⁵⁵. Performance on each task was tracked using a ceiling-mounted camera centered above the apparatus and connected to HVS Image 2020 Plus software (HVS Image Ltd, Twichenham, Middlesex, UK). Performance was also manually recorded by two observers as a backup in case technical problems hindered the automated data collection.

The Elevated Plus Maze (EPM) assesses anxiety-like behavior in rodents⁵⁶. Animals were tested in the morning beginning at 9 AM. The EPM suitable for mice (Stoelting Co., Il, USA) was illuminated evenly by an overhead light and 2 side lamps. Mice were placed in the center of the maze facing the open arm and were given 5 min to freely navigate the apparatus. The maze was cleaned with 70% ethanol between each animal.

The percent of time the animal spent in the open arms, the percent of time the animal spent in the closed arms, and the number of entries into open and closed arms were recorded. An arm entry was defined as all four paws of the animal fully within the arm. Animals that made fewer entries into the open arms or spent less time in the open arms were deemed to be more anxious.

The Morris Water Maze (MWM) is a well-validated task for the study of hippocampal-dependent spatial memory in rodents⁵⁷. A circular white fiberglass pool 1.2 m in diameter with no landmarks on the interior wall was used for this task. The pool was filled with water (19 ± 0.5 °C) to a depth of 20 cm. A clear 12 × 12 cm square platform was placed in the center of a quadrant, 1.5 cm below the surface of the water, and remained in this location throughout all training days. The water was made opaque by adding non-toxic white tempera paint to ensure that the mice could not visually locate the platform in the pool. Several distal cues were placed on the walls of the room. The task was conducted over 5 days in the early part of the light phase (i.e., from about 9 AM to 12 noon) each day. The first 4 days were training days that consisted of 4 trials each day. The fifth day was a single trial testing day with the hidden platform removed (probe trial). For each trial on the training days, the animal was placed into the water facing the pool wall at each of the cardinal compass points (N/E/S/W) in a random order that was the same for each animal that day, but was changed every day. Each animal was given at least 10 min to rest between each trial. Animals were given 60 s to swim around the pool. Once the animal climbed onto the platform, the timer was stopped and the animal was allowed 15 s atop the platform to visually orient itself to its position in the room. Animals that failed to locate the platform within 60 s were gently guided to the platform and were allowed 15 s to orient. Animals were dried and returned to a warm cage to rest until their next trial. The pool was cleaned of any floating particles and the water was stirred between each animal. The time to reach the platform was recorded on each trial.

On the fifth day (probe trial), the platform was removed from the pool and the animal was placed into the water from a novel location opposite of the previous location of the platform. Each animal was given a single trial for a maximum of 60 s to swim freely. For this trial, the initial latency to the platform’s former location, the number of passes through the previous location of the platform, and the time spent in each quadrant were recorded and analyzed.

Delayed Match-to-Sample (DMS) task assesses working memory in rodents^58,59. The protocol was adapted from that described by Bimonte-Nelson⁵⁹. The maze consists of four symmetrical walled arms (10 × 32 cm, and 15 cm in depth) arranged in a “plus” shape that were then placed within the MWM pool. A transparent platform (7 × 8 cm) was placed at the end of one arm chosen in a pseudo-random order. The pool was filled with water (19.5 ± 0.5 °C) to a depth of 9 cm. With this depth, the top of the hidden platform was 2 cm below the surface of the water. The spatial cues in the room were replaced and rearranged between the MWM and DMS tasks. Animals were tested over 8 consecutive days, with 6 trials per day during the first half of the light phase (i.e., between 9 A.M. and 12 noon). For each trial, the animal was placed in an arm adjacent to the arm with the platform. The platform remained in the same location through all trials on a given testing day but was moved to a new arm every testing day. Animals were placed into the water facing the end wall of the arm and were allowed to navigate the maze for 90 s. Once the animal located the platform, it was left atop the platform for 15 s to orient to the spatial cues. If the animal entered the arm containing the platform and touched the platform without climbing onto it, the trial was terminated and the animal was gently guided to the platform to begin its platform time. Animals were dried and given at least 5 min to rest in a dry warm cage between trials. The maze was cleaned of any floating particles and the water was stirred between each animal. On subsequent test days the platform was moved to a new arm. For the DMS task, the latency to the platform and the number of entries into the arms without the platform (errors) was recorded.

A different set of animals was used for the circadian assessment (n = 6 controls, n = 10 disrupted LD animals). These animals were 90–120 days old at the start of the circadian assessment. Animals were housed individually in cages equipped with running wheels. A magnetic switch detected the rotation of the wheel, and this was recorded by a computer running Clocklab (Actimetrics Inc. Wilmette, IL, USA). Animals were housed in 12:12 LD for 4 weeks followed by 2 weeks in constant darkness (DD). Activity levels in LD were assessed from days 17–29. An average waveform was created over these days. Activity was divided into six 4 h bins for analysis. Activity levels in DD were assessed over the first 14 days of DD in a similar fashion. Average waveforms were created based on the individual free-running period during these days. Average waveforms were aligned at CT12 between each animal, with CT12 being defined as the first 10 min bin with at least 20% of the maximum activity observed in any 10 min bin that was followed immediately by another such 10 min bin. Activity duration (alpha) in LD and DD were assessed by calculating the time between Clocklab-identified activity onsets on days 17–29 in LD or days 1–14 in DD. Phase angle of entrainment in LD was assessed by comparing the Clocklab-identified time of activity onset in LD on days 17–29 with the time of dark onset. Negative values represented activity onsets that occurred after dark onset. Free-running period was assessed over the first 14 days in DD using two approaches: 1) a χ² periodogram analysis, and 2) by measuring the slope of a regression line fit to the activity onsets. Additional circadian parameters were assessed in LD and DD using a cosinor analysis (R. Refinetti, https://www.circadian.org/ZIP/cosinor.zip↗) that reports mesor, acrophase, robustness and amplitude. Onset variability in LD and DD was assessed by fitting a regression line to activity onsets in Clocklab 6.1.02 and using the reported error term as a variability measure. Quality of the rhythms was also assessed using the χ² statistic reported in Clocklab, which is a least-squares goodness-of-fit statistic. Rhythms with highly stable outputs will have a larger χ² statistic, while rhythms that are variable and unstable day-to-day will have low χ² statistics.

Adult males were recorded for 30 days in parallel to the litters during either normal (n = 2) or disrupting LD (n = 2). Their locomotor rhythms were assessed using Clocklab’s F-Periodogram routine.

A different set of animals (n = 4 control LD, n = 8 disrupted LD) was used for the Golgi-Cox histological assessment. At PN90 animals were rendered unconscious with isoflurane and were then given an overdose of sodium pentobarbital (Euthanyl, ip, 500 mg/kg). They were then perfused with 0.9% NaCl. Brains were extracted and processed using the protocol described by Zaqout and Kaindl⁶⁰. Briefly, brains were immersed in Golgi-Cox solution for 10 days in darkness at room temperature. Brains were then transferred into tissue protectant where they were stored for 2–3 days in 4 °C temperature in complete darkness. The brains were then sliced using vibratome at 60 Hz and thickness of 200 µm, and the sections were mounted onto gelatin coated slides. At the end of the process, the slides were left to dry for 4–5 days in the dark. Slides were then dehydrated and developed according to the protocol detailed by Gibb and Kolb⁶¹. The slides were then coverslipped with 10 drops of Eukitt solution (Fluka Analytical, Germany). Finally, slides were sealed with nail polish and were left flat to dry in the dark for 48 h before imaging.

For each animal, researchers blinded to the treatment conditions traced 4 pyramidal cells from each of 3 regions of the brain: the mPFC, the CA3 region of the dHC, and the BLA. Tracings were conducted using the 40× objective on an Olympus BX-51 microscope equipped with a camera Lucida tracing arm. The traced neurons selected were well impregnated and not obscured by astrocytes or large clusters of neighboring dendrites. In addition, the apical and the basilar dendrites of the traced neurons were intact and visible in the plane of section. Dendrite length was calculated using Scholl analysis as described by Christensen and colleagues⁶². Complexity of neurons was assessed using the branching order approach described by Coleman and Reisen⁶³. The branching order was measured from both the apical and the basilar dendrites. Two researchers blinded to the treatment assessed each tracing, and the average of the two assessments was used for analysis unless large differences in assessments were calculated, in which case the individual assessments were redone.

Statistical analysis was carried out in SPSS version 26 and SigmaPlot 14.5. To study anxiety-like behavior of the disrupted mice compared to the controls, non-parametric Mann–Whitney U tests were conducted to study entries into, and time spent in, open arms. Latency to the hidden platform on training days of the MWM task was analyzed using a 2 (Groups: disrupted and control animals) × 4 (Days of training) × 4 (Trials) factorial ANOVA. To test MWM spatial memory performance on the probe day (day 5), independent samples t-tests were conducted to analyze latency to the previously located platform, the number of passes made through the previous platform location, and the time spent in the target quadrant. To examine performance in working memory in the DMS task, the latency to platform and the mean number of errors (i.e., entries into the arms other than the target arm with the platform) were analyzed using 2 (groups: disrupted, and control) × 6 (trials per day) × 8 (experiment day) factorial ANOVAs. For the circadian data, a two-way repeated measures ANOVA was conducted to study differences in total activity in light–dark condition (12–12 LD cycle) and in complete darkness (DD). Independent samples t-tests were used to test for differences in alpha in LD and DD, phase angle of entrainment in LD, and free-running period in DD. In cases where homogeneity of variance was not achieved, the non-parametric Mann–Whitney U test was used. Due to the unequal sample sizes, Welch’s F tests were used to analyze dendritic length and dendritic complexity in the Golgi-Cox analysis. Prior to conducting the ANOVA tests, the assumptions of normality, sphericity, and homogeneity of the variance were assessed. In the case where homogeneity of variance was not achieved, a square root transformation was applied to the data prior to running the ANOVA. Square root transformation was used to reduce the variability in the data in an unbiased manner. In cases where sphericity was not achieved, the Greenhouse–Geisser correction was employed. Significant interactions were followed up with pairwise comparisons using the Holm’s correction for multiple comparisons with alpha set at 0.05.