Evaluation of the Digital Ventilated Cage® system for circadian phenotyping

All animals were single housed under a 12:12 LD cycle for a minimum of one week prior to the start of experiments. Ad libitum access to food (Envigo 2916) and water was provided throughout. Wild-type C57BL/6J animals were purchased from Envigo, UK, while Cryptochrome 1 and 2 double knock-out (Cry1^−/−, Cry2^−/−) animals²⁶ were bred in-house on a C57BL/6J background. This study is reported in accordance with the ARRIVE guidelines. All animals were included and were between 3 and 4 months old at the beginning of the study. Control and experimental groups were subjected to the same conditions at the same time to reduce confounding variables. As such, no randomisation or blinding was used. All experimental procedures were carried out at the University of Oxford, UK, in accordance with the University of Oxford Policy on the Use of Animals in Scientific Research and the United Kingdom Animals (Scientific Procedures) Act 1986, under Project License PP0911346 and Personal License I34186130.

The DVC system (Tecniplast SpA, Italy) allows for the continuous recording of home cage locomotor activity⁵². The system comprises 60 or 80 individually ventilated cages (IVCs) of type GM500, made of clear, red or black polycarbonate, offering options for both single and group housing (Fig. 1A). In this study, we utilised black IVCs, which were equipped with individual cool white LED systems known as Leddy. The Leddy lighting system facilitated the entrainment of animals to specific lighting cycles within each cage, eliminating the need for light-tight chambers typically used in circadian and sleep research. Light levels were kept at ~ 100 photopic lux at the centre of the cage (Leddy light intensity level 22), giving < 2 S-cone opic lux, 70 melanopic lux, 75 rhodopic lux and 78 M-cone opic lux⁶¹ (Supp. Fig. 1; Supp. Table 1).

Each IVC was positioned on a capacitive sensor board featuring 12 evenly spaced electrodes (Fig. 1B). This sensor board detected the presence of an animal by monitoring changes in capacitance across electrodes, primarily due to the animal’s physiological composition (70% water) and movement. Subsequently, an Animal Locomotion Index (ALI) was calculated. This is based on the number of electrodes activated, with 0% = no electrodes and 100% = all 12 electrodes simultaneously activated within the time bin (Fig. 1C). The DVC Analytics platform provided real-time access to the data for visual inspection and download. The system’s default minimum activity binning was set to 1 min, although a resolution of 250 msec is possible. We performed daily welfare checks, with the cages being taken out of the rack for animal inspection for less than 5 min.

A total of 12 WT mice (C57BL/6J, 6 males, 6 females) were included in the circadian screen. Home cage locomotor activity was recorded continuously for 10 weeks under various lighting conditions, including 12:12 LD, DD, LL, LD following a 6-hour phase advance (jet-lag), and DD after exposure to a light pulse (LP) at ZT14-16 to measure phase shifting responses. A cage change was performed after the first period of DD (day 29), and before animals were re-entrained to a 12:12 LD cycle.

Cry1^−/−,Cry2^−/− double knockout mice (N = 6, 3 males, 3 females, N = 6 C57BL/6J controls) animals were used as a circadian clock deficient model. Cryptochrome-deficient mice are characterized by the deletion of the core clock genes Cryptochrome 1 and 2²⁶. These mice have no circadian clock and exhibit arrhythmic behaviour under constant conditions. This animal model was employed to assess whether the DVC system could effectively detect phenotypic differences based on locomotor activity, particularly under LD and DD conditions (1 week each).

Raw animal locomotion index data were exported from the DVC Analytics platform in 1 min and 1 h bins. Chi-square periodogram and activity onsets were computed in ActogramJ⁶². A 30-min running average was applied in order to account for missing values (daily checks) and to compute actograms. Inter-daily stability, intra-daily variability and relative amplitude were coded in R based on the equations provided in Witting et al.⁶³. Light phase activity and total activity were calculated in R. One week of data was used to compute the circadian disruption measures for each lighting condition. Due to a user error, one female mice is missing the final 17 h of activity recording during the DD period. Statistical testing and plotting were conducted in Prism 10. One-way or two-way repeated-measures ANOVA, with a Geisser-Greenhouse correction (lack of sphericity) and Holm- Šídák’s multiple comparisons test were computed. When comparing two groups, unpaired nonparametric Mann-Whitney tests or Wilcoxon matched-pairs signed rank tests were used, as the distribution of the data was not Gaussian. α = 0.05 was adopted in all analyses. Mean ± standard error of the mean (SEM) was used for all plots.

To assess the suitability of the DVC system for circadian phenotyping studies, mice were singly housed in DVCs, and locomotor activity was monitored under LD, DD, LL, a 6-hour phase advance, and DD following a LP. A representative double plotted actogram is shown in Fig. 2 and illustrates entrainment to the various light-dark cycles, with orange dots indicating activity onsets. Home cage locomotor activity rhythms in LD, DD and LL, were further analysed using standard measures of circadian disruption¹⁴.

The Chi-square periodogram was employed to quantify the strength and regularity of the activity rhythms (Fig. 3A), expressed through the Qp value, a ratio of variances⁶⁴. A significant effect of lighting condition on the periodogram power was observed (Lighting, F_{(1.565, 17.22)} = 4.507, P = 0.0340), with a higher Max Qp value under DD compared to LL (mean diff = 1464, P = 0.0047; Fig. 3B), indicative of more robust rhythms in DD. Activity patterns under entrained conditions are typically influenced by environmental cues. Under a 24-hour LD cycle, C57BL/6J mice generally display a 24-hour period. However, in DD, the endogenous circadian period is expressed, resulting in a free-running period of approximately 23.5 h⁶⁵. By contrast, constant light suppresses night-time activity, causing the period of activity to lengthen to more than 24 h⁶⁶. Our results showed that the period statistically differed between lighting conditions (Lighting, F_{(1.121, 12.34)} = 106.3, P < 0.0001), and aligned with the outlined patterns, revealing mean periods of 23.9 h (SEM = 0.04) under LD, 23.5 h (SEM = 0.02) under DD, and 25.1 h (SEM = 0.12) under LL, with higher variability under LL (Fig. 3C).

The strength of a rhythm is often associated with its day-to-day consistency and can be measured by its inter-daily stability (IS). IS reveals whether the 24-hour activity pattern remains reproducible from one day to another⁶³. Under a 12:12 LD cycle, high IS values are expected due to the presence of consistent time cues. Under DD, and in the absence of such cues, IS is expected to decrease slightly. Under LL, constant light weakens the circadian clock in the SCN, leading to reduced day-to-day reproducibility. Here, IS values decreased accordingly from LD, to DD and LL (Lighting, F_{(1.437, 15.81)} = 16.71, P = 0.0003; Fig. 3D).

Intra-daily variability (IV) quantifies the fragmentation of activity rhythms, reflecting the number of transitions between activity and rest throughout the day⁶³. Our data showed that activity rhythms remained consolidated, with no significant differences between lighting conditions (Lighting, F_{(1.598, 17.58)} = 0.7966, P = 0.4404; Fig. 3E).

The overall number and distribution of activity bouts did not statistically differ between lighting conditions (Lighting x Bout length, F_{(14, 231)} = 0.8822, P = 0.5789; Fig. 3F). This suggests that the structure of activity patterns was consistent across conditions.

Relative amplitude (RA) is determined by the difference between the period of maximum and minimum activity levels over the 24-hour cycle. It serves as a simple measure of the contrast between periods of high activity and rest⁶³. In the absence of external cues, such as DD, the strength of activity rhythms often declines, leading to a decrease in relative amplitude. This decrease tends to be even more pronounced under LL, where light during the subjective night suppresses mouse activity. Consistent with these patterns, here, the relative amplitude varied between lighting conditions (Lighting, F_{(1.465, 16.12)} = 4.942, P = 0.0295), with a decrease in both DD and LL conditions when compared to LD (Fig. 3G).

We subjected animals to a 6-hour phase advance of the LD cycle and recorded the duration required for them to re-entrain to the new LD cycle. Results showed that it required 6 days for animals to successfully re-entrain to the LD cycle, suggesting that one day of adjustment is required per hour of phase shift (Fig. 4A).

Additionally, we investigated whether a LP administered at the beginning of the night (ZT14-16) could shift the phase of activity²⁸. When comparing the activity onset of animals one day before the LP (Day − 1) and four days after (Day 4), we observed a delay of about 1.3 h, confirming a shift in the phase of activity consistent with an Aschoff type II protocol (Wilcoxon matched-pairs signed rank test, z = -1.375, P = 0.0005; Fig. 4B).

Growing evidence shows that male and female mice may exhibit differences in circadian rhythms. These differences can manifest in various aspects, including the timing and amplitude of activity patterns, sleep-wake cycles, and responses to environmental cues⁶⁷. Despite the tendency of some studies to exclude female cohorts due to concerns about increased variability associated with hormonal fluctuations during estrus, investigating both sexes is essential for gaining a comprehensive understanding of circadian regulation and how it may differ between males and females^68,69. Subtle differences between the activity profiles of male and female mice were observed (Supp. Fig. 2). We observed a significant main effect of Sex on IV and total activity (IV, F_{(1, 10)} = 14.52, P = 0.0034; total activity, F_{(1, 10)} = 14.85, P = 0.0032). Overall, females demonstrated more consolidated rhythms and greater activity levels compared to males. Furthermore, a significant Lighting x Sex interaction was found for both period and RA (Period, F_{(2, 20)} = 3.982, P = 0.0350; RA, F_{(2, 20)} = 4.953, P = 0.0179). Females displayed a shorter period (Mean Diff = -0.095) and higher amplitude (Mean Diff = 0.018) under LD, a longer period (Mean Diff = 0.055) and higher amplitude (Mean Diff = 0.006) under DD, and a longer period (Mean Diff = 0.445) and lower amplitude (Mean Diff = -0.045) under LL. The number and distribution of activity bouts differed between sexes under LD (Bout length x Sex, F_{(7, 70)} = 4.091, P = 0.0008), and LL (Bout length x Sex, F_{(7, 70)} = 5.040, P = 0.0001), but not under DD (Bout length x Sex, F_{(7, 70)} = 1.451, P = 0.1991). Under LD and LL, females showed more long bouts of activity. Females also exhibited a greater delay in activity onset in the days following a LP (Sex, F_{(1, 10)} = 15.47, P = 0.0028), but did not show significant differences in re-entrainment following a 6-hour jet lag (Sex, F_{(1, 10)} = 0.5986, P = 0.4570; Supp. Fig. 3). Together, these findings highlight the importance of considering both male and female animals in circadian phenotyping studies.

In the mammalian circadian clock, CRY1 and CRY2 proteins inhibit the activation of the CLOCK-BMAL1 transcription complex, thereby regulating the expression of clock-controlled genes and ultimately controlling the timing of physiological processes and behaviours⁷⁰. Deletion of both Cry1 and Cry2 results in arrhythmic behaviour under constant conditions such as DD²⁶. Cryptochrome-deficient mice have also been characterized by abnormalities in cognition and habituation behaviour, and increases in immobility, restlessness and anxiety-like behaviour^71,72. Here, we investigate whether home cage activity monitoring using the DVC system could detect arrhythmic mouse models.

Representative actograms of cryptochrome-deficient mice show apparently normal nocturnal activity under LD conditions, but arrhythmic activity under DD in comparison with free-running activity in C57BL/6J wildtype mice (Fig. 5). Chi-square periodogram analysis indicated no significant difference in the period of activity rhythms between wildtype and Cryptochrome-deficient animals under 12:12 LD cycles (M = 23.99, SEM = 0.009, and M = 24.06, SEM = 0.078, respectively; Fig. 6A). However, Cryptochrome-deficient animals displayed lower Max Qp values (Genotype, F_{(1, 10)} = 19.31, P = 0.0013; Fig. 6B), earlier activity onsets (Genotype, F_{(1, 10)} = 259.4, P < 0.0001; Fig. 6C), more light-phase activity (Mann-Whitney, U = 0, P = 0.0022, Fig. 6D), and a higher frequency of short bouts of activity compared to wildtype mice (Genotype x Bout length, F_{(7, 70)} = 4.211, P = 0.0006; 1 min bout, mean diff = -135.3, P < 0.0001; Fig. 6E). This suggests that Cryptochrome-deficient animals exhibited less robust and more fragmented rhythms under LD, with increased daytime activity. Under DD, Cryptochrome-deficient mice showed a lack of rhythmicity, as anticipated for animals lacking a functional endogenous clock in the absence of zeitgebers (Fig. 5). Periodogram analysis identified a high variance in their circadian period and lower Max Qp values (Genotype, F_{(1, 10)} = 19.31, P = 0.0013; Fig. 6B). However, there was no statistical difference between the distribution of their activity bouts compared to WT animals (Genotype x Bout length, F_{(7, 70)} = 0.1472, P = 0.9938; Fig. 6F). Furthermore, standard measures of circadian disruption highlighted lower IS (Genotype, F_{(1, 10)} = 38.95, P < 0.0001; Fig. 6G), higher IV (Genotype, F_{(1, 10)} = 6.571, P = 0.0282; Fig. 6H), and lower RA values (Genotype, F_{(1, 10)} = 122.6, P < 0.0001; Fig. 6I) in Cryptochrome-deficient mice compared to wildtypes under both LD and DD conditions. These findings suggest less reproducible day-to-day rhythms, increased fragmentation, and weaker amplitude in animals lacking a circadian clock. Although Cryptochrome-deficient animals appeared less active overall, the statistical tests did not reach significance, which is likely due to the wildtype’s high variance (Genotype, F_{(1, 10)} = 3.969, P = 0.0743; Fig. 6J). Additionally, sex differences were found for period (Sex, F_{(1, 4)} = 25.88, P = 0.0070) and Max Qp values (Sex, F_{(1, 4)} = 12.64, P = 0.0237), but no statistically significant differences were observed for circadian disruption measures between male and female cryptochrome-deficient mice (Supp. Fig. 4).

Despite its strengths, the current format of the DVC system does have several potential limitations that circadian researchers should be aware of.

As can be seen from the above, the DVC system offers many advantages, including high-capacity monitoring, individual LD cycle control, enhanced biosecurity, environmental monitoring features, high resolution, real-time accessibility and user-friendly online analytics. However, new users should consider potential limitations, such as the absence of light sensors and the charging logistics of the Leddy system, that may affect its performance in circadian rhythm studies. Additionally, research flexibility may be key. The DVC system might not work for researchers aiming to perform complex light cycle manipulations or specific types of studies where an IVC format is unsuitable.

Here we show that the DVC system provides an alternative tool for circadian phenotyping, providing an opportunity to conduct such assessments in IVC cages. The storage and real-time availability of the data on the cloud-based DVC analytics platform facilitates data analysis, as well as ongoing experimental monitoring and interventions, providing an advantage over existing systems. However, researchers should be aware of the current limitations particularly with regard to cost and research flexibility, which may influence the adoption of an IVC based system. In this regard, the DVC system may be more suited for central phenotyping facilities rather than individual circadian labs.

The authors (ST and SNP) have attended meetings funded by Tecniplast. However, Tecniplast had no involvement in the study design, data interpretation, or conclusions drawn from the research findings. As such, the authors declare no conflict of interest regarding the use of the DVC system.

The animal study was reviewed and approved by the Clinical Medicine Animal Welfare and Ethical Review Body (AWERB), University of Oxford.

Raw activity data with metadata for all data are available on GitHub (https://github.com/selmatir/DVC-Circadian-Phenotyping↗).