Edited by: Raffaele d’Isa, San Raffaele Scientific Institute (IRCCS), Italy
Reviewed by: Stefano Gaburro, Tecniplast, Italy; Sophie St-Cyr, Children’s Hospital of Philadelphia, United States; Mali Jiang, Johns Hopkins University, United States
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Monitoring the activity of mice within their home cage is proving to be a powerful tool for revealing subtle and early-onset phenotypes in mouse models. Video-tracking, in particular, lends itself to automated machine-learning technologies that have the potential to improve the manual annotations carried out by humans. This type of recording and analysis is particularly powerful in objective phenotyping, monitoring behaviors with no experimenter intervention. Automated home-cage testing allows the recording of non-evoked voluntary behaviors, which do not require any contact with the animal or exposure to specialist equipment. By avoiding stress deriving from handling, this approach, on the one hand, increases the welfare of experimental animals and, on the other hand, increases the reliability of results excluding confounding effects of stress on behavior. In this study, we show that the monitoring of climbing on the wire cage lid of a standard individually ventilated cage (IVC) yields reproducible data reflecting complex phenotypes of individual mouse inbred strains and of a widely used model of neurodegeneration, the N171-82Q mouse model of Huntington’s disease (HD). Measurements in the home-cage environment allowed for the collection of comprehensive motor activity data, which revealed sexual dimorphism, daily biphasic changes, and aging-related decrease in healthy C57BL/6J mice. Furthermore, home-cage recording of climbing allowed early detection of motor impairment in the N171-82Q HD mouse model. Integrating cage-floor activity with cage-lid activity (climbing) has the potential to greatly enhance the characterization of mouse strains, detecting early and subtle signs of disease and increasing reproducibility in preclinical studies.
Advances in the field of genetics mean that mouse models are increasingly sophisticated, and more closely than ever before the model human disease (Mingrone et al.,
Currently, a number of tests, including grip strength and gait analysis (Tucci et al.,
Investigating perturbations in the home-cage activity of undisturbed mice over extended periods can greatly enrich standard out-of-cage phenotyping and provide novel insights into subtle and progressive conditions at early time points. A number of systems have been developed to investigate motor activity over extended periods of time in single-housed as well as group-housed mice. However, measuring both cage-lid climbing and cage-floor movement simultaneously in group-housed mice remained a technical challenge (Bains et al.,
Over the past few years, there has been a concerted effort toward overcoming these challenges by housing the mice in testing chambers for extended periods of time and automatically measuring non-evoked activity (Bains et al.,
Despite some early promise shown by gene-targeting therapies, there is currently no disease-modifying treatment for HD, and therapy is focused on the management of symptoms and improving quality of life (Kim et al.,
This study used the B6-TgN(HD82Gln)81Dbo/H, also known as the N171-82Q, model of HD, first published in 1999, in which damage to the basal ganglia structures causes a hyperkinetic disorder (chorea) in combination with a loss of voluntary movements (bradykinesia and rigidity). These phenotypes become evident at around 10.5 weeks of age and manifest as abnormal gait and other behavioral and physiological abnormalities, such as lower grip strength, disturbed limb dynamics, and rigidity of the trunk, as well as a tendency toward a lower body weight (Schilling et al.,
Voluntary locomotion in mouse disease models is highly clinically relevant because it provides an insight into the physiology of the condition, as well as the behavioral motivation of the individual, and is a fundamental readout of the phenotype used in the diagnosis in human patients (Karl Kieburtz et al.,
Advanced image analysis, which can highlight changes in the animal’s gait in both the lateral and the ventral plane, has thus far proven to be the most sophisticated way of extracting subtle motor phenotypes earlier than 13 weeks of age in the model used in our study (Preisig et al.,
In this study, we demonstrate a new tool for the automated analysis of motor activity, which encompasses climbing as well as locomotion on the cage-floor, in undisturbed mice over multiple light: dark cycles. Through its application to the N171-82Q model of HD, we have uncovered a robust complex phenotypic profile for disease progression, including early features of motor dysfunction that are fundamental in developing reproducible digital biomarkers for therapeutic testing, especially when targeting the prodromal stages of HD.
In meeting these challenges, we developed an algorithm to automatically annotate climbing behavior from high-definition video captured from a side-on view of the home-cage. The resulting automated climbing behavior annotations provide an important additional parameter set that further enriches the activity profile captured by the Home Cage Analyser system (HCA; Actual Analytics Ltd., UK; Bains et al.,
All mice used in the study were bred in the Mary Lyon Centre at MRC Harwell and were housed in individually ventilated cages (IVCs; Tecniplast BlueLine 1284) in groups of three mice per cage on Eco-pure spruce chips grade 6 bedding (Datesand, UK), with shredded article shaving nesting material and small cardboard play tunnels for enrichment. The mice were kept under controlled light (light 07:00–19:00; dark 19:00–07:00), temperature (22°C ± 2°C), and humidity (55% ± 10%) conditions. They had free access to water (25 p.p.m. chlorine) and were fed
For the first study, 18 male and 18 female mice, in six cages of three mice each, from the inbred strain C57BL/6J were recorded at three time points: 13–14 weeks, 30–31 weeks, and 52–53 weeks of age. For the second study, mice from the mutant strain B6-TgN (HD82Gln) 81Dbo/H (HD), were recorded at three time points: 8 weeks, 13 weeks, and 15–16 weeks of age. Twenty-seven hemizygous (Hemi) males carrying the HD transgene, along with 24 male wild type (WT) littermate controls, and 33 hemizygous females carrying the HD transgene, along with 24 female WT mice, were housed in same-genotype groups of three mice per cage. Using the cage as the experimental unit, a sample number of six was calculated to be the most appropriate sample size based on data from previous studies (
Three days prior to recording sessions, the animals were transferred to clean home cages with fresh bedding, nesting material, and a cardboard rodent tunnel as enrichment material, in line with the standard husbandry procedures for IVC cages. The cages were then placed in an IVC rack in the experimental room for the animals to acclimatize. For each recording, the cages were randomly assigned to an HCA rig. On the first day of recording, each cage was placed onto the ventilation system, within the rig, as would occur during a normal husbandry procedure.
Animal welfare checks were carried out visually twice daily. At the end of the recording period, the home cages were removed from the HCA rigs and returned to their original positions on the IVC racks.
Radio frequency identification microchips were injected subcutaneously into the lower left or right quadrant of the abdomen of each mouse at 12 weeks of age for the C57BL/6J study and 7 weeks of age for the B6-TgN(HD82Gln)81Dbo/H study. These microchips were contained in standard ISO-biocompatible glass capsules (11.5 × 2 mm; PeddyMark Ltd., UK). The procedure was performed on sedated mice (Isoflo; Abbott, UK) after topical application of local anesthetic cream on the injection site prior to the procedure (EMLA Cream 5%; AstraZeneca, UK) as described in Hobson et al. (
Mice were housed in an individually ventilated cage which has a metal wire lid that allows climbing. Mouse activity was videorecorded through a high definition, infra-red camera of the Home Cage Analyser system (HCA; Actual Analytics Ltd., UK), and videos were subsequently analyzed offline through a video-recognition algorithm. Through this method, climbing behavior is measured on a frame-by-frame basis by numerically characterizing the pattern of motion occurring within a pre-defined region around the cage lid and quantifying its similarity to a set of key reference examples of climbing and non-climbing behavior (selected programmatically from a large set of human "training" annotations) to yield a classification decision. More specifically, the local trinary pattern representation proposed by Yeffet and Wolf (
It is however worth noting that there are other ways to combine the human annotations. For example, during training, a frame was treated as climbing if any one annotator deemed it so (the purpose of this was to ensure that no example of climbing was missed). If the test annotations are combined in this way, a greater proportion of the data is considered to be climbing (approximately 30%), and the accuracy scores of the automated climbing annotations change accordingly (85.6% frame-by-frame accuracy with 65.9% of climbing frames, and 94.3% of non-climbing frames correctly classified). Correlation over 5 min time bins is lower, albeit broadly similar (
To account for dependence between data recorded over separate days from the same cage (repeated measures) and to avoid pseudo-replication, statistical analyses were conducted using linear mixed-effects modelling. To adjust for parameters with non-normal distributions, data were box-cox transformed prior to analysis. Any subsequent modelling satisfied assumptions of normally distributed residuals.
We constructed a linear mixed-effects model of either distance travelled or time spent climbing (continuous variables) as a function of the effect of sex, age of caged mice, and genotype (all categorical fixed effects). Cage ID was modelled as the random effect intercept with day of recording as the random effect slope. This structure allows for cages to vary randomly in their baseline distance moved or time spent climbing value, and for this relationship to vary randomly according to the day of recording. It will account for time-dependent and cage-specific fluctuations in activity over the 3 days of recording.
This model was compared to other model iterations with different combinations of sex, age, and genotype with or without the interaction term and random effects structure. An ANOVA test was run to determine the statistical significance of the interaction between Age: Genotype: Sex and to inform model selection. Random effects and fixed effects found not to have a statistically significant contribution to model fit were eliminated. Models were fit using R’s “lmer” function.
The time-frames of interest in the current study were defined as the 30 min directly preceding lights being turned on (06:30–07:00) and 30 min directly preceding lights being turned off (18:30–19:00). This definition was consistent between both parameters of interest: distance travelled (mm) and time spent climbing (seconds). When analyzing activity during the time-frames of interest, we first summed data per time bin (6 min) for the mouse within a cage for distance travelled and cumulative climbing in the cage per time bin (6 min) for time spent climbing. We then calculated the average activity per cage across the 5 time bins.
One of the early indicators in clinical presentation of a number of neurodegenerative conditions is a perturbed circadian rhythm (Carter et al.,
We conducted pairwise
While the algorithm for automated behavior annotation is proprietary, the analysis is openly available as a part of this manuscript; please see “Data availability statement” in this manuscript. The datasets for the experiments in this manuscript are also openly available on request.
All climbing data were converted from a number of frames to time spent climbing in seconds prior to analysis and visualization, as the authors believe that to be a more intuitive parameter. The number of frames is converted to time in seconds as follows:
Each individual frame is 40 ms long.
The data from the females of inbred strain C57BL/6J show significantly higher cage-floor activity, described as distance travelled in mm (
Distance travelled by mice during the dark phase varies according to sex and across age during passive home-cage monitoring.
The data from the females of inbred strain C57BL/6J show significantly higher cage-lid climbing, described as time spent climbing in seconds (
Time spent climbing by mice during dark phase varies according to sex and across age during passive home-cage monitoring.
The data from the HD model recapitulate the clear circadian rhythm and sex differences seen in the inbred strain, in which females were significantly more active than males in both measures of activity (cage-floor activity, as seen in
Distance travelled and climbing activity varies according to genotype in female mice and across age at the conclusion of the dark phase, but not at beginning of the dark phase.
There is, however, a specific time-of-day-dependent deficit in activity seen in Hemi HD mice as compared to WT HD mice, at the transition between dark and light phases for females, not seen in the transition between light and dark phases (18:30–19:00;
This decrease in activity at the end of the dark phase (06:30–07:00) is also seen in male Hemi HD mice as compared to WT HD mice, allowing for the phenotype to be detected as early as 8 weeks of age (
Distance travelled and climbing activity varies according to genotype in male mice and across age at conclusion of dark phase, but not at beginning of dark phase.
This specific time-of-day-dependent deficit in cage-floor activity is mirrored in cage-lid climbing, described as time spent climbing, where female Hemi HD mice show a significant decrease in time spent climbing compared to female WT HD mice, at the transition between dark and light phases (06:30–07:00), not seen in the transition between light and dark phases (18:30–19:00;
As with cage-floor activity, time spent climbing also follows the same pattern in male Hemi HD mice as compared to male WT HD mice, where a statistically significant difference in genotypes is observed at 8 weeks of age (
Classically, with a few exceptions, all behavior testing is carried out during the light phase, in which resting animals are removed from their home cage and placed in a novel environment away from their cage mates (Bains et al.,
The advantage of observing the mice undisturbed within their home cage over multiple light:dark cycles is that, in addition to the observed phenotype, it is also possible to disentangle the temporal appearance of such phenotypic traits by focusing the analysis on specific time-frames of interest. The data from C57BL/6J show that there is a clear sexual dimorphism in the overall activity of the animals and that both males and females show peak activity in the dark phase. These data, therefore, point to a clear circadian influence. This method of analysis allows one to also investigate the influence of ultradian parameters on measures such as cage-floor activity and cage-lid climbing. The importance of understanding these circadian and ultradian effects is highlighted in the HD study, in which the males show very low baseline activity in both mutant and WT strains. We have already shown that the HCA system is capable of detecting statistically significant changes in activity around the light phase changes between various background strains (Bains et al.,
Through the current study, we showcase a recently developed automated behavior annotation tool for climbing behavior in standard IVCs under group-housed conditions. We have previously shown the capabilities of the system in investigating cage-floor activity in group-housed mice in their home cages (Bains et al.,
Sexual dimorphism in climbing behavior has been reported previously in singly tested C57BL/6Ntac mice, using the LABORAS system, in which the main aim of the study was to investigate the difference in response to novelty between the sexes. Furthermore, parameters were only measured for 10 min (Borbélyová et al.,
Serious motor and cognitive deficits that are the hallmark of HD are often preceded, by decades, by more subtle changes in circadian rhythms and motor function (Wang et al.,
The onset of HD is often insidious and progressive and the phenotypes are biphasic; at early stages of the condition involuntary functions are affected and in the later stages the directly controlled, voluntary functions begin to fade. This means that motor phenotypes are often expressed as hyperkinesia in the early stages and akinesia in the later stage (Kim et al.,
The HD data in the current study show a significant increase in signal in both cage-floor and cage-lid climbing activities around the transition between the light-to-dark and dark-to-light phases. There is ample evidence to show that spontaneous cage-lid climbing is mediated through the dopaminergic system and therefore depends on the motivation and arousal state of the mouse (Costall et al.,
The importance of this finding is highlighted in the set of experiments carried out using the mouse model of HD, N171-82Q. These data recapitulate the sex differences seen in the C57BL/6J strain experiments, in which females were significantly more active than males in both measures of activity, with this difference persisting across all time points. However, of note is a specific time-dependent decrease, compared with WT mice, in cage-lid climbing activity at the transition between dark and light phases, which was apparent as early as 8 weeks, when the animals showed no overt signs of the disease onset. The decrease in cage-floor activity at this time was also observed at 8 weeks of age; however, the decrease in cage-floor activity became even more apparent at 13 weeks of age. By 15–16 weeks of age the mice showed clear signs of disease and differences between the two genotypes were distinct. This last finding is of particular interest as clinical case studies, as well as mouse models of HD, are known to present with sleep disturbances, one of the hallmarks of the condition (Pallier et al.,
This study recapitulates the aspect of the disease in which the offset of activity in HD mice is observed sooner than that in their WT counterparts, for both cage-floor activity, as well as cage-lid climbing. Indeed, a 2005 study comparing the circadian activity patterns of human patients with a different mouse model of Huntington’s disease (R6/2), reported a similar pattern of decline in activity towards the end of the active phase with disease progression (Morton et al.,
In progressive degenerative conditions, neuronal dysfunction occurs before any overt signs of the condition become apparent in the behavior. As neurons are unable to regenerate, most therapies under development focus on neuroprotection, with the aim of slowing the progression of the disease and, where possible, delaying the onset (Jin et al.,
It is important to remember, however, that animal behavior is complex, and that external factors, such as the effects of diet and exercise, can have an impact on disease progression (Dutta et al.,
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/
Animal studies described here were subject to the guidance issued by the Medical Research Council in Responsibility in the Use of Animals for Medical Research (July 1993), were dependent on an institutional Animal Welfare and Ethical Review Body evaluation and were carried out in compliance with the Animals (Scientific Procedures) Act 1986, UK, Amendment Regulations 2012 (SI 42012/3039).
RB was responsible for the experimental design, experimental procedure, data collection and manuscript preparation. HF carried out bioinformatics and statistical analysis. RS was responsible for the system design, including automated climbing algorithm. JA contributed to the study design and system design. MS contributed to the study design. PN contributed to the study design, carried out circadian and activity data analysis, and prepared the manuscript. SW contributed to the study design, including animal procedures, and prepared the manuscript. All authors contributed to the article and approved the submitted version.
This work was supported by the Medical Research Council, Strategic Award A410-53658 (RB, HF, MS, PN and SW).
We wish to thank the IT Infrastructure team at the Mary Lyon Centre for their support with the hardware. We also wish to thank the animal care team at the Mary Lyon Centre for their help and technical support. A special thanks to Dr. Louise Tinsley, for copyediting and formatting this manuscript. Finally, we wish to thank the National Centre for 3Rs for their continued support of the home-cage concept.
RS and JA were employed by and were shareholders in Actual Analytics Ltd at the time the research was performed and therefore declare a competing financial interest. Actual HCA is commercially available from Actual Analytics Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
The Supplementary Material for this article can be found online at: