Abstract
Organisms have evolved circadian rhythms in behavior to anticipate daily opportunities and challenges such as mating and predation. However, the ethological investigation of behavioral rhythms has been traditionally limited to studying easy-to-measure behaviors (such as locomotor activity) on a circadian timescale or difficult-to-measure behaviors with limited temporal resolution. Here we sought to examine eight overt behaviors never before studied as a function of time of day, sex, light cycle, and neuropeptide signaling. We hypothesized that sex and neuropeptide signaling-dependent differences in daily behaviors have been largely missed because of the focus on running wheel activity in rodents. To address this hypothesis, we used high-throughput machine learning to automatically score complex behaviors from millions of video frames of singly housed, freely behaving male and female mice. Automated predictions for each of the eight behaviors correlated highly with consensus labels by trained human classifiers. We discovered reliable daily rhythms in eating, drinking, grooming, rearing, nesting, digging, exploring, and resting behaviors that persisted in constant darkness. We found that the overall frequency of most behaviors was predominantly affected by light cycle, but the amplitude and peak time of circadian rhythms in multiple behaviors were each dramatically influenced by neuropeptide signaling and sex. We conclude that machine learning can be used to reveal novel daily rhythms in behaviors that depend on sex, neuropeptide signaling, and ambient light and will allow for the rapid circadian phenotyping of mice with different genotypes or disorders.
Competing Interest Statement
The authors have declared no competing interest.