Thermographic imaging of mouse across circadian time reveals body surface temperature elevation associated with non-locomotor body movements

Abstract

Circadian clocks orchestrate multiple different physiological rhythms in a well-synchronized manner. However, how these separate rhythms are interconnected is not exactly understood. Here, we developed a method that allows for the real-time simultaneous measurement of locomotor activity and body temperature of mice using infrared video camera imaging. As expected from the literature, temporal profiles of body temperature and locomotor activity were positively correlated with each other. Basically, body temperatures were high when animals were in locomotion. However, interestingly, increases in body temperature were not always associated with the appearance of locomotor activity. Video imaging revealed that mice exhibit non-locomotor activities such as grooming and postural adjustments, which alone induce considerable elevation of body temperature. Noticeably, non-locomotor movements always preceded the initiation of locomotor activity. Nevertheless, non-locomotor movements were not always accompanied by locomotor movements, suggesting that non-locomotor movements provide a mechanism of thermoregulation independent of locomotor activity. In addition, in the current study, we also report the development of a machine learning-based recording method for the detection of circadian feeding and drinking behaviors of mice. Our data illustrate the potential utility of thermal video imaging in the investigation of different physiological rhythms.

Fig 1. Infrared video camera-based monitoring of circadian body surface temperature.

(A) Schematic representation of video imaging and representative thermography. The inset indicates temperatures of the top 1000 pixels of the mouse. Arrows indicate the 200th highest temperature. (B) Representative BST records of C57BL/6J mice under light/dark (LD) cycles for 2 days. BST was calculated every 0.5 s and smoothed once with a 120 point moving average. (C) Averages of 3 days recording of BST. Data are presented as mean ± standard error of mean (SEM); n = 4 mice.

Fig 2. Infrared video camera-based monitoring of circadian body movement.

(A) Representative imaging analysis for the determination of mouse position. The process involves binarization, clustering (DBSCAN), and calculating mouse position. A green dot on the red segments indicates the centroid of the area of mouse. A white arrow indicates the areas open but previously warmed by mouse. (B) Representative BST and BM records of C57BL/6J mice under LD for 2 days. (C) Averages of 3 days recording of BM and BST. Data are presented as mean ± SEM. n = 4 mice. (D) Scatter plots showing the relationship between BST and BM. Each data point represents averages of BST (y-axis) and BM (x-axis) for every 1 h. Each color represents an individual mouse (n = 4). Diagonal lines represent linear fitting of the data of each animal.

Fig 3. Identification of non-locomotor activities-associated BST elevation.

(A) Representative BST and BM profiles under different BM states. A period of increase in BST prior to the initiation of locomotor activity is highlighted by a yellow background. Image data on the right show representative mouse positions under resting (#1), non-locomotor movements (#2), and locomotor movements (#3). White arrows on the images indicate the nose-tail directions of the mouse. (B) Alignments of all resting to movement transition events (n = 124 events for four mice). The upper and lower graphs indicate BM and BST data, respectively. For BST, average of 500 s before transition was set to 0. Time 0, time of transition. (C) Time course of average BST increase from resting-to-movement transitions in each cluster. (D) Total BM activity after the transition. *P < 0.05, ****P < 0.0001, Welch ANOVA post-hoc Games-Howell test. (E) Phase distribution of incidence of non-locomotor movements that are not accompanied by locomotor movements (n = 4 mice). Data from independent animals are color-coded. Non-locomotor activities that continued at least 60 s were selected. Arrows in the circle are the Rayleigh plot vector. The black vector indicates the average of all groups.

Fig 4. Infrared video camera-based monitoring of circadian feeding and drinking behavior.

(A) Representative imaging analysis for the determination of feeding and drinking behavior. The process involves thermal imaging, pose estimation (DeepLabCut), and evaluation of access to food or water (tip of water bottle). (B) Representative BST, BM, feeding, and drinking records of C57BL/6J mice under LD cycle for 2 days. (C, D) Averages of 3 days recording of feeding (C) and drinking (D). Data are presented as mean ± SEM. n = 4 mice.