Induction of torpor in response to a common chronic food restriction paradigm: implications for behavioural research using mice

Introduction

The laboratory mouse is a widely used model organism for scientific research, accounting for 54% (934,200) of experimental procedures performed on animals in the UK in 2021 (Annual Statistics of Scientific Procedures on Living Animals Great Britain 2021). As part of many experimental procedures, fasting and food restriction are frequently used, for example in behavioural and sensory neuroscience (Padamsey et al., 2021; Pioli et al., 2014), metabolic studies (Ayala et al., 2010; Berglund et al., 2008), pharmacological studies (Chen et al., 2018; Huang et al., 2015), and circadian phenotyping (Greenwell et al., 2019; Northeast et al., 2019), amongst others.

The potential for fasting or food restriction to be a confound for experimental data is rarely considered within studies that utilise this technique, despite fasting being known to significantly alter physiology (Jensen et al., 2013). Food restriction is especially prevalent in behavioural neuroscience as a method for incentivising engagement with a task. Frequently, mice undergo chronic food restriction in which bodyweight is maintained at ~85-90%. However, behavioural data is subject to high levels of inter- and intra-individual variation, therefore requiring large sample sizes, although reproducibility is often limited (Baker & Penny, 2016; Begley & Ioannidis, 2015; Mandillo et al., 2008).

We hypothesised that torpor, a naturally occurring state of metabolic suppression and hypothermia in response to limited food availability, may be an unaddressed confounding factor in behavioural studies. It has been reported that laboratory mice enter torpor in response to fasting (Ambler et al., 2021; Hudson & Scott, 1979; Jensen et al., 2013), resulting in profound changes to central and peripheral physiology. For example, torpor has been associated with sleep disruption and altered brain activity (Huang et al., 2021; Northeast et al., 2019; Vyazovskiy et al., 2017), and altered levels of circulating hormones such as thyroid hormones (Bank et al., 2015), leptin (Gavrilova et al., 1999), and ghrelin (Gluck et al., 2006).

Here, we aimed to determine the likelihood of torpor induction in response to a food restriction paradigm common in behavioural studies. We focused on the influence of bodyweight due to previous studies demonstrating a link between bodyweight and torpor (Kato et al., 2018; Solymár et al., 2015), and due to weight loss being a key component of food restriction paradigms.

Methods

Animals and recording conditions

Adult, male C57BL/6J mice were obtained from an in-house colony maintained by the University of Oxford Biomedical Services (N=8, aged ~11 weeks at the start of the study). C57BL/6J was chosen as they are a common strain used for behavioural studies, or as a background for genetic mutants. Power calculations were not used for this study, as it was an exploratory study, and we expected that the effect size would be large due to a torpor bout typically involving a considerable (>5°C) drop in body temperature that is easily detected (Huang et al., 2021; van der Vinne et al., 2020). Furthermore, there are few existing studies of a similar nature which limits the ability to accurately perform power calculations.

A single sex was used to maximise the applicability of the data to existing data, as most research to date, unfortunately, has been only conducted in males, especially within the behaviour field. Although we have not performed this study in females to date, we would expect similar outcomes due to torpor being readily induced in female mice elsewhere (Zhang et al., 2020). Mice were individually housed in wire-top cages (48×15×13 cm) on a 12:12 h light-dark cycle for the duration of the experiment. Cages were housed in custom light-tight chambers (LTCs; Figure 1A), with four cages per chamber (Fisher et al., 2012; Tam et al., 2021). LTCs were illuminated by cool white LED strips (Maplin, UK; 200 lux at the cage floor) during the light phase. Ambient room temperature and relative humidity were maintained at 21±1°C and 60±1%, respectively. Water was provided ad libitum throughout.

Figure 1. Experimental design and setup.

(A) Mice were individually housed in standard M3 wire-top cages and placed under thermal imaging cameras (two cages per camera) which were used to continuously monitor environmental and skin temperature. Cameras were placed ~20 cm above the cage floor and positioned to ensure mice were in view. During food restriction, mice were fed at the same time each day and weighed ~2 hours after feeding. Mice were maintained at ~85% of free feeding bodyweight. (B) Thermal images of a mouse during euthermia (left), and a mouse during torpor (right). Skin temperature (Tskin) was measured by recording the warmest pixel in view every 1 s (1 Hz). Ta represents measurement of the ambient temperature (coldest pixel every 1 s). (C) Representative Tskin trace of a torpor bout during food restriction. Torpor bouts were detected by determining when Tskin had decreased by >3 standard deviation below the median euthermic temperature, for >1 h, for each mouse. Red circles indicate where torpor was determined to begin, and black circles indicate where torpor was determined to end using these criteria.

The change in peripheral skin temperature (Tskin) in response to food restriction was recorded using continuous thermal imaging cameras, mounted ~20 cm above the cage base (Optris^® Xi 80 compact spot finder thermal imaging camera with 80° wide angle lens, Optris GmbH, Berlin, Germany; one camera per two animals). Custom Perspex blocks (23×12×3.2 cm, Aquarius Plastics, Surrey, UK) were used to block access to the area beneath the food hopper to minimise mice going out of view of the camera. Additionally, bedding material (sizzle nest) was kept to a minimum, and enrichment items, such as tunnels, were removed to prevent mice from being obscured from view of the thermal cameras, although ~10 g of nesting material was provided to align with the thermal preference of mice (Gaskill et al., 2011).

Tskin was determined by recording the hottest pixel every 1s for the duration of the study using the manufacturer’s software (Optris^® PIX Connect, Optris GmbH). Data were processed by applying a high-pass filter 18°C and a low-pass filter at 35°C to remove artefacts that occurred due to movement and changes in animal posture. The filtered data were binned into 1-minute intervals, and a 20-minute moving average applied to smooth the data and further remove noise ready for analysis (Huang et al., 2021; van der Vinne et al., 2020).

Results

As expected, food restriction resulted in a significant decrease in bodyweight over time (F_{(4.11, 28.8)}=105.9, p<0.0001). Bodyweight dropped until day 8 of food restriction when weight stabilised between 85-90%, as was our aim in line with most behavioural food restriction paradigms (Figure 2A). To determine the effect of food restriction on body temperature, the change in mean Tskin from baseline (day 0, the last day of free feeding) was calculated. This method was chosen due to inter-individual variability in absolute Tskin measurements (van der Vinne et al., 2020). Tskin significantly decreased over days of food restriction, with the mean daily temperature dropping by ~3°C towards the end of the study (F_(3.43,24.0)=34.9, p<0.0001; Figure 2B). The drop in Tskin is likely due to the induction of torpor, and mice spending more time in torpor. This is supported by an increasing proportion of mice entering torpor between days 1–7 of food restriction, and the increasing amount of time spent in torpor each day (Figure 2C, 2F). From day 8 onwards, all mice were entering at least one torpor bout per day. Further, the distribution of Tskin measurements prior to food restriction was unimodal, indicating that body temperature was being maintained with little deviation from a setpoint, as expected for endothermic species. However, after ~1 week of food restriction the distribution shifted to bimodal, with a second peak at a much lower Tskin, corresponding to heterothermy (Figure 2G).

Figure 2. Change in body temperature, bodyweight, and torpor propensity during food restriction.

(A) Change in mean daily Tskin from baseline (day 0) over days of food restriction. Tskin was found to significantly decrease over days of food restriction (F_(3.43,24.0)=34.9, p<0.0001; one-way ANOVA with repeated measures) with a maximal decrease of 3°C. Data are represented as a mean value ± SEM (n=8). (B) Bodyweight, expressed as a percentage of baseline (day 0), significantly decreased over days of food restriction, as expected (F_(4.11,28.8)=105.9, p<0.0001; one-way ANOVA with repeated measures). Bodyweight stabilised at 85-90% on day 8 of food restriction. Data are represented as a mean value ± SEM (n=8). (C) Time course of the percentage of mice entering torpor on each day of food restriction. All mice reliably entered torpor each day from day 8 onwards (n=8) (D) Simple linear regression analysis of bodyweight (as a % of baseline) and change in body temperature compared to baseline (day 0). A strong positive relationship was found between the two parameters (r²=0.8989, p<0.0001). Each data point represents a day of recording. (E) A significant negative relationship was observed between the proportion of mice entering torpor and decrease of bodyweight (as a % of baseline). Note that all mice were entering torpor at 85% and 90% bodyweight, and almost all animals were entering torpor at slightly above 90% bodyweight. Each data point represents a day of recording. (F) The amount of time spent in torpor each day, expressed as a percentage, over days of food restriction. (G) Frequency distribution or recorded Tskin values from a representative mouse before food restriction (left), and during food restriction (right). A unimodal distribution of temperature is observed pre-food restriction, indicating maintenance around a set point. Two peaks are observed during food restriction, indicating a deviation from the euthermic set point, as is typical for heterothermy.

Next, we investigated whether our observations were related to decreasing body weight. A strong positive relationship was found between bodyweight and the change in Tskin, with Tskin decreasing as bodyweight dropped (r²=0.8989, p<0.0001; Figure 2D). Moreover, a strong negative relationship was found between bodyweight and the proportion of mice entering torpor on each day of food restriction (r²=0.8725, p<0.0001). Notably, a high proportion of mice were entering torpor when bodyweight was at ~90% of free feeding bodyweight, indicating that even a weight loss of 10% is sufficient to induce torpor in laboratory mice (Figure 2E).

Discussion

Our results show that torpor is readily induced in response to food restriction, typically occurring in all mice from day 8, although a high proportion of mice began entering torpor before this point. Notably, day 8 coincides with when bodyweight first began to stabilise at 85% bodyweight, which is often the target bodyweight required before behavioural testing commences (Pioli et al., 2014). Although bodyweight in this study was titrated to 85%, we still observed a high occurrence of torpor at bodyweights of 90%. Bodyweight can often increase to 90% during chronic food restriction paradigms; however, the data presented here suggest that even this degree of weight loss would be sufficient to induce torpor. As such, it is likely that torpor will be occurring regularly during the period of behavioural testing. Although it is unlikely that testing would be performed in torpid animals, as they are lethargic and show reduced behavioural responsiveness at low body temperatures, the short-term and delayed effects of torpor induction on brain function and behaviour are largely unknown. More generally, due to the profound changes in physiology associated with torpor, it is possible that variables of interest will be confounded due to previous torpor history. Although limited, some behavioural data gathered from post-torpid mice found that behavioural performance was significantly altered (Nowakowski et al., 2009). Moreover, a study in ground squirrels reported high levels of variability between individuals in memory retention when performing a previously learned task following torpor (Hensleigh et al., 2022).

Here, we focused on a chronic food restriction paradigm at ~85% of free-feeding bodyweight; however, previous work has shown that torpor can be induced after only 7 hours of total food removal (Brown & Staples, 2010; Swoap et al., 2006; Swoap & Weinshenker, 2008). Acute fasting is commonplace as part of metabolic and pharmacological studies (Ayala et al., 2010; Beumer et al., 2006), therefore, based on previous research, we anticipate that our findings are also applicable to studies where acute fasting is used.

Due to technical limitations of the thermal cameras, mice were singly housed throughout this study. Further, single housing allowed for more precise titration of the amount of food provided each day to maintain individual bodyweight at ~85%. Single housing is common practice during torpor studies in mice due to the use of thermal cameras (Hitrec et al., 2019), or due to surgery being required to implant temperature-sensitive telemeters (Oelkrug et al., 2011). However, group housing is often used during behavioural testing to reduce cage costs and for welfare benefits. There is some evidence in other species that torpor is still employed in group settings and can result in increased entries into torpor due to reduced energy savings because of social thermoregulation via huddling (Geiser, 2010; Jefimow et al., 2011). Moreover, a study in group housed sugar gliders found that food restriction resulted in synchronisation of torpor bouts (Nowack & Geiser, 2016). To our knowledge, social torpor has not been investigated in laboratory mice; however, evidence from other species suggests that the results presented here would be applicable to group housed mice.

Based on our data, we recommend that studies using both acute and long-term food restriction in mice are closely monitored for torpor bouts, for example, using continuous non-invasive thermal imaging. Even in group housed settings, thermal imaging could be used to detect sustained drops in body temperature corresponding to mice entering torpor. Doing so will enable data to be compared from cages with similar torpor histories before running a task or taking a sample. Moreover, knowing prior torpor history will allow for additional controls to be implemented and may help to explain unexpected differences between experimental units. From a 3Rs perspective, our recommendations can be used to refine experimental paradigms requiring food restriction, therefore helping to reduce variability.

Author contributions

SLW, SNP, DMB and VVV designed experiments. SLW performed measurements. SLW and VVV performed analysis. SLW wrote the manuscript with input from all other authors.