Hibernation in a primate: does sleep occur?

During hibernation, critical physiological processes are downregulated and thermogenically induced arousals are presumably needed periodically to fulfil those physiological demands. Among the processes incompatible with a hypome tabolic state is sleep. However, one hibernating primate, the dwarf lemur Cheirogaleus medius, experiences rapid eye movement (REM)-like states during hibernation, whenever passively reaching temperatures above 30°C, as occurs when it hibernates in poorly insulated tree hollows under tropical conditions. Here, we report electroencephalographic (EEG) recordings, temperature data and metabolic rates from two related species (C. crossleyi and C. sibreei), inhabiting high-altitude rainforests and hibernating underground, conditions that mirror, to some extent, those experienced by temperate hibernators. We compared the physiology of hibernation and spontaneous arousals in these animals to C. medius, as well as the much more distantly related non-primate hibernators, such as Arctic, golden-mantled and European ground squirrels. We observed a number of commonalities with non-primate temperate hibernators including: (i) monotonous ultra-low voltage EEG during torpor bouts in these relatively cold-weather hibernators, (ii) the absence of sleep during torpor bouts, (iii) the occurrence of spontaneous arousals out of torpor, during which sleep regularly occurred, (iv) relatively high early EEG non-REM during the arousal, and (v) a gradual transition to the torpid EEG state from non-REM sleep. Unlike C. medius, our study species did not display sleep-like states during torpor bouts, but instead exclusively exhibited them during arousals. During these short euthermic periods, non-REM as well as REM sleep-like stages were observed. Differences observed between these two species and their close relative, C. medius, for which data have been published, presumably reflect differences in hibernaculum temperature.

During hibernation, critical physiological processes are downregulated and thermogenically induced arousals are presumably needed periodically to fulfil those physiological demands. Among the processes incompatible with a hypometabolic state is sleep. However, one hibernating primate, the dwarf lemur Cheirogaleus medius, experiences rapid eye movement (REM)-like states during hibernation, whenever passively reaching temperatures above 30°C, as occurs when it hibernates in poorly insulated tree hollows under tropical conditions. Here, we report electroencephalographic (EEG) recordings, temperature data and metabolic rates from two related species (C. crossleyi and C. sibreei), inhabiting high-altitude rainforests and hibernating underground, conditions that mirror, to some extent, those experienced by temperate hibernators. We compared the physiology of hibernation and spontaneous arousals in these animals to C. medius, as well as the much more distantly related non-primate hibernators, such as Arctic, golden-mantled and European ground squirrels. We observed a number of commonalities with non-primate temperate hibernators including: (i) monotonous ultra-low voltage EEG during torpor bouts in these relatively cold-weather hibernators, (ii) the absence of sleep during torpor bouts, (iii) the occurrence of spontaneous arousals out of torpor, during which sleep regularly occurred, (iv) relatively high early EEG non-REM during the arousal, and (v) a gradual transition to the torpid EEG state from non-REM sleep. Unlike C. medius, our study species did not display sleep-like states during torpor bouts, but instead exclusively exhibited them during arousals. During these short euthermic periods, non-REM as well as REM sleep-like stages were observed. Differences . [24]. As a result, it remains unclear if the spontaneous arousals are needed in order to satisfy a need for sleep which is not met during hibernation [25].
All of these studies have been carried out in non-primate temperate hibernators. Within primates, obligate hibernation has so far only been described in dwarf lemurs (genus Cheirogaleus), a group of small-bodied (130-600 g), highly frugivorous, night-active Malagasy lemurs. Despite living in tropical environments, these lemurs are confronted with similar challenges to those of their Arctic and temperate counterparts, such as resource scarcity and cold temperatures during the dry winter season. For instance, the fat-tailed dwarf lemur (Cheirogaleus medius), inhabiting dry forests, exclusively hibernates in tree hollows [12,26], whereas the 'highland' Cheirogaleus (e.g. C, sibreei) retreats underground during the hibernation season [13]. Thus, the animal's body temperature and metabolic rate can remain almost constantly low during hibernation (e.g. underground sites or hollows in large trees), or fluctuate according to the insulation properties of tree hollows. If dwarf lemurs can reach body temperatures above 30°C by passively tracking ambient temperatures, they avoid endogenously generated arousals; if dwarf lemurs hibernate underground in colder areas, they actively undergo energetically expensive arousals about once a week [12,26].
We have previously reported on studies of hibernation in C. medius, which we studied both in Madagascar, as well as at the Duke Lemur Center, NC, USA [27,28]. Cheirogaleus medius hibernates at temperatures between 10 and 30°C, for periods of up to seven months a year in the wild [12]. The EEG of hibernating C. medius, passively warmed without expressing endogenously generated arousals, is marked by rapid-eye movement (REM) sleep-like periods and an absence of non-REM sleep when body temperature is above approximately 25°C. Sleep does not occur at lower temperatures [28]. These findings were at odds with those from Arctic ground squirrels under laboratory conditions (23°C), where periods of such REM-like activity were absent during hibernation bouts [15].
Here, we provide data obtained during hibernation in two related species of dwarf lemurs, C. crossleyi and C. sibreei, which inhabit the colder uplands of Madagascar and hibernate underground, rather than in tree hollows [13]. These underground hibernacula are at a fairly constant temperature of about 12-15°C throughout a 24-h cycle, which is substantially lower than the mean temperatures of the tree hollowdwelling western C. medius at approximately 23°C [12]. We selected these dwarf lemurs because they are biologically close to western C. medius but mirror environmental conditions experienced by temperate hibernators (i.e. underground temperature-buffered hibernacula). Thus, these data will shed light on the possible role of environmental conditions in the expression of hibernation.

Study site and study animals
We conducted this study in a 200 ha forest fragment at Tsinjoarivo, one of the last remaining high-altitude rainforests in eastern Madagascar (19°41 15 S, 47°46 25 E, 1660 m) [13]. At this location, there are two sympatric dwarf lemur species, C. crossleyi (approx. 350 g) and C. sibreei (approx. 250 g). Crossley's dwarf lemurs have a broad geographical distribution and occupy a variety of eastern rainforests including degraded environments and littoral forests, whereas Sibree's dwarf lemurs are high-altitude specialists [29,30]. The duration of hibernation varies between species and across individuals, but it ranges between three and seven months a year. During the hibernation period (April-September), dwarf lemurs of both species retreat into the ground [13].

Trapping schedule and temperature data
Dwarf lemurs were captured in March 2012 and January-February 2013 using Tomahawk traps (Tomahawk Live Trap, Hazelhurst, WI, USA) baited with small pieces of fermented banana. Traps were set between 3-10 m high along known trails at 17.00 and checked the following morning at 4.00. All captured lemurs were brought back to the campsite, where they were identified/marked, weighed and measured. Captured individuals were released at the site of capture later the same day. A subset of captured individuals was collared to attach radio-transmitters with temperature sensors (Advanced Telemetry Systems, Isanti, MN, USA, ARC400, approx. 14 g) that recorded skin temperature (T sk ) hourly. During hibernation, when individuals are curled up in a ball, transmitters are tightly attached to the neck area providing a reliable estimation of body temperature [31]. Data loggers (Maxim DS1922 iButton, Maxim Integrated Products, Inc., San Jose, CA, USA) were placed to record temperature of hibernacula (T h ) as well as T am . Sleeping sites/hibernacula locations were recorded using telemetry (Receiver R410, Advance Telemetry System, Isanti, MN, USA) on a weekly basis between March and July 2012, and March and July 2013.

Settings for hibernation studies
Radio-collared dwarf lemurs were retrieved from their natural hibernacula in July 2012 and 2013 and transferred to experimental boxes to conduct our study. These 'boxes' consisted either of a clear plastic container, approximately 30 × 20 × 10 cm, or a 20 cm length of PVC pipe of approximately 15 cm diameter, which could be sealed at each end. Both of these chambers were equipped with tubes at either end to allow for the air within to be pumped through an oxygen analyzer, as well as inflow of fresh air. Experimental boxes, with study animals connected to EEG electrodes (see below), were returned to hibernation sites and buried to approximate natural conditions of temperature and light levels. The animals were carefully monitored to ensure adequate airflow through the experimental chambers.

Measurement of metabolic rate
Metabolic rate data were also obtained as in our prior work [26]. The rate of metabolism was measured as VO 2 (rate of oxygen consumption) using a portable oxygen analyzer (OxBox, designed and constructed by T. Ruf & T. Paumann, FIWI, University of Veterinary Medicine Vienna) with chemo-electric sensors (Bieler & Lang, Achern, Germany; accuracy less than 0.02 vol%). Oxygen sensors were calibrated directly before the field season in the laboratory using calibration gas made by a gas-mixing pump (Wölsthoff, Bochum, Germany, type G27). The experimental boxes were connected with a gas-tight tube to the oxygen analyzer. Air flow was measured at 50 or 70 l h −1 , depending on the size of the box. The flow rate was continuously monitored by the mass flow meter of the gas analyzer system. Sampled air was dried and filtered with silica gel before entering the gas analyzer. Oxygen content of the sample air was measured once per minute. To control for any drift of the oxygen sensor, reference air (surrounding air) was analysed once per hour for 5 min (zero check). Metabolic rate was calculated as millilitres O 2 per hour. We also calculated mass-specific VO 2 expressed as millilitres O 2 per gram per hour. Torpor bouts could be identified by unambiguously depressed metabolic rate. Operationally, we define torpor when oxygen measurements were under 100 ml h −1 and arousals when oxygen measurements were above 100 ml h −1 , in agreement with values reported previously [26].

Electroencephalographic recording methods
EEG methods were the same as used in our prior work [28]. As an initial step, the animals were removed from their underground hibernacula for insertion of needle EEG electrodes. In one of the animals studied ('Ly'), this was done solely with the aid of a topical anaesthetic, which was possible due to the persistently torpid state of the animal. Otherwise, to minimize distress, animals were anaesthetized using ketamine 10 mg kg −1 (Ketaject, 100 mg ml −1 Bioniche Teoranta, Inverin, Co. Galway, Ireland) and midazolam 0.25 mg kg −1 (midazolam 5 mg ml −1 , Bedford Laboratories, Bedford, OH, USA). We recorded EEG data by placing four subdermal needle electrodes just under the skin of the scalp, and electrocardiogram (EKG) data (only in 2013 individuals) by placing a fifth subdermal needle in the interscapular area (Grass-Telefactor, West Warwick, RI, USA). Because animals have muscle overlying the cranium, electrodes inserted through the skin may have been embedded in the tissue beneath at times. Each electrode was inserted into each of the four quadrants of the cranium: right, left, rostral and caudal. In order to keep the leads in the original location, they were anchored with the use of an adhesive semipermeable membrane (Tegaderm, 3M Health Care, St. Paul, MN, USA). We then wrapped this membrane with flexible cohesive bandage (Webtear, Webster Veterinary) and bandage tape (Wet-Pruf ® The Kendall Co., Mansfield, MA, USA). We left the leads in place and recorded data as long as we were able to record technically acceptable data. The EEG recording device used was a Philips-Respironics DGx with capacity to record four EEG/electromyographic (EMG) channels, two EMG channels and one EKG channel. The subdermal needle electrodes were connected directly to the recording devices. Compumedics hardware and software (Compumedics Inc., Victoria, Australia) were used to record and review the data and to convert the data into European data format (EDF). All EEG records were scored in 30 s epochs by ADK who is Board Certified in Sleep Medicine (American Board of Sleep Medicine, American Board of Psychiatry and Neurology) and EEG (American Board of Clinical Neurophysiology, American Board of Psychiatry and Neurology), as previously described [28]. During scoring procedures, ADK chose the best channels for sleep stage identification and removed all data segments which included movement artefacts. However, we note that unlike the standard scoring criteria, REM sleep was operationalized as 30 s epochs where there was at least one rapid lateral eye movement and there was evidence of diminished EMG activity [28,32]. Because of the absence of electro-oculographic data, lateral eye movements were identified on the basis of their unique topography (a reversal of phase is seen in EEG leads on the left and right sides of the head which share a common reference) and morphology [28]. The artefacted data then underwent spectral analysis using standard methods implemented with software written by ADK which has been validated against MATLAB (MathWorks, Natick, MA, USA) software [33,34].

Sample size
Nine dwarf lemurs (C. crossleyi four females; C. sibreei two females and three males) wearing radiotransmitters were retrieved from their underground hibernacula for this study. Of these nine, seven experienced a period of sustained torpor followed by a spontaneous arousal during the period of evaluation. Owing to the fact that EEG intelligibility varied across individuals, recordings from four dwarf lemurs (three C. crossleyi and one C. sibreei), deemed too fragmentary, were not included in our analysis. As a result, this report is focused on the data collected from five animals for which EEGs could be analysed during torpor bouts as well as during arousal periods (table 1). For each individual, a single arousal episode was measured. For individuals 'Ra', 'Sy' and 'Am' (C. sibreei), data from pre-arousal and post-arousal torpor bouts were available. For individuals 'Ly' (C. crossleyi) and 'Ju' (C. sibreei), data from pre-arousal torpor bouts were only available.

Results
During torpor bouts, dwarf lemurs entered a state marked by a monotonous, ultra-low voltage EEG  (table 4) and following natural transition into torpor were available for three individuals (table 5). Non-REM was the first sleep stage occurring after the arousals began, but there was no evidence of a dramatic clustering that would potentially indicate substantial pent-up sleep need, such as may be expected following 3-5 days of total sleep deprivation. Figure 3 illustrates the continuous transition from relatively high Delta Power to the state of monotonous ultra-low voltage EEG, which occurs during torpor. During the entry into torpor, the EEG gradually transitioned from a pattern of non-REM sleep until achieving a sustained ultra-low voltage state. Because the transition is very gradual, the end of sleep and the beginning of torpor is rather arbitrary. It ranges from up to 5000 µV 2 during the periods of arousal to remaining consistently under 10 µV 2 throughout the periods of torpor. Metabolic rates during the first evidence of sleep is much higher than at the end, and non-REM sleep precedes REM at the onset of the arousal and follows REM-like states at the end (tables 4

Discussion
This study provides the first analysis of EEG data during hibernation, including bouts of torpor and the subsequent spontaneous arousals in a primate hibernator. It also provides the first data on EEG during hibernation in high-altitude dwarf lemurs. The findings suggest that sleep does not occur during torpor bouts in these two lemur species, which are exposed to relatively cold temperatures, in contrast with the western species, C. medius

Comparison with Cheirogaleus medius
Although we observed different findings from what we previously reported with C. medius, we believe that these differences are probably a reflection of the differing T h in the animals studied rather than species differences. Cheirogaleus medius may hibernate in well-insulated tree hollows where temperature remains relatively constant around 23°C and spontaneous arousals have then been reported to occur after roughly 7 days of hibernation bouts [12]. However, we were only able to obtain EEG recordings from C. medius in poorly insulated hibernacula where the temperature fluctuates to a substantially greater degree and reaches much higher temperatures, leaving spontaneous arousals expendable. Unlike C. crossleyi and C. sibreei investigated in this study, the C. medius we studied did not maintain a sustained ultra-low voltage EEG during hibernation; instead, they manifested this monotonous ultra-low voltage pattern only at low ambient temperatures. However, when T am (and consequently T b ) rose to relatively high levels, this pattern would give way to a REM sleep-like state. On this basis, we hypothesize that there may be a link between sleeping and spontaneous arousals which may be driven by T h . If T h is relatively high, at least several days in a week, spontaneous arousals are not needed and sleep occurs, whereas if T h is continuously low, there is a need for the arousals and sleep does not occur during hibernation bouts.

Comparison with non-primate hibernators
The findings of our study suggest that many features of hibernation in C. crossleyi and C. sibreei mirror what is seen in non-primate hibernators such as the Arctic, golden-mantled, and European ground squirrels, including non-REM and REM sleep observed exclusively during thermogenically induced arousals, but not during hibernation bouts [17,20,21]. Studies in the ground squirrels also note a monotonous ultra-low voltage EEG occurring during hibernation bouts and that sleep does not occur unless the T am is raised well above that normally experienced during hibernation [16,24]. In ground squirrels, higher T am during hibernation was noted to be associated with non-REM sleep, but also with lessened EEG Delta Power, which has been postulated as evidence of sleep deprivation during arousals [16,17,35]. Unlike dwarf lemurs (at least the western C. medius), which can continue to hibernate at high T am , squirrels aroused when temperatures were artificially raised up to approximately 25°C [16]. As squirrels are not naturally exposed to these T am during winter, the ability of lemurs to hibernate despite hot T am may indicate inherent biological differences.

Future avenues
A missing link in the study of primate hibernation is the comparison of EEG in dwarf lemurs hibernating under stable but warm conditions with those hibernating in colder and variable environments. Recently, it was shown that tenrecs, the other endemic mammalian group known to hibernate in Madagascar, can do so underground for up to nine months without undergoing endogenously generated arousals [9]. Sandy soils, where tenrecs hibernated, never cooled down under 22°C or heated up above 30°C. Dwarf lemurs hibernating in well-insulated tree hollows (relatively stable T h between 20 and 25°C) in western Madagascar do undergo arousals once a week, although the arousal bouts are shorter (approx. 6 h) than those of Arctic hibernators or high-altitude dwarf lemurs ( [12]; this study). Thus, there may be speciesspecific temperature thresholds that trigger the energetically expensive arousals, which presumably allow critical metabolic processes to resume. These differences, if they exist, may be the result of unrelated evolutionary histories. Although tenrecs may have retained the ability to use torpor from their ancestors [9], cheirogaleids may have diverged from non-heterothermic lemurs and evolved hibernation more recently, after their diversification in Madagascar. If this is the case, investigating hibernation in lemurs, our closest relatives known to hibernate, may be particularly relevant to research concerned with induction of hibernation-like states in humans.