Assessment of Physical Exercise Benefits on Brain Health for Long-Duration Spaceflight Sleep and Health Final Report

Authors: Vladyslav V. Vyazovskiy1, Simon P. Fisher1 , Jessica Gemignani2, Leopold Summerer2

Affiliation: 1University of Oxford, Department of Physiology, Anatomy and Genetics, Sherrington Building, Parks Road, Oxford, OX1 3PT, UK, 2University, 2ESA ACT

Date: 10 April 2017

Contacts:

Vladyslav V. Vyazovskiy Tel: +44 01865 285396 Fax: +44 01865 272420 e-mail: [email protected]

Leopold Summerer (Technical Officer) Tel: +31(0)715654192 Fax: +31(0)715658018 e-mail: [email protected]

Ariadna ID: 15/6301 Ariadna study type: Standard Contract Number: 4000114025/15/NL/LF/as Available on the ACT website http://www.esa.int/act

Table of Content

Summary ...... 3 Introduction ...... 5 Objectives ...... 11 Specific Experiments ...... 12 Methods ...... 12 Results ...... 19 Experiment 1 ...... 19 Experiment 2 ...... 24 Experiment 3 ...... 27 Experiment 4 ...... 28 Discussion ...... 29 Conclusions and future perspectives ...... 32 References ...... 37

Summary Poor sleep on the ISS remains one of the major factors, which affect negatively the performance and psychological resilience of astronauts. Several studies have demonstrated that sleep is severely disrupted during space flights, where factors such as microgravity, workload, noise, light and other environmental influences play a role. Therefore, the development of novel effective countermeasures for sleep loss, performance decrements and psychological deficits associated with space missions is needed. An intriguing possibility, which has not been tested previously, is that the benefits of exercise for the brain can complement cognitive and psychological benefits provided by sleep. Exercise has numerous benefits for health and is used widely as a prevention or treatment strategy for a broad range of diseases. The prevailing view is that regular exercise enhances overall physical fitness and primarily targets muscular strength and the cardiovascular system. However, evidence suggests that exercise also has numerous benefits for mental and psychological health. The proposed mechanisms underlying the positive influence of physical exercise on brain health are likely manifold and the role of improved cerebral oxygen supply, increased neurotransmitter release, neuroproliferation and synaptic plasticity have been studied in this context extensively. Surprisingly, while long-term effects of exercise have been investigated in great detail, few studies have addressed the activity of the brain directly during exercise, and the immediate effects of exercise on brain functioning and sleep remain largely unknown. Recently we obtained evidence that brain activity during wheel running in the mouse shows similarities with certain activity patterns present during sleep. Although traditionally sleep is considered a state markedly different from active behaviours, it appears to share with exercise some of its beneficial effects for brain function, including cognitive enhancement, improved mood and memory. In this study we set out to investigate brain activity during voluntary exercise in mice. In the first experiment, we performed detailed analysis of electroencephalogram (EEG) and neuronal activity in two cortical regions of freely behaving mice during wheel running behaviour. Unexpectedly, we found an overall suppression of cortical firing rates during intense running at a constant speed, which raises an interesting possibility that stereotypic waking behaviours may represent an economical ‘default’ mode of brain functioning. Consistently, we found that running wheel activity correlated with wake

duration, which led to the hypothesis that the build-up of ‘sleep need’ occurs at a slower rate during running, thus enabling longer wakefulness. Next we investigated the relationship between sleep deprivation (SD) and wheel running. The animals were kept awake for 6 hours, when they had a choice to run or explore novel objects. Interestingly, the animals would often prefer to run, and overall running intensity correlated with decreased intrusions of sleep during the 6-h period of SD. Furthermore, the proportion of waking during subsequent sleep opportunity interval was reduced as a function of running intensity during the SD. We interpret this result as a counteraction of sleep pressure provided by stereotypic running. Interestingly, this study also revealed a positive association between the amount of running during waking and faster EEG frequencies during subsequent sleep. Intrusion of faster frequencies into sleep states typically indicates more superficial sleep state, which supports our hypothesis that stereotypic running may reduce physiological sleep need. Furthermore, this result is consistent with our observation that individual neurons, which are activated during running may be also re-activated during sleep spindles: a type of cortical oscillation, which occurs at a faster (typically 10-17 Hz) frequency, and has been linked to learning and cognitive functions. Next we hypothesized that running wheel availability directly affects subsequent sleep, and to address this hypothesis, we performed an experiment where 3-h SD was followed by 3-h waking period either with free access to the running wheel or with novel objects to explore. Interestingly, when the animals were left undisturbed, those animals, which had an opportunity to run, slept significantly less during the following 3-h interval, as compared to animals, which instead explored novel objects. In a subset of animals, we also performed a detailed EEG analysis, and found that running wheel activity was associated with increased faster EEG delta activity (approximately 3-4 Hz) in the occipital derivation, which suggests that wheel running has an influence on local network activity. Finally, we used complex wheels, to investigate whether stereotypic locomotion has differential effects as compared to learning a novel motor skill. We found distinct effects of complex wheel running on wake EEG, manifested in faster and stronger theta (8-10 Hz) EEG power during complex wheel running. Furthermore, 5 out of 8 animals showed shorter sleep latency after a period of running on a complex wheel, and overall sleep was increased by almost one hour. While this result suggests strong interindividual variability in

the effects of voluntary exercise on sleep, it supports our hypothesis that voluntary stereotypic running is associated with decreased sleep need, as compared with waking dominated by demanding cognitive activities. Altogether, our data suggest that stereotypic waking behaviours may either counteract the build-up of sleep pressure, or even provide some of restorative functions, typically provided by sleep. While further studies are undoubtedly necessary, this new perspective opens novel opportunities for renormalisation of brain function in conditions and settings where it is impossible to achieve long consolidated sleep periods, such as throughout long-duration space missions, where confined environment, noise, irregular light-dark exposure, microgravity and intense workload prevent sleep. Understanding the neurophysiological mechanisms underlying the overlap in brain activity patterns between exercise and sleep will allow the development of specific training schedules, which have the potential to improve cognitive functioning and provide psychological benefits to astronauts during long-duration space flights.

Introduction Physical exercise is known to have numerous health benefits and has been shown to be effective in preventing or treating cardiovascular disorders, diabetes and cancer. Recently however, the focus is shifting more towards investigating the benefits of physical exercise on the human brain. There is evidence for beneficial effects of physical exercise in neurodegenerative disorders, such as Alzheimer and Parkinson’s disease (Goodwin et al., 2008; Rolland et al., 2008) and it has been used successfully as a non-pharmacological therapy in mood disorders, such as depression (Strohle, 2009). Even in healthy individuals, exercise can affect cognitive functions substantially. For example, it has been shown that non-strenuous activities, like walking, can boost creativity by about 60% (Oppezzo and Schwartz, 2014). Exercise also has potent psychological effects, which may be associated with changes in brain plasticity and morphology. Specifically, increased grey matter volume in the frontal, temporal, and parietal cortices has been reported in physically fit people (Erickson et al., 2014; Gomez-Pinilla and Hillman, 2013). These parts of the brain are important for decision-making, creativity, problem-solving, executive functions, emotional regulation, memory, confidence, learning languages and processing sensory (e.g. auditory) information. Important insights have been provided from studies in animal

models. For example, it has been shown in mice that running leads to increased neurogenesis in the hippocampus (van Praag et al., 1999) – the brain area crucially implicated in memory and emotions (Buzsáki, 2006). Another recent study found that a muscle secretory factor, cathepsin B (CTSB) protein, is important for the cognitive and neurogenic benefits of running (Moon et al., 2016). Moreover, it has been reported in mice that during locomotion the signal-to-noise ratio in visual stimulus detection increases (Bennett et al., 2013; Niell and Stryker, 2010). The mechanisms underlying the benefits of exercise for the brain are unknown but may be related to specific changes in brain activity. Specifically, it is likely that the mode of brain activity during exercise is fundamentally different from activity patterns during other behaviours, because many types of exercise lack complexity and consist of repetitive stereotypic or rhythmic movements, which are performed automatically without direct volitional control and may involve central pattern generators (Marder and Bucher, 2001). Surprisingly, while long-term effects of exercise have been investigated in great detail, few

Waking NREM sleep REM sleep

EEG

MUA

Spikes

EMG

1 sec Figure 1. EEG and neuronal activity in three basic vigilance states. From top to bottom: surface electroencephalogram (EEG) traces recorded from the somatosensory cortex during waking, NREM sleep and REM sleep. Multiunit activity (MUA) recorded simultaneously from a microwire array. Note high frequency tonic firing in waking and REM sleep, and the periods of population neuronal activity and silence in NREM sleep. Raster plots of spike activity for the same 6 channels (each vertical line is a spike). Note the close temporal relationship between the silent periods and the negative phase of EEG slow waves. Bottom: Corresponding electromyogram (EMG).

studies have addressed the activity of the brain directly during exercise, and the immediate effects of exercise on brain functioning remain largely unknown. Notably, another behaviour that benefits mental health is sleep. A fundamental difference between wakefulness and sleep is the extent to which the brain is engaged in the acquisition and processing of information. In all species carefully studied so far, waking and sleep alternate on a regular basis and continuous wakefulness rarely lasts spontaneously for more than several hours or a few days (Tobler, 2005). This suggests that sleep is necessary and serves a vital role for both the brain and the body as a whole. The maintenance of waking and sleep states is regulated by the activity arising from several subcortical structures in the brainstem, hypothalamus and basal forebrain, which provide a neuromodulatory, such as monoaminergic, orexinergic, glutamatergic, GABAergic and cholinergic, action on the neocortex (Brown et al., 2012; Fort et al., 2009). Importantly, the same neuromodulatory systems are crucially involved in locomotion and other active behaviours (Constantinople and Bruno, 2011; Mileykovskiy et al., 2005; Polack et al., 2013; Wu et al., 1999). During sleep the electroencephalogram (EEG) is dominated by oscillations occurring in a slow (0.5-4 Hz) frequency range, which are associated with an overall decline of cortical neuronal activity (Figure 1). The decrease of spiking and synaptic activity has been implicated in restorative functions of sleep (Vyazovskiy and Harris, 2013). It is obvious that both exercise and sleep are important for a healthy mind and are thus vital for astronauts that have to withstand huge stresses during manned space missions, like in the International Space Station. At the moment physical exercise is only obligatory to counter the effects of osteoporosis during weightlessness in space. However, using exercise in space may also lead to marked improvement of cognitive functions and psychological state, which are crucial during long-term flights. Among the potentially important factors affecting brain health negatively during space mission is a lack of sleep. Several studies have demonstrated that sleep is severely disrupted during space flights, where factors such as microgravity, workload, noise, light and other environmental influences play a role (Barger et al., 2014; Mallis and DeRoshia, 2005). Notably, sleep disturbances are also typical during preparation for space flights on Earth, as was observed during a 520-day high-fidelity ground simulation of a Mars mission where a decline in sleep quality or vigilance deficits were found in the majority of astronauts (Basner et al.,

2013). “Strategic napping” has been recommended as an effective strategy for maintaining performance in commercial aviation pilots, although the negative effects of so-called sleep inertia represents an important confound (Hartzler, 2014). Therefore, the development of novel effective countermeasures for sleep loss, performance decrements and psychological deficits associated with space missions is needed. Important potential confounds for the interpretation of the effects of exercise on sleep and brain function in general, are physical fatigue, which can be incurred during exercise, and the factor of motivation. These aspects are important since the compensatory rebound of sleep or sleep intensity after sleep deprivation may depend on specific types of behaviour, cognitive activities or merely locomotor activity, which is often elevated when the animals or humans are sleep-deprived. The role of sleep in recovery has been studied extensively in relation to exercise, both in animals and humans (Horne, 2013). Early studies suggested that sleep facilitates recovery of the fatigue acquired during exercise. In one of the first experiments, cats that were subjected to treadmill running were found to have compensatory sleep changes which manifested as a decrease in sleep latency and increase in synchronised EEG activity (Hobson, 1968). These data were supported by the observation that rats showed an increased daily sleep time after sleep deprivation by forced treadmill activity (Levitt, 1966). The hypothesis that sleep is associated with physical fatigue was also tested in a study where several different regimens of exercise were studied in two human subjects (Shapiro et al., 1975). This study found a whole-night increase in SWS which was related to the amount of physical fatigue, while the amount of REM sleep was reduced. These results have since been confirmed in more recent studies (Shapiro et al., 1981), while another study found a consistent increase in SWA and stage 2 sleep, accompanied by a reduction in REM sleep in fit subjects across four conditions varying by the amount of exercise (running) (Torsvall et al., 1984). However, a follow up study from a different laboratory, in which eight subjects underwent either a.m. or p.m. exercise identified contradicting results (Horne and Porter, 1976). Specifically, while p.m. exercise resulted in an increase in stage 3 during subsequent sleep, a.m. training did not lead to any changes in sleep, which led the authors to conclude that recovery from muscle fatigue does not require sleep, and can instead be fulfilled during waking.

In rats, sleep deprivation by forced locomotion on a “water wheel” resulted in an increased amplitude of the EEG recordings and intrusion of microsleeps during waking, while the subsequent recovery sleep showed an increased level of high amplitude slow waves during NREM sleep (Friedman et al., 1979). It is important to mention that it is unclear whether locomotor activity or waking duration per se account for this effect. In a control group, which had been awake for the same duration but walked much less, the rebound of sleep was similar, leading the authors to suggest that exercise per se has little influence on sleep. This notion was supported by a study showing that the effects of 12-h sleep deprivation by forced locomotion in rats, was similar irrespective of wheel turning speed (Borbely and Neuhaus, 1979). It should be noted, however, that forced exercise may have different effects from voluntary exercise (Hanagasioglu and Borbely, 1982). In cats, it was found that more physical exercise was associated with a larger increase of SWS during recovery sleep (Susic and Kovacevic-Ristanovic, 1980). It cannot be excluded that differences may be due to alternative factors such as the cognitive and motor effort needed to keep up with the rotation of the treadmill or even attempts to counteract the treadmill and accompanying stress associated with this. A similar conclusion has been reached in a human study where six men marched for 34 km each day with 5 hours rest allowed between the end of exercise and the subsequent sleep opportunity onset (Buguet et al., 1980). The effects on subsequent sleep were variable between individuals however seemed to correlate with the levels of urinary 17-hydroxycorticosteroids. In another study, cycling for 3 h also did not affect subsequent sleep (Youngstedt et al., 1999). An interaction between exercise and sleep deprivation was recently documented in a study where the exercise-related increase in GH was elevated in sleep-deprived individuals (Ritsche et al., 2014). A significant limitation is that obvious technical difficulties prevent EEG recordings in humans during running per se. It was suggested that exercise is closely linked to cognition and does not entail merely physical effort, and this may underlie beneficial effects of exercise on cognition in ageing (Horne, 2013). Interestingly, in overweight adults, exercise led to a reduction in default mode network (DMN) activity in the precuneus (McFadden et al., 2013). Another recent study found that functional connectivity in the DMN was increased following 6-12 months of training, which led the authors to suggest

that it reflects experience-dependent plasticity, as it was associated with an improvement in executive function (Voss et al., 2010). Notably, exercise appears to affect specific aspects of the compensatory sleep response. Specifically, in a rat study it was found that both EEG wave incidence and amplitude are responsive to prior wakefulness, but only incidence appears to be responsive to prior exercise (Mistlberger et al., 1987). This result is interesting given the different changes in network activity associated with specific parameters of EEG slow waves. Increased slow wave incidence may suggest higher excitability or bistability of cortical networks, rather than increased network synchrony, which could be expected to lead to higher slow-wave amplitude as a result of longer and larger population silence (Vyazovskiy, 2013; Vyazovskiy et al., 2009). Rodent studies have shown that access to running wheels can substantially affect the distribution, amount and architecture of vigilance states and activity across 24 h (Edgar et al., 1991; Gu et al., 2015), as well as EEG SWA (Vyazovskiy et al., 2006). Notably, frontal predominance of SWA was enhanced after a period of waking without running wheel activity leading the authors to hypothesize that more diverse waking behaviours and activity led to a higher need for local or global “recovery” (Vyazovskiy et al., 2006). An intriguing observation was that when access to the wheel was not prevented, the duration of the first spontaneous waking period after dark onset was substantially extended compared to the wheel block condition (Vyazovskiy et al., 2006). While this may suggest that in the absence of engaging activity the mice withdraw from active behaviours and therefore fall asleep earlier, it may also suggest that exploratory or diverse behaviours lead to a faster accumulation of sleep need or that voluntary exercise can in part be a substitute for sleep, by providing “rest” to those brain areas which are not involved in stereotypical running behaviour. In summary, it is not surprising that variable and discrepant results have been obtained in studies addressing the effects of exercise on sleep, or assessing the “recovery” role of sleep after physical fatigue. It is likely that a simple relationship cannot be expected, as the effects of exercise may be strongly determined by the kind of exercise, the type of movement/locomotion in general, and the level of motivation. Recent studies suggest that running wheel activity resembles some forms of naturalistic behaviours (Meijer and Robbers, 2014), and may have an intrinsic

reinforcement value (Belke and Pierce, 2014; Belke and Wagner, 2005). In rats, this was demonstrated using conditioned place preference, where prolonged access to the wheel was associated with plasticity in the mesolimbic reward circuitry, including the nucleus accumbens and the ventral tegmental area (Greenwood et al., 2011). Furthermore, it has been reported that wheel running interacts with the rewarding properties or effects of drugs of abuse (Lacy et al., 2014; Zlebnik et al., 2014), suggesting an overlap in their underlying neurophysiological mechanisms. Consistent with this, it has recently been shown that mice have a preference towards wheel running when given a choice between engaging in this behaviour and eating a highly palatable food, and this effect was related to dopaminergic transmission (Correa et al., 2015).

Objectives The overarching aim of this project is to begin addressing the hypothesis that the benefits of exercise for the brain can complement or even partially replace cognitive and psychological benefits provided by sleep. Specifically, we posited that specific types of exercise could compensate for the lack of consolidated sleep during long-duration space missions enabling the preservation of optimal brain functioning. To address this hypothesis, as a first step it is necessary to characterise in detail the patterns of brain activity during waking with and without exercise, and to test whether exercise affects subsequent sleep. Finding similarities in brain activity between exercise and sleep, and understanding their underlying neurophysiological substrate will allow the development of new methodologies for substantially improving the brain function of astronauts. This study aims to investigate the fundamental neurophysiological basis of the beneficial effects of physical exercise on the brain. The first objective is to investigate differences in the patterns of brain activity between periods of restful waking, exercise and sleep in laboratory animals. The second objective is to identify and investigate similarities between the patterns of brain activity during exercise and sleep. The third objective is to develop specific research-based exercise regimens, which would maximize their beneficial effects on specific brain functions. The ultimate goal of this research is to inform future human studies and provide recommendations for astronauts with respect to specific types, duration and intensity of exercise to improve their overall health, cognitive functioning and psychological well-being.

Specific Experiments In this project the following experiments have been conducted:

1. In the first experiment, we performed detailed analysis of electroencephalogram (EEG) and neuronal activity in two cortical regions of freely behaving mice during wheel running behaviour, during waking without running and during sleep. We performed EEG spectral analyses, and analysed the amount and pattern of extracellularly recorded neuronal activity in the primary motor and primary somatosensory cortex.

2. In the second experiment, we performed sleep deprivation (SD), where the animals were kept awake for 6 hours, when they had a choice to run or explore novel objects. We investigated the relationship between wheel running behaviour and several measures of sleep and sleep EEG.

3. In the third experiments we asked whether running wheel availability directly affects subsequent sleep, and to address this hypothesis, we performed an experiment where 3-h SD was followed by 3-h waking period either with free access to the running wheel or with providing the animals with novel objects to explore. We then investigated subsequent sleep and sleep EEG.

4. In the final experimental series we used complex wheels, to investigate whether stereotypic locomotion has differential effects from learning a novel motor skill. This experiment aimed at investigating whether the pattern of activity during running has an influence on brain activity and subsequent sleep.

Methods Experimental animals. Adult male mice, C57BL/6J strain, were used in this study (age range between 5 months and 1 year). All mice were individually housed in custom-made clear plexiglass cages (20.3 x 32 x 35cm) with free access to a running wheel (RW, see below, Figure 2). Cages were housed in ventilated, sound-attenuated Faraday chambers (Campden Instruments, Loughborough, UK, two cages per chamber) under a standard 12:12 h light-dark cycle (lights on 0900, ZT0, light levels ~120-180 lux). Food and water were available ad libitum. Room temperature and relative humidity were maintained at 22

± 1°C and 50 ± 20%, respectively. Mice were habituated to both the cage and recording cables for a minimum of four days prior to recording. All procedures conformed to the Animal (Scientific Procedures) Act 1986 and were performed under a UK Home Office Project Licence in accordance with institutional guidelines.

Surgical procedures and electrode configuration. Surgical procedures were carried out using aseptic techniques under isoflurane anaesthesia (3-5% induction, 1-2% maintenance). During surgery animals were head fixed using a stereotaxic frame (David Kopf Instruments, CA, USA) and liquid gel (Viscotears, Alcon Laboratories Ltd, UK) was applied to protect the eyes. One day prior to surgery animals received dexamethasone (0.2

a b

c Frontal EEG/M1 array SCx array Reference Occipital EEG Ground

Figure 2. Experimental set up and electrode positions. (a) Animals were individually housed in custom-made home cages, providing continuous free access to a running wheel, inside ventilated, sound-attenuated chambers provided with autonomous light-dark (12:12) control. (b) A photograph of a sample 16-ch microwire array, which were used in this study for local field potentials (LFP) and multiunit activity (MUA) recordings in all animals, and a representative stainless screw used to record EEG signals and as a reference and ground. Both devices are shown alongside the mouse brain to illustrate relative size proportions. (c) Schematic depiction of the positions of the 16-ch microwire arrays used to record MUA and LFP in the primary motor cortex (M1) and somatosensory cortex (SCx), the frontal and the occipital EEG screws and the position of reference and ground screws placed above the cerebellum.

mg/kg, i.p) to suppress the local immunological response. Metacam (1-2 mg/kg, s.c., Boehringer Ingelheim Ltd, UK) and dexamethasone (0.2 mg/kg, s.c.) were administered preoperatively (and for at least three days after surgery). Post operatively animals were

administered saline (0.1ml/20g body weight, s.c.) and provided thermal support throughout and following surgery. A minimum two week recovery period was permitted prior to cabling the animals. For this study it was essential to maintain long-term stable recordings and unrestricted movement during spontaneous wheel running, so we avoided using large electrodes (e.g. with a high channel count or laminar probes) or multiple arrays and probes. All mice were implanted with a single polyimide -insulated tungsten microwire array (Tucker-Davis Technologies Inc (TDT), Alachua, FL, USA). The arrays consisted of 16- channels (2 rows each of 8 wires) with properties as follows: wire diameter 33µm, electrode spacing 250 µm, row separation L-R: 375µm, tip angle 45 degrees. We used customized arrays where the lateral row of the wires was longer by 250 µm. A 1x2 mm craniotomy was made using a high-speed drill (carbon burr drill bits, 0.7 mm, InterFocus Ltd, Cambridge, UK) in the region of interest, with the midpoint of the craniotomy relative to bregma as follows: M1: anteroposterior (AP) +1.5-2mm, mediolateral (ML) ~2 mm; SCx: AP -1mm, ML 3.25mm. The dura was dissected using a 25 gauge needle and saline was applied to the cranial opening to keep the exposed brain moist. In most cases, removal of the dura did not cause bleeding (when bleeding occurred, it was stopped with gelfoam soaked in sterile saline). The electrode array was advanced into the brain until the longer row of microwires was at the level of cortical layer 5 (M1: ~0.7-0.8 mm below the pial surface, SCx: ~ 0.5-0.6 mm). A two-component silicon gel (KwikSil; World Precision Instruments, FL, USA) was used to seal the craniotomy and protect the surface of the brain from dental acrylic. After 5-10 min, required for the gel to polymerise, dental acrylic was applied to fix the array to the skull. In all animals, EEG screws were placed in the frontal (motor area, AP +2mm, ML 2mm) and occipital (visual area, V1, AP -3.5-4mm, ML 2.5mm) cortical regions contralateral to the arrays using procedures previously described (Cui et al., 2014). A reference and ground screw electrodes were placed above the cerebellum and an additional anchor screw was placed behind or opposite to the array to ensure implant stability. EEG screws were soldered (prior to implantation) to custom-made headmounts (Pinnacle Technology Inc. Lawrence, USA) and all screws and wires were secured to the skull using dental acrylic. Two single stranded, stainless steel wires were inserted either side of the nuchal muscle to record electromyography (EMG). A schematic diagram of implantation locations is shown in Figure 2.

Signal processing and analysis. Data acquisition was performed using the Multichannel Neurophysiology Recording System (TDT, Alachua FL, USA). Extracellular neuronal spike data were collected continuously (25 kHz, 300 Hz - 5kHz) concomitantly with local field potentials (LFPs) from the same electrodes, and cortical EEG was recorded from frontal and occipital derivations. EEG/EMG data were filtered between 0.1-100 Hz, amplified (PZ5 NeuroDigitizer pre-amplifier, TDT Alachua FL, USA) and stored on a local computer at a sampling rate of 256.9 Hz. LFP/EEG/EMG data were resampled offline at a sampling rate of 256 Hz. Signal conversion was performed using custom-written Matlab (The MathWorks Inc, Natick, Massachusetts, USA) scripts and was then transformed into European Data Format (EDF) using open source Neurotraces software (www.neurotraces.com). For each recording, EEG and LFP power spectra were computed by a Fast Fourier Transform (FFT) routine for 4-s epochs (fast Fourier transform routine, Hanning window), with a 0.25 Hz resolution (SleepSign Kissei Comtec Co, Nagano, Japan).

Scoring and analysis of vigilance states. Vigilance states were scored offline through manual visual inspection of consecutive 4-s epochs (SleepSign, Kissei Comtec Co, Nagano, Japan). Two EEG channels (frontal and occipital), EMG and RW activity were displayed simultaneously to aid vigilance state scoring. Vigilance states were classified as waking (low voltage, high frequency EEG with a high level or phasic EMG activity), NREM sleep (presence of EEG slow waves, a signal of a high amplitude and low frequency) or REM sleep (low voltage, high frequency EEG with a low level of EMG activity). Great care was taken to eliminate epochs contaminated by eating, drinking or gross movements resulting in artifacts in at least one of the two EEG derivations or in any of the multiunit activity (MUA) channels. Detailed analyses of neuronal activity (see below) were based on selected time intervals to ensure stability of neuronal waveforms within the time window specified. Sleep deprivation. Sleep deprivation (SD) was performed during 6 hours starting either at light onset, or starting 3 h before dark onset. During SD, the animals were awake on average >95% of time, and their behavior and polysomnographic recordings were under constant visual observation. In the 2nd experiment, SD was performed in the animal’s home

cage, where they had free access to running wheels throughout the procedure, which they used intermittently. Throughout the SD procedure, the animals were also regularly provided with novel objects to mimic naturalistic conditions of wakefulness an in ethologically relevant manner (Palchykova et al., 2006; Tobler and Borbely, 1986; Vyazovskiy et al., 2007). In the 3rd experiments, the animals did not have access to the running wheel during the first 3 hours of SD, and during subsequent 3 hours, starting from dark onset, were either given access to the wheel or kept awake by providing them constantly with novel objects. All mice were well habituated to the experimenter and to the exposure to the novel objects prior to the experiment. Novel objects included nesting and bedding material from other cages, wooden blocks, small rubber balls, plastic, metallic, wooden, or paper boxes and tubes of different shape and color.

Running-wheel activity. In this study animals had access to running wheels (Campden Instruments, Loughborough, UK, wheel diameter 14 cm, bars spaced 1.11 cm apart inclusive of bars) for at least two weeks prior to the experiment, and were therefore well adapted to the wheels. The wheels were custom made for tethered animals, and did not prevent the animals from running ad libitum. Each running wheel was fitted with a digital counter (Campden Instruments) which uses an infra-red (IR) emitter/receiver to detect each rung passing the IR beam as the wheel rotates. In our study running wheel activity was recorded with a high temporal resolution (one full wheel revolution consists of 38 individually detected rung counts, thus 10 counts per second corresponds to 10.11 centimeters/sec) within the same system used to record electrophysiological signals. This allowed us to achieve a precise synchrony between instantaneous changes in electrophysiological signals/behaviour and RW activity/speed. The wheel counter output is a 5V TTL pulse (0V with no output), that triggers an edge detector in the TDT acquisition system, and in turn creates a time stamp that is stored for each wheel count. For most analyses, RW-activity was analysed in 1-sec epochs that were grouped based on the number of counts during the corresponding epochs. Since the number of 1-sec epochs with progressively higher numbers of counts decreased progressively, we also used progressively wider ranges for the bins (0; 1; 2-3; 4-7; 8-11; 12-32), to obtain a sufficiently high yet reliable number of counts within each speed category.

To obtain an acceleration-deceleration index (ADI) we calculated mean second derivative of the time series corresponding to the timings of consecutive counts for each 1- sec epoch, and subsequently grouped epochs according to ADI from largest negative (acceleration) to largest positive (deceleration) values into ten 10% deciles prior to calculating averages between animals. As a result, the first 10% decile corresponds to fastest acceleration, the tenth decile - to fastest deceleration, while the middle of the distribution (deciles 5-6) correspond mostly to steady running with a close to zero net speed change. In the 4th experiment the animals were recorded for an undisturbed 24-h while provided with regular running wheel. On the next day, the regular wheel was replaced with a complex wheel, where instead of 38 rungs, only 22 rungs were available (Figure 3). The animals were given an opportunity to run on the complex wheel during the 12-h dark period.

Regular Wheel Complex Wheel

Figure 3. A photograph of a regular wheel (38 rungs) and complex wheel (22 rungs)

Analysis of extracellular neuronal activity. Online spike sorting was performed to eliminate artifactual waveforms caused by electrical or mechanical noise. This was performed with OpenEx software (TDT), by manually applying an amplitude threshold for online spike detection. Whenever the recorded voltage exceeded this predefined threshold (at least -25 µV), a segment of 46 samples (0.48 ms before, 1.36 ms after the threshold crossing) was extracted and stored for later use, together with the corresponding time stamps. To investigate the relationship between running behaviour and individual putative single units’ firing, we employed offline spike sorting. Firstly, an artifacts removal procedure was implemented in order to eliminate remaining artifactual waveforms and to

facilitate subsequent clustering. Typical spurious spike waveforms included waveforms which were likely the result of a combination of several spikes from different neurons occurring concurrently and preventing their reliable classification. After removal of the artifactual spikes, principal components were computed using principal component analysis (PCA) on a segment of the spike waveform between the 5th and the 35th time sample, because this is the segment that contains the initial negative peak after threshold crossing and subsequent afterhyperpolarisation, and is therefore more informative about the overall spike waveform. In PCA an orthogonal linear transformation of the original data is performed, through singular value decomposition of the data matrix (Jolliffe, 2002). After performing PCA, each spike between samples 5 and 35 was thus described by 31 variables, each being a linear combination of the original sampling values. Each PC corresponds to an eigenvalue and an eigenvector from the singular value decomposition of the data matrix. The first PC is associated with the eigenvector on the direction of maximal variance of the dataset, and so on in decreasing order for the remaining PCs. The ratio between each eigenvalue and the sum of all eigenvalues represents the ratio of the total variance described by a single PC. Clustering was performed based on the k-means algorithm (Hartigan, 1975). This is a partitioning method that aims to divide n observations into k clusters in which each observation belongs to the cluster with the nearest mean, serving as a prototype of the cluster. There are several k-means algorithms available which are based on different ways to proceed through the iterations. We used the k-means function in Matlab, which was implemented according to Lloyd’s algorithm (Lloyd, 1982). In particular, it chooses k initial cluster centers (centroid) through an initialisation step, and then computes point-to-cluster-centroid Euclidean distances of all observations to each centroid. Based on the distance, each observation is assigned to the cluster with the closest centroid. As a next step, the algorithm computes the average of the observations in each cluster to obtain k new centroid locations. Finally, it repeats previous steps until cluster assignments do not change, or the maximum number of iterations is reached. This approach requires the user to a priori select the number of clusters. Based on visual inspection, it was found unlikely that more than 5 distinct waveforms were present in the same MUA channel, and so the procedure has been performed with k=2, 3, 4 and 5. For each case, several features were extracted within each cluster and were used to validate the quality of clustering and to

select the best outcome of the clustering procedure within each animal and recording channel. These were the average spike waveform with standard deviation, interspike interval distribution, the time course of peak-to-peak amplitude across the recording period, the corresponding time course of average firing rates and the autocorellogram of the spike trains. All selected clusters were screened and visually classified according to their signal to noise ratio, waveshape of the action potential, stability of the amplitude across time and interspike interval (ISI) distribution histogram, and spurious or unstable clusters were excluded from the analysis. We should emphasize that based on our recording and sorting technique it cannot be excluded that two or more neurons having identical spike waveform shapes, amplitude and firing properties are present simultaneously on the same channel.

Statistical analysis. All statistical analyses were performed in Matlab (The MathWorks Inc, Natick, Massachusetts, USA) and all values reported are mean ± SEM. ANOVAs for repeated measures were used to detect statistically significant effects of running wheel speed or acceleration on firing rates. Two-tailed paired t-tests were used for comparisons. In addition, we also employed a normalisation procedure by expressing some of the variables in relative terms (as % of the mean value across all artifact free epochs during the recording period). We have done this in those cases where interindividual variability of absolute values is not relevant for the effects observed, as it allowed us to assess the relative magnitude of the effects. Normalisation of values in this manner is a widely used approach and therefore facilitates comparison between studies, where relative values are reported.

Results Experiment 1 RUN on and RUN off neurons in the neocortex during spontaneous wheel running. In the first experiment, multiunit activity (MUA) in the motor cortex (M1) and EEG were recorded in mice which had free access to a running wheel in their home cage (Figure 4). In every animal, several prolonged waking periods dominated by running wheel (RW) activity were observed during the dark period, interspersed by naps of various durations. As typically observed (Vyazovskiy et al., 2006), during waking epochs with wheel running (wRUN), EEG was dominated by theta-activity (6-9 Hz), which was substantially lower

during waking epochs when the animals did not run (nwRUN). We also observed that during NREM sleep, the cortical LFP recorded from M1 was dominated by high-amplitude positive slow waves (~0.5-4 Hz), associated with a brief reduction of MUA across the majority or even all of the electrodes within the 16-channel array (Figure 4).

400 a c LFP 500 µV 300 b 1 mm 200

SWA (% of mean) SWA 100

0 30 20 10 MUA 0 RW countsRW / sec W N R 2 hours 1 sec p<0.01 e 200f g 160 d RUN off neuron 40 RUN off RUN on 150 140 30 120 100 100 µV RUN on neuron 20 0.5 ms 100

Firing Rates (%) 50 10 Number of Neurons of Number 80 Firing mean) rates (% of 0 0 60 RW-activity 0 50 100 150 200 nwRUN wRUN NREM 0 1 2-3 4-7 8-11 12-32 5 sec Firing Rates (wRUN / nwRUN) RW counts / sec

Figure 4. Cortical neuronal activity in the primary motor cortex (M1) during voluntary wheel running in freely behaving mice. (a) A photograph of the custom-made cage providing continuous free access to a running wheel positioned on the rear. Scale bar: 5 cm. (b) From top to bottom: 12-h profile of EEG slow-wave activity (SWA, EEG power between 0.5-4.0 Hz, represented as % of 12-h mean) recorded in the frontal cortex, running-wheel (RW) activity (counts/sec) and the distribution of sleep-wake stages (W= wakefulness, N= NREM sleep, R= REM sleep) from a representative mouse. (c) Schematic representation of the position of the primary motor cortex (M1, blue area drawn with reference to Paxinos & Franklin, 2001) shown on the dorsal surface of the mouse head, and the position of the 16-ch microwire array above M1 (dots indicate the position of individual wires within the array). Traces on the right: top, representative local field potentials (LFP) recorded during NREM sleep from one row of microwires; bottom, raster plot of multiunit activity (MUA, each vertical line represents a spike) recorded from the same wires. Note the close temporal relationship between positive LFP waves and periods of generalised neuronal silence. (d) Two MUA traces recorded from M1 in the same animal with corresponding RW-activity (bottom, each vertical bar represents a single wheel rung count). Corresponding waveforms of the action potentials recorded extracellularly are shown on the right. (e) The distribution of all putative single units recorded in n=11 mice as a function of the ratio of their average firing rates (FR) during running (wRUN) and non-running waking (nwRUN). Note that a smaller proportion of neurons increase firing during wheel running (red, RUN on neurons), while the majority decrease FR during running (blue, RUN off neurons). (f) Average FR in M1 during nwRUN waking, wRUN waking and NREM sleep. Mean values, SEM, n=11 (individual mice: grey symbols). Significant differences between vigilance states are depicted above the bars (obtained by two-tailed paired t-test following significant one-way ANOVA). (g) FR in M1 shown as function of running speed. Thick lines: mean values, SEM, n=9 mice. Values from individual animals are shown as thin line plots.

In contrast, during waking we observed a striking dissociation between the activities of putative single units depending on the animal’s behaviour. Specifically, in

a b MUA

100 µV

1 mm RW-activity

1 sec

100 µV

1 sec c d 160 25 140 20 120 15

10 100

Number of Neurons of Number 5 80 Firing Rates (% of mean) (% of Rates Firing 0 60 0 50 100 150 200 0 1 2-3 4-7 8-11 12-32 Firing Rates (wRUN / nwRUN) RW counts / sec

Figure 5. Cortical neuronal activity in the somatosensory cortex (SCx) during voluntary wheel running. (a) Schematic representation of the position of the 16-ch microwire array shown relative to the barrel cortex (drawn in reference to Paxinos & Franklin, 2001) shown on the dorsal surface of the mouse head (dots indicate the position of individual wires within the array). (b) Individual representative examples depicting multiunit activity (MUA) in SCx during wheel running. Corresponding running wheel (RW)-activity is shown below each trace (each vertical bar represents a single wheel rung count). (c) The distribution of all putative single units recorded in SCx (in n=5 mice) as a function of the ratio of their average firing rates (FR) during running (wRUN) and non-running waking (nwRUN). Note that a smaller proportion of neurons increase FR during running (red), while the majority decrease spiking activity (blue). (d) FR in SCx shown as a function of running speed (count/sec). Thick lines: mean values, SEM, n=5 mice. Values from individual animals are shown as thin line plots. every mouse we observed clear cut cases when MUA slowed down substantially or ceased altogether during intense running on the wheel (RUN off neurons, Figure 4). Overall, cortical neuronal activity was on average 25.0± 9.3% lower during wRUN compared to nwRUN waking (repeated measures ANOVA, factor ‘vigilance/behavioural state’, F(2,18)=17,23, p=0.0001). Most neurons did not merely change firing when the animals

ran, but they often showed a distinct inverse relationship to running velocity. This resulted in a significant negative relationship between the firing rates in M1 and running speed (Figure 4, F(5,53)=8,61, p=1.3315e-005, ANOVA for repeated measures, factor ‘running speed’). To investigate whether spontaneous wheel running affects firing rates in other cortical areas, we also performed MUA recordings from the somatosensory cortex (SCx) (Figure 5). Although visual inspection did not reveal individual neurons in SCx which would stop firing altogether during running as observed in M1, the majority of neurons also changed their firing activity during running. On average, firing rates in the SCx also showed a decrease with increasing running speed (Figure 5, F(5,29)=19,42, p=4.6534e- 007, ANOVA for repeated measures, factor ‘running speed’).

a a’ M1 SCx MUA

50µV

RW-activity

1 sec 1 sec b Speed Speed Speed b’ Speed Speed Speed 140 140

120 120

100 100

80 80

Firing Rates (% of mean) (% of Rates Firing 0 0 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Acceleration-Deceleration Index (ADI)

Figure 6. The relationship between cortical neuronal activity and changes in wheel running speed. (a, a’) Individual examples depicting cortical firing rates (FR) in the motor cortex (M1, blue) and in the somatosensory cortex (Scx, pink) modulated by running wheel behaviour. From top to bottom: MUA recorded in one representative channel of the 16- channel array, and running-wheel (RW)-activity (each vertical bar represents a single wheel rung count). Note the irregular pattern of spiking associated with variability in running speed. (b, b’) Cortical FR in M1 (blue) and SCx (pink) shown as a function of the wheel acceleration-deceleration index. All 1-sec epochs were subdivided into ten 10% deciles, each consisting of the same number of epochs. These were sorted as a function of the change in wheel running speed within an epoch from fast acceleration to fast deceleration (schematically shown as arrows above), and corresponding average FR were calculated prior to averaging between animals. Mean values, SEM, M1: n=9 mice, SCx: n=5. Mean values are shown as bar plots, the individual values from individual animals are shown as thin line plots.

The decrease in cortical firing rates occurs predominantly during stereotypic running. Notably, spontaneous wheel running is not stable and uniform throughout running bouts. Instead we frequently observed brief pauses in running, and fluctuations in running speed were common, irrespective of the running bout length. On the other hand, moment-to- moment changes in firing rates were frequently observed both within wRUN and nwRUN waking. Most putative single units had a “preferred” frequency of firing, which was affected substantially by wheel running behaviour. However, we noticed that the width of the distribution of firing rates was substantially reduced when the mice engaged in running, especially at a high speed, an effect which was present among the majority of individual putative single units. This suggests that during running a more uniform or stereotyped cortical state is instated, which EEG 200 µV contrasted significantly with the LFPs 500 µV wider dynamic range of firing during nwRUN waking (M1: F(4,44)=93.86, p<0.0001, SCx: F(4,24)=278.88, MUA p<0.0001, ANOVA for repeated measures, factor ‘RW-activity’). RW-activity We observed striking changes in the 1 sec firing activity Figure 7. Cortical activity during voluntary wheel running. Representative examples of ‘sleep-like’ OFF periods (arrow) in the depending on the somatosensory cortex (SCx) occurring during high speed running. Top: EEG recorded in the occipital derivation. Note the presence of pattern and nature of strong EEG theta-activity, typical for running. Traces below running behaviour, represent local field potentials (LFP) recorded concomitantly in all sixteen channels of the 16-channel microwire array. Corresponding such as during initial multiunit activity (MUA) is shown below as a raster (each bar is an individual spike). Bottom: corresponding running-wheel (RW)- acceleration, during activity (each bar represents a single wheel rung count). Note that a steady constant positive LFP wave is accompanied with reduced MUA across most of the 16 channels.

running and during brief of wheel running and firing rates. To address this notion, we calculated wheel acceleration / deceleration index (ADI) based upon the timing of individual RW counts during 1-sec epochs. For most epochs the average ADI was close to 0 representing negligible net speed changes within an epoch as typical for stereotypic steady speed running, although abrupt fluctuations in running speed, corresponding to positive or negative ADI values during a consolidated running bout were also. Firing rates were strongly related to the ADI in both M1 and SCx (M1: F(9,89)=6.31, p=1.5427e-006, SCx: F(9,49)=13,9, p=2.5696e-009). Specifically, firing rates were on average 20-30% lower during epochs with steady running, as compared to epochs with rapid acceleration or deceleration (Figure 6). Calculating the relationship between firing rates and ADI separately for RUN on and RUN off neurons revealed overall lower firing rates during steady running irrespective of whether the neuron increased or decreased its firing rate relative to nwRUN waking in both regions (not shown). In other words, even if a specific neuron fired more strongly during wRUN waking as compared to nwRUN waking, it tended to fire less frequently as running became more stereotypic. It is well established that state-dependent changes in cortical activity are accounted for by a variety of mechanisms, including intrinsic properties of individual neurons, local and global neuromodulation, active inhibition and disfacilitation, as well as connectivity within the network (Bacci et al., 2005; Buzsaki et al., 2007; Connors and Gutnick, 1990; Lemieux et al., 2014; Okun et al., 2015; Rudolph et al., 2007; Timofeev et al., 2001; Zagha and McCormick, 2014). Visual inspection of raw traces revealed the occurrence of isolated brief (~100-200 ms) periods of neuronal silence accompanying positive LFP ‘slow’ waves, even during intense wheel running. Such ‘OFF-periods’ were encountered both in M1 and SCx, and usually encompassed a subset of recording channels, but in some cases were visible across the entire 16-channel array (Figure 7). Such LFP/MUA OFF periods during running were often characterised by a conspicuous brief surge of intense spiking activity immediately preceding and/or following the period of prolonged silence.

Experiment 2 Running wheel activity during sleep deprivation. Since mice use wheels spontaneously during the night, the question remains whether the relationship between cortical activity and running behaviour is similar at different times of day, under different lighting conditions, and when they are given a choice of other behaviours. To address this, fist we

analysed the data collected during the light period, when the animals were kept awake by continuously providing them with novel objects in their home cages (Figure 8). As expected, the animals spent time exploring, but they would also often run on the wheel, running on average >200 meters during a 6 h sleep deprivation protocol. Running was distributed evenly across the 6-h period of sleep deprivation (F(5,53)=1.24, p=0.31, ANOVA for repeated measures, factor ‘1-h interval’), therefore suggesting that increased sleep pressure did not prevent animals from running, even at a time of day when they typically do not run. Notably, in all individual animals we again observed clear cut cases of putative single units decreasing spiking substantially during running. In all animals, firing rates also decreased with increasing running speed during sleep deprivation, similar to spontaneous running during the dark phase (Figure 8). To investigate whether access to running wheels affects sleep propensity, we next quantified the occurrence of brief sleep episodes during the sleep deprivation period. As it is typically observed, the animals tended to fall asleep briefly, especially towards the end of the 6-h SD. The intrusion of sleep into awake state is considered a measure of sleep pressure. Notably, the amount and intensity of wheel running during the 6-h interval showed a negative relationship with the occurrence of sleep, which reached a statistical significance (p<0.05) with the total number of running wheel counts, and a statistical tendency (p<0.1) with the proportion of time spent running (Figure 8). Next, conversely, we calculated the proportion between waking and sleep during subsequent sleep opportunity. Typically after SD sleep pressure is high and the animals spend most of the time asleep. However, if sleep pressure is lower, it is expected that the animals will spend more time awake during this period. We found that running wheel activity during SD correlated positively during the amount of subsequent waking (Figure 8). In other words, the more intense was running during SD, the longer time animals spent awake after they were left undisturbed, suggesting that sleep pressure was reduced.

As well known, physiological sleep pressure is reflected in the predominance of slow EEG frequencies (<=4Hz) in the EEG signal. On the other hand, faster frequencies in sleep EEG are considered a measure of cortical activation, and reflects less deep sleep. We calculated a correlation between running wheel activity during 6-h SD and EEG spectra in NREM sleep during subsequent sleep opportunity. Intriguingly, we found mostly positive

a b 600 120

400 110

200 100 SWA (% of mean) SWA

0 90 W N 80 F(4,44)=19.1 R p<0.001 Firing Rates (% of mean) (% of Rates Firing RW-activity 70 0 1 2-3 4-7 8-32 0 1 2 3 4 5 6 7 RW counts / sec c 0.16 Hours 0.14 d 1.0 0.14 0.12 0.8 0.12 0.10 0.10 0.6 0.08 0.08 0.4 0.06 0.06 r-value 0.2 0.04 r=-0.51 / Sleep Wake r=0.7 0.02 0.04 p=0.016 p=0.03 0.0 0.00 0.02 -0.2 frontal EEG 0 200 400 600 800 1000 0 200 400 600 800 1000 0 5 10 15 20 RW revolutions / 6 h of SD 1.0 0.16 0.14 0.14 0.8

Sleep Sleep during SD (hours) 0.12 0.12 0.6 0.10 0.10 0.08 0.08 0.4 0.06 0.06 r-value 0.2 0.04 Wake / Sleep Wake r=-0.62 r=-0.65 0.04 0.0 0.02 p=0.06 p=0.07 occipital EEG 0.00 0.02 -0.2 0.0 0.4 0.8 1.2 1.6 0.0 0.4 0.8 1.2 1.6 0 5 10 15 20 RW time (hours) Frequency (Hz) Figure 8. Cortical neuronal activity in M1 during wheel running during sleep deprivation. (a) Individual profile of EEG slow-wave activity (SWA, EEG power between 0.5-4.0 Hz) recorded in the frontal cortex, running-wheel (RW) activity and the distribution of sleep-wake stages (W= wakefulness, N = NREM sleep, R= REM sleep) across 6-h of sleep deprivation (SD) and subsequent 1 hour of sleep in one representative mouse. (b) Average cortical firing rates in M1 shown as a function of running speed (x-axis). Thick lines: mean values, SEM, n=9 mice. Values from individual animals are shown as thin line plots. (c) The relationship between the total amount of running and time spent running during SD and the amount of sleep during SD (left) or Wake/Sleep ratio during the 1st hour of recovery. (d) The relationship between RW-activity and EEG power spectra during the first hour of sleep. Triangles above the curves denote frequency bins where Pearson’s correlation was significant (p<0.05). relationship between faster EEG frequencies in the frontal EEG derivation and the amount of running prior to sleep (Figure 8). This suggested a higher level of ‘arousal’ during sleep

after running, which may reflect a lower level of sleep pressure. Notably, this effect was ‘local’, as the relationship was less pronounced in the occipital derivation, where the positive correlation was mostly present in theta-frequency range (Figure 8). Altogether this suggests that wheel running during SD reduced sleep pressure.

Experiment 3 Running wheel activity after sleep deprivation. Next, we performed 6-h SD, starting 3 hours before dark onset. All animals did not have access to running wheels during the first 3 hours of SD, but have been provided with novel objects continuously. Subsequently, the animals were either allowed to run ad libitum on the wheel, or were kept awake by providing them with novel a objects to explore for the next 3 hours. Subsequently in all animals the wheel was blocked and the animals were left undisturbed (Figure 9). We hypothesised that if running discharges sleep need, sleep time will be reduced during

b NREM sleep REM sleep subsequent sleep opportunity. 60 10 noRUN Consistent with this hypothesis, 50 8 RUN 40 we found that the overall 6 30 amount of sleep was reduced in 4 20 p<0.1 2 animals, which had free access % of recording time 10 p<0.05 0 0 to the wheel during the second 0 1 2 3 0 1 2 3 Time [Hours] 3-h interval of SD (Figure 9).

Figure 9. Effects of wheel running on subsequent sleep. (a) Schematic In a subset of animals, diagram of the wheel-block experiment. (b) Time course of NREM and REM sleep during the first 3 hours after the animals were left we also performed detailed EEG undisturbed starting from Hour 15. Note that the amount of sleep is decreased following 6-h waking period when the animal had access to analyses. Interestingly, sleep the wheel. after running was characterised by a prominent increase of fast delta frequencies (3-4 Hz), which suggested that waking experience has an influence on cortical network activity during subsequent sleep (not shown). This effect was mostly pronounced in the occipital cortex, which is located closer

to the hippocampus. It remains to be determined whether the changes in the EEG we observe arise from the cortico-hippocampal interactions, or merely represent a volume conducted signals originating from limbic areas.

Experiment 4 Complex wheel running and subsequent sleep. As well known, hippocampus is strongly implicated in movement and learning motor skills. On the other hand, evidence suggests that motor skill learning, or more generally demanding cognitive activities during waking necessitate sleep. Therefore, we hypothesized that the effects of running on sleep would differ if the animals run voluntarily and stereotypically on a familiar regular wheel, as

nwRUN wRUN a b 140 Frontal 140 Occipital longer waking n=3 120 120

shorter waking 100 100 n=5 (% regular wheel) (% regular EEG power density EEG power

80 80

0 5 10 15 20 0 5 10 15 20 Frequency (Hz)

Figure 10. The effects of running on a complex wheel on wake EEG and subsequent sleep. (a) EEG power density during non-wheel running waking (nwRUN) and wheel running waking (wRUN) during the night with complex wheels expressed as % of corresponding EEG power density during preceding control night, when mice had access to regular running wheels. Note an increase of fast theta activity during the night with complex wheels. Mean values, n=8. (b) Distribution of individual animals as a function of prolongation (n=3) or shortening (n=5) of the main waking period during the night with the complex wheel relative to the previous night with a regular running wheel. compared to a complex wheel, when the animals have to learn a novel motor skill. To address this hypothesis, a group of animals (n=8), well habituated to run on a regular wheel was given a wheel where instead of 38 rungs, only 22 rungs were available (Figure 3). This resulted in irregularly placed gaps between rungs, which required animals to learn to cope with. Interestingly, all animals persisted running, which resulted in approximately the same overall amount of running during the dark period. The wake EEG was affected substantially by learning to run on a complex wheel. Specifically, EEG theta activity was significantly higher during waking with complex wheels, as compared to regular wheels

(Figure 10), which likely reflects the process of learning or higher levels of arousal. Interestingly, we noticed on several occasions that individual animals stopped running and fell asleep substantially earlier when they were provided with complex wheels, as compared to previous night with a regular wheel. Specifically, n=5 out of n=8 (62%) showed longer sleep on the night with the complex wheel, and in total the animals slept almost 1 hour longer (54.4 min) on the night with the complex wheel as compared to the night with a familiar regular wheel. This suggests high interindividual variability in the effects of waking exercise on subsequent sleep, but also indicates that stereotypic locomotion is associated with a reduced sleep need as compared to waking dominated by cognitively or physically demanding activities.

Discussion In this study we investigated changes in the firing rates of cortical neurons recorded from motor (M1) and somatosensory (SCx) cortex in freely moving mice during voluntary unrestricted running-wheel (RW) activity. Importantly, we found that wheel running modulates a substantial proportion of cortical neurons, surprisingly leading to an overall reduction in neuronal activity specifically during high speed and/or stereotypic running. The mechanisms underlying the reduction in neuronal firing in M1 and SCx during high speed and/or stereotypic running may be diverse. As is well known, movement and behavioral state transitions are associated with widespread changes in the activity among several neuromodulatory systems (Aston-Jones and Bloom, 1981; Day et al., 1991; Jacobs and Fornal, 1997; Jones, 2005; Takahashi et al., 2010). Noradrenaline is essential for maintaining a tonic depolarization of pyramidal neurons in the motor, somatosensory and the visual cortex (Constantinople and Bruno, 2011; Polack et al., 2013; Schiemann et al., 2015), and has been implicated in wake-to-sleep transitions (Carter et al., 2010). On the other hand, while wake-promoting effects of acetylcholine are well established (Xu et al., 2015), in vitro studies suggest diverse layer-specific effects of acetylcholine, manifested in hyperpolarisation of layer 4 neurons and excitation of layer 2/3 and layer 5 pyramidal neurons (Eggermann and Feldmeyer, 2009). In addition to “global” neuromodulatory influences, extrinsic to the neocortex, local inhibition is also likely essential. Several studies have highlighted diversity among local GABA-ergic neuronal populations which are recruited in awake mice in a unique manner with respect to the ongoing behaviour,

movement or whisking (Gentet et al., 2010; Kvitsiani et al., 2013; Reimer et al., 2014). Interestingly, pharmacogenetic inhibition of a subpopulation of vasoactive intestinal peptide positive (VIP) GABA-ergic neurons in the visual cortex resulted in reduced cortical activity irrespective of behavioural state (Jackson et al., 2016), while somatostatin- expressing (SOM) GABAergic neurons of layer 2/3 of the barrel cortex were active during spontaneous quiet wakefulness, but reduced activity and hyperpolarised in response to whisking (Gentet et al., 2012). Moreover, again in the visual cortex VIP interneurons were activated by locomotion, and in turn inhibited SOM interneurons, thus disinhibiting excitatory cells (Stryker, 2014). In turn, stimulation of excitatory neurons may lead to a strong activation of GABA-ergic neurons, which precipitates a net inhibitory effect in the local microcircuit (Mateo et al., 2011). The observation of slow LFP waves during running, which were associated with a surge of neuronal spiking and subsequent MUA suppression, may reflect an arrival of extrinsic excitatory inputs recruiting local cortical inhibitory networks (Logothetis et al., 2010; Vyazovskiy et al., 2013). A strong determinant of the overall change in network activity was the pattern of running, such as whether it was stereotypic or not. Intriguingly, a decrease in firing rates during high speed running was observed not only among RUN off neurons, but even among RUN on neurons, which on average fired at a higher rate when the animals ran, yet they were relatively quiescent as the animal engaged in a stable speed running. Notably, in our study the animals were habituated to spontaneous wheel running in their home cages and had access to the wheels for many days prior to the experimental night. It remains to be established whether shifts between high speed stereotypic running and goal directed behaviours in freely behaving animals relate to changes in the levels of arousal. Reduced firing rates in M1 and SCx when animals engaged in high speed running may reflect increased arousal levels, while more stable and uniform firing activity may correspond to an optimal state for sensory processing (Buran et al., 2014; McGinley et al., 2015; Vinck et al., 2015). Our data suggest a distinct role of specific cortical circuits in voluntary locomotion which is dependent on the precise nature or form of the behaviour, for example stereotypic locomotion as compared to purposeful or exploratory behaviour. It is possible that the effects would have been different, if the recordings had been obtained while the animals were learning how to run, as in this case the running would likely be less stereotypic, and active continuous involvement of the motor cortex would be

necessary (Isomura et al., 2009; Kawai et al., 2015). Consistent with this notion, when the animals where provided with complex wheels, which required learning a novel skill, the majority of animals fell asleep earlier, and the overall sleep was increased by on average almost 1 hour. Thus, while our data seems to indicate that stereotypic behaviours may translate into cortical states which are also more homogenous and lack complexity, the possibility remains that more stereotyped waking behaviours also influence the dynamics of the accumulation of sleep need. Strikingly, in our study we observed that mice readily run during sleep deprivation (SD), which is surprising for at least two reasons. First, this procedure was performed during the light period, when mice typically do not run. Second, during SD the animals were constantly provided with novel objects, but rather than spending time exploring, the animals often chose to engage in stereotypical running. While the functional significance of this behaviour remains to be determined, our data suggest that even when wheel running occurs outside of the usual time and setting, it has similar effects on cortical activity as during spontaneous wheel running during the dark phase. It has been previously shown that wheel availability substantially affects the amount and distribution of vigilance states, such that animals would fall asleep later and have less sleep altogether when they are given free access to the wheel (Vyazovskiy et al., 2006; Vyazovskiy and Tobler, 2012). We confirm these observations in our current study. Furthermore, we now show using several complementary approaches that sleep need may indeed be reduced by wheel running. First, we observed that the amount of intrusions of sleep into awake state during sleep deprivation was negatively correlated with the amount of running wheel activity. Second, we observed that the more animals ran during sleep deprivation, the more time they spent awake during subsequent sleep opportunity. Furthermore, sleep EEG after SD was characterised by a higher predominance of faster frequencies if the animals showed more running during preceding waking. Finally, if SD during the dark period was combined with running wheel activity, the amount of sleep in the following 3-h interval was overall lower. Overall, our data suggest that stereotypic waking behaviours, such as wheel running may be associated with a “default” awake state, which in turn permits animals to stay awake longer and/or decrease sleep need. Arguably, this possibility would confer a significant advantage, as it would not only enable animals to remain effectively “online”

and therefore avoid dangers associated with a sensory disconnection during sleep, but it may in some cases be necessary for specific behaviours such as during migration or a mating season when prolongation of wakefulness is favoured ecologically (Lesku et al., 2012; Siegel, 2008). We posit that an important feature of this default awake state is that it may represent waking at a “lower cost”. Such a low cost default wake state is uniquely suited to be effectively placed in the continuum between sleep and purposeful, goal- directed behaviours, and may render substantial adaptive flexibility with respect to changing environmental conditions, time of day and homeostatic needs.

Conclusions and future perspectives Overall, the results presented here provide initial evidence in support of our hypothesis that waking dominated by simple, stereotypic behaviours may be associated with reduced sleep need. Furthermore, our results suggest an intriguing possibility that such “default” wakefulness may even contribute actively to the restorative sleep functions. Our work has relevance for space travel, inasmuch as it suggests a potentially game-changing strategy to mitigate cognitive deficits arising as a result of insufficient or disrupted sleep in space. As well known, disrupted or insufficient sleep is a well-known health risk factor in astronauts during spaceflight and pre-flight training (Pandi-Perumal and Gonfalone, 2016). Light exposure and sleep pills as well as psychostimulants are used extensively to counteract poor sleep and to improve waking performance (Barger et al., 2014; Basner et al., 2013; Brainard et al., 2016). However, since various cognitive, psychological and motor functions are impaired in space (Clement and Ngo-Anh, 2013; De la Torre, 2014; Pagel and Chouker, 2016), the development of novel effective countermeasures for sleep loss, performance decrements and psychological deficits associated with space missions is necessary. One strategy is to invest efforts in the development of interventions or regimens aimed at increasing the duration or quality of sleep. While this research direction is certainly promising, it is important to bear in mind that we do not know yet which aspects of sleep are crucial in this respect and should be targeted in the first place. For example, sleep duration may be less important than sleep intensity, or the same amount of sleep may confer greater benefits if taken during a specific phase of the endogenous circadian rhythm (Borbely et al., 2016). Moreover, it may not be practically achievable to obtain the average

8 hours of consolidated sleep on a daily basis while on ISS, and the negative influence of such factors as microgravity, noise or light on sleep quality may be difficult to eliminate. We proposed an alternative strategy, which consists in targeting wakefulness in the first instance. The central idea of our proposal is that it might be possible to develop schedules of specific waking activities, which cumulatively decrease the need for sleep. According to the traditional view, the need for sleep builds up progressively during wakefulness and needs to be discharged during subsequent sleep (Borbely et al., 2016). This sleep-wake dependent process is called Process S, and mathematical modelling studies have allowed to formally describe its dynamics and to obtain better understanding of the effects of sleep restriction on the amount and distribution of sleep as well as waking performance (McCauley et al., 2009; Van Dongen et al., 2011). Notably, waking is typically assumed to be a homogenous process, which is associated with a continuous increase of sleep pressure largely irrespective of ongoing behaviour or activities. We propose instead, that certain types of wakefulness have a lower ‘cost’ with respect to accumulation of sleep need, and specific activities may, on the contrary, even reduce sleep pressure and improve cognitive performance. This possibility is supported by numerous studies demonstrating positive effects of fitness on cognitive function, with largest exercise-induced benefits reported for executive- control processes, such as such as working memory or planning (Colcombe and Kramer, 2003). Importantly, it was shown that the magnitude of beneficial effects of exercise on cognitive functions was related to a specific type of exercise and the duration of training sessions (Colcombe and Kramer, 2003). Moreover, exercise has been also proposed to hold high promise to counteract or prevent mood changes (Schneider et al., 2010; Strohle, 2009), and there is evidence for beneficial effects of fitness in neurodegenerative disorders, such as Alzheimer and Parkinson’s disease (Goodwin et al., 2008; Rolland et al., 2008) . Notably, even non-strenuous activities, like walking, can boost creativity by about 60% (Oppezzo and Schwartz, 2014). This suggests that it is not physical activity per se, but possibly some other aspects of behaviour, which are relevant for the beneficial effects of exercise. One possibility is that exercise is associated with an occurrence of a unique brain state, which contributes to the beneficial effects of fitness on cognition. Several studies have shown that exercise is associated with enhancement of slow EEG frequencies, such as

in alpha and theta frequency ranges, specifically in the frontal cortex (reviewed in (Dietrich, 2006)). Notably, circadian time and preceding sleep-wake history have also been shown to modulate EEG activities in these frequency bands, both in humans (Cajochen et al., 2002; Finelli et al., 2000; Landolt et al., 2004) and in laboratory animals (Leemburg et al., 2010; Vyazovskiy and Tobler, 2005). Interestingly, the occurrence of slow frequencies in wake EEG is considered a marker of intrusion of sleep-like patterns of activity into awake state (Hung et al., 2013; Vyazovskiy et al., 2011). This indicates that brain systems, which are relevant for sleep-wake control, may be also sensitive to exercise. In other words, evidence suggests that exercise may elicit patterns of brain activity, which are typically encountered during sleep only. We hypothesized that the benefits of exercise, or more specifically, certain types of waking activities, for the brain can, at least in theory, complement cognitive and psychological benefits provided by sleep. It is obvious that both exercise and sleep are important for a healthy mind and are thus vital for astronauts that have to withstand huge stresses during manned space missions, like in the ISS. However, at the moment physical exercise is only obligatory to counter the effects of osteoporosis during weightlessness in space. Moreover, while hundreds of experiments performed on ISS had a neuroscience component, which included testing psychological functions, neurocognitive performance, circadian rhythms, spatial orientation, effect of microgravity upon kinematic and dynamic characteristics of locomotion, vestibular reflexes, sensory functions, and psychomotor vigilance (Clement and Ngo-Anh, 2013), it has never been studied whether specific types of exercise can improve astronauts’ cognitive performance without the need of obtaining longer sleep. One study found increased performance following running sessions on an active treadmill in a simulated space flight to Mars (Schneider et al., 2013). Notably, this study also revealed a decreased activity of the neocortex in the prefrontal area, which is known to be critical for executive brain functions, and also appears to be most vulnerable to sleep loss (Anderson and Horne, 2003; Couyoumdjian et al., 2010; Curcio et al., 2006; Drummond et al., 2000; Goel et al., 2009; Horne, 1993; Thomas et al., 2000). This observation is consistent with the hypothesis that improvements in cognitive performance after exercise is associated with “transient hypofrontality” (Dietrich, 2006). This theory postulates a redistribution of cortical activity from the frontal cortex areas towards areas of

the brain implicated in motor control (i.e. motor cortex and associated areas). While this idea is interesting, it remains unclear why hypofrontality should improve performance, and our results provide important novel insights. In this project we found that stereotypic wheel running is associated with reduced neuronal activity in at least two cortical areas, and the amount of sleep was reduced following an episode of running. Our results thus support our hypothesis and suggest that it is possible to develop a regimen of waking activities, which would decrease the rate of accumulation of sleep need. It is important to note that the subjects would remain effectively awake during this time, rather than engaging in a non-functional ‘hybrid’ or mixed state, which may be mal-functional. Many fundamental questions relating to the link between physical exercise and brain health remain to be addressed. The practical questions include the optimal content of exercise which is most beneficial for the brain (type, duration, and intensity), and specific regimens, which on the one hand should not interfere with the daily routine, but on the other hand provide maximal effects. Next experiments should include human studies, where EEG is recorded continuously during waking dominated by demanding cognitive activities and stereotypic, but engaging tasks. We hypothesize that such experiment will reveal less local signatures of ‘tiredness’ when subjects perform repetitive stereotypic tasks. Ideally, such experiments need to be combined with performance measurements. An intriguing possibility is that performance may improve if a cognitively demanding task is combined with stereotypic movement or sensory stimulation, which we hypothesize could decrease ‘signal-to-noise’ ratio in brain activity, thus allowing to re-allocate cognitive resources to the task. Further animal experiments could in turn provide further insights about specific molecular and cellular mechanisms underlying the effects we observed. It is essential to test other behaviours, which are less physically demanding than wheel running. Furthermore, the factors of arousal and motivation need to be addressed. Next, it remains to be determined whether stereotypic repetitive behaviours benefit brain function in general and performance in cognitive tasks in particular. For example, the question remains whether the animal or a human can learn a novel association while running stereotypically, or whether tasks, which require sleep for memory consolidation (Rasch and Born, 2013) can also benefit from exercise (Rhee et al., 2016; Weinberg et al., 2014). Finally, molecular

markers of cellular function and synaptic plasticity need to be investigated after a period of stereotypic behaviour. It is well known that sleep and sleep deprivation are associated with a marked change in gene expression (Cirelli et al., 2004). However, the possibility that different kinds of waking lead to distinct changes in the transcriptome remain unexplored. Redressing this omission will allow identifying targets for investigating the underlying mechanisms and direct pharmacogenetic manipulations in future studies.

Notes Parts of the introduction were modified from (Vyazovskiy, 2015), and results of the 1st experiments have been reproduced with modifications from Fisher, Cui et al., Nature Communications, in press (DOI: 10.1038/NCOMMS13138).

Acknowledgements In addition to ESA support the study was funded by: MRC NIRG (MR/L003635/1), BBSRC Industrial CASE grant (BB/K011847/1), FP7-PEOPLE-CIG (PCIG11-GA-2012- 322050), John Fell OUP Research Fund Grant (131/032), Wellcome Trust Strategic Award (098461/Z/12/Z). We would like to thank Prof. P. Somogyi, Drs. L. Katona, T. Viney, M. Walton and T. Gheysens for discussions and advice on data analysis and interpretation, and Drs F. Nodal and Z. Molnar for assistance with histology.

References Anderson, C., and Horne, J.A. (2003). Prefrontal cortex: links between low frequency delta EEG in sleep and neuropsychological performance in healthy, older people. Psychophysiology 40, 349-357. Aston-Jones, G., and Bloom, F.E. (1981). Activity of norepinephrine-containing locus coeruleus neurons in behaving rats anticipates fluctuations in the sleep-waking cycle. J Neurosci 1, 876-886. Bacci, A., Huguenard, J.R., and Prince, D.A. (2005). Modulation of neocortical interneurons: extrinsic influences and exercises in self-control. Trends Neurosci 28, 602- 610. Barger, L.K., Flynn-Evans, E.E., Kubey, A., Walsh, L., Ronda, J.M., Wang, W., Wright, K.P., Jr., and Czeisler, C.A. (2014). Prevalence of sleep deficiency and use of hypnotic drugs in astronauts before, during, and after spaceflight: an observational study. Lancet Neurol 13, 904-912. Basner, M., Dinges, D.F., Mollicone, D., Ecker, A., Jones, C.W., Hyder, E.C., Di Antonio, A., Savelev, I., Kan, K., Goel, N., et al. (2013). Mars 520-d mission simulation reveals protracted crew hypokinesis and alterations of sleep duration and timing. Proc Natl Acad Sci U S A 110, 2635-2640. Belke, T.W., and Pierce, W.D. (2014). Effect of sucrose availability and pre-running on the intrinsic value of wheel running as an operant and a reinforcing consequence. Behav Processes 103, 35-42. Belke, T.W., and Wagner, J.P. (2005). The reinforcing property and the rewarding aftereffect of wheel running in rats: a combination of two paradigms. Behav Processes 68, 165-172. Bennett, C., Arroyo, S., and Hestrin, S. (2013). Subthreshold mechanisms underlying state- dependent modulation of visual responses. Neuron 80, 350-357. Borbely, A.A., Daan, S., Wirz-Justice, A., and Deboer, T. (2016). The two-process model of sleep regulation: a reappraisal. J Sleep Res. Borbely, A.A., and Neuhaus, H.U. (1979). Sleep deprivation: effects on sleep and EEG in the rat. Journal of Comparative Physiology 133, 71-87. Brainard, G.C., Barger, L.K., Soler, R.R., and Hanifin, J.P. (2016). The development of lighting countermeasures for sleep disruption and circadian misalignment during spaceflight. Curr Opin Pulm Med 22, 535-544. Brown, R.E., Basheer, R., McKenna, J.T., Strecker, R.E., and McCarley, R.W. (2012). Control of sleep and wakefulness. Physiol Rev 92, 1087-1187. Buguet, A., Roussel, B., Angus, R., Sabiston, B., and Radomski, M. (1980). Human sleep and adrenal individual reactions to exercise. Electroencephalogr Clin Neurophysiol 49, 515-523.

Buran, B.N., von Trapp, G., and Sanes, D.H. (2014). Behaviorally gated reduction of spontaneous discharge can improve detection thresholds in auditory cortex. J Neurosci 34, 4076-4081. Buzsáki, G. (2006). Rhythms of the brain (Oxford ; New York, Oxford University Press). Buzsaki, G., Kaila, K., and Raichle, M. (2007). Inhibition and brain work. Neuron 56, 771- 783. Cajochen, C., Wyatt, J.K., Czeisler, C.A., and Dijk, D.J. (2002). Separation of circadian and wake duration-dependent modulation of EEG activation during wakefulness. Neuroscience 114, 1047-1060. Carter, M.E., Yizhar, O., Chikahisa, S., Nguyen, H., Adamantidis, A., Nishino, S., Deisseroth, K., and de Lecea, L. (2010). Tuning arousal with optogenetic modulation of locus coeruleus neurons. Nat Neurosci 13, 1526-1533. Cirelli, C., Gutierrez, C.M., and Tononi, G. (2004). Extensive and divergent effects of sleep and wakefulness on brain gene expression. Neuron 41, 35-43. Clement, G., and Ngo-Anh, J.T. (2013). Space physiology II: adaptation of the central nervous system to space flight--past, current, and future studies. Eur J Appl Physiol 113, 1655-1672. Colcombe, S., and Kramer, A.F. (2003). Fitness effects on the cognitive function of older adults: a meta-analytic study. Psychol Sci 14, 125-130. Connors, B.W., and Gutnick, M.J. (1990). Intrinsic firing patterns of diverse neocortical neurons. Trends Neurosci 13, 99-104. Constantinople, C.M., and Bruno, R.M. (2011). Effects and mechanisms of wakefulness on local cortical networks. Neuron 69, 1061-1068. Correa, M., Pardo, M., Bayarri, P., Lopez-Cruz, L., San Miguel, N., Valverde, O., Ledent, C., and Salamone, J.D. (2015). Choosing voluntary exercise over sucrose consumption depends upon dopamine transmission: effects of haloperidol in wild type and adenosine AKO mice. Psychopharmacology (Berl). Couyoumdjian, A., Sdoia, S., Tempesta, D., Curcio, G., Rastellini, E., L, D.E.G., and Ferrara, M. (2010). The effects of sleep and sleep deprivation on task-switching performance. J Sleep Res 19, 64-70. Cui, N., McKillop, L.E., Fisher, S.P., Oliver, P.L., and Vyazovskiy, V.V. (2014). Long- term history and immediate preceding state affect EEG slow wave characteristics at NREM sleep onset in C57BL/6 mice. Arch Ital Biol 152, 156-168. Curcio, G., Ferrara, M., and De Gennaro, L. (2006). Sleep loss, learning capacity and academic performance. Sleep Med Rev 10, 323-337. Day, J., Damsma, G., and Fibiger, H.C. (1991). Cholinergic activity in the rat hippocampus, cortex and striatum correlates with locomotor activity: an in vivo microdialysis study. Pharmacol Biochem Behav 38, 723-729. De la Torre, G.G. (2014). Cognitive neuroscience in space. Life (Basel) 4, 281-294. Dietrich, A. (2006). Transient hypofrontality as a mechanism for the psychological effects of exercise. Psychiatry Res 145, 79-83.

Drummond, S.P., Brown, G.G., Gillin, J.C., Stricker, J.L., Wong, E.C., and Buxton, R.B. (2000). Altered brain response to verbal learning following sleep deprivation. Nature 403, 655-657. Edgar, D.M., Kilduff, T.S., Martin, C.E., and Dement, W.C. (1991). Influence of running wheel activity on free-running sleep/wake and drinking circadian rhythms in mice. Physiol Behav 50, 373-378. Eggermann, E., and Feldmeyer, D. (2009). Cholinergic filtering in the recurrent excitatory microcircuit of cortical layer 4. Proc Natl Acad Sci U S A 106, 11753-11758. Erickson, K.I., Leckie, R.L., and Weinstein, A.M. (2014). Physical activity, fitness, and gray matter volume. Neurobiol Aging 35 Suppl 2, S20-28. Finelli, L.A., Baumann, H., Borbely, A.A., and Achermann, P. (2000). Dual electroencephalogram markers of human sleep homeostasis: correlation between theta activity in waking and slow-wave activity in sleep. Neuroscience 101, 523-529. Fort, P., Bassetti, C.L., and Luppi, P.H. (2009). Alternating vigilance states: new insights regarding neuronal networks and mechanisms. Eur J Neurosci 29, 1741-1753. Friedman, L., Bergmann, B.M., and Rechtschaffen, A. (1979). Effects of sleep deprivation on sleepiness, sleep intensity, and subsequent sleep in the rat. Sleep 1, 369-391. Gentet, L.J., Avermann, M., Matyas, F., Staiger, J.F., and Petersen, C.C. (2010). Membrane potential dynamics of GABAergic neurons in the barrel cortex of behaving mice. Neuron 65, 422-435. Gentet, L.J., Kremer, Y., Taniguchi, H., Huang, Z.J., Staiger, J.F., and Petersen, C.C. (2012). Unique functional properties of somatostatin-expressing GABAergic neurons in mouse barrel cortex. Nat Neurosci 15, 607-612. Goel, N., Rao, H., Durmer, J.S., and Dinges, D.F. (2009). Neurocognitive consequences of sleep deprivation. Semin Neurol 29, 320-339. Gomez-Pinilla, F., and Hillman, C. (2013). The influence of exercise on cognitive abilities. Compr Physiol 3, 403-428. Goodwin, V.A., Richards, S.H., Taylor, R.S., Taylor, A.H., and Campbell, J.L. (2008). The effectiveness of exercise interventions for people with Parkinson's disease: a systematic review and meta-analysis. Mov Disord 23, 631-640. Greenwood, B.N., Foley, T.E., Le, T.V., Strong, P.V., Loughridge, A.B., Day, H.E., and Fleshner, M. (2011). Long-term voluntary wheel running is rewarding and produces plasticity in the mesolimbic reward pathway. Behav Brain Res 217, 354-362. Gu, C., Coomans, C.P., Hu, K., Scheer, F.A., Stanley, H.E., and Meijer, J.H. (2015). Lack of exercise leads to significant and reversible loss of scale invariance in both aged and young mice. Proc Natl Acad Sci U S A 112, 2320-2324. Hanagasioglu, M., and Borbely, A.A. (1982). Effect of voluntary locomotor activity on sleep in the rat. Behav Brain Res 4, 359-368. Hartigan, J.A. (1975). Clustering algorithms (New York,, Wiley). Hartzler, B.M. (2014). Fatigue on the flight deck: the consequences of sleep loss and the benefits of napping. Accid Anal Prev 62, 309-318.

Hobson, J.A. (1968). Sleep after exercise. Science 162, 1503-1505. Horne, J. (2013). Exercise benefits for the aging brain depend on the accompanying cognitive load: insights from sleep electroencephalogram. Sleep Med 14, 1208-1213. Horne, J.A. (1993). Human sleep, sleep loss and behaviour. Implications for the prefrontal cortex and psychiatric disorder. Br J Psychiatry 162, 413-419. Horne, J.A., and Porter, J.M. (1976). Time of day effects with standardized exercise upon subsequent sleep. Electroencephalogr Clin Neurophysiol 40, 178-184. Hung, C.S., Sarasso, S., Ferrarelli, F., Riedner, B., Ghilardi, M.F., Cirelli, C., and Tononi, G. (2013). Local experience-dependent changes in the wake EEG after prolonged wakefulness. Sleep 36, 59-72. Isomura, Y., Harukuni, R., Takekawa, T., Aizawa, H., and Fukai, T. (2009). Microcircuitry coordination of cortical motor information in self-initiation of voluntary movements. Nat Neurosci 12, 1586-1593. Jackson, J., Ayzenshtat, I., Karnani, M.M., and Yuste, R. (2016). VIP+ interneurons control neocortical activity across brain states. J Neurophysiol, jn 01124 02015. Jacobs, B.L., and Fornal, C.A. (1997). Serotonin and motor activity. Curr Opin Neurobiol 7, 820-825. Jolliffe, I.T. (2002). Principal component analysis, 2nd edn (New York, Springer). Jones, B.E. (2005). From waking to sleeping: neuronal and chemical substrates. Trends Pharmacol Sci 26, 578-586. Kawai, R., Markman, T., Poddar, R., Ko, R., Fantana, A.L., Dhawale, A.K., Kampff, A.R., and Olveczky, B.P. (2015). Motor cortex is required for learning but not for executing a motor skill. Neuron 86, 800-812. Kvitsiani, D., Ranade, S., Hangya, B., Taniguchi, H., Huang, J.Z., and Kepecs, A. (2013). Distinct behavioural and network correlates of two interneuron types in prefrontal cortex. Nature 498, 363-366. Lacy, R.T., Strickland, J.C., Brophy, M.K., Witte, M.A., and Smith, M.A. (2014). Exercise decreases speedball self-administration. Life sciences 114, 86-92. Landolt, H.P., Retey, J.V., Tonz, K., Gottselig, J.M., Khatami, R., Buckelmuller, I., and Achermann, P. (2004). Caffeine attenuates waking and sleep electroencephalographic markers of sleep homeostasis in humans. Neuropsychopharmacology 29, 1933-1939. Leemburg, S., Vyazovskiy, V.V., Olcese, U., Bassetti, C.L., Tononi, G., and Cirelli, C. (2010). Sleep homeostasis in the rat is preserved during chronic sleep restriction. Proc Natl Acad Sci U S A 107, 15939-15944. Lemieux, M., Chen, J.Y., Lonjers, P., Bazhenov, M., and Timofeev, I. (2014). The impact of cortical deafferentation on the neocortical slow oscillation. J Neurosci 34, 5689-5703. Lesku, J.A., Rattenborg, N.C., Valcu, M., Vyssotski, A.L., Kuhn, S., Kuemmeth, F., Heidrich, W., and Kempenaers, B. (2012). Adaptive Sleep Loss in Polygynous Pectoral Sandpipers. Science. Levitt, R.A. (1966). Sleep deprivation in the rat. Science 153, 85-87.

Lloyd, S.P. (1982). Least Squares Quantization in PCM. IEEE Transactions on Information Theory 28, 129-137. Logothetis, N.K., Augath, M., Murayama, Y., Rauch, A., Sultan, F., Goense, J., Oeltermann, A., and Merkle, H. (2010). The effects of electrical microstimulation on cortical signal propagation. Nat Neurosci 13, 1283-1291. Mallis, M.M., and DeRoshia, C.W. (2005). Circadian rhythms, sleep, and performance in space. Aviat Space Environ Med 76, B94-107. Marder, E., and Bucher, D. (2001). Central pattern generators and the control of rhythmic movements. Curr Biol 11, R986-996. Mateo, C., Avermann, M., Gentet, L.J., Zhang, F., Deisseroth, K., and Petersen, C.C. (2011). In vivo optogenetic stimulation of neocortical excitatory neurons drives brain-state- dependent inhibition. Curr Biol 21, 1593-1602. McCauley, P., Kalachev, L.V., Smith, A.D., Belenky, G., Dinges, D.F., and Van Dongen, H.P. (2009). A new mathematical model for the homeostatic effects of sleep loss on neurobehavioral performance. J Theor Biol 256, 227-239. McFadden, K.L., Cornier, M.A., Melanson, E.L., Bechtell, J.L., and Tregellas, J.R. (2013). Effects of exercise on resting-state default mode and salience network activity in overweight/obese adults. Neuroreport 24, 866-871. McGinley, M.J., David, S.V., and McCormick, D.A. (2015). Cortical Membrane Potential Signature of Optimal States for Sensory Signal Detection. Neuron 87, 179-192. Meijer, J.H., and Robbers, Y. (2014). Wheel running in the wild. Proc Biol Sci 281. Mileykovskiy, B.Y., Kiyashchenko, L.I., and Siegel, J.M. (2005). Behavioral correlates of activity in identified hypocretin/orexin neurons. Neuron 46, 787-798. Mistlberger, R., Bergmann, B., and Rechtschaffen, A. (1987). Period-amplitude analysis of rat electroencephalogram: effects of sleep deprivation and exercise. Sleep 10, 508-522. Moon, H.Y., Becke, A., Berron, D., Becker, B., Sah, N., Benoni, G., Janke, E., Lubejko, S.T., Greig, N.H., Mattison, J.A., et al. (2016). Running-Induced Systemic Cathepsin B Secretion Is Associated with Memory Function. Cell metabolism. Niell, C.M., and Stryker, M.P. (2010). Modulation of visual responses by behavioral state in mouse visual cortex. Neuron 65, 472-479. Okun, M., Steinmetz, N.A., Cossell, L., Iacaruso, M.F., Ko, H., Bartho, P., Moore, T., Hofer, S.B., Mrsic-Flogel, T.D., Carandini, M., et al. (2015). Diverse coupling of neurons to populations in sensory cortex. Nature. Oppezzo, M., and Schwartz, D.L. (2014). Give your ideas some legs: the positive effect of walking on creative thinking. J Exp Psychol Learn Mem Cogn 40, 1142-1152. Pagel, J.I., and Chouker, A. (2016). Effects of isolation and confinement on humans- implications for manned space explorations. J Appl Physiol (1985) 120, 1449-1457. Palchykova, S., Winsky-Sommerer, R., Meerlo, P., Durr, R., and Tobler, I. (2006). Sleep deprivation impairs object recognition in mice. Neurobiol Learn Mem 85, 263-271.

Pandi-Perumal, S.R., and Gonfalone, A.A. (2016). Sleep in space as a new medical frontier: the challenge of preserving normal sleep in the abnormal environment of space missions. Sleep Sci 9, 1-4. Polack, P.O., Friedman, J., and Golshani, P. (2013). Cellular mechanisms of brain state- dependent gain modulation in visual cortex. Nat Neurosci 16, 1331-1339. Rasch, B., and Born, J. (2013). About Sleep's Role in Memory. Physiol Rev 93, 681-766. Reimer, J., Froudarakis, E., Cadwell, C.R., Yatsenko, D., Denfield, G.H., and Tolias, A.S. (2014). Pupil fluctuations track fast switching of cortical states during quiet wakefulness. Neuron 84, 355-362. Rhee, J., Chen, J., Riechman, S.M., Handa, A., Bhatia, S., and Wright, D.L. (2016). An acute bout of aerobic exercise can protect immediate offline motor sequence gains. Psychol Res 80, 518-531. Ritsche, K., Nindl, B.C., and Wideman, L. (2014). Exercise-Induced growth hormone during acute sleep deprivation. Physiol Rep 2. Rolland, Y., Abellan van Kan, G., and Vellas, B. (2008). Physical activity and Alzheimer's disease: from prevention to therapeutic perspectives. J Am Med Dir Assoc 9, 390-405. Rudolph, M., Pospischil, M., Timofeev, I., and Destexhe, A. (2007). Inhibition determines membrane potential dynamics and controls action potential generation in awake and sleeping cat cortex. J Neurosci 27, 5280-5290. Schiemann, J., Puggioni, P., Dacre, J., Pelko, M., Domanski, A., van Rossum, M.C., and Duguid, I. (2015). Cellular mechanisms underlying behavioral state-dependent bidirectional modulation of motor cortex output. Cell Rep 11, 1319-1330. Schneider, S., Abeln, V., Popova, J., Fomina, E., Jacubowski, A., Meeusen, R., and Struder, H.K. (2013). The influence of exercise on prefrontal cortex activity and cognitive performance during a simulated space flight to Mars (MARS500). Behav Brain Res 236, 1- 7. Schneider, S., Brummer, V., Carnahan, H., Kleinert, J., Piacentini, M.F., Meeusen, R., and Struder, H.K. (2010). Exercise as a countermeasure to psycho-physiological deconditioning during long-term confinement. Behav Brain Res 211, 208-214. Shapiro, C.M., Bortz, R., Mitchell, D., Bartel, P., and Jooste, P. (1981). Slow-wave sleep: a recovery period after exercise. Science 214, 1253-1254. Shapiro, C.M., Griesel, R.D., Bartel, P.R., and Jooste, P.L. (1975). Sleep patterns afted graded exercise. J Appl Physiol 39, 187-190. Siegel, J.M. (2008). Do all animals sleep? Trends Neurosci 31, 208-213. Strohle, A. (2009). Physical activity, exercise, depression and anxiety disorders. J Neural Transm 116, 777-784. Stryker, M.P. (2014). A Neural Circuit That Controls Cortical State, Plasticity, and the Gain of Sensory Responses in Mouse. Cold Spring Harb Symp Quant Biol 79, 1-9. Susic, V., and Kovacevic-Ristanovic, R. (1980). Effects of restricted sleep with different exercise loads upon subsequent sleep. Arch Int Physiol Biochim 88, 1-13.

Takahashi, K., Kayama, Y., Lin, J.S., and Sakai, K. (2010). Locus coeruleus neuronal activity during the sleep-waking cycle in mice. Neuroscience 169, 1115-1126. Thomas, M., Sing, H., Belenky, G., Holcomb, H., Mayberg, H., Dannals, R., Wagner, H., Thorne, D., Popp, K., Rowland, L., et al. (2000). Neural basis of alertness and cognitive performance impairments during sleepiness. I. Effects of 24 h of sleep deprivation on waking human regional brain activity. J Sleep Res 9, 335-352. Timofeev, I., Grenier, F., and Steriade, M. (2001). Disfacilitation and active inhibition in the neocortex during the natural sleep-wake cycle: an intracellular study. Proc Natl Acad Sci U S A 98, 1924-1929. Tobler, I. (2005). Phylogeny of sleep regulation. In Principles and Practice of Sleep Medicine K. MH, R. T, and D. WC, eds. (Philadelphia W. B. Saunders). Tobler, I., and Borbely, A.A. (1986). Sleep EEG in the rat as a function of prior waking. Electroencephalogr Clin Neurophysiol 64, 74-76. Torsvall, L., Akerstedt, T., and Lindbeck, G. (1984). Effects on sleep stages and EEG power density of different degrees of exercise in fit subjects. Electroencephalogr Clin Neurophysiol 57, 347-353. Van Dongen, H.P., Belenky, G., and Krueger, J.M. (2011). A local, bottom-up perspective on sleep deprivation and neurobehavioral performance. Curr Top Med Chem 11, 2414- 2422. van Praag, H., Kempermann, G., and Gage, F.H. (1999). Running increases cell proliferation and neurogenesis in the adult mouse dentate gyrus. Nat Neurosci 2, 266-270. Vinck, M., Batista-Brito, R., Knoblich, U., and Cardin, J.A. (2015). Arousal and Locomotion Make Distinct Contributions to Cortical Activity Patterns and Visual Encoding. Neuron. Voss, M.W., Prakash, R.S., Erickson, K.I., Basak, C., Chaddock, L., Kim, J.S., Alves, H., Heo, S., Szabo, A.N., White, S.M., et al. (2010). Plasticity of brain networks in a randomized intervention trial of exercise training in older adults. Front Aging Neurosci 2. Vyazovskiy, V.V. (2013). Cortical neuronal mechanisms of sleep homeostasis. Zh Vyssh Nerv Deiat Im I P Pavlova 63, 13-23. Vyazovskiy, V.V. (2015). Sleep, recovery, and metaregulation: explaining the benefits of sleep. Nat Sci Sleep 7, 171-184. Vyazovskiy, V.V., Achermann, P., and Tobler, I. (2007). Sleep homeostasis in the rat in the light and dark period. Brain Res Bull 74, 37-44. Vyazovskiy, V.V., and Harris, K.D. (2013). Sleep and the single neuron: the role of global slow oscillations in individual cell rest. Nat Rev Neurosci 14, 443-451. Vyazovskiy, V.V., Olcese, U., Cirelli, C., and Tononi, G. (2013). Prolonged wakefulness alters neuronal responsiveness to local electrical stimulation of the neocortex in awake rats. J Sleep Res 22, 264-271. Vyazovskiy, V.V., Olcese, U., Hanlon, E.C., Nir, Y., Cirelli, C., and Tononi, G. (2011). Local sleep in awake rats. Nature 472, 443-447.

Vyazovskiy, V.V., Olcese, U., Lazimy, Y.M., Faraguna, U., Esser, S.K., Williams, J.C., Cirelli, C., and Tononi, G. (2009). Cortical firing and sleep homeostasis. Neuron 63, 865- 878. Vyazovskiy, V.V., Ruijgrok, G., Deboer, T., and Tobler, I. (2006). Running wheel accessibility affects the regional electroencephalogram during sleep in mice. Cereb Cortex 16, 328-336. Vyazovskiy, V.V., and Tobler, I. (2005). Theta activity in the waking EEG is a marker of sleep propensity in the rat. Brain Res 1050, 64-71. Vyazovskiy, V.V., and Tobler, I. (2012). The temporal structure of behaviour and sleep homeostasis. PLoS One 7, e50677. Weinberg, L., Hasni, A., Shinohara, M., and Duarte, A. (2014). A single bout of resistance exercise can enhance episodic memory performance. Acta Psychol (Amst) 153, 13-19. Wu, M.F., Gulyani, S.A., Yau, E., Mignot, E., Phan, B., and Siegel, J.M. (1999). Locus coeruleus neurons: cessation of activity during cataplexy. Neuroscience 91, 1389-1399. Xu, M., Chung, S., Zhang, S., Zhong, P., Ma, C., Chang, W.C., Weissbourd, B., Sakai, N., Luo, L., Nishino, S., et al. (2015). Basal forebrain circuit for sleep-wake control. Nat Neurosci. Youngstedt, S.D., Kripke, D.F., and Elliott, J.A. (1999). Is sleep disturbed by vigorous late- night exercise? Med Sci Sports Exerc 31, 864-869. Zagha, E., and McCormick, D.A. (2014). Neural control of brain state. Curr Opin Neurobiol 29, 178-186. Zlebnik, N.E., Hedges, V.L., Carroll, M.E., and Meisel, R.L. (2014). Chronic wheel running affects cocaine-induced c-Fos expression in brain reward areas in rats. Behav Brain Res 261, 71-78.