Sleep-mediated regulation of reward circuits: implications in substance use disorders

Abstract

Our modern society suffers from both pervasive sleep loss and substance abuse—what may be the indications for sleep on substance use disorders (SUDs), and could sleep contribute to the individual variations in SUDs? Decades of research in sleep as well as in motivated behaviors have laid the foundation for us to begin to answer these questions. This review is intended to critically summarize the circuit, cellular, and molecular mechanisms by which sleep influences reward function, and to reveal critical challenges for future studies. The review also suggests that improving sleep quality may serve as complementary therapeutics for treating SUDs, and that formulating sleep metrics may be useful for predicting individual susceptibility to SUDs and other reward-associated psychiatric diseases.

Introduction

Sleep, as “the main course in life’s feast, and the most nourishing” (—William Shakespeare, Macbeth), powerfully influences our emotional well-being and motivational states. Why is this the case and how does sleep do so? Decades of research in sleep as well as in motivated behaviors have laid the foundation for us to begin to understand the relationship between the two. Studies at circuit, cellular, and molecular levels have lent us increasing insights into their intricate interactions, which have profound implications for questions such as: How does loss of sleep alter motivated behaviors? How does acute sleep loss differ from chronic sleep disturbance? And for our modern society suffering from both pervasive sleep loss and substance abuse—what may be the indications for sleep loss on the disease process of substance use disorders (SUDs) and the individual variations? We will primarily focus on sleep-cocaine interactions to demonstrate the relationships and potential underlying mechanisms, then extend to other substances. Reviews on sleep-opioid or sleep-cannabis interactions can be found elsewhere [1–3].

Sleep composition, function, and association with psychiatric disorders

Sleep is a rapidly reversable and quiescent state characterized by specific sleep postures, reduced response to stimuli, and increased arousal threshold. During sleep, animals cannot forage, reproduce, and are vulnerable to predators. Despite the potential maladaptive nature and disadvantages associated with the prolonged immobility, sleep has been preserved through evolution [4, 5]. Vertebrates such as mammals, birds, and reptiles sleep, and a sleep-like state also exists in invertebrates [6–8]. Sleep can be broadly divided into two main states: non-rapid-eye-movement (NREM) sleep and rapid-eye-movement (REM) sleep (Box 1), which are also preserved across mammalian and avian species [9]. Following acute sleep deprivations (SD), both NREM and REM sleep show rebounds in the duration or intensity during recovery sleep, suggesting that they are under homeostatic regulations [10, 11].

Sleep is important for maintaining various vital physiological functions, including survival [12, 13], restoration of body and mind [4, 14, 15], energy conservation [16], immune functions [17, 18], brain development [19], brain metabolism and waste cleaning [20–22], learning and memory [23], and regulation of emotion and motivation [24, 25]. Good sleep is associated with positive affect and psychophysiological well-being [26], while poor sleep quality is often linked to negative valence and impaired regulation of emotion [27].

Sleep disturbance is a common comorbidity in almost all psychiatric disorders [28, 29]. Insomnia is often observed among patients with mood disorders, anxiety, and SUDs, and thus, chronic sleep disturbance often serves as a diagnostic checkmark for these disorders [30]. In the context of SUDs, sleep problems have been associated with the use or abuse of many substances, including alcohol, nicotine, cannabis, opioids, cocaine, amphetamines, and caffeine (reviewed in [31–33]). Psychostimulants, such as cocaine, nicotine, and amphetamine, may cause sleep loss acutely [33, 34]; narcotics such as opioids can both increase sleepiness and impair sleep quality acutely, and the complex effects with the specific contexts are reviewed elsewhere [1, 2, 35]. Chronic substance uses often lead to persistent sleep disturbances, including difficulty falling asleep, sleep fragmentation, frequent awakenings, reduction of sleep time, poor sleep quality, daytime sleepiness, and abnormalities or shifts in the timing of the cyclic sleep architecture [3, 33, 34]. Following withdrawal or detoxification from alcohol, psychostimulants, or narcotics, sleep disturbances are prevalent [3, 33], with some notable differences. For example, patients with alcohol or cannabis use disorder have higher incidence of sleep-onset insomnia, whereas patients with cocaine or heroin use disorder frequently experience sleep-maintenance insomnia [36]. The extent of sleep disturbances further predicts subsequent use of illicit drugs and alcohol [37]. There have been animal models that recapitulate some of the sleep abnormalities in SUDs, including intermittent ethanol vaper exposure [38], opioid self-administration (SA) [39], and cocaine SA in rodents [40] (Box 2).

Box 1 Sleep states and signature waveforms.

Physiological measurements via electroencephalography (EEG), electrooculography (EOG), and electromyography (EMG) can distinguish sleep versus wakefulness, and further categorize sleep into NREM and REM states [342]. At wakefulness, the cortical EEG exhibits desynchronized low-amplitude waveforms with an alpha-rhythm component while the EMG activity is high. During NREM sleep, EEG exhibits high amplitude and a prominent slow-wave and delta component while EMG activity is reduced. REM sleep is characterized by a low-amplitude EEG with a strong theta component accompanied by muscle atonia (lowest EMG activity), and bursts of rapid-eye movements. In humans, NREM sleep is further subdivided into four stages [343]: Stage 1 is light sleep with mixed frequencies and attenuated alpha rhythms; Stage 2 contains sleep spindles and K-complexes; Stages 3 and 4 are deep sleep, containing high-amplitude slow waves. During a typical night, an average person cycles through different stages of NREM sleep and REM sleep 4–6 times [344]. Similar cyclic sleep-wake architecture is often observed in other mammals [345].

Box 2 Cocaine SA model in rats.

Cocaine self-administration training is conducted in an operant-responding chamber, where the rat learns to obtain infusions of cocaine solution by performing operant respondings (i.e., poking its nose into the active hole or pressing the active lever). The light and/or tone cues are contingently presented with active nose-poking/lever-pressing and cocaine infusions, through which the rat establishes cue-cocaine associations. Nose-poking in the inactive hole or pressing the inactive lever does not result in cue presentations or cocaine infusions.

After acquiring cocaine self-administration over days of repeated training, the rats may undergo drug withdrawal. They are then tested at different times after withdrawal in the same operant chambers with training cues but without cocaine infusions. Presentation of these contingent cues is sufficient to trigger operant responding, a behavioral readout that has been widely used to assess cue-induced cocaine seeking. Over long-term withdrawal, there is progressive intensification of cue-induced cocaine seeking, so called “incubation of cocaine craving”, which suggests growing propensity for drug relapse [54, 346, 347]. Development of cocaine incubation coincides with the gradual accumulation of calcium-permeable AMPA receptors at glutamatergic synapses on nucleus accumbens principal neurons, which critically contributes to the expression of cocaine incubation [210–214].

Sleep-mediated regulation of reward-seeking behaviors

In generally healthy populations, acute sleep loss is often associated with an increase in reward-seeking behaviors. In human adolescents, shorter sleep duration is associated with more snacking and over-consumption of high-calorie food [41]. Pregnant women with poor sleep quality are more likely to have stronger and more frequent food cravings, together with higher hedonic hunger [42]. In young adult males, one night of SD increases the desire for high-calorie foods [43]. Other than food reward, overnight SD increases smoking in healthy cigarette smokers [44], shifts economic preferences toward higher gains [45], and promotes risk taking for higher gains [46]. Similar phenomena have also been observed in animals. Sleep disturbance animal models (Table 1) are often used to examine the impact on a variety of reward-seeking behaviors, including natural and drug reward. In male mice, acute SD (zeitgeber time: ZT0-6, gentle handling) increases the seeking and consumption of sucrose reward but not lab chow [47]. Chronic sleep restriction in male rats (ZT2-6, gentle handling, 4 h/day × 7 days) increases voluntary alcohol consumption [48]. Acute SD (ZT0-4/8, novel object exploration) in male rats increases the rate and efficiency of cocaine infusion during a motivational test, without changing the perceived value of cocaine [49]. Chronic sleep restriction (EEG-based disk-treadmill method, ~25% reduction in baseline sleep over 8 days) in male rats increases the perceived value of cocaine selectively in high drug-taking rats [50]. Chronic REM SD (flowerpot-over-water) in rats lowers the threshold for intra-cranial self-stimulation for reward sensation [51]. While increasing reward seeking during sleep loss may have adaptive values such as consuming calories to sustain energy expenditure, it can also be maladaptive—increasing chances for developing obesity [52, 53], risk taking, and substance use. Meanwhile, the widespread increase in reward-seeking behaviors across species and reward modalities following sleep disturbance also suggests potential common underlying neural substrates.

Table 1.

Sleep disturbance animal models.

Total sleep deprivation
Gentle handling	Rodents are gently handled to be kept awake. The experimenter may pick animals up, gently touch or brush animals’ tails or whiskers, introduce novel objects, shake cages, or brush coat. Total suppression of NREM and REM sleep.	[12, 258]
Automated piezoelectric cages	Automated detection of sleep by piezoelectric signal triggers cage shaking to wake animals up. Total suppression of NREM and REM sleep.	[259]
Automated closed-loop EMG-based running wheel	Automated detection of sleep by EMG signal triggers wheel running to wake animals up. Total suppression of NREM and REM sleep.	[260]
Selective REM sleep deprivation
Flowerpot/single platform (SP)/ modified multiple platform (MMP)	Animals are housed on small platform(s) or flowerpot over water. Entering REM sleep causes the animals to fall into water and wake up. Abolish REM sleep and reduce SWS sleep.	[261, 262]
Disk-over-water method	Automated detection of sleep or REM sleep via EEG/EMG signals triggers rotation of the disk placed over water and forces the animal on top of the disk to move to avoid falling into water. Chronic total or partial sleep/REM sleep deprivation.	[262, 263]
Pendulum/swing	Continuously oscillating apparatus keeps the animals from entering REM sleep through imbalanced postures. Abolish REM sleep and reduce significant NREM sleep	[264, 265]
Cold environmental exposure	Cold exposure to low ambient temperature < 10 °C. REM sleep loss proportional to cold exposure	[266]
Chronic sleep restriction
Automated closed-loop rotating disc	Automated detection of sleep via EEG/EMG signals triggers rotations of the bottom plate in pulses to wake up the animals, reducing a proportion of their sleep time. Chronic sleep restriction/partial sleep deprivation.	[50]
Chronic sleep fragmentation
Treadmill	Treadmill belt with a cylinder-shaped object underneath moving in both directions to wake up the animal. REM sleep restriction in cocaine withdrawn rats; Chronic REM sleep fragmentation in mice.	[40, 74]
Sleep enhancement
Environmental warming	Increasing ambient temperature towards thermal neutral zone increases sleep. Maximal REM sleep at ~29 °C.	[215, 216]
Various region or cell type-specific stimulation/inhibition	Stimulation of NREM or REM sleep-promoting brain areas or specific cell types to promote NREM or REM sleep; or suppression of wake-promoting mechanisms

It has been challenging to understand the function of sleep in psychiatric disorders. Much of previous work performs correlational analysis to determine the relationship between sleep phenotypes and psychiatric disorders with difficulty addressing causality. Animal models are designed to address causality with mechanistic insights by introducing SD or sleep restriction, often using rodents (reviewed in [267]). Some of these procedures inevitably involve additional stress, which may confound the interpretation of results [180].

The brain reward circuitry—composition and functional interaction with sleep

The mesolimbic reward pathway is an interconnected neurocircuit that regulates reward-cue encoding, reward evaluation, and execution of reward seeking. The nucleus accumbens (NAc) resides in the ventral striatum and serves as a limbic-motor interface that integrates and prioritizes emotional and motivational inputs for motor outputs [54–57]. The NAc receives convergent glutamatergic inputs from the medial prefrontal cortex (mPFC), hippocampus, amygdala, among other regions, which carry various information on reward-associated cues, context, and executive controls. Moreover, the NAc is an important target of the mesolimbic dopamine (DA) projection, which carries information on reward-cue salience and reward prediction error [58, 59]. In addition, NAc neurons express a rich repertoire of neuropeptide receptors, including opioid, hypocretin/orexin (Hcrt), and melanin-concentrating hormone (MCH) receptors and many more, relaying information from the hypothalamus, thalamus, midbrain, brain stem, etc. to influence various reward-associated behaviors [60]. Thus, the NAc is a converging hub where top-down controls from the cortex interact with a variety of bottom-up emotional and motivational drives (Fig. 1a). Increasing evidence suggests that the reward circuit presents multi-layered targets for sleep and sleep disturbance to regulate reward-associated emotional and motivational responses.

Fig. 1. Sleep–reward interactions and implications in substance use disorders.

a Acute sleep loss often results in reduced top-down controls, contrasting with increased bottom-up drives. Following acute SD, PFC shows reduced coupling with the NAc and amygdala (AMG), possibly through adenosine build-up and reduced glutamatergic transmission efficacy; some subregional increase in activity is also observed. Hippocampus (Hipp) shows deficits in glutamatergic signaling and synaptic plasticity. NAc shows increased reactivity to reward or risk taking, possibly through decreased dopamine D2/3 receptor signaling, thus biasing toward D1-receptor signaling. The VTA and AMG show increased reactivity to emotional stimuli. The AMG also shows increased reactivity to food reward-associated cues. The hypocretin (Hcrt)/orexin (Orx) system based in the LH may represent a regulatory hub. It shows enhanced activity following sleep disturbances through increase in the activity of Hcrt neurons, increase in prepro-orexin synthesis and orexin release, and increase in OX1R and OX2R receptor expressions in the target regions. Hcrt signaling may orchestrate various neurotransmitter and modulator systems, including glutamate, GABA, acetylcholine, DA, norepinephrine, and 5-HT, and promote reward-associated sign-tracking (PVT-NAc), attention (basal forebrain-PFC), impulsivity and risk taking (PFC, NAc), and reward perception (NAc). These effects may, together, lead to typical increases in reward-seeking behaviors, including palatable food wanting and consumption, alcohol intake, cigarette smoking, and illicit drug use. DR: dorsal raphe nucleus (a main source of 5-HT neurons). Details see Tables 2 and 3. b Withdrawal from chronic drug use affects not only the reward circuit, but also the sleep-regulatory mechanisms, often resulting in persistent sleep disturbances at night and excessive sleepiness during the day, which may, in turn, facilitate drug relapse. In rats trained to self-administer cocaine, drug withdrawal induces excessive activation of Hcrt neurons in the LH, which may contribute to sleep instability as well as increase in drug seeking. Moreover, cocaine withdrawal dampens MCH neuron intrinsic membrane excitability and impairs glutamatergic transmissions, which may compromise MCH neuron’s role in REM sleep regulations. Hypofunction of MCH neurons may result in dis-inhibition of downstream targets, worsening cocaine-induced hyperactivity of Hcrt neurons in LH and exacerbating CP-AMPAR accumulation at NAc principal neuron synapses, which together facilitate incubation of cocaine craving. REM sleep interventions, including REM sleep restriction-rebound and sleep-warming, engage MCH neuron activities in sleep, and promote REM sleep through downstream targets such as medial septum (MS), tuberomammillary nucleus (TMN), and dorsal raphe (DR), among others. MCH neuron activities in REM sleep may also produce anti-relapse effects hypothetically through MCH-Hcrt neuron interactions in the LH as well as downstream mutual targets including the NAc, hippocampus, amygdala, and VTA.

The NAc-interconnected reward circuit has substantial anatomical overlap with the sleep-regulatory network (Table 2; details reviewed in [1, 61]). Some regions regulate both reward and sleep, including the PFC, NAc, ventral tegmental area (VTA), habenula (Hb), and lateral hypothalamus (LH); most of them are known to be affected by sleep disturbance in various ways (Table 2). How these regional changes induced by sleep disturbances may orchestrate to produce behavioral outcome is not fully understood. One example comes from human functional MRI studies. Following acute SD, the medial frontal cortex shows reduced coupling with the amygdala and NAc in response to pleasure-evoking stimuli, suggesting reduced top-down controls, whereas amygdala, NAc, and VTA activities show increase to various reward stimuli, suggesting increased bottom-up drives [62]. Thus, the overall compromised top-down controls combined with increased bottom-up drives may synergize to result in biased reward processing favoring risk taking and reward seeking after acute SD [63]. This notion would be consistent with (1) increased impulsivity and decreased inhibitory control sometimes observed after acute SD [64–67]; and (2) increased subjective value of the reward or reward-associated cues following acute sleep disturbances. In rats and mice, acute SD increases the preference to contextual cues associated with cocaine or amphetamine experience [68, 69]. In humans, preference for the stimulant methylphenidate is driven by sleepiness [70]; and higher perception of cocaine strength is reported after 24 h of SD [71]. Consistent with these changes in top-down versus bottom-up drives, SD can shift the decision-making strategy from loss-defending toward gain-seeking [45, 72].

Table 2.

Overlapping sleep-reward brain regions.

Brain region	Reward function	Impact by sleep disturbance	Sleep/arousal function
PFC	Executive evaluation of reward value and outcome [268–270]	Reduced coupling with the amygdala, with subregional increase in activity [62]; reduced synaptic release of glutamate onto NAc [47]; both following acute SD	Generation of slow-wave activity [271, 272]
Hippocampus	Memory and learning, attribution of context to reward [273]	Impaired NMDA receptor-dependent synaptic plasticity; deficit in hippocampus-dependent memory tasks following acute or chronic sleep disturbance [274]	Reactivate memory in sleep through sharp-wave ripples to communicate with prefrontal cortex [275, 276]
NAc	Reward prediction and translation, goal orientation, reward-associated learning [270]	Increased reactivity to monetary reward or high risk in adolescents/young adults following acute SD [196, 277]; Reduced probability of glutamate release at mPFC and rostral amygdala inputs onto principal neurons following acute SD [47, 278]	Slow-wave sleep initiation via A2A /D2 receptor-expressing neurons [279, 280]; arousal through A2A receptor inhibition [281], or D1-receptor neuron activation [280, 282].
VTA	Reward signaling and prediction. Approach, reward-related learning, and conditional acquisition [283]	Increased reactivity to positive emotional stimuli following acute SD [62]	VTA DA neuron activity promotes arousal [284]
Amygdala	Valence encoding of stimuli and cue association [270, 285]	Increased reactivity to food reward anticipation or negative emotional stimuli, decreased coupling to mPFC and orbital frontal cortex following acute SD in humans [43, 62]
Habenula	Reward value-based decision-making, signaling of aversion [286, 287]	Increase in pace-making spontaneous firing of cholinergic projection neurons in medial habenula, following 5–7 days of chronic REM sleep fragmentation [74]	Suppression of motor movements related to circadian control of behavior and REM sleep generation via melatonin and cytokine interleukin-18 [286, 287]
Lateral hypothalamus	Reward evaluation and motivation for food and drug reward [288]	Increased activity of Hcrt neurons or orexin levels following acute to chronic sleep restriction [80, 220, 289–292]; increased MCH neural activity following recovery sleep [220]	Initiation of sleep on/ off mechanisms associated with orexin/MCH neuronal activation [293]
Dorsal raphe nucleus	Activity is correlated with reward values, reward delay, task progress, and reward absence; modulation of reward waiting and reinforcement [294, 295]	Total SD increases dorsal raphe neuron mean firing rates in cats, maximal increase at 15 h of total SD [296]	Promotion of waking and inhibition of REM sleep via serotonergic inputs [297]

It should be noted that the above results in humans are mostly obtained from adults or young adults, and there may be important differences in the adolescent brain regarding the sleep-reward responding relationship. In early through late pubertal adolescents, there is less activation in the caudate during both reward anticipation and reward outcome in individuals with shorter sleep/later onset and earlier offset and lower sleep quality [73]. However, there have been limited sleep manipulation studies in adolescents—it is not clear whether acute SD in adolescents may similarly blunt subcortical reward responding, and whether reward “insensitivity” in this context may also contribute to risk taking and high-reward seeking.

Compared to acute SD, less is known about the impact of chronic sleep disturbance on the reward circuit. Recently, using a low-stress sleep fragmentation paradigm in mice, it was shown that the medial Hb (mHb) exhibits increased spontaneous pace-making activity following 5–7 days of chronic REM sleep fragmentation [74]. This is postulated to promote mHb-mediated anxiety and anhedonia [75–79] following chronic sleep disturbance. The potential molecular and cellular mechanisms mediating these sleep disturbance-induced changes will be discussed in the next section.

Molecular and cellular mechanisms that contribute to sleep-mediated regulation of reward function

Neurotransmitter and neuromodulator systems

Our insights into sleep-mediated regulation of neurotransmitter/modulator systems predominantly come from acute SD studies, with limited results from chronic sleep manipulations. Acute SD has broad impact across brain regions and over many neurotransmitter systems, including SD-induced extracellular accumulation of metabolites (e.g., adenosine, ceremide) that promote sleep, and changes in pro-wakefulness neurotransmissions (e.g., DA, Hcrt). When these changes occur in the reward circuit, they may result in acute changes in reward processing and thus impact associated behaviors. Summarized in Table 3 are details on the sleep manipulation paradigms, impacted neurotransmitter/modulator systems, brain regions, cellular effects, and outcome or expected outcome on reward-associated behaviors. A few are highlighted below.

Table 3.

Sleep disturbance affects neurotransmitter/modulator systems that impact reward function.

Neurotransmitter/modulator system		Sleep manipulation	Species	Brain region sampled	Cellular/ Molecular effect	Impact on reward function
Glutamatergic	Presynaptic glutamate release	Acute SD 6 h by gentle handling (ZT0-6)	Mouse (male)	mPFC-to-NAc and rostral amygdala-to-NAc inputs	Reduce presynaptic release probability [47, 278]	Reduction in top-down inhibition; increase in sucrose SA [47, 278]
	mGluR5	Acute SD 33–40 h (human); 6 h (ZT0-6, mouse)	Human (male); Mouse (male)	Whole brain	Increase in mGluR5 receptor availability, no change at mRNA level [298,299]	Anti-depressive effect in mouse, through increase in mGluR5-mTOR signaling; increase in AMPA receptor function [300]
	Homer1a	Acute SD 1, 3, or 6 h	Mouse (male, female)	Cortex, hippocampus, claustrum, cingulate, piriform cortex	Increase in homer1a [301, 302]; homer1a uncouples mGluR5 from effectors [303], constitutively activates mGluR1a and mGluR5 [304]	Anti-depressive effect [300, 305, 306]
Dopaminergic	D2/D3 receptors	Acute SD for 1 night	Human	Ventral striatum	Reduced availability of D2/D3 receptors	Similarly observed in chronic cocaine users [307–309]; indicated in increase of risk taking, impulsivity, and compulsive drug taking (human and rat) [137, 138]
Serotonergic	Serotonin (5-HT)	Acute SD 24 h (human) 6–8 h (mouse) 3–12 h (rat)	Human Mouse Rat	Cortex, Hippocampus	Increase cerebral 5-HT 2A receptor binding in humans [310]; increase 5-HT 2A receptors in the frontal cortex of mice [311]; increase in 5-HT turnover in frontal cortex and hippocampus etc. in rats [312] (but see [313])	Reduced 5-HT turnover may render susceptibility to depression [314]; the impact of SD-induced increase in 5-HT tone on reward-associated behaviors is not clear
Cholinergic	Acetylcholine	Chronic selective REM sleep fragmentation 5–7 d	Mouse	mHb	Increase in tonic firing; reduced KCNK9 activity in cholinergic neurons [74]	Not reported in [74]; but increase in mHb cholinergic activity promotes anxiety and depressive-like behaviors [75–79]
Peptidergic	Hypocretin/Orexin	Acute SD 6 h (ZT0-6) by different stimuli	Rat (male)	LH	Increase in hypocretin-1 levels (i.e., orexin-A) [315]	Not reported
		Acute SD 24 h by gentle handling, sound, or short-walk	Dog (normal/narcoleptic)	CSF	Increase in CSF HCRT-1 levels in both normal and narcoleptic dogs [92]	Positive correlation between HCRT-1 level and motor activity during SD [92]
		Acute SD 6 h (ZT0-6) by gentle handling	Rat (male)	Hypothalamus	D-type orexin neurons show increased membrane excitability. Both D-/H-type show inhibition of excitatory synaptic inputs [316]	Not reported
		SD 12 h/day for 5 weeks using a Columbux device	Rat (unknown sex)	Hippocampus	Increase in relative protein level of orexin-A, OX1R, and OX2R [317]	Not reported
		4-h sleep/night for 5 nights (partial SD)	Human (Male, female)	CSF	Increase in orexin levels [289]	Not reported
		REM SD for 4 days; flowerpot method	Male Wistar rats	Locus coeruleus, cortex, hippocampus, pedunculo-pontine tegmentum	Increase in orexin-A levels [290]	Not reported
		SD 3 h (ZT5–8); gentle handling	Rat (male)	Hypothalamus	Increase in c-Fos immuno-reactive orexin cells [220]	Not reported
		SD 72 h; treadmill	Rat (male adolescent)	LH, posterior hypothalamus	Increase in c-Fos immuno-reactive cells in which some are orexin cells [291]	Increase in energy expenditure and anxiety; decrease in locomotion
		SD 24–96 h or SD 18 h/day for 21 days; platform method	Rat (male)	Hypothalamus	Increase in prepro-orexin mRNA levels [318]	Decrease in food intake at SD-72 h and SD-96 h [318]
		REM SD 6 h or 96 h; platform method	Rat (male)	CSF	Increase in hypocretin-1 levels [292]	Not reported
		SD 4 h (begins around ZT0–ZT2); treadmill method	Mouse (male)		No cellular or molecular measurements	Administration of an OX1R antagonist before cocaine conditioning trials or before postconditioning trials prevents SD-enhanced cocaine CPP [68]
	MCH	REM sleep increase by environmental warming [215,216]; REM sleep rebound [220]	Rat (male, cocaine SA and withdrawal)	LH MCH neurons and projection targets	Increase in calcium activity in MCH neurons during REM sleep; intra-NAc MCH receptor signaling reduces synaptic calcium-permeable AMPA receptors in MSNs [207]	Reduces incubation of cocaine craving after long-term withdrawal [207]
Metabolites	Adenosine	Acute SD 2, 6, 12 h	Human Rat Mouse	Cortex, thalamus, striatum, cerebellum, basal forebrain	Increase in adenosine release and A1 receptor expression [319–321]; homeostatic slow-wave regulation [322]; increase in mGluR5-homer1a activity [305]	Reduction in top-down inhibition; anti-depressive effect [323–325]
	Ceramide	Acute SD 12 h	Mouse (male, female)	Thalamic reticular nucleus (microglia)	Increase in ceramide; decrease in thalamic reticular nucleus neuron membrane excitability [139]; Ceramide modulates DA transporter [326]	Decrease in dorsal hippocampus ceramide is associated with faster extinction of learned appetitive behaviors in rats [142]; inhibition of ceramide biosynthesis attenuates the development of tolerance to morphine [143]

Hcrt

Hcrt signaling not only sustains wakefulness and promotes arousal [80–100], but also facilitates reward seeking including natural and drug rewards [68, 98, 101–124]. Acute SD and chronic sleep restriction both enhance Hcrt signaling through increasing prepro-orexin synthesis, orexin-A release, and/or OX1 and OX2 receptor expressions (Table 3). Sleep disturbance-induced increase in the activities of Hcrt system is thought to critically contribute to dysregulated reward seeking ranging from binge eating disorder to SUDs (reviewed in [123–125]). However, direct assessment of such causal relationships has been limited. Recently, it was shown in male mice that acute SD (4 h starting at ZT0-ZT2) enhances cocaine conditioned place preference (CPP), and that systemic administration of an orexin 1 receptor antagonist attenuates the SD effect [68], suggesting that SD-induced increase in Hcrt signaling can indeed enhance drug-reward seeking. The Hcrt system may regulate diverse aspects of reward-seeking behaviors through its wide projection targets and versatile cellular effects. For example, Hcrt through receptor signaling in the paraventricular thalamus (PVT) critically regulates reward-associated sign-tracking conditioned responses [126]; Hcrt signaling in the VTA increases glutamate release and transmission efficacy onto DA neurons, facilitates DA release (reviewed in [127]), and promotes impulsive behaviors [128]; in the basal forebrain, Hcrt increases cholinergic cell activity, resulting in robust acetylcholine release in the PFC and sustained attention [129]; in the NAc, Hcrt increases GABAergic transmission onto the principal neurons (tested in dissociated neurons [130]), and increases hedonic “liking” and/or “wanting” of palatable food reward [131]. In addition, Hcrt also regulates serotonergic (5-HT) system in dorsal raphe, norepinephrinergic system in the locus coeruleus, and histaminergic system in the tuberomammillary nucleus (TMN) (reviewed in [1]). Thus, the Hcrt system orchestrates a full collection of neurotransmitter and modulator systems to promote motivational activation in various circumstances (reviewed in [132]). Their up-regulation in response to sleep disturbances may tip the top-down (e.g. attention) versus bottom-up balance (e.g. reward-cue salience, hedonic values etc.), resulting in reward dysfunction.

DA

There are mixed results regarding DA release following sleep disturbance. Acute SD (6 h, gentle handling) in rats increases extracellular DA metabolites in the basal forebrain [133]; acute SD (4 h, gentle handling) in hamsters increases hypothalamic DA and its metabolites [134]; chronic REM SD (flowerpot-over-water; 96 h) in rats increases DA metabolites in the striatum but not cortex [135]; yet REM SD for 16 h in rats does not appear to modulate overall DA release in the NAc assessed by microdialysis [136], suggesting procedure-dependent, and brain region-specific regulations. Furthermore, acute SD decreases the availability of DA D2/3 receptors in the NAc in humans [136]. This is thought to result in a D1-D2/3 imbalance, which may increase the tendency for impulsivity and risk taking [137, 138].

Adenosine

The accumulation of extracellular adenosine upon acute SD suppresses neurotransmitter release, and, when occurs in the cortex, would “tune down” the top-down control. On the other hand, adenosine modulates postsynaptic mGluR5-homer1a-mTOR signaling as shown in the cortex, producing anti-depressive effects (Table 3).

Ceramide

A recent study identified a lipid product ceramide, which accumulates during acute SD. It promotes sleep through direct inhibition of thalamic reticular neuron firing [139]. Interestingly, the source of ceramide is thought to be microglia [139], the CNS immune cells that play active roles in synaptic pruning during sleep [140, 141]. Related to reward and drug seeking, a decrease in ceramide in dorsal hippocampus is associated with faster extinction of learned appetitive behaviors in rats [142]. Moreover, inhibition of ceramide biosynthesis attenuates the development of tolerance to morphine [143]. Whether and how ceramide accumulation from SD may impact reward function is not known.

In summary, changes in neurotransmitter systems underlie the reduced top-down controls and increased bottom-up drives, which may together increase reward motivation and seeking (Fig. 1a). This will have important implications for SUDs (see below).

Synaptic plasticity

In addition to acute modulation of neurotransmissions, sleep can also impose enduring effects by inducing long-term synaptic plasticity. A great deal of insight was obtained from learning-and-memory studies in the visual, somatosensory, and motor cortices as well as the hippocampal circuits, where diverse mechanisms are engaged in sleep to produce long-term synaptic potentiation or depression as well as homeostatic synaptic scaling [23, 144–153]. On the one hand, slow waves during NREM sleep and theta waves in REM sleep sort different neuronal ensembles across the peaks and throughs, enabling spike timing-dependent synaptic potentiation or depression [149, 154, 155]; on the other hand, NREM and REM sleep states provide dramatic fluctuations in neuromodulators such as increase in acetylcholine and MCH in REM sleep, decrease in 5-HT, and almost complete silence of norepinephrine transmission in REM sleep, which further modulate synaptic plasticity [145, 153, 155–163]. Furthermore, imposed upon these background fluctuations are the “memory replays” that occur both during NREM and REM sleep, which combine with the synchronous population activity and fluctuation of neuromodulators to influence synapses and neurocircuits. These sleep effects are postulated to mediate distinct aspects of learning, memory, as well as creativity [151].

Whereas most of sleep-assisted synaptic plasticity is studied in the context of sensory cortex development, cognition, or motor learning and memory, similar cellular processes may occur in the limbic circuit to regulate emotion and motivation. Many “signature” sleep waves occur in the reward circuit. Slow-wave activity in NREM sleep, the most prominent sleep EEG feature, is frequently generated in the PFC-orbital frontal cortices and propagates as a wave [164]; theta waves as a prominent feature in REM sleep are found in the hippocampus, amygdala, and cortex (reviewed in [165]). These waves could provide opportunities for spike timing-dependent plasticity to take place in large scales in the reward circuit. Moreover, “memory replays” are often observed in hippocampal “place” cells during sleep [166–168]. In mice acquired cocaine CPP, the hippocampal “place” cells are coupled to NAc principal neurons in wake and sleep through theta oscillations, which is thought to underlie the potentiation of hippocampal-NAc coupling and the increase in NAc principal neuron firing in the cocaine zone [169]. In addition to promoting synaptic strengthening, sleep may also facilitate learning-associated synaptic weakening. In the mouse frontal association cortex, auditory-cued fear conditioning induces spine eliminations. This process is dependent on REM sleep through a calcium-dependent mechanism [170]. Finally, opposite to promoting “memory”, sleep may also enable “forgetfulness”. For example, in the mouse hippocampus, MCH neuron terminal activity suppresses pyramidal neuron firing, and MCH neuron activities during REM sleep promote the elimination of contextual fear memory [171]. Thus, REM sleep may serve to modify contextual memory, which together with reducing the affective tone of emotional memory [25, 63], may provide protections against the development of post-traumatic stress disorder phenotypes in rodents and humans [172–174]. In summary, sleep is integral to synaptic plasticity across cortex, hippocampus, and the interconnected circuit. It is fully capable of playing diverse roles in the formation and modification of emotional memories. However, direct demonstrations of sleep-specific synaptic plasticity that regulates natural or drug reward-seeking behaviors have been limited.

Sleep-mediated regulation of gene expression

Another avenue for sleep to impose long-term changes is through regulation of gene expressions. This is reflected in sleep disturbance-induced changes in the transcriptome [175–177] or epigenome. The latter is just beginning to be understood [178]. Changes in gene expression and regulation could be the result of neuronal and glial ensemble activity in sleep or sleep disturbance combined with neurochemical changes discussed above, which may, in turn, feedback to sustained regulation of neural activity. Indeed, comparing mouse cortical transcriptome between sleep, 6-h-SD, and 4-h-SD + 2-h-sleep recovery conditions, more than half of the SD-altered genes continue to be differentially expressed after recovery sleep, suggesting sustained changes induced by SD [175]. Although acute SD profoundly alters the cortical transcriptomes [175, 177], whether and how these changes alter subsequent reward processing is poorly understood. Notably, a few of the SD-sensitive transcriptional hubs in the cortex or basal forebrain are similarly affected by repeated drug exposures, including MEF2C, PER2, JUNB, NFκB, FOSB, and EGR1 etc., some of which have been examined for potential implications in SUDs (Table 4). Thus, sleep disturbance may interact with drug experience at the level of transcription regulation to exert long-term impact on cellular functions and behaviors, including predisposition to drug use, drug experience, and drug withdrawal. Nonetheless, the transcriptomic data from sleep studies are mostly focused on the cortex, whereas data from SUD patients or addiction animal models are overwhelmingly focused on the NAc [179]. Thus, future studies need to bridge the two by examining overlapping brain regions shared by sleep and reward regulations.

Table 4.

Transcription regulations shared by sleep disturbances and substance use experience.

Transcription factor	Sleep manipulation	Species	Assay	Region-specific changes		Drug regimen	Species	Assay	Region-specific changes
MEF2C	Acute SD (ZT0-6)	Mouse	Anti-phospho-MEF2C antibody labeling	Decrease in MEF2C P-S396, suggesting increase in activity in frontal cortex [175]		Cocaine SA (0.33 mg/kg/infusion; 2 h/day × 7day)	Rat	RT-qPCR [327]; IHC [328]	mRNA increase in cingulate cortex [327], and striatum [179, 328]
PER2	Acute SD by gentle handling (ZT0-8)	Rat	Micro-array + qPCR	mRNA increase in frontal cortex [177]; reviewed in [329]		• Per2 regulates MAOA, DRD2, and VGlut1 etc. expression [330] • Per2 mutant mice show altered cocaine-, methamphetamine-, morphine-, and alcohol-elicited reward-associated behaviors [330, 331] • Per2 polymorphism is associated with cocaine addiction in human [256]
JUNB	Acute SD by gentle handling (ZT0-8)	Rat	Micro-array + qPCR	mRNA increase in frontal cortex [177]		Cocaine intoxication; Cocaine dependence	Human	Total RNA sequencing on single neuronal nuclei isolated from post-mortem brain [332]; RNA sequencing [333]	mRNA increase in dlPFC [332, 333]
NFκB	Acute SD by gentle handling (ZT0-6)	Mouse	Protein electrophoresis	Protein increase in cortex [334]		Opioid use disorder	Human	RNA sequencing	Increase in NFκB signaling in dlPFC and NAc [179, 335]
FOSB	Acute SD by gentle handling (ZT0-6)	Mouse	RT-qPCR	mRNA increase in cortex, Basal forebrain [336];		Cocaine dependence	Human	RNA sequencing [333]	mRNA increase in dlPFC [333] and NAc [179]
EGR1	Acute SD by gentle handling (ZT0-8)	Rat	Micro-array + qPCR	mRNA increase in frontal cortex, basal forebrain [177]		Cocaine or amphetamine	Rat	In situ hybridization	Induced in ventral and dorsal striatum [337–341]

Chronic sleep disturbance may affect the reward circuit and cellular processes in ways distinct from acute SD effects. Research in this area has been limited, and mostly relies on high-stress sleep disturbance models (Table 1; [180]). Using disk-over-water method for chronic sleep restriction in rats, it was shown that long-term sleep restriction alters cortical transcriptome qualitatively different from that following acute SD [176]. In a separate study, rats were sleep-restricted using a periodically rotating wheel for 18 h daily, and adrenergic receptors of different subtypes in the basal forebrain and anterior cingulate cortex show differential expressions following 1 day versus 3–5 days of sleep restriction [181], consistent with the notion that sleep undergoes allostasis as sleep debt accumulates [182, 183]. However, it is not known how the specific changes may affect cortical function or reward processing. Using a customized treadmill system for chronic selective REM sleep fragmentation with minimum stress, it was shown that the mHb cholinergic neurons in mice increase tonic firing through reduced activity of an acid-sensing potassium channel KCNK9 [74]. The behavioral consequences of this effect are not known, although increase in mHb activity may modulate a variety of affect/reward functions ranging from aversion, anxiety, anhedonia, to nicotine and alcohol abuse, as well as reinstatement of cocaine CPP [77–79, 184–187]. Understanding the impact of chronic sleep disturbance on the reward circuit is of high clinical significance. This will be greatly facilitated by the development of robust, automated, and noninvasive chronic sleep disturbance models that introduce minimum procedural stress.

Implications in SUDs

Initial drug use

In humans, initiation of drug use often occurs during adolescence [188–190]. The median ages for first opportunity and first use of illicit drugs are 13 years and 14 years, respectively, in the US as of 2011 [191]. In 2019, 1.8 million among 12–17-year olds had first time use of alcohol, 1 million for marijuana, 385,000 for cigarette smoking, and 82,000 for cocaine [192]. Moreover, based on a 2010 survey in a nationally representative sample (N = 2524) of 10th graders in the US, 8% were “predominant polysubstance users” [193]. This is greatly concerning because of the association between younger initiation and poorer outcome in developing SUDs [188–190]. There are many reasons that set adolescence a vulnerable period for initiating drug use [194], and compromised sleep is an unequivocal one. Although 8–10 h of sleep per night is needed for optimal function in adolescents, only 29% of US high school students reach this amount, and 44% of US high schoolers sleep for less than 6 h per night [195]. Acute sleep loss weakens top-down controls and increases bottom-up drives for reward, which may result in biases toward reward-associated sign-tracking (speculated above), increases in impulsivity, risk taking, and reward responding [46, 196] (Fig. 1a). All these aspects, combined with the fast-developing mesolimbic system and slowly maturing prefrontal cortex in the adolescent brain [194], may facilitate drug seeking and taking behaviors.

Compared to acute SD, implications of chronic sleep disturbance in SUDs are less explored. A longitudinal study shows that over 70% of individuals with insomnia or hypersomnia develop some forms of psychiatric problems, including major depression and SUDs, in their lifetime [197]. Insomnia is also a commonly reported cause for self-medication with alcohol, which often precedes the development of alcohol use disorder [198]. It is not clear, however, whether it is chronic sleep problems that lead to substance use as self-medication, or there could be a third causal factor—e.g., genetic-environmental predisposition—that leads to both sleep problems and SUDs, with sleep problems manifesting first.

Relapse to drug use after withdrawal—preclinical studies targeting sleep improvement

Sleep-drug use interactions take place since the initial drug use and persist through long-term withdrawal [2, 3, 199]. On the one hand, sleep dysregulation persists, indicating drug withdrawal-induced prolonged neurophysiological changes; and conversely, chronic sleep disturbance is thought to precipitate relapse to drug use [2, 49, 50, 199–205]. The relationship between sleep disturbance and SUDs is thus thought to be bi-directional, which forms a vicious cycle that drives continued substance use and relapse [200]. Sleep disturbance may tilt the balance between top-down controls and bottom-up drives that contribute to drug craving and relapse after withdrawal. However, the bottom-up drive at this stage is likely to be more about alleviating negative affect than obtaining reward [206].

Sleep disturbance resulting from drug withdrawal could be pathophysiologically different from externally imposed SD. For example, withdrawal-induced sleep disturbance may reflect sleep allostasis, and experimentally induced sleep disturbance in either naïve or drug-experienced individuals is a deviation from homeostasis. Considering this difference, it is important to use animal models that recapitulate the long-term sleep disturbances following substance use and withdrawal, and examine whether increasing or improving sleep may be beneficial.

Increasing sleep is not the mathematical opposite of SD, as sleep is not simply a lack of wakefulness, but rather an active process of similar scales. This is clearly demonstrated in transcriptomic studies, which show a large number of sleep-specific gene expressions qualitatively distinct from those during wakefulness or SD [177]. In this regard, it is equally important to complement sleep disturbance studies by addressing whether increasing sleep time or improving sleep quality after drug withdrawal affects relapse.

To tackle this problem, we have used a rat cocaine SA model. Following SA training (0.75 mg/kg/infusion, 1 overnight + 5 day × 2 h/day) and withdrawal, rats experience sustained sleep loss and fragmentation [40], qualitatively recapitulating the human situations [3]. Moreover, this cocaine procedure induces “incubation of cocaine craving”—a progressive intensification of cue-induced cocaine seeking after withdrawal, which indicates increased propensity for drug relapse (Box 2). Additionally, this cocaine procedure induces a persistent synaptic accumulation of calcium-permeable AMPA receptors (CP-AMPARs) after withdrawal in NAc principal neurons [40, 207–209], which are key synaptic substrates for cocaine incubation [210–214]. These features allow for behavioral and electrophysiological assessment of the effects of sleep interventions after cocaine experience.

First, using a sleep restriction-rebound strategy to increase and consolidate light-phase (inactive phase) sleep in rats, we found that selective REM sleep restriction-rebound during withdrawal day 22–42 leads to decreased incubation of cocaine craving tested on withdrawal day 45. This is accompanied by decreased CP-AMPARs at NAc principal neuron synapses [40]. Next, focusing on the role of REM sleep in this regulation, we used environmental warming to selectively increase REM sleep. Warming to near thermoneutrality preferentially increases REM sleep in drug naïve animals [215,216]. In rats after cocaine withdrawal, warming leads to a selective increase in total REM sleep time and decrease in REM sleep fragmentation, and the REM sleep effects are accompanied by reduced incubation of cocaine craving as well as decrease in NAc CP-AMPARs [207]. Together, these results suggest a potential causal relationship between REM sleep and relapse to cocaine use.

LH MCH system comes into play—contrasting with Hcrt system

LH MCH neurons critically regulate REM sleep onsets and/or bout durations [217–219]. Both REM sleep restriction-induced rebound in REM sleep and warming-induced increase of REM sleep engage the activity of LH MCH neurons [216, 220]. These neurons are predominantly REM sleep-active [221], with greater population activities during long-bout versus short-bout REM sleep episodes [207]. Cocaine withdrawal-induced REM sleep fragmentation concurs with persistent decrease in the membrane excitability of these neurons, as well as impairment in glutamatergic transmission efficacy [222]. To counteract cocaine-induced deficits in MCH neurons, enhancing MCH neuron activities in sleep by environmental warming or direct chemo- or optogenetic stimulations similarly decreases cocaine incubation [207]. Moreover, locally infusing MCH peptide into NAc decreases cocaine-induced synaptic accumulation of CP-AMPARs and reduces incubation [207], suggesting that MCH-to-NAc projections could play an important role. In the NAc, cocaine-induced CP-AMPARs are enriched in GluA1 subunits [210, 211, 223, 224]. MCH receptor activation in the NAc dephosphorylates GluA1 at Serine 845 and facilitates their removal from synapses [225]. Thus, consolidated REM sleep may engage MCH neurons to regulate cocaine seeking after withdrawal in part through regulating NAc CP-AMPARs (Fig. 1b).

In contrast to the hypoactivity of MCH neurons, the LH Hcrt neurons show enhanced functionality following cocaine experience. After rats learn to self-administer cocaine with intermittent access for 14 d (0.2 mg/infusion, 5 min/30 min × 12 epochs/d), they undergo withdrawal for >150 d, followed by a re-exposure to the SA environment. The rats show an increase in the number of Hcrt neurons in LH and increase in Hcrt neuron activities, as measured by the proportion of Hcrt neurons that show Fos immunoreactivity, which further show a significant correlation with incubation of cocaine craving [101]. Excessive activation of Hcrt neurons may promote drug seeking through diverse projection targets as discussed above. In addition, Hcrt neurons may drive sleep instability after cocaine withdrawal, as they do in the aging brain [100].

In summary, repeated cocaine exposure and withdrawal can lead to reduced activities of MCH neurons and enhanced activities of Hcrt neurons in the LH, which may synergistically impair sleep and promote drug seeking (Fig. 1b).

Targeting sleep for SUDs

Through diverse neurotransmitter and modulator systems, all substance uses result in sleep changes, and persistent sleep disturbances are a common complaint following drug and alcohol withdrawal [3, 33, 34, 36, 226]. Sleep problems are thought to fuel drug relapse, and could indeed predict the outcome of drug and alcohol relapse in human patients [37, 227]. This may also be directly relevant to polysubstance uses, where patients may intend to self-medicate and resort to alcohol, marijuana, or opioids to mend impaired sleep from various drug withdrawal; or alternatively, use psychostimulants to promote wakefulness to counteract the withdrawal-induced excessive sleepiness during the day [228]. Polysubstance use is increasingly gaining attention [229,230], for which animal models are beginning to be developed (reviewed in [230]). In summary, targeting sleep may be applicable to both mono- and poly-SUDs, and may benefit drug relapse far beyond sleep improvement per se.

The Hcrt system represents a prominent candidate for this goal. Excessive activation of Hcrt system is commonly observed during withdrawal from a variety of substances, including cocaine, opioids, nicotine, and alcohol [1, 123, 124, 231], suggesting the possibility of targeting Hcrt system in general either for individual substance use or polysubstance use (e.g. cocaine + alcohol [232]). Recent studies focused on suvorexant, an FDA-approved, insomnia-treating dual Hcrt/orexin receptor OX1R/OX2R antagonist, show promising results for improving withdrawal symptoms in various preclinical models, raising the possibility that the reward circuit can be targeted through its shared components with the sleep-regulatory machinery [233]. A few other sleep-promoting medications have been proposed to be tested for treating SUDs, including topiramate, tiagabine, gaboxadol, vigabatrin, and lisuride [3]. The reported sleep effects are either a selective increase in slow-wave sleep time, or a reduction in REM sleep. While some have safety issues as GABA enhancers, none are shown to be particularly effective yet in reducing drug relapse [3]. More recently, modafinil, a mild psychostimulant acting on DA transporters, shows promising results in reducing cocaine relapse. In the study, modafinil is taken in the morning to increase alertness during the day, resulting in improved nighttime sleep. Patients with this modafinil regimen exhibit reduced rate of cocaine relapse, with the improvement positively correlated with the amount of N3 stage NREM sleep; and the positive correlation persists even after correction for modafinil treatment [234]. However, modafinil does not alter REM sleep architecture in this study (but see [235]).

Could REM sleep-MCH neuron activity benefit SUDs in general? The notion of enhancing REM sleep to achieve benefits may appear counter-intuitive, as many antidepressants that increase brain monoaminergic signaling suppress or delay REM sleep [236]. However, REM sleep may have both pro-relapse (i.e. sleep fragmentation) and anti-relapse components, and MCH neuron activity in REM sleep may represent a key to the anti-relapse effects [207]. MCH neurons project throughout the brain, and many limbic structures important for reward processing have moderate to high levels of MCH receptor expressions, including NAc, hippocampus, VTA, PFC, amygdala, and LH [237, 238]. Many of these regions also express Hcrt receptors and receive Hcrt fiber projections [239–242]. Thus, in the LH, MCH neurons inhibit Hcrt neurons through local inhibitory circuits, which has been detailed in sleep and energy expenditure regulations [243, 244]. In ventral hippocampus, MCH neuron terminal activities and MCH receptor signaling regulate impulsivity [245]; stimulating MCH neuron axon terminals suppresses pyramidal neuron firing, and REM sleep-active MCH neuron activity facilitates the erasure of contextual fear memory [171]. This may be in contrast with the role of Hcrt in facilitating hippocampus-dependent memory formation [246]. Other brain regions receiving MCH and Hcrt dual innervations include PFC, amygdala, NAc, VTA, among others [107, 237–242, 247, 248]. Therefore, it is possible that MCH neuron activities modulate Hcrt effects from both the source and the target regions, shaping the reward network through synaptic and cellular plasticity mechanisms in REM sleep (Fig. 1b). It will be meaningful to test whether REM sleep-enhancing medications, particularly those that can increase MCH neuron activity or MCH signaling during REM sleep and prolong REM sleep episodes without causing REM sleep fragmentation, alleviate withdrawal symptoms and reduce relapse to drug use.

Individual variations

Both sleep phenotypes and SUD development exhibit substantial individual variations. Sleep time, architecture, and waveforms are influenced by genetic and environmental factors [249]. SUD development is also complex, involving personality and physiological traits as well as individual experiences [250]. Genetic factors may strongly influence personality and physiological traits such as impulsivity, risk taking, and reward/stress responsivity, which tap into the vulnerability to SUDs [250]. Environmental factors include drug availability, peer influence, stress, lifestyle, and others, to which circadian misalignment and sleep disturbance could also be contributors [194]. Specifically, how may sleep interact with genetics to produce diverse individual variations in SUDs?

An inspiring example comes from rodent studies of sign-tracking versus goal-tracking behaviors. These are naturally occurring behaviors through Pavlovian learning, in which reward-associated cues acquire incentive salience in sign-tracking animals, whereas in goal-tracking ones the cues obtain predictive value that predominates over incentive salience [251]. It is thought that sign-trackers may be more vulnerable to addiction, as their reaction to reward-associated cues is more relying on subcortical drive rather than cortical executive control [252]. Mechanistically, sign-tracking relies on PVT Hcrt receptor signaling [126] and (presumed downstream) phasic DA release in the NAc core [58]. In response to reward-associated cues, both c-Fos activities in LH-PVT and PVT-NAc projection neurons and phasic DA release in the NAc are more prominent in sign-trackers than in goal trackers [58, 253]. Thus, comparing the top-down versus bottom-up dynamics, sign-trackers may have stronger bottom-up drive than goal trackers at baseline, which may render sign-trackers higher susceptibility to acute SD-induced weakening in top-down controls (reviewed in [61]).

Although sign-tracking may not be directly applied to humans, certain genetic polymorphisms affecting Hcrt or DA systems are associated with SUDs and/or SD susceptibility. For example, in a genome-wide association study, a Hcrt receptor 2 (HCRTR2) gene polymorphism shows association with nicotine dependence as well as the age for initial methamphetamine use [254]. HCRTR2 is also a candidate gene in sertraline- (a selective serotonin reuptake inhibitor antidepressant) associated insomnia in depressed patients [255]. Regarding the DA system, a variable number tandem repeat polymorphism in the human PER2 gene is associated with lower availability of striatal D2 receptors and cocaine abuse [256]. Moreover, a nine-repeat DA transporter (DAT) allele in human is linked to higher phasic DA activity and increased striatal responsivity to reward anticipation following SD, and a ten-repeat DAT allele is linked to lower phasic DA activity and increased striatal responsivity to punishment following SD [257]. These examples highlight the interaction between sleep and genotypes in producing individual variations.

Based on the sleep-reward interactions at circuit, cellular, and molecular levels discussed above, we suggest that future efforts be focused on the following three aspects to further dissect sleep-based individual variations in SUDs: (1) to explore the potential correlation between the sensitivity to acute SD in reward-associated tasks and susceptibility to drug-taking and seeking tests at the level of individual subjects; (2) to explore the potential correlation between key sleep parameters (e.g., REM sleep fragmentation and microstructures) and scores in SUD tests; (3) to explore the potential correlation between sleep-associated genetic polymorphisms and SUD phenotypes.

Conclusions and future directions

By summarizing recent results, this review discusses how sleep influences reward processing through circuit, cellular, and molecular substrates. This sleep-mediated regulation and the underlying mechanisms offer a conceptual possibility not only to improve the wellbeing of healthy populations and safe-guard adolescence development, but also develop treatment for SUDs. Considering sleep disruption both as a comorbidity to SUDs and a potential causal factor for drug relapse, manipulating sleep quality may serve as complementary therapeutics treating SUDs and mood disorders. Moreover, by integrating key sleep parameters, a sleep metrics may be formulated to predict individual susceptibility to SUDs and other reward-associated psychiatric diseases.

Author contributions

YH designed the framework; RG, DV, AA, and YH collected the literature and wrote the review. We agree to be accountable for all aspects of the work.

Funding

This work was supported by the National Institutes of Health under Award Numbers DA043826 (YH), DA046491 (YH), AA028145 (YH), DA046346 (YH).

Competing interests

The authors declare no competing interests.