Sleep regulates drug seeking and relapse – a perspective

Abstract

The recent decade has seen an increasing recognition of the importance of sleep in substance use disorders (SUD) in both research and clinical settings. Not only is sleep disturbance a common comorbidity in SUD, but it may provide a causal link and druggable targets for complementary treatment. The interpersonal variation in sleep also provides opportunities for developing biomarkers and individualized medicine. This review is focused on the key neurotransmitter and neuromodulator systems through which sleep loss may bias reward seeking, increasing the risks for initial drug exposure as well as relapse after drug withdrawal. The review summarizes sleep changes following acute or long-term drug exposure and withdrawal, and current understanding of sleep-mediated regulation of glutamatergic, dopaminergic, and peptidergic transmissions importantly indicated in SUD research.

Introduction

Over the past two decades, there is growing recognition of the critical role that sleep plays in the development and potential treatment of substance use disorders (SUD)^1–9. Sleep disturbances are not only highly prevalent as a consequence of repeated substance use^{5, 10}, but they often persist during abstinence and may contribute to relapse — thus, reinforcing the sleep impairment-addiction positive feedback loop^11–16. This highlights the importance of understanding how sleep-related mechanisms influence reward processing in the brain. In this review, we focus on key neurotransmitter and neuromodulator systems that dually regulate sleep and reward, in an attempt to understand how their dysregulation under sleep disturbances may bias aspects of reward seeking. We do not discuss the role of sleep in memory consolidation related to reward learning. Our goal is to provide a mechanistic basis for assessing vulnerability to initial drug use or relapse during abstinence; strategic targeting sleep physiology as an adjunctive therapeutic approach; and identifying sleep-based biomarkers that may inform risk stratification, prognosis, and personalized treatment. Together, these efforts aim to advance prevention strategies and precision medicine in SUD care.

The sleep impairment-addiction positive feedback loop

It is postulated that repeated drug exposure and subsequent withdrawal results in persistent impairment of sleep time and quality, which imposes negative physiological and psychological consequences and precipitate drug-seeking and relapse^11–16. The hypothesis is based on the observations that sleep disturbances commonly occur in humans both during and after cessation of drug and alcohol use (Supplemental Tables 1–3). Approximately 80% of individuals with SUD also experience sleep difficulties that negatively impact their quality of life^{17, 18}. In human adolescents, self-reported sleep problems were significantly associated with alcohol use disorder (AUD), after controlling for other associated factors such as negative emotionality¹⁹. The positive feedback loop also fits the allostasis hypothesis of addiction²⁰, where sleep impairment may contribute to the brain stress responses that drive the persistent deviation of reward set point. Consistent with this notion, protracted sleep impairment is postulated to contribute to relapse across SUD²¹. Indeed, sleep disturbance is the second most common reason for relapse given by individuals with opioid use disorder (OUD) or AUD²², and difficulty sleeping is one of the most common symptoms in OUD patients who drop out of naltrexone treatment²³. Moreover, poor sleep quality predicts relapse in AUD^{24, 25}, OUD^{26, 27} and cannabis use²⁸, and high rapid eye movement sleep (REMS) pressure is shown to predict relapse in AUD patients^{29, 30}. Admittedly, the human data is predominantly correlative and does not answer whether sleep disturbances and SUD share common predisposition factors rather than driven by each other. To answer potential causal relationships, animal models are developed which recapitulate drug-induced short-term and long-term sleep disturbances to certain extents (Supplemental Tables 4,5). Furthermore, acute or chronic sleep disturbances in rats or mice increase voluntary drug- or alcohol-seeking and intake^31–36. Hence, both arms of the positive feedback loop find some evidence in animal studies. Together, these results suggest that strategically treating sleep impairment in SUD may break the positive feedback loop and facilitate recovery from SUD.

An important implication of the positive feedback loop is whether sleep loss may also increase vulnerability to the initial reinforcing phase of drug use prior to withdrawal. Evidence from rodent studies provides some support for this possibility^33–37. This is especially concerning in the context of human adolescence – a developmental stage characterized by heightened reward responsiveness relative to cognitive control and greater susceptibility to substance use^38–46, and a chronic state of partial sleep deprivation exacerbated by early school start times misaligned with delayed biological rhythms^{47, 48}. The implications on adolescent onset of substance use are reviewed in detail elsewhere² and remains an area of active investigation.

Diverse strategies are deployed in humans and animals to study sleep-addiction relationships. While it is not feasible to directly assess sleep-addiction causal relationships in humans, it is, nonetheless, possible to assess sleep disturbance-induced changes in brain activities that accompany features associated with drug seeking risks; these include impulsive and risk-taking behaviors, reward wanting and liking, positive or negative affective states, reward evaluation, and habitual behaviors. Moreover, animal studies probe for the molecular, cellular, and circuit level mechanisms underlying sleep-mediated regulation of reward processing and drug seeking behaviors, and vice versa – mechanisms underlying drug-induced changes in sleep. Here, we attempt to integrate human and animal studies, highlighting recent advances in the mechanistic understanding of how sleep disturbances impact the reward circuit and related behaviors. We will focus on the dopaminergic, glutamatergic, and peptidergic systems.

Neural mechanisms contributing to sleep-reward interactions

Dopamine

The mesolimbic dopamine (DA) system is not only the backbone of brain reward circuitry^{49, 50} but also plays a key role in regulating sleep and wakefulness. Both hypo- and hyper-DA states impair sleep architecture as shown in mice⁵¹, and hypo- and hyper-DA states are each associated with sleep disturbances in human diseases^{52, 53}. Moreover, genetic variants of the DA transporter DAT1 and the receptor DRD2 have both been associated with habitual sleep duration, shown in a large European population⁵⁴.

Animal studies have revealed anatomical details in the dual regulation (i.e., hypo- and hyper-DA states), highlighting the midbrain ventral tegmental area (VTA) and the nucleus accumbens (NAc). Whereas their critical roles in regulating drug reward and motivation are thoroughly reviewed elsewhere^55–58, VTA DA neuron activity is also required for maintaining wakefulness, and VTA-to-NAc DAergic projection promotes wakefulness, as shown in mice using chemogenetic and optogenetic manipulations⁵⁹. Moreover, intra-NAc infusion of DA receptor D1/D3 agonists or D2/D3 receptor antagonists increase wakefulness and locomotion in rats⁶⁰. Furthermore, optogenetic activation of NAc D1 receptor-expressing neurons induces immediate transitions from non-REMS (NREMS) to wakefulness, and chemogenetic stimulation of these neurons prolongs arousal in mice⁶¹. Conversely, chemo- or optogenetic stimulation of adenosine A2A receptor-expressing (presumed D2 receptor-expressing) neurons in the core region of the NAc strongly induces slow-wave sleep in mice⁶². Thus, VTA, NAc, and the VTA-to-NAc DAergic projection powerfully regulate sleep and wakefulness.

On the other hand, human and animal literature suggests that sleep and sleep disturbances can impact DA signaling at multiple levels, affecting DA release, receptor signaling, and transporter function. For example, overnight sleep deprivation in adult men induces an increase in DA release in the striatum and thalamus, measured by positron emission tomography (PET)⁶³. This was recapitulated in male and female mice where acute sleep deprivation for 6 hours from the start of light phase increases spontaneous DA release in NAc^{64, 65}; and the increase is thought to mediate the affective state transitions in the mice toward elevated social and sexual behaviors and diminished depressive-like behaviors⁶⁴. Furthermore, at the DA receptor level, overnight sleep deprivation reduces D2/D3 receptors in the ventral striatum as measured by PET in healthy young men, which is associated with worsened cognitive performance and increased sleepiness⁶⁶. Relatedly, sleep duration in cocaine abusers statistically predict their striatal D2/D3R availability measured by PET⁶⁷. Thus, sleep disruptions may shift the D1-D2 balance favoring D1 receptor signaling in the ventral striatum, which may consequently bias toward impulsive reward-seeking^68–71. Additionally, regarding DA reuptake, a study in male rats found that normal periods of sleep promote DA uptake by the DA transporter DAT, measured by ex vivo fast scan cyclic voltammetry in the NAc core; the increase is accompanied by enhanced phosphorylation of DAT at threonine 53 – a site known for facilitating dopamine transport, and increased potency for cocaine to inhibit DAT⁶⁵. To summarize these findings in a hypothetical scenario, sleep loss induces an increase in DA release at the target regions, including the NAc, and decreases D2 receptor availability, which may promote reward wanting and seeking; meanwhile sleep deprivation reduces DAT reuptake, which may enhance the potency of the rewarding effect of drugs and exacerbate drug intake (Figure 1).

Figure 1. Hypotheses concerning the positive feedback loop between sleep and drug exposure.

The dopamine (DA) hypothesis posits that sleep loss induces an increase in DA release at the target regions, including the nucleus accumbens (NAc), and decreases D2 receptor availability, which may promote reward wanting and seeking; meanwhile sleep deprivation reduces dopamine transporter (DAT) reuptake, which may reduce the potency of the rewarding effect of drugs and exacerbate drug intake. The glutamate (Glu) hypothesis posits that sleep deprivation changes cortical dynamics and cortical-subcortical interactions, resulting in reduced top-down control, increased impulsivity, facilitation of habitual behaviors, bias toward positive affect and reward, and NAc-mediated relapse-like behaviors. Orexin/hypocretin (Orx/Hyr) signaling may be engaged both following sleep loss and during drug withdrawal to facilitate drug seeking. Rapid eye movement sleep may reduce drug craving after withdrawal through activating the melanin-concentrating hormone (MCH) system. ACC, anterior cingulate cortex; CP-AMPA receptors, calcium-permeable AMPA receptors; DMN, default mode network; LH, lateral hypothalamus; mPFC, medial prefrontal cortex; OFC, orbitofrontal cortex; Ox1R, orexin receptor 1; Ox2R, orexin receptor 2; PFC, prefrontal cortex; VTA, ventral tegmental area.

In line with this reasoning, examining individuals with genetic variants of the dopamine transporter or receptors may offer valuable insights into whether such variants confer differential sensitivity to sleep deprivation–induced changes in reward-seeking behaviors^{72, 73}.

Glutamate

The glutamatergic system is essential for cortical processing and cortical-subcortical coordination^{74, 75}. Drug-induced plasticity at glutamatergic synapses is widely observed in rodent models throughout the reward circuitry, including prefrontal cortex (PFC), NAc, and VTA etc. Many of these changes are persistent and thought to critically mediate drug-reward memory or withdrawal-induced drug craving and relapse-like behaviors^76–83.

Acute sleep deprivation induces large-scale changes at transcriptional, translational, posttranslational, and epigenetic levels in the rodent neocortex^84–96, and sometimes assessed in human blood or peripheral tissues^97–100. A large proportion of sleep deprivation-induced changes in the cortical transcriptome or synaptic proteome/phosphoproteome in rats and mice affect glutamatergic synapses^{84–87, 89–95}, with some persistent well beyond recovery sleep⁸⁴. Notably, sleep deprivation in mice abolishes 98% of all phosphorylation cycles in synaptoneurosomes⁹², and abolishes proteome (but not transcriptome) oscillations in synaptoneurosomes⁹¹. These changes suggest drastic and possibly long-lasting cellular and circuit level reorganizations. Whereas cellular level assessments are limited in humans, functional imaging studies have provided substantial insights – revealing sleep deprivation-induced loss of functional connectivity within frontal cortex¹⁰¹ and between frontal and posterior areas of the Default Mode Network (DMN)¹⁰², as well as an imbalance within cortical-striatal networks^{103, 104}. These changes are often thought to mediate impaired cognitive performance in humans and impaired emotion regulation across species^{69, 105–108}. Both aspects may influence natural reward or drug seeking behaviors (Figure 1).

Functional magnetic resonance imaging (fMRI) studies in humans have outlined cortical frameworks through which sleep loss may alter motivated behaviors, which can be related to detailed anatomical studies in drug-seeking animal models. For example, acute sleep deprivation reduces the functional connectivity within the DMN during rest and task performance in healthy young men¹⁰⁹, including changes in anterior cingulate cortex (ACC) interconnections¹¹⁰. Considering the role of ACC in top-down inhibitory control^111–113, it is not surprising that the sleep deprivation-induced changes in ACC interconnections are correlated with impaired impulsivity control, measured by a go/no-go task¹¹⁰ (although sleep deprivation-induced impulsivity change may be task-dependent¹¹⁴). Related to drug-seeking, recent studies in rats showed that the two major types of projecting pyramidal neurons, intratelencephalic (IT) and pyramidal tract (PT) neurons in ACC synergistically regulate drug seeking. Specifically, inhibition of IT neurons reduces the aversive properties of cocaine, and suppression of PT neurons enhances the rewarding properties of cocaine during a conditioned place preference test¹¹⁵. Therefore, it may be speculated that acute sleep deprivation-induced impairment in ACC connectivity may enhance the rewarding effects of drugs through both a reduction of PT activity to enhance reward and a reduction of IT activity to reduce aversion, thus synergistically biasing behavior towards drug seeking.

An additional insight comes from studies of goal-directed versus habitual behaviors. Using fMRI in young adult men and women, it was shown that one night of sleep deprivation results in deficient formation (but not expression) of goal-directed behaviors and an overreliance on habits. Moreover, goal-directed learning mainly recruits the ventromedial PFC, which shows reduced activity during goal-directed behavior following sleep deprivation¹¹⁶. Thus, sleep deprivation may increase drug seeking in drug-experienced individuals, in part through reducing ventromedial PFC activities and promoting habitual behaviors. To our knowledge, there are limited animal studies in this area.

A broader cortical-striatal network has been proposed as neural substrates that may contribute to additional sleep deprivation-induced decision-making bias. In a study of food choices in healthy men and women, it was shown that one night of total sleep deprivation increases the fat consumption compared to normal sleep controls, which is correlated with increased brain connectivity in the salience network from the dorsal ACC to bilateral putamen and bilateral anterior insula¹⁰³. Sleep deprivation may also bias decision making toward gains over losses. In healthy young men performing a gambling task, it was shown that 24 hours of sleep deprivation decreases activities in insula and orbitofrontal cortices in response to losses, whereas increasing activities in the NAc to gains. However, such regional activity differences are not reflected at the behavioral level¹⁰⁴. Thus, sleep deprivation may intensify expectation of gains while attenuating response to losses during risky decision making. We have not found related animal studies.

Sleep deprivation affects subcortical regions involved in reward seeking, though findings differ between human and rodent studies. For example, the NAc receives convergent glutamatergic inputs from PFC, amygdala, hippocampus, and thalamus, and serves as a limbic-motor interphase for regulating reward-seeking behaviors^117–119. Human fMRI imaging studies in young adults have consistently shown amplified reactivity to reward in parts of the mesolimbic system following sleep deprivation, including the NAc and VTA^{104, 120, 121} (but also see opposite changes in healthy adolescents¹²²). By contrast, in adult mice, acute sleep deprivation reduces the NAc activities during sucrose self-administration measured by population NAc Ca²⁺ activities¹²³. Consistent with this observation, slice electrophysiology in mice shows dampened glutamatergic inputs onto NAc following sleep deprivation. Specifically, both medial PFC-to-NAc and rostral amygdala-to-NAc glutamatergic transmissions show a reduction in release probability following acute sleep deprivation^{124, 125}. Moreover, reversing these effects by optogenetic stimulation of the axon terminals at either of the projections effectively counteracts sleep deprivation-induced increase in sucrose reward seeking^{124, 125}. It is not clear if the human versus mouse difference is due to differences in the types of tasks (cue-induced versus self-initiated reward seeking), age/developmental stages, or species. It is also not clear if the altered NAc activities following sleep deprivation may be due to interactions between glutamate, DA, and other neuromodulators. Nonetheless, these results suggest that acute sleep deprivation impairs NAc function. It remains to be determined whether these changes contribute to promoting drug seeking under sleep loss.

NAc glutamatergic transmission may also be sensitive to chronic sleep disturbance. For example, chronic sleep fragmentation in rats undergoing long-term withdrawal from repeated cocaine self-administration facilitates the accumulation of calcium-permeable (CP)-AMPA type glutamate receptors on NAc principal neuron synapses³¹. These receptors gradually and persistently accumulate at NAc principal neuron synapses over long-term withdrawal from repeated cocaine self-administration^{82, 83, 126–129}, which critically contribute to the progressive intensification of cue-induced cocaine seeking – termed incubation of cocaine craving^{130, 131}. Moreover, reducing CP-AMPA receptor transmission on NAc principal neurons either pharmacologically, electrophysiologically, or by sleep manipulations that consolidate REMS (discussed below), reduces incubation of cocaine craving in rats^{31, 82, 126, 132–134}. It remains to be determined whether such sleep-mediated regulation of NAc CP-AMPA receptor dynamics may similarly regulate drug-seeking behaviors of other addictive substances.

In summary, the glutamate hypothesis posits that sleep deprivation changes cortical dynamics and cortical-subcortical interactions, resulting in reduced top-down control, increased impulsivity, facilitation of habitual behaviors, bias toward positive affect and reward, and NAc-mediated relapse-like behaviors (Figure 1). While substantial progress has been made at analyzing both gene expression and circuit level changes, a deeper understanding at the cellular level is needed to contextualize and bridge the two, highlighting the need for future animal studies.

Orexin/hypocretin

Orexin (also called hypocretin) is a neuropeptide produced in the lateral hypothalamus (LH) and released throughout the brain that both promotes wakefulness^135–139 and facilitates natural and drug reward seeking^{37, 140–162}. Orexin neurons are most active during wakefulness and suppressed during sleep^163–166. Orexin’s pro-wakefulness mechanisms include stimulation of the locus coeruleus norepinephrinergic neurons, dorsal raphe serotoninergic neurons, tuberomammillary nucleus histaminergic neurons, basal forebrain cholinergic neurons, ventrolateral periaqueductal gray and lateral pontine tegmentum areas, and disynaptic inhibition of ventrolateral preoptic nucleus GABA/galanin sleeping-promoting neurons, as well as relaying inputs from paraventricular nucleus of hypothalamus containing stress information^167–170. In addition to regulating wakefulness and arousal, orexin neurons influence a broad network of the reward circuitry. They show increased activity in response to reward-associated cues, and orexin receptor signaling through Ox1R and Ox2R along the mesolimbic pathway has been implicated in regulating food, alcohol, cocaine, and opioid reward seeking and drug reinstatement after withdrawal^{140–144, 146–152, 154–163, 171–176}.

Following acute or chronic sleep disturbances, there is often elevated prepro-orexin synthesis, orexin-A release, and/or Ox1R and Ox2R gene expression as measured in the rat, dog, or human (recently reviewed in³). Moreover, functional studies in mice have shown that orexin signaling through Ox1R is required in male mice for the acute sleep deprivation-induced increase in cocaine conditioned place preference³⁷, whereas Ox2R signaling is required in female mice for acute sleep deprivation-induced increases in sucrose reward seeking¹⁷⁷. Thus, the orexin system can be recruited by sleep deprivation to regulate natural or drug reward.

Orexin signaling is also engaged during drug withdrawal as shown in rodents. For example, chronic morphine and antagonist-precipitated withdrawal increases the immediate early gene c-Fos and orexin gene expression in orexin neurons, whereas orexin knock-out mice have attenuated withdrawal symptoms¹⁷⁸. Similar benefits are also obtained in mice and rats using an Ox1R selective antagonist^{179, 180}. Moreover, in rats undergoing withdrawal from repeated cocaine self-administration using an intermittent access procedure, there is an increase in the number and activity of LH orexin neurons, which is accompanied by increased cue-induced cocaine seeking under protracted withdrawal. Furthermore, antagonizing Ox1R or knocking down orexin in the LH reduced cocaine reinstatement and motivation for cocaine, respectively¹⁴⁰. It is not known to what extent that the elevated orexin neuron number and/or function may be the cause or consequence of the persistent sleep impairment following either morphine or cocaine withdrawal. Nevertheless, these results from rats suggest that dampening orexin receptor signaling may be a promising strategy to reduce relapse after drug withdrawal (Figure 1; related human studies discussed below).

Melanin concentrating hormone

Melanin concentrating hormone (MCH) is another example of a sleep/reward dual-regulatory peptide. MCH-producing neurons reside in the LH and zona incerta and project throughout the brain^181–184, similar to orexin neurons. MCH receptors (MCHR1 and MCHR2 in human, MCHR1 in rodents) are widely expressed in the neocortex and limbic structures^185–188. MCH neurons are most active during REMS, modestly active during explorative behaviors in a familiar environment or during feeding, and quiescent during NREMS or quiet wakefulness^189–192. MCH peptide is co-released with GABA or glutamate^{182, 193} and powerfully regulates feeding¹⁹⁴ and REMS initiation and maintenance^{182, 184, 195}.

A series of studies in rats trained to self-administer cocaine revealed an intimate relationship between REMS and drug-seeking behavior. Whereas both NREMS and REMS show persistent fragmentation long after cessation of cocaine exposure, selectively prolonging REMS bouts in the light (inactive) phase reduced incubation of cocaine craving. This is thought to be at least in part mediated by a reduction in synaptic CP-AMPA receptors in the NAc principal neurons³¹. A similar relationship between REMS quality and relapse to drug use has yet to be thoroughly examined in human subjects, though REMS pressure in non-depressed AUD patients is shown to predict relapse to alcohol use at 3-month follow-up²⁹. Our knowledge gap regarding REMS quality and relapse to drug use in humans may partly arise from limitations in current tools to enhance REMS, as well as from an incomplete understanding of what constitutes REMS quality.

Subsequent work in rats focusing on REMS-active LH MCH neurons identified its role in reducing incubation of cocaine craving. Specifically, LH MCH neurons are predominantly active during prolonged REMS episodes, with much reduced population activities during short REMS episodes¹³⁴. Repeated cocaine self-administration followed by long-term withdrawal reduces MCH neuron membrane excitability and impairs glutamatergic transmission onto these neurons. Counteracting these cocaine-induced effects by stimulating MCH neurons chemo- or optogenetically during sleep after long-term withdrawal increases REMS¹⁹⁶ and reduces incubation of cocaine craving¹³⁴. Thus, the REMS-active MCH system may present a key functional component of REMS, a marker for REMS quality, and a target for sleep-based therapeutics (Figure 1).

It is important to highlight that the above-mentioned systems do not operate independently but are intimately interconnected with one another. For example, VTA DA neurons receive a large number of glutamatergic inputs from a cortical-subcortical continuum, including the PFC^{197, 198}. Conversely, DA release in the cortex and NAc also regulates glutamatergic transmission through DA receptor signaling^199–205. Moreover, both orexin and MCH directly impact DA neuron firing and interact with glutamatergic transmission onto DA neurons^{147, 150, 206, 207}. Finally, sleep deprivation regulates glutamatergic inputs onto LH orexin and MCH neurons through redistribution of the glial glutamate transporter 1 (GLT1) on astral glial cell processes, as shown in rats, such that it decreases perisomatic GLT1 apposition with orexin neurons and increases GLT1 apposition with MCH neurons. This reorganization is accompanied by enhanced presynaptic inhibition of excitatory transmission onto orexin neurons and reduced a form of slow EPSCs in MCH neurons²⁰⁸. Thus, DAergic, glutamatergic, and peptidergic transmissions interact within mesolimbic, cortical, and LH circuits, which together mediate the complex effects of sleep deprivation. Future studies combining brain region and projection target-specificity with single-cell omics analysis may be especially powerful in elucidating the molecular, cellular, and circuit mechanisms of sleep deprivation effects on drug seeking^{87, 94, 209}.

Sleep-based therapeutics and biomarkers

Both NREMS and REMS can be disrupted persistently after repeated drug use and withdrawal (Supplemental Tables 1–3). There is evidence suggesting that both may provide important measures for developing biomarkers and targets for therapeutics.

NREMS

Conventional NREMS-enhancing drugs often have various caveats. For example, many hypnotics promote NREMS in humans by effectively lengthening the shallow stage N2 at the expense of the deeper slow wave sleep (N3) (e.g. benzodiazepines^{210, 211}). Some others induce non-physiological slow-waves (e.g. sodium oxybate²¹²) or unnatural sleep architecture (e.g. lorazepam and tiagabine⁸), have safety concerns (e.g. GABA_A receptor agonist²¹⁰), or are associated with abuse liability (e.g. benzodiazepines²¹³). Nonetheless, it was noted that taking psychostimulants during the day can promote nighttime slow-wave activities in NREMS in abstinent cocaine users^{7, 9}. Indeed, subsequent clinical studies in an inpatient setting revealed that modafinil, a stimulant with low propensity for abuse, effectively increases slow-wave sleep time at night in abstinent cocaine users⁷, and the same dose in normal populations does not alter sleep architecture²¹⁴. Importantly, this effect is a statistical mediator of improved clinical outcomes, i.e. higher rate of %drug-negative urines and greater maximum number of consecutive days abstinent during the following outpatient phase. Furthermore, the change in stage N3 NREMS time before versus after modafinil treatment during inpatient treatment phase was correlated with cocaine abstinence measures in the same patients during subsequent outpatient phase⁷. Thus, the deepest NREMS and slow-wave activities may be important in the effective treatment of cocaine use disorder (CUD) and possibly other SUD as well.

REMS

Although pharmacological interventions for insomnia predominantly target NREMS, some drugs for psychiatric interventions effectively affect REMS, giving an opportunity to examine the relationship between REMS and SUD.

Many antidepressants are known REMS suppressants. For example, most selective serotonin receptor inhibitors, selective serotonin/norepinephrine reuptake inhibitors, and monoamine oxidase inhibitors delay REMS onset and reduce the overall REMS²¹⁵. Because comorbid depression is common among individuals with SUD, antidepressants are frequently prescribed as part of the treatment, which may help reduce the likelihood of relapse. However, the efficacy of antidepressants in preventing drug and alcohol relapse varies²¹⁶, raising the question whether suppressing REMS may be an undesired side effect.

Recently, OxR antagonists are gaining momentum as a treatment option for insomnia^217–220. The National Institute on Drug Abuse (NIDA) also included OxR antagonists as one of the “10 most wanted” high-priority medications for breaking the insomnia-addiction positive feedback loop for the treatment of OUD¹⁶. Compared to classical hypnotics that target GABA_A receptors, OxR antagonists do not have the side effects of gross suppression of brain activity and cognitive functions^221–223, nor do they have high abuse liability^224–228 (albeit abuse potential still exists^229–231). Different OxR antagonists produce varied effects on sleep. For example, Ox2R antagonists promote both REMS and NREMS in a proportional and physiologically balanced manner, as observed in rodents and humans²³². By contrast, dual orexin receptor antagonists (DORAs) strongly and selectively promote REMS without increasing deep (NREM) sleep^{233, 234}.

Two recent clinical studies examined the effect of suvorexant, a DORA, on subjects with CUD. In one study involving 20 subjects of mixed age, sex, and ethnicity in an outpatient setting, suvorexant or placebo was taken orally at bedtime daily for two weeks; sleep was assessed using actigraphy and PSQI, and inhibitory control and drug cue-reactivity were assessed using a saccade-based eye-tracking task. The results suggest that suvorexant improves sleep actigraphy, increases inhibitory control, and reduces craving²³⁵. However, the study was of limited scale and did not distinguish NREMS versus REMS. By contrast, another study tested the effect of suvorexant on cocaine self-administration in an inpatient setting involving 7 CUD subjects of mixed sex and ethnicity; the authors reported that a similar dose of suvorexant taken at bedtime increased intravenous self-administration of cocaine on the following day without altering other subjective and physiological effects of cocaine. Sleep was not measured in this study²³⁶. In addition to CUD, DORA was also proposed for treating OUD⁶. In a recent clinical trial on OUD patients, suvorexant prolonged total sleep time and decreased subjective opiate withdrawal rating, with no evidence of abuse potential²²⁸. Together, these results suggest that orexin receptor antagonists may promote recovery by improving sleep and inhibitory control, while reducing drug cue-reactivity and withdrawal symptoms. However, they may also enhance the reinforcing properties of certain addictive drugs. Future studies should carefully evaluate these specific aspects in larger populations and explore strategies to stratify SUD patients – potentially using sleep measures and behavioral assessments in relation to drug paraphernalia – to optimize treatment outcomes.

Sleep-based biomarkers

What may be considered candidates for sleep-based biomarkers for drug addiction? The modafinil human study described above identified stage N3 NREMS time and related slow-wave activities as potential candidates for biomarker development. On the other hand, studies in rats identify a relationship between REMS and drug seeking³¹, though an initial attempt using conventional REMS bout and waveform measures was not successful in identifying potential REMS biomarkers¹³⁴. Nonetheless, the results prompted the effort to use REMS-active MCH neuron activities in the search for biomarkers associated with cocaine craving. Although MCH neuron activities are difficult to access in vivo, further analysis of rat scalp EEG signals revealed a close correlation between an EEG theta-delta-theta peak (EEG_TDT) ratio measure and the MCH neuron population Ca²⁺ activity¹⁹². Moreover, it was shown that long-clusters of EEG_TDT ratio events over 24 hours, which represent prolonged MCH neuron population activities, are inversely correlated with cue-induced cocaine seeking in rats; further, in drug-naïve rats, greater EEG_TDT ratio events in the light (inactive) phase relative to dark (active) phase was correlated with less cocaine intake during subsequent self-administration training¹⁹². Thus, EEG_TDT ratio may offer a unique opportunity for developing REMS-based biomarkers for drug-seeking and relapse risks, as well as to evaluate the effectiveness of therapeutic interventions. One main challenge is to translate EEG_TDT ratio from rat to human.

Sleep versus circadian rhythm: to entangle or not

One caveat in examining sleep-mediated regulation of reward processing is its inherent entanglement with circadian rhythms, some of which may be introduced by the experimental design. In human studies, sleep and circadian influences can be separated to a certain extent – e.g. acute sleep deprivation occurs overnight, and behavioral tests are conducted on the following day in line with the circadian active phase. However, nighttime sleep deprivation often involves some level of light exposure, which may perturb circadian rhythms^{237, 238}. In rodent studies, acute sleep deprivation typically starts at the beginning of the light phase and lasts for a few hours. This design maximizes the accumulation of sleep pressure, thus the preferred way by convention. However, after sleep deprivation, the behavioral tests are conducted in the light phase, which is the circadian “inactive” phase for rodents. Indeed, most results of sleep-mediated regulation of natural reward or drug seeking are based on such or similar schedules^{34, 37, 124, 125, 239}, raising the question to what extent that sleep or circadian rhythm disruptions contribute to enhanced reward seeking. Whereas 12-hour sleep deprivation across the light phase is technically challenging in rodents, sleep deprivation and behavioral tests in the dark phase may be better suited for circadian considerations, though this approach may be less effective in building up the sleep pressure.

Further complicating the issue is that sleep is regulated by two interacting processes: homeostatic pressure and circadian rhythms^{240, 241}. Moreover, sleep deprivation per se also impacts circadian rhythms by altering circadian gene expression. For example, in healthy male shift workers, overnight wakefulness induces epigenetic and transcriptional changes to core circadian clock genes measured in adipose tissue and skeletal muscle⁹⁷. In animal models, changes in circadian gene expression are also characterized in brain tissues. In the medial PFC of young mice, 10 out of 12 circadian core clock genes are altered following a few hours of sleep deprivation in the light phase²⁴². Similarly, in adult male mice a single acute sleep deprivation in the light phase induces long-lasting suppression of most core clock gene oscillations in the cortex²⁴³. Moreover, medial PFC molecular clock is shown to mediate the effects of sleep deprivation on depressive-like behaviors in mice²⁴⁴. Thus, aside from the timing issue discussed above, sleep deprivation per se may engage circadian mechanisms to impact subsequent behaviors.

How might changes in circadian gene expression relate to drug seeking? Accumulating evidence from both human and animal studies suggests a bidirectional relationship. On the one hand, repeated drug exposure not only changes sleep homeostasis, but also extensively alters circadian rhythms – measured at both transcriptional and protein levels in humans and animals^245–248. On the other hand, animal studies reveal that an impaired circadian rhythm either at the systemic level, in selective reward-regulatory brain regions, and/or in specific cell types can lead to changes in reward seeking behaviors including cocaine self-administration and conditioned place preference^249–252. For example, either environmental or genetic disruption of circadian rhythms leads to increased drug reward and motivation for drugs in animal models²⁵². Although such manipulations are not feasible in humans, studies of teens and young adults reveal that having an evening chronotype and greater circadian misalignment with the environment (i.e. social jet lag) is associated with increased drug and alcohol use, along with other risky behaviors (reviewed in²). Thus, restoring circadian rhythms – via sleep manipulations or other approaches – may help mitigate drug seeking behaviors.

Often, sleep-targeting therapeutics have a built-in circadian component. For example, the above-mentioned modafinil treatment in humans abstinent from cocaine use is directed to be given in the morning^{7, 9}, which is expected to be mildly stimulating early in the day, resulting in better sleep in the evening. From a circadian point of view, this treatment may effectively increase the circadian amplitude. Similarly, in rats undergoing long-term cocaine withdrawal, dark phase sleep restriction followed by light phase sleep rebound – which effectively increased REMS bouts in the light phase³¹ – may also enhance circadian amplitude. Thus, the benefits could result from both an improvement in sleep and an enhancement of circadian rhythms. From a basic science perspective, a future challenge is to disentangle sleep versus circadian mechanisms. From a therapeutic perspective, synergistically targeting sleep and circadian rhythm restoration may be especially beneficial to reduce drug seeking and relapse.

Conclusions

Here, we integrate current findings from human and animal studies on sleep-mediated regulation of the reward circuitry to propose the following hypothesis: persistent sleep disturbances following exposure to addictive drugs disrupt glutamatergic, dopaminergic, and peptidergic systems within the reward circuitry, thereby altering key aspects of reward processing, increasing vulnerability to drug seeking and relapse and perpetuating the addiction-sleep cycle (Figure 1). Moreover, targeting neural circuits that jointly regulate sleep and reward may offer therapeutic potential – interrupting the vicious cycle, improving sleep, and facilitating recovery.

We acknowledge limitations to this synthesis. Human versus animal studies often employ distinct methodologies to interrogate the central nervous system (e.g. non-invasive versus invasive), or apply similar techniques to different tissues (e.g., peripheral versus central), each presenting unique interpretive challenges. As such, it often takes a leap of faith to connect the dots for a cohesive understanding of the framework. Additionally, we would like to highlight the following priorities for future research: 1) development of animal models to investigate the mechanisms underlying persistent sleep disturbances following chronic drug exposure; 2) elucidation of cellular and circuit mechanisms through which acute or chronic sleep disruption influences reward seeking – findings that may contextualize the human fMRI results; 3) studies in both humans and animal models that clearly dissociate sleep versus circadian rhythm-mediated effects on reward regulation; 4) validation of selective NREMS and REMS features as candidate biomarkers for SUD risks; 5) assessment of the efficacy of sleep-targeting therapeutics in preventing relapse in SUD. The expected outcomes will not only inform treatment strategies for SUD but also lay the foundation for exploring interactions between sleep and other psychiatric or neurodegenerative disorders.