Interaction between cocaine use and sleep behavior: A comprehensive review of cocaine's disrupting influence on sleep behavior and sleep disruptions influence on reward seeking

Dopamine, orexin (hypocretin), and adenosine systems have dual roles in reward and sleep/arousal suggesting possible mechanisms whereby drugs of abuse may influence both reward and sleep/arousal. While considerable variability exists across studies, drugs of abuse such as cocaine induce an acute sleep loss followed by an immediate recovery pattern that is consistent with a normal response to loss of sleep. Under more chronic cocaine exposure conditions, an abnormal recovery pattern is expressed that includes a retention of sleep disturbance under withdrawal and into abstinence conditions. Conversely, experimentally induced sleep disturbance can increase cocaine seeking. Thus, complementary, sleep-related therapeutic approaches may deserve further consideration along with development of non-human models to better characterize sleep disturbance-reward seeking interactions across drug experience.


General introduction
As a psychostimulant with strong reinforcing properties (Ritz et al., 1987), cocaine acutely increases behavioral arousal (Zubrycki et al., 1990;Kiyatkin and Smirnov, 2010). Accordingly, cocaine can induce sleep disruption if taken near the habitual bedtime (evening in humans and other primates, morning in most rodents). More chronic cocaine use has also been associated with sleep disturbance in people with cocaine use disorder (for review, Gawin, 1991;Morgan and Malison, 2007;Valladares and Irwin, 2007) with sleep disturbance persisting into withdrawal and abstinence suggesting that cocaine use may cause alterations in sleep/waking control beyond the direct pharmacological action of cocaine-induced psychostimulation. The mechanism by which cocaine alters sleep/waking control is yet unknown, but dopamine, orexin (hypocretin), and adenosine systems are well-positioned based on their dual roles in reward and sleep/arousal behavior. This review will begin with a description of the role of these neuromodulators in reward and sleep/arousal behavior. Next, the effect of acute or limited cocaine on sleep will be discussed followed by the effect of chronic cocaine on sleep. Most of the acute cocaine research has featured non-human animal models, while the chronic cocaine research has largely focused on cocaine-dependent human subjects. The ability of sleep disruption to modulate cocaine reward-related behavior will also be covered since long term changes in sleep behavior may influence reward responding thereby conferring possible vulnerability to addiction/relapse. Finally, the possibility of treating cocaine use disorder through sleep-based pharmaceuticals will be discussed followed by limitations and possible future directions.
Our goal was to include previous research articles describing sleeprelated changes (or the absence thereof) following cocaine exposure in both human and non-human models. These papers were collected via search of the Pubmed database using keywords 'cocaine and sleep', along with relevant papers cited within articles found by keyword search. Articles were excluded if they did not have an explicit sleep component, did not include cocaine exposure or cocaine-experienced participants, or were not written in English. Case reports, Letter to the Editor-style reports, and book chapters were excluded. For rewardrelated research, articles were excluded if they did not have an explicit reward component (for example, research describing sleep deprivation-related changes in locomotor sensitization are not reviewed here).
2. Dopamine, orexin, and adenosine systems are involved in cocaine reward and modulated by cocaine exposure Dopamine (DA) is a monoaminergic neurotransmitter with cell bodies primarily in the Ventral Tegmental Area (VTA) and Substantia Nigra (SN) of which the VTA located neurons have been heavily implicated in reward behavior, including for stimulant drugs of abuse such as cocaine (Baik, 2013;Volkow and Morales, 2015) with the mesocorticolimbic DA system hypothesized to be a common pathway for drug reinforcement (Wise, 1980;Koob, 1992). Cocaine, a psychostimulant with high abuse potential due to its strong reinforcing properties (Johanson et al., 1976;Bozarth and Wise, 1985), blocks monoaminergic transporters of which dopamine transporter (DAT) blockade is predominantly responsible for these reinforcing properties (Ritz et al., 1987). A full review of the role of DA in cocaine reward is beyond the scope of the current review, but has been previously reviewed (Kuhar et al., 1991;Kiyatkin, 1994;Baik, 2013;Volkow and Morales, 2015;Francis et al., 2019) as have alterations in the DA system in response to chronic cocaine exposure (for review, Porrino et al., 2004;Luscher, 2016;Wolf, 2016;Nestler and Luscher, 2019).
Orexin (also known as hypocretin) is a peptide expressed in a subset of glutamatergic neurons within the lateral hypothalamus and perifornical areas. These neurons send widespread projections throughout the brain, targeting various neuronal populations including those involved in motivation (Gonzalez et al., 2012). In rodents, orexin administration increases cocaine self-administration (Espana et al., 2011), while orexin antagonists decrease cocaine self-administration (Hollander et al., 2012;Gentile et al., 2018). Stimulation of orexin neurons is sufficient to induce conditioned place preference (Taslimi et al., 2011 [rodents]), while dual orexin receptor antagonism reduces expression of cocaine conditioned place preference (Steiner et al., 2013 [rodents]). Conversely, loss of orexin through genetic mutation in rodents prevents the acquisition of conditioned place preference to cocaine (Shaw et al., 2017), reduces cocaine self-administration at high doses (Steiner et al., 2018), and reduces DA response to cocaine (Shaw et al., 2017), as does orexin antagonism (Prince et al., 2015). Similar to the ability of chronic cocaine to modulate the DA system, chronic cocaine also modulates orexin signaling with changes persisting days to weeks following termination of exposure. Aston-Jones and colleagues, using a rodent model, reported that cocaine self-administration on an intermittent schedule increases orexin neuronal activity as determined by the number of neurons expressing orexin and co-expressing an immediate early gene with increases in response to re-exposure to the drug context lasting at least 5 months into withdrawal (James et al., 2019). Repeated cocaine exposure also increases experience-dependent potentiation of excitatory synapses on orexin neurons and facilitates induction of long term potentiation in orexin neurons, an effect that persists several days into withdrawal (Rao et al., 2013 [rodents]). Furthermore, stimuli previously conditioned to mark the availability of cocaine retain the ability to increase orexin neuronal activity at least three weeks following conditioning (Martin-Fardon et al., 2018 [rodents]).
Adenosine is a purine neuromodulator and a component of ATP such that adenosine levels are influenced by metabolic activity, with adenosine, acting through adenosine A1 receptors, providing feedback inhibition to reduce subsequent neuronal activity (for review, Fredholm and Dunwiddie, 1988;Greene and Haas, 1991). All drugs of abuse directly or indirectly modulate adenosine signaling (Hack and Christie, 2003) with cocaine increasing adenosine tone (Bonci and Williams, 1996;Fiorillo and Williams, 2000). In turn, adenosine is thought to be able to increase DA release from the VTA by facilitating burst pause firing of glutamate afferents via inhibition of mGluR inhibitory postsynaptic potentials (Fiorillo and Williams, 2000). The effects of adenosine modulators on cocaine seeking behavior are quite complicated as multiple factors can influence outcomes. In rodents, acute caffeine (mixed A1 and A2 receptor antagonist) exposure increases responding for cocaine, while chronic caffeine exposure decreases responding (Kuzmin et al., 2000). Interestingly, cocaine dependent humans show lower caffeine use prevalence than the general public and caffeine use is associated with less frequent cocaine use (Budney et al., 1993), suggestive of substitution. Furthermore, caffeine and cocaine generalize to each other; in rodents trained to discriminate caffeine from saline, cocaine presentation results in caffeine-appropriate responding (Holtzman, 1986), while in cocaine experienced humans trained to discriminate cocaine from placebo, caffeine presentation results in cocaineappropriate responding (Oliveto et al., 1998). Other non-caffeine A1 and A2 antagonists also partially substitute for cocaine in a discrimination task in rodents (Justinova et al., 2003). Another complicating factor is that A1 (Gi-coupled) and A2 (Gs-coupled) receptor modulators can have similar behavioral outcomes despite opposing effects on intracellular signaling as in the case of A1 and A2 agonists decreasing responding during extinction of cocaine self-administration (O'Neill et al., 2014 [rodents]). Furthermore, adenosine receptors directly influence DA signaling through the formation of heteromers (A1/D1, A2/ D2, and less commonly, A1/D2; for review, Franco et al., 2007). As with DA and orexin systems, chronic cocaine also induces plastic changes in the adenosine system in terms of increases in adenosine A1 receptor expression in the nucleus accumbens (Toda et al., 2003) and decreases in adenosine-mediated inhibition of presynaptic glutamate release due to increases in adenosine uptake (Manzoni et al., 1998).

Dopamine, orexin, and adenosine systems influence sleep behavior
While the role of DA in reward behavior has been well-characterized, until recently a potential role of DA in sleep/waking behavior received much less attention. Dopaminergic neurons do not change firing rate across state (Trulson and Preussler, 1984), though the pattern of firing does change as demonstrated by prominent burst firing during REM sleep (Dahan et al., 2007) which would be expected to increase DA release. Additionally, there are brain region specific changes in DA level across states (Lena et al., 2005), increases in DA and metabolites in various brain regions following sleep deprivation (Asikainen et al., 1995;Farooqui et al., 1996;Zant et al., 2011), and reports that DA receptor agonists and antagonists influence sleep/waking state (for review, Monti and Monti, 2007). More recently, optogenetic and chemogenetic stimulation of dopaminergic VTA neurons was found to strongly promote arousal (Eban-Rothschild et al., 2016;Oishi et al., 2017). Classic and relatively new stimulants, such as methamphetamine and Modafinil respectivley, require DAT to promote waking (Wisor et al., 2001). Furthermore, subjective sleep duration is shorter in cocaine users and D2/D3 receptor availability is reduced with a mediation analysis suggesting that sleep duration mediates the relationship between cocaine use and D2/D3 receptor availability (Wiers et al., 2016). In sum, cocaine-induced modulation of the DA system could underlie changes in sleep disturbance, particularly in regards to the increase in waking as a direct pharmacological effect, but also through alterations in the DA system in response to chronic cocaine exposure (for review, Porrino et al., 2004;Luscher, 2016).
In addition to orexin neuronal projections to motivation-related targets, orexin neurons also activate neuronal populations involved in arousal/waking including, noradrenergic neurons in the locus coeruleus (Hagan et al., 1999), histaminergic neurons in the tuberomammillary nucleus (Eriksson et al., 2001), and cholinergic neurons in the ascending reticular activating system (Eggermann et al., 2001;Burlet et al., 2002). Orexin neuronal activity is high during waking (Lee et al., 2005) and increases further during sleep deprivation (Estabrooke et al., 2001). Similarly, orexin levels are high during typical waking and increase further during sleep deprivation (Yoshida et al., 2001). Furthermore, a subset of orexin neurons show increased excitability during sleep deprivation which may serve to promote continued arousal despite high sleep need (Briggs et al., 2019). Orexin administration increases waking (Akanmu and Honda, 2005), while orexin antagonist administration increases sleep (Winrow et al., 2011). Loss of orexin, through the disorder narcolepsy in humans or genetic modification in mice, causes sleep fragmentation (Chemelli et al., 1999;Hungs and Mignot, 2001) due to an inability to maintain waking. In sum, the orexin system has received widespread attention as a potential hub intermediating motivation and arousal (for review, Tyree et al., 2018).
As with orexin, adenosine is also modulated by cocaine exposure and, unlike orexin, is heavily implicated in the homeostatic control of sleep Porkka-Heiskanen and Kalinchuk, 2011) which is the portion of sleep need that is influenced by prior waking time. Adenosine levels build up during waking (Porkka-Heiskanen et al., 1997); acting through Gi-coupled adenosine A1 receptors, adenosine inhibits wake active neurons in various brain regions (Rainnie et al., 1994;Alam et al., 1999;Thakkar et al., 2003;Liu and Gao, 2007) and disinhibits sleep active neurons (Chamberlin et al., 2003;Morairty et al., 2004), while acting through Gs-coupled adenosine A2 receptors adenosine excites sleep active neurons in the hypothalamus (Gallopin et al., 2005) and disinhibits wake active neurons in the tuberomammillary nucleus (Hong et al., 2005), with all these actions promoting sleep. Homeostatic sleep rebound is absent in mutant mice that lack adenosine A1 receptors in forebrain glutamatergic neurons . Furthermore, increases in adenosine tone through reductions in adenosine kinase increase delta power under baseline conditions and slow the decay of delta power during NREM sleep indicative of enhanced sleep drive (Bjorness et al., 2016), while reductions in adenosine tone caused by overexpression of the cytoplasmic form of adenosine kinase reduce sleep and the homeostatic response to sleep deprivation (Palchykova et al., 2010). Interestingly, individual differences in the adenosine system of humans may be more pronounced than with other systems based on polymorphisms in adenosine A2a receptor (Retey et al., 2007;Erblang et al., 2019) and adenosine deaminase (Retey et al., 2005;Mazzotti et al., 2012) genes, which have functional consequences on sensitivity to caffeine and sleep loss. In sum, the adenosine system is positioned to influence and be influenced by prolonged waking and cocaine exposure, though interpretation of effects may be more complicated as compared to DA and orexin mechanisms.

Cocaine-induced alterations of sleep: limited exposure
Acute cocaine exposure increases sleep latency (SL) and REM sleep latency (REML) compared to vehicle or placebo in rodents (Hill et al., 1977;Bjorness and Greene, 2018;Dokkedal-Silva et al., 2020) and increases waking (Hill et al., 1977;Gruner et al., 2009) in a dosedependent manner (Knapp et al., 2007;Bjorness and Greene, 2018 [rodents]) with higher doses increasing waking over longer durations, as expected, and cocaine and caffeine combinations increasing waking to a greater degree than cocaine alone (Schwarzkopf et al., 2018 [rodents]). Increased waking reflects decreases in both SWS and REM, though REM disruptions are prolonged compared to SWS (Knapp et al., 2007;Bjorness and Greene, 2018). Furthermore, this increase in waking is due to an increase in episode duration, while the decrease in SWS and REM is due to decreases in both episode duration and number (Bjorness and Greene, 2018). The immediate cocaine-induced increase in waking is followed by a delayed rebound in sleep resulting in modest wake surplus at 22 h post-injection (Gruner et al., 2009) and full recovery within the 24 h period (Bjorness and Greene, 2018). In rhesus monkeys, a primate species with consolidated night-time sleep patterns like that of humans, cocaine early in the day reduces sleep efficiency during the subsequent night, but only in response to the preferred dose (based on relative intake versus food reward) in that lower and higher doses do not significantly modulate sleep efficiency (Brutcher and Nader, 2013). Overall, these results indicate a direct pharmacological wake promotion that is non-linearly modulated by dose and subjective effects.
As determined by EEG spectral analysis, delta activity decreases and gamma (30-50 Hz) activity increases immediately following cocaine administration (Bjorness and Greene, 2018 [rodents]) with subsequent delta rebound over the next several hours. Protracted rebound increases in slow band power occur following binge cocaine (15 mg/kg delivered once per hour for 3 consecutive hours) with delta remaining elevated at 24 h post-injection (Urbano et al., 2009 [rodents]) suggesting that while sleep time recovers within a 24 h period following cocaine, subtle alterations in neuronal activity during sleep states may show longer lasting alterations.
In humans, an evening of recreational cocaine use increases REML and decreases REM time compared to a pre-cocaine baseline night with subsequent rebound decreases in REML and increases in REM during recovery, while NREM and SWS are unchanged (Watson et al., 1992). Cocaine administration in non-users (severely depressed individuals) also decreases REM and total sleep time (TST), both of which subsequently rebound (Post et al., 1974a). These results are consistent with the pattern of sleep disturbance followed by recovery pattern observed following limited cocaine exposure in non-humans.

Cocaine-induced alterations of sleep: chronic exposure and withdrawal
In rodents, cocaine chronic exposure (20 mg/kg for 5d, 30 mg/kg for the next 5d), the pattern of increased waking immediately following cocaine administration with subsequent rebound resulting in unchanged sleep/waking time over the 24 h period persists (Dugovic et al., 1992). Conversely, across repeated cocaine self-administration exposures, sleep efficiency reductions moderate indicative of tolerance (Cortes et al., 2016 [rhesus macaque]). Discrepancy between these outcomes may be due to methodological differences, including route of administration (IP vs IV), volition (experimenter-administration vs self-administration), pattern (bolus vs repeated small volumes), relative timing (within subjective inactive phase vs within subjective active phase) along with species difference (rat vs rhesus macaque). Upon withdrawal from experimenter-delivered chronic cocaine, in which the term withdrawal refers to the termination of daily exposure and not a specific physiological state, increased waking in the light phase and increased SWS and REM in the dark phase continues for several days (Dugovic et al., 1992), while increased waking is observed following withdrawal from a combined cocaine plus caffeine exposure (Rivero-Echeto et al., 2021).
Persistent alterations in sleep/waking behavior following withdrawal from chronic self-administered cocaine (self-administration during the dark phase for 6d followed by weekly polysomnography at withdrawal days 1, 7, 14, 21; Chen et al., 2015 [rodents]) have also been demonstrated. Specifically, delayed decreases in NREM (withdrawal days 14, 21) and protracted decreases in REM (withdrawal days 1, 7, 14, 21) are seen (Chen et al., 2015) with shorter NREM and REM episode durations and fewer NREM to REM transitions alongside more NREM to waking transitions. Furthermore, spectral power including the delta band is unchanged at a time point when NREM and REM are decreased (withdrawal day 21, Chen et al., 2015) indicating that decreased sleep time is not counteracted by increased sleep intensity and thereby suggesting a net loss of absolute sleep activity. Thus, duration of cocaine exposure influences the pattern of sleep disturbance with more limited exposure resulting in a sleep disturbance-recovery pattern compared to protracted sleep disturbance (without demonstration of recovery within the recording period) following longer cocaine exposure.
The effect of cocaine use or withdrawal/abstinence from cocaine use on sleep/waking behavior in humans has been investigated for decades, however, considerable variability in the design of experiments alongside variability in outcomes impedes conclusive interpretation of the literature as a whole. The majority of studies have found some measure of sleep disturbance, though the nature of the disturbance varies. In lieu of describing these disparate findings, details are summarized in a table. See Table 1 for a brief description of the experimental design with notable features, outcome measures used, population of cocaine use and control group details, and notable outcomes. This table includes published articles of original research or meta-analysis featuring human subjects with sleep-related outcomes under cocaine use/withdrawal/ abstinence.
A few brief notes on terminology: 1) several groups divide abstinence into early and late phases with early abstinence generally <2 weeks post cocaine and late abstinence generally 2-4 weeks post cocaine, 2) the terms withdrawal and abstinence are sometimes distinguished and other times used interchangeably with additional terms such as subacute used occasionally, 3) conventions of NREM staging have changed over time from NREM1 (Stage 1), NREM2 (Stage 2), NREM3 (Stage 3), NREM4 (Stage 4) with NREM3,4 (Stage 3-4) used synonymously with SWS (Rechtschaffen and Kales, 1968) to N1, N2, N3 with N3 representing SWS (Iber et al., 2007); the authors' terms are kept. It is important to note that the term 'withdrawal' was generally not defined on the basis of DSM-defined withdrawal symptoms such as dysphoric mood, fatigue, vivid, unpleasant dreams, insomnia or hypersomnia, increased appetite, or psychomotor retardation or agitation (DSM-V, American Psychiatric Association, 2013) but rather on a time basis following the last cocaine exposure, which itself was inconsistently applied across studies. Additionally, about 20% of long term cocaine users do not meet DSM-criteria for cocaine withdrawal on the basis of symptoms present (Sofuoglu et al., 2003) indicating that cocaine withdrawal is not consistently observed. There are several previous reviews of sleep/waking outcomes following cocaine use in humans (Gawin, 1991;Morgan and Malison, 2007;Valladares and Irwin, 2007) along with reviews featuring cocaine amongst other drugs of abuse (Schierenbeck et al., 2008;Garcia and Salloum, 2015;Angarita et al., 2016;Gordon, 2019).
In summary, sleep disruptions following cocaine use, withdrawal, and abstinence in humans have been observed along with a small portion of reports indicating a lack of disturbed sleep compared to controls. Methodological reasons may account for the lack of consistency in specific sleep outcomes, including: • cocaine timing; varied (morning, daytime, evening) relative to nighttime sleep/waking measurements • polydrug issues; some reports explicitly mention that subjects regularly used other drugs, most commonly nicotine, alcohol, and marijuana, that themselves influence objective and subjective sleep outcomes (Schierenbeck et al., 2008;Garcia and Salloum, 2015;Angarita et al., 2016) • caffeine exposure; limited in some inpatient studies and presumably not limited elsewhere. Caffeine is a mixed adenosine A1/A2 receptor antagonist commonly consumed for psychostimulation, caffeine intake is variable across individuals (for example, range of 0-1180 mg/d in a group of individuals recruited for a caffeine-related study [Garrett and Griffiths, 1998]), sleep and delta power response to caffeine is modulated by a polymorphism in DAT1 (Holst et al., 2014) and a polymorphism in the adenosine A2 receptor (Retey et al., 2007), habitual caffeine consumption is influenced by a polymorphism in the adenosine 2A receptor (Cornelis et al., 2007), and caffeine is a common adulterant in street cocaine (Kudlacek et al., 2017) • napping; discouraged or prohibited in many studies to reduce daytime sleep, but napping was occasionally reported (Weddington et al., 1990) even when explicitly discouraged (Kowatch et al., 1992;Pace-Schott et al., 2005) • inpatient vs outpatient design; differing advantages/disadvantages in oversight and experimental control vs familiarity of sleeping environment and naturalistic setting • route of use; frequent coughing throughout the night was noted as a possible source of disturbed sleep during the early part of a study featuring individuals whose typical intake was by smoking crack (Kowatch et al., 1992) • controls (excluding binge cocaine use portion in which healthy control use is precluded); some studies feature healthy controls, some used healthy control data previously collected, some used normative data acquired from published materials, and others had no healthy control comparison Finally, while circadian rhythm and drug use interactions are outside the scope of the current review (see Kosobud et al., 2007;Webb et al., 2015;DePoy et al., 2017 for previous reviews) it is important to note that changes in the circadian timing of sleep may contribute to sleep disturbance. Delayed bed times during active cocaine use periods and early withdrawal have been reported (Kowatch et al., 1992) as have later bed times across abstinence (Morgan et al., 2006) indicative of phase delay which may partially drive sleep disturbance given the relative night-time consolidation of typical sleep periods in adult humans alongside the commonality of morning start times of work and school in the United States. In sum, alterations in circadian rhythm may drive or exacerbate sleep disturbance following cocaine exposure.

Sleep loss mediated alterations of cocaine reward
Sleep disruption has been indirectly linked to relapse behavior in humans in which an increase in N3 during abstinence is associated with percentage of urine screens negative for cocaine metabolites and maximum consecutive days abstinent (Angarita et al., 2014a). Direct tests of the ability of sleep disturbance to influence cocaine reward have been mixed suggesting that sleep disturbance can influence cocaine reward, but does not necessarily do so for every component of reward or in all individuals.
Using a conditioned place preference paradigm in which the rewarding properties of cocaine are inferred by time spent in a cocainepaired context relative to time spent in a neutral (saline-paired) context, sleep deprivation prior to cocaine-conditioning trials or prior to the postconditioning test enhances preference to the cocaine-paired location for a moderate dose of cocaine (8 mg/kg) and induces preference to a low dose of cocaine (3 mg/kg) when deprivation occurs prior to cocaine conditioning trials (Bjorness and Greene, 2020 [rodents]). Sleep deprivation-induced preference to a low dose of cocaine is consistent with reports that sleep deprivation induces preference to a low dose of amphetamine (Berro et al., 2018 [rodents]) or methylphenidate (Roehrs et al., 1999 [humans]) and that one night of sleep deprivation increases the perception of the strength of cocaine and reverses sleep deprivationinduced vigilance impairments in humans (Fischman and Schuster, 1980), effects that are lost with two nights of sleep deprivation.
Several experiments have used self-administration, in which nonhumans lever-press or nose-poke to receive cocaine rewards, to assess the effect of sleep disturbance on cocaine reward. Acute sleep deprivation of 4-8 h reduces reinstatement (a model for relapse [for review, Shaham et al., 2003]) in a population of rats that self-administer relatively low amounts of cocaine, but not high amounts of cocaine (Puhl et al., 2009; rats are divided into high or low cocaine taking based on their self-administrative behavior relative to the entire population). Furthermore, sleep deprivation reduces inter-infusion interval and increases active/inactive lever press ratio in low cocaine-taking animals (Puhl et al., 2009) indicating that acute sleep deprivation can subtly influence cocaine seeking behavior. Acute sleep deprivation does not influence motivation to seek cocaine as determined by a progressive Non-treatment group design: accommodation (cocaine tolerance test plus off days) -binge cocaine or placebo (midday for 3d) -abstinenceplacebo or binge cocaineabstinence (some subjects accommodationbinge cocaineearly abstinenceplacebolate abstinence, others accommodation placebo/early abstinencelate abstinencebinge cocaineearly abstinence). PSG 16× over 23d which were binned into accommodation, binge days, and 5 bins of abstinence.
Cocaine dose was 32 mg/70 kg and was self-administered by IV route. Binge duration was 2 h (5 min lockout period between injections with safety lockouts based on heart rate and blood pressure as needed). Placebo was saline. Non-treatment group: REM and TST late in abstinence negatively correlated with fraction of available cocaine used, while those whose SWS increased from early to late abstinence showed a trend towards less cocaine self-administered.
In treatment-seeking group: REM time late in abstinence positively correlated with negative urine screens and max consecutive days abstinent, percent change SWS from early to late abstinence positively correlated with percentage of negative urine screens, and those whose SWS time increased from early to late abstinence had higher percentage of negative urine screens and higher max consecutive abstinence days than those whose SWS did not increase. Angarita et al., 2014b 12d inpatient, 6 week outpatient with PSG measured during inpatient week 1 and 2 and outpatient week 3 and 6, subjective overall sleep quality on 0-100 measured scale during inpatient phase.  Includes self-reported TST/ PSG TST comparison to previously published insomnia data sets.

No HC
Time-dependent sleep state misperception in abstinent cocaine users was reported. TST values from self-report and PSG correlated more strongly during week 1 vs week 2 and there was a significant difference in TST values between these sources [PSG TST decreased from week 1 to 2 while self-report TST did not. During week 2, selfreport SL and WASO were underestimated compared to PSG measures]. A median split for high and low mis-reporters resulted in no difference in baseline PSQI between groups, but REM minutes were lower during week 2 of the high misreporters. There was also a correlation between minutes misreported during week 2 and both TST and REM time.
(continued on next page) T.E. Bjorness and R.W. Greene Pharmacology, Biochemistry and Behavior 206 (2021) 173194  Cocaine dose was 32 mg/70 kg and was self-administered by IV route. Binge duration was 2 h (5 min lockout period between injections with safety lockouts based on heart rate and blood pressure as needed). Placebo was not explicitly defined.
Included cognitive testing (simple reaction time, digit vigilance reaction time) and procedural learning (motor sequence task).
Inpatient on Caffeine Free Unit, some polydrug use.
> binge, 5th abs <1st, 2nd, 3rd, 4th abs, SL: binge >1st and 2nd abs and 4th, 5th abs >1st abs, Slow band activity: binge >2nd, 3rd abs, 2nd abs <4th, 5th abs, subjective sleep quality binge <3rd, 5th abs bin and 5th abs >1st, 2nd, 4th abs bin). Thus, the timing of increased TST early during abstinence was not accompanied by an increase in sleep intensity and the subjective improvement overlapped with an objective decrease in TST which was labeled 'occult insomnia'. SWS did not vary from cocaine through abstinence, though it was noted that several subjects had little SWS. Digit vigilance reaction times were faster during binge compared to all abs bins, but slower during the 5th abs bin compared to 4th abs bin.  Accommodation (cocaine tolerance test plus off days) -binge cocaine or placebo (midday for 3d) -abstinenceplacebo or binge cocaineabstinence (some subjects accommodationbinge cocaineabstinenceplaceboabstinence, others accommodationplaceboabstinencebinge cocaineabstinence. PSG 16× over 23d was binned into accommodation, binge days, and 5 bins of abstinence. Cocaine dose was 32 mg/70 kg and was self-administered by IV route. Binge duration was 2 h (5 min lockout period between injections with safety lockouts based on heart rate and blood pressure as needed). The placebo was saline.
Included visual overnight learning (texture discrimination task).  With post-hoc assessment separated male (21 placebo, 23 Modafinil) and female (6 placebo, 7 Modafinil) participants No HC An increase in N3 from abstinent week 2 to 3 was associated with percentage of urine screens negative for cocaine metabolites and maximum consecutive days abstinent.
Modafinil increased N3 from abstinence week 2 to 3 and resulted in a higher mean rate of cocaine negative urine screens.
When dividing participants by sex, sleep outcomes and differences between Modafinil and placebo remained in males, but not females. Female participants showed a similar increase in N3 time following Modafinil, but had more N3 during baseline compared to male participants which likely explains the lack of significant increase. However, Modafinil also did not increase mean rate of cocaine negative urine screens in female participants suggesting that the positive group-level clinical outcome was driven by male participants. Pace-Schott et al., 2005 Initial abstinence (3d) -binge cocaine (midday for 3d) -abstinence (15d divided into early and late phases SE decreased from binge cocaine exposure to late abstinence, with a non-significant trend towards decreased TST and REML and increased SL. SL and REML were significantly different between binge cocaine and early abstinence (decreased and increased, respectively). Subjective SL was stable across binge through abstinence, but compared to objective SL individuals overestimated SL during binge phase but underestimated SL during abstinence. This is similar to the 'occult insomnia' described by Morgan et al., 2006. Post et al., 1974a Placebo (average 4d) -cocaine (average 6d) -placebo (average 3d).
Cocaine dose varied across days REM, TST, SL, REML, NREM, Stage3, Intermediate wakefulness, movement time, early morning awakening, total recording period, REM activity, REM 5, moderate to severely depressed non-users No HC REM and TST were decreased followed by subsequent rebound.
Subset of population included those with dual diagnosis of mental disorder and substance use disorder. In non-statistical comparisons with HC group, TST was within normal range during acute withdrawal, but a standard deviation lower relative to control participants during subacute withdrawal. REM was higher relative to control group during acute withdrawal, but lower during subacute withdrawal. Trksak et al., 2013 Baseline ( No difference in sleep response or subjective sleepiness following 1 night of SD between cocaine users and non-users and no difference in sleep between groups prior to SD. In response to SD, SL and WASO decreased while SE, TST, Stage 2, and REM increased suggesting sleep time homeostasis was maintained in a population of cocaine users in which sleep was similar to control values. Cocaine users showed lower evening sleepiness and higher difficulty waking under the baseline conditions, but showed similar sleepiness responses as controls following SD. Performance on a continuous performance task and digit symbol substitution task was impaired in cocaine users under non-deprived conditions, but SD did not influence performance. Walsh et al., 2009 Binge cocaine (morning for 4d)placebo (4d, considered acute withdrawal) -binge cocaine (morning for 4d) -placebo (28d, considered protracted withdrawal). Subjective Self-reported TST, time to wake up, time to get up 9, cocaine users Placebo condition; no HC Subjective TST decreased on the cocaine exposure phase compared to placebo during the acute withdrawal phase, with decreases in sleep due to less sleep during the day.
(continued on next page) ratio schedule in which progressively more lever presses are needed to acquire successive infusions (Puhl et al., 2009) nor does overnight sleep disturbance (hourly awakening) increase cocaine choice relative to food reward in rhesus monkeys (Brutcher and Nader, 2013). However, recently it was shown that pharmacodynamics influence food vs drug choice in non-humans with faster dynamics supporting relative preference for food rewards (Canchy et al., 2021) suggesting the reward timing is an important consideration of choice designs. With chronic sleep reduction (40% reduction of sleep during light phase, 80% reduction of sleep during dark phase, for two sessions of 4d sleep restriction separated by 2d recovery), motivation to seek cocaine increases in high cocaine-taking rats during the second sleep restriction period with near significant increases in seeking during the first sleep restriction period (Puhl et al., 2013). Chronic sleep restriction did not influence motivation for cocaine in low cocaine-taking rats. Reduction of cocaine withdrawal-induced REM sleep fragmentation (via sleep restriction during the dark phase) decreases cocaine craving as determined by cue-induced cocaine seeking, while chronic sleep fragmentation across the 24 h period for 7d accelerates the time-dependent incubation of cocaine craving (Chen et al., 2015 [rodents]).
In sum, there is both direct and indirect support for the ability of sleep disturbance to influence cocaine reward behavior. Interestingly, with self-administration, acute sleep deprivation modulates cocaine seeking in low taking animals, while chronic sleep disruption modulates Also included physiological measurements and a performance battery (digit symbol substitution task, digit enter and recall task).
Inpatient on a Caffeine Free Unit.
Conversely, subjective waking early was increased on cocaine exposure phase compared to placebo during the acute withdrawal phase. Time to wake up and time to get up increased from cocaine exposure phase to the protracted withdrawal phase. Washton and Tatarsky, 1984 Telephone interview based on callers to a cocaine abuse helpline.
Self-reported sleep problems included as a checklist item

55, cocaine users
No HC Sleep difficulties endorsed by 58% of individuals; excessive sleeping following a binge was one of the most common physical complaints Watson et al., 1992 Baseline (1d) -recreational use (binge, 1d) -recovery (3d).
Cocaine dose was unknown, but estimated to be 1-1.5 g by intranasal route.
Mixed inpatient with leaving facility for cocaine use. cocaine seeking in high taking animals suggesting differential sensitivity to differing amounts of sleep loss within subpopulations.

Therapeutics targeting sleep disturbance and their reward outcomes
There is currently no FDA approved pharmacotherapy for cocaine use treatment, though dopaminergic, GABAergic, and orexinergic agents have been investigated for efficacy in modulating sleep in cocaine dependent individuals or maintenance of abstinence. See Table 1 for full details. Modafinil (weak, selective DAT inhibitor used to treat excessive daytime sleepiness) has been shown to improve sleep during abstinence (Morgan et al., 2010 and provides some useful secondary abstinence outcomes such as reducing the number of cocaine use days, though it does not improve cocaine abstinence or treatment retention rates (meta-analysis, Sangroula et al., 2017). Recently, a small pilot experiment has provided preliminary support for use of Suvorexant (dual orexin receptor antagonist approved for insomnia treatment), in cocaine-dependent individuals finding a slight increase in sleep via actigraphy and a decrease in craving (Suchting et al., 2020). A Phase I clinical trial investigating the effects of Suvorexant on cocaine reinforcement in humans is ongoing (PI, Dr. William Stoops, University of Kentucky). Orexin receptor antagonism (SB-334867, selective orexin receptor 1 antagonist) reduces sleep deprivation-induced enhancement of cocaine conditioned place preference in rodents (Bjorness and Greene, 2020). While outside the scope of the current review focusing on cocaine reward, orexin receptor antagonists may also provide therapeutic benefits for alcohol, opioid, and polydrug abuse (for review, Moorman, 2018;James et al., 2020;Zarrabian et al., 2020). Currently, dual orexin receptor antagonists are promising with a clinical trial currently underway.
Adenosinergic modulators have been tested for sleep-related therapeutic effects in disorders such as restless leg syndrome (Decerce et al., 2007) and Parkinson's disease , with subjective improvements in sleep-related symptoms. These modulators have not been investigated for sleep-related treatment to reduce relapse rates in a cocaine experienced population. Targeting the adenosine system brings challenges due to issues such as tolerance, receptor heterodimerization, and caffeine use (for review, Chen et al., 2013). Nevertheless, alternate mechanisms of adenosine modulation including allosteric modulators or indirect modulation of the adenosine system have gained attention as possible therapeutic avenues (Peleli et al., 2017;Jacobson et al., 2019). Modulation of the adenosine system may be particularly useful in promoting delta activity within SWS thereby increasing the intensity of SWS; low levels of SWS (N3) have been reported in a subset of cocainedependent individuals (Morgan et al., 2006;Irwin et al., 2016), while increases in SWS over the course of abstinence have been correlated with reduced relapse risk (Angarita et al., 2014a). It is currently unclear whether increasing delta activity alone is sufficient to reduce relapse risk; however, modulators that either increase adenosine tone or activity through A1 receptors could provide a direct test. Thus, as yet, cocaine use outcomes following sleep interventions are encouraging based on reductions in cocaine use days and craving.

Limitations and future directions
A main limitation for interpreting the body of work investigating changes in sleep following cocaine use/withdrawal/abstinence is the wide variability in experimental designs used across experiments. While results in human subjects have an advantage in relevancy to the human condition, there are myriad challenges in designing experiments due to the variability in cocaine use patterns, polydrug issues, and genetic variability. Additionally, long term objective sleep measurements are expensive to conduct and difficult to control in humans. Thus, nonhuman experiments would be useful to characterize long term changes in sleep (and the response to challenges such as sleep deprivation in order to assess possible changes in homeostatic sleep control) and would provide the additional benefit of being able to use within-subjects designs by comparing drug naïve and post-chronic cocaine exposure sleep parameters, which is precluded in human subjects. Rodent studies provide the advantages of being comparatively less expensive, easier to control, and include the option for genetic sameness; however, while many sleep parameters are similar across rodent and primate species, one major difference is the timing of sleep across the 24 h period. Excluding early development, primates are largely diurnal, while rodents are typically nocturnal dominant and polyphasic. Furthermore, humans may face social pressures (work/school timing) that are not easily modeled in non-humans which may limit interpretation of cocaine-mediated changes in the circadian timing of sleep.
Additionally, investigation of sleep/waking behavior following drug combination exposure such as cocaine with caffeine (coca paste formulations as in Schwarzkopf et al., 2018 andRivero-Echeto et al., 2021) and cocaine with heroin ("speedball" as in Duvauchelle et al., 1998), which represent cocaine use variations that occur in the human population, would be informative based on the opposing influence (stimulant vs sedative) of the co-factors with respect to sleep. Polydrug exposure may also provide insight into different potential mechanisms by which drug use can influence sleep behavior and may explain some populationlevel variability in sleep outcomes under drug use and withdrawal.
In regards to the ability of sleep disturbance to influence cocaine reward, additional studies are necessary to provide a mechanistic understanding and to better characterize the conditions under which sleep disturbance alters cocaine reward. Determining the mechanism/s by which cocaine alters sleep and sleep disturbance alters cocaine reward may provide insight into possible novel therapeutic options to reduce the vulnerability to addiction or reduce relapse risk. Here, the use of non-human subjects provides advantages in terms of types of data that can be collected (such as brain tissue), but has the disadvantage that reward may be in some ways fundamentally different in humans and in non-humans, particularly rodents, (for example, the timescale of delay discounting influencing food vs drug choices as hypothesized by Ahmed (2018)).
Finally, while this review was intended to be comprehensive for sleep-related cocaine research, the constraint of English-only papers and use of a Pubmed-based keyword search may have missed important contributions. As interest in research involving sleep and cocaine use increases future reviews will be warranted.

Conclusions
Based on their dual roles in sleep/arousal and reward behaviors, DA, orexin, and adenosine systems are well placed to mediate cocaineinduced alterations of reward and sleep behavior. Further, cocaineinduced plasticity of the DA, orexin, and adenosine systems may be the means by which sleep disturbance continues into withdrawal and abstinence conditions, long after any disturbance can be attributed to direct cocaine effects. The protracted sleep disturbance has implications for maintenance of abstinence based on correlational evidence of longlasting risk for relapse (measured up to 6 weeks; Angarita et al., 2014a). We now have direct evidence that this sleep disturbance increases the rewarding properties of cocaine (Bjorness and Greene, 2020). Thus, treatment of sleep disturbance, including with dopaminergic, orexinergic, or adenosinergic modulators, may be expected to have potential for reducing the risks of relapse, a highly valued clinical outcome for the patient with cocaine use disorder.

Declaration of competing interest
The authors have nothing to declare.