Introduction

Significant advances have been made in understanding the brain mechanisms that regulate sleep and wakefulness1,2. The widely accepted two-process model of sleep regulation3 proposes that sleep timing, depth, and duration result from the interaction between circadian rhythms and a homeostatic drive that accumulates during wakefulness4. Yet, growing evidence shows that sleep need is not uniform across all waking states. Instead, it is shaped by the nature and intensity of waking experiences, particularly those involving high arousal or strong motivational and emotional salience2,5. Parallel neural circuits recruited during such behaviors have been shown to influence the onset and duration of both non-rapid eye movement (NREM) and rapid eye movement (REM) sleep. However, how these circuits process experience-dependent signals and translate them into sleep-promoting activity remains poorly understood. It is also possible that these same circuits contribute to sleep-related processes beyond state regulation itself.

The basal ganglia (BG), and particularly the ventral striatum including the nucleus accumbens (NAc), has emerged as a key site for integrating motivational state, arousal regulation, and sleep homeostasis. By contributing to the fine-tuning of arousal across behavioral states, the NAc may regulate transitions between sleep and wakefulness. This review provides an updated overview of experimental evidence from rodent models implicating the BG, focusing in particular on the NAc and its interconnected circuits, in the modulation of vigilance states. It further examines how sleep may intersect with NAc functions related to valence processing and associative learning, and considers the NAc’s emerging, though still incompletely defined, role in sleep-associated cognitive functions. In doing so, the review aims to outline key mechanisms and highlight unresolved questions that warrant further investigation.

From movement to motivation: the Basal Ganglia as a state control network

Motor and limbic balance in Basal Ganglia organization

The BG is structurally organized into input, intrinsic and output components. These components form parallel circuits, including motor, associative, and limbic loops, each supporting distinct behavioral functions6,7.

The dorsal striatum (DS), which includes the caudate nucleus and putamen in rodents, serves as the main input region for the motor and associative circuits6. It receives excitatory input from the sensorymotor and prefrontal cortices, as well as from the intralaminar and motor thalamic nuclei, along with dopaminergic modulation primarily from the substantia nigra pars compacta (SNc). Signals are processed within the external globus pallidus (GPe) and subthalamic nucleus (STN), and then relayed through the output nuclei, the internal globus pallidus (GPi) and the substantia nigra pars reticulata (SNr), to the thalamus and back to the cortex. This circuit supports movement control and behavioral flexibility.

The limbic circuit is centered on the NAc, which is divided into core and shell compartments. It receives converging inputs from limbic cortices and subcortical structures, including the extended amygdala, hippocampus, ventral tegmental area (VTA) and midline thalamic nuclei. Notably, it maintains reciprocal connections, particularly with many of these regions, especially through its shell compartment8. Within this circuit, the primary targets of the NAc are the ventral pallidum (VP) and the ventral mesencephalon (VM), which includes the SNr and VTA. The VP also serves as a relay, transmitting signals from the NAc to the VM. In addition, the NAc projects to multiple forebrain and brainstem regions involved in arousal and sleep-wake regulation9, such as the lateral hypothalamic area (LHA), lateral preoptic area, pedunculopontine tegmental nucleus (PPN), periaqueductal gray, and parabrachial area. Through this extensive connectivity, the NAc acts as a central hub that integrates reward-related signals with motivational, emotional, and motor systems to coordinate adaptive behavioral responses.

Throughout the striatum, the principal projection neurons are GABAergic medium spiny neurons (MSNs), which are broadly divided into two main types based on their dopamine receptor expression and their connectivity10. These include D1-MSNs, which express Gs-coupled D1 receptors, and D2-MSNs, which express Gi/o-coupled D2 receptors, with dopamine acting on these receptors to oppositely regulate neuronal excitability through divergent effects on cAMP signaling11. Additionally, D2-MSNs uniquely co-express Gs-coupled adenosine A2A receptors12,13, which, among other functions, are implicated in sleep promotion14. In the DS, D1-MSNs form the direct pathway to the GPi and SNr, whereas D2-MSNs contribute to the indirect pathway via the GPe and STN10. These pathways exert opposing effects on motor output. In the NAc, this anatomical segregation is less clear15,16. D1-MSNs project either directly to the VM or indirectly via the VP, forming both direct and indirect pathways. Similarly, D2-MSNs influence the VM primarily through the indirect route via VP neurons, but they also engage an alternative ‘direct pathway’ by targeting VP neurons that project directly to the thalamus.

More recently, a third MSN subtype has been identified that co-expresses both D1 and D2/A2A receptors. In the DS, this minority population projects to the GPe, similar to D2-MSNs, and may serve a buffering role that helps coordinate or stabilize the motor functions of the two classical MSN types17. These neurons also display unique physiological properties and differentially integrate dopamine signals, suggesting a distinct modulatory function within striatal circuits. However, their contribution to NAc functions, including their distribution and connectivity within this ventral region, remains unclear and requires further investigation, particularly in the context of reward processing and arousal regulation, given their ambivalent sensitivity to dopamine.

D1- and D2-MSNs in the NAc: Valence, encoding and motivational control

Reward can be defined both as an external stimulus that reinforces approach behavior and as an internal state that motivates such behavior. While positive reinforcement involves learning the relationship between environmental cues and outcomes, reward also encompasses the arousal and motivational drive that enhance attentional engagement with salient stimuli18,19. The NAc plays a central role in these processes, functioning as a key integrative hub that links sensory input, motivational salience, and motor output to coordinate goal-directed actions18,20,21. It is particularly recruited when behavior must be flexibly guided in the face of uncertainty or competing options, helping the organism weigh potential rewards, costs, and risks. In such contexts, the NAc supports adaptive decision-making by distinguishing between stimuli based on their valence (emotional value, e.g., pleasant vs. unpleasant) and salience (noticeability or importance, e.g., novelty or unpredictability). This includes resolving approach-avoidance conflicts, comparing actions based on reward value or effort, and dynamically adjusting behavior as contingencies shift. When NAc function is disrupted, this ability to resolve ambiguity and select appropriate responses is often impaired, underscoring its importance in integrating motivational information to guide behavior.

Within the NAc, D1- and D2-MSNs have been shown to exert distinct, often opposing effects on motivation, reward processing, and learning, with foundational studies linking D1-MSNs to the encoding of rewarding and positively reinforcing stimuli, and D2-MSNs to processing aversive or punishing experiences22,23,24,25,26. However, this traditional dichotomy has been increasingly challenged by evidence supporting a more nuanced interplay, in which D1- and D2-MSNs are co-activated in response to both aversive and rewarding stimuli, and their recruitment may produce overlapping or complementary behavioral outcomes18,20,21,27,28. Their relative contributions also appear to shift depending on context and changing contingencies. D1-MSN activation consistently drives robust motivation for self-stimulation, regardless of the paradigm, whether through actions like touching a spout to trigger optogenetic stimulation or exploring a location paired with stimulation29. They have also been proposed to encode perceived salience, consistent with findings showing that they respond broadly to stimuli regardless of valence or contingency, and that D1 receptor signaling is required for reward-induced stimulus generalization30,31,32. In contrast, although NAc D2-MSN optogenetic stimulation can also drive self-administration behavior29,32, it tends to be less effective and is associated with slower response rates29. Interestingly, this initially reinforcing effect can shift toward neutral or even negative valence depending on several factors, including the experimental paradigm, the phase of motivated behavior (e.g., reward prediction vs. reward consumption), and specific stimulation parameters. For instance, brief stimulation pulses tend to promote motivation, whereas prolonged stimulation can induce avoidance in a laser-paired location test18,20,28,29,33,34,35. Beyond reinforcement, NAc D2-MSNs play a critical role in outcome-based learning and behavioral inhibition29,30,32,36,37,38,39. Specifically, they are activated by negative outcomes and stimuli omissions in operant conditioning, and their stimulation can shift to risk avoidance behavior, suggesting that D2-MSNs promote adaptive decision-making by encoding prior unfavorable outcomes and shaping both individual risk preferences and real-time choices30,32,36,37,38,39. D2-MSNs have been shown to dynamically track learning. In the NAc core, their activity in response to conditioned cues scales up with learning, a pattern not observed in D1-MSNs32. In contrast, in the shell, D2-MSNs are primarily recruited by unexpected omissions of rewarding or aversive stimuli and are essential for extinction learning39. Together, these findings support a distinct role for D2-MSNs in error signaling and updating outcome expectations, depending on the NAc subregion recruited. This underscores the broader functional divergence between D1- and D2-MSNs, which extends beyond reinforcement to include specialized contributions to behavioral monitoring, adaptation, and flexibility in response to changing environmental demands.

Importantly, such differences often reflect the specific NAc subregion targeted40,41,42. MSNs in the NAc core may be more prominently involved in salience detection and attentional prioritization, particularly D1-MSNs, as well as in signaling prediction errors, especially through D2-MSNs. In contrast, medial shell MSNs may be more specialized in processing valence and guiding approach or avoidance behaviors, with D2-MSNs playing a prominent role in updating responses when contingencies shift. These functional distinctions likely stem from differences in afferent inputs and projection targets, both across NAc subregions8,40,43 and between D1- and D2-MSN populations44. Another major source of divergence is their differential sensitivity to dopamine and other neuromodulators. D1 and D2 receptors differ in dopamine affinity: D2 receptors respond preferentially to tonic dopamine levels, while D1 receptors are primarily activated by phasic bursts45. This contributes to distinct modes of dopaminergic processing, influencing how each MSN type shapes behavior. While both D1- and D2-MSNs can respond to dopamine transients46,47, D2-MSNs uniquely detect dips in tonic dopamine, a property consistent with their recruitment during the omission of expected rewards or aversive stimuli associated with dopamine dips48. These differences are further modulated by receptor-specific adaptations to sustained dopamine exposure and by the influence of additional neuromodulators47, most notably adenosine, which plays a key role in tuning arousal states through its action on A2A receptors. This suggests a potential state-dependent modulation of NAc D2-MSNs in particular. Such a mechanism aligns with growing evidence that D1- and D2-MSNs exert distinct influences on sleep-wake regulation, supporting the view that NAc circuits are well-positioned to integrate motivational drive with vigilance state control.

State signals in striatal circuits: correlates across sleep and wake

Motor pathways and state-dependent activity

In the DS, early in vivo recording studies revealed that the activity of undifferentiated MSNs fluctuates across vigilance state49, indicating dynamic modulation during sleep and wakefulness. During wakefulness, MSNs exhibit irregular spiking and remain mostly hyperpolarized. In NREM sleep, their firing becomes fast and rhythmic, tightly synchronized with cortical slow waves, alternating between hyperpolarized and depolarized phases that generate bursts of action potentials, likely driven by high-frequency glutamatergic cortical input49,50. In REM sleep, MSNs display slow ( ~1-2 Hz) oscillations with brief, high-frequency depolarizing events, whose origin remains unclear but appears independent of cortical activity. Focusing on MSN subtype specificity in the DS, fiber photometry recordings have shown that D1-MSNs increase their activity specifically during transitions from NREM sleep to wakefulness and are most active during spontaneous wake51. Their activity begins to rise just before NREM-to-wake transitions and decreases prior to wake-to-NREM transitions, occurring simultaneously with the activation of prefrontal cortex and mediodorsal thalamic neurons. In contrast, while no significant changes in D1-MSN activity are observed during transitions between NREM and REM sleep or from REM to wakefulness, their mean activity during REM sleep is significantly higher than during NREM sleep. These patterns suggest that D1-MSNs are more selectively engaged in wake promotion specifically at the exit of NREM sleep. For D2-MSNs, while causal manipulations suggest that their sustained activation promotes sleep52, direct recordings showing dynamic activity changes across sleep/wake states are still lacking.

In downstream BG nuclei, GPe neurons show a higher mean firing rate during wakefulness than during NREM sleep, while maintaining relatively stable firing across both states53. In contrast, STN neurons consistently switch from a random discharge pattern during wakefulness to rhythmic bursting during NREM sleep, without significant changes in their mean firing rate53. Notably, both structures exhibit marked increases in firing rate during REM sleep, a pattern that diverges from the classical view of reciprocal activity between the STN and GP in motor loop processing. This suggests that mechanisms beyond the well-known inhibitory input from the GP may modulate STN activity, particularly during REM sleep, and remain to be elucidated. In parallel, the SNr presents a more intricate organization, comprising distinct GABAergic subpopulations and circuits that are differentially engaged in motor control and brain state regulation54. Notably, a neuronal population expressing glutamic acid decarboxylase 2 (GAD2) shows a marked increase in activity at the transition from wakefulness to NREM sleep and during states of low motor activity and arousal, while remaining silent during movement. In sharp contrast, SNr neurons expressing parvalbulmin (PV) are predominantly active during locomotion and show a dramatic decrease in firing at transitions to quiet wakefulness and sleep. Both populations receive inputs from upstream BG nuclei, including the STN, GPe, and DS. However, GAD2 neurons show distinct, reciprocal connectivity with brainstem and hypothalamic wake-promoting centers such as the locus coeruleus, dorsal raphe and LHA.

Limbic loop engagement during sleep: spotlight on the NAc

Early evidence linking NAc activity to behavioral state regulation emerged from seminal in vivo electrophysiological recordings, which showed that firing rate of non-classified NAc neurons varies across vigilance states, being generally higher during EEG-desynchronized states such as wakefulness and REM sleep compared to NREM sleep55,56,57. Although only a subset of recorded neurons was tracked across vigilance state transitions, those that were observed longitudinally tended to decrease their firing at the onset of NREM sleep and occasionally exhibited higher activity during REM than wakefulness. In parallel, high-frequency oscillatory activity within the NAc was found to diminish during NREM sleep56, a pattern that stands in contrast to brain regions like the hippocampus and cortex, where such oscillations are typically enhanced during NREM. This dissociation suggests that high-frequency NAc rhythms may be preferentially engaged during active behavioral states rather than during offline consolidation processes58. Nonetheless, given the similarity in synaptic inputs and intrinsic membrane properties between NAc MSNs and their counterparts in the DS49, it is plausible that NAc MSNs also participate in structured, rhythmic activity during NREM sleep. Supporting this idea, slow delta-range oscillations (0.5–3 Hz) have been observed in the NAc, where they are suppressed by dopaminergic signaling through D2 receptors and reach their lowest amplitude during movement and REM sleep57.

Cell-type-specific activity patterns have revealed a markedly different picture from the earlier notion of the NAc being generally more active during states of EEG desynchronization. In fact, early studies appear more consistent with the activity profile of D1-MSNs rather than D2-MSNs. Bulk population signals from D1-MSNs, measured using fiber photometry, show that these neurons exhibit minimal activity during NREM sleep, followed by a sharp increase immediately preceding transitions into wakefulness or REM sleep, and a subsequent reduction in firing during wakefulness before returning to NREM sleep59. In contrast, NAc D2-MSNs exhibit heightened activity during the transition from wakefulness to NREM sleep and remain more active across both NREM and REM sleep compared to wakefulness60. These opposing activity patterns point to a functional dissociation, with D1-MSNs likely involved in initiating arousal and REM-related activation, while D2-MSNs may contribute more directly to the initiation and maintenance of sleep, especially NREM sleep.

Flipping the switch: causal insights into BG and NAc sleep control

Motor circuits as arousal modulators: beyond movement

Causal studies targeting the motor BG circuit have shown that its interconnected components collectively modulate sleep-wake states (Fig. 1, Table 1). Early loss-of-function approaches first revealed that lesions within this circuit alter global arousal stability and cortical activation61. Excitotoxic (ibotenic acid) lesion of the DS decreases total wake duration, increases sleep-wake fragmentation, and slows cortical EEG rhythms61. Although these effects are moderate compared to those seen after pallidal manipulations61, they demonstrate that striatal integrity contributes to maintaining stable vigilance states. Because dopamine from the SNc exerts a major modulatory influence on DS activity and downstream BG output, its contribution is of particular relevance for understanding how this circuit regulates sleep-wake states. Lesions of SNc dopaminergic neurons, including their inputs to the DS, increase wakefulness and reduce both NREM and REM sleep, accompanied by higher state fragmentation62. Consistently, optogenetic excitation of SNc dopaminergic terminals in the DS promotes NREM sleep, but only when performed under low sleep pressure, during the transition to the dark active phase62. At light onset, when sleep pressure is high, no sleep-promoting effect is observed62. On the contrary, stimulation during this period decreases EEG delta power, increases EMG tone, and raises the probability of transitioning to wakefulness with shorter latency51,63, consistent with the well-established arousal-promoting role of VM dopaminergic neurons. This role is further supported by evidence that dopamine release in the DS and activity of VM dopaminergic neurons are elevated during wakefulness, and that their activation initiates and maintains wakefulness, whereas inhibition suppresses it and promotes NREM sleep63,64,65,66. Together, these findings underscore the complex, context-dependent role of the DS, and particularly dopaminergic modulation within its intricate neuronal network, in regulating sleep and wakefulness. The impact of dopamine within this structure likely varies across circadian phase, sleep pressure, and prior wake experience, reflecting shifts in network state, neuromodulatory tone, and circuit-level interactions. Although these mechanisms remain speculative and require further investigation, a dynamic balance between D1- and D2-MSN activity may help mediate such state-dependent transitions, enabling dopamine to support both arousal and sleep depending on internal conditions.

Fig. 1: Causal influence of basal ganglia pathways on sleep and wake states: limbic and motor circuitry.
figure 1

This schematic illustrates the functional organization of the basal ganglia’s limbic pathway (left) and motor pathway (right) in relation to sleep–wake control. Solid red and blue arrows represent causally demonstrated wake- and sleep-promoting circuits, respectively. Dashed gray arrows indicate anatomically established connections that are potentially relevant to sleep/wake regulation, though their functional impact remains to be determined. Cx cortex, D1- MSNs Medium spiny neurons-expressing dopamine D1 receptors, D2-MSNs Medium spiny neurons-expressing dopamine D2 receptors, DA Dopaminergic neurons, DS Dorsal striatum, GAD2 Glutamate decarboxylase 2 expressing neurons, Glu Glutamatergic neurons, GPe External part of the Globus Pallidus, GPi Internal part of the Globus Pallidus, HA Histaminergic cells, Hcrt Hypocretin-expressing neurons, Hipp Hippocampus, LHA Lateral Hypothalamic Area, MCH Melanin concentrating hormone-expressing neurons, NAc Nucleus Accumbens, SNc Substantia Nigra pars Compacta, SNr Substantia Nigra pars Reticulata, TMN Tuberomammillary Nucleus, VP Ventral Pallidum, VTA Ventral Tegmental Area, PV Parvalbumin neurons.

Full size image
Table 1 Cell-Type and pathway-specific manipulations in basal ganglia circuits relevant to sleep–wake regulation
Full size table

Consistent with a role in promoting wakefulness, optogenetic activation of D1-MSNs triggers immediate transitions from NREM sleep to wakefulness, accompanied by rapid suppression of cortical delta activity51. This effect appears even stronger than that observed with stimulation of SNc dopaminergic terminals in the DS51. Conversely, inhibition of D1-MSNs reduces wakefulness and promotes NREM sleep, without affecting REM sleep51. These findings underscore the essential role of the DS in maintaining arousal and suggest that the D1-MSN pathway is central to this function.

In opposition to D1-MSNs influence, chemogenetic activation of D2-MSNs in the DS increased sleep by extending NREM episodes at the expense of wakefulness, while their inhibition reduced NREM sleep without affecting REM sleep52. These D2-MSNs driven sleep-promoting effects seem to be efficient only when neurons are activated during the active dark period, at low sleep pressure. They are also recapitulated by optogenetic activation of D2-MSN terminals in the GPe52, which involves the prototypic neurons of GPe expressing PV67. Indeed, the selective lesion of GPe PV-expressing neurons prevents the sleep-promoting effects of D2-MSNs activation52. Among BG subnuclei tested with lesioning approaches, the GPe stands out as exerting the most pronounced impact on vigilance states, underscoring its pivotal position as a relay integrating sleep-related striatal signals61. However, its lesion produces opposite effects depending on the extent: complete lesions induce severe insomnia61, whereas selective ablation of PV-expressing prototypic neurons increases NREM sleep at the expense of wakefulness and tends to reduce REM sleep52. These latter effects are consistent with the firing patterns observed in GPe neurons across sleep–wake states53. Together, these findings reveal the complexity of the GPe network, whose heterogeneous neuronal subpopulations can exert opposing influences on sleep and wake regulation, much like their divergent roles in motor control, highlighting the need for finer dissection of this structure’s functional architecture68.

Progressing further downstream in the motor circuit, although lesions of the STN, GPi, and SNr do not markedly alter overall sleep-wake architecture61, targeted neuronal manipulations within these structures have also revealed a modulatory role in sleep-wake regulation. While circuit-level dissection of the STN’s role in sleep remains limited in rodent models, clinical studies show that high-frequency stimulation of the STN not only improves motor symptoms in Parkinson’s disease but also alleviates sleep disturbances and enhances sleep quality69,70,71. Similarly, recent circuit-based investigations in the SNr have uncovered distinct GABAergic pathways involved in either motor control or sleep regulation54. Specifically, optogenetic activation of SNr GAD2-positive neurons, reciprocally connected with brainstem and hypothalamic arousal centers, promotes sleep initiation and prolongs sleep duration. In contrast, stimulation of PV-expressing SNr neurons, which are more exclusively linked to the motor circuit, induces motor initiation without affecting sleep state, consistent with their silenced activity during sleep54.

This functional segregation at the level of the BG output nucleus provides a compelling framework to interpret how motor-related circuits may intersect with sleep regulation. However, this does not fully account for the distinct roles of upstream striatal D1- and D2-MSNs, both converging onto SNr populations, in promoting wakefulness and sleep, respectively, suggesting the involvement of additional mechanisms, likely influencing thalamocortical dynamics. In this context, a broader view of the basal ganglia’s contribution reveals a complementary top-down mechanism of arousal control originating from the striatum, contrasting with the classical bottom-up pathway ascending from the pons through midbrain and diencephalic relays to the cortex72. Importantly, the effectiveness of striatum-driven motor functions appears strongly influenced by both circadian phase and sleep pressure. Recent results show that manipulations of D1- or D2-MSNs across the circadian cycle do not produce consistent motor outcomes17, an aspect often neglected in motor studies. Optogenetic stimulation of the DS direct pathway robustly increased locomotor speed at the beginning of the light phase but failed to produce consistent effects at the onset of the dark period. These findings emphasize the importance of internal brain-state variables in shaping the motor consequences of striatal activation. Building on this, the opposing functions of D1- and D2-MSN-derived pathways in arousal may together facilitate D1-driven behavioral and motor engagement, alongside the action inhibition and adaptive properties supported by D2-MSNs. To further unravel how sleep and motor functions interact at the circuit level, future studies should examine the temporal dynamics between cortical, thalamic, and BG activity under precise manipulation of D1- or D2-striatal pathways during behavior and concurrent monitoring of arousal state.

NAc control of brain state: D1 for drive, D2 for disengagement

Early lesion studies point to distinct contributions from each region: while DS lesions reduce overall wakefulness and abolish long wake episodes, NAc lesions have the opposite effect, increasing wakefulness at the expense of sleep by shortening sleep episode durations, without affecting the length of wake bouts61. This finding stands in contrast to early in vivo electrophysiological recordings of unidentified NAc neurons, which showed the highest firing rates during wakefulness55. Based on these recordings, the NAc was initially thought to play a role similar to the DS in initiating and maintaining arousal. However, more recent studies have revealed a functional opposition within the NAc itself, with D1-MSNs promoting arousal and D2-MSNs facilitating sleep. In addition to this cell-type dichotomy, subregional differences between the NAc core and shell also appear to influence sleep regulation, although their specific contributions remain incompletely understood. Lesions targeting the core result in greater wakefulness and a more pronounced reduction in the homeostatic response to sleep deprivation compared to medial shell lesions73, underscoring the core’s importance in sleep regulation. At the same time, core lesions also abolish the arousal-promoting effects of modafinil, indicating that this region is essential for the expression of dopaminergic arousal. In contrast, caffeine-induced arousal relies on signaling within the medial shell through D2/A2A-MSNs, rather than those in the core74, suggesting that this subregion contributes to, and may be more sensitive to, adenosine-related modulation of arousal. These findings highlight distinct neuromodulatory mechanisms across NAc subregions and caution against a simplistic anatomical division of arousal and sleep regulation.

In the following sections, we separately examine the roles of D1-MSNs and D2-MSNs in sleep-wake regulation, along with the specific circuits through which they exert their influence (Table 1, Fig. 1).

a D1-MSNs in the NAc: arousal drivers in conflict with REM sleep?

Optogenetic and chemogenetic activation of D1-MSNs in the NAc robustly promotes arousal, inducing rapid transitions from NREM sleep to wakefulness and sustaining prolonged wake states59,60. This activation also reduces sensitivity to anesthesia and enhances cortical activation, facilitating emergence from anesthetized states75. Importantly, the wake-promoting effects of D1-MSN activation may not solely reflect a shift in vigilance state, but also a mobilization of motivational resources. D1-MSNs are strongly implicated in encoding reward salience, facilitating approach behavior, and driving goal-directed engagement: functions that inherently require heightened arousal and behavioral readiness. Thus, D1-MSN-induced wakefulness may reflect not just cortical activation per se, but the induction of a motivated state, optimized for behavioral engagement in response to salient internal or external cues. In this view, arousal driven by D1-MSN activity could be understood as a behavioral mode switch: transitioning the brain into a “ready-to-engage” state that supports exploratory behavior, decision-making, or reward pursuit. This interpretation aligns with evidence that D1-MSN activation supports reward-seeking regardless of contingency or valence, and responds broadly to salient stimuli, suggesting a role in energizing behavior beyond simple sleep suppression. The wake-promoting profile of NAc D1-MSNs is further supported by studies showing that activation of VM dopaminergic neurons innervating the NAc, robustly increases arousal63,65.

While this framing supports a tight link between D1-MSNs, arousal, and motivated behavior, their role in REM sleep introduces a more complex and paradoxical dynamic. REM is not a passive state, it is marked by intense cortical activation, elevated limbic engagement, and internal motivational processes such as emotional memory consolidation. Yet, despite D1-MSNs being highly active during REM sleep, their activation through optogenetic or chemogenetic methods leads to its suppression, whereas inhibition results in enhancement59,60,76. This effect is consistent across stimulation protocols, from brief pulses (1–16 s) delivered during NREM59,60 to prolonged chemogenetic activation60,76. Depending on the ligand type and dosage, such prolonged activation can either selectively inhibit REM sleep without affecting other states76 or promote sustained wakefulness at the expense of both NREM and REM sleep60. It remains unclear whether optogenetic temporal control on D1-MSNs was ever aligned precisely with REM episodes, an important experimental gap. Time-locked stimulation during REM could help determine whether D1-MSN activation actively disrupts this state or might paradoxically support it once underway. This apparent contradiction raises several important considerations. Given that D1-MSNs are principal targets of dopamine and key mediators of its excitatory effects, the dynamics of dopaminergic signaling offer a compelling framework to interpret these findings. Due to their particular sensitivity to phasic dopamine motifs45, D1-MSNs may contribute to REM induction under specific dopaminergic conditions. This effect may depend on the timing and pattern of dopamine release, as tonic signaling has been shown to suppress REM sleep, while phasic bursts promote it77. Moreover, the effect of dopamine on REM sleep appears to vary by NAc subregion: phasic stimulation of dopaminergic terminals in the NAc core increases REM, while similar stimulation in the medial shell has no effect77. Circadian timing may further influence these dynamics. In the DS, stimulation of SNc dopaminergic terminals promotes NREM sleep at the end of the light cycle, but not at lights-on51,62,63, which is when most D1-MSN manipulation studies were conducted. In addition, dopamine levels in the NAc gradually rise across NREM and peak during REM, following a pattern modulated by REM pressure: dopamine release is suppressed after long REM bouts but not after short ones78. These observations have led to the hypothesis that under REM-sufficient conditions, dopamine may only weakly support REM sleep, whereas under REM-deficient conditions (e.g., after deprivation), dopamine released during NREM may more strongly promote subsequent REM, a possibility that remains to be directly tested. D1-MSNs are thus well positioned to mediate these REM pressure-dependent effects, a possibility that remains to be directly tested using state-specific manipulations and monitoring, such as time-locked targeting of D1-MSNs during REM episodes with and without prior selective deprivation, across distinct NAc regions, including the core, the medial shell, and the lateral shell. Another explanation may lie as well in functional heterogeneity among D1-MSNs across but also within a given region. Just as distinct D1-MSNs mediate reward and aversion encoding depending on their projection targets, such as the VM or VP, respectively79, it is plausible that only a subset of D1-MSNs contributes specifically to REM sleep regulation. These cells may possess distinct input-output connectivity, separate from the dominant population implicated in wake promotion. This hypothesis aligns with growing evidence for functionally distinct dopaminergic subcircuits within the VTA, whose projections to targets like the prefrontal cortex, lateral hypothalamus, amygdala, DS, or NAc exert divergent influences on sleep-wake control, including REM modulation80,81,82. It also resonates with findings demonstrating hemispheric specialization in amygdala-NAc circuits, where interhemispheric inputs to D1- and D2-MSNs encode opposing emotional valences83. Given the well-established engagement of the amygdala during REM sleep84, future studies should investigate the reciprocal connectivity between the amygdala and NAc D1-MSNs in this context.

To resolve these complexities, cell-type and projection-specific tools will be essential. Approaches such as calcium imaging in freely moving animals or projection-specific optogenetics will allow researchers to track the real-time activity of D1-MSNs across sleep states and identify REM-active subpopulations. These methods will also help determine how D1-MSNs integrate motivational, dopaminergic, and glutamatergic inputs to shape brain-state transitions, ultimately clarifying their role at the intersection of motivation and sleep regulation, and their contribution to state-dependent control and behavioral readiness.

b D2-MSNs: signaling when to stop—circuit for sleep and withdrawal

Causal studies focusing on D2-MSNs in the NAc consistently demonstrate that their activation promotes NREM sleep at the expense of wakefulness, particularly in the dark (active) phase59,60,76,85. Notably, this sleep-promoting effect appears to arise predominantly from D2-MSNs in the NAc core, rather than the medial shell85. While D2-MSNs in the NAc also show activation during both NREM and REM sleep, their causal involvement in REM regulation appears limited59,60,76,85. Stimulation of these neurons does not produce substantial changes in REM sleep, or the effect is less consistently reported across studies. Conversely, inhibition or ablation of D2-MSNs leads to reduced total sleep and prolonged wakefulness60,85. These findings stand in contrast to earlier studies examining D2 receptor signaling through full knockout or systemic antagonists, which reported decreased wakefulness and increased sleep86. These manipulations also led to fragmented arousal without significantly affecting homeostatic sleep responses, indicating that D2 receptors play a broader role in arousal maintenance86. This discrepancy suggests that NAc D2-MSNs may promote sleep in a highly localized and circuit-specific manner, distinct from the global wake-promoting effects typically associated with D2 receptor signaling in other brain regions.

The NREM sleep-promoting influence of D2-MSNs may reflect either a direct induction of sleep, as suggested by the rapid sleep onset following stimulation60, or an indirect consequence of increased drowsiness that naturally leads to sleep initiation. Importantly, this effect is not uniform across all states. D2-MSN activation does not appear to induce sleep when homeostatic pressure is already high, such as during the light (inactive) phase. Instead, these neurons may facilitate sleep under permissive behavioral conditions, where wake-promoting systems are less active or disengaged9. Some clues to the mechanism underlying D2-MSN-mediated sleep promotion can come from anatomical tracing studies. D2-MSNs from all NAc subregions project to the VP, the nucleus of the diagonal band, and the substantia innominata87. While strong projections from the core and ventral shell reach the LHA and lateral preoptic area, regions central to sleep/wake state switching88, inputs from the medial shell to these areas are comparatively weaker87. The cellular targets of these projections remain uncertain but could reveal mechanisms for D2-MSN-driven sleep promotion. These neurons may influence sleep–wake regulation either by directly inhibiting wake-promoting populations, such as GABAergic neurons in the lateral hypothalamus (LHA) or glutamatergic neurons in the preoptic area, or by disinhibiting sleep-promoting neurons, including melanin-concentrating hormone (MCH) cells or preoptic GABAergic neurons88. Although the precise identity of the LHA GABAergic neurons involved remains unclear, anatomical and functional studies have shown that D1-MSNs from the NAc shell project robustly to the LHA, where they preferentially target GABAergic neurons rather than directly innervating Hcrt or MCH neurons89. In addition, D1-MSNs from the NAc core have been shown to indirectly influence LHA glutamatergic neurons, potentially including hypocretin (Hcrt)/orexin cells, via a disinhibitory mechanism involving local GABAergic interneurons90. Whether D2-MSNs engage similar or distinct downstream pathways in sleep–wake regulation remains to be determined through targeted circuit-tracing and functional studies, although input from shell D2-MSNs appears minimal compared to that of D1-MSNs89.

The VP remains the central downstream target of NAc D2-MSNs. It is a heterogeneous basal forebrain region involved in motivation, reward, and arousal regulation91. The VP contains GABAergic, glutamatergic, and cholinergic neurons91 and projects to brain regions critical for cortical activation during both wake and REM sleep1,92. Distinct VP subpopulations promote arousal via different pathways: GABAergic neurons increase wakefulness by disinhibiting cortical and VTA circuits93,94,95, while glutamatergic neurons project to the LHA and promote arousal96. In this context, NAc core D2-MSNs appear to promote NREM sleep by inhibiting VP GABAergic neurons, thereby suppressing their arousal-promoting effects85. This local inhibitory mechanism may serve as a gating function, reducing ascending arousal signals and enabling sleep onset. Thus, NAc D2-MSN input to the VP may represent a key convergence point for balancing sleep and arousal states.

The role of NAc D2-MSNs in sleep promotion may be best understood not in isolation, but as deeply intertwined with their broader function in adaptive behavioral regulation. These neurons are recruited by negative outcomes, unexpected reward omissions, and behavioral states where continued effort no longer yields beneficial outcomes30,32,38,39. Within this context, their activation during low-arousal periods and subsequent inhibition of arousal-promoting VP neurons may represent a circuit-level implementation of a broader decision-making strategy: disengagement from unproductive or unfavorable environmental interaction. Just as D2-MSNs shape behavioral choices under conditions of risk, uncertainty, or negative reinforcement, their role in promoting NREM sleep or lowering arousal may serve to withdraw the organism from ineffective goal pursuit and facilitate a restorative internal state. This suggests that sleep induction via D2-MSN activity is not simply a homeostatic reflex, but a motivationally regulated process aligned with behavioral optimization. In this view, D2-MSNs may contribute to arousal tuning by encoding the motivational salience, valence, and predictive value of external stimuli, whether appetitive, aversive, or uncertain, and adjusting arousal levels accordingly when such stimuli no longer justify sustained attention, effort, or behavioral engagement.

The ability of D2-MSNs to tune arousal likely depends not only on their circuit specificity but also on their unique sensitivity to both inhibitory dopamine and excitatory adenosine signals, which fluctuate dynamically across behavioral states, sleep stages, and circadian rhythms. Moreover, A2A receptors can form heteromeric complexes with D2 receptors, enabling reciprocal antagonistic modulation through Gs- and Gi/o-coupled signaling, respectively, and thereby allowing precise tuning of D2-MSN excitability in response to changing adenosine and dopamine levels97. This interplay may underlie a dynamic push-pull mechanism, in which adenosine and dopamine exert opposing, state-dependent influences within the same neuronal population. These complexes also interact with glutamatergic and cannabinoid receptors, influencing both excitability and gene expression in D2-MSNs97. Such multimeric assemblies are thought to support flexible striatal responses, instrumental learning, and adaptive switching between goal-directed and habitual behaviors via the indirect pathway98. These mechanisms may also contribute to arousal recalibration. A2A signaling exhibits temporal specificity, as demonstrated by optogenetic activation of striatopallidal A2A receptor pathways using a rhodopsin-A2A receptor chimera in the dorsomedial striatum, precisely timed with reward delivery99. This finding suggests that A2A receptor activation can exert time-locked effects on behavior, enabling the indirect pathway to dynamically constrain goal-directed actions in response to specific motivational contexts. Further investigation into adenosine dynamics across NAc subregions, particularly during behaviorally relevant transitions into sleep, will be essential, especially given that the primary sources of striatal adenosine remain poorly defined and may involve astrocytic mechanisms that are emerging as potentially relevant to sleep regulation. In this way, A2A signaling may act as a temporally gated modulatory brake, selectively tuning behavioral engagement according to the timing of environmental cues and internal states. Thus, the co-expression of D2 and A2A receptors equips D2-MSNs to integrate motivational and arousal-related inputs with high temporal precision, supporting flexible behavioral regulation across changing physiological and environmental conditions.

Future studies should more directly test whether D2-MSN-induced sleep represents a motivationally adaptive behavioral switch. One promising approach would be to combine real-time sleep-state monitoring with tasks that vary in reward predictability, cost-benefit ratios, or environmental controllability. If D2-MSN activation increases under conditions of low reward utility or high effort cost, and precedes transitions into sleep, this would support the hypothesis that sleep initiation reflects a strategic withdrawal from unfavorable engagement. Manipulating task parameters (e.g., reducing outcome predictability or increasing response cost) while recording D2-MSN activity could help determine whether their recruitment corresponds to a shift in behavioral state or internal disengagement. In addition, longitudinal imaging of D2-MSNs at single-cell resolution across behavioral and sleep/wake cycles could reveal whether distinct subpopulations are specialized for initiating sleep versus responding to aversive outcomes during wakefulness. These functionally defined ensembles may fluctuate dynamically depending on task demands, internal state, or recent experience. Such approaches could also clarify whether D2-MSNs contribute to sleep-active processes such as memory consolidation, particularly for negatively valenced or effortful learning episodes, further bridging their role in motivation and sleep. Overall, the unique neuromodulatory sensitivity of D2-MSNs positions them as state integrators, regulating when to persist in goal-directed activity and when to disengage into sleep, balancing behavioral cost-benefit computations with the organism’s internal need for recovery and recalibration.

NAc contributions to sleep-driven memory and emotion

Motivation plays a pivotal role in shaping what we remember, with rewarding experiences typically leading to stronger and more persistent memories100,101. Memory consolidation, the process that stabilizes and stores recent experiences into long-term memory, is believed to occur predominantly during sleep102,103. It is thought to occur through a process called “replay” or neuronal reactivation. This process engages neuronal populations that were active during learning, which are reactivated shortly after the experience, particularly during subsequent sleep. Such replay has been observed in multiple brain regions, including the hippocampus, neocortex, amygdala, and striatum100.

Reward further enhances this process: for example, expectancy of reward can promote the offline consolidation of skill learning during sleep104. Reactivations in both the hippocampus and the NAc have been shown to increase following reward, suggesting that mesolimbic engagement enhances consolidation101,105,106. A particularly interesting study in this context demonstrated that artificially pairing hippocampal place-cell activity with reward during NREM sleep, via stimulation of the medial forebrain bundle targeting the VTA-NAc mesolimbic pathway, was sufficient to create a spatial memory for a location the animal had never previously associated with reward. This finding provides causal evidence that activation of reward circuits during sleep can drive memory formation and potentially shape future behavior107. However, the precise mechanisms and circuit-level pathways mediating these effects remain largely undefined. The NAc is well positioned to participate in such processes. It shares strong reciprocal connections with the Hippocampus, which is central to new memory formation108. Furthermore, hippocampal place-cell replays were shown to precede NAc reward-related reactivation, suggesting a coordinated interaction between spatial and emotional memory systems during consolidation105,106. Similarly, reactivation in the NAc during sleep has been linked to spontaneous reward-related firing patterns, which are positively modulated by hippocampal sharp-wave ripples105.

In emotional memory formation, the hippocampus works in concert with cortical areas and the amygdala, regions reactivated during REM sleep109. Given its integration into the limbic system and its high activity during REM sleep, the NAc may also contribute to this emotional reprocessing. It is plausible that the NAc participates in transforming emotionally charged experiences into emotionally neutral memories over time, a process believed to occur during REM sleep109. Dopamine, long recognized for its role in memory acquisition110,111 and consolidation112, also modulates synaptic plasticity in the hippocampus113, but also in the striatum114,115,116. In addition, dopamine may influence memory selection by promoting the retention of salient information while facilitating the pruning of less relevant content, aligning with recent findings on active forgetting mechanisms117.

While D1- and D2-MSNs have opposing roles in sleep-wake regulation, with D1-MSNs promoting wake and D2-MSNs promoting NREM sleep, their differential sensitivity to dopamine may also shape sleep-dependent plasticity. Dopamine release during NREM sleep, shown to increase under certain conditions, could act on D2-MSNs to modulate the reactivation of specific neuronal ensembles, possibly contributing to the consolidation or sorting of low-arousal, emotionally salient memories. Conversely, heightened dopamine levels and elevated NAc activity during REM sleep may act predominantly on D1-MSNs, facilitating plasticity mechanisms associated with emotional memory updating or re-evaluation. This suggests that dopamine’s influence on memory processes may be partly dissociable from its role in sleep induction per se, instead modulating the qualitative content and salience of reactivated memories via cell-type-specific pathways. Support for this idea also comes from invertebrate models. Evidence from Drosophila melanogaster has identified similar mechanisms for dopamine-regulated forgetting during sleep. Specific subsets of dopamine neurons in flies regulate forgetting experience of negative valence, and their activity is tightly linked to arousal levels118,119. More specifically, sleep following aversive learning enhances memory retention by suppressing this dopamine-signaling, whereas heightened arousal before sleep accelerates forgetting by increasing the activity of these neurons. These findings suggest a conserved mechanism whereby dopamine interacts with sleep state to regulate memory fate, possibly reinforcing or destabilizing specific traces depending on motivational and arousal signals. This arousal-gated dopamine modulation of memory dynamics may therefore extend to the vertebrate basal ganglia, including the NAc, with cell-type-specific dopamine signaling contributing to selective memory consolidation and forgetting during distinct sleep stages.

Future studies should explore how the NAc contributes to the dynamic interplay between arousal states and memory-related processes during sleep, including consolidation, forgetting, and emotional modulation. Rodent models, combined with circuit- and cell-type-specific approaches, particularly those distinguishing D1- from D2- MSNs, offer a promising avenue to uncover both causal and correlative links between NAc activity and memory processes shaped by sleep. Investigating sleep-dependent synaptic plasticity, including dendritic spine dynamics and synaptic strength alterations in hippocampal and cortical networks, could shed light on these interactions120,121. Overall, such studies may help clarify how motivational and emotional salience shapes the fate of memories during sleep, potentially conditioning future behaviors.

Conclusion

Altogether, the NAc emerges as a central integrator at the intersection of motivational drive, arousal regulation, and sleep-related cognitive functions. Through complementary actions of D1- and D2-MSNs, it supports a dynamic control of behavioral engagement, promoting arousal and motivated wakefulness via D1-MSNs, while facilitating disengagement and NREM sleep initiation via D2-MSNs. This functional dichotomy reflects a state-dependent modulation of accumbal output, in which dopamine and adenosine may play a critical role in tuning activity according to internal needs and external demands.

Although detailed circuit-level mechanisms were addressed in earlier sections, the reviewed evidence shows that MSN subpopulation dynamics, regional specialization (core vs. shell), and projection targets all contribute to the differential roles of D1- and D2-MSNs in vigilance state regulation. Still, critical gaps remain in understanding how these same neuronal populations contribute to sleep-active processes, such as memory consolidation and emotional re-evaluation. In this regard, insights from broader BG literature point to potential functional heterogeneity within both D1- and D2-MSN populations. Such diversity may shape their contributions to arousal modulation and sleep-related processes, including the permissive gating of sleep via D2-MSNs and the possible involvement of D1-MSNs in REM sleep, potentially through specific input–output connectivity. Moreover, traditional stimulation approaches such as optogenetics may not fully replicate the nuanced, state-dependent modulation mediated by endogenous dopamine and adenosine. To address these complexities, future studies should combine temporally precise MSN manipulation with simultaneous EEG and thalamocortical recordings to capture downstream effects and better characterize how NAc-driven activity shapes global brain states.

Dopaminergic signaling during NREM and REM sleep may not only regulate arousal thresholds but also guide plasticity mechanisms within D1- and D2-MSN populations. This suggests that dopamine’s role extends beyond sleep-wake transitions to include the shaping of memory salience and emotional tone during offline states. Whether the NAc operates as a passive relay or actively selects memory traces for consolidation or forgetting remains a key open question. Understanding these mechanisms is especially urgent in the context of psychiatric disorders where motivation, arousal, and sleep are profoundly disrupted. In conditions such as substance use disorder (SUD)122,123,124, schizophrenia125, and depression126, the NAc plays a central role within dysfunctional reward circuitry. SUD, in particular, illustrates how maladaptive learning and motivational processes can become decoupled from the regulation of internal states. Chronic drug exposure is frequently associated with disrupted sleep architecture, including insomnia, sleep fragmentation, difficulties falling asleep, and poor overall sleep quality123,124. More critically, these disturbances often persist into abstinence, with adaptative impairments in REM sleep that remain poorly understood. Such alterations may contribute to cognitive deficits and increase vulnerability to relapse127,128,129. SUD can be conceptualized as a disorder of dysregulated learning, in which the consolidation and reinforcement of drug-associated memories during sleep may heighten the salience of these cues and promote recurrence of addictive behaviors. This process likely involves the brain’s core reward systems, including the NAc. Future studies focusing on how the NAc shapes sleep-dependent plasticity could illuminate novel intervention points, not only for SUD but for a range of disorders where motivation and sleep intersect. Targeting the sleep phase as a window for therapeutic modulation of NAc circuits may offer novel avenues to restore cognitive-emotional balance in disorders marked by disrupted motivation and arousal.