Emerging Functions of Neuromodulation during Sleep

Introduction

Sleep is a fundamental process of animal life, manifested at most system levels, from cellular to organismal (Frank and Heller, 2019; Sulaman et al., 2023). Characterized by behavioral quiescence and reduced responsiveness to internal and external stimuli, sleep encompasses two main states: rapid eye movement sleep (REMS) and non-rapid eye movement sleep (NREMS). These states differ across multiple levels, including oscillatory patterns, body temperature, heart and breathing rates, and muscle tone.

The link between neuromodulator systems and sleep–wake states has been studied for decades. Traditionally, neuromodulators were thought to generate the characteristics of the awake state through their presence and the sleep state through their absence, as well as to control arousal-state transitions (Saper et al., 2001; Jones, 2020). However, while neuromodulator levels generally decrease during sleep, they remain present and even undergo regular fluctuations in extracellular levels at an infraslow rhythm (∼1–2 cycles per minute; Fig. 1a).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Neuromodulatory signaling during sleep. a, Rhythms in neuromodulatory tone control various facets of sleep/wake states. b, The synaptic milieu, which is determined by neuromodulatory tone as well as intrinsic cellular properties, generates system substates. c, There is a complex relationship between various facets of sleep and neuromodulatory levels, which challenges linear causal interpretations.

Monoamines, including noradrenaline/norepinephrine (NE), histamine, serotonin, and dopamine (DA), along with acetylcholine (ACh), are produced by various neuronal populations with long-range projections from the brainstem to the forebrain. Early studies revealed that neuromodulator-producing neurons display arousal-state–dependent alterations in neuronal activity: they are generally active during the awake state, show reduced activation during NREMS, and are mostly silent during REMS (with some exceptions; Eban-Rothschild and de Lecea, 2017; Jones, 2020). Loss- and gain-of-function experiments have further demonstrated that their activation drives sleep–wake transitions and is necessary for behavioral and electrocortical arousal, in addition to controlling neuronal computations, learning, and behavioral performance during wakefulness (Lee and Dan, 2012; Grossman and Cohen, 2022).

The influential “flip-flop” switch model for sleep–wake transitions (Saper et al., 2001) proposed that bidirectional interactions between wake-promoting and sleep-promoting neuronal populations create distinct, mutually exclusive states. State transitions occur when one group suppresses the other, with decreased neuromodulator activity essential for sleep initiation and maintenance. Consequently, neuromodulator activation during sleep was considered sleep-disruptive and indicative of sleep fragmentation.

Until recently, several technical challenges have limited the detailed characterization of neuromodulatory release patterns during sleep. Targeting specific subsets of neurons in heterogeneous brain regions regulating different aspects of sleep has been difficult. Additionally, it was not possible to monitor neuromodulator release with high substance specificity and the temporal and spatial resolution relevant to spontaneous sleep–wake transitions. Recent technological advances now enable cell-type–specific interrogations and recordings, along with real-time monitoring of neuromodulator release using genetically encoded fluorescent sensors (Patriarchi et al., 2018; Feng et al., 2024; Simpson et al., 2024). These advancements have revealed new insights into the temporal dynamics and functions of neuromodulator signaling during sleep, particularly in orchestrating sleep structure, oscillatory activity, and memory consolidation, as detailed bellow.

Neuronal oscillations, which are rhythmic patterns of extracellular electrical activity, act as pacemakers by providing a temporal window to synchronize activity across different brain regions. Neuronal oscillations during NREMS include cortical slow waves including slow (0.1–1 Hz) and delta (1–4 Hz) oscillations], infraslow oscillations (0.01–0.1 Hz), thalamocortical spindles (waxing and waning 10–15 Hz oscillations lasting 0.5–2 s), and hippocampal sharp wave ripples (SWRs; 100–250 Hz oscillations lasting 50–100 ms). REMS neuronal oscillations include hippocampal and cortical (in rodents) theta oscillations (5–9 Hz).

The coordination of neuronal oscillations during sleep is crucial for memory consolidation—the transformation and stabilization of internal representations of experiences over time. Slow waves, spindles, and SWRs play significant roles in this process (Girardeau et al., 2009; Andrillon et al., 2011; Latchoumane et al., 2017; Adamantidis et al., 2019; Rothschild, 2019; Fernandez and Lüthi, 2020). The bidirectional flow of information between the hippocampus and the cortex—particularly during SWR events within specific phases of slow oscillations and spindles—is critical for memory consolidation.

Until recently, our understanding of how neuromodulation, neuronal oscillations, and memory consolidation during sleep are interconnected has been limited. Recent studies suggest that neuromodulators shape neuronal oscillatory activity during NREMS, including the suppression of SWRs, slow waves, and spindles (Aston-Jones and Bloom, 1981; Eschenko et al., 2012; Vandecasteele et al., 2014; Ma et al., 2020; Turi et al., 2024). This capacity, combined with the infraslow fluctuations of extracellular neuromodulator levels, times oscillatory events to specific phases of NREMS, consequently generating different substates (Fig. 1b). Moreover, multiple neuromodulators can be released in the same brain region during NREMS, with their extracellular levels being synchronized or unsynchronized at different times (Jones, 2020; Zhang et al., 2024). This phenomenon further segregates NREMS into distinct substates that may serve different functions, characterized by the neuromodulatory milieu, brain oscillatory signature, and potentially other features (Fig. 1b). Together, these findings provide a novel understanding of the complexity of sleep states.

Our goal in this review is to highlight novel insights, primarily derived from research on adult rodents, into the roles of NE, DA, and ACh in controlling various aspects of sleep. These topics were discussed at our 2024 Society for Neuroscience Mini-Symposium. The key insights are that (1) neuromodulator levels are not merely low during NREMS but display regular fluctuations that organize other aspects of sleep across the brain and body, such as spindle oscillations and heart rate; (2) neuromodulator levels control NREMS substates, which are associated with different probabilities of transitioning into wakefulness or REMS; (3) the release of certain neuromodulators can regulate the levels of other neuromodulators during sleep; and (4) neuromodulators participate in memory consolidation processes during sleep (Fig. 1a,b).

The Noradrenergic System

Locus coeruleus noradrenergic (LC^NE) neurons are widely recognized as a key arousal-promoting system, facilitating rapid activation of the brain and autonomic nervous system through their extensive projections. Activity in LC^NE neurons is strongly correlated with levels of behavioral and autonomic arousal and can induce sleep–wake transitions, maintain wakefulness, promote cortical activation, and facilitate motor activity during wakefulness (Fung et al., 1991; Carter et al., 2010; C. W. Berridge et al., 2012; Burgess and Peever, 2013; Liang et al., 2021). LC^NE neurons exhibit tonic and phasic activation during wakefulness, intermittent or low activation during NREMS, and no firing during REMS (Aston-Jones and Bloom, 1981). While the roles of LC^NE neuronal activity in attention, stress, sensory-motor processing, and other waking processes are well established (Sara, 2009; Poe et al., 2020), until recently, the release of NE during sleep was believed to be disruptive to sleep processes, mainly causing sleep fragmentation and awakening. This view has gradually changed following the identification of intermittent burst firing in LC^NE neurons during NREMS. The bursts in activity occur approximately once or twice per minute, generating an infraslow rhythm in extracellular NE levels. Subsequently, different research groups have explored the consequences of these rhythmic changes in extracellular NE levels on neuronal oscillatory activity during NREMS, autonomic state, and sleep microstructure and functions.

LC^NE Function in Controlling Neuronal Oscillatory Patterns during Sleep, Microarousals, and Memory Consolidation

Early studies have identified that during NREMS, LC^NE neurons demonstrate slow oscillations in their bursting activity, which are coupled with cortical slow waves and spindles (Aston-Jones and Bloom, 1981; Eschenko et al., 2012). Additionally, there has been evidence supporting the role of LC neuronal activity during sleep in memory consolidation (Gais et al., 2011; Novitskaya et al., 2016). Together, these findings suggested that not only is NE signaling during sleep not absent, but it also controls important functions of sleep. Yet, the precise dynamics of LC^NE activity and its precise functions during NREMS were, until recently, not well understood. Moreover, while LC activity during NREMS has been previously implicated in spindle termination (Aston-Jones and Bloom, 1981; Novitskaya et al., 2016; Lecci et al., 2017; Swift et al., 2018), the temporal organization of this interaction remained unclear.

To address these knowledge gaps, the Nedergaard and Lüthi groups have recently employed the genetically encoded fluorescent noradrenaline sensor, GRAB_NE, along with fiber photometry and electroencephalogram (EEG)/electromyogram (EMG) recordings in freely behaving and sleeping mice. They demonstrated that LC^NE neurons and NE levels in the medial prefrontal cortex (mPFC) and somatosensory thalamus exhibit regular phasic activity during NREMS, creating an infraslow oscillation of extracellular NE levels at ∼1 cycle per minute (Osorio-Forero et al., 2021; Kjaerby et al., 2022). These infraslow oscillations were found to be anticorrelated with the occurrence of sleep spindles, which are often identified by increases in sigma band power (10–15 Hz; Osorio-Forero et al., 2021, 2024). Specifically, the peak of NE infraslow oscillations was correlated with low spindle power, while the trough was associated with high spindle power. NE depolarizes thalamocortical and thalamic reticular neurons, suppressing burst firing and the capacity of thalamic circuits to generate sleep spindles (Fernandez and Lüthi, 2020). Subsequent optogenetic manipulations revealed the causal role of LC^NE neurons and their projections to the somatosensory thalamus in controlling spindle clustering by actively suppressing spindles via NE release and clustering spindles to coincide with periods of low NE activity. This generated a ∼50-s-long rhythm composed of spindle-rich and spindle-poor phases. Notably, prolonged mPFC NE descents were followed by significant LC^NE cell body activation and prolonged awakenings. In contrast, shorter mPFC NE descents were followed by more subtle LC^NE activation and microarousals—defined as epochs of desynchronized EEG and elevated EMG activity lasting under 15 s—rather than full awakenings (Kjaerby et al., 2022). Together, these findings suggest that infraslow oscillations in extracellular NE levels result in the clustering of sleep spindles into defined intervals, potentially facilitating spindle-related functions.

Recently, the Lüthi group has further explored the relationship between LC^NE activity and microarousals (defined as epochs of desynchronized EEG and elevated EMG activity lasting under 12 s). Osorio-Forero et al. (2024) found that only 30% of the surges in LC^NE activity during NREMS were associated with concurrent EEG desynchronization and elevated EMG activity, indicative of microarousals, as also reported by Kjaerby et al. (2022). LC^NE surges that coincided with a microarousal were characterized by electrophysiological signatures of cortical, cardiac, and thalamic activation. In contrast, surges not associated with microarousals involved cardiac and thalamic activation but cortical deactivation. These findings imply the existence of at least two distinct states during NREMS that are generated by LC^NE activation patterns (Osorio-Forero et al., 2024). This rhythmic activation of LC^NE neurons during NREMS imposes an autonomic, subcortical, and cortical rhythm, with broader consequences for physiology and sleep functions.

To further explore the functions of NE-mediated spindle clustering in memory consolidation during sleep, Kjaerby et al. (2022) utilized optogenetic techniques. They inhibited LC^NE neurons, thus generating cortical NE descents, during post-learning sleep. They employed the “novel object recognition” task and discovered that suppressing LC^NE activity at regular intervals during post-learning sleep enhances the capacity to recognize the novel object (Kjaerby et al., 2022). Additionally, the magnitude of sigma power during sleep, when subjected to LC^NE suppression, correlated with subsequent memory recall in individual animals. Notably, closed-loop optogenetic stimulation of LC^NE neurons, which decreases the amplitude of cortical NE descents, led to reduced sigma activity and decreased memory recall, suggesting the critical role of NE descent amplitude in spindle clustering and memory consolidation during sleep. Altogether, these findings suggest that LC-driven NE oscillations create temporal windows during NREMS that promote the synchronization of neuronal oscillations and facilitate memory consolidation processes, highlighting an important function for neuromodulatory signaling during sleep.

Given the central role of LC^NE activity in regulating sleep structure and the stress response, two recent studies (Antila et al., 2022; Osorio-Forero et al., 2024) have investigated whether stress-induced modifications in LC^NE firing patterns disrupt sleep. Osorio-Forero et al. (2024) used a stimulus-enriched sleep deprivation paradigm in which mice were exposed to diverse, mildly aversive sensory stimuli that increased their corticosterone levels. Conversely, Antila and colleagues employed a model of acute social stress. Both studies found that subjecting mice to stress prior to sleep initiation increased transient activation of LC^NE neurons and disrupted sleep, characterized by increased microarousals and suppressed REMS. Furthermore, both the optogenetic (Osorio-Forero et al., 2024) and chemogenetic (Antila et al., 2022) inhibition of these neurons after stress effectively suppressed these sleep disturbances, suggesting that abnormal LC^NE activity following stress leads to sleep quality deterioration. Together, these findings demonstrate that natural fluctuations in LC^NE activity are susceptible to prior acute stressful experiences, which can impact post-stress sleep quality.

In addition to stress, LC^NE neuronal functions are crucial in aging and neurodegenerative disorders, which often feature sleep disturbances such as increased microarousals and decreased spindle density and clustering (Bonnet and Arand, 2007; Cooke and Ancoli-Israel, 2011; Martin et al., 2013; Champetier et al., 2023). Loss of LC neurons is an early marker of neurodegenerative disorders (Mann et al., 1980; Grudzien et al., 2007), and maintaining LC neural density can reduce neurodegeneration (Wilson et al., 2013). Despite cell loss, the remaining LC neurons may become hyperactive (Engberg et al., 1987; Raskind et al., 1999), potentially disrupting the NE oscillatory pattern controlling spindle clustering and contributing to memory impairments.

LC^NE Function in Organizing NREMS–REMS Cycles

The sleep phase is characterized by cyclic alternations between NREMS and REMS episodes, interspersed with either short or long wake bouts. In healthy individuals, the transition into REMS exclusively occurs from NREMS. While much is known about the neuronal regulation of NREMS and REMS separately, the mechanisms controlling the NREMS–REMS cycle remain largely unknown. REMS is generated by complex interactions among distributed neuronal populations, with LC^NE neurons long suggested to function as REMS gatekeepers (McCarley and Hobson, 1975; Brown et al., 2012), preventing its onset during wakefulness and NREMS (Saper et al., 2010; Sulaman et al., 2023). Nonetheless, the precise mechanisms permitting a gatekeeping function were, until recently, undetermined. It was known that the initiation of REMS is dependent on low LC^NE activity (McCarley and Hobson, 1975; Brown et al., 2012; Osorio-Forero et al., 2021, 2024; Antila et al., 2022; Kjaerby et al., 2022) and the optogenetic activation of LC^NE neurons, even during periods of high REMS pressure following REMS restriction, prevents REMS initiation (Osorio-Forero et al., 2024). Additionally, optogenetic inhibition of LC^NE neurons during NREMS has been found to decrease mPFC NE levels, increase sigma power, and facilitate transitions into REMS (Cardis et al., 2021; Kjaerby et al., 2022).

Utilizing a closed-loop optogenetic strategy, Osorio-Forero and colleagues demonstrated that stimulating LC^NE neurons during the rising phase of sigma oscillations—which corresponds to the trough of LC^NE oscillations—reduced NREMS-to-REMS transitions (Osorio-Forero et al., 2024). Conversely, optogenetic inactivation of LC^NE neurons during sigma troughs–which correspond to LC^NE activity peaks–led to an increase in NREMS-to-REMS transitions. Notably, the selective inhibition of LC^NE activity either early in NREMS epochs (within 100 s after NREMS onset) or later (after NREMS duration exceeded twice the length of the preceding REMS bout) revealed a temporal dimension to LC function in REMS transitions: only inhibitions during the later stages of NREMS facilitated NREMS-to-REMS transitions, while earlier inhibition had no effect (Osorio-Forero et al., 2024). These findings suggest that transiently low LC activity during a specific phase of NREMS episodes creates windows of opportunity for transitions into REMS. Notably, this work significantly advances our current understanding of the function of LC^NE neurons as REMS gatekeepers. It demonstrates that “gate opening” alone is not sufficient for a NREMS-to-REMS transition; the rest of the system must also be prepared for the transition to take place. It remains to be elucidated what other factors need to properly align during late NREMS in order for NE descent to facilitate a NREMS-to-REMS transition. Osorio-Forero et al. (2024) also found that under elevated REMS pressure, the interval between REMS bouts stabilizes at ∼50 s, matching the period of infraslow oscillations in LC^NE activity. This suggests that these natural fluctuations during NREMS help set a minimum inter-REMS interval. Together, these findings underscore a role of LC^NE neurons in regulating the NREMS–REMS cycle and suggest a potential function for LC^NE activity in partitioning NREMS into two alternating brain-autonomic states: one that is permissive to REMS and another that promotes wakefulness.

LC^NE Function in Sensory Processing during Sleep

Sleep is marked by reduced responsiveness to the external environment, manifested as an elevated arousal threshold. While animals show little behavioral response to sensory stimuli during sleep, they can promptly rouse to salient ones, such as unexpected and novel sounds (Neckelmann and Ursin, 1993). This capacity provides significant fitness benefits, as failure to mount an arousal response to cues predicting salient events, such as danger, could be detrimental.

The ability to respond selectively to stimuli during sleep suggests an active salience processing system. While sensory processing during wakefulness is well understood, it is less so during sleep. Though sensory cortices are responsive during sleep, the processing of sound salience and the generation of wakefulness were, until recently, less understood. Recently, LC^NE neurons have emerged as key regulators of sensory-evoked arousal from sleep.

As mentioned, LC^NE neurons show strong activation to salient stimuli during wakefulness, supporting autonomic and behavioral responses (Uematsu et al., 2017; Gelbard-Sagiv et al., 2018; Bornert and Bouret, 2021). In 2020, Hayat et al. (2020) demonstrated that LC^NE neurons regulate sound-evoked awakening from sleep. Bidirectional optogenetic manipulations revealed that LC^NE neurons controlled the propensity to arouse following sound delivery (Hayat et al., 2020). Furthermore, they found that the probability of sound-evoked awakening from sleep was influenced by the activity of LC neurons prior to sound delivery; higher tonic activity resulted in a higher probability of arousal (Hayat et al., 2020). Together, these findings expand our understanding of LC functions during sleep, suggesting that LC neurons continue to serve as central regulators of sensory processing and arousal; however, the specific circuitry involved remains elusive.

An intriguing question remains about the extent to which LC^NE activity is homogeneous especially in the context of its roles in arousal and sleep functions. NE is secreted from varicosities along unmyelinated, widely distributed axons of highly interconnected neurons. Individual LC neurons can integrate information from, and project to, many reciprocally interconnected brain regions (Nagai et al., 1981; Room et al., 1981; Schwarz et al., 2015). Electrical coupling can also promote synchronous LC activity (Alvarez et al., 2002; Rash et al., 2007). While LC neuromodulation has traditionally been regarded as homogenous (Aston-Jones and Waterhouse, 2016), recent studies have identified some segregation of LC outputs and related functional diversity (Chandler et al., 2019). For instance, spinal-projecting LC neurons exert analgesic actions, whereas ascending LC projections have a pronociceptive effect (Hickey et al., 2014; Hirschberg et al., 2017). Similarly, LC neurons projecting to the prefrontal versus motor cortex differ in their molecular profiles, excitability, and activity across vigilance states (Bellesi et al., 2016). Additionally, LC neuronal firing has been shown to be only sparsely synchronized (Totah et al., 2018). These findings raise the question of whether projection-specific LC^NE subpopulations differentially regulate responsiveness during sleep. To address this, ongoing work in the Nir laboratory has utilized circuit-specific fiber photometry recordings of GRAB_NE, in conjunction with optogenetic manipulations and EEG/EMG recordings in freely behaving and sleeping mice. The research focuses on forebrain and brainstem projections, which are extensively innervated by LC^NE neurons and implicated in arousal regulation (C. W. Berridge et al., 2003; Dergacheva et al., 2004). The Nir group discovered that projection-defined subpopulations of LC^NE neurons are partially distinct regarding their activity and function and differentially regulate sensory-evoked awakenings from sleep, highlighting a more nuanced and specialized role of LC^NE functions during sleep.

To summarize, significant conceptual advancements have been made in recent years in our understanding of NE activation during sleep. It has become clear that LC^NE neurons not only control sleep–wake transitions but also partition NREMS into different substates, each associated with distinct brain-autonomic activation and varying probabilities of transitioning into other states. The infraslow rhythm in extracellular NE levels during sleep imposes an autonomic, subcortical, and cortical rhythm, with broader consequences for physiology and sleep functions (Fig. 1a).

The Dopaminergic Systems

DA is another key neuromodulator implicated in sleep–wake regulation, in addition to playing central roles in motivated behaviors, movement, learning, and even immune responses (K. C. Berridge and Robinson, 1998; Wise, 2004; Hosp et al., 2011; Ben-Shaanan et al., 2016; Hughes et al., 2020). DA is produced by several neuronal populations distributed from the hypothalamus to the brainstem. Among these, the ventral tegmental area (VTA^DA) and the dorsal raphe/ventral periaqueductal gray matter (DR^DA/vPAG^DA) populations are especially implicated in sleep–wake regulation. Both populations target cortical and subcortical areas. Early pharmacological and lesion studies have suggested a role of DA in wake promotion, NREMS suppression, and a nuanced function in REMS (Khanday et al., 2016). These neurons are strongly activated during wakefulness and show reduced activity during NREMS (Lu et al., 2006; Eban-Rothschild et al., 2016; Cho et al., 2017). Moreover, the optogenetic and chemogenetic activation of VTA^DA and DR^DA neurons induce sleep–wake transitions and maintain wakefulness (Eban-Rothschild et al., 2016; Cho et al., 2017; Oishi et al., 2017). However, while DR^DA neuronal activity is suppressed during REMS, VTA^DA neurons exhibit burst firing (Dahan et al., 2007; Eban-Rothschild et al., 2016; Cho et al., 2017). Though research into the functions of dopaminergic activity during sleep has been sparse, potential roles suggested for VTA^DA neurons include the regulation of REMS initiation and memory consolidation. The functions of DR^DA neurons during sleep await further investigations.

VTA^DA Function in REMS

The function for VTA^DA neurons in REMS was suggested by the observed reduction in REMS duration following the pharmacological blockade of DA signaling pathways (Dzirasa et al., 2006; Lima et al., 2008). In 2016, Eban-Rothschild and colleagues revealed that VTA^DA neurons are not only strongly activated during REMS, as reported by Dahan and colleagues (Dahan et al., 2007), but also exhibit increased population activity ∼20 s before the transition into REMS (Eban-Rothschild et al., 2016), further implicating VTA^DA neurons in the regulation of NREMS-to-REMS transitions. However, the causal consequences of this ramping activity and the circuitry involved were undetermined.

In 2022, Hasegawa and colleagues explored the role of DA in REMS and discovered a function for the VTA^DA–basolateral amygdala (BLA) circuitry in the state generation (Hasegawa et al., 2022). Utilizing the GRAB^DA sensor, they found that DA release in the BLA, but not in the lateral hypothalamus and mPFC, increases prior to REMS initiation. Using optogenetics, they demonstrated that stimulation of VTA^DA-BLA projections reduces the latency to REMS and increases its duration, while inhibition of these projections increases REMS latency and reduces its duration. Since the latency from VTA^DA–BLA stimulation to REMS initiation was prolonged (over a minute), the precise actions induced by DA leading to the eventual transition remain to be elucidated. It would be intriguing to examine whether, similar to NE, DA can facilitate the transition only during a specific substate of NREMS (e.g., during NE descent). Moreover, while the VTA is innervated by REMS regulatory populations [such as those residing in the LC, lateral hypothalamus, basal forebrain (BF), and pedunculopontine tegmental (PPT) and laterodorsal tegmental (LDT) nuclei], the functions of these in REMS initiation and maintenance, as well as the causal role of SNc^DA neurons, await further investigation. Together, these findings highlight a novel understanding of DA's functions in sleep–wake regulation beyond wake promotion, implicating it in the NREMS-to-REMS transition (Fig. 1a).

VTA^DA Function in Memory Consolidation

DA signaling has long been implicated in learning processes, from the acquisition of associative and motor skill learning to the synaptic plasticity mechanisms necessary for retaining memories (Wise, 2004). Although post-learning DA signaling has been suggested to facilitate memory consolidation, the necessity of DA during post-learning sleep to memory consolidation processes was undetermined. Several lines of evidence suggest the role of VTA^DA neurons in these processes. VTA neurons have been found to undergo reactivation of taste-related activity during NREMS (Valdés et al., 2015). In addition, VTA^DA activity during wakefulness influences hippocampal reactivation events during NREMS and potentially the rate of hippocampal SWRs (McNamara et al., 2014; Karunakaran et al., 2016). Lastly, pharmacological blockade of DA signaling following learning impairs long-term memory formation (Rossato et al., 2009; Gonzalez et al., 2014).

Ongoing work in the Eban-Rothschild laboratory is investigating the functions of DA signaling during sleep, employing fiber photometry recordings of VTA^DA neurons, optogenetic and chemogenetic manipulations, and EEG/EMG along with local field potential (LFP) recordings in freely behaving and sleeping mice. Their findings suggest that VTA^DA neuronal activity during NREMS is modulated by positive, but not negative, valence waking experiences, including those involving motor skill learning and place–reward association. Moreover, they found evidence suggesting a role for VTA^DA activity in motor memory consolidation during sleep, but not place–reward association. To further investigate this link between VTA^DA neuronal activity and memory consolidation processes, the laboratory is examining how VTA^DA neuronal activity relates to hippocampal and cortical oscillatory patterns associated with memory consolidation. Preliminary data suggest that VTA^DA activity is anticorrelated with both SWRs and spindle events, similar to findings with NE in relation to spindles and ACh in relation to SWRs. While VTA^DA neurons (Dahan et al., 2007; Fujisawa and Buzsáki, 2011) and DA release in the striatum (Krok et al., 2023) have been found to show fluctuations in the delta range (∼4 Hz), their functions in synchronizing other oscillatory events during sleep remain to be elucidated. Together, this work suggests that VTA^DA neuronal activity during NREMS is experience-dependent, is modulated around memory-related oscillatory patterns, and may facilitate memory consolidation processes (Fig. 1a).

Altogether, recent research has revealed that dopaminergic signaling is consequential not only during wakefulness but also during sleep. As alterations in dopaminergic signaling are linked to various pathological conditions, including Parkinson's disease, depression, and schizophrenia (Klein et al., 2019), a better understanding of DA functions during sleep could illuminate overlooked consequences of their dysregulation. These insights have the potential to significantly improve treatment options and the well-being of individuals affected by these disorders.

The Cholinergic Systems

ACh is an additional key neuromodulator that has long been known to participate in sleep–wake regulation, as well as regulating various cognitive processes (Hasselmo and Giocomo, 2006; Ananth et al., 2023). ACh is produced by neurons in the BF, LDT, and PPT nuclei. ACh neurons are strongly activated during wakefulness and REMS but show little activation during NREMS (Sakai, 2012; Boucetta et al., 2014; Ma et al., 2020). Moreover, ACh signaling controls hippocampal oscillations, synaptic plasticity, and hippocampal-dependent memory (Hasselmo, 2006). BF^ACh neurons suppress cortical synchronization and delta activity during wakefulness, and their optostimulation induces cortical activation and NREMS-to-wake transitions (Han et al., 2014; Irmak and de Lecea, 2014; Xu et al., 2015). Moreover, cholinergic signaling during sleep has been suggested to facilitate the initiation and maintenance of REMS (Grace et al., 2014; Van Dort et al., 2015; Lohani et al., 2022). However, the dynamics of ACh release at different brain regions, as well as its relationship with neuronal oscillatory patterns during sleep, remain poorly understood. Recent studies by the Buzsáki group addressed these knowledge gaps by recording ACh release, as well as single-unit and LFP signals, from the hippocampus to characterize the nature of neuromodulation during sleep in a central brain region implicated in memory consolidation (Zhang et al., 2021, 2024). Moreover, since the hippocampus is innervated by multiple neuromodulatory systems, including the oxytocin (OXT) system—which is critical for hippocampus-dependent social memory (Raam et al., 2017)—they further characterized the temporal relationship between ACh, OXT, and hippocampal oscillations associated with memory consolidation (Zhang et al., 2024).

Dynamics of ACh and OXT Levels in the Hippocampus during NREMS

The hippocampus, innervated by BF medial septum (MS)^ACh neurons, plays a crucial role in forming short-term memory during wakefulness and transferring these memories to long-term storage in the cortex during sleep (Buzsáki and Moser, 2013). Hippocampus-dependent memories, particularly episodic (in humans) or spatial (in rodents), are typically sleep-dependent (Krause et al., 2017; Rothschild, 2019; Pronier et al., 2023; Giri et al., 2024). Although much is known about neuromodulatory control of hippocampal functions during wakefulness, its impact during sleep remains poorly understood. Since neuromodulators can control neural oscillations and influence system states, understanding their temporal dynamics and interactions with neural oscillations is crucial for comprehending how the hippocampus processes and consolidates various types of experiences during sleep.

To characterize the temporal relationship between neuromodulator levels, sleep–wake states, and oscillatory patterns, the Buzsáki laboratory utilized GRAB sensors to record extracellular ACh and OXT levels in the dorsal hippocampus, along with electrical recordings in freely moving mice. While ACh and OXT levels were found to be anticorrelated during wakefulness and REMS, the signals were positively correlated during NREMS (Zhang et al., 2024). Cross-correlation analysis further revealed that ACh modulation precedes OXT by ∼2 s. The extracellular levels of both neuromodulators fluctuated with an infraslow rhythm (∼1 cycle per 40 s) and were relatively low during NREMS episodes. However, while ACh levels were modulated around transitions into and out of NREMS, OXT levels exhibited a gradual, continuous decrease during NREMS (Zhang et al., 2024). Together, these results suggest a differentiation in ACh and OXT neuromodulation in the hippocampus during wakefulness, REMS, and arousal-state transitions.

Bidirectional Interactions between ACh, OXT, and Hippocampal Oscillatory Activity

Optogenetic manipulations have revealed that the activation of MS^ACh neurons during NREMS can suppress ripples, spindle power, and slow oscillations while promoting theta oscillations (Vandecasteele et al., 2014; Ma et al., 2020). Zhang et al. (2024) demonstrated that hippocampal ACh levels are anticorrelated with SWR power during NREMS, with SWRs occurring during the troughs of the ACh oscillatory phase. Furthermore, by optogenetically inhibiting MS^ACh neurons, the group established a causal link between ACh signaling and hippocampal oscillatory activity (Zhang et al., 2024). Both hippocampal ACh and OXT levels were found to be anticorrelated with sigma power during NREMS. Notably, the effects of ACh levels on hippocampal gamma oscillatory activity depended on the sleep–wake state: ACh surges during NREMS coincided with decreased gamma power, while during wakefulness, they coincided with increased gamma power. Together, these findings highlight a function for neuromodulators in the hippocampus during NREMS in controlling oscillatory activity related to memory consolidation, similar to the relationship found for NE and DA with spindles and SWRs, and highlight the influence of the sleep–wake state in determining the effects of neuromodulators on network activity (Fig. 1a,b).

In addition to the capacity of the neuromodulators to control oscillatory hippocampal activity, the latter seems to modulate extracellular levels of the neuromodulators themselves (Fig. 1c). Fluctuations in SWR power were found to precede the troughs and peaks of hippocampal OXT levels during NREMS, with OXT levels reaching a minimum ∼2 s after SWRs (Zhang et al., 2024). Likewise, during wakefulness increases in synchronous hippocampal population activity preceded decreases in OXT levels by a similar timescale. Furthermore, the induction of artificial ripples during NREMS leads to a partial reduction in hippocampal OXT levels that is dependent on activity in the lateral septum, suggesting a causal relationship between synchronous hippocampal activity and reduced OXT levels. Notably, the optogenetic stimulation of MS^ACh neurons also decreased OXT levels in the hippocampus, suggesting that levels of one neuromodulator may influence levels of another. These findings indicate that changes to synchronous population activity in the hippocampus influence OXT levels during NREMS. Furthermore, the data suggest that ACh may modulate OXT release through a long-loop pathway involving the hippocampus and the lateral septum.

In summary, these findings underscore the state-specific effects of neuromodulators on hippocampal activity and provide important insights into the intricate relationships among neuromodulators and neuronal oscillatory activity during sleep (Fig. 1). The role of neuromodulatory coordination in the hippocampus during NREMS, and its relation to subsequent circuit and behavioral functions, remains to be fully understood.

Serotonin and Histamine

While the functions of serotonin and histamine are outside the scope of our minisymposium, it is worth noting that these two monoamines have long been implicated in sleep–wake regulation, although their precise roles remain under discussion (Fujita et al., 2017; Oikonomou et al., 2019; Venner et al., 2019). Recent findings revealed that extracellular serotonin levels in the hippocampal dentate gyrus (DG) fluctuate at an infraslow rhythm (∼1 cycle per minute), are anticorrelated with DG glutamatergic neuronal activity, and are modulated around microarousals (<15 s of desynchronized EEG and EMG movement; Turi et al., 2024). As with NE, ACh, and DA, synchronous neuronal oscillatory activity occurs during the troughs of serotonin fluctuations. Further research is needed to determine the functions of serotonin signaling during NREMS.

Conclusion and Future Directions

Recent findings have unveiled the nuanced dynamics and functions of neuromodulatory signaling during sleep, highlighting their roles in shapeing sleep processes such as the orchestration of neuronal oscillatory patterns, control of sleep architecture, regulation of behavioral responsiveness, and facilitation of memory consolidation (Fig. 1a).

Contrary to the traditional view that neuromodulatory tone is merely low or absent during NREMS, several neuromodulatory systems exhibit infraslow oscillations during NREMS, with various aspects of sleep occurring at specific phases of these oscillations (Fig. 1a). A common finding across systems is that synchronous neuronal oscillatory activity, such as spindles and SWRs, occurs during the troughs of oscillations in extracellular neuromodulator levels. These fluctuations in neuromodulator levels constrain the timing of oscillatory events to specific phases of their release patterns, generating distinct substates of NREMS (Fig. 1b). Moreover, multiple neuromodulators can be released in the same brain region during NREMS and can regulate each other, resulting in altered network/synaptic activity. The extracellular levels of different neuromodulators can be either synchronized or unsynchronized at different times during NREMS, further generating distinct substates. Together, these findings provide a novel understanding of the complexity of sleep states, which, rather than being homogeneous, contain several substates. These substates are partially defined by the neuromodulatory milieu (Fig. 1b). While most research to date has focused on NREMS substates, expanding this line of research to REMS, which is known to include at least two substates, would be intriguing. Additionally, the relationship between neuromodulation during NREMS and subsequent REMS and a broader characterization of NREMS neuromodulation across different developmental stages, brain regions, and substances (such as melanin-concentrating hormone and hypocretin) warrants further investigation.

A major challenge in the field is the extensive interconnectedness between systems. For example, NE regulates arousal, microarousals, oscillatory activity (spindles, slow waves, delta oscillations), and the parasympathetic system (Fig. 1c). Furthermore, findings suggest that behaviors or neuronal oscillations may also influence neuromodulatory signaling, indicating bidirectional interactions (Fig. 1c). This interconnectedness challenges the use of gain-/loss-of-function manipulations to establish linear causal relationships, as changes in one aspect affect all others. Future research should disentangle these relationships or acknowledge their complexity.

Lastly, a few open questions and considerations: while progress has been made in understanding neuromodulators’ influence on circuit function during sleep, our understanding of the synaptic and cellular consequences remains limited. Moreover, we know little about interactions between different neuromodulators in specific brain regions on fast timescales. Given the heterogeneous nature of neuromodulatory neuronal populations, it would be beneficial to further dissect the contributions of subpopulations to different phenomena. Finally, the sleep field will benefit from clearly defining the term “microarousals,” which various research groups use differently.