Synchrony of neuromodulatory systems during NREM sleep

Significance

Neuromodulatory systems regulate arousal and sleep. Dysfunctions in these systems have been implicated in neurological and psychiatric disorders. However, their interactions during sleep have not been directly studied in vivo. Here, we show that the releases of three major neuromodulators, norepinephrine, serotonin, and acetylcholine are synchronized in the infraslow frequencies during non–rapid eye movement (NREM) sleep. The level of synchrony increased with the increased level of arousal from NREM sleep. Our pharmacological and optogenetic experiments further showed that the release pattern of one neuromodulator is interdependent on others. Together, our study revealed an important interaction among different neuromodulatory systems during sleep. This interaction might provide a mechanism for the coordinated regulation of sleep and sleep-related neural processes.

Abstract

Neuromodulatory systems play an essential role in regulating brain states and functions. The canonical view is that the release of monoamines including norepinephrine (NE) and serotonin (5-HT) is high during wakefulness and attenuated during sleep, particularly during rapid eye movement (REM) sleep, whereas the cholinergic system is active during both wakefulness and REM sleep and quiescent during non–REM (NREM) sleep. Recent studies have revealed a slow and rhythmic release pattern of neuromodulators during NREM sleep that drives infraslow oscillation (ISO) (0.02 to 0.03 Hz) in the brain. A key question is whether/how the release of different neuromodulators during sleep is coordinated. In this study, we combined 2-site fiber photometry with electroencephalogram/electromyography recording to monitor the release of NE, 5-HT, and acetylcholine in the cortex and hippocampus during sleep and wake cycles. We found that the ISO of these neuromodulators is synchronized during NREM sleep. The synchrony between neuromodulatory systems increases in the oscillatory cycles leading to arousal. Furthermore, pharmacological inhibition of either the 5-HT or NE system eliminates the oscillation of other neuromodulators during NREM sleep. Optogenetic activation of 5-HT or NE neurons during NREM sleep induces the release of other neuromodulators in the absence of sleep-to-wake transitions. These results suggest that the rhythmic neuromodulator releases are highly coordinated in the brain. The synchrony among multiple neuromodulatory systems across brain regions provides a powerful neural mechanism to orchestrate sleep architecture and sleep-related neural processes.

Neuromodulatory systems are networks of neurons that release neuromodulators to globally regulate the activity of widespread brain regions. These neuromodulators include norepinephrine (NE) primarily released by neurons in the locus coeruleus (LC), serotonin (5-HT) mostly by the raphe nuclei, dopamine (DA) mostly by the ventral tegmental area (VTA), histamine by the tuberomammillary nucleus (TMN), and acetylcholine (ACh) by the basal forebrain (BF) and other cholinergic nuclei in the brainstem (1). Despite their localized origins, these systems diffusely innervate the cerebral cortex, hippocampus, thalamus, hypothalamus, and many other regions (2–5), and they are essential for regulating a range of behavioral and physiological processes including motivation, emotion, and cognition (5–8). Dysregulation of these systems has been implicated in many neurological and psychiatric disorders, including Parkinson’s and Alzheimer’s diseases, depression, and anxiety (7, 9–12). Thus, understanding the patterns of neuromodulator release is important for both basic and translational research.

One of the primary functions of Neuromodulatory systems is the regulation of brain states, particularly sleep and wakefulness (1, 13). For instance, phasic optogenetic stimulation of LC-NE neurons induces immediate transition from sleep to wakefulness (14). Interestingly, a more recent study demonstrates that repeated optogenetic activation of LC-NE neurons can induce rebound sleep following transient wakefulness, suggesting its role in regulating homeostatic sleep pressure (15). Similarly, optogenetic stimulation of dorsal raphe (DRN) neurons has bidirectional mode-dependent effects on sleep in mice: Burst stimulation causes immediate wake while tonic stimulation slowly promotes sleep (16). Together, these findings underscore the dynamic role of monoaminergic systems in regulating sleep and wakefulness.

Despite their complexity, a consistent observation over decades is that both NE and 5-HT neurons are most active during wakefulness, less active during non–rapid eye movement (NREM) sleep, and largely silent during REM sleep (16–19). Recent advances in genetically encoded neuromodulator sensors enable the measurement of neuromodulator releases in vivo during sleep and wake cycles, which reveal a novel infraslow oscillation (ISO, 0.02 to 0.03 Hz) of neuromodulator release during NREM sleep, which is partially associated with microarousals (MA) (20). For instance, several groups have identified the ISO of NE release or LC-NE neuronal activity in various brain regions, including the locus coeruleus (LC) (19, 21), the thalamus (22), and the medial prefrontal cortex (mPFC) (23). Our recent study revealed a similar ISO of 5-HT release during NREM sleep in the hippocampus and raphe nuclei (24). These ISOs of neuromodulatory systems during NREM sleep are thought to be the drive of the ISO of electroencephalogram (EEG) signals, particularly in the sleep spindles-related sigma power range (22, 23, 25, 26).

In contrast to the monoaminergic systems, the cholinergic system exhibits a more consistent role in promoting arousal (1). For instance, optogenetic or chemogenetic activation of cholinergic neurons in the BF reliably promotes wakefulness (27, 28). Unlike the monoaminergic systems, BF cholinergic neurons display increased activity during both wakefulness and REM sleep, compared to NREM sleep (27, 29, 30). Notably, Zhang et al. recently identified an ISO pattern of ACh release during NREM sleep in the hippocampus (31).

The striking similarity in ISO patterns across different neuromodulators raises an intriguing possibility that their releases during sleep are highly coordinated in the brain. To test this hypothesis, we combined 2-site fiber photometry with EEG/EMG recording to examine the release pattern of NE, 5-HT, and ACh during sleep and wake cycles. Our results demonstrate that the infraslow release of these neuromodulators during NREM sleep are tightly synchronized across cortical and hippocampal regions, particularly in the oscillatory cycles leading to MA/wakefulness. Furthermore, pharmacological inhibition of either the NE or 5-HT system impaired the ISO of other neuromodulators during NREM sleep, whereas optogenetic activation of NE or 5-HT neurons during NREM sleep induced the release of other neuromodulators without changing brain states. Together, these results suggest the coordination and interdependence among these neuromodulatory systems.

Results

Synchrony of NE and 5-HT Releases during NREM Sleep.

To investigate the release pattern of NE and 5-HT in the brain during sleep and wake cycles, we injected AAV9-hSyn-GRAB_NE2m (32) in the mPFC, and AAV9-hSyn-GRAB_5-HT2h (33) in the hippocampus in wildtype mice (C57BL/6J) to express genetically encoded fluorescence sensors (Fig. 1A). The individual mice were then implanted with two fiber photometry probes above the injection sites, EEG and EMG electrodes for brain state classification. Two weeks after recovery, we conducted 2-site photometry and EEG recordings while the animals experienced natural wake/sleep cycles in a behavioral chamber. Consistent with previous studies (23, 24), we observed ISO of NE in the mPFC and 5-HT in the hippocampus during NREM sleep (Fig. 1B). Quantification of photometric signals showed highest levels of NE and 5-HT during wakefulness, significantly lower levels during NREM sleep, and lowest levels during REM sleep (Fig. 1C), recapitulating the canonical view on the dynamics of monoamines during sleep and wake cycles.

Fig. 1.

Synchronized release of NE and 5-HT during NREM sleep. (A) Left, Schematic representation of the 2-site photometry experimental design. Right, Expression of GRAB_NE (green) and GRAB_5HT (green) and fiber placement in the mPFC and hippocampus (HPF), respectively. (Scale bar, 0.5 mm.) (B) A representative example of concurrent recording of the mPFC and hippocampus in a mouse during sleep and wake cycles. From Top to Bottom: brain states, EEG power spectrogram (0 to 25 Hz), EMG trace, photometric signals in the mPFC (dark green) and HPF (light green). The white trace superimposed on the spectrogram indicates the sigma power. (C) Quantification of NE and 5-HT activity during different brain states (12 sessions from 4 mice). (D) Left, Spectral coherence between NE and 5-HT during NREM sleep. Right, Cross-correlograms between NE and 5-HT during NREM sleep (102 NREM episodes in 4 mice). The 5-HT signal was used as the reference. (E) NE (dark green) and 5-HT (light green) signals during the transition of brain states (12 sessions from 4 mice). The dashed lines at time 0 indicated the onset of wake, NREM, REM, and MA, respectively. Data are mean ± SEM. The shadows around the traces indicate the SEM.

Then, we performed correlation analysis between NE and 5-HT signals. We found that the releases of NE and 5-HT during NREM sleep are highly correlated in the infraslow frequency (~0.02 Hz) range indicated by the increased spectral coherence (Fig. 1 B and D). Cross-correlation analysis further revealed that NE and 5-HT signals are highly synchronized (lag < 1 s, Fig. 1D). Previous studies show that the ISO of NE is negatively correlated with the EEG sigma power during NREM sleep (21–23). The analysis of our data also supported this observation (SI Appendix, Fig. S1 A–C). In addition, we also confirmed that 5-HT is anticorrelated with the sigma power during NREM sleep (SI Appendix, Fig. S1 B and C). Next, to examine whether the lead-lag relationship between NE and 5-HT is specific to the brain regions, we performed 2-site photometry in different combinations. First, we switched the regions for these two neuromodulators, i.e., 5-HT in the mPFC and NE in the hippocampus. We observed synchrony, and importantly similar lag relationship between 5-HT and NE during NREM sleep (lag < 1 s, SI Appendix, Fig. S1 D–F). Second, we recorded 5-HT and NE in the hippocampus, either using two green GRAB sensors in the left and right hemispheres (SI Appendix, Fig. S2 A–C), or using green and red GRAB sensors in the same hemisphere (SI Appendix, Fig. S2 D and E). We validated the same lead-lag relationship (SI Appendix, Fig. S2). Together, these results demonstrate the synchronized release of 5-HT and NE during NREM sleep in the brain.

Finally, we examined the NE and 5-HT signals during the transitions of brain states. As expected, both NE and 5-HT signals greatly increased during sleep-to-wake transitions, and decreased during wake-to-NREM or NREM-to-REM transitions (Fig. 1E). As the ISO of NE or 5-HT is partially coupled with MA (23, 24), we also examined their activity and observed an increase in both NE and 5-HT levels when animals entered MA (Fig. 1E).

Synchronized Release of ACh and Monoamines during NREM Sleep.

Next, we examined the release relationship between ACh and monoamines (either NE or 5-HT) during sleep. To monitor ACh in the brain, we used AAV expressing GRAB_ACh3.0 (34, 35). In the first experiment, we expressed GRAB_ACh3.0 in the mPFC and GRAB_5-HT2h in the hippocampus and recorded their activity during sleep and wake cycles (Fig. 2A). As expected, we observed higher ACh levels during wakefulness and REM sleep, compared to NREM sleep (Fig. 2B). Then, our correlation analysis showed synchrony of the ISO between ACh and 5-HT during NREM sleep (Fig. 2C). Cross-correlation between two signals revealed that the ACh signal preceded the 5-HT signal by 2 to 4 s (Fig. 2C). This result was supported by the cross-correlation analysis between neuromodulator signals and EEG sigma power. As shown in SI Appendix, Fig. S3, the ACh signal preceded the sigma (~2 s), whereas the 5-HT signal succeeded the sigma (~2 s). Consistent with a previous study (31), spectral analysis between ACh and sigma power revealed a coherence peak around the ISO range (~0.02 Hz) during NREM sleep. To confirm the lead-lag relationship between ACh and 5-HT, we switched the brain regions, and recorded 5-HT in the mPFC and ACh in the hippocampus. Again, we observed that ACh signal preceded 5-HT signal by 2 to 4 s (SI Appendix, Fig. S4). Finally, we examined the levels of ACh and 5-HT during the transitions. Compared to 5-HT, ACh increased rapidly during sleep-to-wake and NREM-to-MA transitions, and declined rapidly during wake-to-NREM sleep (Fig. 2D). During NREM-to-REM transition, the ACh level increased, consistent with the canonical view on the ACh dynamics during REM sleep.

Fig. 2.

Synchronized release of ACh and 5-HT during NREM sleep. (A) A representative example of concurrent recording of ACh in the mPFC and 5-HT in the hippocampus (HPF) in a mouse during sleep and wake cycles. From Top to Bottom: brain states, EEG power spectrogram (0 to 25 Hz), EMG trace, photometric signals in the mPFC and in the hippocampus. The white trace superimposed on the spectrogram indicates the sigma power. (B) Quantification of ACh and 5-HT activity during wake (W) and NREM (N) and REM (R) sleep (16 sessions from 7 mice). (C) Left, Spectral coherence between ACh and 5-HT during NREM sleep. Right, Cross-correlograms between ACh and 5-HT signals during NREM sleep (130 NREM episodes in 7 mice). Note that ACh leads 5-HT signal by about 4 s. (D) ACh (dark green) and 5-HT (light green) signals during the transition of brain states (16 sessions in 7 mice). The dashed lines at time 0 indicated the onset of wake, NREM, REM, and MA, respectively. Data are mean ± SEM. The shadows around the traces indicate the SEM.

In the second experiment, we examined the synchrony between NE and ACh. We expressed GRAB_NE2m in the mPFC and GRAB_ACh3.0 in the hippocampus and recorded their signals during sleep and wake cycles (Fig. 3A). The ACh release in the hippocampus displayed a similar pattern to the mPFC across sleep and wake cycles (Fig. 3 A and B), in line with earlier observation (31). Similar to the dynamics between ACh and 5HT, we observed synchronized ISO between NE and ACh (Fig. 3C). Cross-correlation between NE and ACh showed that NE lagged ACh (2 to 4 s, Fig. 3C). This lead-lag relationship was confirmed when we recorded ACh and NE signals both in the hippocampus (SI Appendix, Fig. S5). During sleep-to-wake and particularly NREM-to-MA transitions, ACh rose more quickly to the peak, compared to NE (Fig. 3D). As expected, two signals diverged during NREM-to-REM transitions, with ACh increasing and NE decreasing (Fig. 3D). Together, our data revealed a synchronized release of ACh and monoamines across cortical and hippocampal regions during NREM sleep.

Fig. 3.

Synchronized release of NE and ACh during NREM sleep. (A) A representative example of concurrent recording of NE in the mPFC and ACh in the hippocampus in a mouse during sleep and wake cycles. From Top to Bottom: brain states, EEG power spectrogram (0 to 25 Hz), EMG trace, photometric signals in the mPFC and hippocampus (HPF). (B) Quantification of NE and ACh activity during wake (W) and NREM (N) and REM (R) sleep (13 sessions from 6 mice). (C) Left, Spectral coherence between NE and ACh during NREM sleep. Right, Cross-correlograms between NE and ACh during NREM sleep (102 NREM episodes in 6 mice). Note that NE lags ACh signal by 2 to 4 s. (D) NE (dark green) and ACh (light green) activity during the transition of brain states (13 sessions in 6 mice). The dashed lines at time 0 indicated the onset of wake, NREM, REM, and MA, respectively. Data are mean ± SEM. The shadows around the traces indicate the SEM.

Synchrony of Multiple Neuromodulatory Systems Promotes Arousal.

Each neuromodulatory system can promote arousal during sleep. We asked whether synchrony between neuromodulatory systems can contribute to this transition process. To address this question, we analyzed the release of neuromodulators and their correlation during the NREM-ISO cycles that led to different outcomes. As reported, the sigma power in the EEG displays ISO during NREM sleep and the sigma trough is associated with the peak of neuromodulator release (22, 26, 36). Thus, we used the sigma ISO during NREM sleep as a reference and aligned neuromodulator release to the sigma trough. There are four possible outcomes around the sigma trough from an NREM-ISO cycle: NREM (without MA), MA, wake, and REM sleep (Fig. 4 A and B). Among them, ~60% is associated with NREM sleep (i.e., NREM-to-NREM), 25 to 30% with MA (NREM-to-MA, SI Appendix, Fig. S6). To examine the activity of neuromodulatory systems associated with these four different outcomes, we separated ISO cycles and aligned all neuromodulator signals during NREM sleep to the sigma trough. We found that the amplitude of NE, 5-HT, and ACh releases during NREM-ISO cycles increases with the increased level of arousal from NREM, to MA, and to full wake (Fig. 4 C–E and SI Appendix, Fig. S6). The NE activity associated with different arousal levels is consistent with a previous study in the mPFC (23). As expected, the releases of monoamines and ACh differed during the transition to REM sleep: NE and 5-HT decreased while ACh increased around the sigma trough. This decrease of NE during the ISO cycles is consistent with a previous study showing that decreased LC-NE activity during NREM sleep allows REM entries (19).

Fig. 4.

Synchronized release of multiple neuromodulators promotes arousal. (A) A representative example showing synchrony among EEG sigma power, NE, and 5-HT during NREM sleep and MAs. Gray and pink arrows indicate the peaks and troughs of the sigma power, respectively. The sigma power was smoothed across 10 s to capture the ISO. (B) The sigma power in the ISO cycles associated with different outcomes of brain states from NREM sleep (12 sessions from 4 mice). Time 0 indicates the trough of sigma power. (C) Schematic of experimental design of photometry recordings in the brain. (D and E) Left, the activity of neuromodulator release in the cortex (D) and hippocampus (E) associated with different outcomes of brain states from NREM sleep. Right, quantitation of Z-scored activity during different outcomes. The amplitude of activity from the time of sigma peak (T1) to the time of sigma trough (T0) in each trial was quantified. The shadows around the traces indicate the SEM. P < 0.001 (ANOVA) for all groups. (F) Left, Quantification of Pearson correlation coefficient between two neuromodulators associated with different outcomes of brain states from NREM sleep. Right, Quantification of covariance between two neuromodulators. The activity in the ISO cycles from the time of sigma peak (T1) to the time of sigma trough (T0) in each trial was used for correlation analysis. ***P < 0.001 unpaired t test between NREM and MA (N = 10 sessions in 4 mice for NE-5HT, N = 16 sessions in 7 mice for ACh-5HT, N = 13 sessions in 6 mice for NE-ACh). Data are mean ± SEM.

Next, we performed the pair-wise correlation analysis between two neuromodulators to examine their synchrony prior to the brain state transition (i.e., from a prior sigma peak to a sigma trough, Fig. 4A). Strikingly, we found that both Pearson correlation coefficient and covariance between neuromodulator signals increased significantly in the ISO cycles that led to the increased level of arousal from NREM to MA/wake: Lower when animals remained in NREM sleep, higher when animals entered MA or wake (Fig. 4F). Furthermore, linear regression analysis confirmed that the synchrony, especially covariance, can predict the outcome of brain states from NREM sleep (SI Appendix, Tables S1–S3). Together, our data suggest that increased amplitude and increased synchrony of multiple neuromodulatory systems during sleep promote arousal.

ISO Requires Coordination of Neuromodulatory Systems.

Given the cross-region synchrony among NE, 5-TH, and ACh during NREM sleep, a key question is what coordinates this synchrony. A more specific question is whether the ISO of one neuromodulator is regulated by another neuromodulator. To address this question, we used a pharmacological approach to inhibit one neuromodulatory system and then examine its effect on the ISO of other neuromodulatory systems during NREM sleep. Since NE, 5-TH, and ACh are anticorrelated with sigma power during NREM sleep (23, 31, 36) (SI Appendix, Figs. S1 and S3), we performed cross-correlation and spectral coherence analysis between each neuromodulator and the sigma power to examine the effect of pharmacological inhibition of the ISO. Activation of inhibitory autoreceptors in neuromodulatory systems offers a powerful way to inhibit neuromodulator release. First, we administered 8-hydroxy-DPAT (8-OH-DPAT), a selective 5-HT1a agonist (i.p. 1 mg/kg). Since Htr1a receptors act as autoreceptors on raphe 5-HT neurons (37), this treatment effectively silences the 5-HT system, thereby “removing” 5-HT signaling from the brain. Indeed, this pharmacological manipulation abolished the ISO of 5-HT during NREM sleep (Fig. 5 A–C). Early pharmacological studies showed that systemic administration of 8-OH-DPAT increases wakefulness and suppresses both NREM and REM sleep (38, 39). We confirmed this effect during the 2-h period immediately after 8-OH-DPAT injection (SI Appendix, Fig. S7 A–D). We then examined the dynamics of NE and ACh in the brain. Strikingly, we found that inhibition of the 5-HT system also abolished the ISO of NE and ACh during NREM sleep indicated by the absence of spectral peaks around 0.02 Hz (Fig. 5 A–C). The treatment also abolished the anticorrelation between neuromodulators and sigma power during NREM sleep. Next, we administered dexmedetomidine (40, 41), a selective adrenergic α2 agonist (i.p. 50 μg/kg) to silence the NE system through the inhibitory autoreceptors. As reported in both animals and humans, dexmedetomidine induces a sedative response and strongly promotes sleep (42–44). Indeed, we observed that systemic administration of dexmedetomidine significantly suppressed wakefulness and increased NREM sleep (SI Appendix, Fig. S7 E–H). Interestingly, REM sleep was strongly suppressed during the 2-h period after dexmedetomidine injection (SI Appendix, Fig. S7F). Next, we examined the effect of dexmedetomidine on the releases of neuromodulators. Similarly, we found that inhibition of the NE system abolished the ISO of NE, 5HT, and ACh during NREM sleep (Fig. 5 D–F). Finally, we attempted to use the same agonist strategy to silence the ACh system. Due to lack of selective agonists for inhibitory muscarinic autoreceptors (e.g., M2R and M4R), we tried pilocarpine, a nonselective muscarinic agonist (45). We found that pilocarpine (10 mg/Kg, i.p.) largely abolished both NREM and REM sleep within the 2-h window (SI Appendix, Fig. S7 I–L). Due to lack of sleep, we could not examine its effect on ISO during NREM sleep. Together, our pharmacological data indicated that the ISO of one neuromodulator requires coordinated action of other neuromodulatory systems.

Fig. 5.

ISO during NREM sleep requires coordination among neuromodulatory systems. (A) Representative examples showing EEG sigma power and neuromodulator signals during NREM sleep before (pre), during, and after (post) pharmacological treatments of Htr1a agonist, 8-OH-DPAT (1 mg/Kg, i.p.). (B) Cross-correlograms between neuromodulators and sigma power before, during, and after Htr1a agonist treatments during NREM sleep. The sigma power was used as the reference. (C) Spectral coherence between neuromodulators and sigma power before, during, and after Htr1a agonist treatments (Pre, During, After: N = 32, 23, 58 NREM episodes from 5 mice for 5-HT recorded in the hippocampus; N = 26, 30, 31 NREM episodes from 4 mice for NE recorded in the mPFC; N = 35, 33, 24 NREM episodes from 4 mice for ACh recorded in the hippocampus). (D) Representative examples showing sigma power and neuromodulator signals during NREM sleep before (pre), during, and after (post) pharmacological treatments of Adra2a agonist, Dexmedetomidine (50 μg/Kg, i.p.). (E) Cross-correlograms between neuromodulators and sigma power before, during, and after Adra2a agonist treatments during NREM sleep. The sigma power was used as the reference. (F) Spectral coherence between neuromodulators and sigma power before, during, and after Adra2a agonist treatments (Pre, During, After: N = 32, 17, 32 NREM episodes from 4 mice for NE recorded in the mPFC; N = 36, 12, 46 NREM episodes from 5 mice for 5-HT recorded in the hippocampus; N = 43, 22, 37 NREM episodes from 5 mice for ACh recorded in the mPFC). Data are mean ± SEM. The shadows around the traces indicate the SEM.

An intriguing question is whether activation of one neuromodulatory system is capable to induce the release of other neuromodulators. To test this, we performed optogenetic activation of LC-NE and raphe 5-HT neurons while recording different neuromodulators. First, we injected AAV1-DIO-ChR2 in the LC, and either GRAB_5-HT2h or GRAB_ACh3.0 in the hippocampus of Dbh-Cre mice (Fig. 6A). Previous studies report that phasic optogenetic stimulation causes immediate sleep-to-wake transitions, whereas long-term or repeated stimulation subsequently induces rebound NREM sleep (14, 15). The change of brain states likely influences the release of neuromodulators. To avoid this compound effect, we used low frequency optogenetic stimulation (1 Hz) for 10 s. This stimulation protocol did not cause immediate sleep-to-wake transitions (Fig. 6 B and C and SI Appendix, Fig. S8 A–D). Strikingly, optogenetic activation of LC-NE neurons during NREM sleep induces the release of 5-HT in the brain (Fig. 6 B–D). In contrast, we observed very mild increase of 5-HT level following optogenetic stimulation during wakefulness (Fig. 6D). Importantly, the stimulation during NREM sleep did not induce MAs, as indicated by the unchanged EMG amplitude (Fig. 6E). We also examined the effect on sigma power during sleep. We found that optogenetic activation significantly reduced sigma power during NREM sleep (SI Appendix, Fig. S8E), similar to the sigma decrease in spontaneous ISO cycles. Then, we repeated the optogenetic experiment to examine ACh signals in a different cohort of mice. Similarly, we demonstrated that optogenetic activation of LC-NE neurons during NREM sleep is sufficient to induce the release of ACh while not inducing sleep-to-wake transitions (Fig. 6 F–J and SI Appendix, Fig. S8 F–I). EMG amplitude was not affected by stimulation during both wakefulness and NREM sleep (Fig. 6J), whereas sigma power was decreased during NREM sleep (SI Appendix, Fig. S8J).

Fig. 6.

Optogenetic activation of neuromodulatory systems. (A) Top, schematic of experimental design in Dbh-Cre mice. Bottom, Expression of GRAB_5HT (green) in the hippocampus (HPF) and ChR2-EYFP (green) in the LC. (Scale bar, 0.5 mm.) (B) A representative example showing the effect of optogenetic activation (1 Hz, 10 s) of LC-NE neurons on 5-HT release during sleep and wake cycles. From Top to Bottom, brain states, EEG spectrogram (0 to 25 Hz), EMG, and photometry. The light blue lines indicate laser stimulation. Two trials (one during NREM sleep, the other during wake) were enlarged below. (C) Left, Probability of wake (black), NREM (orange), and REM sleep (purple) before, during, after laser stimulation (blue lines) in all trials (255 trials from 4 Dbh-Cre mice). Right, Z-scored 5-HT activity in the hippocampus before, during, and after optogenetic activation in NREM trials and wake trials (N = 4 Dbh-Cre mice). T0 indicates the laser onset. (D) Quantification of 5-HT activity before (pre) and during (ChR2) stimulation in NREM and wake trials. (E) Quantification of EMG amplitude before and during stimulation in NREM and wake trials. (F) Schematic of experimental design. (G) representative examples showing the effect of optogenetic activation (1 Hz, 10 s) of LC-NE neurons on ACh release during NREM sleep and wake. (H) Left, Probability of brain states before, during, after laser stimulation (120 trials from 3 Dbh-Cre mice). Right, Z-scored ACh activity in the hippocampus before, during, and after optogenetic activation in NREM trials and wake trials (N = 3 Dbh-Cre mice). (I) Quantification of ACh activity in NREM and wake trials. (J) Quantification of EMG amplitude in NREM and wake trials. (K) Top, schematic of experimental design in Slc6a4-Cre mice. Bottom, Expression of GRAB_NE (green) in the mPFC and ChR2-EYFP (green) in the DRN. (Scale bar, 0.5 mm.) (L) A representative example showing the effect of optogenetic activation (1 Hz, 10 s) of DRN 5-HT neurons on NE release during sleep and wake cycles. (M) Left, Probability of wake (black), NREM (orange), and REM sleep (purple) before, during, after laser stimulation (blue lines) in all trials (300 trials from 3 Slc6a4-Cre mice). Right, Z-scored NE activity in the mPFC before, during, and after optogenetic activation in NREM trials and wake trials (N = 3 Slc6a4-Cre mice). T0 indicates the laser onset. (N) Quantification of NE activity in NREM and wake trials. (O) Quantification of EMG amplitude in NREM and wake trials. (P) Schematic of experimental design. (Q) representative examples showing the effect of optogenetic activation (1 Hz, 10 s) of 5-HT neurons on ACh release. (R) Left, Probability of brain states before, during, after laser stimulation (187 trials from 3 Slc6a4-Cre mice). Right, Z-scored ACh activity in the hippocampus in NREM trials and wake trials (N = 3 Slc6a4-Cre mice). (S) Quantification of ACh activity in NREM and wake trials. (T) Quantification of EMG amplitude in NREM and wake trials. *P < 0.05, ***P < 0.001, n.s., not significant, paired t test.

Next, we used the same strategy to optogenetically activate 5-HT neurons by expressing ChR2 in the raphe nuclei of Slc6a4-Cre (also named SERT-Cre) mice while recording NE or ACh in the mPFC. Again, the low frequency stimulation (1 Hz) of 5-HT neurons did not induce sleep-to-wake transitions (Fig. 6 M and R and SI Appendix, Fig. S9). Importantly, optogenetic stimulation of 5-HT neurons during NREM sleep induced the release of NE or ACh in the brain (Fig. 6 L–N and Q–S). In contrast, stimulation during wakefulness had no significant effect on NE and ACh level (Fig. 6 M, N, R, and S). Moreover, stimulation during NREM sleep had no significant effect on EMG amplitude (Fig. 6 O and T), but decreased sigma power (SI Appendix, Fig. S9 E and J). Together, these results indicated that activation of one neuromodulatory system is sufficient to coordinate the release of other neuromodulators, further demonstrating the interdependence among neuromodulatory systems.

Discussion

In this study, we characterized the release patterns of NE, 5-TH, and ACh in the brain across sleep and wake cycles. By using concurrent recordings of two neuromodulators in two brain regions, we revealed a cross-region synchrony of ISO during NREM sleep. The synchrony between neuromodulatory systems further increases in the NREM-ISO cycles leading to arousal. Moreover, our pharmacological and optogenetic experiments confirmed the coordinated and interdependent releases of multiple neuromodulators in the brain. These findings revealed a previously unrecognized pattern of neuromodulator releases during sleep states. This synchronized ISO of multiple neuromodulatory systems enables robust regulation of sleep and arousal; and might further contribute to sleep-related functions, such as memory consolidation (23, 24, 46).

While fiber photometry has limited temporal resolution, our cross-correlation analysis between neuromodulators in different configurations did suggest a sequential order of releases during NREM sleep: ACh>NE>5-TH. In particular, while the time-lag between NE and 5-HT is short (<1 s), ACh release preceded both NE and 5-HT (2 to 4 s). The kinetics of GRAB_NE and GRAB_5HT sensors are highly similar (32, 33), whereas the GRAB_ACh sensor does display slight faster kinetics, particularly off-kinetics, compared to the GRAB_NE and GRAB_5HT sensors (34). However, the time-lag (2 to 4 s) between ACh and NE/5HT is much larger than the kinetics difference of the sensors (Δτ_on < 0.1 s, Δτ_off < 1 s). Importantly, this lead-lag relationship between ACh and monoamines is independent on brain regions, which is an important observation. Thus, this action sequence might carry some biological meaning in regulating brain states, particularly during MAs and wakening. On the other hand, our pharmacological experiments indicated that neuromodulatory systems are interdependent, which allows coordinated regulation of sleep and arousal. An interesting question is how one neuromodulatory system regulates another. A key point to distinguish the direct vs. indirect effect is whether ACh and monoaminergic receptors are expressed in neurons that release neuromodulators. A recent study performed single-nucleus RNA-sequencing and spatially resolved transcriptomics in human LC, and their data showed that no 5-HT inhibitory autoreceptors (e.g., Htr1a and Htr1b) are expressed in LC-NE neurons (47). While the expression needs validation in rodents, this result suggests that the effect of our Htr1a pharmacological inhibition on NE release might be indirect. A similar indirect mechanism was reported that ACh modulates another neuromodulator oxytocin through a long-loop hippocampal-lateral septum path (31). On the other hand, Ren et al. reported that ACh and monoamine receptors including Adra2a are expressed in raphe 5-HT neurons (48), suggesting that our Adra2a pharmacological manipulation could directly inhibit the 5-HT system. Anatomical studies showed that the BF receives direct serotonergic innervation from the dorsal raphe nucleus (DRN) (49, 50). Electrophysiological studies demonstrated that Htr1a and Htr1b receptors mediate both presynaptic and postsynaptic inhibition on BF cholinergic neurons (51). These findings along with our data further support a direct interaction between 5-HT and ACh systems. Between NE and ACh, an electron microscopy study confirmed the presence of Adra2a receptors in BF cholinergic neurons (52). Furthermore, a c-Fos study showed that sleep-active GABAergic neurons in the BF are immunostained for Adra2a receptors (53). These data also support the direct modulation of NE on the ACh system.

An interesting observation in our pharmacological experiments is that while inhibitions of 5-HT and NE systems have different effect on overall sleep behaviors (i.e., Htr1a and Adra2a agonists suppressed and increased NREM sleep, respectively), both manipulations effectively abolished the ISO of neuromodulators during NREM sleep. The different effect on sleep might be related to the unspecific manipulations, dose effects, and complicated expression patterns of these neuromodulatory receptors in sleep and/or arousal-related brain regions (3, 5, 38, 54). In particular, our data demonstrated that dexmedetomidine (10 mg/Kg, a rather lower concentration) induced different spectral profiles during NREM sleep and in some cases prior to NREM sleep (resembling a cataplexy-like state). Consistently, previous studies show that decreased excitatory noradrenergic drive is implicated in cataplexy (55, 56). Moreover, the role of the 5-HT system in promoting arousal and sleep is still controversial (16, 39). Nevertheless, inhibition of either NE or 5-HT system was sufficient to impact the ISO of other neuromodulators during sleep. A related question is what circuits are involved in this process. Targeted inhibitions with optogenetic tools in LC-NE or Raphe-5HT neurons might provide more insights into the circuit mechanisms.

We demonstrated that optogenetic activation of LC-NE neurons or raphe 5-HT neurons, in the absence of sleep-to-wake transitions, is sufficient to induce the release of other neuromodulators during NREM sleep (e.g., 5-HT and ACh for NE activation, and NE and ACh for 5-HT activation). A question is whether this action is via direct projection or indirect pathways. Future ex vivo experiments combining optogenetics and GRAB sensors in brain slices might provide evidence to address this question. Notably, optogenetic activation during wakefulness largely did not induce detectable increase of neuromodulator levels. One possibility is that the high levels of neuromodulators during wake periods might mask the small increase evoked by the weak stimulation. Due to relatively less time of REM sleep, we obtained much less REM trials for optogenetic stimulation. Our preliminary data suggested that optogenetic activation of either LC-NE or raphe 5-HT neurons likely induced awakening in some trials. However, given the short duration of REM epoch, some could be spontaneous REM-to-wake transitions. Together with pharmacological inhibition, these optogenetic results highlight the coordination among neuromodulatory systems.

One limitation of this study is lack of manipulation of the ACh system. We tried pilocarpine, a muscarinic agonist. We found that pilocarpine strongly induces prolonged wakefulness. This is likely due to nonselective activation of muscarinic receptors. While pilocarpine activates inhibitory muscarinic autoreceptors M2 and M4, it also binds to Gq-coupled M1 and M3 receptors (57, 58). Due to lack of sleep following pilocarpine treatments, we could not examine its effect on NREM-related ISO. A related experiment is optogenetic activation of cholinergic neurons. Unlike NE or 5-HT neurons, cholinergic neurons are more broadly distributed in the brains including the BF (e.g., medial septum and nucleus basalis), the brainstem (e.g., pedunculopontine nucleus and lateral dorsal tegmental nucleus), and the striatum (2, 4, 59). Considering this complexity and the scope of this study, we did not perform optogenetic activation of cholinergic neurons. Future studies in this direction are needed to examine the role of the ACh system in regulating the release of NE and 5-HT during sleep.

In this study, we recorded neuromodulators in the mPFC and hippocampus, two key brain regions involved in memory (60), a process heavily affected by sleep (61). Given the global projections of neuromodulatory systems, we speculate that this synchrony of neuromodulator releases can be generalized to other brain regions, such as thalamus and hypothalamus. Indeed, recent studies have reported NE-related ISO in the thalamus (22), and in the preoptic area of the hypothalamus (62). Simultaneous recordings of these brain regions are needed to validate the speculation. Moreover, the findings of cross-region synchrony among neuromodulatory systems raise many questions. For instance, what signals initiate the ISO? What brain functions require synchronized release of multiple neuromodulators during sleep? Further studies at the molecular, circuit, and behavioral levels are needed to address these questions.

Materials and Methods

Animals.

All procedures were carried out in accordance with the US NIH guidelines for the care and use of laboratory animals and approved by the Animal Care and Use Committees of Columbia University. Male and female Dbh-Cre (JAX #033951, 12 to 20 wk old), Slc6a4-Cre (also named SERT-Cre, JAX #014554, 12 to 24 wk old), male C57BL/6J mice (JAX #000664, 12 to 16 wk old) were used for all experiments. Mice were housed in 12-h light–dark cycles (lights on at 07:00 am and off at 07:00 pm, temperatures of 65 to 75°F with 40 to 60% humidity) with free access to food and water.

Viral Constructs.

AAV9-hSyn-GRAB_NE2m was obtained from Biohippo Inc (Cat# BHV12400445). AAV9-hSyn-GRAB_5-HT2h was obtained from Vigene Biosciences (Cat# YL10097-AV9). AAV9-hSyn-GRAB_ACh3.0 and AAV9-hSyn-GRAB_r5-HT1.0 were obtained from WZ Biosciences (Cat# YL h-A06). AAV1-EF1α-double-floxed-hChR2(H134R)-EYFP-WPRE-HGHpA (AAV1-DIO-ChR2) was obtained from Addgene (Cat# 20298).

Surgical Procedures.

EEG and fiber implants.

Mice were anesthetized with a mixture of ketamine and Xylazine (100 mg/kg and 10 mg/kg, intraperitoneally), then placed on a stereotaxic frame with a closed-loop heating system to maintain body temperature. After asepsis, the skin was incised to expose the skull and a small craniotomy (~0.5 mm in diameter) was made on the skull above the regions of interest. For EEG and EMG recordings, a reference screw was inserted into the skull on top of the cerebellum. EEG recordings were made from two screws on top of the cortex 1 mm from midline, 1.5 mm anterior to the bregma and 1.5 mm posterior to the bregma, respectively. Two EMG electrodes were bilaterally inserted into the neck musculature. EEG screws and EMG electrodes were connected to a PCB board which was soldered with a 5-position pin connector. For optogenetic stimulation or fiber photometry recording, an optical fiber (0.2 mm diameter, 0.39 NA, Thorlabs) was implanted into the target region with the tip 0.1 mm above the virus injection site. All the implants were secured onto the skull with dental cement (Lang Dental Manufacturing). After surgery, the animals were returned to their home cages for recovery for at least 2 wk before any experiment.

Virus injection.

A solution containing 150 to 200 nL viral construct was loaded into a pulled glass capillary and injected into the target region using a Nanoinjector (WPI). For 2-site fiber photometry, two AAV expressing different GRAB sensors were unilaterally injected in the hippocampus (AP −1.9 mm, ML 1.5 mm, DV 1.7 mm) and the mPFC (AP +1.8 mm, ML 0.3 mm, DV 2.4 mm), respectively. For optogenetic experiments, 200 nL AAV1-EF1α-double-floxed-hChR2(H134R)-EYFP-WPRE-HGHpA was unilaterally injected into the LC in Dbh-Cre mice (AP −5.5 mm, ML −0.9 mm, DV 3.4 mm) or DRN in Slc6a4-Cre mice (AP −4.5 mm, ML 0 mm, DV 3.2 mm), The DV Coordinates listed above are relative to the pial surface.

Sleep Recording.

Mouse sleep behavior was monitored using EEG and EMG recording along with an infrared video camera at 30 frames per second. Recordings were performed across 24 h (light on at 7:00 am and off at 7:00 pm) in a behavioral chamber inside a sound attenuating cubicle (Med Associated Inc.). Animals were habituated in the chamber for at least 4 h before recording. EEG and EMG signals were recorded, bandpass filtered at 0.5 to 500 Hz, and digitized at 1,017 Hz with 32-channel amplifiers (TDT, PZ5, and RZ5D). Spectral analysis was carried out using fast Fourier transform over a 5 s sliding window, sequentially shifted by 2 s increments (bins). Brain states were semi-automatically classified into wake, NREM sleep, and REM sleep states using a custom-written MATLAB (MathWorks) program (24) (wake: desynchronized EEG and high EMG activity; NREM: synchronized EEG with high-amplitude, delta frequency (0.5 to 4 Hz) activity and low EMG activity; REM: high power at theta frequencies (6 to 9 Hz) and low EMG activity). Semi-auto classification was validated manually by trained experimenters. As previously described (24), we categorized wake bouts of <15 s as MAs in the following way: the baseline of the EMG was calculated by averaging the signal, then a threshold of 0.5 SD above the baseline was used to detect the MA during NREM sleep. Thus, the presence of EMG events was needed for MA. Finally, each MA event was further validated manually by a trained experimenter. Relative sigma power (9 to 15 Hz) was calculated by dividing the sigma power in the 2-s bins by the total EEG power averaged across the recording session. To detect the sigma trough in Fig. 4, we first calculated a moving baseline by smoothing the sigma signals over 60 s, then set a threshold (0.2 SD from the moving baseline) for events of sigma decrease, and finally detected the minimum point in each event as the sigma trough. A maximum point between two troughs was considered the sigma peak.

Fiber Photometry.

Fiber photometry recordings were performed as previously described (24, 63). In brief, the GRAB sensor fluorescence was excited by sinusoidal modulated LED light (for green sensor: 465 nm 220 Hz, 405 nm 350 Hz; for green and red sensors: 465 nm 350 Hz, 560 nm 230 Hz, 405 nm 430 Hz) and detected by femtowatt silicon photoreceivers (NewPort #2151 and TDT RZ10x). Photometric signals and EEG/EMG signals were simultaneously acquired by a real-time processor (RZ5D, TDT) and synchronized with behavioral video recording. A motorized commutator (ACO32, TDT) was used to route electric wires and optical fiber. The collected data were analyzed by custom MATLAB scripts. They were first extracted and subject to a low-pass filter at 2 Hz. A least-squares linear fit was then applied to produce a fitted 405 nm signal. The DF/F was calculated as (F-F0)/F0, where F0 was the fitted 405 nm signal. To compare activity across animals, photometric data were further normalized using Z-score calculation in each mouse. To analyze the ISO of photometric signals and their interaction, data were first downsampled to 1 Hz, we performed spectral coherence analysis at the infraslow frequency range (0 to 0.5 Hz) and cross-correlation analysis with a 60-s time-lag. Only wake, NREM and REM episodes longer than 60 s were included for spectral coherence and cross-correlation analysis due to the time-lag. In Fig. 4, the Pearson correlation coefficient and covariance were calculated between two signals at a time window from a prior sigma peak to the sigma trough during NREM sleep and then grouped by the different brain state outcome associated with a sigma trough. To examine the activity of neuromodulators during the brain state transitions, the photometric signals were aligned to the onset of a brain state (wake, NREM, REM, and MA) to generate the per-stimulus time histogram. The Z-scored signals in each transition were further calibrated by subtracting the baseline, defined as the average of photometric signals in the first 5-s window (from −15 s to −10 s).

Optogenetic Manipulation.

All optogenetic stimulation were conducted unilaterally. Mice were habituated in the behavioral chamber for at least 4 h before the experiment. Light pulses (1 Hz, 10 ms) with a duration of 10 s from a 473 nm laser diode (Shanghai Laser & Optics Century Co., Ltd.) were controlled by a microcontroller board (Arduino Mega 2560, Arduino). Interstimulation interval for optogenetic stimulation is 10- or 5-min. Laser power is set to 6 to 8 mW. Optogenetic stimulation was evenly applied at both light and dark cycles while conducting fiber photometry recordings (1-h session with an interval of 1 to 3 h).

Pharmacology.

Adrenergic α2 agonist Dexmedetomidine hydrochloride (Cat#2749), Htr1a agonist 8-Hydroxy-DPAT hydrobromide (Cat#0529), and muscarinic agonist Pilocarpine hydrochloride (Cat#0694) were obtained from Tocris Bioscience. On the day of the experiment, mice were habituated and recorded in a behavioral chamber for 1 to 2 sessions (prerecording). Following the prerecording, 8-OH-DPAT (1 mg/Kg in 0.9% NaCl, 0.2 µL) or Dexmedetomidine (50 μg/Kg in 0.9% NaCl, 0.2 µL) or Pilocarpine (10 mg/Kg in 0.9% NaCl, 0.2 µL) was administered intraperitoneally. Immediately from injection, mice were recorded for another 2-h session. At least 6 h after injection, mice were recorded again for 1 to 2 sessions (postrecording).

Histology.

Viral expression and placement of optical implants were verified at the termination of the experiments using DAPI counterstaining of 100 μm coronal sections (Prolong Gold Antifade Mountant with DAPI, Invitrogen). Images were acquired using a Zeiss 810 confocal microscope.

Statistics.

No statistical methods were used to predetermine sample size, and investigators were not blinded to group allocation. No method of randomization was used to determine how animals were allocated to experimental groups. Mice in which post hoc histological examination showed viral targeting or fiber implantation was in the wrong location were excluded from analysis. Two-sided paired t test, unpaired t test, and one-way ANOVA were used and indicated in the respective Figure legends. All analyses were performed in MATLAB. Data are presented as mean ± SEM.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

MATLAB scripts for fiber photometry and EEG/EMG analysis are available at GitHub (https://github.com/thepenglab/TDTEEG) (64). All other data are included in the manuscript and/or supporting information.