Research Articles, Systems/Circuits

Oscillatory Population-Level Activity of Dorsal Raphe Serotonergic Neurons Is Inscribed in Sleep Structure

Tomonobu Kato, Yasue Mitsukura, Keitaro Yoshida, Masaru Mimura, Norio Takata and Kenji F. Tanaka
Journal of Neuroscience 21 September 2022, 42 (38) 7244-7255; https://doi.org/10.1523/JNEUROSCI.2288-21.2022

Abstract

Dorsal raphe (DR) 5-HT neurons regulate sleep–wake transitions. Previous studies demonstrated that single-unit activity of DR 5-HT neurons is high during wakefulness, decreases during non-rapid eye movement (NREM) sleep, and ceases during rapid eye movement (REM) sleep. However, characteristics of the population-level activity of DR 5-HT neurons, which influence the entire brain, are largely unknown. Here, we measured population activities of 5-HT neurons in the male and female mouse DR across the sleep–wake cycle by ratiometric fiber photometry. We found a slow oscillatory activity of compound intracellular Ca2+ signals during NREM sleep. The trough of the concave 5-HT activity increased across sleep progression, but 5-HT activity always returned to that seen during the wake period. When the trough reached a minimum and remained there, REM sleep was initiated. We also found a unique coupling of the oscillatory 5-HT activity and wideband EEG power fluctuation. Furthermore, optogenetic activation of 5-HT neurons during NREM sleep triggered a high EMG power and induced wakefulness, demonstrating a causal role of 5-HT neuron activation. Optogenetic inhibition induced REM sleep or sustained NREM, with an EEG power increase and EEG fluctuation, and pharmacological silencing of 5-HT activity using a selective serotonin reuptake inhibitor led to sustained NREM, with an EEG power decrease and EEG fluctuation. These inhibitory manipulations supported the association between oscillatory 5-HT activity and EEG fluctuation. We propose that NREM sleep is not a monotonous state, but rather it contains dynamic changes that coincide with the oscillatory population-level activity of DR 5-HT neurons.

SIGNIFICANCE STATEMENT Previous studies have demonstrated single-cell 5-HT neuronal activity across sleep–wake conditions. However, population-level activities of these neurons are not well understood. We monitored DR 5-HT population activity using a fiber photometry system in mice and found that activity was highest during wakefulness and lowest during REM sleep. Surprisingly, during non-REM sleep, the 5-HT population activity decreased with an oscillatory pattern, coinciding with EEG fluctuations. EEG fluctuations persisted when DR 5-HT neuron activity was silenced by either optogenetic or pharmacological interventions during non-REM sleep, suggesting an association between the two. Although oscillatory DR 5-HT neuron activity did not generate EEG fluctuations, it provides evidence that non-REM sleep exhibits at least binary states.

Introduction

The dorsal raphe (DR) nucleus in the hindbrain contains about one third of the 5-HT neurons of the brain (Müller and Cunningham, 2009). DR 5-HT neurons innervate mainly the forebrain, including the cortex and striatum (Gaspar and Lillesaar, 2012). The roles of DR 5-HT neurons vary from regulating physical activities to regulating emotional states (Müller and Cunningham, 2009).

In regard to sleep–wake regulation by DR 5-HT neurons, studies using DR 5-HT neuron ablation or other methods that induce a loss of function of these neurons demonstrated disturbances in sleep–wake structure. DR ablation reduced sleep in cats (Jouvet, 1968) and fish (Oikonomou et al., 2019). Administration of the irreversible inhibitor of tryptophan hydroxylase (Tph; the rate-limiting enzyme in 5-HT synthesis) reduced sleep in monkeys (Weitzman et al., 1968), cats (Koella et al., 1968), rats (Torda, 1967; Mouret et al.,1968), and fish (Oikonomou et al., 2019). Knockout of Tph2 (the gene encoding the CNS isoform of Tph) reduced sleep (siesta) in mice (Whitney et al., 2016) and fish (Oikonomou et al., 2019). All these long-term loss-of-function manipulations of the DR 5-HT neurons resulted in decreased sleep. In contrast, temporary cooling of the DR induced sleep in cats (Cespuglio et al., 1976), indicating an opposing outcome after acute loss of function of DR 5-HT neurons. Although the outcome of loss-of-function studies targeting DR 5-HT neurons is controversial, it is widely accepted that DR 5-HT neurons are causally involved in the sleep–wake structure.

Unlike the interventional studies modulating DR 5-HT neuronal activity, observational electrophysiological studies monitoring DR neuronal activities have reported consistent results (Mcginty and Harper, 1976; Trulson and Jacobs, 1979; Urbain et al., 2006; Sakai, 2011). Single-unit recording from the DR across the sleep–wake cycle revealed two major types of neurons. One type—the 5-HT neuron cell type—tonically fires during wakefulness, is less active during non-rapid eye movement (NREM) sleep, and is mostly silent during rapid eye movement (REM) sleep. The other cell type, non-5-HT neurons such as dopaminergic or GABAergic neurons, does not modulate its firing rate across the sleep–wake cycle (Sakai, 2011). These pioneering studies encouraged us to monitor the population-level activity of DR 5-HT neurons across the sleep–wake cycle because the population 5-HT neuron activity, rather than individual neuron activity, could mediate a wide range of cortical activity patterns.

In this study, we sought to examine the dynamics of the population-level activity of DR 5-HT neurons during sleep in mice and to address how such dynamics were correlated to EEG and EMG changes. We addressed the causal relationship between DR 5-HT neuron activity and EEG/EMG changes across sleep stages using optogenetics and pharmacological experiments.

Materials and Methods

Ethics statement

All animal procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Keio University Animal Experiment Committee in compliance with the Keio University Institutional Animal Care and Use Committee (approval numbers 12035 and 14027). This study was conducted following the principles of the Declaration of Helsinki. Informed consent was obtained from all participants.

Animals

Experiments were conducted with 8- to 14-month-old male and female mice. All mice were maintained on a 12 h light/dark cycle (lights on at 8:00 A.M.), and polysomnographic recordings were performed during the light phase [start time at 9:00–11:00 A.M. (Zeitgeber time 1–3)]. Tph2 Yellow Cameleon (YC) mice (Tph2-tTA::tetO-YC-nano50 double transgenic mice) were obtained by crossing tetO-YC-nano50 mice, Tph2-tTA mice (Miyazaki et al., 2014), and tetO-YC-nano50 mice (Kanemaru et al., 2014). Tph2-ChR2 mice (Tph2-tTA::tetO-ChR2(C128S)-EYFP double transgenic mice) were obtained by crossing Tph2-tTA mice and tetO-ChR2 mice (Tanaka et al., 2012). Tph2-Archaerhodopsin T (ArchT) mice (Tph2-tTA::tetO-ArchT-EGFP double transgenic mice) were obtained by crossing Tph2-tTA mice and tetO-ArchT-EGFP mice (Tsunematsu et al., 2013). All mouse lines were sourced from the RIKEN Center for Biosystems Dynamics Research. The genetic background of all transgenic mice was mixed with C57BL6 and 129 SvEvTac. Genotyping for Tph2-tTA, tetO-YC-nano50, tetO-ChR2(C128S), and tetO-ArchT has been described previously (Tanaka et al., 2012; Tsunematsu et al., 2013; Kanemaru et al., 2014; Miyazaki et al., 2014).

Surgical procedure

Surgeries were performed using a stereotaxic apparatus (SM-6M-HT, Narishige). Mice were anesthetized with a mixture of ketamine and xylazine (100 mg/kg and 10 mg/kg, respectively). Body temperature during surgery was maintained at 37 ± 0.5°C using a heating pad (FHC-MO, Muromachi Kikai). An optic fiber for optogenetics or photometry was inserted into the DR at the following coordinates relative to bregma: anteroposterior (AP), −4.3 mm; ML, 0.0 mm; DV, 3.0 mm, all while tilted at 10° relative to the vertical axis (SM-15R, Narishige). The mice received permanent EEG and EMG electrode implants for polysomnography. Using a carbide cutter (drill size diameter 0.8 mm), three pits were drilled into the skull, while avoiding penetration of the skull to prevent brain damage. Each implant had a 1.0 mm diameter stainless steel screw that served as an EEG electrode; one implant was placed over the right frontal cortical area (AP, +1.0 mm; ML, +1.5 mm) as a reference electrode and the other over the right parietal area (AP, +1 mm anterior to lambda; ML, +1.5 mm) as a signal electrode. Another electrode was placed over the right cerebellar cortex (AP, −1.0 mm posterior to lambda; ML, +1.5 mm) as a ground electrode. Two silver wires (catalog #AS633, Cooner Wire) were placed bilaterally into the trapezius muscles and served as EMG electrodes. Finally, the electrode assembly and optical fiber cannula were anchored and fixed to the skull with Super-Bond (Sun Medical).

EEG/EMG recordings

The EEG/EMG signals were amplified (gain ×1000) and filtered (EEG, 1–100 Hz; EMG, 10–100 Hz) using a DC/AC differential amplifier (3000, A-M Systems). The input was then received via an input module (NI-9215, National Instruments), digitized at a sampling rate of 1000 Hz by a data acquisition module (cDAQ-9174, National Instruments), and recorded by a custom-made LabVIEW program (National Instruments). We habituated mice sufficiently; in other words, REM sleep (see below, Mouse vigilance state assessment) was often observed, at which point we started recording. Measurements were collected from five mice for 1–4 h/sessions, with one two sessions/mice (total seven sessions).

Mouse vigilance state assessment

EEG/EMG signals were analyzed using MATLAB (MathWorks). The power spectral data of the EEG were obtained using the multispectrogram method. A power spectral profile over a 1–50 Hz window was used for the analysis. We detected each sleep–wake state scored off-line by characterizing 1 s epochs, as previously described (Funato et al., 2016). A wake state was characterized by low-amplitude fast EEG and high-amplitude variable EMG. NREM sleep was characterized by a high-amplitude delta (1–4 Hz) frequency EEG and a low-amplitude tonus EMG. REM sleep was staged based on theta (4–9 Hz)-dominant EEG and EMG atonia.

Fiber photometry

The method for ratiometric fiber photometry has been described previously (Natsubori et al., 2017). An excitation light (435 nm; silver light-emitting diode, Prizmatix) was reflected off a dichroic mirror (DM455CFP, Olympus), focused with a 2× objective lens (numerical aperture 0.39, Olympus) and coupled into an optical fiber (400 μm diameter, 0.39 numerical aperture; catalog #M79L01, Thorlabs) through a pinhole (diameter 400 μm). The light-emitting diode power was <200 μW at the fiber tip. The cyan and yellow fluorescence emitted by YC-nano50 was collected via an optical fiber cannula, divided by a dichroic mirror (DM515YFP, Olympus) into cyan (483/32 nm band path filters, Semrock) and yellow (542/27 nm), and detected by a photomultiplier tube (H10722-210, Hamamatsu Photonics). The fluorescence signals were digitized using a data acquisition module (cDAQ-9174, National Instruments) and simultaneously recorded using a custom-made LabVIEW program (National Instruments). Signals were collected at a sampling frequency of 1000 Hz.

Optogenetic manipulation

An optical fiber (numerical aperture 0.39, Thorlabs) was inserted through the guide cannula. Blue (470 nm) and yellow (575 nm) light was generated using a SPECTRA 2-LCR-XA light engine (Lumencor). The blue and yellow light power intensity at the tip of the optical fiber was 0.5–1 mW and 6–8 mW, respectively. During EEG and EMG monitoring, we illuminated ChR2-expressing mice during the wake, NREM, and REM periods. We illuminated Arch-T-expressing mice during the NREM period. For ChR2 activation, blue and yellow light (1 s and 5 s duration, respectively) was used to open and close the step-function type opsin ChR2(C128S) (Berndt et al., 2009). In the control trials, yellow light was used instead of blue light in Tph2-ChR2 mice. For ArchT activation, a 120 s duration of yellow (inhibition) light was used in Tph2-ArchT mice. In control trials, yellow light was used in wild-type mice.

Escitalopram administration and recording

Escitalopram (Tokyo Chemical Industry) was dissolved in normal saline and injected intraperitoneally at 15 mg/kg (Burstein et al., 2017). We used well-habituated mice for drug administration; mice had at least five previous recordings without any prior injection. We obtained EEG/EMG and YC signals for 1 h before drug administration and compared them with signals from 2 to 3 h after administration. To compare REM sleep latency, another cohort was treated with escitalopram or saline after habituation.

Histology

Mice were deeply anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg) and perfused with 4% paraformaldehyde phosphate buffer solution. Brains were removed from the skull and postfixed in the same fixative overnight. Subsequently, the brains were cryoprotected in 20% sucrose overnight, frozen, and cut at a 2 μm thickness on a cryostat. Sections were mounted on silane-coated glass slides (Matsunami Glass). The sections were incubated overnight with anti-GFP antibodies (1:200; goat polyclonal, Rockland) at room temperature and then incubated with anti-goat IgG antibody conjugated to Alexa Fluor 488 (1:1000; Invitrogen) for 2 h at room temperature. Fluorescence images were obtained using an all-in-one microscope (BZ-X710, Keyence).

Data processing

All animals and trials were randomly assigned to an experimental condition. Experimenters were not blinded to the experimental conditions during data collection and analysis. Mice were excluded when the optical fiber position was not correctly targeted. Fiber photometry and EEG/EMG signals were analyzed using custom-made programs in MATLAB. Yellow and cyan fluorescence signals were fitted using an exponential function to counteract the fading of fluorescent proteins and autofluorescence of optical fibers. We then used the YC ratio (R), which is the ratio of yellow to cyan fluorescence intensity, for calculating neural activity. We derived the value of the photometry signal (Δ R/R0) by calculating (R – R0) / R0, where R0 was the baseline fluorescence signal (signals in the wake state). For normalization of activity intensity, population 5-HT activities during the wake period were regarded as zero, and the REM period was regarded as −1. We defined the baseline at −0.05 to omit small fluctuations and/or baseline trends. We defined a single concave wave as an event that had a trough below −0.4. EEG signals were bandpass [finite impulse response (FIR), Kaiser window] filtered. We set the EEG frequency bands as follows: delta, 1-4 Hz; theta, 4–9 Hz; alpha, 9–12 Hz; beta, 12–30 Hz; and gamma, 30–50 Hz (Choi et al., 2010). We obtained the power of each EEG frequency band by the square of the amplitude (see Figs. 2b, 4e, 5g, 6e,f). The amplitude of the EMG signal was squared to obtain EMG power (see Figs. 3, 4d, 6f).

Power spectrum analysis of fluorescence signal

Power spectrogram of a 10 Hz resampled YC ratio was calculated with a wavelet transform using a Morse wavelet with a symmetry parameter (γ = 3) and time-bandwidth parameter (P2 = 60).

Analysis of infraslow oscillations of EEG power

We selected uninterrupted NREM sleep bouts that lasted >120 s. We used EEG signals from the last 120 s of NREM bouts. EEG signals were filtered in a wideband (1–50 Hz) or each frequency band (see above, Data processing) using FIR filter (Kaiser window). EEG power time series were obtained by squaring the preprocessed EEG signal, downsampled to 10 Hz, Z-scored, and high-pass filtered with a 0.01 Hz cutoff frequency. Power spectrum of the EEG power fluctuation was calculated with a fast Fourier transform using a Hamming window (see Figs. 2d, 5h, 6d). The strength of the infra slow EEG power fluctuation was quantified as the area of the power underneath the Fourier transform from 0.01 to 0.04 Hz, subtracting the mean power between 0.08 and 0.12 Hz (Lecci et al., 2017; Osorio-Forero et al., 2021; see Fig. 2e).

Statistical analysis

Data for all experiments were analyzed using the following parametric statistics: Student's t test (independent samples t test), paired t test, one-way ANOVA followed by the Tukey–Kramer post hoc test, and linear mixed-effect (LME) model.

Linear mixed-effect model analysis

We conducted an LME model analysis (Yu et al., 2022) to test the effects of blue and yellow light illumination on EMG amplitudes of individual mice as grouping variables, formulated as follows: Yij=β0 + xij,1β1 + xij,2β2 + zij,1u1 + zij,2u2+... + zij,7u7 + εij, where Yij is the EMG amplitude of a mouse j (= 1, …, 7) in an observation i (= 1, …, nj); β0, β1, and β2 are the fixed-effects coefficients for the intercept; blue and yellow light illumination, u1–7, is the random effect for each mouse; εij is the observation error for observation i of a mouse j; and xij,1-2 and zij,1-7 are dummy variables for fixed and random effects, respectively.

Data availability

The datasets generated during and/or analyzed in the current study are available from the corresponding author on reasonable request.

Results

Dorsal raphe 5-HT neurons showed oscillatory population-level activity during NREM sleep

To investigate population-level activity of 5-HT neurons in the DR during the sleep–wake cycle, we used a fiber photometry system and monitored intracellular calcium signals from 5-HT neurons in the DR of freely moving mice (Fig. 1a). We used transgenic mice expressing a FRET-based ratiometric Ca2+ indicator, YC-nano50 (Horikawa et al., 2010), in 5-HT neurons under the control of the Tph2 promoter (Tph2-YC mice; Fig. 1b,c). The ratio of yellow to cyan fluorescence intensities (YC ratio) represents a compound Ca2+ activity of 5-HT neurons. Because fluorescence intensities of these two colors exhibit inversely proportional dynamics according to changes in Ca2+ concentration, the YC ratio is suited for detecting both decreases and increases in Ca2+ concentration (Tsutsui-Kimura et al., 2017; Yoshida et al., 2019).

Figure 1.
Figure 1.

DR 5-HT neurons in mice showed oscillatory population activity during NREM sleep. a, Schematic illustration of the fiber photometry system used for monitoring DR 5-HT neuron activity. EEG and EMG were recorded simultaneously. PMT, Photomultiplier tube. b, Genetic construct of Tph2-tTA::tetO-YC-nano50 double transgenic mice. c, The compound Ca2+ signal in the DR was monitored via an optic fiber (asterisk indicates the tip of the fiber). A fluorescence image shows YC-nano50 expression in the raphe nucleus. Scale bar, 1 mm. d, Representative examples of acquired data, including a trace of the EEG signal, a relative EEG power spectrogram, a trace of the EMG signal, the sleep–wake states (wake, green; REM, red; NREM, blue), population Ca2+ dynamics of DR 5-HT neurons, and the wavelet spectrogram of 5-HT population-level activity. Light blue lines indicate NREM→Wake epochs, purple lines indicate NREM→REM epochs, and red lines indicate REM epochs. Numbers along the bottom indicate time. Red triangles indicate the time when the infaslow power of 5-HT activity was high. e, Normalized DR 5-HT neuron activities across the sleep–wake cycle (7 sessions from 5 mice). Averaged DR 5-HT neuron activity during NREM sleep was lower than during the wake stage but higher than during REM sleep (one-way ANOVA followed by Tukey–Kramer post hoc test, F(2,12) = 138.4, wake vs NREM, p = 5.9 × 10−4; NREM vs REM, p = 3.3 × 10−7; wake vs REM, p = 4.0 × 10−9). f, g, Representative population activity dynamics of DR 5-HT neurons during the NREM→Wake epoch (f) and the NREM→REM epoch (g). The solid and dashed lines show −0.05 and −0.4 of normalized activities, respectively. Waves that exceeded −0.4 are marked by black triangles. Time 0 in (f) corresponds to 365 s in (d), and time 0 in (g) corresponds to 775 s in (d). h, Durations of the NREM→Wake epoch (n = 29 epochs from 5 animals) and the NREM→REM epoch (n = 38 epochs from 5 animals) were comparable (independent t test, p = 0.74, df = 8, t = −0.33). i, Numbers of concave waves during NREM→Wake and NREM→REM were comparable (independent t test, p = 1.0, df = 8, t = 1.3 × 10−3). j, Wavelet spectrogram of infraslow 5-HT population activity during an NREM state (<0.1 Hz, 7 sessions from 5 mice, total 21 epochs). Time is normalized from 0 (start of NREM sleep) to 1 (end of NREM). Power is normalized across time 0–0.1. Error bars indicate SD; *p < 0.05.

We observed changes in the population activity of DR 5-HT neurons across the sleep–wake cycle (Fig. 1d). Sleep–wake stages were identified with EEG and EMG measurements (Fig. 1d). During wake periods (Fig. 1d, green bars in the hypnogram), population activity of the DR 5-HT neurons showed small amplitude fluctuations. During NREM sleep (dark blue bars), the average population activity was lower than that during the wake period. During REM sleep (red bars), population activity of the DR 5-HT neurons was at a minimum (Fig. 1e; normalized 5-HT activities of seven sessions from five mice, 1–4 h per sessions for a total 17 h; mean 5-HT activities during wake and REM states were normalized to 0 and −1, respectively). These observations were consistent with previous findings obtained through electrophysiological studies examining single neuron activity (Jacobs and Azmitia, 1992).

We found an oscillatory population activity of DR 5-HT neurons during NREM sleep, regardless of the type of NREM epoch; one type of epoch was NREM flanked by wake periods (NREM→Wake; Fig. 1f), and the other was NREM flanked by a wake period on one side and a REM period on the other (NREM→REM; Fig. 1g). A small concave wave of 5-HT activity appeared during the first third to half of either the NREM→Wake or NREM→REM epoch (Fig. 1f,g). The 5-HT population activity of each concave wave returned from the trough to its baseline, which was comparable to the activity level observed during a wake period. Hereafter, we defined both types of NREM epochs as a period including more than three concave waves (downward Ca2+ waves) with a greater than −0.4 trough. Using these criteria, the duration of the NREM→Wake and NREM→REM epochs was similar (Fig. 1h; mean ± SD, 207 ± 95 s and 222 ± 36 s, respectively; p = 0.74, independent t test, n = 29 and 38 epochs, respectively, from five animals) and the number of concave waves during an epoch of NREM→Wake or NREM→REM was also similar (mean ± SD, 6.1 ± 2.3 and 6.1 ± 1.6, respectively; p = 1.0, independent t test; Fig. 1i). A difference in 5-HT neuron activity during NREM was found for the trough level; in an NREM→Wake epoch the trough level of the last concave wave did not reach −1 (mean ± SD, −0.87 ± 0.07; Fig. 1f), but in an NREM→REM epoch the activity of 5-HT neurons decreased further to the lowest level of 5-HT activity (−1; Fig. 1g).

We conducted a wavelet analysis to describe the more detailed characteristics of DR 5-HT neuron activity fluctuation during the NREM state (Fig. 1d). Wavelet analysis found slow waves oscillated at 0.01–0.3 Hz (Fig. 1d). The power of infraslow waves gradually increased and reached a peak at the end of the NREM epoch (Fig. 1d, red triangles). We further focused on infraslow oscillation (<0.1 Hz) and averaged the power after normalization (Fig. 1j). The power of waves (0.02–0.05 Hz) increased in the first half of the NREM sleep, and in the second half toward the end of NREM sleep, there was another increase in the power of waves (0.01–0.03 Hz). In summary, the trough of concave waves deepened, and their duration elongated across time, reaching a maximum by the end of the NREM sleep period.

Low DR 5-HT neuron population activity was accompanied by a wideband EEG power increase, and high DR 5-HT neuron activity was accompanied by an EMG power increase

Low 5-HT neuron activity may result in altered cortical EEG signals because DR 5-HT neurons directly innervate most cortical regions or affect them via the relay of subcortical brain regions (Muzerelle et al., 2016). We noticed vertical stripes in the heatmap of the EEG spectrogram during NREM sleep (Fig. 2a, top; see also Fig. 1d second row), and thus we asked whether cortical EEG fluctuation was associated with the oscillatory 5-HT activity during NREM sleep. To visually enhance the vertical stripes mentioned above, we normalized EEG power at frequencies up to 50 Hz at intervals of 1 Hz (Fig. 2a, bottom). The time course of EEG power in wideband (1–50 Hz) and each frequency band ranged from gamma to delta waves, exemplifying the periodic increase during NREM sleep (Fig. 2b) and showing a strong inverse correlation to 5-HT activity (correlation coefficient of −0.7 or lower; Fig. 2b,c). The EEG power fluctuated at ∼0.01–0.05 Hz (Fig. 2d,e). In summary, we found a transient and repetitive wideband EEG power increase that was associated with lowered 5-HT activity during NREM sleep.

Figure 2.
Figure 2.

The increase in wideband EEG power repeatedly coincided with a decrease in DR 5-HT neuron activity during NREM sleep. a, The spectrogram of relative EEG power during NREM sleep. Top, Periodic yellow-green stripes were observed. Bottom, Normalized EEG power for every 1 Hz between 0 and 50 Hz. b, Temporal changes of wide (1–50 Hz), gamma, beta, sigma, theta, and delta frequency bands. R shows the correlation coefficient between each frequency of EEG power and 5-HT activity (n = 6 mice, mean ± SD). c, Temporal changes in population DR 5-HT neuronal activity. Vertical dotted lines indicate troughs of DR 5-HT neuron activities. Time 0 s corresponds to 775 s in Figure 1d. d, Power spectra of EEG power fluctuations from wide gamma, beta, sigma, theta, and delta frequency bands during NREM sleep (7 sessions from 5 mice, total 53 epochs). Gray shade represents SD. e, Strengths of infraslow EEG fluctuation patterns. There was no significant difference between each frequency band (Friedman rank sum test, χ2 = 4.32, df = 4, p = 0.36, n = 5 mice). Error bars indicate SD.

We found occasional EMG amplitude increases during NREM sleep in addition to during the transition from sleep to wake (Fig. 1d). EMG amplitude increases appear at the transition from sleep to wake and last for >4 s at the beginning of a wake period. Such an EMG amplitude increase is always associated with an increase in population-level DR 5-HT neuron activity as seen at the end of both NREM→Wake and REM sleep periods. During NREM sleep, the EMG amplitude increase did not last for >4 s, but this type of EMG amplitude increase was also associated with an increase in DR 5-HT neuron activity. We aligned the 5-HT neuron activity, EEG power, and EMG power during an NREM state and found that 5-HT neuron activity was elevated concurrently with an EMG power increase (Fig. 3a). After the sleep–wake transition, 5-HT neuron activity remained high, the magnitude of the EMG power increase was high, the duration of EMG power increase was long, and the EEG spectrum showed a wake state (Fig. 3b, right). On the other hand, EMG amplitude increases during NREM were lower than those seen during wake periods (mean ± SD, 0.074 ± 0.067 vs 0.25 ± 0.14 mV2; p = 0.02, df = 4, t = −3.5, paired t test), and these events were also associated with increased 5-HT neuron activity, but 5-HT neuron activity levels declined instantly after the peak of 5-HT activity (Fig. 3b, left). The EEG delta power decrease did not last for >4 s. These latter features fit with the concept of microarousal or microawakenings during NREM sleep (Halász et al., 2004; Oikonomou et al., 2019). These results indicate that an increase in population-level DR 5-HT neuron activity during sleep is associated with either the sleep–wake transition or microawakenings.

Figure 3.
Figure 3.

Transient surges of DR 5-HT population neuron activities were associated with microawakenings during NREM sleep or during the sleep–wake transition. a, Temporal changes in population DR 5-HT neural activity (top) and EMG power during NREM) sleep (bottom). Vertical dotted lines indicate the time points when the declined DR 5-HT neuron activities returned to their peak levels. Time 0 s corresponds to 2655 s in Figure 1d. b, Mean relative EEG power (top), DR 5-HT population activity (middle), and EMG power (bottom) aligned at the initial time point of the EMG signal elevation during NREM sleep (left column) or during the transition to wake (right column; 143 time points of microawakenings or 79 time points of NREM→Wake from 5 mice). Gray shade represents SD.

Activation of the 5-HT neurons induced wakefulness

To determine the causal relationship between DR 5-HT neuron activity and the EMG amplitude increase during sleep, we used transgenic mice in which only 5-HT neurons express the step-function-type variant of ChR2 (Tph2-ChR2(C128S); Miyazaki et al., 2014; Fig. 4a) and artificially activated their DR 5-HT neurons for 10 s during NREM sleep. Mice received 1 s of blue light illumination to open ChR2, followed by 5 s of yellow light illumination to close ChR2, 10 s after the blue light illumination (Fig. 4a, middle). Optogenetic activation during NREM and REM sleep increased the EMG amplitude (Fig. 4b–d; n = 7 mice, NREM, 57 stimulations; REM, 18 stimulations) and decreased the delta power of the EEG (Fig. 4e), indicating a transition to the wake state. Regardless of sleep type, optogenetic activation always induced wakefulness, but the degree of increase in EMG power was low when the REM sleep was targeted (Fig. 4d). An induced wake state persisted after optogenetic activation during NREM or REM state (mean ± SD, 77 ± 25; n = 6 mice, 48 stimulations or 62 ± 31 s, n = 4 mice, 10 stimulations; Fig. 4f). This duration was comparable to that of the wake period seen in control conditions during the light phase of the circadian cycle (mean ± SD, 66 ± 30 s from eight mice, 212 periods, Fig. 4f). However, application of the control yellow light (Fig. 4a, bottom) to Tph2-ChR2(C128S) mice during their natural sleep state did not induce EEG or EMG changes (Fig. 4d,e; n = 5 and 3 mice; NREM, 62 stimulations; REM, 16 stimulations). Blue light illumination in wild-type mice did not induce EEG or EMG changes during sleep (n = 3 mice, wake, 13 stimulations; NREM, 22 stimulations; REM, 17 stimulations). Artificial activation of DR 5-HT neurons during the wake state did not change EEG or EMG amplitude (n = 6 mice, 25 stimulations; Fig. 4d).

Figure 4.
Figure 4.

Activation of the DR 5-HT neurons induced an EMG amplitude increase and wakefulness. a, Schematic illustration of optogenetic activation of DR 5-HT neurons in a Tph2-ChR2(C128S) mouse. Scale bar, 1 mm. Blue and yellow shade indicates illumination times. b, c, EEG, EEG power spectrogram, and EMG before and after optogenetic activation during NREM sleep (b) and REM sleep (c). The vertical yellow lines indicate the illumination times. d, Comparison of mean EMG power during a 10 s period before (preillumination), during (illumination), and after (postillumination) activation. An EMG power increase was triggered by optogenetic activation and sustained afterward [LME, NREM, n = 7, df = 168, p (preillumination-illumination) = 1.8 × 10−8, t = 5.92, p (preillumination-postillumination) = 2.5 × 10−8, t = 5.85; REM, n = 4 mice, df = 51, p (preillumination-illumination) = 0.02, t = 2.31, p (preillumination-postillumination) = 5.0 × 10−3, t = 2.93]. DR 5-HT optogenetic activation during a wake period did not alter EMG power (paired t test, n = 6 mice, df = 5, p = 0.72, t = −0.37). Yellow light illumination did not alter EMG power (paired t test, NREM, n = 7 mice, p = 0.20, df = 6, t = −1.44; REM, n = 3 mice, p = 0.10, df = 2, t = −2.9). EMG power of control mice was used as the average EMG power during a 10 s period immediately after spontaneous arousal from NREM or REM sleep in Tph2-YC mice (n = 7 and 9 sessions, respectively). Optogenetic 5-HT activation during NREM sleep increased the EMG power to a level comparable to control mice (independent t test, p = 0.35, df = 10, t = 0.98), whereas activation during REM sleep did not increase the EMG power to the level of control mice (independent t test, p = 0.01, df = 7, t = 3.30). e, Mean EEG delta power for the 10 s before, during, and after optogenetic activation with cyan light during NREM. A significant delta power decline was induced and sustained [LME, n = 6 mice, df = 156, p (preillumination-illumination) = 1.5 × 10−7, t = −5.50, p (preillumination-postillumination) = 2.4 × 10−3, t = −3.08]. Yellow light illumination did not alter EEG delta power (paired t test, n = 5 mice; p = 0.11, df = 4, t = 2.1). f, Duration of a wake period in Tph2-YC mice (left, n = 8, 1–2 sessions per mouse, 212 periods) and duration of the induced wake period after light illumination in Tph2-ChR2 mice during a NREM period (middle, n = 6, 3 sessions per mouse, 48 stimulations), and during an REM period (right, n = 4, 3 sessions per mouse, 10 stimulations). There were no significant differences between the three groups (one-way ANOVA, F(2,15) = 0.41, p = 0.67). Error bars indicate SD; *p < 0.05.

Inhibition of 5-HT neurons occasionally induced REM

At the transition from NREM to REM sleep, DR 5-HT neuron activity no longer oscillated, and the lowered DR 5-HT neuron activity during NREM shifted to that observed in REM sleep. We sought to determine whether a continuous inhibition of DR 5-HT neuron activity induced REM sleep. For this inhibition, we used transgenic mice (Tph2-ArchT) harboring an inhibitory opsin, ArchT, in 5-HT neurons (Fig. 5a). To choose the duration of illumination, we measured the latency from the starting time point of the last concave wave during an NREM→REM epoch to the trough time point of the following REM (Fig. 5b) and found that it ranged from 31 to 65 s (mean ± SD, 46 ± 16 s, n = 61 transitions; Fig. 5c). We thus chose 120 s as the illumination duration to allow for a margin of error.

Figure 5.
Figure 5.

Inhibition of DR 5-HT neuron activity increased the probability of transition to REM or sustained NREM. a, Schematic illustration of optogenetic inhibition of DR 5-HT neurons in Tph2-ArchT mice. Scale bar, 1 mm. Yellow shade indicates light illumination. b, Latency from the start time of the last concave wave of an NREM→REM epoch to the time at the trough of the REM state that follows (gray shade). c, Quantification of b (n = 5 mice, 61 transitions from NREM to REM). Error bars indicate SD. d, f, Representative time courses of EEG, EEG power spectrogram, normalized EEG power (f), ratio of theta power to delta power of EEG, and EMG, with optogenetic inhibition (yellow), respectively, from top to bottom. REM-induced trial (d), NREM-sustained trial (f). e, Percentage of REM-sleep-induced trials in Tph2-ArchT mice (80 sessions from 7 mice) and control mice (40 sessions from 4 mice). g, Optogenetic inhibition induced a wideband EEG power increase (paired t test, n = 7 mice, 7 sessions; delta, p = 0.03, df = 6, t = −2.86; theta, p = 4.9 × 10−2, df = 6, t = −2.46; delta, p = 0.04, df = 6, t = −2.65; beta, p = 5.2 × 10−3, df = 6, t = −4.27; gamma, p = 0.02, df = 6, t = −3.22) and did not lead to changes in wild-type mice (paired t test, n = 4 mice, 4 sessions; delta, p = 0.08, df = 3, t = −2.63; theta, p = 0.28, df = 3, t = 1.32; sigma, p = 0.75, df = 3, t = 0.35; beta, p = 0.68, df = 3, t = 0.46; gamma, p = 0.30, df = 3, t = −1.26) in NREM-sleep-sustained trials. Error bars indicate SD; *p < 0.05. h, Comparable fluctuation of a wideband EEG power during (yellow) and before (black) inhibition of 5-HT neurons by yellow light illumination (13 sessions from 7 mice, total 45 epochs each, paired t test for every 0.01 Hz with a Bonferroni correction for a familywise error rate at p = 0.05). Gray and yellow shades represent SD.

We then applied optogenetic inhibition during a 3 h polysomnography recording during the light phase of the circadian cycle. We identified NREM sleep on-line and initiated 120 s of illumination at the first trial. We had an interval of at least 20 min before the next trial. As a result, we applied illumination 10 times on average during two recording sessions from each animal. In Tph2-ArchT mice (n = 7), of 80 trials, 12 trials induced REM sleep (15%; Fig. 5e, left). In controls (n = 4), of 40 trials, 1 trial induced REM sleep (3%; Fig. 5e, right). In comparing these two stimulation conditions, we found that there is a higher probability of REM sleep induction by optogenetic inhibition versus control light stimulation (p = 0.04, Fisher's exact test). Nonetheless, the REM sleep induction rate was low, and NREM sleep was sustained in most trials (83% for Tph2-ArchT mice and 88% for wild-type mice). Of note, optogenetic inhibition after the third session did not increase the REM induction probability.

As we previously showed, oscillatory population DR 5-HT neuron activity corresponds to periodic EEG power fluctuations. Thus, we asked whether EEG power fluctuations could be perturbed by 120 s of optogenetic inhibition, a duration that should contain 3–4 concave waves of DR 5-HT neuron activity. The other possibility is that oscillatory 5-HT neuron activity does not underlie EEG power fluctuations. We consistently found a wideband EEG power increase during the 120 s of optogenetic DR 5-HT silencing (Fig. 5f,g); however, DR 5-HT silencing sustained infraslow EEG power fluctuations (Fig. 5h). Regarding the sustained EEG power fluctuations, we found a comparable occurrence of microawakenings during the 120 s of illumination (41% of trials for Tph2-ArchT mice and 55% for wild-type mice, p = 0.16, Fisher's exact test). These results indicated that DR 5-HT neuron inhibition during NREM is associated with an EEG power increase and is not associated with perturbed fluctuation of the infraslow EEG power and the occurrence of microawakenings.

EEG fluctuation persisted in the absence of oscillatory DR 5-HT neuron activity

To confirm the persistence of EEG fluctuations under the loss of DR 5-HT neuron activity fluctuations, we intervened pharmacologically. Serotonin selective reuptake inhibitors (SSRIs) lead to elevated extracellular 5-HT levels, and 5-HT neuron activity is in turn silenced via activation of 5-HT1A autoreceptors (Gartside et al., 1995). As expected, SSRI treatment eliminated population-level Ca2+ dynamics of DR 5-HT neurons during sleep periods (Fig. 6a). Under this acute SSRI treatment condition, REM sleep latency was significantly delayed compared with saline controls (mean ± SD, 93 ± 24–149 ± 35 min from six saline- and seven SSRI-treated mice, respectively; p = 6.1 × 10−3, df = 11, t = −3.4, independent t test), supporting previous findings (Bridoux et al., 2013). We assumed that the level of population DR 5-HT neuron activity was at a minimum during REM sleep, even under SSRI treatment (Fig. 6b, B) and considered that the population DR 5-HT neuron activity remained at a lower level across the sleep–wake transition (Fig. 6b,c). As a result, acute SSRI treatment led to high extracellular 5-HT levels, low DR 5-HT neuron activity, and a loss of oscillatory DR 5-HT neuron activity.

Figure 6.
Figure 6.

SSRI administration abolished oscillatory DR 5-HT neuron activity but not EEG fluctuations. a, Representative data including a trace of the EEG signal, a relative EEG power spectrogram, a trace of the EMG signal, the sleep–wake states (wake, green; REM, red; NREM, blue), and population Ca2+ dynamics of DR 5-HT neurons. Under control conditions (left), oscillatory activity is observed in DR 5-HT neurons. Following SSRI administration (right), apparent concave waves disappeared. Black triangles show the wideband EEG power increase. b, Enlargement of a. A, Time at wake; B, time at REM sleep. c, The drop rate of population Ca2+ signals at REM sleep. Following SSRI administration, the degree of the Ca2+ signal decline was less than that in controls (independent t test, p = 0.03, df = 4, t = −3.42; 3 sessions from 3 mice). d, Comparable fluctuation of a wideband EEG power after (blue) and before (black) SSRI administration (6 sessions from 6 mice, total 39 epochs each, paired t test for every 0.01 Hz with a Bonferroni correction for a familywise error rate at p = 0.05). Gray and blue shades represent SD. e, SSRI induced a wideband EEG power decrease (paired t test, 5 sessions from 5 mice; delta, p = 0.21, df = 4, t = 1.49; theta, p = 0.01, df = 4, t = 4.37; sigma, p = 0.06, df = 4, t = 2.64; beta, p = 0.04, df = 4, t = 2.95; gamma, p = 0.02, df = 4, t = 3.78) during NREM sleep. f, Optogenetic DR 5-HT activation following SSRI administration induced wakefulness [EMG power, LME, df = 81, p (preillumination-illumination) = 2.3 × 10−15 t = 9.78, p (preillumination-postillumination) = 1.3 × 10−8, t = 6.33; EEG delta power, LME, df = 81, p (preillumination-illumination) = 6.4 × 10−8 t = −5.96, p (preillumination-postillumination) = 3.8 × 10−7, t = −5.53. 3 sessions from 3 mice]. Error bars indicate SD; *p < 0.05.

We observed infraslow periodic EEG power fluctuations under SSRI treatment (Fig. 6a,d), supporting the idea that EEG fluctuations persist under a loss of oscillatory DR 5-HT neuron activity. The difference here, from the optogenetic silencing experiments, was that SSRI treatment induced a wideband EEG power decrease (Fig. 6e), suggesting that extracellular 5-HT controls wideband EEG power. An NREM-to-wake transition also occurred under SSRI treatment, suggesting that the surge of DR 5-HT neuron activity is not engaged in this mechanism. The frequency of microawakenings remained the same before and after SSRI treatment (0.29 ± 0.12 vs 0.28 ± 0.18 times per 1 min from six mice, six sessions, p = 0.94, df = 5, t = 0.08, paired t test), suggesting that a surge of DR 5-HT neuron activity is not necessary to induce microawakenings under SSRI treatment or that undetectable fluctuations of DR 5-HT neuron activity can induce microawakenings during NREM sleep.

Finally, we evaluated responses to optogenetic activation of DR 5-HT neurons under SSRI treatment. Every stimulation induced wake and was associated with an EMG power increase and an EEG delta power decrease (Fig. 6f), suggesting that artificial DR 5-HT neuron activation always induces wake, but other systems continue to operate during the sleep–wake transition under SSRI treatment.

Discussion

The aim of this study was to understand the population-level activities of DR 5-HT neurons during the sleep–wake cycle. Once we identified the population-level DR 5-HT activity patterns, we manipulated them to address the role of serotonin in sleep. We demonstrated that population-level 5-HT activity in the DR was, on average, high during wake states, intermediate during NREM sleep, and low during REM sleep. We found a slow oscillatory population-level activity of 5-HT neurons (0.01–0.05 Hz) during NREM sleep. Oscillatory changes of population-level 5-HT activities coincided with dynamics of EEG power fluctuations in an antiparallel manner.

Other groups have also described population-level 5-HT neuron activities by fiber photometry during sleep. Monitoring of DR 5-HT neuron activity using GCaMP6 revealed oscillatory patterns during NREM sleep (Oikonomou et al., 2019). The duration of the waves was similar to our findings; however, the signal did not return to the level seen in wakefulness periods. This difference may be because of the nature of the Ca2+ sensors used; YC is useful for detecting downward Ca2+ changes, whereas GCaMP was developed to efficiently detect spikes in Ca2+. Further study is needed to address this difference in dynamics using a distinct methodology. For example, the GPCR activation-based (GRAB)5-HT sensor, a probe for extracellular 5-HT, would be ideal. This was recently examined in a study monitoring extracellular 5-HT levels in the basal forebrain, which revealed similar oscillatory dynamics during NREM sleep (Wan et al., 2021). Similar to the GCaMP6 study, although the released 5-HT levels fluctuated, the extracellular 5-HT levels did not return to the levels seen at periods of wakefulness. With recent developments in GRAB5-HT sensor technology, an improved version of it has been reported, having a similar Ca2+ detection pattern as that observed with YC (Japan Neuroscience meeting 2021, symposium); comprehensive results of this new technology are awaited.

In our photometry setup, we believe we captured a sufficient spatial range across the DR and were able to detect YC signals in 5-HT neurons in an unbiased manner. We used an optic fiber with a 400 µm diameter and a 0.39 numerical aperture. According to our previous estimation (Natsubori et al., 2017), this system could detect signals up to 700 µm beneath the tip of fiber. As a result, the shape of the range would be a conical frustrum, with 200 and 470 µm radii and a 700 µm height, suggesting that we covered most of the DR and did not detect signals from the median raphe. Recent single-cell RNAseq studies in DR/median raphe 5-HT neurons identified nine clusters (Huang et al., 2019; Ren et al., 2019; Okaty et al., 2020). Each cluster was located with some spatial bias in DR, but the range defined by the conical frustrum would roughly cover all clusters. Further, previous single-unit recordings from DR 5-HT neurons revealed four types of wake-activating neurons (that is, sleep-inhibiting neurons) and demonstrated their locations (Sakai, 2011). The location of each type varied, but the detection range of our photometry setup would cover most of DR neurons recorded. Together, we assume that we monitored most of DR 5-HT neurons in an unbiased manner in our photometry setup, including all the major clusters identified by single-cell RNAseq and the major cell types identified by single-cell recording. In fact, the patterns of DR 5-HT neuron population activities were indistinguishable between animals, supporting that the same population was targeted across animals in our study. One caveat regarding the range of our system was that we failed to capture the lateral wings of B7 DR 5-HT neurons, which preferentially innervate the lateral thalamus, mamillary nucleus, and cerebellum, and we captured caudal B6 DR 5-HT neurons, which preferentially innervate the hippocampus and lateral septum (Muzerelle et al., 2016).

The population activity of DR 5-HT neurons was lowest during REM sleep. Does this mean that none of the DR 5-HT neurons fire during REM sleep? We believe this is not the case because we did not observe the lowest level when REM sleep first began (Figs. 1g, 5b). It was only after a substantial delay that the population activity level reached its lowest, and still, it exhibited minor fluctuations afterward. To reconcile this decline at the initial phase of REM sleep, we reanalyzed a previous report (Sakai, 2011) in which the author monitored activities from a total of 229 DR 5-HT neurons by single-unit recordings and described varied firing patterns during sleep. The author identified 195 wake-activating 5-HT neurons (4–5 Hz firing at wake periods), 9 wake/REM-activating probable 5-HT neurons (4–5 Hz firing at wake periods), and 25 sleep-activating probable 5-HT neurons (<0.5 Hz firing at wake periods). The author further classified wake-activating 5-HT neurons into four subtypes. Type I and II cells (n = 115, 50%) completely ceased to fire at the REM sleep stage and type III and IV cells (n = 80, 35%) fired at the beginning of REM sleep (<1 Hz) but decreased their firing rates (<0.5 Hz) afterward. Wake/REM-activating neurons (4%) increased their firing rates from 2 Hz at the initial phase of REM to 4 Hz during REM sleep. Sleep-activating neurons (11%) fired at 2 Hz in both NREM and REM sleep. Because the population-level 5-HT neuron activity was higher at the beginning of REM sleep than the level at the nadir during REM sleep, the decrease in firing rate in wake-activating type III and IV cells can be attributed to the decline of population 5-HT neuron activity after the transition to REM. During REM sleep, wake/REM-activating neurons and sleep-activating neurons (total 15%) tonically fire at 2–4 Hz, and these neuron activities may induce fluctuations during REM sleep.

REM sleep induction by optogenetic inhibition of DR 5-HT neurons is controversial. We were only able to induce REM sleep at the first and second sessions of illumination of ArchT. Illumination after the third session failed to induce REM sleep. These data suggested that silencing of DR 5-HT neurons facilitates the induction of REM sleep, but the effect of artificial silencing may be canceled adaptively. Although we induced REM sleep by optogenetic inhibition, the success rate was at most 15%, suggesting that another mechanism was required to induce REM sleep. In the future, it may be worth trying to inhibit the locus ceruleus (LC) noradrenergic (NA) neurons because they fire at 1–3 Hz during the wake period, have reduced activity during NREM sleep, and cease firing during REM sleep (Aston-Jones and Bloom, 1981). Further studies are required to explore the conditions required to induce REM sleep efficiently.

We also found interesting results with optogenetic activation of DR 5-HT neurons during REM sleep. Artificial activation of DR 5-HT neurons induced wakefulness, as we observed decreased power in EEG slow waves and increased EMG power. However, the degree of the EMG power increase was significantly lower than that seen in both the natural wake transition and in an artificial wake transition from NREM sleep (Fig. 4d). These data suggest that artificial DR 5-HT neuron activation is sufficient to induce a sleep–wake transition, but it cannot fully control REM sleep atonia. REM sleep atonia is mediated by the activation of the sublaterodorsal nucleus (SLD; Arrigoni et al., 2016). DR 5-HT neurons regulate SLD activation by direct connection to SLD (Semba, 1993) or by the DR–laterodorsal tegmental nucleus–SLD indirect pathway (Jones, 1990). These anatomic data partially support the EMG increase we observed during REM sleep by artificially activating DR 5-HT neurons and imply a circuit mechanism that recapitulates the same EMG increase at the REM to wake natural transition.

Results with our optogenetic and pharmacological interventions on DR 5-HT neuron activity did not support the idea that DR 5-HT neurons generate EEG fluctuations during NREM as EEG fluctuations persisted during the silencing of DR 5-HT neurons. These data indicated the presence of another source that yields EEG fluctuations. The NA system would be a good candidate source for the shaping of EEG fluctuations. A recent study reported infraslow oscillations of extracellular NA levels during NREM sleep (Osorio-Forero et al., 2021) and demonstrated an inverse correlation with the sigma band power of EEG and a causal relationship between LC NA neuron activity and σ-band power (Osorio-Forero et al., 2021). This suggested that the oscillatory activity of DR 5-HT neurons partially mirror the effects of the NA system on NREM sleep structure. Indeed, there is a reciprocal functional connection between LC NA neurons and DR 5-HT neurons (Pudovkina et al., 2002), yielding the hypothesis that there is cooperation of NA and 5-HT systems in forming EEG fluctuations across the whole spectrum. It would be worth monitoring and/or manipulating both systems simultaneously during NREM sleep in the future.

Several questions regarding serotonergic mechanisms in sleep–wake regulation remain. One long-lasting dichotomous debate regarding the role of serotonin is that it is sleep promoting rather than wake promoting (Ursin, 2008; see above, Introduction). Sleep/wake-promoting roles were originally designated by long-term loss-of-function studies; for example, ablation, and the designation sleep promoting meant that the proportion of sleep bouts was enhanced by serotonin. In light of this point, Oikonomou et al. (2019) designed an optogenetic experiment to tonically activate DR 5-HT neurons for 12.5 min and demonstrated that DR 5-HT neurons promote sleep during the illumination period. We also conducted optogenetic activation of DR 5-HT neurons; however, our illumination time was 10 s, and we only examined the sleep/wake state change. Thus, it should be noted that our optogenetic experiments were not designed to address the sleep/wake-promoting effects of DR 5-HT neurons. Instead, our short-term optogenetic activation experiment provided evidence that the artificial activation of DR 5-HT neurons during sleep is sufficient to induce a wake state (Fig. 4).

In conclusion, the population-level activity of DR 5-HT neurons fluctuated during NREM sleep in mice. The temporal association between population DR 5-HT neuron activity and wideband EEG power, as well as the outcomes of optogenetic manipulation of DR 5-HT neurons, indicate that mouse NREM sleep is not monotonous but rather a dynamic state associated with oscillatory population activity of DR 5-HT neurons.

Footnotes

  • This work was supported by Ministry of Education, Culture, Sports, Science, and Technology Grant 20H05894 to K.F.T., Japan Agency for Medical Research and Development Grant JP21dm0207069 to K.F.T., and Japan Society for the Promotion of Science Grants 21H00212 and 19K06944 to N.T.

  • The authors declare no competing interests.

  • Correspondence should be addressed to Kenji F. Tanaka at kftanaka{at}keio.jp

References

  1. ↵
    2. Arrigoni E,
    3. Chen MC,
    4. Fuller PM
    (2016) The anatomical, cellular and synaptic basis of motor atonia during rapid eye movement sleep. J Physiol 594:5391–5414. doi:10.1113/JP271324 pmid:27060683
    OpenUrlCrossRefPubMed
  2. ↵
    2. Aston-Jones G,
    3. Bloom FE
    (1981) Activity of norepinephrine-containing locus coeruleus neurons in behaving rats anticipates fluctuations in the sleep-waking cycle. J Neurosci 1:876–886. doi:10.1523/JNEUROSCI.01-08-00876.1981 pmid:7346592
    OpenUrlAbstract/FREE Full Text
  3. ↵
    2. Berndt A,
    3. Yizhar O,
    4. Gunaydin LA,
    5. Hegemann P,
    6. Deisseroth K
    (2009) Bi-stable neural state switches. Nat Neurosci 12:229–234. doi:10.1038/nn.2247 pmid:19079251
    OpenUrlCrossRefPubMed
  4. ↵
    2. Bridoux A,
    3. Laloux C,
    4. Derambure P,
    5. Bordet R,
    6. Charley CM
    (2013) The acute inhibition of rapid eye movement sleep by citalopram may impair spatial learning and passive avoidance in mice. J Neural Transm (Vienna) 120:383–389. doi:10.1007/s00702-012-0901-0 pmid:23053350
    OpenUrlCrossRefPubMed
  5. ↵
    2. Burstein O,
    3. Franko M,
    4. Gale E,
    5. Handelsman A,
    6. Barak S,
    7. Motsan S,
    8. Shamir A,
    9. Toledano R,
    10. Simhon O,
    11. Hirshler Y,
    12. Chen G,
    13. Doron R
    (2017) Escitalopram and NHT normalized stress-induced anhedonia and molecular neuroadaptations in a mouse model of depression. PLoS One 12:e0188043. doi:10.1371/journal.pone.0188043 pmid:29141007
    OpenUrlCrossRefPubMed
  6. ↵
    2. Cespuglio R,
    3. Walker E,
    4. Gomez ME,
    5. Musolino R
    and (1976) Cooling of the nucleus raphe dorsalis induces sleep in the cat. Neurosci Lett 3:221–227. doi:10.1016/0304-3940(76)90077-X pmid:19604890
    OpenUrlCrossRefPubMed
  7. ↵
    2. Choi JH,
    3. Koch KP,
    4. Poppendieck W,
    5. Lee M,
    6. Shin HS
    (2010) High resolution electroencephalography in freely moving mice. J Neurophysiol 104:1825–1834. doi:10.1152/jn.00188.2010 pmid:20610789
    OpenUrlCrossRefPubMed
  8. ↵
    2. Funato H, et al
    . (2016) Forward-genetics analysis of sleep in randomly mutagenized mice. Nature 539:378–383. doi:10.1038/nature20142 pmid:27806374
    OpenUrlCrossRefPubMed
  9. ↵
    2. Gartside SE,
    3. Umbers V,
    4. Hajós M,
    5. Sharp T
    (1995) Interaction between a selective 5-HT1A receptor antagonist and an SSRI in vivo: effects on 5-HT cell firing and extracellular 5-HT. Br J Pharmacol 115:1064–1070. doi:10.1111/j.1476-5381.1995.tb15919.x pmid:7582504
    OpenUrlCrossRefPubMed
  10. ↵
    2. Gaspar P,
    3. Lillesaar C
    (2012) Probing the diversity of serotonin neurons. Philos Trans R Soc Lond B Biol Sci 367:2382–2394. doi:10.1098/rstb.2011.0378 pmid:22826339
    OpenUrlCrossRefPubMed
  11. ↵
    2. Halász P,
    3. Terzano M,
    4. Parrino L,
    5. Bódizs R
    (2004) The nature of arousal in sleep. J Sleep Res 13:1–23. doi:10.1111/j.1365-2869.2004.00388.x pmid:14996030
    OpenUrlCrossRefPubMed
  12. ↵
    2. Horikawa K,
    3. Yamada Y,
    4. Matsuda T,
    5. Kobayashi K,
    6. Hashimoto M,
    7. Matsu-Ura T,
    8. Miyawaki A,
    9. Michikawa T,
    10. Mikoshiba K,
    11. Nagai T
    (2010) Spontaneous network activity visualized by ultrasensitive Ca(2+) indicators, yellow Cameleon-Nano. Nat Methods 7:729–732. doi:10.1038/nmeth.1488 pmid:20693999
    OpenUrlCrossRefPubMed
  13. ↵
    2. Huang KW,
    3. Ochandarena NE,
    4. Philson AC,
    5. Hyun M,
    6. Birnbaum JE,
    7. Cicconet M,
    8. Sabatini BL
    (2019) Molecular and anatomical organization of the dorsal raphe nucleus. Elife 8:e46464. doi:10.7554/eLife.46464
    OpenUrlCrossRef
  14. ↵
    2. Jacobs BL,
    3. Azmitia EC
    (1992) Structure and function of the brain serotonin system. Physiol Rev 72:165–229. doi:10.1152/physrev.1992.72.1.165 pmid:1731370
    OpenUrlCrossRefPubMed
  15. ↵
    2. Jones BE
    (1990) Immunohistochemical study of choline acetyltransferase-immunoreactive processes and cells innervating the pontomedullary reticular formation in the rat. J Comp Neurol 295:485–514. doi:10.1002/cne.902950311 pmid:2351765
    OpenUrlCrossRefPubMed
  16. ↵
    2. Jouvet M
    (1968) Insomnia and decrease of cerebral 5-hydroxytryptamine after destruction of the raphe system in the cat. Adv Pharmacol (1962) 6:265–279. doi:10.1016/s1054-3589(08)60326-9 pmid:5301639
    OpenUrlCrossRefPubMed
  17. ↵
    2. Kanemaru K,
    3. Sekiya H,
    4. Xu M,
    5. Satoh K,
    6. Kitajima N,
    7. Yoshida K,
    8. Okubo Y,
    9. Sasaki T,
    10. Moritoh S,
    11. Hasuwa H,
    12. Mimura M,
    13. Horikawa K,
    14. Matsui K,
    15. Nagai T,
    16. Iino M,
    17. Tanaka KF
    (2014) In vivo visualization of subtle, transient, and local activity of astrocytes using an ultrasensitive Ca(2+) indicator. Cell Rep 8:311–318. doi:10.1016/j.celrep.2014.05.056 pmid:24981861
    OpenUrlCrossRefPubMed
  18. ↵
    2. Koella WP,
    3. Feldstein A,
    4. Czicman JS
    (1968) The effect of para-chlorophenylalanine of the sleep of cats. Electroencephalogr Clin Neurophysiol 25:481–490. doi:10.1016/0013-4694(68)90158-2 pmid:4182602
    OpenUrlCrossRefPubMed
  19. ↵
    2. Lecci S,
    3. Fernandez LMJ,
    4. Weber FD,
    5. Cardis R,
    6. Chatton J-Y,
    7. Born J,
    8. Luthi A
    (2017) Coordinated infraslow neural and cardiac oscillations mark fragility and offline periods in mammalian sleep. Sci Adv 3:e1602026.
    OpenUrlFREE Full Text
  20. ↵
    2. Mcginty DJ,
    3. Harper RM
    (1976) Dorsal raphe neurons: depression of firing during sleep in cats. Brain Res 101:569–575. doi:10.1016/0006-8993(76)90480-7 pmid:1244990
    OpenUrlCrossRefPubMed
  21. ↵
    2. Miyazaki KW,
    3. Miyazaki K,
    4. Tanaka KF,
    5. Yamanaka A,
    6. Takahashi A,
    7. Tabuchi S,
    8. Doya K
    (2014) Optogenetic activation of dorsal raphe serotonin neurons enhances patience for future rewards. Curr Biol 24:2033–2040. doi:10.1016/j.cub.2014.07.041 pmid:25155504
    OpenUrlCrossRefPubMed
  22. ↵
    2. Mouret J,
    3. Bobillier P,
    4. Jouvet M
    (1968) Insomnia following parachlorophenylalanin in the rat. Eur J Pharmacol 5:17–22. doi:10.1016/0014-2999(68)90151-9
    OpenUrlCrossRefPubMed
  23. ↵
    2. Müller CP,
    3. Cunningham KA
    (2009) Handbook of the behavioral neurobiology of serotonin. London: Academic.
  24. ↵
    2. Muzerelle A,
    3. Scotto-Lomassese S,
    4. Bernard JF,
    5. Soiza-Reilly M,
    6. Gaspar P
    (2016) Conditional anterograde tracing reveals distinct targeting of individual serotonin cell groups (B5-B9) to the forebrain and brainstem. Brain Struct Funct 221:535–561. doi:10.1007/s00429-014-0924-4 pmid:25403254
    OpenUrlCrossRefPubMed
  25. ↵
    2. Natsubori A,
    3. Tsutsui-Kimura I,
    4. Nishida H,
    5. Bouchekioua Y,
    6. Sekiya H,
    7. Uchigashima M,
    8. Watanabe M,
    9. de Kerchove d'Exaerde A,
    10. Mimura M,
    11. Takata N,
    12. Tanaka KF
    (2017) Ventrolateral striatal medium spiny neurons positively regulate food-incentive, goal-directed behavior independently of D1 and D2 selectivity. J Neurosci 37:2723–2733. doi:10.1523/JNEUROSCI.3377-16.2017 pmid:28167674
    OpenUrlAbstract/FREE Full Text
  26. ↵
    2. Oikonomou G,
    3. Altermatt M,
    4. Zhang R-W,
    5. Coughlin GM,
    6. Montz C,
    7. Gradinaru V,
    8. Prober DA
    (2019) The serotonergic raphe promote sleep in zebrafish and mice. Neuron 103:686–701. doi:10.1016/j.neuron.2019.05.038 pmid:31248729
    OpenUrlCrossRefPubMed
  27. ↵
    2. Okaty BW,
    3. Sturrock N,
    4. Escobedo Lozoya Y,
    5. Chang Y,
    6. Senft RA,
    7. Lyon KA,
    8. Alekseyenko OV,
    9. Dymecki SM
    (2020) A single-cell transcriptomic and anatomic atlas of mouse dorsal raphe Pet1 neurons. Elife 9:e55523. doi:10.7554/eLife.55523
    OpenUrlCrossRef
  28. ↵
    2. Osorio-Forero A,
    3. Cardis R,
    4. Vantomme G,
    5. Guillaume-Gentil A,
    6. Katsioudi G,
    7. Devenoges C,
    8. Fernandez LMJ,
    9. Lüthi A
    (2021) Noradrenergic circuit control of non-REM sleep substates. Curr Biol 31:5009–5023. doi:10.1016/j.cub.2021.09.041 pmid:34648731
    OpenUrlCrossRefPubMed
  29. ↵
    2. Pudovkina OL,
    3. Cremersab TIFH,
    4. Westerinka BHC
    (2002) The interaction between the locus coeruleus and dorsal raphe nucleus studied with dual-probe microdialysis. Eur J Pharmacol 445:37–42. doi:10.1016/s0014-2999(02)01663-1 pmid:12065192
    OpenUrlCrossRefPubMed
  30. ↵
    2. Ren J,
    3. Isakova A,
    4. Friedmann D,
    5. Zeng J,
    6. Grutzner SM,
    7. Pun A,
    8. Zhao GQ,
    9. Kolluru SS,
    10. Wang R,
    11. Lin R,
    12. Li P,
    13. Li A,
    14. Raymond JL,
    15. Luo Q,
    16. Luo M,
    17. Quake SR,
    18. Luo L
    (2019) Single-cell transcriptomes and whole-brain projections of serotonin neurons in the mouse dorsal and median raphe nuclei. Elife 8:e49424. doi:10.7554/eLife.49424
    OpenUrlCrossRef
  31. ↵
    2. Sakai K
    (2011) Sleep-waking discharge profiles of dorsal raphe nucleus neurons in mice. Neuroscience 197:200–224. doi:10.1016/j.neuroscience.2011.09.024 pmid:21958868
    OpenUrlCrossRefPubMed
  32. ↵
    2. Semba K
    (1993) Aminergic and cholinergic afferents to REM sleep induction regions of the pontine reticular formation in the rat. J Comp Neurol 330:543–556. doi:10.1002/cne.903300410 pmid:7686567
    OpenUrlCrossRefPubMed
  33. ↵
    2. Tanaka KF,
    3. Matsui K,
    4. Sasaki T,
    5. Sano H,
    6. Sugio S,
    7. Fan K,
    8. Hen R,
    9. Nakai J,
    10. Yanagawa Y,
    11. Hasuwa H,
    12. Okabe M,
    13. Deisseroth K,
    14. Ikenaka K,
    15. Yamanaka A
    (2012) Expanding the repertoire of optogenetically targeted cells with an enhanced gene expression system. Cell Rep 2:397–406. doi:10.1016/j.celrep.2012.06.011 pmid:22854021
    OpenUrlCrossRefPubMed
  34. ↵
    2. Torda C
    (1967) Effect of brain serotonin depletion on sleep in rats. Brain Res 6:375–377. doi:10.1016/0006-8993(67)90204-1 pmid:6056718
    OpenUrlCrossRefPubMed
  35. ↵
    2. Trulson ME,
    3. Jacobs BL
    (1979) Raphe unit activity in freely moving acts: correlation with level of behavioral arousal. Brain Res 163:135–150.
    OpenUrlCrossRefPubMed
  36. ↵
    2. Tsunematsu T,
    3. Tabuchi S,
    4. Tanaka KF,
    5. Boyden ES,
    6. Tominaga M,
    7. Yamanaka A
    (2013) Long-lasting silencing of orexin/hypocretin neurons using archaerhodopsin induces slow-wave sleep in mice. Behav Brain Res 255:64–74. doi:10.1016/j.bbr.2013.05.021 pmid:23707248
    OpenUrlCrossRefPubMed
  37. ↵
    2. Tsutsui-Kimura I,
    3. Natsubori A,
    4. Mori M,
    5. Kobayashi K,
    6. Drew MR,
    7. de Kerchove d'Exaerde A,
    8. Mimura M,
    9. Tanaka KF
    (2017) Distinct roles of ventromedial versus ventrolateral striatal medium spiny neurons in reward-oriented behavior. Curr Biol 27:3042–3048. doi:10.1016/j.cub.2017.08.061 pmid:28966085
    OpenUrlCrossRefPubMed
  38. ↵
    2. Urbain N,
    3. Creamer K,
    4. Debonnel G
    (2006) Electrophysiological diversity of the dorsal raphe cells across the sleep-wake cycle of the rat. J Physiol 573:679–695. doi:10.1113/jphysiol.2006.108514 pmid:16613874
    OpenUrlCrossRefPubMed
  39. ↵
    2. Ursin R
    (2008) Changing concepts on the role of serotonin in the regulation of sleep and waking. In: Serotonin and sleep: molecular, functional and clinical aspects (Monti JM, Pandi-Perumal SR, Jacobs BL, Nutt DJ, eds), pp 3–21. Basel: Birkhäuser.
  40. ↵
    2. Wan J,
    3. Peng W,
    4. Li X,
    5. Qian T,
    6. Song K,
    7. Zeng J,
    8. Deng F,
    9. Hao S,
    10. Feng J,
    11. Zhang P,
    12. Zhang Y,
    13. Zou J,
    14. Pan S,
    15. Shin M,
    16. Venton BJ,
    17. Zhu JJ,
    18. Jing M,
    19. Xu M,
    20. Li Y
    (2021) A genetically encoded sensor for measuring serotonin dynamics. Nat Neurosci 24:746–752. doi:10.1038/s41593-021-00823-7 pmid:33821000
    OpenUrlCrossRefPubMed
  41. ↵
    2. Weitzman ED,
    3. Rapport MM,
    4. Mcgregor P,
    5. Jacoby J
    (1968) Sleep patterns of the monkey and brain serotonin concentration: effect of p-chlorophenylalanine. Science 160:1361–1363. doi:10.1126/science.160.3834.1361 pmid:4967983
    OpenUrlAbstract/FREE Full Text
  42. ↵
    2. Whitney MS,
    3. Shemery AM,
    4. Yaw AM,
    5. Donovan LJ,
    6. Glass JD,
    7. Deneris ES
    (2016) Adult brain serotonin deficiency causes hyperactivity, circadian disruption, and elimination of siestas. J Neurosci 36:9828–9842. doi:10.1523/JNEUROSCI.1469-16.2016 pmid:27656022
    OpenUrlAbstract/FREE Full Text
  43. ↵
    2. Yoshida K,
    3. Drew MR,
    4. Mimura M,
    5. Tanaka KF
    (2019) Serotonin-mediated inhibition of ventral hippocampus is required for sustained goal-directed behavior. Nat Neurosci 22:770–777. doi:10.1038/s41593-019-0376-5
    OpenUrlCrossRefPubMed
  44. ↵
    2. Yu Z,
    3. Michele G,
    4. Grieco SF,
    5. Chen L,
    6. Holmes TC,
    7. Xu X
    (2022) Beyond t test and ANOVA: applications of mixed-effects models for more rigorous statistical analysis in neuroscience research. Neuron 110:21–35. doi:10.1016/j.neuron.2021.10.030 pmid:34784504
    OpenUrlCrossRefPubMed
Back to top

In this issue

Email
Print
View Full Page PDF
Citation Tools
Respond to this article
Oscillatory Population-Level Activity of Dorsal Raphe Serotonergic Neurons Is Inscribed in Sleep Structure
Tomonobu Kato, Yasue Mitsukura, Keitaro Yoshida, Masaru Mimura, Norio Takata, Kenji F. Tanaka
Journal of Neuroscience 21 September 2022, 42 (38) 7244-7255; DOI: 10.1523/JNEUROSCI.2288-21.2022
Twitter logo Facebook logo Mendeley logo

Jump to section

Keywords

Responses to this article

Respond to this article

Jump to comment:

No eLetters have been published for this article.