The Serotonergic Raphe Promote Sleep in Zebrafish and Mice

Grigorios Oikonomou

1,x

Michael Altermatt

1,x

Rong wei Zhang

Gerard M. Coughlin

Christin Montz

Viviana Gradinaru

1,*

David A. Prober

1,*,#

Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA

91125, USA.

SUMMARY

The role of serotonin (5-HT) in sleep is controversial: early studies suggested a sleep-promoting

role, but eventually the paradigm shifted towards a wake-promoting function for the serotonergic

raphe. Here we provide evidence from zebrafish and mice that the raphe are critical for the

initiation and maintenance of sleep. In zebrafish, genetic ablation of 5-HT production by the raphe

reduces sleep, sleep depth and the homeostatic response to sleep deprivation. Pharmacological

inhibition or ablation of the raphe reduces sleep, while optogenetic stimulation increases sleep.

Similarly, in mice, ablation of the raphe increases wakefulness and impairs the homeostatic

response to sleep deprivation, whereas tonic optogenetic stimulation at a rate similar to baseline

activity induces sleep. Interestingly, burst optogenetic stimulation induces wakefulness in

accordance with previously described burst activity of the raphe during arousing stimuli. These

results indicate that the serotonergic system promotes sleep in both diurnal zebrafish and nocturnal

rodents.

eTOC blurb

The wake-active serotonergic system (STS) has been considered part of the ascending arousal

system that promotes wakefulness. Using zebrafish and mice, Oikonomou, Altermatt et al.

demonstrate that the STS promotes sleep, potentially by generating homeostatic sleep pressure

during wakefulness.

Keywords

Sleep; arousal; serotonin; 5-HT; raphe; optogenetics; fiber photometry

Correspondence: viviana@caltech.edu (V.G.), dprober@caltech.edu (D.A.P.).

Author Contributions

G.O., M.A., D.P. and V.G. designed experiments. G.O. performed zebrafish experiments. M.A. performed mouse experiments. R.Z.

performed zebrafish electrophysiology. G.M.C. performed mouse histology and quantification. C.M. isolated the zebrafish

tph2

promoter. V.G. supervised rodent aspects of the project. D.A.P. supervised zebrafish aspects of the project. G.O., M.A., V.G. and

D.A.P. wrote the paper with input from R.Z. and G.M.C.

These authors contributed equally

Lead contact

Publisher's Disclaimer:

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our

customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of

the resulting proof before it is published in its final citable form. Please note that during the production process errors may be

discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Declaration of Interests

The authors declare no competing interests.

HHS Public Access

Author manuscript

Neuron

. Author manuscript; available in PMC 2020 August 21.

Published in final edited form as:

Neuron

. 2019 August 21; 103(4): 686–701.e8. doi:10.1016/j.neuron.2019.05.038.

Author Manuscript

INTRODUCTION

The role of serotonin (5-HT) in sleep has been debated for over 50 years (

Ursin, 2008

). The

multiple roles of the serotonergic system (STS) in animal physiology and behavior (

Monti et

al., 2008

;

Müller and Jacobs, 2010

) make dissection of any specific aspect a challenge. In

the central nervous system, the raphe, a diffuse network of brainstem nuclei, synthesize 5-

HT and send projections to almost every region of the brain (

Azmitia and Segal, 1978

Initial studies reported that ablation of the raphe in cats resulted in reduced 5-HT levels and

sleep, in proportion to the size of the lesion (

Jouvet, 1968

). Intraperitoneal injections of

para

-chlorophenylalanine (pCPA), an irreversible inhibitor of tryptophan hydroxylase (TPH,

the rate limiting enzyme in 5-HT synthesis), also reduced sleep (

Koella et al., 1968

;

Mouret

et al., 1968

;

Torda, 1967

;

Weitzman et al., 1968

). This work, combined with studies of other

neurotransmitters, gave rise to the monoaminergic theory of sleep (

Jouvet, 1972

), in which

5-HT plays an important sleep-promoting role. However, this paradigm eventually came into

doubt (

Monti, 2011

;

Ursin, 2008

). Cooling of the dorsal raphe was shown to induce sleep in

cats (

Cespuglio et al., 1976

), while in rats, raphe lesions were reported to cause hyperactivity

with no effect on sleep (

Bouhuys and Van Den Hoofdakker, 1977

). The finding that most

serotonergic raphe neurons are active during wakefulness, less active during non-rapid eye

movement (NREM) sleep and mostly silent during rapid eye movement (REM) sleep

(

McGinty and Harper, 1976

;

Trulson and Jacobs, 1979

), was taken as support for a wake-

promoting role for 5-HT and the raphe. In addition, microdialysis studies showed that brain

5-HT levels follow a similar pattern with levels highest during waking, lower during NREM,

and lowest during REM (

Portas et al., 2000

). Currently, the raphe are widely thought to

promote wakefulness as part of the ascending arousal system (

Saper et al., 2005

;

Scammell

et al., 2017

), although the debate is far from resolved (see Discussion).

We decided to investigate the role of 5-HT and the raphe in sleep using modern optogenetic

and chemogenetic tools that enable neuronal manipulation with superior spatial and

temporal precision. The conserved nature of both the STS and the sleep/wake cycle suggests

that using a simple vertebrate system, which lacks the complex behavioral repertoire of

murine systems that can confound interpretation of sleep studies, could provide insights

translatable to mammals. The zebrafish is a well-established model system for the study of

sleep (

Oikonomou and Prober, 2017

) with extensive anatomical and neurochemical

homology to mammals, and has proven useful in uncovering conserved neuromodulatory

circuits (

Lovett-Barron et al., 2017

). Here we show using both zebrafish and mice that the

serotonergic raphe is a sleep-promoting system that is required for sleep homeostasis, but

can also induce wakefulness in specific contexts. Our results support an evolutionarily

conserved and mode-dependent role for the STS in regulating sleep.

RESULTS

Serotonin receptor agonists increase sleep and inhibition of 5-HT synthesis reduces sleep

in zebrafish

We first sought to verify and expand upon previous pharmacological studies demonstrating a

role for the STS in zebrafish sleep (

Rihel et al., 2010

). We treated zebrafish larvae with the

Oikonomou et al.

Page 2

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

broad spectrum serotonergic receptor agonist quipazine, while monitoring their behavior

using a videotracking system (

Prober et al., 2006

). Animals treated with quipazine showed a

reduction in locomotor activity and increase in sleep compared to vehicle treated controls

(Figures S1A-S1E). A reduction in waking activity (i.e. locomotor activity while awake) was

also observed, indicating that activation of the STS induces an overall reduction in

locomotion during wakefulness. Treatment with buspirone, a broad serotonergic receptor

agonist with partial specificity for the inhibitory receptor 5-HT1A, gave similar results

(Figures S1F-S1J). In both cases, night sleep phenotypes were associated with longer sleep

bouts and reduced sleep latency (time to first sleep bout at night).

To broadly impair serotonergic signaling we used pCPA. Treated larvae had reduced 5-HT

immunoreactivity (Figures S2A and S2B) (

Airhart et al., 2012

), as well as a significant

increase in locomotor activity and decrease in sleep, due to fewer and shorter sleep bouts

and a longer sleep latency (Figures S2C-S2G). These experiments suggest that the STS

promotes sleep in zebrafish.

tph2−/−

zebrafish do not produce 5-HT in the raphe and lack serotonergic brain innervation

In most vertebrates 5-HT is synthesized in several brain regions, including the pineal gland,

preoptic area, posterior tuberculum, hypothalamus, midbrain/pons and medulla oblongata,

with the raphe nuclei spanning the midbrain/pons boundary and medulla (

Lillesaar, 2011

). In

placental mammals, brain 5-HT synthesis is limited to the pineal gland and the raphe nuclei.

Hypothalamic (

Vanhatalo and Soinila, 1998

) and thalamic (

Lebrand et al., 1996

) neurons in

mammals are 5-HT positive although they do not express TPH; instead they appear to take

up 5-HT through the action of serotonin transporters (

Vanhatalo and Soinila, 1998

Staining for 5-HT in the larval zebrafish brain (Figure 1A) revealed several populations of 5-

HT positive cells (

Lillesaar et al., 2009

;

McLean and Fetcho, 2004

). These include a small

pretectal population, and large groups in the posterior tuberculum/hypothalamus which form

a prominent horseshoe-like pattern. The pineal gland, where 5-HT serves as a substrate for

the production of melatonin (Figure S3A), is also 5-HT positive (Figure 1A). In mammals,

the raphe is subdivided into a rostral or superior group that lies on the midbrain/pons

boundary (groups B5-B9,

Dahlstroem and Fuxe, 1964

) and includes the DRN and the

median raphe nuclei (MRN), and a caudal or inferior group in the medulla (groups B1-B3).

In zebrafish larvae, the raphe nuclei are also subdivided into superior raphe (SRa) and

inferior raphe (IRa) (Figure 1A), while the dispersed medullary 5-HT-positive neurons have

been excluded from the raphe designation, presumably because they do not express

tryptophan hydroxylase

(

tph

The zebrafish genome contains three

tph

paralogs (

Bellipanni et al., 2002

;

Teraoka et al.,

2004

), of which only

tph2

is expressed in the raphe (

Teraoka et al., 2004

). We used genome

editing to generate a zebrafish

tph2

predicted null mutant line (

Chen et al., 2013

tph2−/−

mice show severe developmental retardation and postnatal lethality (

Alenina et al., 2009

)

which make interpretation of behavioral assays problematic. In contrast,

tph2−/−

zebrafish

are healthy, develop normally and are born in Mendelian ratios, although some adults

eventually develop scoliosis (data not shown).

tph2

mutant zebrafish lacked 5-HT

immunoreactivity in the raphe and showed a dramatic reduction of 5-HT positive fibers

Oikonomou et al.

Page 3

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

throughout the brain (Figure 1A).

tph2

is also expressed in the pretectal group and the pineal

gland (

Teraoka et al., 2004

). In

tph2−/−

animals, 5-HT immunoreactivity was absent from

the pretectal group, but the pineal gland still contained 5-HT, presumably due to expression

tph1a

(

Bellipanni et al., 2002

), although the signal was reduced compared to sibling

controls (Figure 1A). Interestingly, the medullary cells also lost 5-HT signal in

tph2−/−

animals even though these cells do not express

tph2

, suggesting that they take up

extracellular 5-HT, similar to thalamic and hypothalamic populations in mammals. These

observations corroborate previous studies demonstrating that the

tph2

-expressing raphe are

the main source of serotonergic innervation in the zebrafish brain, similar to mammals

(

Lillesaar et al., 2009

tph2−/−

zebrafish show reduced sleep and altered sleep architecture

We next asked whether genetic loss of 5-HT affects sleep by in-crossing

tph2+/−

animals

and monitoring their progeny.

tph2−/−

animals showed increased locomotor and waking

activity and reduced sleep (Figures 1B-1F) compared to sibling controls. We observed no

such differences between

tph2+/−

and

tph2+/+

animals. The mutant phenotype at night was

due to shorter sleep bouts, with no effect on the number of sleep bouts (Figure 1F),

suggesting that

tph2−/−

mutants enter sleep as often as controls but fail to maintain the sleep

state for the normal amount of time. The mutants also showed an increase in sleep latency at

night (Figure 1F). Thus,

tph2−/−

animals are defective in both initiating and maintaining the

sleep state.

In the pineal gland, 5-HT is converted by arylalkylamine N-acetyltransferase 2 (AANAT2)

to N-acetylserotonin, the substrate for the biosynthesis of melatonin (Figure S3A), a

hormone that regulates sleep in mammals (

Fisher et al., 2013

) and zebrafish (

Zhdanova et

al., 2001

;

Gandhi et al., 2015

). We hypothesized that the reduced sleep in

tph2

mutants could

be caused not by reduced 5-HT in the raphe, but rather by reduced melatonin levels in the

pineal gland due to reduced 5-HT. To test this hypothesis, we used

aanat2

mutant animals,

which fail to synthesize melatonin and exhibit reduced sleep at night (

Gandhi et al., 2015

We reasoned that if the

tph2

phenotype is caused solely by reduced melatonin levels,

aanat2

should be epistatic to

tph2,

that is,

aanat2

;

tph2

double mutants should show the same sleep

phenotype as

aanat2

single mutants. Instead, we found that animals mutant for both

tph2

and

aanat2

slept less than either single mutant (Figures S3B-S3F), indicating that the sleep

phenotypes of

tph2

and

aanat2

are additive and not epistatic to each other. This demonstrates

that the

tph2−/−

phenotype is not due to reduced melatonin levels.

tph2−/−

zebrafish show increased arousal, lighter sleep and attenuated sleep homeostasis

Since

tph2

mutants show reduced sleep, we asked whether they also exhibit a change in

overall arousal. To assess this, we delivered mechano-acoustic stimuli of variable intensities

at one-minute intervals during the night, while monitoring behavior (

Singh et al., 2015

). We

determined the fraction of animals that responded to each stimulus, constructed stimulus-

response curves (Figure 2A), and calculated the tapping power that resulted in half-maximal

response (effective power 50, EP

). The EP

tph2−/−

animals was not significantly

different from that of sibling controls, suggesting that although

tph2

mutants sleep less, they

have a normal arousal threshold. However, the

tph2−/−

response curve had an elevated

Oikonomou et al.

Page 4

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

plateau (Figure 2A), suggesting that

tph2

mutants exhibit increased arousal in response to

strong stimuli.

To assay sleep depth, we applied the mechano-acoustic stimulus with an inter-trial interval

of five minutes. This allows the animals to enter the sleep state between trials; thus, we were

able to assay the response of awake or sleeping fish to a mild stimulus. We found that awake

tph2−/−

animals responded as efficiently as controls (Figure 2B left), but sleeping

tph2−/−

animals showed a significantly higher probability of responding to the stimulus (Figure 2B

right). The observation that

tph2

mutants exit the sleep state more readily suggests a

reduction in sleep depth.

The STS has been implicated in sleep homeostasis (see Discussion), so we asked whether

zebrafish

tph2

mutants are defective in their response to sleep deprivation. To assay this, we

crossed

tph2+/−

tph2−/−

animals and monitored the sleep profile of their progeny over a

48-hour period. On the beginning of the second night we subjected the animals to 6 hours of

sleep deprivation (SD) by maintaining full daylight conditions, and then allowed them to

enter recovery sleep by turning off the lights (Figure 2C). The Normalized SD Response (see

STAR Methods) was significantly reduced in

tph2−/−

animals compared to sibling controls

(Figure 2C). This reduced homeostatic response to sleep deprivation is in accordance with a

previously-proposed role for the STS in sleep homeostasis (

Jouvet, 1999

). According to this

hypothesis, STS activity and concomitant 5-HT release during the wake period are part of

the build-up of homeostatic sleep pressure (see Discussion).

Zebrafish raphe neurons have higher firing rates during the day

Electrophysiological studies have shown that mammalian DRN serotonergic neurons have

higher firing rates during wakefulness than during sleep (

McGinty and Harper, 1976

;

Trulson and Jacobs, 1979

). Serotonergic neurons have a “distinctive neuronal signature” that

is used to identify them (

Jacobs and Fornal, 2010

), consisting of slow tonic firing (1-6 Hz)

that is highly regular, and a long-duration action potential (2-3 ms). We asked whether larval

zebrafish raphe 5-HT neurons show similar properties. We used a

Tg(tph2:eGFP)

line and

fluorescence microscopy to identify such neurons and performed

in vivo

cell-attached

recording to examine their spontaneous activity (Figure 2D). We found that they fire at 1-2

Hz (Figure 2E), with an action potential half-duration of 1-1.5 ms during the day (Figure

2F), similar to mammals. The firing rate was significantly reduced at night (Figure 2E), the

main rest phase of zebrafish, similar to sleeping mammals. We also observed a reduction in

the length of spontaneous action potentials during the night (Figure 2F). These observations

demonstrate conserved electrophysiological properties between raphe 5-HT neurons of

zebrafish and mammals.

Ablation of the raphe results in reduced sleep in zebrafish

Having established a sleep-promoting role for

tph2

in zebrafish, we asked whether cellular

ablation of the raphe has a similar effect on behavior as genetic ablation of

tph2.

The

tph2

promoter is subject to positional effects (

Yokogawa et al., 2012

); thus using a 1.2 kb

promoter fragment and screening multiple lines we generated a

Tg(tph2:eNTR-YFP)

line in

which the

tph2

promoter drives expression of enhanced nitroreductase (

Mathias et al., 2014

;

Oikonomou et al.

Page 5

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

Tabor et al., 2014

) in the raphe, with no expression observed in the pineal gland or pretectal

area (Figures 3A and 3A’). As previously described (

Montgomery et al., 2016

;

Yokogawa et

al., 2012

), we observed sparse expression in the spinal cord for this and other

tph2

transgenic lines (data not shown). Unlike a previous transgene (

Yokogawa et al., 2012

expression was not limited to the SRa, but instead included both the SRa and the IRa (Figure

3A). Nitroreductase converts the inert pro-drug metronidazole (MTZ) into a cytotoxic

compound that causes cell-autonomous ablation (

Curado et al., 2008

). Following treatment

with MTZ (Figures 3D-3F’), but not vehicle (Figures 3A-3C’),

Tg(tph2:eNTR-YFP)

animals

lost both YFP and 5-HT immunoreactivity within the raphe soma and projections. Some 5-

HT immunoreactive projections remained in the medulla, suggesting that these projections

originate from medullary 5-HT positive cells and not the raphe (Figure 3E). After treatment

with MTZ,

Tg(tph2:eNTR-YFP)

animals showed increased locomotor and waking activity,

and reduced sleep, compared to sibling controls (Figures 3G-3K), similar to

tph2

mutants.

These behavioral phenotypes were not simply due to the presence of the transgene, as

vehicle-treated

Tg(tph2:eNTR-YFP)

animals behaved similarly to non-transgenic siblings

(Figures S4A-S4E).

In order to verify that this phenotype is specific to loss of the raphe, we performed 2-photon

laser ablation of the superior and inferior raphe at 4 dpf in

Tg(tph2:eGFP)

animals (Figure

S4F) and assayed their behavior the next day. These animals showed increased locomotor

activity and reduced sleep (Figure S4G-S4K), similar to the chemogenetic ablation

phenotype. These results suggest that the raphe are necessary for normal sleep patterns, and

have an overall sleep-promoting role in zebrafish.

Optogenetic stimulation of the raphe promotes sleep in zebrafish

Having shown that the raphe are necessary for normal amounts of sleep, we next asked

whether their stimulation is sufficient to drive sleep. To address this question, we created a

Tg(tph2:ChR2-YFP)

transgenic line in which ChR2 (

Lin et al., 2009

) is expressed in the

raphe (Figures 4A-4C). We used a modification of the videotracking system (

Singh et al.,

2015

) to expose

Tg(tph2:ChR2-YFP)

animals and non-transgenic sibling controls to 30

minute pulses of blue light during the night, with a 90 minute inter-trial interval. These

animals were maintained under dim light during the day in order to minimize background

activation of ChR2; this still allows for normal wake/sleep cycles (Figures S5A-S5E). Upon

exposure to blue light both populations of animals showed a brief spike in locomotor activity

(presumably due to a startle response) followed by a return to baseline, and then a steady

increase in activity during the period of illumination (Figure 4D). Transgenic animals

showed reduced locomotor activity (Figure 4D) and increased sleep (Figure 4E) during

illumination compared to sibling controls, suggesting that stimulation of the raphe promotes

sleep. Interestingly, when we repeated this experiment using

tph2−/−

animals, stimulation of

the raphe resulted in reduced locomotor activity (Figure 4F) but no increase in sleep (Figure

4G). A fraction of larval zebrafish raphe neurons produce GABA (

Kawashima et al., 2016

which could mediate 5-HT-independent inhibition of locomotor activity. These results

demonstrate that stimulation of the raphe is sufficient to both suppress locomotor activity

and increase sleep, and that the sleep-promoting effect requires 5-HT.

Oikonomou et al.

Page 6

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

The

Tg(tph2:ChR2-YFP)

line that we used showed YFP immunoreactivity in the pineal

gland (Figures 4A’-4C’) in addition to the raphe, although no expression was observed in

the pretectal

tph2

-expressing population (Figures 4A-4C). Melatonin has high membrane

permeability (

Yu et al., 2016

) and is thought to be released by passive diffusion

(

Simonneaux and Ribelayga, 2003

), as opposed to being stored in vesicles which fuse with

the cell membrane upon membrane depolarization. It is therefore unlikely that the sleep

increase observed during exposure of

Tg(tph2:ChR2-YFP)

animals to blue light is due to an

increase in melatonin release by the pineal gland. However, to experimentally address this

possibility, we created

Tg(aanat2:ChR2-YFP)

animals, which express ChR2 exclusively in

the pineal gland (Figures S5F-S5H’), and repeated the optogenetic experiment. We observed

no difference in behavior between transgenic animals and non-transgenic sibling controls

during blue light exposure (Figures S5I and S5J).

DRN neuronal activity correlates with sleep-wake states in mice

Encouraged by our zebrafish studies, we decided to revisit the question of sleep regulation

by the STS in mammals. Previous studies used electrophysiology to record the activity of a

few 5-HT neurons across sleep-wake states (

McGinty and Harper, 1976

) with cell-type

determination based on electrophysiological characteristics, which do not allow for

unequivocal identification of serotonergic neurons (

Allers and Sharp, 2003

;

Kirby et al.,

2003

). To confirm these findings using genetically identified serotonergic neurons and

investigate how the combined activity of large numbers of these neurons changes across

sleep-wake states, we performed simultaneous polysomnographic (EEG/EMG) and fiber

photometry (GCaMP) recordings in the DRN

SERT-cre mice (

Zhuang et al., 2005

), in which the serotonin transporter (SERT) promoter

drives Cre expression specifically in 5-HT neurons, were implanted with EEG and EMG

electrodes and injected with AAV5-Syn-Flex-GCaMP6s or AAV5-EF1a-DIO-eYFP,

followed by optical fiber implantation targeting the DRN (Figures 5A, 5B and S6A). We

quantified the specificity and efficiency of GCaMP6s+ neurons in the DRN of SERT-cre

mice (DRN

SERT-GCaMP6s

), and found that 93±1% of GCaMP6s-expressing neurons co-

expressed TPH2, and that 81±3% of TPH2-expressing neurons co-expressed GCaMP6s. A

representative example of an EEG spectrogram, behavioral state classification, EMG and

DRN

SERT-GCaMP6s

signal is shown in Figure 5C. In agreement with electrophysiological

studies, we observed the highest activity of DRN

SERT

during wakefulness, followed by

NREM and REM. Even though the baseline GCaMP6s signal decreased when animals

transitioned from wake to NREM, we still observed bouts of increased DRN

SERT

activity

during periods classified as NREM. However, spectral analysis of EEG signals at a finer

time scale suggests that these fluorescence peaks actually mark micro-awakenings (Figure

S6B). Due to the short duration (<2 s; window size used for sleep/wake classification is 5 s)

and small magnitude of changes in EMG, these periods were classified as NREM.

DRN

SERT-eYFP

fiber photometry recordings displayed no such fluorescent variations (Figure

S6C). To examine the dynamics of DRN

SERT

population activity over the course of a

particular state, fluorescent signals were extracted for each state, normalized with regard to

time, and averaged across animals (Figure 5D). The summed neuronal activity level

normalized by time was higher during wakefulness than in NREM and REM (Figure S6D).

Oikonomou et al.

Page 7

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

Changes over time in DRN

SERT-GCaMP6s

fluorescence was observed for all states (Figure

5D), but only the decreases over the wake and NREM states were statistically significant

(Figure S6E).

Next, we focused on changes in DRN

SERT

activity at state transitions. Fluorescent signal of

DRN

SERT-GCaMP6s

gradually decreased when an animal transitioned from wake to NREM

(Figure 5E). Conversely, a time-locked increase in DRN

SERT

activity occurred at the

transition from NREM to wake (Figure 5F). When considering the activity during NREM

>90 s before the transition to REM, DRN

SERT

activity was lower during REM than in

NREM (Figure 5G). Interestingly, DRN

SERT-GCaMP6s

fluorescence was lowest just before

the transition from NREM to REM (Figure 5G), in agreement with single unit recordings

(

Trulson and Jacobs, 1979

). When an animal transitioned from REM to wake, there was a

time-locked increase in fluorescence (Figure 5H), similar to the NREM-to-wake transition.

Taken together, these results verify and expand upon observations of putative 5-HT DRN

neurons identified based on electrophysiological criteria, and provide insights into

population activity across states and state transitions.

Ablation of serotonergic neurons leads to increased wakefulness and impairs the

homeostatic response to sleep deprivation in mice

Early raphe physical ablation experiments resulted in reduced sleep (

Jouvet, 1968

), but this

phenotype is complicated by the role of the medullary raphe in thermoregulation (

Tan and

Knight, 2018

). Indeed, animals genetically engineered to lack all 5-HT neurons sleep less at

an ambient temperature of 23°C but not at 33°C (

Buchanan and Richerson, 2010

presumably due to shivering at 23°C, but not at the thermoneutral temperature of 33°C. We

reasoned that targeted ablation of the superior raphe (groups B5-B9) while sparing the

medullary populations (groups B1-B3) would avoid this complication. Therefore, we

implanted SERT-cre mice with EEG and EMG electrodes and performed injections targeting

the superior raphe with either AAV5-EF1a-mCherry-FLEX-dtA, which ablates cells

expressing Cre recombinase, or a negative control virus (AAV5-EF1a-DIO-eYFP or AAV5-

CAG-GFP) (Figures 6A and 6B). Two weeks later, sleep-wake patterns were recorded for 24

h. Subsequent histological analysis showed that only 5-HT neurons in the B5-B9 groups

were ablated (B5-B9

SERT-dtA

: 4±1%, n=8; B5-B9

SERT-eYFP

or B5-B9

GFP

100±3%, n=9;

unpaired t-test, p<0.001), whereas no change was observed in the B1-B3 groups (B1-

SERT-dtA

: 104±6%, n = 8; B1-B3

SERT-eYFP

or B1-B3

GFP

100±5%, n=9; unpaired t-test,

p>0.05). The amount of wakefulness in experimental animals was increased at the expense

of NREM and REM sleep during both the day and night (Figure 6C), in accordance with a

sleep-promoting role for the superior raphe. B5-B9

SERT-dtA

animals had fewer wake bouts

than controls (Figure 6E), but these wake bouts were longer than those of controls (Figure

6D, note log scale for Y-axis), resulting in increased time spent in wakefulness (Figure 6C).

To investigate the consequences of STS ablation on sleep homeostasis, animals were

subjected to 6 h of SD at the beginning of the light phase following 24 h of baseline

polysomnographic recordings. As expected, both B5-B9

SERT-dtA

and B5-B9

SERT-eYFP

or B5-

GFP

animals showed an increase in sleep following SD (recovery sleep), and the amount

of time spent in wake, NREM or REM state during recovery was similar in both populations

Oikonomou et al.

Page 8

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

(data not shown). A critical feature of recovery sleep is an increase in delta power during

NREM (

Borbely and Neuhaus, 1979

). Indeed, both B5-B9

SERT-dtA

and B5-B9

SERT-eYFP

B5-B9

GFP

animals showed a significant increase in delta power during recovery sleep

compared to baseline, although this effect was only significant for the first 2 h of recovery

sleep for B5-B9

SERT-dtA

animals, whereas it remained significant for at least 6 h for B5-

SERT-eYFP

or B5-B9

GFP

animals (Figures S7J and S7K). Furthermore, B5-B9

SERT-dtA

animals showed a significantly smaller increase in delta power compared to B5-B9

SERT-eYFP

or B5-B9

GFP

(Figure 6F), suggesting that SD led to a smaller increase in homeostatic sleep

pressure in B5-B9

SERT-dtA

animals. These results suggest that STS activity is normally

involved in generating homeostatic sleep pressure, in accordance with the hypothesis put

forward by Michel Jouvet (

Jouvet, 1999

; see Discussion).

Optogenetic stimulation of DRN neurons has bidirectional mode-dependent effects on

mouse sleep

In addition to a tonic (~1-6 Hz) baseline firing pattern during wakefulness (

McGinty and

Harper, 1976

), 5-HT neurons also fire in bursts (up to ~30 Hz) (

Cohen et al., 2015

;

Liu et al.,

2014

;

Schweimer and Ungless, 2010

;

Veasey et al., 1995

). Therefore, we asked whether

different stimulation modes can have different effects on sleep/wake states. To answer this

question, we implanted SERT-cre mice with EEG and EMG electrodes, performed bilateral

injections with either AAV5-EF1a-DIO-ChR2-eYFP or AAV5-EF1a-DIO-eYFP, and then

implanted two optical fibers targeting the DRN (Figures 7A, 7B and S6A). We quantified the

specificity and efficiency of ChR2+ neurons in the DRN of SERT-cre mice, and found that

94±1% of ChR2-expressing neurons co-expressed TPH2, and that 90±1% of TPH2-

expressing neurons co-expressed ChR2. We then stimulated the raphe with either a burst or a

tonic pattern (Figure 7C). Importantly, the number of stimuli per trial was the same for both

stimulation modes. Based on results from preliminary experiments, burst stimulation

experiments were conducted during the light phase, when mice are mostly asleep, and tonic

stimulation experiments were conducted during the dark phase, when wake states dominate.

Burst optogenetic stimulation of DRN

SERT-ChR2

animals caused a sudden decrease in

NREM and REM probability, and an increase in wake probability (Figure 7D). Changes in

probability of sleep-wake states during the first minute of optogenetic stimulation were

significantly different between DRN

SERT-ChR2

and DRN

SERT-eYFP

mice (Figure 7F). Tonic

optogenetic stimulation of DRN

SERT-ChR2

caused an inhibition of REM similar to burst

stimulation, however, it also caused an increase in NREM probability at the expense of wake

probability over the course of stimulation (Figure 7G). Changes in probability of sleep-wake

states during the last 5 minutes of optogenetic stimulation were different between

DRN

SERT-ChR2

and DRN

SERT-eYFP

mice (Figure 7I). In DRN

SERT-eYFP

animals, neither

burst nor tonic optogenetic stimulation affected behavioral states (Figures 7E, 7F, 7H and

7I). In order to determine whether optogenetic stimulation promotes initiation of a particular

state, we investigated the latency to state transitions. We found that burst stimulation reduced

the latency from NREM or REM to wake (Figure 7J), while tonic stimulation reduced the

latency from wake to NREM (Figure 7J). We also asked whether optogenetic stimulation

affects the duration of a particular state. Burst stimulation increased the duration of wake

bouts (Figure S7A) and decreased the duration of NREM bouts (Figure S7B) while tonic

Oikonomou et al.

Page 9

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

stimulation increased the duration of NREM bouts (Figure S7B) and decreased the duration

of wake bouts (Figure S7A).

Spectral analysis demonstrated that burst stimulation of DRN

SERT-ChR2

animals reduced

delta power, whereas tonic stimulation increased delta power, compared to DRN

SERT-eYFP

controls (Figure 7K). Conversely, burst stimulation increased, and tonic stimulation

decreased, high frequency power (Figure 7L). These effects on patterns of brain activity are

consistent with wake- and NREM-promoting effects of burst and tonic stimulation,

respectively. Furthermore, as delta power is a measure of homeostatic sleep pressure

(

Borbély and Neuhaus, 1979

), the increase observed under tonic stimulation is consistent

with the idea that baseline STS activity (predominately 1-6 Hz) serves to build homeostatic

sleep pressure (

Jouvet, 1999

To corroborate our results with an independent transgenic line, we performed similar

experiments using the ePet-cre transgenic mouse line (

Scott et al., 2005

) (Figures S7E, S7F

and S7H). Like the SERT-cre line (

Zhuang et al., 2005

), Cre recombinase is specifically

expressed in CNS 5-HT neurons in ePet-cre mice. Surprisingly, neither burst nor tonic

optogenetic stimulation of DRN

ePet-ChR2

animals had an effect on wake or sleep probability.

The ePet-cre mouse line can show imperfect recombination efficiency (

Narboux-Neme et

al., 2013

), so it is possible that fewer neurons were activated in this transgenic background.

Indeed, histological quantification revealed that even though 96±1% ChR2-expressing

neurons co-expressed TPH2, similar to the SERT-cre line (Figure S7C), only 28±2% TPH2-

expressing neurons co-expressed ChR2 in ePet-cre mice, compared to 90±1% in SERT-cre

mice (Figure S7D). This suggests that the lack of behavioral phenotypes during stimulation

of DRN

ePet-ChR2

animals could be due to low levels of 5-HT release compared to SERT-cre

mice. To test this hypothesis, we injected the selective serotonin reuptake inhibitor (SSRI)

fluoxetine in DRN

ePet-ChR2

animals immediately prior to optogenetic stimulation to boost

the optogenetically-induced increase in extracellular 5-HT concentration (

Marcinkiewcz et

al., 2016

). In this context, burst and tonic stimulation resulted in increased wake and NREM

probability (Figures S7G and S7I), respectively, similar to SERT-cre animals.

DISCUSSION

The debate regarding the role of the STS in sleep regulation has been carefully chronicled

elsewhere (

Ursin, 2008

). In short, initial studies based on lesions and pharmacological

inhibition of TPH supported a sleep-promoting role, but later the paradigm shifted in support

of a wake-promoting role, mostly due to the wake-active nature of 5-HT DRN neurons. The

debate has continued in the modern era with pharmacological work generating conflicting

results. For example, ritanserin, an antagonist of the excitatory 5-HT2 receptor family,

promotes sleep in humans (

Idzikowski et al., 1986

) and rats (

Dugovic and Wauquier, 1987

but wakefulness in cats (

Sommerfelt and Ursin, 1993

), while agonists of the inhibitory 5-

HT1A receptor decrease wakefulness when administered locally, but increase wakefulness

when administered systemically (

Portas and Grønli, 2008

). Genetic models have often

contradicted pharmacology (

Adrien, 2008

): mice mutant for 5-HT2A show reduced sleep,

but treatment with a 5-HT2A receptor antagonist increases sleep (

Popa, 2005

). However,

genetic studies can also be difficult to interpret: while embryonic ablation of

tph2

was

Oikonomou et al.

Page 10

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

reported to not alter total sleep levels in mice (

Solarewicz et al., 2015

), these results are

complicated by severe developmental retardation and postnatal lethality in these animals

(

Alenina et al., 2009

). Genetic ablation of

tph2

in the pons/midbrain raphe nuclei (groups

B5-B9) of adult mice was shown to cause hyperactivity and eliminate siestas (

Whitney et al.,

2016

), although ablation of the majority of serotonergic neurons in both the midbrain/pons

and medulla of adults using an ePet-cre line caused a reduction only in REM sleep (

Iwasaki

et al., 2018

). Optogenetic inhibition of 5-HT DRN neurons using ePet-cre blocked arousal in

response to high levels of CO

(

Smith et al., 2018

), suggesting that the DRN is required for

wakefulness induced by hypercapnia.

To limit confounds generated by the diverse role of the STS in animal behavior (

Müller and

Jacobs, 2010

), we first investigated sleep-regulating aspects of the STS in the larval

zebrafish, a diurnal vertebrate which displays robust sleep cycles (

Prober et al., 2006

) but

lacks a complicated behavioral repertoire (

Dreosti et al., 2015

). Our larval zebrafish work

supports a sleep-promoting role for the STS. Serotonin receptor agonists increase sleep and

TPH inhibition reduces sleep.

tph2

mutant animals exhibit reduced and lighter sleep,

increased maximal arousal, and reduced homeostatic response to sleep deprivation. Ablation

of the raphe phenocopies the

tph2

mutant phenotype, while optogenetic stimulation

increases sleep in a

tph2

-dependent fashion. Interestingly, the electrophysiological properties

of the 5-HT raphe neurons of larval zebrafish are similar to those of mammals: they fire at a

regular and slow rate which is elevated during the wake period, and their action potentials

are long-lasting.

In mammals, early studies describing the firing pattern of 5-HT neurons were based on

single-unit recordings and used electrophysiological criteria to differentiate between

serotonergic neurons and other DRN cell types. However, these criteria do not allow for

unequivocal identification of serotonergic neurons as non-serotonergic DRN neurons can

have similar firing properties (

Allers and Sharp, 2003

;

Kirby et al., 2003

). We therefore used

fiber photometry to record the activity of genetically identified 5-HT neurons. Our findings

confirm results from single-unit recordings (

McGinty and Harper, 1976

;

Trulson and Jacobs,

1979

), wherein serotonergic DRN neurons display the highest activity level during the wake

state, followed by NREM and REM. Early studies in which 5-HT production was

pharmacologically inhibited or the DRN were physically ablated showed a reduction in sleep

(

Jouvet, 1968

;

Mouret et al., 1968

), but this phenotype was later attributed to increased

shivering due to disrupted thermoregulation (

Buchanan and Richerson, 2010

;

Murray et al.,

2015

). To avoid this complication, we selectively targeted the superior raphe, while leaving

intact the thermoregulatory populations in the medulla, and observed a reduction in sleep.

Optogenetic activation of the DRN in a tonic pattern, at a frequency similar to that of

endogenous baseline activity, reduced wake and REM probability with a concurrent increase

in NREM probability. Thus, in both zebrafish and mice the STS appears to play a sleep-

promoting role.

Our work in zebrafish and mice, as well as classical studies in mammals (

McGinty and

Harper, 1976

;

Trulson and Jacobs, 1979

) generate an apparent paradox: how can a sleep-

promoting system be more active during wakefulness? Michel Jouvet sought to reconcile

this paradox by proposing a role for the STS in sleep homeostasis (

Jouvet, 1999

). According

Oikonomou et al.

Page 11

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript

to the two-process model (

Borbély, 1982

), sleep is regulated by a circadian component

(process C), which gates the timing of sleep, and a homeostatic component (process S),

which builds sleep pressure during wakefulness. The sleep pressure generated by process S

is eventually relieved when the animal enters the sleep state at the time prescribed by

process C. Jouvet proposed that the activity of the STS and concomitant release of 5-HT

during wakefulness are part of process S and serve to build sleep pressure by measuring the

duration and intensity of waking (

Jouvet, 1999

), presumably along with other somnogenic

molecules such as adenosine (

Porkka-Heiskanen et al., 1997

). Indeed in

Drosophila,

5-HT

promotes sleep and

5HT1

mutants show reduced sleep, as well as reduced rebound sleep

following sleep deprivation, in accordance with a role for 5-HT in measuring sleep debt

(

Yuan et al., 2006

). Flies mutant for

trh

(equivalent to

tph

) or the serotonin receptor

5HT2b

have reduced homeostatic response to sleep deprivation, and a small group of

5HT2b

expressing neurons in the dorsal fan-shaped body is necessary for rebound sleep (

Qian et al.,

2017

). Mice mutant for

5-HT2A

sleep less and, following sleep deprivation, show a smaller

delta power increase compared to controls (

Popa, 2005

), suggesting a reduced homeostatic

sleep drive. Mice mutant for

5-HT2C

also show reduced sleep and disrupted sleep

homeostasis (

Frank et al., 2002

). Here we show that

tph2

mutant zebrafish have reduced

response to sleep deprivation, that raphe-ablated mice show reduced increase in delta power

after sleep deprivation, and that tonic optogenetic activation of the raphe in mice increases

delta power. These results support Jouvet’s hypothesis that the STS promotes sleep, despite

being wake-active, by forming part of the sleep homeostasis system. Thus, the induction of

sleep during optogenetic activation of the STS in zebrafish and mice, described here, could

be due to a transient increase in sleep pressure.

Distinct and even opposing effects during tonic versus burst activity have been reported for

other monoaminergic systems, including dopamine (

Goto et al., 2007

) and noradrenaline

(

Aston-Jones and Cohen, 2005

). Recently, tonic activation of centromedial thalamic neurons

in mice was shown to induce wakefulness, while burst activation induced slow-wave-like

activity and enhanced cortical synchrony (

Gent et al., 2018

). These studies suggest that

different activity modes of the same circuit can have opposite behavioral outcomes. Indeed,

contrary to tonic stimulation, burst stimulation of the DRN in mice increased wake

probability at the expense of NREM and REM. Whereas tonic activity is the typical mode

observed in the raphe (

Jacobs and Azmitia, 1992

), burst activity occurs in specific contexts

including reward/punishment (

Cohen et al., 2015

;

Liu et al., 2014

), treadmill-induced

locomotion (

Veasey et al., 1995

) and noxious stimuli (

Schweimer and Ungless, 2010

), all of

which are presumably arousing. Thus, we propose that the endogenous baseline mode of

tonic activity mediates the sleep-promoting role of the STS, while burst activity is a

component of arousal-inducing behaviors. The correlation between burst activity and

noxious stimuli suggests that hypercapnia-induced arousal might involve burst firing of STS

neurons. The mechanisms that enable different modes of activity by the STS to generate

opposing behavioral outcomes are unclear. In the leech Retzius neuron, tonic stimulation at

1 Hz generates only synaptic release of 5-HT, but burst stimulation at 10 or 20 Hz generates

extrasynaptic release of 5-HT from the soma and axonal varicosities (

Trueta and De-Miguel,

2012

). Thus, different modes of STS activity could translate into different spatial or

quantitative patterns of 5-HT release. Alternatively, tonic and burst modes of activity could

Oikonomou et al.

Page 12

Neuron

. Author manuscript; available in PMC 2020 August 21.

Author Manuscript