Cholinergic and noncholinergic brainstem neurons expressing Fos after paradoxical (REM) sleep deprivation and recovery.

Abstract

It is well accepted that populations of neurons responsible for the onset and maintenance of paradoxical sleep (PS) are restricted to the brainstem. To localize the structures involved and to reexamine the role of mesopontine cholinergic neurons, we compared the distribution of Fos- and choline acetyltransferase-labelled neurons in the brainstem of control rats, rats selectively deprived of PS for approximately 72 h and rats allowed to recover from such deprivation. Only a few cholinergic neurons from the laterodorsal (LDTg) and pedunculopontine tegmental nuclei were Fos-labelled after PS recovery. In contrast, a large number of noncholinergic Fos-labelled cells positively correlated with the percentage of time spent in PS was observed in the LDTg, sublaterodorsal, alpha and ventral gigantocellular reticular nuclei, structures known to contain neurons specifically active during PS. In addition, a large number of Fos-labelled cells were seen after PS rebound in the lateral, ventrolateral and dorsal periaqueductal grey, dorsal and lateral paragigantocellular reticular nuclei and the nucleus raphe obscurus. Interestingly, half of the cells in the latter nucleus were immunoreactive to choline acetyltransferase. In contrast to the well-accepted hypothesis, our results strongly suggest that neurons active during PS, recorded in the mesopontine cholinergic nuclei, are in the great majority noncholinergic. Our findings further demonstrate that many brainstem structures not previously identified as containing neurons active during PS contain cholinergic or noncholinergic neurons active during PS, and these structures may therefore play a key role during this state. Altogether, our results open a new avenue of research to identify the specific role of the populations of neurons revealed, their interrelations and their neurochemical identity.

Full text

Page 1

Cholinergic and noncholinergic brainstem neurons
expressing Fos after paradoxical (REM) sleep deprivation
and recovery
Laure Verret, Lucienne Le ´ger, Patrice Fort and Pierre-Herve ´ Luppi
CNRS UMR 5167, Institut Fe ´de ´ratif des Neurosciences de Lyon (IFR 19), Faculte ´ de me ´decine RTH Laennec, 7, rue Guillaume
Paradin, 69372 Lyon Cedex 08, France
Keywords: acetylcholine, muscle atonia, rat, REM sleep, reticular
Abstract
It is well accepted that populations of neurons responsible for the onset and maintenance of paradoxical sleep (PS) are restricted to
the brainstem. To localize the structures involved and to reexamine the role of mesopontine cholinergic neurons, we compared the
distribution of Fos- and choline acetyltransferase-labelled neurons in the brainstem of control rats, rats selectively deprived of PS for
?72 h and rats allowed to recover from such deprivation. Only a few cholinergic neurons from the laterodorsal (LDTg) and
pedunculopontine tegmental nuclei were Fos-labelled after PS recovery. In contrast, a large number of noncholinergic Fos-labelled
cells positively correlated with the percentage of time spent in PS was observed in the LDTg, sublaterodorsal, alpha and ventral
gigantocellular reticular nuclei, structures known to contain neurons specifically active during PS. In addition, a large number of Fos-
labelled cells were seen after PS rebound in the lateral, ventrolateral and dorsal periaqueductal grey, dorsal and lateral
paragigantocellular reticular nuclei and the nucleus raphe obscurus. Interestingly, half of the cells in the latter nucleus were
immunoreactive to choline acetyltransferase. In contrast to the well-accepted hypothesis, our results strongly suggest that neurons
active during PS, recorded in the mesopontine cholinergic nuclei, are in the great majority noncholinergic. Our findings further
demonstrate that many brainstem structures not previously identified as containing neurons active during PS contain cholinergic or
noncholinergic neurons active during PS, and these structures may therefore play a key role during this state. Altogether, our results
open a new avenue of research to identify the specific role of the populations of neurons revealed, their interrelations and their
neurochemical identity.
Introduction
It is now well accepted that populations of neurons localized in the
dorsal part of the pontine reticular formation, the mesopontine
cholinergic nuclei and the ventromedial medullary reticular formation
play a crucial role in the genesis of paradoxical sleep (PS), also termed
rapid eye movement (REM) sleep (Jouvet, 1962; Jones, 1991). Indeed,
long periods of PS have been elicited in cats and rats by pharmaco-
logical stimulation of the dorsal part of the pontine reticular formation
named the sublaterodorsal nucleus (SLD) in the latter species
(Baghdoyan et al., 1987; Vanni-Mercier et al., 1989; Yamamoto
et al., 1990; Boissard et al., 2002). Further, it contains neurons
specifically active during PS (review in Sakai et al., 2001). Such type
of neurons has been also recorded in the laterodorsal (LDTg) and
pedunculopontine (PPTg) tegmental cholinergic nuclei (El Mansari
et al., 1989; Steriade et al., 1990; Kayama et al., 1992; Sakai &
Koyama, 1996) and the alpha and ventral gigantocellular reticular
nuclei (GiA and GiV), located in the ventromedial medullary reticular
formation (Kanamori et al., 1980). It has been hypothesized that LDTg
and PPTg neurons are cholinergic and responsible for EEG activation
during PS (review in Jones, 1991) while those of the GiA and GiVare
glycinergic and responsible for the muscle atonia of PS (Sakai, 1988;
Fort et al., 1993; Rampon et al., 1996; Boissard et al., 2002).
The lack of information concerning the activity of neurons in other
brainstem structures with regards to PS has been partially overcome
by the use of the immunohistochemical detection of Fos, the protein
product of the immediate–early gene c-fos (Dragunow & Faull, 1989;
Morgan & Curran, 1991). Activated neurons can be mapped at a
larger scale than with electrophysiology and their neurochemical
nature determined. Using this method, authors have shown that the
numbers of Fos-positive cells counted in some discrete nuclei of the
brainstem are positively or negatively correlated with the amount of
PS induced either by carbachol microinjection (Shiromani et al.,
1992; Yamuy et al., 1993; Shiromani et al., 1996; Yamuy et al.,
1998) or by PS deprivation (Merchant-Nancy et al., 1995; Maloney
et al., 1999, 2000), in both cat and rat. However, there are significant
discrepancies in these studies. In particular, two of them reported that
in the LDTg and PPTg the number of Fos-expressing cells identified
as cholinergic is positively correlated with the amount of PS
(Shiromani et al., 1996; Maloney et al., 1999) while another study
failed to confirm these findings (Yamuy et al., 1998). In addition, the
SLD and the GiV were reported to contain a number of Fos-labelled
cells positively correlated with PS only in two of these studies
(Yamuy et al., 1993; Merchant-Nancy et al., 1995).
Therefore, the aim of the present study was to resolve previous
discrepancies by mean of a careful analysis in rats of the distribution
Correspondence: Dr P. H. Luppi, as above.
E-mail: luppi@sommeil.univ-lyon1.fr
Received 13 August 2004, revised 25 January 2005, accepted 2 February 2005
European Journal of Neuroscience, Vol. 21, pp. 2488–2504, 2005
ª Federation of European Neuroscience Societies
doi:10.1111/j.1460-9568.2005.04060.x

Page 2

of the neurons expressing Fos in the brainstem following PS
deprivation, PS recovery after deprivation and control conditions. In
addition, it was aimed to determine whether the cholinergic neurons
from the LDTg and PPTg are Fos-immunoreactive after PS recovery
and therefore indeed correspond to the PS-active neurons recorded in
these nuclei.
Materials and methods
Animals and surgery
Male Sprague-Dawley rats (280–320 g, IFFA Credo, L’Arbresle,
France) were anaesthetized with chloral hydrate (1 g⁄kg, i.p.) and
mounted conventionally in a stereotaxic frame (David Kopf,
Epinay-sur-Seine, France) with ear bars and a head holder. The
bone was exposed and cleaned. Three stainless steel screws were
fixed in the parietal and frontal parts of the skull and three wire
electrodes inserted into the neck muscles to monitor the electro-
encephalogram (EEG) and the electromyogram (EMG), respect-
ively.
The bone was then covered with a thin layer of acrylic cement
(Superbond; Sun Medical Co, Moriyama, Shiga, Japan). A six-pin
connector was embedded in a mount of dental cement with the EEG
screws and wires. Animals were allowed 2 days recovery from
surgery in individual Plexiglas containers before starting the habitu-
ation to the recording cable. Animals were cared for according to the
Guide for the Care and Use of Laboratory Animals (NIH Publication
80–23; Authorization n? 03–505 of the French Ministry of Agricul-
ture).
Recording and PS deprivation
Each rat was kept in its container and connected to a cable attached to
a commutator and suspended with a balanced boom to allow free
movement of the animal within the container. During the baseline day
and in the control condition, the floor of the container was covered
with woodchips. The animal had ad libitum access to food and water
in dispensers that hung within easy reach on the side of the container.
A 12-h light–dark cycle was maintained in the recording room (with
lights on 06.00–18.00 h). The rats were placed in the recording
container (30 cm diameter, 40 cm high) and connected to the cable
8 days before baseline recording to allow for habituation to the
recording environment. EEG and EMG recordings were collected on a
computer via a CED interface using Spike 2 software (Cambridge
Electronic Design, UK).
PS deprivation was performed for ?72 h using the flowerpot
technique, which has previously been shown to cause a selective
deprivation of PS in rats (Mendelson et al., 1974; Maloney et al.,
1999). Each rat was placed on a platform which was surrounded by
water and large enough (6.5 cm in diameter) to hold the animal during
waking (W) and slow wave sleep (SWS), but preventing it from
engaging in PS bouts because of the loss of muscle tone accompany-
ing this state and the correlated risk of falling into the water. Food and
water containers were positioned so as to be easily accessible to the
animal on the platform.
The experimental protocol was performed over a 5-day period in
three groups of four rats: control (PSC), PS deprivation (PSD) and
PS rebound (PSR) groups. On the first day, a baseline recording was
performed for all animals. During the last 4 days of the experiment,
the animals from the PSC group remained on a bed of woodchips in
their container before being anaesthetized for histological perfusion
(at 17.00 h). In the PSD condition, the animals were placed on the
platform on the second day of the experiment (at 13.00 h) until they
were killed, at 16.30 h. On the two first days of deprivation, the
animals were removed from the platform for 30 min to permit
cleaning of their containers. In the PSR condition the animals were,
like the PSD animals, placed on the platform. After 72 h of PS
deprivation they were put back at 13.00 h into control conditions, i.e
in a container with a dry bed of woodchips to allow for recovery of
PS. After ?30 min of exploration and grooming, PSR animals fell
asleep. They were anaesthetized for perfusion 150 min after the
beginning of the first PS phase, which appeared with a latency of
37.5 ± 5.2 min.
Polygraphic recordings
Vigilance states were discriminated with the cortical EEG and neck
EMG. During W, desynchronized (or activated) low-amplitude EEG
was accompanied by a sustained EMG activity with phasic bursts.
SWS was clearly distinguished by high-voltage slow waves (1.5–
4.0 Hz) and spindles (10–14 Hz) and the disappearance of phasic
muscular activity in an immobile animal with closed eyes. A decrease
in the EEG amplitude associated with a flat EMG (i.e. muscle atonia)
signalled the onset of PS episodes further characterized by a
pronounced theta rhythm (5–9 Hz). For each vigilance state, a
spectral analysis of the EEG was performed on-line using the fast-
Fourier transform.
Histological and immunohistochemical procedures
The animals were perfused with a Ringer’s lactate solution containing
0.1% heparin, followed by 500 mL of a fixative composed of 4%
paraformaldehyde and 0.2% picric acid in 0.1 m phosphate buffer
(PB; pH 7.4). The brains were postfixed for 2 h in the same fixative at
4 ?C and then stored at 4 ?C for at least 2 days in 0.1 m PB with 30%
sucrose. They were rapidly frozen with CO2gas. Coronal sections
25-lm-thick were obtained with a cryostat and stored in 0.1 m PB,
pH 7.4, containing 0.9% NaCl, 0.3% Triton X-100 (PBST) and 0.1%
sodium azide (PBST-Az).
For double immunostaining experiments, free-floating sections were
successively incubated in: (i) a rabbit antiserum to Fos (1 : 5000;
Ab-5; Oncogene, CA, USA) in PBST-Az for 3 days at 4 ?C; (ii) a
biotinylated goat antirabbit IgG (1 : 2000; Vector Laboratories,
Burlingame, CA, USA) for 90 min at room temperature; and (iii) an
ABC-HRP solution (1 : 1000; Elite kit, Vector Laboratories) for
90 min at room temperature. The sections were immersed in a 0.05-m
Tris-HCl buffer (pH 7.6) containing 0.025% 3,3¢-diaminobenzidine-4
HCl (DAB; Sigma, Saint Quentin Fallavier, France), 0.003% H2O2
and 0.6% nickel ammonium sulphate for 10 min at room temperature.
The reaction was stopped by two rinses in PBST-Az. The free-floating
sections were then incubated in: (i) a goat antiserum to choline
acetyltransferase (ChAT; 1 : 2000; Chemicon, CA, USA) in PBST-Az
for 3 days at 4 ?C; (ii) a biotinylated donkey antigoat IgG (1 : 2000;
Vector Laboratories, CA, USA) for 90 min at room temperature; and
(iii) an ABC-HRP solution (1 : 1000; Elite kit, Vector Laboratories)
for 90 min at room temperature. Finally, the sections were immersed
for 15 min at room temperature in the same DAB solution as above
but without nickel. The sections were then mounted on gelatin-coated
slides, dried, dehydrated and coverslipped with DePex.
Controls in the absence of primary antibodies were routinely run to
ensure the absence of nonspecific single or dual immunostaining in the
material. As specified by the supplier, the Fos antiserum was made
against a synthetic peptide corresponding to the N-terminal part
Reticular neurons responsible for REM sleep2489
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 3

(residues 4–17) of human Fos. This part of the protein displays 100%
homology between human and rat (van Straaten et al., 1983; Curran
et al., 1987) and no homology with Fos-related antigens such as Fos B,
Jun B, Fra-1 and Fra-2 (Blast 2 sequences, NCBI). In agreement with
previousstudies,veryfewimmunoreactive(IR)nucleiwereobservedin
theratbrainwiththisFosantiserumfollowingperfusionduringdaylight
(Semba et al., 2001; Ro et al., 2003; Gong et al., 2004).
Cell counts and analysis of immunohistochemical data
Because the distribution of the Fos-IR and Fos–ChAT double-labelled
neuronswasidenticalonbothsidesofthebrain,onlyhemi-sectionswere
analysed. Drawings of double-immunostained sections were made with
a Axioscope microscope (Zeiss, Germany) equipped with a motorized
X–Y-sensitive stage and a video camera connected to a computerized
Fig. 1. Schematic distribution of ChAT-IR (small grey dots), Fos-IR (small black dots) and Fos–ChAT (large black dots) double-labelled neurons in 15 coronal
sections taken at 400-lm intervals through the full rostrocaudal extent of the pons and medulla in a representative animal for control (left hand side), PS-deprivation
(middle) and PS-rebound (right hand side) conditions. (A) Sections from )7.40 and )7.80 from Bregma. (B) Sections from )8.20 to )9.00 from Bregma.
(C) Sections from )9.40 to )10.20 from Bregma. (D) Sections from )10.60 to )11.80 from Bregma. (E) Sections from )12.20 to )13.00 from Bregma.
Abbreviations for the names of the structures are in the main list.
2490L. Verret et al.
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 4

image analysis system (Mercator; ExploraNova, La Rochelle, France).
Single- and double-labelled neurons were plotted on sections taken at
400-lm intervals through the full rostrocaudal extent of the pons and
medulla(15sectionsbetweenAP)7.40and)13.00fromBregma).The
three categories of neurons were counted per structure automatically
usingMercator.Whenastructurewaspresentonmore thanonesection,
Fig. 1. Continued
Reticular neurons responsible for REM sleep 2491
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 5

the sum of all the neurons counted on all the sections was made. The
countswerenotperformedinnucleiunlikelytoplayanyroleintheonset
and maintenance of PS, in particular the motor and sensory nuclei such
as the superior and inferior colliculus, the pontine nuclei, nuclei of the
lateral lemniscus, cochlear nuclei, cranial motor nuclei, sensory
trigeminal nuclei, cerebellum, superior and inferior olivary complexes
Fig. 1. Continued
2492L. Verret et al.
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 6

and vestibular nuclei. However, to provide qualitative information we
decided to display Fos-IR neurons localized in these structures in the
drawings of Fig. 1.
The atlas of Paxinos & Watson (1997) was used as a reference for
all structures except the medial PPTg (PPTgM), which was delineated
according to Maloney et al. (1999), the SLD which was named and
Fig. 1. Continued
Reticular neurons responsible for REM sleep2493
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 7

delineated according to Swanson (1998) and the lateral paragiganto-
cellular (LPGi) and rostroventrolateral reticular nuclei, which were
grouped under the name of LPGi.
Statistical analysis
anova tests were performed on the different vigilance states across
experimental conditions. A post hoc PLSD Fisher test was used to
identify significant pairwise differences. The same tests were
performed to compare cell counts among the animals groups in the
different brainstem structures. Cell counts were correlated with
physiological variables using a simple linear regression analysis. All
statistics were performed using Statview 5.0.
Results
Quantification of the sleep and waking phases
The PS deprivation procedure was effective in producing a near
complete elimination of PS for ?75 h in the PSD and 72 h in the PSR
groups. During the 150 min prior to killing, the quantities of PS
differed significantly between the group of control animals (PSC) and
the PSD and PSR groups (all P < 0.05). The state of PS constituted
16% of the last 150 min prior to killing in the PSC group, 0% in the
PSD group and 49% in the PSR group (Table 1). The PSD animals
spent 30% and the PSR 31% of their time in SWS, significantly less
than those from the PSC group (48%; P < 0.05). The PSD group
presented a large quantity of W (70%) compared to the PSC (36%)
and PSR (20%) groups. The increase in PS quantities during the last
Fig. 1. Continued
2494 L. Verret et al.
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 8

150 min prior to killing in the PSR group compared to the PSC group
was due to a significant increase in the duration of the PS bouts
(Table 1; P < 0.01). The number of bouts did not vary significantly
between the two groups (Table 1).
Distribution of the Fos-immunoreactive neurons
As expected, the Fos immunolabelling was restricted to the nucleus of
the labelled cells. The diameter of the labelled nuclei was of 8–15 lm.
The distribution of Fos-IR neurons in the brainstem greatly differed
between the three experimental groups of rats (Table 2 and Fig. 1). In
the PSC group, only a few to a moderate number of cells was found in
all brainstem structures while in the PSD and PSR groups large
numbers of neurons were seen with a globally different distribution
(Table 2 and Fig. 1). In the PSD group, no significant Fos labelling was
seen in brain areas activated in response to acute stress (Cullinan et al.,
1995), areas such as the dorsomedial nucleus of the hypothalamus, the
lateral portion of the retrochiasmatic area (not illustrated) and the locus
coeruleus (LC). Altogether, these results confirm previous ones
showing that, in the brainstem, very few neurons express Fos during
the rest phase when the animals are maintained in their familiar
environment (Herdegen et al., 1995; Semba et al., 2001; Ro et al.,
2003; Gong et al., 2004). They further indicate that the Fos
immunostaining in the PSD group is not due to stress.
The number of Fos-IR neurons was quantified in 27 brainstem
structures including all structures from the reticular formation, the
periaqueductal grey and the cholinergic and monoaminergic nuclei
(Tables 2 and 3 and Fig. 1). Correlations between the number of
Fos-IR and Fos–ChAT double-labelled neurons and the amount of the
three vigilance states were calculated for all structures (Table 4). To
simplify the text, only the significant correlations are mentioned.
Three structures displayed a significantly higher number of Fos-IR
neurons in the PSD condition than in the PSR and PSC conditions.
These were the dorsal raphe (DRN), lateral parabrachial (PBL) and
raphe pallidus (RPa) nuclei (Table 2 and Figs 2C and D, and 4G
and H). In the DRN, the number of Fos-IR cells was significantly and
positively correlated with the percentage of time spent in W during
the final 150-min recording period (R ¼ 0.65, P < 0.05). In the PBL,
the Fos-IR neurons were spread over all subdivisions in the PSD
condition whereas many of them were concentrated in the central
portion of the central lateral subnucleus (Bester et al., 1997) in the
PSR condition (Figs 1C, and 2C and D). The number of Fos-labelled
cells was significantly higher in the PSD condition than the PSC
condition in the deep mesencephalic reticular nucleus (DPMe). The
nucleus of the solitary tract (Sol) displayed a higher number of Fos-IR
neurons in the PSD condition, but this did not reach significance
because of large variability among the animals (Fig. 1E).
In 12 brainstem structures, the number of Fos-IR neurons was
significantly higher in the PSR condition than in the PSD condition
and moreover the PSC condition (Table 2). A very high level of
significance (P < 0.001) between PSR and PSD conditions and a
positive correlation with the percentage of time spent in PS during the
final 150-min recording period was reached in three nuclei, namely the
SLD (R ¼ 0.83, P < 0.001), dorsal paragigantocellular nucleus
(DPGi; R ¼ 0.84, P < 0.001) and GiA (R ¼ 0.65, P < 0.05; Table 4).
This is due to a substantial number of Fos-IR neurons labelled
specifically in these three nuclei after PS rebound (Figs 1C–E, 2E and
F, and 4A and B) and to moderate interindividual variability (Table 2).
Interestingly, the number of Fos-IR cells in the DPGi was also
negatively correlated with the percentage of time spent in W
Table 1. Time spent in waking, slow-wave sleep and paradoxical sleep
PSCPSD PSR
Vigilance states
W (%)
SWS (%)
PS (%)
35.6 ± 2.4
48.0 ± 2.6
16.4 ± 0.3
69.6 ± 6.8**
30.4 ± 6.8*
0 ± 0*
20.4 ± 3.9*,###
30.6 ± 4.8*
49.1 ± 6.4**,###
PS bouts
Number (n)
Duration (min)
21.3 ± 2.3
1.2 ± 0.1
–
–
23.8 ± 5.3
2.3 ± 0.3**
Percentage of time spent in waking (W), slow wave sleep (SWS) and para-
doxical sleep (PS) scored per 10-s epoch over the last 150 min before killing,
and number and duration of PS bouts during the same period for Control (PSC,
n ¼ 4), PS-deprived (PSD, n ¼ 4) and PS-recovery (PSR, n ¼ 4) rats. Signi-
ficance valuesindicated forindividual
**P < 0.01 vs. PSC,###P < 0.001 between PSR and PSD.
pointsare*P < 0.05 and
Table 2. Numbers of Fos-IR neurones in the brainstem
n
PSC PSDPSR
Periaqueductal grey
dPAG
lPAG
vlPAG
CGPn
4
4
5
3
8.0 ± 5.4
9.3 ± 4.8
15.5 ± 5.7
4.3 ± 2.3
29.8 ± 5.6
50.0 ± 9.6*
68.5 ± 3.9**
21.8 ± 14.7
144.5 ± 56.2*,#
47.8 ± 11.5*
85.3 ± 16.0***
35.5 ± 13.3
Rostral raphe
DRN
MnR
4
4
4.0 ± 2.1
0.3 ± 0.3
31.0 ± 7.9**
3.3 ± 1.3
7.5 ± 4.7#
15.3 ± 4.2**,##
Pontomesencephalic tegmenti
PPTg
PPTgM
PPTg + PPTgM
LDTg
LDTgV
LDTg + LDTgV
5
3
3.0 ± 2.4
0 ± 0
3.3 ± 2.3
13.5 ± 7.0
1.8 ± 0.5
15.3 ± 7.1
20.0 ± 3.9
0 ± 0
23.3 ± 4.1
22.5 ± 2.6
11.0 ± 5.4
33.5 ± 6.0
24.3 ± 9.1*
11.8 ± 3.4**,#
37.0 ± 12.3*,#
63.0 ± 15.6**,#
23.3 ± 6.0**
86.3 ± 21.0**,#
5
3
Mesopontine reticular formation
DPMe
SLD
PnO
PnC
4
3
6
3
3.8 ± 1.0
0.5 ± 0.5
6.8 ± 1.4
6.5 ± 1.9
28.0 ± 7.2*
6.0 ± 1.4
34.3 ± 16.7
18.0 ± 4.1
59.5 ± 17.8**
46.8 ± 9.9***,###
66.3 ± 26.7*
64.3 ± 18.3**,#
Dorsolateral pons
PBL
LC
KF
PCRt
Sol
1
3
1
6
5
6.8 ± 4.2
0.3 ± 0.3
2.0 ± 0.9
7.8 ± 3.6
8.3 ± 4.3
73.8 ± 1.7***
1.3 ± 0.8
6.5 ± 2.0
51.0 ± 19.8
51.5 ± 28.7
42.5 ± 8.3**,##
0.8 ± 0.5
4.0 ± 2.4
63.8 ± 29.4
13.0 ± 6.8
Medullary reticular formation
Gi
GiA
GiV
DPGi
LPGi
6
4
3
6
6
3.3 ± 1.3
0.5 ± 0.3
0.3 ± 0.3
3.3 ± 1.1
6.0 ± 1.9
14.5 ± 9.9
10.3 ± 3.7
4.5 ± 1.9
1.8 ± 0.6
52.3 ± 8.1*
33.8 ± 10.7*
31.8 ± 4.1***,###
35.3 ± 9.9**,##
49.0 ± 10.9***,###
99.5 ± 23.3**,#
Caudal raphe
RMg
ROb
RPa
5
5
5
0.5 ± 0.5
0.3 ± 0.3
0.3 ± 0.3
1.5 ± 0.5
3.8 ± 1.1
27.0 ± 7.0**
6.5 ± 2.3*,#
27.3 ± 6.0***,##
7.8 ± 2.8#
Number (mean ± SEM) of Fos-IR neurons calculated in the brainstem for
control (PSC, n ¼ 4), PS-deprived (PSD, n ¼ 4) and PS-recovery (PSR,
n ¼ 4) rats. Cells were counted from a total of 15 sections at 400-lm intervals
through the full rostrocaudal extent of the pons and medulla. The values dis-
played are the average across four animals in each group of the sum of Fos-
labelled neurons counted on one or several sections (column n) depending on
the rostrocaudal extent of the structures. Significance values indicated for
individual points are: *P < 0.05, **P < 0.01 and ***P < 0.001 vs. PSC;
#P < 0.05,
##P < 0.01 and###P < 0.001 between PSR and PSD.
Reticular neurons responsible for REM sleep2495
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 9

(R ¼ –0.61, P < 0.05). A high level of significance (P < 0.01) and a
strong positive correlation with the percentage of time spent in PS
during the final 150-min recording period was seen in three other
nuclei, namely the median raphe (MnR; R ¼ 0.75, P < 0.01), ventral
gigantocellular (GiV; R ¼ 0.71, P < 0.01) and raphe obscurus (ROb;
R ¼ 0.76, P < 0.01) nuclei (Table 4). The MnR contained a moderate
number of Fos-IR neurons in the PSR condition (Fig. 1B), whereas the
GiV and ROb showed a substantial number of Fos-positive neurons
(Fig. 1E). In the GiV, a conspicuous group of neurons showing an
intensely labelled nucleus was systematically seen among the exiting
fibres of the hypoglossal nerve (Fig. 4C and D).
A low level of significance (P < 0.05) was obtained in six
additional areas, which are the dorsal periaqueductal grey (dPAG),
pedunculopontine tegmental (PPTg together with PPTgM), laterodor-
sal tegmental (LDTg and LDTgV, the ventral part of the LDTg),
caudal pontine reticular (PnC), raphe magnus (RMg) and lateral
paragigantocellular nuclei (LPGi) (Table 2). The dPAG and LPGi
displayed the highest number of Fos-IR neurons among all the
structures quantified while the RMg contained the lowest number. In
between were the LDTg (Figs 1B and C, and 3) and the PnC (Fig. 1C
and D), which contained a substantial number of Fos-IR neurons
(Table 2). Across conditions, the numbers of Fos-IR cells were
significantly positively correlated with the percentage of time spent in
PS during the final 150-min recording period in the PPTgM
(R ¼ 0.71,
(R ¼ 0.63, P < 0.05; Table 4).
It is worth noting that in four additional regions the number of Fos-
IR neurons was significantly higher in the PSR condition than in the
PSC condition but only tended to be higher than the PSD condition.
These were the DPMe, oral pontine (PnO) and gigantocellular reticular
nuclei and the LDTgV, which all contained substantial numbers of
Fos-IR neurons in the PSR condition (Table 2, Fig. 1).
In two structures, the number of Fos-IR neurons did not differ
between the PSD and PSR conditions but was much higher than in the
PSC condition. These structures are the lateral and ventrolateral
periaqueductal grey (lPAG and vlPAG; Fig. 2A and B). In two areas,
more Fos-IR cells were seen in the PSR and PSD conditions than in
the PSC condition but this did not reach statistical significance. These
were the central grey of the pons and the parvicellular reticular nucleus
P < 0.01), PnC(R ¼ 0.65,
P < 0.05)andLDTg
Table 3. Double-labelled Fos–ChAT neurones in the brainstem: numbers and
as percentages of singly ChAT- and Fos-labelled neurones
PSCPSD PSR
PPTg + PPTgM
Fos–ChAT
ChAT (%)
Fos (%)
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
2.3 ± 1.7
0.9 ± 0.6
3.8 ± 2.7
LDTg + LDTgV
Fos–ChAT
ChAT (%)
Fos (%)
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
3.8 ± 1.8*,#
2.3 ± 1.3
3.5 ± 1.4
PCRt
Fos–ChAT
ChAT (%)
Fos (%)
0 ± 0
0 ± 0
0 ± 0
1.5 ± 0.9
6.3 ± 3.7
6.3 ± 4.3
0.5 ± 0.3
1.0 ± 0.6
2.8 ± 1.7
Gi
Fos–ChAT
ChAT (%)
Fos (%)
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0.8 ± 0.3**,##
5.5 ± 2.1
3.7 ± 2.2
DPGi
Fos–ChAT
ChAT (%)
Fos (%)
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
3.8 ± 1.4**,##
55.1 ± 5.8
12.5 ± 8.2
LPGi
Fos–ChAT
ChAT (%)
Fos (%)
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
8.0 ± 1.7***,###
23.8 ± 13.6
8.3 ± 1.7
ROb
Fos–ChAT
ChAT (%)
Fos (%)
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
0 ± 0
11.5 ± 3.1**,##
66.8 ± 11.6
42.7 ± 7.8
Number (mean ± SEM) of double-labelled neurons (Fos–ChAT) and mean
percentage (± SEM) of Fos–ChAT neurons vs. singly ChAT-IR (%ChAT)
and Fos-IR (%Fos)cellsforcontrol
(PSD, n ¼ 4)
ing, see Table 2. *P < 0.05, **P < 0.01 and ***P < 0.001 vs. PSC;
#P < 0.05,
(PSC,
n ¼ 4)
n ¼ 4),
rats.
PS-deprivation
Forcell andPS-rebound (PSR,count-
##P < 0.01 and###P < 0.001 between PSR and PSD.
Table 4. Relationship between the number of Fos-IR and Fos–ChAT cell
counts and the states of sleep
Regression coefficient (R)
Fos-IRFos–ChAT
WSWS PSW SWSPS
Periaqueductal grey
dPAG
lPAG
vlPAG
CGPn
–0.02
0.20
–0.04
–0.15
–0.04
–0.50
–0.38
–0.20
0.07
0.05
0.23
0.27
–
–
–
–
–
–
–
–
–
–
–
–
Rostral raphe
DRN
MnR
0.64*
–0.49
–0.23
–0.45
–0.54
0.75**
–
–
–
–
–
–
Pontomesencephalic tegmenti
PPTg
PPTgM
PPTg + PPTgM
LDTg
LDTgV
LDTg + LDTgV
0.09
–0.38
–0.06
–0.43
–0.34
–0.42
–0.56
–0.57
–0.59
–0.34
–0.36
–0.36
0.22
0.71**
0.40
0.63*
0.55
0.63*
–0.37
–0.14
–0.38
–0.41
–0.37
–0.40
–0.12
–0.22
–0.15
–0.31
–0.38
–0.34
0.44
0.27
0.47
0.59
0.59
0.61*
Mesopontine reticular formation
DPMe
SLD
PnO
PnC
–0.15
–0.54
–0.17
–0.38
–0.59
–0.48
–0.35
–0.46
0.51
0.83***
0.37
0.65*
–
–
–
–
–
–
–
–
–
–
–
–
Dorsolateral pons
PBL
LC
KF
0.56
0.21
0.48
–0.59
–0.34
–0.51
–0.25
–0.02
–0.22
–
–
–
–
–
–
–
–
–
Medullary reticular formation
PCRt
Sol
Gi
GiA
GiV
DPGi
LPGi
0.02
0.30
–0.32
–0.44
–0.45
–0.84*
–0.22
–0.17
0.12
–0.24
–0.35
–0.45
–0.37
–0.47
–0.11
–0.38
0.47
0.65?
0.71**
0.84***
0.49
0.49
–
–0.50
–
–
–0.55
–0.56
–0.50
–
–0.10
–
–
–0.03
–0.04
–0.23
–
0.58
–
–
0.60*
0.88****
Caudal raphe
RMg
ROb
RPa
–0.34
–0.57
0.50
–0.42
–0.30
–0.17
0.58
0.76**
–0.42
–––
0.83***
–
–0.59
–
–0.26
–
Relationship between the number of Fos-IR and Fos–ChAT cell counts
and states of sleep as assessed by a simple linear regression (testing
for W, SWS and PS amounts). The sign + or – of the regression
oefficient (R) indicates the direction of the covariation. *P < 0.05;
**P < 0.01; ***P < 0.001; ****P < 0.0001.
?P < 0.02;
2496L. Verret et al.
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 10

of the medulla oblongata (Table 2 and Fig. 1). Finally, two nuclei, the
Ko ¨lliker–Fuse and the LC, showed a small or very small number of
Fos-IR neurons in all conditions (Table 2).
Distribution of the ChAT-immunoreactive neurons
expressing Fos
The distribution of ChAT-IR neurons, as observed in our experiment,
was similar to that described previously (Tatehata et al., 1987; Jones,
1990; Ruggiero et al., 1990). Apart from the well-known mesopontine
group and cranial motor nuclei, ChAT-IR cells were visible in several
medullary nuclei, namely the periolivary nuclei, medial vestibular
nucleus, reticular formation close to the ventromedial solitary tract
nucleus, DPGi, LPGi and ROb, and in a band of neurons which
stretched dorsoventrally across both the parvicellular reticular nucleus
(PCRt) and the lateral part of the gigantocellular nucleus (Fig. 1C–E).
In the LPGi, a medial group of relatively small and moderately ChAT-
IR neurons was clearly differentiated from the larger and more IR cells
belonging to the external nucleus ambiguus (Fig. 4E and F). Neurons
Fig. 2. Photomicrographs of Fos (nuclear staining) and ChAT (diffuse cytoplasmic staining) double-immunostained sections at the pontine level, from PSD (left
hand side) and PSR (right hand side) rats. (A and B) Photomicrographs showing the vlPAG in (A) PSD and (B) PSR animals. Numerous Fos-labelled neurons are
observed in this structure in both conditions. Aq, Sylvius aqueduct. (C and D) Photomicrographs showing the PBL in (C) PSD and (D) PSR animals. In the PSD
condition, Fos-IR cells are spread over all the PBL while they form a compact group in the central part of the nucleus in the PSR condition. (E and F)
Photomicrographs showing the SLD in (E) PSD and (F) PSR rats. Note the presence of a large number of Fos-labelled cells specifically in the PSR rat. These cells
are not immunoreactive to ChAT and are therefore not cholinergic. Scale bars, 100 lm.
Reticular neurons responsible for REM sleep2497
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 11

Fig. 3.
Fig. 4.
2498L. Verret et al.
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 12

stretching vertically along the midline and approaching the fourth
ventricle (Fig. 4A and B) were considered as belonging to ROb.
Double-labelled Fos–ChAT neurons were almost exclusively ob-
served in the PSR group (Table 3 and Fig. 1). The only region showing
double-labelled neurons in both PSD and PSR conditions was the PCRt
in which very few Fos–ChATcells were observed (Table 3). In the PSR
group, occasional cholinergic motoneurons were double-labelled in the
oculomotor nuclei. The periolivary nuclei and the medial vestibular
nucleus also contained a few double-labelled cells (Fig. 1C–E).
Surprisingly, the mesopontine group (PPTg, PPTgM, LDTg and
LDTgV) contained a very small number of double-labelled cells
(Fig. 3). These neurons represented a small percentage (3–5%) of the
singly Fos-IR cells and an even lower percentage (0.8–4%) of the
singly ChAT-IR cells (Table 3). In contrast, medullary structures such
as the ROb, the LPGi and to a lesser extent the DPGi contained a
substantial number of double-labelled cells (Table 3 and Figs 1 and
4B). In the DPGi, double-labelled cells displayed a preferential dorsal
location, near the ependymal layer (Figs 1D and E, and 4B). In the
LPGi, their distribution was restricted to the medial half of the nucleus
(Figs 1E and 4F). In the ROb, 43% of the singly Fos-IR and 67% of
the singly ChAT-IR neurons were double-labelled (Table 3). In the
DPGi and LPGi, the double-labelled neurons corresponded to 12 and
8% of the singly Fos-IR neurons and to 55 and 24% of the singly
ChAT-IR neurons, respectively.
Across experimental conditions, the number of double-labelled cells
was significantly positively correlated with the percentage of time
spent in PS during the final 150-min recording period in the LPGi
(R ¼ 0.88, P < 0.001), ROb (R ¼ 0.83, P < 0.001), LDTg + LDTgV
as a whole (R ¼ 0.61, P < 0.05) and DPGi (R ¼ 0.59, P < 0.05;
Table 4).
Discussion
In the present study, by comparing Fos staining in control rats, rats
subjected to a selective PS deprivation and rats allowed to recover from
such deprivation, we identified at the cellular level populations of
brainstem neurons with an activity positively or negatively correlated
with PS. We first confirmed that the SLD, LDTg and GiV, three
structures previously identified by lesion, pharmacological and elec-
trophysiological studies as being essential components of the network
responsible for PS genesis, contained a large number of Fos-labelled
neurons following PS hypersomnia. In addition, we further reveal that
the dPAG, the DPGi and the LPGi, structures unknown to be active
during PS, contained a large number of Fos-labelled cells following PS
hypersomnia. In contrast, structures such as the DRN, the PBL and the
RPa contained Fos-labelled neurons specifically after PS deprivation,
indicating that these neurons might be very active in the absence of PS.
Finally, we found that cholinergic mesopontine neurons were rarely
immunoreactive to Fos following PS hypersomnia, suggesting that the
vast majority of these neurons are not activated during PS. In contrast,
numerous cholinergic neurons immunoreactive to Fos were seen in
several medullary structures such as the LPGi and ROb following PS
hypersomnia, indicating that these medullary cholinergic neurons
might play an important role during PS. In the following, we first
discuss the methodological issues of our work before confronting our
results with the literature and discussing their functional significance.
Methodological considerations
Our study is based on the assumption that activated cells display Fos
immunoreactivity after a delay of ?1 h with a peak 1–3 h after the
stimulation followed by a progressive decrease and a disappearance of
staining after 4 h (Dragunow & Faull, 1989). We choose the Fos
method to identify the cells activated during PS because, despite some
drawbacks, it still constitutes the best method available to identify at a
cellular level activated neurons. Indeed, although neurons with low or
phasic discharge rates do not always express Fos or, conversely, Fos
labelling reflects in some neurons changes in calcium-mediated
cellular processes other than increases in neuronal discharge (Morgan
& Curran, 1986; Dragunow & Faull, 1989), it is still accepted that
most of the activated neurons express Fos.
To maximize our chance of obtaining Fos immunostaining reflecting
activity positively or negatively correlated with PS, we submitted the
rats for a long period of time (72 h) to a very effective and selective
PS-deprivation method (the flowerpot method). During the deprivation,
the animals presented 69% of W, 30% of SWS and 0% of PS. We then
allowed a group of these deprived animals to recover for 3 h, a period
during which they presented 49% of PS, 30% of SWS and 20% of W.
By comparison, control rats exhibited 16% of PS, 48% of SWS and
35% of W. Because none of these three groups of animals presented
only one vigilance state during the 3 h preceding perfusion, one might
conclude that the Fos labelling observed is not specific to any of the
states of vigilance. However, several points strongly suggest that the
Fos-labelled cells after PS recovery correspond to neurons active
during PS. First, we have demonstrated that the Fos labelling obtained
in the PSR group is positively correlated with the amount of PS for
numerous structures but not with that of SWS and W. Second,
numerous Fos-labelled cells were observed in three structures already
known to contain PS-active cells, namely the LDTg, SLD and GiV,
while a few or no cells were observed in W- or SWS-active structures
such as the LC (present results), the hypocretin-containing cells (Verret
et al., 2003b) or the ventrolateral preoptic area (Verret et al., 2003a).
Third, the distribution of labelled cells was significantly different
between the PSD and PSR groups in the large majority of brainstem
structures. Altogether, these points strongly suggest that the Fos-
labelled cells in the PSR group correspond to neurons active during PS.
It is important to stress, however, that it is not possible with the Fos
method to determine whether or not the labelled neurons are active
Fig. 4. Photomicrographs of medullary sections double-immunostained with Fos (black nuclear staining) and ChAT (brown cytoplasmic staining) of a PSD (left
hand side) and a PSR (right hand side) rat. (A and B) Photomicrographs showing the DPGi and ROb in (A) a PSD and (B) a PSR rat. Note the presence of Fos-IR
and Fos–ChAT double-labelled cells (arrows) in these two nuclei specifically in the PSR rat. 4V, fourth ventricle; mlf, medial longitudinal fasciculus. (C and D)
Photomicrographs showing the GiV in (C) a PSD and (D) a PSR rat. A cluster of Fos-IR neurons is localized in the GiV region just dorsal to the inferior olivary
complex in between the fibres of the hypoglossal nerve (12n) specifically in the PSR condition. (E and F) Photomicrographs showing the LPGi in (E) a PSD and
(F) a PSR rat. Note that more Fos-IR neurons are visible in the LPGi in the PSR rat than in the PSD rat. Two double-immunostained neurons (designated by arrows
and enlarged) are visible in the medial part of the LPGi specifically in the PSR condition. (G and H) Photomicrographs showing the RPa in (G) a PSD and
(H) a PSR rat. Note the presence of numerous Fos-IR cells in the RPa specifically in the PSD rat. Scale bars, 100 lm.
Fig. 3. Photomicrographs showing the LDTg on three sections double-immunostained with Fos (black nuclear staining) and ChAT (brown cytoplasmic staining) in
(A) PSC, (B) PSD and (C) PSR rats. A large number of singly Fos-IR neurons intermingled with cholinergic neurons are observed in the PSR condition, whereas a
few to occasional Fos-labelled cells are present in PSD and PSC conditions, respectively. Only one double-immunostained neuron is visible in the PSR rat (arrow).
Aq, Sylvius aqueduct. Scale bars, 100 lm.
Reticular neurons responsible for REM sleep 2499
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 13

selectively during PS. Further, they could be responsible for PS onset
and maintenance or could be involved in physiological changes taking
place during this state.
By contrast, most of the Fos labelling obtained in the PSD group
does not seem to be correlated with one of the vigilance states or with
the chronic stress induced by the procedure. Indeed, excepting for the
DRN, neurons previously shown to be Fos-immunoreactive after Wor
stress (Pompeiano et al., 1992; Tononi et al., 1994; Cullinan et al.,
1995) were not labelled in our PSD group. These results suggest that
W-active and stress-related neurons are not active during PS
deprivation although we cannot completely exclude the possibility
that such neurons are active but not labelled with Fos due to
autoregulation of Fos expression known to occur in the case of
exposure to chronic stimuli (Sassone-Corsi et al., 1988; Schonthal
et al., 1989; Herdegen et al., 1995; Herdegen & Leah, 1998). The
most likely explanation is therefore that the Fos-positive neurons in
the PSD group correspond to neurons increasing their activity during
the deprivation to avoid the onset of PS. These neurons can be either
directly responsible for the inhibition of the onset of PS via inhibition
of PS-executive neurons or indirectly involved in such inhibition, for
example by increasing the postural tonus of the animals to prevent
them falling into the water.
In summary, it seems likely that the Fos-labelled neurons in our PSR
group correspond to neurons active during PS while those labelled in
our PSD group correspond to neurons inhibiting the onset of PS.
Functional significance of the Fos and Fos–ChAT labellings
visualized
Structures known to be implicated in PS onset and maintenance
containing Fos-labelled cells specifically after PS recovery
Six pontine and two medullary structures previously hypothesized to
be implicated in PS onset and maintenance, namely the LDTg,
LDTgV, PPTg, PPTgM, SLD, PnC, GiA and GiV contained a
substantial to large number of Fos-labelled neurons in the PSR group.
The number of neurons was highly significantly increased in the PSR
group compared to the two other conditions in the SLD, GiA and GiV.
In the cholinergic nuclei, only a few of the Fos-labelled neurons were
immunoreactive to ChAT. In addition, although the number of Fos-
labelled neurons was high in the PnC it must be stressed that the
labelled neurons were in fact dispersed in this large structure,
suggesting that only a small percentage of the neurons from this
nucleus are active during PS.
Our results are only in partial agreement with the two previous
similar studies. Fos-labelled neurons were indeed also observed in the
PPTg, LDTg and SLD but not in the PnC, GiA and GiV by Merchant-
Nancy et al. (1995) after 60–90 min recovery from a 48-h PS
deprivation performed with the flowerpot technique. At variance with
these and our results, Maloney et al. (1999) observed with the same
protocol as ours excepting a shorter PS deprivation (48 h in place of
72 h) significantly more Fos-labelled neurons only in the LDTg and
the PnO in the PSR group compared to the PSC and PSD groups. They
further showed significantly more ChAT–Fos double-labelled cells in
the LDTg and PPTg in the PSR group than the PSC and PSD groups.
The differences between these and our study could be due to the
shorter duration of the PS deprivation used and therefore the smaller
amount of PS in their recovery group compared to our study
(Merchant-Nancy et al., 19% of PS; Maloney et al., 28%; our study,
49%). It could also be due to a lower specificity and sensitivity of the
Fos antibodies previously used. As stated by Merchant-Nancy et al.
(1995), their Fos antibody indeed also recognizes Fos-related antigens
well known to be strongly expressed in basal daylight condition in
contrast to Fos which is not expressed in significant amounts in that
condition. Such cross-reactivity could also explain why Maloney et al.
(1999) did not observe an increase in the number of Fos-labelled cells
in their PSR groups in the SLD, GiVand GiA. Indeed, they reported a
large number of Fos-labelled cells in their control group in all counted
brainstem structures while we found only a few or a small number of
such cells in those structures in our control group (for example they
counted 66 Fos-labelled cells in the GiV per section in their PSC
group whereas we found less than one cell per section in the same
condition). Such high numbers in their PSC group probably prevented
them from reaching statistically significant differences between
experimental groups for several structures.
In addition, our findings that the SLD, GiA and GiV contained Fos-
positive cells specifically after PS recovery are strongly supported by
our recent results in rats and previous ones in cats. We indeed recently
showed that desinhibition or excitation of the SLD by iontophoretic
administration of GABAAantagonists (bicuculline or gabazine) or
kainic acid, a glutamate agonist, induces a state resembling PS
(Boissard et al., 2002). In contrast, administration of the same
compounds in the adjacent PnO and PnC produced active waking
(Boissard et al., 2002). These and our present results altogether highly
suggest that the SLD contains a large number of neurons specifically
active during PS, playing a key role in its onset and maintenance,
whereas the PnO and PnC are structures primarily involved in other
functions with a small subset of neurons playing a role in the state of
PS.
In agreement with previous studies in cats (Sakai, 1988; Yamuy
et al., 1993; Morales et al., 1999), we also recently demonstrated, with
a combination of Fos staining after 90 min of a state of PS induced by
bicuculline injection in the SLD and anterograde PHA-L tracing, that
the SLD provides an excitatory projection to the GiA and GiV
(Boissard et al., 2002). Indeed, PHA-L anterogradely labelled fibres
coming from the SLD were seen around numerous Fos-labelled cell
bodies in the GiA and GiV. The location of the Fos-labelled cells in
that previous study was similar to that found here in the PSR group,
confirming that these two structures contain numerous neurons active
during PS. Our and other previous studies (Holstege & Bongers, 1991;
Fort et al., 1993; Rampon et al., 1996; Boissard et al., 2002) further
suggest that these neurons are glycinergic and are responsible for the
hyperpolarization of motoneurons during PS, leading to the muscle
atonia occurring during this state.
We observed only a small number of Fos–ChAT double-labelled
neurons in the mesopontine cholinergic nuclei (LDTg, LDTgV, PPTg
and PPTgM) in the PSR group and none in the PSC and PSD groups.
Of great interest, the LDTg contained a large number of ChAT-
negative Fos-labelled neurons and the other nuclei a substantial
number of such neurons specifically in the PSR group. Maloney et al.
(1999) found a large number of Fos and a substantial number of Fos–
ChAT double-labelled cells in these structures in all experimental
groups, with significantly more Fos-labelled neurons in the PSR group
than the PSD group only in the LDTg. They also found slightly but
significantly more Fos–ChAT neurons in the LDTg and the PPTgM in
the PSR group than the PSD group. In two additional studies in cats
examining immunostaining for Fos and ChAT following a strong
increase in PS quantities induced by carbachol injection into the dorsal
pontine tegmentum, either a small proportion of the cholinergic cells
were Fos-positive (Shiromani et al., 1996) or no ChAT cells were
double-labelled (Yamuy et al., 1998). In the latter study, a large
number of ChAT-negative Fos-labelled cells were, however, observed
in the LDTg and PPTg. The authors further demonstrated that a minor
2500 L. Verret et al.
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 14

proportion of these neurons were immunoreactive to GAD and
therefore GABAergic. These results are in agreement with those of
Maloney et al. (1999) showing a significantly higher number of GAD–
Fos-labelled neurons within the LDTg, LDTgV, PPTgM and PPTg in
the PSR group than in the PSD and PSC groups. These and our results
indicate that the LDTg and to a minor extent the other cholinergic
nuclei contain a substantial to large number of noncholinergic Fos-
labelled cells specifically after PS recovery, a minor proportion of
which are GABAergic and the majority of which use another
neurotransmitter. These results further importantly suggest that the
neurons active during PS recorded previously in these nuclei (El
Mansari et al., 1989; Kayama et al., 1992) are in the vast majority
noncholinergic. The present results are at variance with the hypothesis
that cholinergic mesopontine neurons play a crucial role in PS onset
and maintenance. This hypothesis was based on results obtained in
cats showing that direct administration of cholinomimetics into the
pontine reticular formation causes a dose-dependent enhancement of
PS (review in Baghdoyan, 1997). Further, it has been shown in the
same species that neurotoxic lesion of the LDTg–PPTg decreases the
amount of PS (Jones & Webster, 1988) and that acetylcholine release
is significantly increased in the pontine reticular formation over Wand
SWS (Kodama et al., 1990; Leonard & Lydic, 1995). However, direct
administration of cholinomimetics in the rat pontine reticular
formation was recently shown to have either no effect on PS or to
induce only a small increase in PS quantities (review in Boissard et al.,
2002). These and our present results suggest that mesopontine
cholinergic neurons might not play a crucial role in PS onset and
maintenance in rats, in contrast to cats.
Structures unknown to be implicated in PS containing
Fos-labelled cells specifically after PS recovery
Several structures previously not identified as containing neurons
active during PS contained a substantial to large number of Fos-
labelled neurons specifically in the PSR group. These structures were
the dPAG, the LPGi, the DPGi, the MnR and the ROb. A small to a
large proportion of the Fos-labelled neurons from some of the
medullary structures were immunoreactive to ChAT.
To our knowledge, only the DPGi has been implicated in the control
of PS. Indeed, we and others recently hypothesized that it contains the
GABAergic neurons responsible for the cessation of activity of the
noradrenergic LC neurons during PS (Luppi et al., 1999; Kaur et al.,
2001). Supporting this claim, the DPGi constitutes a major GABA-
ergic afferent to the LC (Aston-Jones et al., 1986; Ennis & Aston-
Jones, 1989; Luppi et al., 1995; Peyron et al., 1995), and application
of bicuculline to noradrenergic LC neurons during PS restores their
activity, indicating that GABA is responsible for their inactivation
during this state (Gervasoni et al., 1998). In addition, it has been
shown that bilateral stimulation of the DPGi significantly increased PS
quantities, an effect blocked by picrotoxin application into the LC
(Kaur et al., 2001). Further, we found that the DPGi contained a large
number of cells labelled both with Fos and retrogradely in animals
that received an injection of cholera toxin b subunit (CTb) in the LC
and then were subjected to 72 h of PS deprivation and 3 h of PS
recovery (Verret et al., 2003a). In these experiments, a smaller but
significant number of CTb–Fos double-labelled cells was also
observed in the LPGi, indicating that it could also contain a subset
of the GABAergic neurons responsible for the inactivation of
noradrenergic LC neurons during PS. Supporting this claim, neurons
with an activity specific to PS have been recorded in the cat LPGi
(Sakai, 1988). It is nevertheless unlikely that all Fos-labelled neurons
found specifically in the LPGi in the PSR group play such role. In
addition, because the LPGi also contained significantly more Fos-
labelled cells in the PSD than in the PSC group, some of the Fos-
labelled neurons in the PSR group could be active during both PS
deprivation and PS recovery. Studies of the afferents and efferents of
the LPGi Fos-labelled neurons after PS deprivation and recovery are
necessary to determine their role in particular with regards to the
sympathetic nervous system because the LPGi is known to play a
crucial role in its control (review in Sun, 1995).
The ROb contained a substantial number of Fos-labelled neurons
specifically in the PSR group. This nucleus is known to contain
serotonergic neurons innervating the spinal cord (Arvidsson et al.,
1992). However, it has been shown that these neurons are not active
during PS (Sakai et al., 1983; Jacobs et al., 2002). Nevertheless,
neurons specifically active during PS have also been recorded in the
cat ROb (Sakai, 1988). In the present study, we found that half of the
Fos-labelled neurons in the PSR group were immunoreactive to ChAT
and therefore cholinergic. We also found that a smaller subset
(? 10%) of the Fos-labelled neurons from the LPGi and DPGi were
immunoreactive to ChAT in the PSR group. Although these choliner-
gic neurons have already been described, their physiological role is
still unknown. Of relevance to this matter, electrical stimulation of the
nucleus ROb has been shown to induce a pattern of sympathetic
activity similar to that occurring naturally during PS (Futuro-Neto &
Coote, 1982). Additional neuroanatomical and physiological studies
are needed to determine whether cholinergic neurons of the ROb are
indeed involved in such phenomena during PS.
Finally, the dPAG was the brainstem structure containing the largest
number of Fos-labelled neurons following PS rebound. This is quite
surprising because this structure has never been implicated in PS
regulation and is classically involved in fear, anxiety and vocalization
(review in Behbehani, 1995). Additional studies are again needed to
determine the role of these neurons during PS.
Structures containing Fos-positive neurons after both PS
deprivation and PS recovery
Four structures contained more Fos-labelled cells in the PSR and PSD
groups than the PSC group, namely the lPAG and vlPAG, the DPMe
and the PBL. The location of the neurons within the PBL differed
between the two conditions, indicating that two different populations
of neurons are active during PS deprivation and PS recovery. The
majority of labelled neurons were indeed clustered in the central lateral
subnucleus in the PSR group while they were more diffusely localized
in the PBL in the PSD group. Because these structures were not
previously implicated in PS control, additional studies are necessary to
determine their role.
Numerous Fos-labelled neurons were observed with a similar
distribution in the lPAG and vlPAG and the DPMe in both PSR and
PRD groups. They might therefore correspond to neurons active
during both PS deprivation and PS recovery or to two different types
of neurons active specifically during one of these conditions. A
number of previous studies suggest that the lPAG and vlPAG contain
neurons specifically active during PS while the most ventral part of the
periaqueductal grey and the adjacent DPMe contain GABAergic
neurons ceasing firing during PS. Indeed, we have shown that the
lPAG and vlPAG contain GABAergic neurons projecting to the DRN
and LC nuclei (Peyron et al., 1995; Luppi et al., 1999; Gervasoni
et al., 2000). Further, we showed that the lPAG and vlPAG contain
CTb–Fos double-labelled cells in animals that received an injection of
CTb in the LC and then were subjected to 72 h of PS deprivation and
Reticular neurons responsible for REM sleep2501
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504

Page 15

3 h of PS recovery (Verret et al., 2003a). Besides, it has been shown in
cats and rats that ejection of muscimol, a GABA agonist, in the most
ventrolateral part of the periaqueductal grey induces a strong increase
in PS quantities (Sastre et al., 1996; Boissard et al., 2000). Further, we
showed that this area and the adjacent DPMe contain GABAergic
neurons projecting to the SLD and we hypothesized that these neurons
are responsible for the inactivation of the PS-active neurons of the
SLD during W and SWS (Boissard et al., 2003). Altogether these and
our present results suggest that the lPAG and vlPAG and the DPMe
contain a mixture of neurons specifically active or inactive during PS.
Additional electrophysiological, neuroanatomical and pharmacologi-
cal studies are necessary to confirm the existence of these two types of
neurons.
Structures containing Fos-labelled neurons specifically after PS
deprivation
The DRN and RPa contained a substantial number of Fos-labelled
cells specifically in the PSD group. These results suggest that these
neurons might be strongly active in the absence of PS. They could
be serotonergic because these nuclei are known to contain seroton-
ergic neurons which show tonic activity during W, which is
decreased during SWS, and no activity during PS (Trulson &
Jacobs, 1979; Heym et al., 1982). Double-labelling experiments with
serotonin and Fos are necessary to test this hypothesis. However, not
all brainstem serotonergic neurons seem to be active in the PSD
group because only a small proportion of DRN neurons were
immunoreactive to Fos and only a few Fos-labelled neurons were
seen in other raphe nuclei such as the MnR and the ROb in this
experimental group. Studies of the Fos-labelled cells found in the
DRN and RPa might be important because they could play a crucial
role in the inhibition of PS-executive neurons during W and SWS,
although recent pharmacological results suggest that serotonergic
neurons from the DRN play no role in PS generation (Sakai &
Crochet, 2001). In that context, neurons from the LC were not Fos-
labelled in any of our experimental groups. This is not surprising for
the PSR group because noradrenergic LC neurons are known to be
inactive during PS (Aston-Jones & Bloom, 1981; Gervasoni et al.,
1998). However, it has been hypothesized that these neurons play a
key role in the inhibition of PS during W and SWS. Moreover,
Maloney et al. (1999) reported, using nearly the same protocol than
ours, a decrease in the number of Fos–tyrosine hydroxylase-positive
neurons in the PSR group compared to the PSC and PSD groups.
However, the difference found was only weakly significant and
concerned a very small number of LC noradrenergic neurons. In
addition, Merchant-Nancy et al. (1995) found no significant change
in the number of Fos-labelled neurons in the LC between their
control and PS recovery groups. Similarly, Yamuy et al. (1998)
found no significant change in the number of Fos–tyrosine
hydroxylase double-labelled cells between control and animals with
large amounts of pharmacologically induced PS. From these and our
results, it might therefore be concluded that noradrenergic LC
neurons are not strongly active during PS deprivation and therefore
might play a rather minor role in the inhibition of PS onset.
In conclusion, our results confirm that the SLD, LDTg, GiA and
GiV play a crucial role in PS onset and maintenance. In contrast,
they suggest that the cholinergic mesopontine neurons might not
play an important role. In addition, our results reveal that many
additional brainstem structures contain a large number of neurons
active during PS. We in particular found that cholinergic neurons
from the ROb, neurons from the dLPG, lPAG, vlPAG, LPGi and
PBL were immunoreactive to Fos after PS recovery. All these
nuclei might therefore also play a crucial role in PS onset and
maintenance.
These results strongly suggest that the neuronal network implicated
in PS genesis involves many more populations of neurons than
previously thought. They open the way to neuroanatomical and
physiological studies to identify the specific role of each population
revealed. Studies combining Fos staining with that of neurotransmit-
ters, enzymes, receptors, anterograde and retrograde tracers are already
under way in our laboratory to more fully identify the populations
revealed. In vivo and in vitro electrophysiological recordings are also
now necessary to characterize the firing activity of these populations
of neurons across the sleep–waking cycle.
Acknowledgements
This work was supported by CNRS (CNRS UMR5167) and Universite ´ C.
Bernard Lyon I. L.V. received a PhD grant from the French Ministe `re de la
Recherche and Lilly Institute. We also thank Anne Be ´rod for her photographic
assistance.
Abbreviations
3, oculomotor nucleus; 4, trochlear nucleus; 5, trigeminal nucleus; 6, abducens
nucleus; 7, facial nucleus; 7n, facial nerve; 10, vagus nucleus; 12, hypoglossal
nucleus; Amb, ambiguus nucleus; CGPn, central grey of the pons; ChAT,
choline acetyltransferase; CnF, cuneiform nucleus; CTb, cholera toxin b
subunit; dPAG, dorsal periaqueductal grey; DPGi, dorsal paragigantocellular
nucleus; DPMe, deep mesencephalic nucleus; DRN, dorsal raphe nucleus;
DTg, dorsal tegmental nucleus; EEG, electroencephalogram; EMG, electromy-
ogram; g7, genu of the facial nerve; Gi, gigantocellular reticular nucleus; GiA,
gigantocellular reticular nucleus, alpha part; GiV, gigantocellular reticular
nucleus, ventral part; IC, inferior colliculus; IO, inferior olivary nucleus; IP,
interpeduncular nucleus; IR, immunoreactive; KF, Ko ¨lliker–Fuse nucleus; LC,
locus coeruleus; LDTg, laterodorsal tegmental nucleus; LDTgV, laterodorsal
tegmental nucleus, ventral part; lfp, longitudinal fasciculus of the pons; LL,
lateral leminiscus; lPAG, lateral periaqueductal grey; LPGi, lateral paragigan-
tocellular and rostroventrolateral reticular nuclei; mlf, medial longitudinal
fasciculus; MnR, median raphe nucleus; MVe, medial vestibular nucleus; PB,
phosphate buffer; PBL, lateral parabrachial nucleus; PBST, PB containing 0.9%
NaCl and 0.3% Triton X-100; PBST-Az, PBSTwith 0.1% sodium azide; PCRt,
parvicellular reticular nucleus; Pn, pontine nuclei; PnC, pontine reticular
nucleus, caudal part; PnO, pontine reticular nucleus, oral part; PPTg,
pedunculopontine tegmental nucleus; PPTgM, pedunculopontine tegmental
nucleus, medial part; PS, paradoxical sleep; PSC, PS control; PSD, PS
deprivation; PSR, PS rebound; py, pyramidal tract; RMg, raphe magnus
nucleus; ROb, raphe obscurus nucleus; RPa, raphe pallidus nucleus; rs,
rubrospinal tract; scp, superior cerebellar peduncle; SLD, sublaterodorsal
nucleus; SO, superior olivary nucleus; Sol, nucleus of the solitary tract; Sp5,
spinal trigeminal nucleus; sp5, spinal trigeminal tract; SWS, slow-wave sleep;
Tz, nucleus of the trapezoid body; vlPAG, ventrolateral periaqueductal grey;
VTg, ventral tegmental nucleus; W, waking.; xscp, decussation of the superior
cerebellar peduncle.
References
Arvidsson, U., Cullheim, S., Ulfhake, B., Ramirez, V., Dagerlind, A., Luppi,
P.H., Kitahama, K., Jouvet, M., Terenius, L., Aman, K. & Ho ¨kfelt, T. (1992)
Distribution of enkephalin and its relation to serotonin in cat and monkey
spinal cord and brain stem. Synapse, 11, 85–104.
Aston-Jones, G. & Bloom, F.E. (1981) Activity of norepinephrine-containing
locus coeruleus neurons in behaving rats anticipates fluctuations in the sleep-
waking cycle. J. Neurosci., 1, 876–886.
Aston-Jones, G., Ennis, M., Pieribone, V.A., Nickell, W.T. & Shipley, M.T.
(1986) The brain nucleus locus coeruleus: restricted afferent control of a
broad efferent network. Science, 234, 734–737.
Baghdoyan, H.A. (1997) Cholinergic mechanisms regulating REM sleep. In
Schwartz, W.J. (ed.), Sleep Science: Integrating Basic Research and Clinical
Practice. S. Karger Publishing, Basel, pp. 88–116.
2502L. Verret et al.
ª 2005 Federation of European Neuroscience Societies, European Journal of Neuroscience, 21, 2488–2504