Article (PDF Available)

Homer1a is a core brain molecular correlate of sleep loss

Center for Integrative Genomics and Lausanne DNA Array Facility, University of Lausanne, Génopode, CH-1015 Lausanne, Switzerland.
Proceedings of the National Academy of Sciences (Impact Factor: 9.67). 01/2008; 104(50):20090-5. DOI: 10.1073/pnas.0710131104
Source: PubMed

ABSTRACT

Sleep is regulated by a homeostatic process that determines its need and by a circadian process that determines its timing. By using sleep deprivation and transcriptome profiling in inbred mouse strains, we show that genetic background affects susceptibility to sleep loss at the transcriptional level in a tissue-dependent manner. In the brain, Homer1a expression best reflects the response to sleep loss. Time-course gene expression analysis suggests that 2,032 brain transcripts are under circadian control. However, only 391 remain rhythmic when mice are sleep-deprived at four time points around the clock, suggesting that most diurnal changes in gene transcription are, in fact, sleep-wake-dependent. By generating a transgenic mouse line, we show that in Homer1-expressing cells specifically, apart from Homer1a, three other activity-induced genes (Ptgs2, Jph3, and Nptx2) are overexpressed after sleep loss. All four genes play a role in recovery from glutamate-induced neuronal hyperactivity. The consistent activation of Homer1a suggests a role for sleep in intracellular calcium homeostasis for protecting and recovering from the neuronal activation imposed by wakefulness.

Full-text (PDF)

Available from: Bruce F O'Hara
Homer1a
is a core brain molecular correlate
of sleep loss
Ste
´
phanie Maret*, Ste
´
phane Dorsaz*, Laure Gurcel*, Sylvain Pradervand
, Brice Petit*, Corinne Pfister*,
Otto Hagenbuchle
, Bruce F. O’Hara
, Paul Franken*, and Mehdi Tafti*
§
*Center for Integrative Genomics and
Lausanne DNA Array Facility, University of Lausanne, Ge´ nopode, CH-1015 Lausanne, Switzerland; and
Department
of Biology, University of Kentucky, Lexington, KY 40506-0225
Communicated by Michael Rosbash, Brandeis University, Waltham, MA, October 24, 2007 (received for review October 15, 2007)
Sleep is regulated by a homeostatic process that determines its need
and by a circadian process that determines its timing. By using sleep
deprivation and transcriptome profiling in inbred mouse strains, we
show that genetic background affects susceptibility to sleep loss at
the transcriptional level in a tissue-dependent manner. In the brain,
Homer1a expression best reflects the response to sleep loss. Time-
course gene expression analysis suggests that 2,032 brain transcripts
are under circadian control. However, only 391 remain rhythmic when
mice are sleep-deprived at four time points around the clock, sug-
gesting that most diurnal changes in gene transcription are, in fact,
sleep–wake-dependent. By generating a transgenic mouse line, we
show that in Homer1-expressing cells specifically, apart from
Homer1a, three other activity-induced genes (Ptgs2, Jph3, and Nptx2)
are overexpressed after sleep loss. All four genes play a role in
recovery from glutamate-induced neuronal hyperactivity. The consis-
tent activation of Homer1a suggests a role for sleep in intracellular
calcium homeostasis for protecting and recovering from the neuronal
activation imposed by wakefulness.
homeostasis microarray mRNA tagging sleep deprivation
sleep function
T
wo main processes regulate sleep. A homeostatic process
regulates sleep need and intensity according to the time
spent awake or asleep. A circadian process regulates the appro-
priate timing of sleep and wakefulness across the 24-h day. A
highly reliable index of the homeostatic process is provided by
the amplitude and prevalence of delta (1- to 4-Hz) oscillations
in the electroencephalogram (EEG) of nonrapid eye movement
(NREM) sleep (hereafter, ‘‘delta power’’). Delt a power is high
at sleep onset and decreases during sleep, in parallel with sleep
depth. Sleep deprivations and naps induce a predictable increase
or decrease, respectively, in delta power during subsequent
sleep. The interaction between homeostatic and circadian pro-
cesses is mathematically described in the two-process model of
sleep regulation, which provides a f ramework for prediction and
interpret ation of a large body of experimental data (1).
Among hypotheses concerning the physiological function of
waking-induced changes in sleep, the most compelling suggests that
sleep plays a key role in synaptic plasticity (2, 3). More specifically,
EEG delta power during NREM sleep has been shown to play a
critical role in learning-induced plasticity (46). In general, the
prediction is that local neural activation due to specific behavioral
(cognitive) demands imposes a burden on the brain which neces-
sitate s sleep and which is reflected by the EEG delta power.
On the basis of mathematical modeling and experimental data,
we have shown that sleep need, as indexed by the EEG delta power,
is under genetic control (7), which is of direct relevance for
explaining the interindividual vulnerability to sleep loss in human
subjects (8, 9). However, deciphering the molecular bases of sleep
need is rendered difficult because the contributions of the homeo-
static and circadian processes are difficult to separate and because
the impact of genetic background on brain gene expression is poorly
understood. From a series of gene-profiling experiments, we here
report a comprehensive transcriptome analysis that specifically
take s these interacting factors into account. We show that short-
term sleep loss induces changes in brain gene expression for a few
genes only. The se genes are all part of a highly specific pathway
involved in neuronal protection and recovery after waking-induced
glutamate overstimulation.
Results
We have previously reported (7) that the dynamics of sleep need
varies strongly among inbred mouse strains, with AKR/J (AK) mice
showing a dramatic increase in delta power after 6-h sleep depri-
vation, whereas DBA/2J (D2) mice have a blunted response.
Through quantitative linkage analysis, a significant quantitative
trait locus (QTL) was identified on mouse chromosome 13 (Dps1:
delta power in slow-wave sleep 1) that explains 50% of variance
in delta power after sleep deprivation in BXD recombinant inbred
lines (RIs) derived from inbred mouse strains C57BL/6J (B6) and
D2 (7). The best polymorphic marker associated was D13Mit126 at
46.5 cM (95% CI 25 cM), suggesting a large QTL region (38
Mb). However, based on the most recent high-resolution, single-
nucleotide polymorphism (SNP) genetic map of BXD RIs (10)
(http://gscan.well.ox.ac.uk/gs/strains.cgi), the smallest differential
region corre sponds to an 11-Mb sequence flanked by SNPs
rs3669221 and rs397202. According to the Ensemble database (Mus
musculus release 46), this region contains 33 known genes, 15
potential unknown coding sequences, and three pseudogenes.
Among these, the short splice variant of the Homer1 (Homer1a)
gene is the only transcript in the region that was previously reported
to be among the up-regulated genes after sleep deprivation (11–14).
However, the specificity of this finding, compared with other gene
expre ssion changes after sleep loss, has not yet been established.
To confirm and verify whether differences in sleep need between
mouse strains are regulated at the transcriptional level, mice of
three genotypes (AK, B6, and D2) were deprived of sleep by gentle
handling [see supporting information (SI) Materials and Methods]
for 6 h, starting at light onset, and killed at the same time of day
[Zeitgeber time (ZT)6] with their non-sleep-deprived controls.
Because it is believed that sleep fulfills a brain-specific function, we
also sampled the liver as a peripheral reference organ. Microarray
results were analyzed by a two-way ANOVA, with strain and
condition as main factors. Venn diagrams summarizing the data are
shown in Fig. 1 A and B and sugge st that very few gene s show
consistent changes in expression across genetic backgrounds. We
use the terms ‘‘consistent’’ and ‘‘reliable’’ hereafter only for tran-
Author contributions: S.D. and L.G. contributed equally to this work; O.H., P.F., and M.T.
designed research; S.M., S.D., L.G., B.P., C.P., O.H., and B.F.O. performed research; S.M., S.D.,
L.G., S.P., B.F.O., P.F., and M.T. analyzed data; and S.M., S.D., L.G., S.P., P.F., and M.T. wrote
the paper
The authors declare no conflict of interest.
Data deposition: The data reported in this paper have been deposited in the Gene
Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE9444).
§
To whom correspondence should be addressed. E-mail: mehdi.tafti@unil.ch.
This article contains supporting information online at www.pnas.org/cgi/content/full/
0710131104/DC1.
© 2007 by The National Academy of Sciences of the USA
20090–20095
PNAS
December 11, 2007
vol. 104
no. 50 www.pnas.orgcgidoi10.1073pnas.0710131104
Page 1
scripts either increased or decreased in all three inbred strains and
across experimental conditions.
Sleep deprivation induced changes in only 42 brain probe sets in
B6, 92 in D2, and 188 in AK, among which 52 sets were affected by
genotype. To our surprise, sleep loss induced almost three times
more transcriptional changes in liver compared with whole brain
(Fig. 1A). As in brain, among all transcripts with significant changes
in liver, 50% changed as a function of genetic background. Thus,
for genes differentially expre ssed in each tissue of the three strains
in response to sleep deprivation, we have focused on those showing
a significant two-way ANOVA interaction (see SI Tables 1 and 2).
As reported in refs. 11, 15, and 16, sleep deprivation most signifi-
cantly up-regulated the expression of heat shock protein genes in
both brain and liver, strongly suggesting that this group of genes is
part of a general stress–re sponse pathway induced by enforced
wakefulness in most organs, and, therefore, is not specific to the
brain. The most significant decrease was found for the cold-induced
RNA binding protein (Cirbp) in both tissue s, again suggesting a
general rather than a brain-specific implication (SI Tables 1 and 2).
In brain, the expression of Homer1a was most affected by both
genotype and sleep loss. In situ hybridization analysis confirmed the
strong induction of Homer1a transcript after sleep deprivation and
indicated a restricted up-regulation in cortex, striatum, and hip-
pocampus (Fig. 1C). Homer1 encodes several transcripts by alter-
native splicing, among which long forms are constitutively expressed
while two short forms, Homer1a and Ania-3, are activity-induced.
These postsynpatic density proteins bind calcium-signaling mole-
cules and have been implicated in synaptic plasticity.
In contrast to most other activity-induced genes, Homer1a over-
expre ssion seemed highly specific. For instance, brain-derived
neurotrophic factor (Bdnf ), a plasticity-related gene, showed a
significant increase in AK and D2 only; the immediate early gene
Fos did not show a significant increase in D2; and Arc, Egr1, and
Egr3 were induced by sleep deprivation mainly in AK (SI Table 1).
Thus, comparisons among genotypes, and between brain and liver,
identified Homer1a as the most specific transcriptional index of the
whole brain in response to sleep loss.
We then analyzed the time course of Homer1a induction by
real-time quantitative RT-PCR in a dose–response experiment.
Mice of all three strains were sleep-deprived for 1, 3, and6hand
killed together with their time-matched controls. An additional
group, which was also sleep-deprived for 6 h, was killed2hinto
recovery sleep. As shown in Fig. 1D, Homer1a is rapidly and
strongly induced by sleep deprivation in a dose-dependent manner.
Parallel to this increase, Homer1a expression decreases with in-
creasing accumulated sleep in non-sleep-deprived animals, and its
relative level remains significantly higher than in the time-matched
controls after only2hofrecovery sleep, indicating that the time
course of Homer1a expre ssion closely parallels sleep need. How-
ever, the fold change of Homer1a expression after 6-h sleep
deprivation was similar between D2 and AK, which represent the
two extreme strains in their response to sleep deprivation, and was
significantly higher than in B6 mice (Fig. 1E). In accord with our
findings, sequence comparisons in public databases for the Homer1
region indicated that D2 and AK share the same SNP haplotype,
which is different from that of B6, although no coding sequence
difference s between any of the three strains can be identified. This
finding suggested that, even though Homer1a is the best candidate
for the Dps1 QTL segregating with response to sleep loss between
D2 and B6 mice, other genes may affect sleep need in other inbred
strains, or posttranscriptional or translational changes of Homer1a
may differ between D2 and AK mice.
Time-of-Day Effects of Sleep Loss on Brain Transcriptional Changes. As
shown in Fig. 1D, the level of Homer1a expre ssion under baseline
conditions shows significant variation that might be either due to a
direct circadian effect or driven by the diurnal sleep–wake distri-
bution, as we suggest. To separate time-of-day and homeostatic
effects on brain gene expression, we have simulated the time course
of sleep need (quantified as EEG delta power) according to
published assumptions and parameters (7) under entrained base-
line conditions and after 6-h sleep deprivation at four time points
around the 24-h day. If a gene is implicated in the processe s
underlying sleep need, then changes in its expre ssion can be
expected to parallel the predicted time course of sleep need (Fig.
2A). Under control conditions, sleep need is high at sleep onset,
which coincides with light onset (ZT0), and during the latter part
of the active period (ZT18), whereas sleep need is lowest between
ZT6 and ZT12 (i.e., dark onset) (Fig. 2 A). Under sleep deprivation
conditions, these pronounced sleep–wake-dependent changes are
expected to be strongly reduced. Thus, we sleep-deprived mice of
the three strains for 6 h, starting at ZT0, ZT6, ZT12, or ZT18, and
Fig. 1. Transcriptome analysis of the brain and liver after sleep deprivation
in three inbred mouse strains indicates that Homer 1a is specifically upregu-
lated in the brain. (A and B) Venn diagrams of a two-way ANOVA microarray
data analysis of total RNA extracted from brain (A) and liver (B) in AK, B6, and
D2 inbred mouse strains at ZT6 after a 6-h sleep deprivation. Results are
reported as a function of genetic background (genotype), experimental con-
dition (sleep deprivation), and their interaction. (C) Homer1a is overexpressed
after sleep deprivation (SD) in the cortex and caudate putamen (Left) and in
the hippocampus (Right). (D) Mean (1 SEM) forebrain mRNA levels of
Homer1a for AK, B6, and D2 mice after SDs of 1, 3, and 6 h (pink bars), with
their time-matched controls (gray bars). All SDs started at light onset (ZT0).
Effects of recovery sleep on expression were assessed by allowing2hof
recovery after 6-h SD (ZT8). Homer1a expression was affected by SD, geno-
type, and time of day (P 0.05; three-way ANOVA). (E) Fold change increase
of Homer1a expression level after6hofSDvaries among strains. Ratios
between SD and control mRNA levels at ZT 6 were calculated as described in
SI Materials and Methods, with their standard errors. The effect of SD on
Homer1a expression is statistically different between B6 and the two other
strains (, P 0.006; two-way ANOVA).
Maret et al. PNAS
December 11, 2007
vol. 104
no. 50
20091
NEUROSCIENCE
Page 2
killed them together with their home-cage controls (n 9 per strain
per time per condition; i.e., total 216). A linear statistical model was
first used (see SI Materials and Methods) and identified 2,540 probe
sets that were significantly changed at any time point (false-positive
rate; false discovery rate 5%). The se probe sets were then
assessed for time-of-day variation separately in the baseline and
sleep-deprivation conditions. Under the baseline condition, 8%
(2,032; see SI Table 3) of probe sets detected in the brain showed
a significant time-of-day pattern of expre ssion (Fig. 2 C and D), with
two major, opposite phases of peak expre ssion (between ZT6 and
ZT12 and between ZT18 and ZT0, re spectively; Fig. 2D). Under the
sleep-deprivation condition, only 391 (SI Table 4) of the 2,032 sets
at baseline were still significantly affected by time of day, indicating
that most others changed according to the sleep–wake distribution,
rather than as a result of (or in addition to) a direct circadian effect.
Among the 391 transcripts that maintained a significant cycling
pattern after sleep deprivation, a large majority reached maximum
levels of expression between ZT0 and ZT6 and between ZT12 and
ZT18 (Fig. 2 C and D). Among all rhythmic transcripts under
control conditions, Homer1a clearly showed the largest amplitude
of variation (Fig. 2 B, D, and E), beyond that of all canonical
circadian genes. Time course of expre ssion of six representative
genes under the two conditions is depicted in Fig. 2E. As predicted
by the simulation analysis (Fig. 2 A), activity-regulated genes such
as Homer1a, Arc, and Egr3 show a high amplitude variation (up to
6-fold for Homer1a), closely following the diurnal sleep–wake
distribution, and no, or greatly reduced, variation after sleep
deprivation. Although the pattern of expression of circadian genes
remains little affected by sleep deprivation, their relative levels can
be significantly affected (unchanged for Bmal1, increased for Per2,
and decreased for Dbp;Fig.2E), as we also reported in refs. 17
and 18.
Comparison of sleep-deprived and control mice at all four time
points again revealed that 2% (343 probe sets) of the expressed
genes in the brain are up-regulated (70%, or 249 probe sets) or
down-regulated (30%, or 94 probe sets) by sleep loss (SI Table 5).
Significant interaction between condition and time of day was
detected for 585 probe sets. Again, the most significantly overex-
pressed gene after sleep deprivation was Homer1a, followed by
those belonging to stress–response and synaptic-plasticity gene
groups. The major functional gene groups that reduced their
expre ssion after sleep deprivation concern protein synthesis, mem-
brane trafficking, and protein transport (SI Table 5) (12). Real-time
quantitative RT-PCR verification for 41 candidate genes (found
here and by others) at ZT6 confirmed our microarray findings at
this reference time point (SI Table 6).
Cell-Specific Transcriptional Changes Due To Sleep Loss. Previous
studies (12, 19, 20) investigated transcriptomes of different brain
ZT 0 6 12 18 0 6 12 18 0 6 12 18
Homer1a control
AK B6 D2
Control
SD
D
B
A
E
ZT 6
ZT12
ZT18
Control
Cycling
in SD
C
AK B6 D2AK B6 D2AK B6 D2
min
SD
max
sin coeff
f
feoc
s
oc
ZT 0
ZT 6
ZT12
ZT18
ZT 0
ZT0
ZT6
ZT12
ZT18
ZT
6
Z
T 6
ZT0
ZT6
ZT12
ZT18
erusserp peelS
ZT [h]
langiS ANRm
)X001(
ZT [h]
Homer1a
Egr3
Arc
Bmal1
Per2
Dbp
Fig. 2. Around-the-clock transcriptome
analysis of the brain after sleep deprivation
indicates that most changes are affected by
behavioral states. (A) Simulation of
sleep need in the around-the-clock sleep-
deprivation experiment. Sleep-need dynam-
ics in the control (gray) were modeled
mathematically according to Franken et al.
(7): lights on at ZT0, lights off at ZT12. The
increase in sleep-need during the four 6-h
sleep deprivations (SD) was determined by
assuming a saturating exponential increase
during wakefulness (red dashed lines). The
red solid line connects values of sleep need
reached at the end of the SDs. (B and C)SD
affects the cycling pattern of gene expression
in the brain. Rhythmic transcripts were se-
lected, as outlined in SI Materials and Meth-
ods, and their temporal expression patterns
were aligned according to phase. (B) Enlarge-
ment of heat map for Homer1a gene in con-
trol condition. For AK, B6, and D2, the four
time points (ZT0, 6, 12, and 18) are repre-
sented in triplicate. Green and red represent
minimal and maximal expression levels, re-
spectively. (C Left) The 2,032 genes cycling in
controls are depicted. The peak time of ex-
pression is indicated at left. (Center) The
same genes are represented after SD. Note
that the rhythmic accumulation of most tran-
scripts is severely blunted after SD. (Right)
The 391 probe sets for which cycling expres-
sion profiles are not affected by SD (with the
peak of expression indicated at right). (D)
Cycling genes shown in C are plotted accord-
ing to their amplitude and phase. Homer1a,
outlined in green, is the most rhythmic tran-
script under control conditions. (E) Temporal
expression profiles of representative homeo-
static (Left) and circadian (Right) genes in
whole brain. Normalized expression signals
from microarray experiments are plotted as a
function of time for control (solid lines) and
SD (dashed lines) mice of each strain. Each
point is the mean SEM of three pools of
three animals.
20092
www.pnas.orgcgidoi10.1073pnas.0710131104 Maret et al.
Page 3
structure s, rather than the whole brain. In contrast to most other
organs, the brain is a highly heterogeneous tissue, with specialized
regions and nuclei having very different functional roles. Whole-
brain transcriptome, therefore, suffers several potential limitations,
including dilution of low copy-number transcripts and missing those
with opposing patterns of transcriptional changes among brain
regions. On the other hand transcriptome analysis of selected
regions also has drawbacks because of our limited knowledge of
exactly where functionally significant molecular changes can be
expected and because, even in well defined regions, cell type s vary
greatly. To overcome these limitations, we chose to analyze changes
in mRNA profiles of those neurons that are selectively and specif-
ically activated by sleep deprivation. To this end, we used a modified
mRNA tagging technique originally established in Caenorhabditis
elegans and Drosophila (21–23).
We have shown here that Homer1a is the transcript most
consistently increased by enforced wakefulness. The Homer1 gene
is therefore a good marker for neuronal populations activated by
sleep loss but not restricted to a single structure. We thus generated
BAC-based transgenic mice by replacing the first five exons of
Homer1 (corresponding to activity-induced Homer1a transcript;
Fig. 3A) by a FLAG-tagged poly(A) binding protein 1 (PABP)
followed by internal ribosome entry site (IRES)-eGFP (Fig. 3B).
Because PABP binds poly(A) tails of mRNAs, affinity-purification
of FLAG-tagged PABP proteins from whole-brain lysates is ex-
pected to coprecipitate all mRNAs from neurons expressing
Homer1a. Seven transgenic lines were obtained, and analysis indi-
cated that at least three lines expressed the transgene at very similar
amounts, at both mRNA and protein levels (Fig. 3 C and D). Also,
the induction of the transgene by 6-h sleep deprivation was very
similar to that of endogenous Homer1a, indicating that our BAC
construction contains the regulatory elements for the correct
functional expression of Homer1a (Fig. 3D).
The expre ssion of this construct was also verified by in situ
hybridization (Fig. 4A). Although eGFP inserted after the IRES did
not result in reliable signal under epifluorescence, riboprobe s
against Homer1a and eGFP clearly indicated that both are similarly
coexpressed in the same brain regions (Fig. 4A). Double fluores-
cent in situ hybridization (FISH) with riboprobes against Homer1a
and eGFP revealed that, at least in two transgenic lines, eGFP was
almost exclusively expressed in Homer1a-expre ssing neurons (Fig.
4B), which represented 40% of the neurons in the cingulate cortex
and 30% in the dorsal striatum (not counted in the hippocampus).
All mRNA immunoprecipitations and microarray data presented
below are from a single transgenic line (line 36). Sleep recordings
in Homer1-PABP transgenic mice indicated that they react to sleep
deprivation similar to their wild-type littermate s (SI Fig. 5). The
specificity of this mRNA pull-down method was also verified by
quantitative RT-PCR with probes specific for eGFP, FLAG,
Homer1a, and Hcrt (a gene expressed only in the lateral hypothal-
amus, where Homer1a expre ssion is very low). The results indicated
that eGFP, FLAG, and the endogenous Homer1a were enriched
and induced by sleep deprivation, whereas only trace amounts of
Hcrt could be detected (SI Fig. 6).
Immunoprecipitated mRNAs were prepared from Homer1-
PABP transgenic mice with (n 6) or without (n 6) a 6-h sleep
deprivation at light onset for gene expression profiling. For com-
parison, RNA extractions were also made from supernatants (n
4) after immunoprecipitation and from transgenic whole brains
(n 8). To test for transcriptional changes after sleep deprivation
in Homer1-expressing cells, we proceeded in two steps: (i)we
identified probe sets enriched in the pull-down extracts and (ii)
among those probe sets, we compared sleep deprivation with
control condition in both pull-down (6 vs. 6-chip comparison) and
whole-brain (4 vs. 4-chip comparison) extracts. We found that 4,728
probe sets were significantly enriched at 5% false discovery rate
when pull-downs were compared with both supernatant and whole-
brain extracts (SI Materials and Methods). Again, very few genes
were identified (SI Table 7), among which the most significant
ones—Homer1a, Egr2 (NGFI-B), and Fosl2 (Fos-like antigen 2)—
were similarly induced after sleep loss in the pull-down and the
whole-brain extracts, suggesting that gene expression changes in
the pull-down samples recapitulate the most significant changes at
the whole-brain level. In addition, several unique immediate-early
genes were specifically identified in pull-down samples that might
be coinduced with Homer1a, namely prostaglandin–endoperoxide
synthase 2 (Ptgs2), junctophilin 3 (Jph3), and neuronal pentraxin 2
(Nptx2). Interestingly, both Jph3 and Nptx2 are activity-induced
through either ryanodine receptor-mediated intracellular calcium
mobilization (Jph3) or activity-induced AMPA receptor synaptic
clustering (Nptx2). Among the very few down-regulated transcripts
(SI Table 7), we have identified another activity-induced gene,
4-nitrophenylphosphatase domain and nonneuronal SNAP25-like
protein homolog 1 (Nipsnap1), suggesting that plasticity genes can
be up- or down-regulated by sleep deprivation.
B
C
PABPFLAG
500
300
700
Tg mice
+-+- +-
eGFP
D
eGFP
36
22
98
64
142
Flag
Sleep Deprived
Tg mice
-+-+
Control
5’ 6’
Recombination
Flip-out of Zeo
IRES eGFPPABP
F
l
a
g
ZEO
179Kb 36Kb
BAC
BXD-F2 pronucleus injection
BAC
IRES eGFPPABP
Flag
ZEO
123 54
BAC
IRES eGFPPABP
F
la
g
11
22
33
Chrom.
IRES eGFPPABP
Flag
A
Ania-3
1
H1a H1b/c
23 6789 1054Exons
~100kb
Loading control
Fig. 3. A transgenic mouse model to analyze neuron-specific gene expres-
sion. (A) Schematic representation of the genomic structure of Homer1. The
putative transcriptional initiation site is depicted by a bent arrow at the
beginning of exon 1; the translational stops for short activity-induced
Homer1a (H1a) and Ania-3, as well as for long constitutively expressed
Homer1b/c (H1b/c) are indicated by black circles. Intron 5 is here divided into
four segments (4.4 kb of Homer1a 3 UTR, 5.7 kb up to Ania-3, 1.4 kb of
Ania-3-specific sequence, and 18.8 kb to exon 6). The Homer1a-specific exon
5 extends exon 5 by the intron 5 sequence. The Ania-3-specific exon 6 sits
within intron 5 (adapted from ref. 25). (B) General strategy for generating
Homer1a-PABP transgenic mice. The relative position of the Homer1a gene in
the fully sequenced 227,644-bp RP23–262I3 BAC clone (GenBank accession no.
AC120347), and the different steps used to introduce a PCR-amplified con-
struction containing PABP cDNA, followed by IRES-eGFP and the zeocine (ZEO)
selection cassette flanked by FRT sites (hashed boxes) by BAC recombination
in the EL250 bacterial stain (see SI Materials and Methods). (C) Transgenic mice
(Tg) were identified by RT-PCR with the primer pairs depicted in B for the
presence of the FLAG (primers 1/1), PABP (primers 2/2), and eGFP (primers 3/3).
(D) Western blot verification of transgene expression in transgenic mice line
36 (Tg) indicated the presence of the FLAG and eGFP and the up-regulation of
the FLAG after sleep deprivation. The loading control is a nonspecific band
generated by the GFP antibody.
Maret et al. PNAS
December 11, 2007
vol. 104
no. 50
20093
NEUROSCIENCE
Page 4
Discussion
The results presented here demonstrate that6hofsleepdepriva-
tion, which importantly impacts sleep physiology and behavior,
results in only minimal changes in brain transcriptional adaptation.
As reported for changes in delta power (7), we also showed here
that sleep loss-induced transcriptional changes are largely affected
by genetic background. As opposed to other studies that did not
take genetic background into account and did not contrast their
findings to peripheral tissue (12, 15, 20), our results indicate that
only a few genes reliably change expre ssion after sleep deprivation.
The surprising finding that sleep loss induced a larger number of
changes in liver than in brain suggests either that sleep deprivation
might have a specific impact on the liver or that the brain might be
protected against major transcriptional changes.
Another important aspect of the present study is the interaction
between the homeostatic and circadian processe s. Although we did
not sleep-deprive the animals under constant conditions, and
therefore the direct and indirect effects of light, for instance, on
gene expre ssion could not be accounted for, we have shown that a
large majority (80%) of changes in gene expression were driven
by the prior sleep–wake history.
We also adopted, and further developed, a reliable mRNA
tagging technique to investigate gene expre ssion changes in neu-
rons. This technique can be used to evaluate different neuronal
subpopulations without the burden of sampling several structures
or using labor-intensive laser microdissection to isolate neurons.
The results of this technique confirmed that sleep loss-induced
transcriptional changes occur for very few genes, among which
Homer1a remains the most specific.
In addition to Homer1a, we identified overexpression of other
genes involved in synaptic plasticity, but only Egr2 and Homer1a
were found to consistently change across experiments. Others
reported overexpre ssion for a number of plasticity-related genes,
and these observations are commonly used in support of a func-
tional role for sleep in plasticity. Because the expre ssion of most
plasticity-related gene s was not reliably changed, our findings do not
support such a general conclusion and instead suggest that the
molecular mechanisms might not be identical for most plasticity-
regulated genes. In this context it is important to note that
Nipsnap1, one of the proposed plasticity genes (24), is actually
down-regulated after sleep deprivation in our mRNA tagging
experiment.
Three different genes in mammals encode Homer proteins.
Homer1 encodes constitutively expressed long-form proteins,
whereas short-form Homer1a is activity-induced (25). Homer1
long-form proteins dimerize and interact with metabotropic gluta-
mate receptors and increase calcium from intracellular stores.
Short-form proteins, which lack the dimerization domain, function
as natural activity-dependent dominant negative forms that regu-
late the scaffolding and signaling capabilities of the long forms and
reduce glutamate-induced intracellular calcium release (26, 27). We
have recorded sleep, and the response to a 6-h sleep deprivation, in
Homer1 (all forms) mutant mice but found a very similar pattern
compared with their wild-type littermates (data not shown). This
finding could be expected due to the fact that, because Homer1a
functions as a dominant negative form of the long forms, consti-
tutive loss of all isoforms might not result in any specific sleep
phenotype.
Conceptually, spontaneous or enforced wakefulness repre-
sents a stressor activating a series of stress–response mecha-
n isms of the organism, which, at the transcriptional level, could
be translated into the induction of genes such as heat shock
proteins in most tissues. However, unlike in other organs,
brain-specific stress–response pathways are primarily triggered
by glut amate. Glutamate is the major excit atory neurotrans-
mitter in the central nervous system and acts through either
ionotropic or met abotropic receptors (mGluRs). Long Hom-
er1 tetramers bind g roup I mGluRs and inositol 1,4,5-
triphosphate receptors, thus enabling ef ficient calcium release
f rom intracellular stores, whereas monomeric Homer1a c om-
petitively disr upts synaptic glutamatergic signaling c omplexes
to reduce glut amate-induced intracellular calcium release (26,
27). Homer1 also activates ryanodine receptors and L-t ype
calcium channels (28, 29). Interestingly, Jph3, which was
identified by our mRNA-t agging strateg y as being up-
regulated by sleep deprivation, has been shown to play a major
role in ryanodine receptor-mediated, calcium-induced open-
ing of small-c onductance, calcium-activated pot assium (SK)
channels (28, 30). SK channels are responsible for the gener-
ation of slow afterhyperpolarizations in neurons of the nucleus
reticularis thalami and thus contribute to the EEG slow waves
characteristic of NREM sleep (31).
According to Tononi and Cirelli (3), plastic processe s occurring
during wakefulness result in increased synaptic strength, whereas
the role of sleep is to downscale synaptic strength to a basal level.
Homer1a transcription is rapidly up-regulated in neurons in re-
sponse to synaptic activity induced by long-term potentiation,
seizure, inflammation, stimulant drugs, or even selectively in the
hippocampus of rodents by exploratory behavior (32, 33). In this
context, Homer1a, by buffering intracellular calcium and disassem-
bling synaptic glutamatergic signaling complexes, could play a
pivotal role in synaptic downscaling. Because Homer1a and the
other genes identified here (Egr2, Fosl2, Ptgs2, Jph3, and Nptx2) are
all induced by stressful conditions such as seizure, stroke and
hypoxia, and inflammation, an alternative, complementary view
could be that they play a primary brain-protective or recovery role.
This view is also of relevance for the etiology of neuropsychiatric
disorders because it is increasingly recognized that stress is impli-
cated in many of such disorders (34). Both environmental and
pharmacological stressors up-regulate Homer1a mRNA in key
structure s involved in higher brain functions (35): the same struc-
tures in which Homer1a is up-regulated after sleep deprivation. It
has also been shown that overexpre ssion of Homer1a after inflam-
mation, seizure, and psychostimulant or antipsychotic drug use
plays a major role in neuroprotection (28, 35). It is tempting to
relate the dramatic improvement in depression in humans after
sleep-deprivation (36) to the sleep deprivation-induced up-
esnesitnAesneS
A Homer1a eGFP
Pir
CPu
Cg
d
e
gr
e
M
IPA
D
PFG
a1remoH
B Antisense Sense
5µm
Fig. 4. Colocalization of Homer 1a and FLAG-tagged PABP eGFP transcripts.
(A) In situ hybridization with Homer1a and eGFP antisense riboprobes indi-
cated that both are expressed in similar brain structures. Cg, cingulate cortex;
CPu, caudate putamen; Pir, piriform cortex. (B) Confocal images of FISH
analysis. FISH experiments were performed with both Homer1a (red) and eGFP
(green) riboprobes at the same time and revealed that almost all positive
neurons were double-labeled, indicating the colocalization of the endoge-
nous Homer1a and the transgene. Negative control conditions were obtained
with both sense riboprobes.
20094
www.pnas.orgcgidoi10.1073pnas.0710131104 Maret et al.
Page 5
regulation of Homer1a. In conclusion, converging evidence strongly
implicates Homer1a as a brain-coping marker against stressors, and
our findings suggest that Homer1a might represent the molecular
link between sleep, cognition, and neuropsychiatric disorders.
Materials and Methods
Animal Handling. A ll ex periments were perfor med in ac c or-
dance with the protoc ols approved by the Ethical Committee
of the State of Vaud Veterinary Office, Switzerland. Sleep-
deprivation and sleep-recording procedures are described in SI
Mater ials and Methods.
cRNA Preparation, cDNA Microarray Hybridization, and Real-Time
RT-PCR.
For the first experiment, we isolated total RNA from whole
brain and liver by using the RNAXEL kit (Eurobio), treated the
RNA with DNase, and cleaned it using RNeasy columns (Qiagen).
Equal quantities of total RNA from three individual mice of each
strain were pooled in triplicate (nine mice of each strain in each
condition). A hybridization mixture containing 15
g of biotin-
ylated cRNA was hybridized to GeneChip Mouse Expression Set
430. Chips were washed, scanned, and analyzed with Affymetrix
GeneSpring software.
For the around-the-clock microarray experiment, RNA from
whole brain was isolated and purified with the RNeasy Lipid Tissue
Midi kit (Quiagen) and DNase-treated. All RNA quantities were
assessed with a NanoDrop ND-1000 spectrophotometer, and the
quality of RNA was controlled on Agilent 2100 bioanalyzer chips.
Equal amounts of total RNA were pooled from three mice within
each of the 24 experimental groups (three strains, two conditions,
4ZT 24; in triplicate: 24 3 72 chips). Three micrograms of
each of these pools were used to perform the chip array experiment,
according to the Affymetrix Gene Expression procedure. Twelve
micrograms of biotinylated cRNA from each sample were frag-
mented and hybridized to GeneChip Mouse 430 2.0 arrays, accord-
ing to standard procedures. Microarray analyses and qPCR verifi-
cations were performed as reported in SI Materials and Methods.
Immunoblot and
in Situ
Hybridization. Total protein extract was
prepared with RIPA lysis buffer. Protein concentration was calcu-
lated by using the bicinchoninic acid assay (Pierce) with BSA as a
standard. Eighty micrograms of each fraction were analyzed by
SDS/PAGE, followed by We stern blotting using antibodies as
follows: mouse anti-tubulin 1/1,000 (Santa Cruz), goat anti-
Homer1a 1/200 (Santa Cruz), mouse anti-Flag M2 1/300 (affinity-
purified; Sigma), and rabbit anti-GFP 1/2,500 (AbCam). Secondary
antibodies were all coupled with HRP, except for the anti-goat
antibody, which was IRDye800-conjugated for Lycor analysis.
In situ hybridizations with coronal cryosections of 12
m were
performed according to Allen Brain Atlas protocols (enzymatic
BCIP/NBT revelation) (37). All reagents and solutions were pur-
chased and prepared based on Eurexpress II in situ hybridization
consortium instructions. GFP and Homer1a riboprobes were syn-
thesized by in vitro transcription on a linearized pGEM-Easy vector
(Promega) containing the corresponding sequences. The cDNA
insert of this plasmid was generated by RT-PCR from mouse brain
RNA, using the following primers: Homer1a for ward, 5-
GCTGTCAGAAGCTTAGGATGTG-3; Homer1a reverse,
5-AAAGTGCAGAAAGTCCAGCAGC-3; GFP forward, 5-
GAGCTGGACGGCGACGTAAACG-3; and GFP reverse, 5-
AGGACCATGTGATCGCGCTTCTC-3.
FISH was perfor med as described in ref. 38, using anti-DIG-
POD 1/600 (Roche), anti-FLU-AP 1/100 (Roche), and SA-A lexa
488 (Molecular Probes) and counterstained with DAPI (Sigma).
Transgenic and mRNA Tagging. Transgenic mice were generated as
described in Fig. 3. See SI Materials and Methods for details.
We thank K. Harshman, A. Paillusson, and M. Bueno for assistance in
microarray and real-time RT-PCR analyses at the Lausanne DNA Array
Facility; P. Descombes, M. Docquier, D. Chollet, and C. Delucinge for
assistance in microarray and real-time RT-PCR analyses at the Geneva
Genomics Platform, National Center for Competence in Research Frontiers
in Genetics; S. Excoffier for help in the transgenic construction; P. Seeburg,
M. Schwarz (Max Planck Institute, Heidelberg, Germany), and P. Worley
(Johns Hopkins School of Medicine, Baltimore, MD) for providing Homer1
mutant mice; and A. Vassali for constructive discussions. This work was
supported by the Swiss National Science Foundation and the State of Vaud
(M.T.) and in part by National Institutes of Mental Health Grant MH67752
(to P.F.).
1. Dijk DJ, Franken P (2005) in Principles and Practice of Sleep Medicine, eds
Kryeger TH, Roth TH, Dement W (Saunders, Philadelphia), pp 418 434.
2. Krueger JM, Obal F (1993) J Sleep Res 2:63–69.
3. Tononi G, Cirelli C (2006) Sleep Med Rev 10:49 62.
4. Huber R, Ghilardi MF, Massimini M, Ferrarelli F, Riedner BA, Peterson MJ,
Tononi G (2006) Nat Neurosci 9:1169–1176.
5. Marshall L, Helgadottir H, Molle M, Born J (2006) Nature 444:610613.
6. Molle M, Marshall L, Gais S, Born J (2004) Proc Natl Acad Sci USA
101:13963–13968.
7. Franken P, Chollet D, Tafti M (2001) J Neurosci 21:2610–2621.
8. Van Dongen HP, Baynard MD, Maislin G, Dinges DF (2004) Sleep 27:423–433.
9. Tucker AM, Dinges DF, Van Dongen HP (2007) J Sleep Res 16:170–180.
10. Shifman S, Bell JT, Copley RR, Taylor MS, Williams RW, Mott R, Flint J
(2006) PLoS Biol 4:e395.
11. Cirelli C, Gutierrez CM, Tononi G (2004) Neuron 41:35–43.
12. Mackiewicz M, Shockley KR, Romer MA, Galante RJ, Zimmerman JE,
Naidoo N, Baldwin DA, Jensen ST, Churchill GA, Pack A (2007) Physiol
Genomics 31:441–457.
13. Nelson SE, Duricka DL, Campbell K, Churchill L, Krueger JM (2004) Neurosci
Lett 367:105–108.
14. Huber R, Tononi G, Cirelli C (2007) Sleep 30:129–139.
15. Terao A, Steininger TL, Hyder K, Apte-Deshpande A, Ding J, Rishipathak D,
Dav is RW, Heller HC, Kilduff TS (2003) Neuroscience 116:187–200.
16. Wisor JP, Morairty SR, Huynh NT, Steininger TL, Kilduff TS (2006) Neuroscience
141:371–378.
17. Franken P, Thomason R, Heller HC, O’Hara BF (2007) BMC Neurosci 8:87.
18. Wisor JP, O’Hara BF, Terao A, Selby CP, Kilduff TS, Sancar A, Edgar DM,
Franken P (2002) BMC Neurosci 3:20.
19. Cirelli C, Faraguna U, Tononi G (2006) J Neurochem 98:1632–1645.
20. Terao A, Grec o MA, Davis RW, Heller HC, Kilduff TS (2003) Neuroscience
120:1115–1124.
21. Roy PJ, Stuart JM, Lund J, Kim SK (2002) Nature 418:975–979.
22. Kunitomo H, Uesugi H, Kohara Y, Iino Y (2005) Genome Biol 6:R17.
23. Yang Z, Edenberg HJ, Davis RL (2005) Nucleic Acids Res 33:e148.
24. Satoh K, Takeuchi M, Oda Y, Deguchi-Tawarada M, Sakamoto Y, Matsubara
K, Nagasu T, Takai Y (2002) Genes Cells 7:187–197.
25. Bottai D, Guzowski JF, Schwarz MK, Kang SH, Xiao B, Lanahan A, Worley
PF, Seeburg PH (2002) J Neurosci 22:167–175.
26. Worley PF, Zeng W, Huang G, Kim JY, Shin DM, K im MS, Yuan JP, Kiselyov
K, Muallem S (2007) Cell Calcium 42:363–371.
27. Xiao B, Tu JC, Worley PF (2000) Curr Opin Neurobiol 10:370 –374.
28. Kakizawa S, Kishimoto Y, Hashimoto K, Miyazaki T, Furutani K, Shimizu H,
Fukaya M, Nishi M, Sakagami H, Ikeda A, et al . (2007) EMBO J 26:1924–1933.
29. Yamamoto K, Sak agami Y, Sugiura S, Inokuchi K, Shimohama S, Kato N
(2005) Eur J Neurosci 22:1338–1348.
30. Moriguchi S, Nishi M, Komazak i S, Sakagami H, Miyazaki T, Masumiya H,
Saito SY, Watanabe M, Kondo H, Yawo H, et al. (2006) Proc Natl Acad Sci USA
103:10811–10816.
31. Blethyn KL, Hughes SW, Toth TI, Cope DW, Crunelli V (2006) J Neurosci
26:2474–2486.
32. Tappe A, Klugmann M, Luo C, Hirlinger D, Agarwal N, Benrath J, Ehren-
gruber MU, During MJ, Kuner R (2006) Nat Med 12:677–681.
33. Kato A, Ozawa F, Saitoh Y, Fukazawa Y, Sugiyama H, Inokuchi K (1998) J Biol
Chem 273:23969–23975.
34. Southw ick SM, Vythilingam M, Charney DS (2005) Annu Rev Clin Psychol
1:255–291.
35. Szumlinski KK, Kalivas PW, Worley PF (2006) Curr Opin Neurobiol 16:251–
257.
36. Giedke H, Schwarzler F (2002) Sleep Med Rev 6:361–377.
37. Lein ES, Hawrylycz MJ, Ao N, Ayres M, Bensinger A, Bernard A, Boe AF,
Boguski MS, Brockway KS, Byrnes EJ, et al. (2007) Nature 445:168–176.
38. Ishii T, Omura M, Mombaerts P (2004) J Neurocytol 33:657–669.
Maret et al. PNAS
December 11, 2007
vol. 104
no. 50
20095
NEUROSCIENCE
Page 6
  • Source
    • "Recent animal and human data provide strong evidence that the timing of sleep and sleep deprivation can have a profound influence on this rhythmicity in the peripheral transcriptome (for a recent review, see Archer and Oster, 2015). In mice, sleep deprivation can lead to an 80% reduction in rhythmic transcripts in the brain (Maret et al., 2007). When sleep occurs at night, when temperature is low and melatonin is high, 6.4% of the human blood transcriptome is rhythmic. "
    [Show abstract] [Hide abstract] ABSTRACT: Although the functions of sleep remain to be fully elucidated, it is clear that there are far-reaching effects of its disruption, whether by curtailment for a single night, by a few hours each night over a long period, or by disruption in sleep continuity. Epidemiological and experimental studies of these different forms of sleep disruption show deranged physiology from subcellular levels to complex affective behavior. In keeping with the multifaceted influence of sleep on health and well-being, we illustrate how the duration of sleep, its timing, and continuity can affect cellular ultrastructure, gene expression, metabolic and hormone regulation, mood, and vigilance. Recent brain imaging studies provide some clues on mechanisms underlying the most common cause of disrupted sleep (insomnia). These insights should ultimately result in adequate interventions to prevent and treat sleep disruption because of their high relevance to our most prevalent health problems.
    Full-text · Article · Oct 2015 · The Journal of Neuroscience : The Official Journal of the Society for Neuroscience
  • Source
    • "The same categories are enriched during sleep in oligodendrocytes, and the rate of proliferation of oligodendrocyte precursors doubles in sleep relative to wake [37] . Cirp, coding for cold inducible RNA binding protein, is also upregulated by sleep in astrocytes and oligodendrocytes [37] , and was identified as a sleep gene in previous studies that pooled transcripts from all brain cells414243. CIRP is required for high-amplitude circadian gene expression in fibroblasts, but daily oscillations in Cirp mRNA levels are driven by changes in temperature, not by the circadian clock [44]. "
    [Show abstract] [Hide abstract] ABSTRACT: Astrocytes can mediate neurovascular coupling, modulate neuronal excitability, and promote synaptic maturation and remodeling. All these functions are likely to be modulated by the sleep/wake cycle, because brain metabolism, neuronal activity and synaptic turnover change as a function of behavioral state. Yet, little is known about the effects of sleep and wake on astrocytes. Here we show that sleep and wake strongly affect both astrocytic gene expression and ultrastructure in the mouse brain. Using translating ribosome affinity purification technology and microarrays, we find that 1.4 % of all astrocytic transcripts in the forebrain are dependent on state (three groups, sleep, wake, short sleep deprivation; six mice per group). Sleep upregulates a few select genes, like Cirp and Uba1, whereas wake upregulates many genes related to metabolism, the extracellular matrix and cytoskeleton, including Trio, Synj2 and Gem, which are involved in the elongation of peripheral astrocytic processes. Using serial block face scanning electron microscopy (three groups, sleep, short sleep deprivation, chronic sleep restriction; three mice per group, >100 spines per mouse, 3D), we find that a few hours of wake are sufficient to bring astrocytic processes closer to the synaptic cleft, while chronic sleep restriction also extends the overall astrocytic coverage of the synapse, including at the axon-spine interface, and increases the available astrocytic surface in the neuropil. Wake-related changes likely reflect an increased need for glutamate clearance, and are consistent with an overall increase in synaptic strength when sleep is prevented. The reduced astrocytic coverage during sleep, instead, may favor glutamate spillover, thus promoting neuronal synchronization during non-rapid eye movement sleep.
    Full-text · Article · Aug 2015 · BMC Biology
  • Source
    • "Unfortunately, these effects have not yet translated into robust responses in clinical trials (Boyle et al. 2012; Wesensten et al. 2007 ), possibly due to lack of a direct measure of target engagement to inform human dosing regimens. Beyond glutamate , interactions of the mGlu5 receptor with other mechanisms postulated to play a role in sleep-wake regulation may also be important, for example Homer 1a (Ango et al. 2001; Ango et al. 2002; Maret et al. 2007 ) or adenosinergic transmission (Bachmann et al. 2012; Bodenmann et al. 2012; Gallopin et al. 2005; Okada et al. 2003). Finally, a recent PET imaging study in humans using the selective radioligand 11 C-ABP688 has demonstrated that mGlu5 receptor availability is increased in several brain regions after one night of sleep deprivation (Hefti et al. 2013), which suggests that dynamic changes in mGlu5 receptor expression may play a fundamental role in sleep/wake homeostasis. "
    [Show abstract] [Hide abstract] ABSTRACT: While treatment options are available, excessive daytime sleepiness (EDS) remains a significant unmet medical need for many patients. Relatively little rodent behavioural pharmacology has been conducted in this context to assess potential pro-vigilant compounds for their ability to restore functional capacity following experimentally induced sleep loss. Male Wistar rats were prepared for electroencephalographic (EEG) recording and subject to 11 h of sleep restriction using a biofeedback-induced cage rotation protocol. A simple response latency task (SRLT) was used to behaviourally index sleep restriction and the effects of pro-vigilant compounds: modafinil, d-amphetamine, caffeine, and the mGlu5-positive allosteric modulator LSN2814617. Sleep restriction resulted in a consistent, quantified loss of non-rapid eye movement (NREM) and REM sleep that impaired SRLT performance in a manner suggestive of progressive task disengagement. In terms of EEG parameters, all compounds induced wakefulness. Amphetamine treatment further decreased SRLT performance capacity, whereas the other three compounds decreased omissions and allowed animals to re-engage in the task. Caffeine and modafinil also significantly increased premature responses during this period, an effect not observed for LSN2814617. While all compounds caused compensatory sleep responses, the magnitude of compensation observed for LSN2814617 was much smaller than would be predicted to result from the prolongation of wakefulness exhibited. Using simple response latencies to index performance, an mGlu5 PAM dramatically increased wakefulness and improved functional capacity of sleep-restricted animals, without eliciting a proportionate compensatory sleep response. This effect was qualitatively distinct from that of amphetamine, caffeine and modafinil.
    Full-text · Article · Apr 2015 · Psychopharmacology
Show more