Abstract

Prolonged rapid-eye-movement (REM) sleep deprivation has long been used to study the role of REM sleep in learning and memory processes. However, this method potentially induces stress and fatigue that may directly affect cognitive functions. Here, by using a short-term and nonstressful REM sleep deprivation (RSD) method we assessed in rats the bidirectional influence of reduced and increased REM sleep amount on hippocampal-dependent emotional memory and plasticity. Our results indicate that 4 h RSD impaired consolidation of contextual fear conditioning (CFC) and induction of long-term potentiation (LTP), while decreasing density of Egr1/Zif268-expressing neurons in the CA1 region of the dorsal hippocampus. LTP and Egr1 expression were not affected in ventral CA1. Conversely, an increase in REM sleep restores and further facilitates CFC consolidation and LTP induction, and also increases Egr1 expression in dorsal CA1. Moreover, CFC consolidation, Egr1 neuron density, and LTP amplitude in dorsal CA1 show a positive correlation with REM sleep amount. Altogether, these results indicate that mild changes in REM sleep amount bidirectionally affect memory and synaptic plasticity mechanisms occurring in the CA1 area of the dorsal hippocampus.

Introduction

Mammalian brain activity during sleep is characterized by the alternation of 2 distinct states, rapid-eye-movement (REM), or paradoxical sleep, and non-REM (NREM), or slow-wave sleep. Although total sleep loss has been repeatedly reported to impair the acquisition (Van Der Werf et al. 2009; Abel et al. 2013) and the consolidation stages of hippocampus-dependent memories (Graves et al. 2003; Vecsey et al. 2009; Hagewoud et al. 2010; Inostroza and Born 2013; Prince et al. 2014) little is known about the specific contribution of REM sleep in hippocampal function. It has long been hypothesized that REM sleep promotes learning and memory by facilitating consolidation of newly acquired information into long-term storage (Maquet 2001; Siegel 2001; Hobson and Pace-Schott 2002). In rodents, evidence for this hypothesis is based on early physiological observations and more recent findings suggesting a REM sleep-dependent modulation of the cellular and molecular mechanisms underlying forms of synaptic plasticity such as long-term potentiation (LTP) (Benington and Frank 2003). REM sleep amount has been shown to be transiently increased within a short time window after training in various memory tasks, during which consolidation is sensitive to REM sleep loss (Smith and Lapp 1986; Smith and Rose 1996; Datta et al. 2004). Additionally, expression of the immediate-early gene Egr1 (also known as Zif268, Krox-24, NGF1-A, Zenk), an indirect marker of neuronal activity and synaptic plasticity (Veyrac et al. 2014), was augmented in hippocampal and neocortical neurons when exploration of a novel environment or the induction of hippocampal LTP were followed by REM sleep (Ribeiro et al. 1999, 2002). Intriguingly, replay of activity patterns has been observed in hippocampal neurons during REM sleep episodes following spatial learning (Louie and Wilson 2001), although similar features also occur during NREM sleep and awake states (Ji and Wilson 2007; Karlsson and Frank 2009). Finally, it has been recently shown that REM sleep contributes to homeostatic regulation of hippocampal neuronal activity (Grosmark et al. 2012). All these studies point toward a specific function for REM sleep in facilitating hippocampal memory and plasticity.

To reveal a causal link between REM sleep and memory, we designed an experimental strategy to assess the impact of a mild modulation in REM sleep amount on hippocampal-dependent forms of memory and synaptic plasticity. We hypothesized that an increase in REM sleep quantity should facilitate memory performance and synaptic plasticity whereas a decrease in REM sleep amount should reduce memory performance and plasticity. By this bidirectional REM sleep modulation, we will be able to determine, for the first time to our knowledge, whether REM sleep is able to regulate memory encoding, consolidation, and synaptic plasticity. REM sleep restriction was induced by REM sleep deprivation (RSD) and REM sleep increase was obtained by a REM sleep rebound (RSR). As REM sleep is under homeostatic regulation, RSR was obtained from recovering of prior RSD.

Prolonged RSD have been extensively used by previous studies to determine the role of this sleep state (McDermott et al. 2003; Wetzel et al. 2003; Alvarenga et al. 2008; Ravassard et al. 2009; Pinho et al. 2013). However, methods inducing such RSD are also known to induce stress (Rechtschaffen et al. 1999; Vertes 2004) as well as unspecific effects on NREM sleep (Verret et al. 2006; Ravassard et al. 2009). The specific contribution of REM sleep still remains elusive due to these confounding factors. Methods aimed to discard nonspecific effects by using mild and short protocols have been developed to address the impairment of hippocampal function by total sleep loss (Romcy-Pereira and Pavlides 2004; Kopp et al. 2006; Romcy-Pereira et al. 2009; Vecsey et al. 2009; Prince et al. 2014) or REM sleep loss (Datta et al. 2004; Romcy-Pereira and Pavlides 2004). Supplementary advantages of short deprivations are the ability to assess the effects of more ecological sleep loss, to test whether these effects are reversible, and to temporally target different stages of the memory process using a short time window.

We assessed encoding and consolidation stages of hippocampal-dependent contextual fear conditioning (CFC), and CA1 synaptic plasticity using Egr1 expression and LTP induction after a mild and short RSD (4 h) and RSR. Our results showed that in the absence of confounding factors, CFC consolidation and CA1 LTP are bidirectionally regulated by modest changes in REM sleep quantity.

Materials and Methods

Animals

Male Sprague-Dawley rats (n = 129) were purchased at age P21 (45–55 g) from Charles River Laboratories (Le Genest St. Isle, France). Rats were housed individually in recording barrels and kept on a 12 h light–dark cycle with lights-on at Zeitgeber 0 (8:00 AM). Room temperature was maintained at 21 ± 1°C, and standard rodent food and water were available ad libitum throughout all experiments. All procedures were approved by the institutional animal care and use committee of the University of Lyon 1 (protocols BH2006-09 and BH2006-10) and were conducted in accordance with the French and European Community guidelines for the use of research animals.

Surgery and Polygraphic Recordings

All animals used in this study were implanted for polygraphic recordings. Surgery was performed under anesthesia provided by an i.p. injection of chloral hydrate (320 mg/kg). Atropine sulfate (0.4 mg/kg) was administered to minimize respiratory distress. Rimadyl and xylocain (5 and 10 mg/kg) were provided before incision for analgesia. The skull was exposed, and burr holes were made for the insertion of silver plated wires (30 AWG, Kynar) with 1 mm of insulation stripped off for subdural EEG recordings. Electrodes were located over the left prefrontal cortex (AP +1.2; ML −1.2), the left parietal cortex (AP −3.7; ML −3.0), the right parietal cortex (AP −1.7; ML +1.2), and the reference over the cerebellum (AP, −9.0; ML, 0.0). Electromyogram (EMG) activity was assessed from the dorsal neck muscles by 2 embedded wires tipped with a ∼1 mm diameter tin sphere. Electrodes were linked to a miniature connector cemented to the skull. After a 4-day recovery period, animals were placed in a sound-attenuated, ventilated, and electrically isolated chamber (40 × 40 × 60 cm) and connected by a tether to a rotating connector (Plastics One Inc., CT) for a 2-day habituation period. EEG/EMG activity was amplified (500×), filtered (1–100 Hz for EEG and 10–100 Hz for EMG, Alpha-Omega, Israel) and recorded continuously using a CED Power1401 converter and Spike2 software (Cambridge Electronic Design, UK). The signals were sampled at 512 Hz. Data were stored on a computer for off-line analysis. The vigilance states were scored in 5 s epochs. Double-blind off-line analysis was independently carried out using the following criteria: for wakefulness (Wk), low voltage/fast cortical EEG and high amplitude EMG; for NREM sleep, high voltage (>200 µV) slow-wave EEG (1–5 Hz) and low amplitude EMG; and for REM sleep, low voltage and predominant theta frequency (5–10 Hz) with an absence of muscle tone. Spectral analysis (Fast Fourier Transform, FFT) was performed using 5-s time windows on raw parietal EEG signals from identified REM and NREM sleep episodes of substantial length (>10 s). FFTs were then normalized by their integral and averaged by animals and conditions.

REM Sleep Deprivation and Rebound

Baseline vigilance states were recorded for at least 24 h prior to sleep manipulations. RSD was performed on rats starting at light-on (Zeitgeber 0 set at 8:00 AM—except if indicated otherwise) for 4 h by continuously monitoring polygraphic recordings online. When REM sleep was detected for one or two 5 s windows, the cage was manually slightly inclined to wake up the rat, and the time of each intervention was recorded. RSR was elicited in rats during the 2 h 30 min following this 4 h RSD. Control (RSC) rats remained undisturbed for 4 h.

Hippocampal Slice Electrophysiology

A subset of RSC, RSD, and RSR animals (n = 52) were used for electrophysiology and immunochemistry. Immediately following the respective sleep manipulations described above, they were euthanized with pentobarbital i.p. injection (150 mg/kg, Sigma). Under anesthesia, the heart was exposed, cardiac blood was withdrawn in a subset of animals (n = 20) as described below for corticosterone assay, then animals were perfused with cold, oxygenated artificial cerebrospinal fluid (ACSF for perfusion and dissection, in mM: 85 NaCl, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 4 MgCl2, 0.5 CaCl2, 25 Glucose, and 75 Sucrose). After brain extraction, at least one hemisphere was used for electrophysiology experiments (n = 52) while a subset of animals had their second hemisphere used for immunochemistry staining (n = 17, see below). Coronal slices (350 μm thick) containing dorsal and ventral hippocampus were obtained with a Vibratome (Leica VT1000S). Brain slices were then left to recover for 1 h at room temperature in incubation ACSF (124 NaCl, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 4 MgCl2, 1 CaCl2, and 25 Glucose). Recordings were performed in a submerged chamber perfused with ACSF (124 NaCl, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 1.3 MgCl2, 2.5 CaCl2, and 10 Glucose) maintained at 32°C and constantly gassed with 95% O2–5% CO2. A cut was applied between CA3 and CA1 to prevent epileptic discharges before picrotoxin (GABA-A receptor antagonist, PTX, 100 μM) was added to the ACSF. Orthodromic stimulations were delivered by bipolar tungsten electrodes, and field excitatory postsynaptic potentials (fEPSPs) were recorded with an ACSF-filled glass microelectrode (∼1 MΩ). Stimulations (100 μs duration) were applied at a frequency of 0.1 Hz via a stimulus isolator (World Precision Instruments). Stimulating and recording electrodes were placed in the middle of CA1 stratum radiatum and were separated by 100–200 μm. The fEPSP were carefully adjusted to obtain a similar amplitude in each hippocampal slice (fEPSP amplitude in dorsal hippocampus from 10 RSC, 11 RSD, and 12 RSR slices, in µV ± SD: 112 ± 35, 121 ± 22, 115 ± 31; in ventral hippocampus from 12 RSC, 14 RSD, and 11 RSR slices: 111 ± 23, 113 ± 21, 119 ± 31; P > 0.4 between groups in both cases). Input/output (I/O) curves were assessed by progressively raising the stimulation intensity every 3–5 data points from 40 to 200 μA. The input is given by the amplitude of the fiber volley (FV, Schaffer collateral presynaptic action potentials) that precedes the postsynaptic response, whereas the output is given by the amplitude or initial slope of the fEPSP. Paired-pulse facilitation (PPF) was tested by evoking fEPSP with 2 stimulations separated by 10, 50, or 100-ms intervals for 5 min each. PPF ratio was calculated using the initial fEPSP slope. LTP induction protocols consisted in a high-frequency train of either 10 or 100 Hz stimulation for 90 s and 1 s, respectively. LTP amplitude (measured using the initial fEPSP slope) was normalized with the average fEPSPs recorded during the baseline period (−10 to 0 min before LTP induction).

Immunohistochemistry

Brain hemispheres extracted from RSC, RSD, and RSR animals following perfusion as described above (n = 17) were submerged in a fixative solution (4% paraformaldehyde in 0.1 M phosphate buffer). Twenty micrometer thick coronal sections containing the hippocampus were obtained with a cryostat and immerged in ammonium chloride solution (50 mM NH4Cl, 15 min) to reduce endogenous fluorescence, then incubated with Egr1 (1:250, Santa Cruz) rabbit polyclonal antibody for 72 h at 4°C with constant rotation. Sections were then washed and incubated with donkey anti-rabbit AlexaFluor-488 IgG (1:500, Invitrogen) at room temperature for 2 h. Sections were then rinsed in PBS, mounted, and visualized under a fluorescent microscope equipped with a digital camera and an image analyzing system (Soft Imaging System). Egr1-positive neurons were quantified double-blind using contrast threshold analysis on every 3–5 or 2–3 adjacent sections from the dorsal and ventral hippocampus, respectively.

Corticosterone Radioimmunoassay

Cardiac blood (∼400 µl) was collected immediately following the end of control sleep (RSC), RSD, and RSR procedures using heparinized syringes in a subset of rats (n = 20). Samples were centrifuged at 2000 rpm for 15 min at 4°C, and the plasma fraction was isolated and stored at –80°C. Samples were blindly analyzed by radioimmunoassay at the Centre de Médecine Nucléaire (HCL, Lyon) directed by Pr. Bruno Claustrat.

Behavior

CFC was carried out on 77 rats following methods described elsewhere (Shumyatsky et al. 2005; Saxe et al. 2007). For training, the rats were placed in a conditioning chamber (40 × 40 × 60 cm, custom-made) that they were allowed to freely explore during 4 min prior to the onset of the unconditioned stimulus (US), a 0.3 mA, 2-s electric footshock. This low intensity footshock was chosen in order to reduce the amount of stress in our subjects. After an additional 2 min, the rats were returned to their home cage. CFC was tested either 1 or 24 h after training to test acquisition or consolidation, respectively. During testing, the rats were reexposed to the conditioning chamber (context) for 6 consecutive minutes without footshock delivery. Freezing behavior, defined as a complete absence of movement, was assessed pre-, post-training and during testing by a custom video-based software (camera specification: definition 640 × 480 pixels at 15 frames per second, and video resolution of 11.5 pixels/cm, custom software were developed under Matlab® from the Mathworks company). The animal actimetry was estimated by calculating the number of pixels whose intensity changed more than 10 gray levels between 2 successive frames. A freezing epoch was automatically scored when the actimetry was below the threshold of 150 changing pixels/sec. Video actimetry was performed with a side view for better detection. The conditioning chamber was cleaned after each training/testing session using 70% ethanol.

Statistical Analysis

Statistical analyses were performed using Origin software. Comparisons with multiple groups were initially analyzed with a nonparametric Kruskal–Wallis test followed by post hoc Mann–Whitney test for individual comparisons. Intragroup comparisons were analyzed with a nonparametric paired sample Wilcoxon signed rank test. Error bars on the figures correspond to standard error of the mean (SEM) and n values correspond to the number of animals for all experiments. Significant pairwise differences between groups are illustrated in figures by an asterisk indicating the significance level (* for a P-value P < 0.05, ** for P < 0.01, and *** for P < 0.001 or lower) centered between the 2 experimental groups for which the comparison was made.

Results

Short and Gentle RSD Induces a Selective REM Sleep Deprivation and Rebound

We first tested whether our gentle deprivation method was efficient to produce a selective REM sleep reduction. We monitored EEG/EMG signals for 4 h, and when REM sleep episodes were detected for more than one 5-s analysis window, the home cage was slightly tilted to awaken the animal (Fig. 1A). RSD rats showed a significant decrease in REM sleep quantity during the 4-h deprivation as compared with undisturbed control (RSC) (P < 0.0001, n = 29 in each group, Fig. 1B). This REM sleep reduction was characterized by a decrease in the duration of REM sleep episodes (P < 0.0001, Fig. 1C), while their number was comparable (P > 0.5, Fig. 1D). Similar NREM sleep amount (P > 0.5, Fig. 1B), episode duration (P > 0.5, Fig. 1C) and episode number (P > 0.5, Fig. 1D) were obtained in both groups, indicating that the RSD method was selective of REM sleep. Furthermore, power spectrum analysis showed no difference during NREM sleep episodes in RSD animals compared with controls (P > 0.5 at delta and spindle frequencies, Fig. 1F). In the third experimental group, we studied the effects of RSR, which was achieved by allowing the animals to rest for 2 h 30 min in their home cage following a 4-h REM sleep deprivation comparable to the RSD group. RSR animals showed a significant increase in REM sleep quantity (also termed rebound) as compared with both RSD and RSC groups (P < 0.0001, RSR: n = 20, Fig. 1B). This rebound in RSR animals was also specific for REM sleep as NREM sleep quantity and NREM sleep power spectrum analysis in this group were similar to those of the other groups (Fig. 1BD,F). REM sleep rebound was characterized by longer REM sleep episodes duration (P < 0.001 compared with RSC and P < 0.0001 compared with RSD, Fig. 1C and Supplementary Fig. 1), but unchanged number of episodes (P > 0.5, Fig. 1D). In addition, the power spectrum analysis performed on the parietal EEG recording of RSR animals showed no difference between RSC and RSR groups, in particular in the theta band (5–10 Hz, P > 0.5, Fig. 1E), showing that the rebound produced no change in theta oscillation during REM sleep, a result consistent with a previous study using longer RSD (Maloney et al. 1999). Our results thus indicate that REM sleep homeostasis after 4-h RSD elicited longer REM sleep episodes without changing its key electrophysiological features. We also found that the number of interventions during deprivation increased throughout the 4-h RSD (2.9 ± 0.8 interventions during the first hour, against 13 ± 1.8 interventions during the fourth hour, P < 0.0001, n = 33 rats), but their total number were not significantly different from the number of REM sleep episodes in RSC animals during the same period (34.7 ± 2.4 interventions in RSD, n = 29 and RSR, n = 20 animals, compared with 29.2 ± 1.7 REM sleep episodes in RSC animals, P > 0.05), suggesting only a mild increase in REM sleep pressure. Finally, since stress can have a deleterious effect on LTP (Kim and Diamond 2002; Yang et al. 2004), we assessed the plasma corticosterone level as an indicator of stress potentially induced during RSD and found comparably low concentrations for all 3 groups (P > 0.5, Fig. 1G), indicative of a nonstressful RSD procedure.

Figure 1.

Short, selective and nonstressful REM sleep deprivation induces a 2-fold increase .in REM sleep rebound. (A) Schema (left panel) and representative polygraphic recordings (right panel) illustrating the REM sleep deprivation method achieved by terminating REM sleep after detection of 5–10 s episodes (black arrows signal intervention times on the recordings) with a slight inclination of the home cage as represented by the schema. Scale bars: 2 mV, 10 s. (B) Quantification of the vigilance states (in percentage of recording time) for REM sleep deprived (RSD, white bars), control (RSC, gray bars), and REM sleep rebound (RSR, black bars) groups. A significant decrease in REM sleep amount was observed in RSD compared with RSC (2.6 ± 0.3% vs. 9.5 ± 0.6%, P < 0.0001, n = 29 in each group). In contrast, a significant increase in REM sleep amount was observed in RSR (17.9 ± 1.1%, n = 20, P < 0.0001). No significant difference in NREM sleep amount was found between groups. (C) Mean sleep episode duration. REM sleep deprivation selectively decreased REM, but not NREM, sleep episode duration (for REM sleep, RSD: 9.7 ± 0.6 s, RSC: 30.2 ± 3.1 s, P < 0.0001; for NREM sleep, RSD: 52.7 ± 3.3 s, RSC: 55.2 ± 2.5 s, P > 0.05). REM sleep episode duration was significantly increased in RSR group compared with RSD and RSC groups (RSR: 50.3 ± 3.9 s, RSC vs. RSR P < 0.001; RSD vs. RSR: P < 0.0001). (D) Mean sleep episode number per hour. Same REM and NREM sleep episode number was observed in each group of rats. (E) Spectral analysis (fast Fourier transform) performed on parietal EEG during REM sleep episodes showed similar power spectra for RSR and RSC groups (n = 29 RSC and 20 RSR rats). (F) Spectral analysis (Fast Fourier Transform) performed on parietal EEG during NREM sleep episodes showed almost identical power spectra for RSC, RSD, and RSR groups (n = 29 RSD, 29 RSC, and 20 RSR rats). (G) Blood plasma corticosterone levels showed no significant difference between the 3 groups (n = 6 RSD, 8 RSC, and 6 RSR animals).

Figure 1.

Short, selective and nonstressful REM sleep deprivation induces a 2-fold increase .in REM sleep rebound. (A) Schema (left panel) and representative polygraphic recordings (right panel) illustrating the REM sleep deprivation method achieved by terminating REM sleep after detection of 5–10 s episodes (black arrows signal intervention times on the recordings) with a slight inclination of the home cage as represented by the schema. Scale bars: 2 mV, 10 s. (B) Quantification of the vigilance states (in percentage of recording time) for REM sleep deprived (RSD, white bars), control (RSC, gray bars), and REM sleep rebound (RSR, black bars) groups. A significant decrease in REM sleep amount was observed in RSD compared with RSC (2.6 ± 0.3% vs. 9.5 ± 0.6%, P < 0.0001, n = 29 in each group). In contrast, a significant increase in REM sleep amount was observed in RSR (17.9 ± 1.1%, n = 20, P < 0.0001). No significant difference in NREM sleep amount was found between groups. (C) Mean sleep episode duration. REM sleep deprivation selectively decreased REM, but not NREM, sleep episode duration (for REM sleep, RSD: 9.7 ± 0.6 s, RSC: 30.2 ± 3.1 s, P < 0.0001; for NREM sleep, RSD: 52.7 ± 3.3 s, RSC: 55.2 ± 2.5 s, P > 0.05). REM sleep episode duration was significantly increased in RSR group compared with RSD and RSC groups (RSR: 50.3 ± 3.9 s, RSC vs. RSR P < 0.001; RSD vs. RSR: P < 0.0001). (D) Mean sleep episode number per hour. Same REM and NREM sleep episode number was observed in each group of rats. (E) Spectral analysis (fast Fourier transform) performed on parietal EEG during REM sleep episodes showed similar power spectra for RSR and RSC groups (n = 29 RSC and 20 RSR rats). (F) Spectral analysis (Fast Fourier Transform) performed on parietal EEG during NREM sleep episodes showed almost identical power spectra for RSC, RSD, and RSR groups (n = 29 RSD, 29 RSC, and 20 RSR rats). (G) Blood plasma corticosterone levels showed no significant difference between the 3 groups (n = 6 RSD, 8 RSC, and 6 RSR animals).

Egr1 Expression is Correlated with REM Sleep Quantity

We sought to determine how REM sleep could regulate plasticity at the cellular level. We assessed by immunofluorescence the expression of the immediate early gene (IEG) Egr1 (also known as Zif268 or Krox-24), given its role in long-term plasticity and memory (Bozon et al. 2002). Figure 2A illustrates representative distributions of the neuronal populations expressing Egr1 in the CA1 area of the dorsal hippocampus from animals of all 3 groups. The density of Egr1-positive neurons was significantly decreased in RSD rats as compared with the RSC group and significantly increased in RSR animals as compared with both RSC and RSD rats (P < 0.01, RSD and RSC: n = 6, RSR: n = 5, Fig. 2B). We also found a highly positive correlation between the density of Egr1-positive neurons and the amount of REM sleep (r = 0.88, P < 0.0001, n = 17, Fig. 2C), but not NREM sleep (r = 0.2, P > 0.05, n = 17, not shown). Conversely, a negative correlation was observed when wakefulness (Wk) quantity was considered (r = −0.53, P < 0.03, n = 17). In contrast, in the CA1 area of the ventral hippocampus, the density of Egr1-positive cells was not different between RSC and RSD groups (P > 0.05, RSD and RSC: n = 6 in both groups). Thus, Egr1 mapping in the dorsal and ventral hippocampus showed different profiles in the RSD group. This suggests that the lack of REM sleep differentially affected synaptic plasticity across the dorso-ventral axis of the hippocampus. However, REM sleep rebound could enhance plasticity in the whole hippocampus as RSR rats showed a dramatic increase in Egr1-positive cell density in the CA1 area of the ventral hippocampus (Fig. 2B). Furthermore, a positive correlation was also observed between the density of Egr1-positive cells in the ventral CA1 area and REM sleep amount (r = 0.80, P < 0.001, n = 16) but not with the NREM sleep (r = 0.23, P > 0.05) and the Wk quantity (r = −0.49, P > 0.05). Finally, to determine whether stress could modulate the expression of Egr1 in our conditions, we calculated the correlation coefficient between the density of Egr1-positive neurons in dorsal or ventral CA1 and the corticosterone level measured in each animal across conditions, and found no significant correlation (r = 0.1, P > 0.05, n = 17 in dorsal, r = 0.2, P > 0.05, n = 16 in ventral).

Figure 2.

REM sleep amount regulates the density of Egr1-positive neurons in the CA1 area of the hippocampus. (A) Microphotographs depicting representative labeling of Egr1-positive neurons in the CA1 area of the dorsal hippocampus in RSD, RSC and RSR rats. sp, stratum pyramidale; so, stratum oriens; sr, stratum radiatum. Scale bar: 100 µm. (B) Quantification of the density of Egr1 pyramidal cells in the CA1 area of the dorsal and ventral hippocampus for each group. The mean density for each animal was normalized to the control group. In the CA1 area of the dorsal hippocampus, the density of Egr1-positive cells was significantly lower in RSD compared with RSC (70.8 ± 3.6% vs. 100 ± 1.3%, n = 6 in each group, P < 0.005). Strikingly, the density was significantly stronger in RSR as compared with both RSC and RSD rats (127.1 ± 6.2%, n = 5, P < 0.01 for both comparisons). In contrast, in the CA1 area of the ventral hippocampus, the density of Egr1-positive cells was not different between RSD and RSC (RSD: 93.1 ± 18.5 vs. RSC: 100 ± 20.8%, P > 0.05). In ventral CA1 of RSR rats, a significant increase in the density of Egr1-positive cells was found compared with RSC and RSD rats (RSR: 204.5 ± 6.0%, P < 0.001). (C) A highly positive correlation was observed between the density of Egr1-positive cells in the CA1 area of the dorsal hippocampus and the amount of REM sleep (r = 0.88, P < 0.0001, n = 17).

Figure 2.

REM sleep amount regulates the density of Egr1-positive neurons in the CA1 area of the hippocampus. (A) Microphotographs depicting representative labeling of Egr1-positive neurons in the CA1 area of the dorsal hippocampus in RSD, RSC and RSR rats. sp, stratum pyramidale; so, stratum oriens; sr, stratum radiatum. Scale bar: 100 µm. (B) Quantification of the density of Egr1 pyramidal cells in the CA1 area of the dorsal and ventral hippocampus for each group. The mean density for each animal was normalized to the control group. In the CA1 area of the dorsal hippocampus, the density of Egr1-positive cells was significantly lower in RSD compared with RSC (70.8 ± 3.6% vs. 100 ± 1.3%, n = 6 in each group, P < 0.005). Strikingly, the density was significantly stronger in RSR as compared with both RSC and RSD rats (127.1 ± 6.2%, n = 5, P < 0.01 for both comparisons). In contrast, in the CA1 area of the ventral hippocampus, the density of Egr1-positive cells was not different between RSD and RSC (RSD: 93.1 ± 18.5 vs. RSC: 100 ± 20.8%, P > 0.05). In ventral CA1 of RSR rats, a significant increase in the density of Egr1-positive cells was found compared with RSC and RSD rats (RSR: 204.5 ± 6.0%, P < 0.001). (C) A highly positive correlation was observed between the density of Egr1-positive cells in the CA1 area of the dorsal hippocampus and the amount of REM sleep (r = 0.88, P < 0.0001, n = 17).

LTP is Bidirectionally Modulated by Short REM Sleep Deprivation and Rebound

Given the decreased expression of Egr1 in the dorsal hippocampus after a short REM sleep deprivation, we then tested whether synaptic plasticity could be differentially impaired by RSD when inducing LTP in the CA1 field of slices from the dorsal or ventral hippocampus (Fig. 3A). We used a mild LTP-induction protocol (10 Hz, 900 pulses) mimicking theta activity in the hippocampus (Hu et al. 2007). As shown in Figure 3B, this 10-Hz induction protocol elicited a modest, yet significant, LTP in the RSC group in the dorsal hippocampus. In contrast, LTP induction was impaired in the RSD group in the dorsal hippocampus (P < 0.05, n = 8 in each group, Fig. 3B,D), and was significantly increased in the RSR group as compared with RSC rats (P < 0.05, RSR: n = 8, Fig. 3B,D). Such an effect was absent in the ventral hippocampus (Fig. 3D). Could a stronger LTP induction protocol elicit a normal LTP in RSD animals? We tested a 100-Hz induction (1 s) protocol to address this question (Fig. 3C). Using this protocol, we found that LTP amplitude in the RSD group was still significantly lower than in the RSC group in the dorsal hippocampus (P < 0.01, n = 8 in each group, Fig. 3C,D). However, no significant difference in LTP amplitude was observed between RSR and RSC rats (P > 0.05, RSR: n = 8, Fig. 3C,D). No significant change between groups was found in the ventral hippocampus (P > 0.05, n = 6 in each group, Fig. 3D). Given the increased LTP amplitude observed in the dorsal hippocampus of RSR rats, we performed a linear regression between REM sleep quantity and LTP amplitude induced at 10 Hz (Fig. 3E). We found a positive correlation between LTP amplitude in dorsal CA1 (expressed as the percentage of fEPSP slope change from baseline) and REM sleep amounts (r = 0.61, P < 0.003, n = 22, Fig. 3E). In contrast, no correlation was observed between LTP amplitude in dorsal CA1 and NREM sleep amounts (r = 0.3, P > 0.05, n = 22) or between LTP amplitude in ventral CA1 and REM sleep amount (r = 0.12, P > 0.05, n = 17). This result suggests that REM sleep amount could regulate the threshold for LTP induction selectively in the dorsal hippocampus. Such a correlation was not observed with the 100 Hz induction protocol (r = 0.36, P > 0.05, n = 22), indicating that this strong induction protocol might have overshadowed an LTP that naturally occurred during rebound, leading to LTP saturation in slices from RSR animals. Before LTP induction, baseline fEPSP amplitude was comparable across groups (see Methods). Additionally, I/O curves (Fig. 3F,G), as well as PPF measured at several interpulse intervals showed no difference between conditions in the dorsal and ventral hippocampus (in the dorsal and ventral hippocampus, between RSD and RSC: P > 0.05 and between RSR and RSC: P > 0.05, n = 7 in each group, data not shown). These results suggest that RSD did not alter cooperativity and short-term plasticity, but specifically impaired LTP induction in the dorsal hippocampus. Last, we compared the LTP amplitude induced at 100 Hz to the corticosterone level measured in a subset of animals and found no correlation (r = 0.08, P = 0.79, n = 12 in dorsal CA1, r = −0.12, P = 0.73, n = 10 in ventral CA1), discarding a direct contribution of stress in LTP amplitude in our experimental conditions.

Figure 3.

REM sleep amounts modulate LTP in the CA1 region of the dorsal hippocampus. (A) Coronal cerebral slices representing the position of recording and stimulating electrodes in the dorsal (top) and ventral (down) hippocampus to elicit CA3-to-CA1 LTP. (B) Time course of 10-Hz LTP in RSD, RSC, and RSR groups in dorsal CA1. 10-Hz LTP was impaired by short-term RSD as compared with controls. In contrast, RSR increased the amplitude of CA1 LTP compared with controls. Insert: sample average fEPSP traces 5–0 min before induction protocol (gray line) and 55–60 min after (black line). Scale bars: 0.2 mV, 10 ms. (C) Time course of 100-Hz LTP in RSD, RSC, and RSR groups in dorsal CA1. One hundred hertz LTP was altered in RSD rats compared with controls. One hundred hertz LTP was restored in RSR group. Insert: sample average fEPSP traces 5–0 min before induction protocol (gray line) and 55–60 min after (black line). Scale bars: 0.2 mV, 10 ms. (D) Bar graphs showing the LTP amplitude at 55–60 min from 10-Hz and 100-Hz induction protocols for each group in the CA1 area of the dorsal and ventral hippocampus. In the dorsal hippocampus, 10-Hz LTP was significantly lower in RSD group than in RSC (105.4 ± 8.1% vs. 123.9 ± 3.4%, P < 0.05, n = 8 in each group) and significantly higher in RSR group than in RSC (147.8 ± 9.5%, RSR vs. RSC P < 0.05, n = 8). One hundred hertz LTP in the dorsal hippocampus was significantly lower in RSD group than in RSC (183.4 ± 19.7% vs. 124.8 ± 8.1%, P < 0.01, n = 8 in each group) but not significantly higher in RSR group than in RSC (161.8 ± 13.9%, RSR vs. RSC, P > 0.05, n = 8). In the ventral CA1, no difference was found between groups at 10-Hz and 100-Hz LTP (P > 0.05, n = 6 in each group). (E) In the CA1 area of the dorsal hippocampus, a positive correlation was observed between REM sleep amount and 10-Hz LTP amplitude (r = 0.61, P < 0.003, n = 22). (F) Typical I/O curve obtained from slices of CA1 area of the dorsal hippocampus in each group. The fiber volley amplitude (FV, in mV) was computed against the initial slope of the field EPSP (in mV/ms) at varying stimulus intensity. The slope of the linear regression of the I/O curve gave an estimate of the synaptic efficacy (I/O slope index). Insert: superimposed fEPSP traces elicited at increasing stimulus intensities in an experiment in RSC group (star: stimulus artifact removal, a: fiber volley, b: fEPSP). Scale bars: 0.2 mV, 10 ms. (G) I/O slope index in the dorsal and ventral hippocampus was not different in the 3 conditions (n = 7 in each group).

Figure 3.

REM sleep amounts modulate LTP in the CA1 region of the dorsal hippocampus. (A) Coronal cerebral slices representing the position of recording and stimulating electrodes in the dorsal (top) and ventral (down) hippocampus to elicit CA3-to-CA1 LTP. (B) Time course of 10-Hz LTP in RSD, RSC, and RSR groups in dorsal CA1. 10-Hz LTP was impaired by short-term RSD as compared with controls. In contrast, RSR increased the amplitude of CA1 LTP compared with controls. Insert: sample average fEPSP traces 5–0 min before induction protocol (gray line) and 55–60 min after (black line). Scale bars: 0.2 mV, 10 ms. (C) Time course of 100-Hz LTP in RSD, RSC, and RSR groups in dorsal CA1. One hundred hertz LTP was altered in RSD rats compared with controls. One hundred hertz LTP was restored in RSR group. Insert: sample average fEPSP traces 5–0 min before induction protocol (gray line) and 55–60 min after (black line). Scale bars: 0.2 mV, 10 ms. (D) Bar graphs showing the LTP amplitude at 55–60 min from 10-Hz and 100-Hz induction protocols for each group in the CA1 area of the dorsal and ventral hippocampus. In the dorsal hippocampus, 10-Hz LTP was significantly lower in RSD group than in RSC (105.4 ± 8.1% vs. 123.9 ± 3.4%, P < 0.05, n = 8 in each group) and significantly higher in RSR group than in RSC (147.8 ± 9.5%, RSR vs. RSC P < 0.05, n = 8). One hundred hertz LTP in the dorsal hippocampus was significantly lower in RSD group than in RSC (183.4 ± 19.7% vs. 124.8 ± 8.1%, P < 0.01, n = 8 in each group) but not significantly higher in RSR group than in RSC (161.8 ± 13.9%, RSR vs. RSC, P > 0.05, n = 8). In the ventral CA1, no difference was found between groups at 10-Hz and 100-Hz LTP (P > 0.05, n = 6 in each group). (E) In the CA1 area of the dorsal hippocampus, a positive correlation was observed between REM sleep amount and 10-Hz LTP amplitude (r = 0.61, P < 0.003, n = 22). (F) Typical I/O curve obtained from slices of CA1 area of the dorsal hippocampus in each group. The fiber volley amplitude (FV, in mV) was computed against the initial slope of the field EPSP (in mV/ms) at varying stimulus intensity. The slope of the linear regression of the I/O curve gave an estimate of the synaptic efficacy (I/O slope index). Insert: superimposed fEPSP traces elicited at increasing stimulus intensities in an experiment in RSC group (star: stimulus artifact removal, a: fiber volley, b: fEPSP). Scale bars: 0.2 mV, 10 ms. (G) I/O slope index in the dorsal and ventral hippocampus was not different in the 3 conditions (n = 7 in each group).

Short REM Sleep Deprivation Impairs Consolidation of CFC

Given that RSD causes lower expression of Egr1 and impaired LTP in the hippocampus, what are the behavioral implications of this altered plasticity? To answer this question, we tested animals in a CFC task known to depend on both sleep and hippocampal functions (Graves et al. 2003). We first examined the effect of a 4-h RSD immediately following training to assess changes in the consolidation of contextual information (Fig. 4A). RSD animals showed a significant reduction in memory performance (expressed as the percentage of time spent in freezing) when reexposed in the conditioning chamber 24 h after training as compared with RSC animals (P < 0.002, RSD: n = 8 and RSC: n = 10, Fig. 4B). As expected, quantification of vigilance states showed a significant decrease in REM sleep, but not NREM sleep, for the RSD group immediately after training (REM sleep: P < 0.001; NREM sleep: P > 0.05, Fig. 4C). Interestingly, 4 h after training (Zeitgeber (ZT) 4–8, 12:00–4:00 PM), a significant sleep rebound (combining both REM and NREM sleep) was observed in both groups as compared with the baseline period assessed the day prior to training (REM sleep: P < 0.05 in both groups; NREM sleep: P < 0.05 in both groups), in agreement with previous studies (Lucero 1970; Hennevin et al. 1995; Hellman and Abel 2007). The EEG spectral analysis during post-learning REM sleep did not show any differences in the power or peak frequency of the theta band between both groups (P > 0.5, n = 18). A positive correlation was found between the percentage of freezing and REM (but not NREM) sleep quantity during the first 4-h post-training period (REM sleep: r = 0.57, P < 0.01; NREM sleep: r = 0.1, P > 0.05, n = 18, Fig. 4D), indicating that the more an animal spent in post-training REM sleep, the stronger was the consolidation of fear conditioning. Such correlation with REM sleep was not observed for the second 4-h post-training period (r = 0.2, P > 0.05, n = 18). In addition, no correlation was found between the first 4 h post-training REM sleep latency and the percentage of freezing during testing (r = 0.14, P > 0.05, n = 18). These results not only show that short RSD impaired CFC consolidation, but also suggest a link between post-training REM sleep quantity and performance level at testing 24 h after.

Figure 4.

Consolidation of CFC was altered by REM sleep deprivation after training. (A) Timeline of the protocol: REM sleep-deprived group rats were subjected to a 4-h RSD immediately after training given at Zeitgeber 0 (8:00 AM, light-on onset), as compared with a nondeprived control group (RSC). The training protocol consisted of a unique electrical footshock (0.3 mA during 2 s). Both groups were tested 24 h later. (B) CFC performances were assessed by measuring the level of freezing during training and testing. RSD led to a decrease in the amount of freezing assessed during testing 24 h after training (13.8 ± 3.4% vs. 41.6 ± 6.6%, P < 0.003, n = 8 and 10 for RSD and RSC, respectively). (C) Vigilance states quantity during ZT 0–4 and ZT 4–8 periods in the 2 groups during baseline (24 h before training, dashed bars) and after training (white bars for RSD, and gray bars for RSC). As expected, REM sleep amount in the RSD group was significantly reduced as compared with RSC and baseline immediately after training (ZT 0–4, for REM sleep, RSD and RSC, respectively: 2.9 ± 0.5% vs. 9.8 ± 1.2%, P < 0.001; for NREM sleep, RSD and RSC, respectively: 45.2 ± 2.6% vs. 44.7 ± 1.1%, P > 0.05). The training significantly reduced the amount of NREM sleep in both groups as compared with baseline. For ZT 4–8 period, REM and NREM sleep amounts significantly increased in RSD and RSC groups as compared with baseline (REM sleep for, RSD, RSC and baseline period: 19.3 ± 1.1% and 18.6 ± 0.8%, 13.0 ± 0.8%, P < 0.05 in both comparisons; NREM sleep: 51.5 ± 2.6% and 50.2 ± 1.8%, 42.7 ± 1.7%, P < 0.05 in both comparisons). (D) Linear regression between REM sleep amount during the 4-h post-training period and the level of freezing at testing. The figure shows a significant positive correlation between these 2 parameters (r = 0.57, P < 0.01, n = 18).

Figure 4.

Consolidation of CFC was altered by REM sleep deprivation after training. (A) Timeline of the protocol: REM sleep-deprived group rats were subjected to a 4-h RSD immediately after training given at Zeitgeber 0 (8:00 AM, light-on onset), as compared with a nondeprived control group (RSC). The training protocol consisted of a unique electrical footshock (0.3 mA during 2 s). Both groups were tested 24 h later. (B) CFC performances were assessed by measuring the level of freezing during training and testing. RSD led to a decrease in the amount of freezing assessed during testing 24 h after training (13.8 ± 3.4% vs. 41.6 ± 6.6%, P < 0.003, n = 8 and 10 for RSD and RSC, respectively). (C) Vigilance states quantity during ZT 0–4 and ZT 4–8 periods in the 2 groups during baseline (24 h before training, dashed bars) and after training (white bars for RSD, and gray bars for RSC). As expected, REM sleep amount in the RSD group was significantly reduced as compared with RSC and baseline immediately after training (ZT 0–4, for REM sleep, RSD and RSC, respectively: 2.9 ± 0.5% vs. 9.8 ± 1.2%, P < 0.001; for NREM sleep, RSD and RSC, respectively: 45.2 ± 2.6% vs. 44.7 ± 1.1%, P > 0.05). The training significantly reduced the amount of NREM sleep in both groups as compared with baseline. For ZT 4–8 period, REM and NREM sleep amounts significantly increased in RSD and RSC groups as compared with baseline (REM sleep for, RSD, RSC and baseline period: 19.3 ± 1.1% and 18.6 ± 0.8%, 13.0 ± 0.8%, P < 0.05 in both comparisons; NREM sleep: 51.5 ± 2.6% and 50.2 ± 1.8%, 42.7 ± 1.7%, P < 0.05 in both comparisons). (D) Linear regression between REM sleep amount during the 4-h post-training period and the level of freezing at testing. The figure shows a significant positive correlation between these 2 parameters (r = 0.57, P < 0.01, n = 18).

REM Sleep Rebound Improves the Consolidation of Contextual Information

Because LTP amplitude and Egr1 density increases in the hippocampus of RSR rats, we asked whether REM sleep rebound could improve memory consolidation. RSR animals were first REM sleep deprived for 4-h prior training (Fig. 5A) and were allowed to recover their REM sleep debt immediately after training. As shown in Figure 5B, during testing (24 h after training), RSR rats exhibited improved performance as compared with controls (P < 0.003, RSR: n = 8 and RSC: n = 14). As expected, REM sleep quantity for these rats was significantly higher immediately after training (ZT 4–8) as compared with the RSC group (P < 0.02, Fig. 5C). We found no correlation in RSC and RSR animals between REM sleep quantity during the 4 h post-training and the performance during testing (r = 0.38, P > 0.05, n = 21). In control animals, we observed a post-training sleep rebound (as compared with prior 24 h baseline sleep quantity, P < 0.02) similar to that already found in the previous experiment at ZT 4–8 (Fig. 4C), which could contribute to the lack of correlation. However, a significant positive correlation could be drawn using the percentage of REM sleep rebound for RSC and RSR rats (calculated as the ratio of REM sleep quantity during the 4 h post-training divided by the REM sleep quantity during the 4 h pretraining period) and the performance (freezing) at testing (r = 0.56, P < 0.01, n = 21, Fig. 5D). We also obtained a negative correlation between post-training REM sleep latency and the level of freezing during testing (r = −0.49, P < 0.02, n = 21, not shown) suggesting that the faster a rat fell in REM sleep, the better was its memory performance. We did not find such correlation between performance and NREM sleep quantity (r = 0.17, P > 0.05, n = 21) or latency (r = −0.217, P > 0.05, n = 21). We also found no difference between groups in the theta band for REM sleep episodes during the 4 h post-training period (P > 0.05, n = 21). Altogether, these results show that REM sleep rebound has a specific and facilitating effect on the consolidation of hippocampal-dependent memory and suggest an influence of REM sleep latency immediately following training on subsequent memory performance. Alternatively, one could attribute the improved performance seen in the RSR group to a positive influence of the pretraining REM sleep deprivation rather than the post-training REM sleep rebound. The increased memory performance at test would then be due to a positive effect of RSD on the encoding of contextual information during training. To discard this possibility, we performed an additional experiment consisting of a pretraining RSD and a test only 1 h following training (Fig. 5E). In addition, between training and testing, sleep was inhibited in both experimental groups in order to avoid any further effect of post-training rebound on memory formation. We found no significant difference between RSD and RSC rats in the percentage of freezing during training and testing at 1 and 24 h (P > 0.05, RSD: n = 10 and RSC: n = 8, Fig. 5F), although the RSD group was significantly deprived of REM sleep compared with RSC animals (Fig. 5G). This result demonstrates that the improved performance in the RSR group in the previous experiment (Fig. 5B) was specifically due to a post-training REM sleep increase rather than a pretraining REM sleep decrease. Additionally, it shows that a short REM sleep deprivation does not alter CFC acquisition, arguing against a fatigue effect generally attributed to sleep deprivation procedures.

Figure 5.

Increase in REM sleep after training enhances the consolidation of CFC. (A) Timeline of the protocol: animals from the RSR group were subjected to a 4-h RSD before training and then allowed to recover (REM sleep rebound) from this deprivation immediately after training. This group is compared with a nondeprived control group (RSC). For both groups, training occurred at ZT 4 and testing 24 h later. (B) CFC performance assessed by the percentage of freezing during training and test periods. A significant difference was obtained between RSC (gray bars) and RSR groups (black bars) at test (10.4 ± 2.0% vs. 30.9 ± 5.7%, P < 0.003, n = 14 and 8). (C) Vigilance state quantity during ZT 0–4 (deprivation period) and ZT 4–8 (rebound period) in the 2 groups (black bars for RSR, gray bars for RSC) and during baseline (24 h before training, dashed bars). During deprivation period, REM sleep amount in the RSR group was significantly reduced as compared with RSC and baseline. For the post-training period, REM sleep amount significantly increased in RSR as compared with baseline and RSC (P < 0.01 and P < 0.03 respectively). During the same period, REM sleep amount significantly increased in RSC group compared with baseline (P < 0.05). (D) Linear regression between REM sleep rebound amount during the 4-h post-training period and the level of freezing at testing. A significant positive correlation was found between these 2 parameters (r = 0.56, P < 0.01, n = 21). (EG) Effect of pretraining RSD on the encoding of CFC. (E) Timeline of the protocol: animals of the RSD group were subjected to a 4 h REM sleep deprivation before training, as compared with a nondeprived control group (RSC). For both groups, training occurred at ZT 4 and testing 1 h and 24 h later. (F) CFC performance assessed by the percentage of freezing during training and test periods. No difference was found between RSD and RSC groups, during post-training period, nor at testing 1 and 24 h after conditioning (15.2 ± 6.7% vs. 10.6 ± 7.7%, P > 0.05, n = 10 and 8, respectively, at 1 h test). (G) Vigilance states quantity during the pretraining deprivation period in the 2 groups (white bars for RSD, gray bars for RSC). During this period, REM sleep amount was significantly reduced in RSD as compared with RSC group. No difference of NREM sleep amount was found between these 2 groups.

Figure 5.

Increase in REM sleep after training enhances the consolidation of CFC. (A) Timeline of the protocol: animals from the RSR group were subjected to a 4-h RSD before training and then allowed to recover (REM sleep rebound) from this deprivation immediately after training. This group is compared with a nondeprived control group (RSC). For both groups, training occurred at ZT 4 and testing 24 h later. (B) CFC performance assessed by the percentage of freezing during training and test periods. A significant difference was obtained between RSC (gray bars) and RSR groups (black bars) at test (10.4 ± 2.0% vs. 30.9 ± 5.7%, P < 0.003, n = 14 and 8). (C) Vigilance state quantity during ZT 0–4 (deprivation period) and ZT 4–8 (rebound period) in the 2 groups (black bars for RSR, gray bars for RSC) and during baseline (24 h before training, dashed bars). During deprivation period, REM sleep amount in the RSR group was significantly reduced as compared with RSC and baseline. For the post-training period, REM sleep amount significantly increased in RSR as compared with baseline and RSC (P < 0.01 and P < 0.03 respectively). During the same period, REM sleep amount significantly increased in RSC group compared with baseline (P < 0.05). (D) Linear regression between REM sleep rebound amount during the 4-h post-training period and the level of freezing at testing. A significant positive correlation was found between these 2 parameters (r = 0.56, P < 0.01, n = 21). (EG) Effect of pretraining RSD on the encoding of CFC. (E) Timeline of the protocol: animals of the RSD group were subjected to a 4 h REM sleep deprivation before training, as compared with a nondeprived control group (RSC). For both groups, training occurred at ZT 4 and testing 1 h and 24 h later. (F) CFC performance assessed by the percentage of freezing during training and test periods. No difference was found between RSD and RSC groups, during post-training period, nor at testing 1 and 24 h after conditioning (15.2 ± 6.7% vs. 10.6 ± 7.7%, P > 0.05, n = 10 and 8, respectively, at 1 h test). (G) Vigilance states quantity during the pretraining deprivation period in the 2 groups (white bars for RSD, gray bars for RSC). During this period, REM sleep amount was significantly reduced in RSD as compared with RSC group. No difference of NREM sleep amount was found between these 2 groups.

REM Sleep Rebound Improves Acquisition of CFC

We finally examined the possibility that, unlike RSD, REM sleep rebound could affect the encoding of contextual information. In the experiment illustrated in Figure 6A, RSR animals were REM sleep deprived and allowed to recover for 2 h 30 min before being trained in CFC. Tests were performed 1 and 24 h after training to respectively assess acquisition and consolidation of contextual information. As above (Fig. 5E), animals of both groups were sleep-deprived during the 1 h training-test interval. The results showed that pretraining REM sleep rebound potentiated CFC acquisition as illustrated by an increased freezing response during both training (postshock: P < 0.05, RSR: n = 9 and RSC: n = 10, Fig. 6B) and reexposure to the conditioning chamber 1 h after training (P < 0.01, Fig. 6B). The RSR group showed a significant REM sleep increase (rebound) during the recovery period following RSD, without modification in NREM sleep quantity (Fig. 6C). Furthermore, REM sleep amount during rebound was correlated with the freezing response at the 1 h test (Fig. 6D, r = 0.63, P < 0.004, n = 19). No significant correlation was obtained between NREM sleep quantity and the level of freezing, suggesting a specific effect of REM sleep on CFC encoding.

Figure 6.

REM sleep rebound facilitated CFC encoding. (A) Timeline of the protocol: animals from the RSR group were subjected to a 4-h REM sleep deprivation and then allowed to recover (REM sleep rebound) from this deprivation. 2 h 30 min following the first REM sleep episode onset, training session was conducted and testing were performed 1 h and 24 h later. (B) CFC performance assessed by the level of freezing during training and test periods. RSR led to an increase in the amount of freezing assessed during the post-training period and during testing performed 1 h after conditioning (post-training: 12.3 ± 4.3% vs. 29.8 ± 9.3%, P < 0.05, n = 10 and 9 for RSC and RSR groups; 1 h test: 7.8 ± 4.5% vs. 20.3 ± 6.4%, P < 0.01). This effect was not maintained at the 24 h test. (C) Pretraining vigilance states quantity. REM sleep amount in the RSR group was significantly reduced during deprivation (6.3 ± 0.8% vs. 3.6 ± 0.5%, P < 0.03, RSC and RSR, respectively) and then increased during rebound (11.5 ± 0.9% vs. 19.8 ± 1.9%, P < 0.005). (D) Linear regression between REM sleep amount during the rebound period and the level of freezing at 1 h test post-training. A significant positive correlation was found between these 2 parameters (r = 0.63, P < 0.004, n = 19).

Figure 6.

REM sleep rebound facilitated CFC encoding. (A) Timeline of the protocol: animals from the RSR group were subjected to a 4-h REM sleep deprivation and then allowed to recover (REM sleep rebound) from this deprivation. 2 h 30 min following the first REM sleep episode onset, training session was conducted and testing were performed 1 h and 24 h later. (B) CFC performance assessed by the level of freezing during training and test periods. RSR led to an increase in the amount of freezing assessed during the post-training period and during testing performed 1 h after conditioning (post-training: 12.3 ± 4.3% vs. 29.8 ± 9.3%, P < 0.05, n = 10 and 9 for RSC and RSR groups; 1 h test: 7.8 ± 4.5% vs. 20.3 ± 6.4%, P < 0.01). This effect was not maintained at the 24 h test. (C) Pretraining vigilance states quantity. REM sleep amount in the RSR group was significantly reduced during deprivation (6.3 ± 0.8% vs. 3.6 ± 0.5%, P < 0.03, RSC and RSR, respectively) and then increased during rebound (11.5 ± 0.9% vs. 19.8 ± 1.9%, P < 0.005). (D) Linear regression between REM sleep amount during the rebound period and the level of freezing at 1 h test post-training. A significant positive correlation was found between these 2 parameters (r = 0.63, P < 0.004, n = 19).

Discussion

We here report that a short, mild, and nonstressful RSD was sufficient to alter CA3-to-CA1 synaptic plasticity and Egr1 neuronal expression in the CA1 area of the dorsal hippocampus and to cause a decrease in the consolidation of hippocampal-based memory as assessed by CFC. Strikingly, both LTP level and density of Egr1-expressing neurons were increased in the CA1 area of the dorsal hippocampus after REM sleep gain and correlated with REM sleep quantity, indicating a finely tuned regulation of plasticity processes during this vigilance state. In addition, CFC encoding and consolidation were restored and facilitated by REM sleep rebound and correlated with REM (but not NREM) sleep amount. Altogether, our results suggest a synaptic mechanism by which REM sleep bidirectionally regulates aversive contextual memory.

Methodological Considerations

Classically, the RSD was obtained when a rat is placed on top of an upside down flowerpot which is located in a bucket of water (Mendelson et al. 1974). This method, which was designed to prevent the generation of REM sleep episodes while allowing NREM sleep, induces in rats a stress by immobilization. The modified flowerpot technique using multiple platforms was then used to prevent this form of stress (McDermott et al. 2003; Ravassard et al. 2009). Although, the modified flowerpot technique was reported not to produce significant stress increase after 72 h or more of RSD, we recently found that short exposure (30 min) to this RSD method induced major stress (Clement et al. 2010) that may alter memory and LTP induction. This late form of stress is likely to be caused by the water surrounding the platforms. To avoid these methodological caveats, we developed a short and gentle RSD inducing a selective REM sleep reduction, without affecting NREM sleep or causing alteration in basic sleep EEG features, and without increasing stress. In addition, this method prevented direct exposure to the experimenter or to a novel environment (Kopp et al. 2006) that could produce stress (Beerling et al. 2011). We thus believe that this RSD method is more ecological and suitable to ascertain the specific role of REM sleep in synaptic plasticity and memory processes.

Our CFC protocols limited circadian variations between animals. However, we found that control animals conditioned at the beginning of the light phase (ZT 0, Fig. 4B) showed higher freezing level at testing than those conditioned later (ZT 4, Fig. 5B). In accordance with previous studies (Rudy and Pugh 1998; Devan et al. 2001; Loh et al. 2010), circadian cycle could have differentially influenced consolidation of our control groups. However, CFC consolidation was shown to depend more on absolute sleep loss during deprivation than on circadian rhythm (Hagewoud et al. 2010), and we found similar sleep quantity in both control groups after training. Thus, post-training sleep latency or REM/NREM sleep ratio following conditioning could have greater influence on CFC consolidation than circadian rhythm.

Short RSD Alters Memory Consolidation of CFC and Plasticity in the CA1 Area of the Dorsal Hippocampus

Our results revealed that short RSD impaired both synaptic plasticity in the CA1 area of the dorsal hippocampus and CFC consolidation. Previous findings point to the existence of a critical time window for CFC consolidation, which is only sensitive to gentle total sleep deprivation (TSD) when performed within 5 h post-training and not 5 h after CFC (Graves et al. 2003). Does such a time window specifically apply to REM sleep? Previous investigation of the RSD effect on memory using the flowerpot method led to contradictory results. While CFC consolidation was found to be impaired after prolonged RSD (Alvarenga et al. 2008), it seemed insensitive to short post-training RSD (5 h) (Silvestri 2005). The difference between this last result and ours may be due to the conditioning protocol used by these authors (10 footshocks delivered), rather than a difference in RSD method or duration. This multiple-trial protocol might have overshadowed potential RSD effects on hippocampal function. We used here a mild protocol (one shock) enabling the rat to exhibit mild freezing response and allowing REM sleep modulation to be expressed bidirectionally. Indeed, using a multiple-trial training (2 shocks), we did not observe any impairment on consolidation after 4 h RSD (testing at 24 h, P > 0.05, n = 7 in each group). CFC studies with lesion of the dorsal hippocampus (Wiltgen et al. 2006) and ablation of dentate gyrus neoneurogenesis (Drew et al. 2010) have shown a significant effect between control and experimental groups with a single shock, but no difference with 2 shocks. An explanation for these studies and our current study is that single-trial training requires the dorsal hippocampus, unlike multiple-trial training, which seems to be independent of the dorsal hippocampus (Nolan et al. 2004; Malleret et al. 2010). Consistent with the hypothesized critical post-training time window, a short and mild RSD (6 h) completely blocked the improvement of memory performance in an avoidance task in rats (Datta et al. 2004). These results, together with ours, show that REM sleep is required during a short time window to regulate emotional memory consolidation, without confounding effects from stress or other unspecific impairments related to the RSD procedure.

It is well demonstrated that CFC acquisition and consolidation require the dorsal hippocampus (Phillips and LeDoux 1992, 1994; Sanders et al. 2003) and involve long-term synaptic plasticity mechanisms requiring PKA and ERK-MAPK activation (Abel et al. 1997; Sindreu et al. 2007). We thus examined the effect of RSD on CA1 LTP and Egr1 expression, an upstream effector of ERK-MAPK pathway. We found that dorsal CA1 LTP induced at 10 and 100 Hz was severely decreased after short RSD. This result is in agreement with previous studies showing that prolonged RSD impairs CA1 LTP in the dorsal hippocampus (McDermott et al. 2003; Ravassard et al. 2009). Importantly, previous findings reported impaired LTP induced by 10 and 100-Hz protocols after 4 h gentle TSD (Kopp et al. 2006). This suggests that RSD was as powerful as TSD to increase LTP induction threshold. In vivo studies using a short and mild RSD (4 h) have also shown the importance of REM sleep for the long-term maintenance of the late phase of LTP at the perforant path-dentate gyrus synapse (Romcy-Pereira and Pavlides 2004), suggesting that REM sleep may facilitate LTP induction and expression in the hippocampus. Furthermore, we found that the density of Egr1-positive cells was also reduced after RSD, supporting a role for REM sleep in ERK-MAPK activation, further evidenced by previous findings (Ribeiro et al. 1999, 2002; Ravassard et al. 2009). Overall, these data suggest that the impairment in CFC consolidation seen after short RSD is caused by an alteration of hippocampal synaptic plasticity. Since amygdala-dependent cued fear conditioning was shown to be insensitive to prolonged RSD (McDermott et al. 2003), it would be of interest to determine whether plasticity is impaired in the amygdala after a short REM sleep loss. For instance, lack of REM sleep could lead to mechanisms increasing plasticity in other brain areas to compensate for the loss of dorsal hippocampus function, as shown for striatum-dependent learning (Watts et al. 2012).

Increase in REM Sleep Amount Improves Memory Encoding and Consolidation of CFC and LTP in the CA1 Area of the Dorsal Hippocampus

Another major finding of the present study is that REM sleep increase enhanced CFC encoding and consolidation, as well as 10 Hz-induced LTP and Egr1 expression in the dorsal CA1 area of the hippocampus, and that these 3 parameters were positively correlated with REM sleep amount. Previous studies using the flowerpot technique to induce prolonged RSD before CFC training either did not report a positive effect of subsequent REM sleep rebound on performance at test (McDermott et al. 2003; Li et al. 2009), or showed only a restoration without any enhancement on memory (Alvarenga et al. 2008; Pinho et al. 2013) or LTP (Ishikawa et al. 2006; Ravassard et al. 2009). A plausible explanation is that stress, fatigue accumulation or other compensatory mechanisms occurring during chronic RSD disrupted further potential contribution of REM sleep increase. In contrast, a study reported a positive effect on the consolidation of a Y-maze discrimination task after enhancing REM sleep amount either pharmacologically or following prolonged RSD (Wetzel et al. 2003). Since the training has been performed 15 h after RSD in order to eliminate unspecific effects of the flowerpot technique, it remained unclear if RSR improves the acquisition or the consolidation stages. In our experimental design, we were able to show that both encoding and consolidation stages of CFC were facilitated by a short window of REM sleep increase, respectively, occurring prior or post-training. In addition, performance at 24 h test was positively correlated with post-training REM sleep amount and negatively correlated with REM sleep latency. In accordance with a study showing that consolidation of human emotional memory is dependent on REM sleep amount and latency (Nishida et al. 2009), our latter findings indicate that the first REM sleep episode immediately following learning might constitute a transient temporal window during which consolidation is facilitated. Our study also suggests a synaptic mechanism by which RSR enhances CFC consolidation and learning. For the first time, an increase in REM sleep amount is shown to enhance LTP induced at 10 Hz. What could be the mechanism responsible for this RSR-dependent LTP facilitation? One possibility is that REM sleep may rescale or desaturate the synaptic efficacy of pyramidal cells. However, this seems unlikely as the I/O curves in RSD and RSR rats were not different from controls. In addition, whole-cell patch clamp recordings showed no difference in AMPAR-mediated synaptic responses between groups (unpublished data). Alternatively, REM sleep may influence the NR2A/NR2B ratio of NMDA receptor subunits known to modulate CA1 LTP induction, similarly to what was found after a short TSD (Kopp et al. 2006). Interestingly, the density of Egr1-positive cells was increased in the RSR group, suggesting that ERK-MAPK could be activated by REM sleep. Thus, the improvement seen in CFC consolidation by RSR may be linked to an enhancement of hippocampal plasticity.

Functional Dissociation Between Dorsal and Ventral Hippocampus

We found that REM sleep amount selectively modulated CA1 LTP in the dorsal, but not ventral hippocampus, confirming previous results using prolonged RSD (Ravassard et al. 2009). The dorsal hippocampus is more interconnected with cortical areas as compared with the ventral hippocampus, which is more linked to the amygdala and the hypothalamus. As mentioned above, RSD impairs CFC, but not cued fear conditioning (McDermott et al. 2003) which specifically relies on the functional integrity of the amygdala and the ventral part of the hippocampus (Phillips and LeDoux 1992, 1994). In contrast with cued fear conditioning, CFC requires the activation of both the ventral and dorsal hippocampus (Ruskin et al. 2004). As we found no alteration in LTP in the CA1 area of the ventral hippocampus, our results could explain the selective impact of REM sleep loss on CFC but not on cued fear conditioning (Ruskin et al. 2004).

In conclusion, our results suggest that REM sleep might be a key regulator of hippocampal-dependent plasticity and emotional memory. As REM sleep amount is increased in psychiatric diseases such as major depression and post-traumatic stress disorder (Walker and van der Helm 2009), as well as in models of depression (Popa et al. 2006; Watts et al. 2012), it is now of interest to investigate the role of REM sleep in these diseases.

Authors’ Contributions

P.A.S., P.R., G.M., and A.M.H. designed research; P.R., A.M.H., M.A.J., N.F, L.L., S.A., P.A.S., and G.M. performed research; C.M and M.T. contributed new reagents/analytic tools; P.R., A.M.H., P.A.L, M.A.J., N.F, L.L., P.A.S., C.M, and M.T. analyzed data; P.A.S., P.R., G.M., and A.M.H. wrote the manuscript.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.

Funding

P.R. and the project were supported by the Cluster11 Rhône-Alpes grant to P.A.S.; N.F. was supported by an ARC2 (Région Rhône-Alpes) grant.

Notes

We thank Ashley L. Kees for careful reading of the manuscript, Dr Anne-Marie Mouly for very helpful comments, Dr Bruno Claustrat for corticosterone analysis and members of the laboratory for useful discussions. Conflict of Interest: None declared.

References

Abel
T
Havekes
R
Saletin
JM
Walker
MP
.
2013
.
Sleep, plasticity and memory from molecules to whole-brain networks
.
Curr Biol
 .
23
:
R774
R788
.
Abel
T
Nguyen
PV
Barad
M
Deuel
TA
Kandel
ER
Bourtchouladze
R
.
1997
.
Genetic demonstration of a role for PKA in the late phase of LTP and in hippocampus-based long-term memory
.
Cell
 .
88
:
615
626
.
Alvarenga
TA
Patti
CL
Andersen
ML
Silva
RH
Calzavara
MB
Lopez
GB
Frussa-Filho
R
Tufik
S
.
2008
.
Paradoxical sleep deprivation impairs acquisition, consolidation, and retrieval of a discriminative avoidance task in rats
.
Neurobiol Learn Mem
 .
90
:
624
632
.
Beerling
W
Koolhaas
JM
Ahnaou
A
Bouwknecht
JA
de Boer
SF
Meerlo
P
Drinkenburg
WHIM
.
2011
.
Physiological and hormonal responses to novelty exposure in rats are mainly related to ongoing behavioral activity
.
Physiol Behav
 .
103
:
412
420
.
Benington
JH
Frank
MG
.
2003
.
Cellular and molecular connections between sleep and synaptic plasticity
.
Prog Neurobiol
 .
69
:
71
101
.
Bozon
B
Davis
S
Laroche
S
.
2002
.
Regulated transcription of the immediate-early gene Zif268: mechanisms and gene dosage-dependent function in synaptic plasticity and memory formation
.
Hippocampus
 .
12
:
570
577
.
Clement
O
Charrie
P
Salvert
D
Mandier
N
Libourel
P-A
Luppi
P-H
Salin
P-A
Malleret
G
.
2010
.
Paradoxical (REM) sleep, BDNF and the reprocessing of stressful experience
 .
San Diego
,
Calif
.
Datta
S
Mavanji
V
Ulloor
J
Patterson
EH
.
2004
.
Activation of phasic pontine-wave generator prevents rapid eye movement sleep deprivation-induced learning impairment in the rat: a mechanism for sleep-dependent plasticity
.
J Neurosci
 .
24
:
1416
1427
.
Devan
BD
Goad
EH
Petri
HL
Antoniadis
EA
Hong
NS
Ko
CH
Leblanc
L
Lebovic
SS
Lo
Q
Ralph
MR
et al
2001
.
Circadian phase-shifted rats show normal acquisition but impaired long-term retention of place information in the water task
.
Neurobiol Learn Mem
 .
75
:
51
62
.
Drew
MR
Denny
CA
Hen
R
.
2010
.
Arrest of adult hippocampal neurogenesis in mice impairs single- but not multiple-trial contextual fear conditioning
.
Behav Neurosci
 .
124
:
446
454
.
Graves
LA
Heller
EA
Pack
AI
Abel
T
.
2003
.
Sleep deprivation selectively impairs memory consolidation for contextual fear conditioning
.
Learn Mem
 .
10
:
168
176
.
Grosmark
AD
Mizuseki
K
Pastalkova
E
Diba
K
Buzsáki
G
.
2012
.
REM sleep reorganizes hippocampal excitability
.
Neuron
 .
75
:
1001
1007
.
Hagewoud
R
Whitcomb
SN
Heeringa
AN
Havekes
R
Koolhaas
JM
Meerlo
P
.
2010
.
A time for learning and a time for sleep: the effect of sleep deprivation on contextual fear conditioning at different times of the day
.
Sleep
 .
33
:
1315
1322
.
Hellman
K
Abel
T
.
2007
.
Fear conditioning increases NREM sleep
.
Behav Neurosci
 .
121
:
310
323
.
Hennevin
E
Hars
B
Maho
C
Bloch
V
.
1995
.
Processing of learned information in paradoxical sleep: relevance for memory
.
Behav Brain Res
 .
69
:
125
135
.
Hobson
JA
Pace-Schott
EF
.
2002
.
The cognitive neuroscience of sleep: neuronal systems, consciousness and learning
.
Nat Rev Neurosci
 .
3
:
679
693
.
Hu
H
Real
E
Takamiya
K
Kang
MG
Ledoux
J
Huganir
RL
Malinow
R
.
2007
.
Emotion enhances learning via norepinephrine regulation of AMPA-receptor trafficking
.
Cell
 .
131
:
160
173
.
Inostroza
M
Born
J
.
2013
.
Sleep for preserving and transforming episodic memory
.
Annu Rev Neurosci
 .
36
:
79
102
.
Ishikawa
A
Kanayama
Y
Matsumura
H
Tsuchimochi
H
Ishida
Y
Nakamura
S
.
2006
.
Selective rapid eye movement sleep deprivation impairs the maintenance of long-term potentiation in the rat hippocampus
.
Eur J Neurosci
 .
24
:
243
248
.
Ji
D
Wilson
MA
.
2007
.
Coordinated memory replay in the visual cortex and hippocampus during sleep
.
Nat Neurosci
 .
10
:
100
107
.
Karlsson
MP
Frank
LM
.
2009
.
Awake replay of remote experiences in the hippocampus
.
Nat Neurosci
 .
12
:
913
918
.
Kim
JJ
Diamond
DM
.
2002
.
The stressed hippocampus, synaptic plasticity and lost memories
.
Nat Rev Neurosci
 .
3
:
453
462
.
Kopp
C
Longordo
F
Nicholson
JR
Luthi
A
.
2006
.
Insufficient sleep reversibly alters bidirectional synaptic plasticity and NMDA receptor function
.
J Neurosci
 .
26
:
12456
12465
.
Li
S
Tian
Y
Ding
Y
Jin
X
Yan
C
Shen
X
.
2009
.
The effects of rapid eye movement sleep deprivation and recovery on spatial reference memory of young rats
.
Learn Behav
 .
37
:
246
253
.
Loh
DH
Navarro
J
Hagopian
A
Wang
LM
Deboer
T
Colwell
CS
.
2010
.
Rapid changes in the light/dark cycle disrupt memory of conditioned fear in mice
.
PLoS One
 .
5
:
1
12
.
Louie
K
Wilson
MA
.
2001
.
Temporally structured replay of awake hippocampal ensemble activity during rapid eye movement sleep
.
Neuron
 .
29
:
145
156
.
Lucero
MA
.
1970
.
Lengthening of REM sleep duration consecutive to learning in the rat
.
Brain Res
 .
20
:
319
322
.
Malleret
G
Alarcon
JM
Martel
G
Takizawa
S
Vronskaya
S
Yin
D
Chen
IZ
Kandel
ER
Shumyatsky
GP
.
2010
.
Bidirectional regulation of hippocampal long-term synaptic plasticity and its influence on opposing forms of memory
.
J Neurosci
 .
30
:
3813
3825
.
Maloney
KJ
Mainville
L
Jones
BE
.
1999
.
Differential c-Fos expression in cholinergic, monoaminergic, and GABAergic cell groups of the pontomesencephalic tegmentum after paradoxical sleep deprivation and recovery
.
J Neurosci
 .
19
:
3057
3072
.
Maquet
P
.
2001
.
The role of sleep in learning and memory
.
Science (80-)
 .
294
:
1048
1052
.
McDermott
CM
LaHoste
GJ
Chen
C
Musto
A
Bazan
NG
Magee
JC
.
2003
.
Sleep deprivation causes behavioral, synaptic, and membrane excitability alterations in hippocampal neurons
.
J Neurosci
 .
23
:
9687
9695
.
Mendelson
WB
Guthrie
RD
Frederick
G
Wyatt
RJ
.
1974
.
The flower pot technique of rapid eye movement (REM) sleep deprivation
.
Pharmacol Biochem Behav
 .
2
:
553
556
.
Nishida
M
Pearsall
J
Buckner
RL
Walker
MP
.
2009
.
REM sleep, prefrontal theta, and the consolidation of human emotional memory
.
Cereb Cortex
 .
19
:
1158
1166
.
Nolan
MF
Malleret
G
Dudman
JT
Buhl
DL
Santoro
B
Gibbs
E
Vronskaya
S
Buzsaki
G
Siegelbaum
SA
Kandel
ER
et al
2004
.
A behavioral role for dendritic integration: HCN1 channels constrain spatial memory and plasticity at inputs to distal dendrites of CA1 pyramidal neurons
.
Cell
 .
119
:
719
732
.
Phillips
RG
LeDoux
JE
.
1992
.
Differential contribution of amygdala and hippocampus to cued and contextual fear conditioning
.
Behav Neurosci
 .
106
:
274
285
.
Phillips
RG
LeDoux
JE
.
1994
.
Lesions of the dorsal hippocampal formation interfere with background but not foreground contextual fear conditioning
.
Learn Mem
 .
1
:
34
44
.
Pinho
N
Moreira
KM
Hipolide
DC
Sinigaglia-Coimbra
R
Ferreira
TL
Nobrega
JN
Tufik
S
Oliveira
MGM
.
2013
.
Sleep deprivation alters phosphorylated CREB levels in the amygdala: relationship with performance in a fear conditioning task
.
Behav Brain Res
 .
236
:
221
224
.
Popa
D
El Yacoubi
M
Vaugeois
J-M
Hamon
M
Adrien
J
.
2006
.
Homeostatic regulation of sleep in a genetic model of depression in the mouse: effects of muscarinic and 5-HT1A receptor activation
.
Neuropsychopharmacology
 .
31
:
1637
1646
.
Prince
T-M
Wimmer
M
Choi
J
Havekes
R
Aton
S
Abel
T
.
2014
.
Sleep deprivation during a specific 3-hour time window post-training impairs hippocampal synaptic plasticity and memory
.
Neurobiol Learn Mem
 .
109
:
122
130
.
Ravassard
P
Pachoud
B
Comte
JC
Mejia-Perez
C
Scote-Blachon
C
Gay
N
Claustrat
B
Touret
M
Luppi
PH
Salin
PA
.
2009
.
Paradoxical (REM) sleep deprivation causes a large and rapidly reversible decrease in long-term potentiation, synaptic transmission, glutamate receptor protein levels, and ERK/MAPK activation in the dorsal hippocampus
.
Sleep
 .
32
:
227
240
.
Rechtschaffen
A
Bergmann
BM
Gilliland
MA
Bauer
K
.
1999
.
Effects of method, duration, and sleep stage on rebounds from sleep deprivation in the rat
.
Sleep
 .
22
:
11
31
.
Ribeiro
S
Goyal
V
Mello
CV
Pavlides
C
.
1999
.
Brain gene expression during REM sleep depends on prior waking experience
.
Learn Mem
 .
6
:
500
508
.
Ribeiro
S
Mello
CV
Velho
T
Gardner
TJ
Jarvis
ED
Pavlides
C
.
2002
.
Induction of hippocampal long-term potentiation during waking leads to increased extrahippocampal zif-268 expression during ensuing rapid-eye-movement sleep
.
J Neurosci
 .
22
:
10914
10923
.
Romcy-Pereira
R
Pavlides
C
.
2004
.
Distinct modulatory effects of sleep on the maintenance of hippocampal and medial prefrontal cortex LTP
.
Eur J Neurosci
 .
20
:
3453
3462
.
Romcy-Pereira
RN
Erraji-Benchekroun
L
Smyrniotopoulos
P
Ogawa
S
Mello
CV
Sibille
E
Pavlides
C
.
2009
.
Sleep-dependent gene expression in the hippocampus and prefrontal cortex following long-term potentiation
.
Physiol Behav
 .
98
:
44
52
.
Rudy
JW
Pugh
CR
.
1998
.
Time of conditioning selectively influences contextual fear conditioning: further support for a multiple-memory systems view of fear conditioning
.
J Exp Psychol Anim Behav Process
 .
24
:
316
324
.
Ruskin
DN
Liu
C
Dunn
KE
Bazan
NG
LaHoste
GJ
.
2004
.
Sleep deprivation impairs hippocampus-mediated contextual learning but not amygdala-mediated cued learning in rats
.
Eur J Neurosci
 .
19
:
3121
3124
.
Sanders
MJ
Wiltgen
BJ
Fanselow
MS
.
2003
.
The place of the hippocampus in fear conditioning
.
Eur J Pharmacol
 .
463
:
217
223
.
Saxe
MD
Malleret
G
Vronskaya
S
Mendez
I
Garcia
AD
Sofroniew
MV
Kandel
ER
Hen
R
.
2007
.
Paradoxical influence of hippocampal neurogenesis on working memory
.
Proc Natl Acad Sci USA
 .
104
:
4642
4646
.
Shumyatsky
GP
Malleret
G
Shin
RM
Takizawa
S
Tully
K
Tsvetkov
E
Zakharenko
SS
Joseph
J
Vronskaya
S
Yin
D
et al
2005
.
stathmin, a gene enriched in the amygdala, controls both learned and innate fear
.
Cell
 .
123
:
697
709
.
Siegel
JM
.
2001
.
The REM sleep-memory consolidation hypothesis
.
Science (80-)
 .
294
:
1058
1063
.
Silvestri
AJ
.
2005
.
REM sleep deprivation affects extinction of cued but not contextual fear conditioning
.
Physiol Behav
 .
84
:
343
349
.
Sindreu
CB
Scheiner
ZS
Storm
DR
.
2007
.
Ca2+ -stimulated adenylyl cyclases regulate ERK-dependent activation of MSK1 during fear conditioning
.
Neuron
 .
53
:
79
89
.
Smith
C
Lapp
L
.
1986
.
Prolonged increases in both PS and number of REMS following a shuttle avoidance task
.
Physiol Behav
 .
36
:
1053
1057
.
Smith
C
Rose
GM
.
1996
.
Evidence for a paradoxical sleep window for place learning in the Morris water maze
.
Physiol Behav
 .
59
:
93
97
.
Van Der Werf
YD
Altena
E
Schoonheim
MM
Sanz-Arigita
EJ
Vis
JC
De Rijke
W
Van Someren
EJW
.
2009
.
Sleep benefits subsequent hippocampal functioning
.
Nat Neurosci
 .
12
:
122
123
.
Vecsey
CG
Baillie
GS
Jaganath
D
Havekes
R
Daniels
A
Wimmer
M
Huang
T
Brown
KM
Li
X-YY
Descalzi
G
et al
2009
.
Sleep deprivation impairs cAMP signalling in the hippocampus
.
Nature
 .
461
:
1122
1125
.
Verret
L
Fort
P
Gervasoni
D
Léger
L
Luppi
P-HH
Leger
L
.
2006
.
Localization of the neurons active during paradoxical (REM) sleep and projecting to the locus coeruleus noradrenergic neurons in the rat
.
J Comp Neurol
 .
495
:
573
586
.
Vertes
RP
.
2004
.
Memory consolidation in sleep; dream or reality
.
Neuron
 .
44
:
135
148
.
Veyrac
A
Besnard
A
Caboche
J
Davis
S
Laroche
S
.
2014
.
The transcription factor zif268/egr1, brain plasticity, and memory
.
Prog Mol Biol Transl Sci
 .
122
:
89
129
.
Walker
MP
van der Helm
E
.
2009
.
Overnight therapy? The role of sleep in emotional brain processing
.
Psychol Bull
 .
135
:
731
748
.
Watts
A
Gritton
HJ
Sweigart
J
Poe
GR
.
2012
.
Antidepressant suppression of non-REM sleep spindles and REM sleep impairs hippocampus-dependent learning while augmenting striatum-dependent learning
.
J Neurosci
 .
32
:
13411
13420
.
Wetzel
W
Wagner
T
Balschun
D
.
2003
.
REM sleep enhancement induced by different procedures improves memory retention in rats
.
Eur J Neurosci
 .
18
:
2611
2617
.
Wiltgen
BJ
Sanders
MJ
Anagnostaras
SG
Sage
JR
Fanselow
MS
.
2006
.
Context fear learning in the absence of the hippocampus
.
J Neurosci
 .
26
:
5484
5491
.
Yang
CH
Huang
CC
Hsu
KS
.
2004
.
Behavioral stress modifies hippocampal synaptic plasticity through corticosterone-induced sustained extracellular signal-regulated kinase/mitogen-activated protein kinase activation
.
J Neurosci
 .
24
:
11029
11034
.

Author notes

P.R., A.M.H., G.M., and P.-A.S. contributed equally to the work.