Abstract
Stress profoundly affects sleep and memory processes. Stress impairs memory consolidation and, similarly, disruptions in sleep compromise memory functions. Yet, the neural circuits underlying stress-induced sleep and memory disturbances are still not fully understood. Here, we show that activation of corticotropin-releasing hormone neurons in the paraventricular nucleus of the hypothalamus (CRHPVN), similar to acute restraint stress, decreases sleep and impairs memory in a spatial object recognition task in male mice. Conversely, inhibiting CRHPVN neurons during stress reduces stress-induced memory deficits while slightly increasing the amount of sleep. We found that both stress and stimulation of CRHPVN neurons activate neurons in the lateral hypothalamus (LH) and that CRHPVN projections to the LH regulate stress-induced memory deficits and sleep disruptions. Our results suggest that CRHPVN neuronal pathways regulate the adverse effects of stress on memory and sleep—an important step toward improving sleep and ameliorating cognitive deficits associated with stress-related disorders.
Significance Statement
Stress significantly affects both sleep and memory, with spatial memory being particularly vulnerable. In this study, we combine acute restraint stress with optogenetic manipulations and a spatial object recognition task to investigate how corticotropin-releasing hormone neurons in the paraventricular nucleus of the hypothalamus (CRHPVN), and their projections to the lateral hypothalamus (LH), influence memory performance and sleep–wake states following stress. Our findings reveal that activating CRHPVN neurons impairs memory performance and increases wakefulness, whereas inhibiting CRHPVN neurons during stress improves memory and sleep. Inhibiting CRHPVN neuronal projections to the LH similarly improves memory performance and sleep. This work highlights the role of CRHPVN neurons and their projections to the LH in modulating stress-induced alterations in memory and sleep–wake states.
Introduction
Stress disrupts memory and sleep (Cheeta et al., 1997; Roozendaal, 2002; Joëls et al., 2006; Kim and Dimsdale, 2007; Sandi and Pinelo-Nava, 2007; Pawlyk et al., 2008; Conrad, 2010; Han et al., 2012; Park et al., 2015; Lo Martire et al., 2020; Schwabe et al., 2022). In particular, spatial memory is susceptible to the effects of stress as well as sleep disturbances (Smith and Rose, 1997; Graves et al., 2003; Palchykova et al., 2006; Binder et al., 2012; Li et al., 2012; Ishikawa et al., 2014; Prince et al., 2014; Lopes da Cunha et al., 2019; Valdivia et al., 2024). Although numerous studies have established links between stress, memory, and sleep, the neural mechanisms underlying the impact of stress on both memory and sleep remain poorly understood.
Corticotropin-releasing hormone (CRH) plays a crucial role in regulating autonomic, endocrine, and behavioral responses to stress (Bale and Vale, 2004; de Kloet et al., 2005; Herman and Tasker, 2016). It has been shown to be involved in regulating spontaneous wakefulness and stress-induced arousal (Chang and Opp, 1998, 1999, 2001, 2002, 2004; Sanford et al., 2015). Furthermore, activation of CRH neurons in the paraventricular nucleus of the hypothalamus (CRHPVN) strongly promotes wakefulness and is associated with triggering stress-related behaviors, such as grooming, in the absence of stress (Füzesi et al., 2016; Li et al., 2020; Ono et al., 2020; Mitchell et al., 2024). Optogenetic activation of CRHPVN neurons causes an increase in the release of corticosterone, a primary stress hormone, which further emphasizes the importance of these neurons in regulating physiological responses to stress (Füzesi et al., 2016). The role of CRHPVN neurons in modulating stress responses and sleep–wake patterns highlights their potential impact on cognitive functions. However, the specific neural pathways involved and whether targeted interventions in these circuits can mitigate stress-induced disruptions in memory and sleep are still unknown.
CRHPVN neurons project to the lateral hypothalamus (CRHPVN→LH), a brain region implicated in the regulation of sleep–wake states and memory (Rho and Swanson, 1987; Graebner et al., 2015; Bonnavion et al., 2016; Füzesi et al., 2016; Li et al., 2020; Ono et al., 2020; Mitchell et al., 2024). Optogenetic activation of CRHPVN neurons has been shown to increase c-Fos labeling in the lateral hypothalamus (LH), confirming the functional connection between these two brain regions (Füzesi et al., 2016; Mitchell et al., 2024). This pathway has been studied for its role in modulating wakefulness, stress-induced behavioral changes, and alterations in motivational drive (Füzesi et al., 2016; Li et al., 2020; Ono et al., 2020; Mitchell et al., 2024). Different populations of neurons within the LH have been linked to the regulation of memory processes. Optogenetic activation of hypocretin/orexin neurons in the LH after the learning phase of the novel object recognition (NOR) task causes sleep fragmentation and impairs memory (Rolls et al., 2011). Similarly, activating melanin-concentrating hormone (MCH) neurons in the LH significantly impairs memory in the NOR task, underscoring the critical role of these neurons in memory functions (Izawa et al., 2019). Although the CRHPVN→LH pathway has been studied for its role in regulating sleep–wake cycles and stress responses, how this projection influences both sleep and memory following stress is largely unclear, constituting a major gap in our understanding of the neural circuitry involved in the interplay between stress, sleep, and cognitive functions.
By employing acute restraint stress coupled with optogenetic manipulations as well as a spatial object recognition (SOR) task, we investigated the role of CRHPVN neurons and their projections to the LH in regulating memory performance and sleep–wake states following stress.
Materials and Methods
Mice
All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC reference no. 806197) at the University of Pennsylvania and conducted in compliance with the National Institutes of Health Office of Laboratory Animal Welfare Policy. Experiments were performed in male CRH-IRES-Cre mice (#012704, The Jackson Laboratory) or C57BL/6J mice (#000664, The Jackson Laboratory) aged 6–18 weeks, weighing 18–25 g at the time of surgery. Mice were randomly assigned to experimental and control groups. Animals were group-housed with littermates on a 12 h light/dark cycle (lights on 7 A.M. and off at 7 P.M.) with ad libitum access to food and water.
Surgical procedures
Mice were anesthetized with isoflurane (1–4%) during the surgery and placed on a stereotaxic frame (Kopf) while being on a heating pad to maintain body temperature. The skin was incised, and small holes were drilled for virus injections and implantations of optic fibers and EEG/EMG electrodes. Two stainless steel screws were inserted into the skull 1.5 mm from midline and 1.5 mm anterior to the bregma and 2.5 mm from midline and 2.5 mm posterior to the bregma. The reference screw was inserted on top of the cerebellum. Two EMG electrodes were inserted into the neck musculature. Insulated leads from the EEG and EMG electrodes were soldered to a 2 × 3 pin header, which was secured to the skull using dental cement. After surgery, mice were monitored for any signs of pain or distress until fully recovered from anesthesia and ambulatory.
For optogenetic activation experiments (Figs. 1 and 2, Extended Data Fig. 2-1), AAV2-Ef1α-DIO-hChR2-eYFP [for the activation group—stabilized step-function opsin (SSFO)] or AAV2-Ef1α-DIO-eYFP (for the control group—eYFP) was unilaterally injected into the PVN (300 nl, AP −0.7 mm; ML ±0.4 mm; DV −4.6 to −4.7 mm from the cortical surface) of CRH-Cre mice, and an optic fiber (200 μm diameter) was implanted above the injection site in the PVN (AP −0.7 mm; ML ±0.3; DV −4.5 mm).
For optogenetic inhibition experiments (Figs. 4–7, Extended Data Fig. 5-1), AAV2-Ef1α-DIO-SwiChR++-eYFP (for the inhibition group) or AAV2-Ef1α-DIO-eYFP (for the control group) was bilaterally injected into the PVN (200 nl) of CRH-Cre mice, and an optic fiber was implanted above the middle of the PVN (AP −0.7 mm; ML ±0.00 mm; DV −4 mm).
For retrograde optogenetic inhibition experiments (Figs. 8 and 9, Extended Data Fig. 9-1), rAAV2-retro-Ef1α-DIO-SwiChR++-eYFP (for the inhibition group) or rAAV2-retro-Ef1α-DIO-eYFP (for the control group) was bilaterally injected into the LH (100–300 nl, AP −1.2 to −1.4 mm; ML ±1 mm; DV −5 to −5.2 mm from the cortical surface) of CRH-Cre mice, and an optic fiber was implanted above the middle of the PVN (AP −0.7 mm; ML ±0.00 mm; DV −4 mm).
After surgery, mice were allowed to recover for at least 2–3 weeks before beginning experiments.
For collateralization experiments (Fig. 10), AAVrg-EF1α-DIO-FlpO (50 nl) was bilaterally injected into the LH, and AAV8-hSyn-FLExFRT-mGFP-2A-synaptophysin-mRuby (50 nl) was bilaterally injected into the PVN of CRH-Cre mice. Mice were perfused 3 weeks later for histology.
Histology
Mice were deeply anesthetized and transcardially perfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA) in PBS. Brains were fixed overnight in 4% PFA and then transferred to a 30% sucrose in PBS solution for at least one night. Brains were embedded and mounted with Tissue-Tek OCT compound (Tissue-Tek, Sakura Finetek) and frozen. Subsequently, 40–60 μm sections were cut using a cryostat (Thermo Fisher Scientific HM525 NX) and mounted onto glass slides. Brain sections were washed with PBS followed by counterstaining with Hoechst solution (#33342, Thermo Fisher Scientific).
For GFP staining to visualize virus expression (Figs. 1, 2, 8, 9, 10), brain sections were washed in PBS for 5 min, permeabilized using PBST (0.3% Triton X-100 in PBS) for 30 min, and incubated in blocking solution (5% normal donkey serum in 0.3% PBST; 017-000-001, Jackson ImmunoResearch Laboratories) for 1 h. Brain sections were incubated with chicken anti-GFP antibody (1 : 1,000; GFP8794984, Aves Lab) in the blocking solution overnight (Figs. 1, 2, 8, 9) or for two nights (Fig. 10) at 4°C. The following morning, sections were washed in PBS and incubated at room temperature for 2 h with the donkey anti-chicken secondary antibody conjugated to A594 or A488 (1 : 500; 703-585-155 or 703-545-155, Jackson ImmunoResearch Laboratories). Afterward, sections were washed with PBS followed by counterstaining with Hoechst solution (#33342, Thermo Fisher Scientific). Slides were coverslipped with mounting medium (Fluoromount-G, SouthernBiotech) and imaged using a fluorescence microscope (microscope, Leica DM6B; camera, Leica DFC7000GT; LED, Leica CTR6 LED) to verify virus expression and optic fiber placement. Collateralization slides (Fig. 10) were imaged using a Zeiss LSM 980 confocal microscope at the Cell and Developmental Biology Microscope Core (RRID:SCR_022373).
Animals were excluded if no virus expression was detected, or the virus expression/optic fiber tips were not properly localized to the targeted area.
For c-Fos staining (Fig. 8), brain sections were washed in PBS for 5 min, permeabilized using PBST (0.3% Triton X-100 in PBS) for 30 min, and incubated in blocking solution (5% normal donkey serum in 0.3% PBST; 017-000-001, Jackson ImmunoResearch Laboratories) for 1 h. Brain sections were incubated with rabbit anti-c-Fos antibody (1 : 1,000; 2250S, Cell Signaling Technology) in the blocking solution for 2 d at 4°C. Following primary antibody incubation, sections were washed in PBS and incubated at room temperature for 2 h with the donkey anti-rabbit secondary antibody conjugated to A594 (1 : 500; A21207, Thermo Fisher Scientific). Afterward, sections were washed with PBS followed by counterstaining with Hoechst solution (#33342, Thermo Fisher Scientific). Slides were coverslipped with mounting medium (Fluoromount-G, SouthernBiotech) and imaged using a fluorescence microscope. The c-Fos–positive cells were counted from sections containing the LH.
Sleep recordings
Sleep recordings were carried out in the animal’s home cage or in a recording cage to which the mouse had been previously habituated. EEG and EMG electrodes were connected to flexible recording cables via a mini connector. EEG and EMG signals were recorded using an RHD2132 amplifier (Intan Technologies; sampling rate, 1 kHz) connected to the RHD USB interface board (Intan Technologies). EEG and EMG signals were referenced to a ground screw placed on top of the cerebellum. To determine the sleep–wake state of the animal, we first computed the EEG and EMG spectrogram for sliding, half-overlapping 5 s windows, resulting in a 2.5 s time resolution. To estimate within each 5 s window the power spectral density (PSD), we performed Welch’s method with the Hanning window using sliding, half-overlapping 2 s intervals. Next, we computed the time-dependent δ (0.5–4 Hz), θ (5–12 Hz), σ (12–20 Hz), and high γ (100–150 Hz) power by integrating the EEG power in the corresponding ranges within the EEG spectrogram. We also calculated the ratio of the θ and δ power (θ/δ) and the EMG power in the range 50–500 Hz. For each power band, we used its temporal mean to separate it into a low and high part (except for the EMG and θ/δ ratio, where we used the mean plus one standard deviation as threshold). REM sleep was defined by a high θ/δ ratio, low EMG, and low δ power. A state was set as NREM sleep if the δ power was high, θ/δ ratio was low, and EMG power was low. In addition, states with low EMG power, low δ, but high σ power were scored as NREM sleep. Wake encompassed states with low δ power and high EMG power and each state with high γ power (if not otherwise classified as REM sleep). Our algorithm has been published and has a 90.256% accuracy compared with manual scoring by expert annotators (Weber et al., 2015, 2018; Chung et al., 2017; Antila et al., 2022; Stucynski et al., 2022; Schott et al., 2023; Maurer et al., 2024; Smith et al., 2024). We manually verified the automatic classification using a graphical user interface visualizing the raw EEG and EMG signals, EEG spectrograms, EMG amplitudes, and the hypnogram to correct for errors, by visiting each single 2.5 s epoch in the hypnograms. The software for automatic sleep–wake state classification and manual scoring was programmed in Python (available at https://github.com/tortugar/Lab/tree/master/PySleep).
Mice exhibiting two faulty channels of the same type (EEG or EMG) or a faulty ground wire were excluded from sleep analysis due to the challenges in accurately determining sleep states.
Optogenetic manipulation
Light pulses (1 s step pulses, 4 mW) for SSFO and SwiChR++ experiments were generated by a blue laser (473 nm, Laserglow) and sent through the optic fiber (200 μm diameter, Thorlabs) that connects to the ferrule on the mouse’s head. The timing of laser stimulation for optogenetic activation was randomized within a range of intervals (120–240 s intervals). This laser stimulation protocol was rationally designed based on previous studies (Berndt et al., 2016; Iyer et al., 2016; Selimbeyoglu et al., 2017; Wiegert et al., 2017; Stucynski et al., 2022). TTL pulses to trigger the laser were controlled using a Raspberry Pi, which was controlled by a custom user interface programmed in Python. Optogenetic manipulations were conducted during the light period for 1 h. Sleep recordings with stress were performed once to ensure that experimental mice were exposed to acute stress only once.
Spatial object recognition task
Mice were habituated to being handled twice daily for ∼1–2 min per handling for 5 d prior to beginning the memory task. Mice were also habituated to their recording cages during this time. During the open field habituation session (ZT1–1.5, Days 1 and 2), mice were habituated to the training context (13” × 13” open field arena) for two 5 min sessions. Between habituation sessions, the mice were returned to their individual recording cages for ∼1–2 min while the field was wiped with 70% ethanol and then placed back into the field for the next habituation session. A visual cue (a rectangle containing alternating black and white stripes) was attached to one wall of the open field to help the mice orient themselves within the open field. During the training session (ZT1–1.5, Day 3), mice were placed in the arena with two identical objects (two small glass bottles) for three 5 min training sessions. Between training sessions, the mice were returned to their individual recording cages for ∼1–2 min while the field was wiped with 70% ethanol and then placed back into the field for the next training session. Immediately following the final training session, the mice were returned to their recording cages. In Figures 1, 2, and 4⇓⇓⇓⇓–9, optogenetic manipulation was performed for 1 h with the mice in their recording cages with (Figs. 4, 5, 8, 9) or without (Figs. 1, 2, 6, 7) restraint stress (ZT1.5–2.5). In Figure 3, control and restraint stress mice were not implanted, and therefore their sleep was not recorded. However, they underwent identical treatment to the implanted mice, except that they were not tethered during the post-habituation and post-training periods. During the test session (ZT1–1.5, Day 4), mice were placed in the arena with the two familiar objects, one displaced to a new location, for a single 5 min test session. Exploration of the objects was defined as the amount of time the mouse had its nose oriented toward the object, within ∼2 cm of the object. Grooming near the object was not counted as exploration (Leger et al., 2013).
Preference (%) is calculated as follows:
Locomotion trajectory
For body part tracking, we used DeepLabCut (version 2.3.9; Mathis et al., 2018; Nath et al., 2019). Specifically, we labeled 20 frames/video taken from 37 SOR training session videos, and then 95% was used for training the model. We used a ResNet50-based neural network with default parameters for 100,000 training iterations (He et al., 2016; Insafutdinov et al., 2016). This network was then used to analyze videos from similar experimental settings during the SOR test session. We generated representative trajectory line traces based on the nose x- and y-coordinates generated by DeepLabCut. Coordinates that were outside of the arena (i.e., the nose was obscured and could not be tracked) were excluded from the final traces.
Restraint stress
Mice were restrained in modified plastic conical tubes (Falcon 50 ml Conical Centrifuge Tubes) with a hole cut in the front for ventilation and a hole cut in the back (lid) for the tail. For optogenetic experiments, the mice were first gently handled while the EEG/EMG and optogenetic cables were plugged into the implants on their heads. Once the cables were attached, the mice were restrained in a modified conical tube containing a hole cut in the front for ventilation, a thin (∼1 cm wide) rectangular window cut out on the top (spanning from the front of the tube to the back/lid) to allow the implant and cables to fit comfortably, and a hole in the back (lid) for the tail. Mice were gently removed from the restraint stress tubes after the hour was finished. Optogenetic mice remained attached to the EEG/EMG and optogenetic cables after being removed from their tubes.
Behavioral analysis
Mice were recorded by an overhead camera (Chameleon3 USB3, Teledyne with 2.8–8 mm varifocal CS-mount lens, Fujinon) during sleep recordings following the spatial object recognition task on the training day. The first 15 min of behaviors during eYFP/SSFO stimulation (Fig. 1F) or immediately following eYFP/SwiChR++ inhibition during restraint stress on the training day (Fig. 4F) was annotated. The behaviors annotated included grooming, walking, rearing, surveying (mouse was actively looking around/moving head, but not walking), and inactive (mouse was stationary). If the mouse spent the majority of the recording in an area of the recording cage not captured by the camera, the behavioral analysis was excluded from the final analysis.
Blood collection
CRH-Cre mice with AAV2-Ef1α-DIO-hChR2-eYFP (SSFO) injected in their PVN were habituated to the room where blood collection would occur for 1 h (ZT0–1) in individual recording cages (Fig. 1A). An optic fiber cable was then connected to the ferrule implanted on each mouse’s head to allow the mice to habituate to being tethered (ZT1–2). This protocol was repeated for at least 3 d prior to blood collection. At the time of blood collection, mice were restrained in a tube, and a sterile needle was used to nick the tail vein. A small volume of blood (15–20 µl) was collected into heparinized tubes (Microvette CB 300), and pressure was applied to stop any bleeding before mice were returned to their recording cages. Mice were restrained for blood collection for <3 min. Each mouse had 1–2 baseline blood samples collected on different days (Fig. 1G). All blood samples (baseline and stimulation) were collected at the same time of day (ZT2–3) to account for circadian variations in corticosterone levels. On the stimulation day, SSFO mice were habituated to the room where blood collection would occur for 1 h (ZT0–1) in individual recording cages and then were tethered and received laser stimulation (1 s step pulses, 4 mW, 120–240 s intervals) for 1 h (ZT1–2). Immediately afterward, a small volume of blood was collected (ZT2–3; as described above; Fig. 1G).
Corticosterone assay
The blood samples were centrifuged at 2,000 × g for 5 min at room temperature and stored at −80°C until running the corticosterone ELISA assay (Kim et al., 2018). Plasma corticosterone concentrations were measured using the DetectX Corticosterone Immunoassay kit (Arbor Assays) according to the manufacturer's instructions (Fig. 1G).
Statistical tests
Statistical analyses were performed using the Python modules (scipy.stats, https://scipy.org; pingouin, https://pingouin-stats.org; statsmodels, https://www.statsmodels.org/) and Prism v10.3.1 (GraphPad Software). We did not predetermine sample sizes, but cohorts were similarly sized as in other relevant sleep and memory studies (Rolls et al., 2011; Izawa et al., 2019). All data collection was randomized and counterbalanced. All data are reported as mean + SEM. A (corrected) p-value of <0.05 was considered statistically significant for all comparisons. Data were compared using unpaired t tests, paired t tests, mixed model ANOVA followed by t tests with Bonferroni's multiple-comparisons test, and a linear mixed model. Statistical results and parameters (exact value of n and what n represents) are presented in Extended Data Figure 1-1, figure legends, and results.
In Figures 1 and 2 and Extended Data Figure 2-1, n = 10 eYFP and 14 SSFO mice were used.
n = 1 eYFP mouse and 4 SSFO mice were excluded from Figure 1, D and E, for low exploration (n = 2 SSFO mice) or having a preference during the training session (n = 1 eYFP and 2 SSFO mice).
n = 2 SSFO mice were excluded from Figure 1F because their behaviors were not able to be annotated (mice were not visible on camera for much of the recording). These mice were not excluded from sleep recording analysis unless otherwise noted.
n = 1 eYFP mouse and 4 SSFO mice were excluded from Figure 2E and Extended Data Figure 2-1D due to having shorter sleep recordings (less than the full 6 h).
n = 1 SSFO mouse was excluded from Figure 2 and Extended Data Figure 2-1 for having a faulty ground wire, making accurate sleep annotations difficult.
In Figure 3, n = 19 control and 13 stress mice were used.
n = 3 control mice and 1 stress mouse were excluded from Figure 3, C and D, for low exploration (n = 1 stress mouse) or having a preference during the training session (n = 3 control mice).
In Figures 4 and 5 and Extended Data Figure 5-1, n = 13 eYFP-stress and 14 SwiChR++-stress mice were used.
n = 2 eYFP-stress mice and 3 SwiChR++-stress mice were excluded from Figure 4, D and E, for low exploration (n = 1 SwiChR++-stress mouse) or having a preference during the training session (n = 2 eYFP-stress and 2 SwiChR++-stress mice).
n = 4 SwiChR++-stress mice were excluded from Figure 4F because their behaviors were not able to be annotated (mice were not visible on camera for much of the recording). These mice were not excluded from sleep recording analysis unless otherwise noted.
n = 1 eYFP-stress mouse was excluded from Figure 5 and Extended Data Figure 5-1 for having a faulty ground wire, making accurate sleep annotations difficult.
In Figures 6 and 7, n = 17 eYFP and 19 SwiChR++ mice were used.
n = 3 eYFP mice and 7 SwiChR++ mice were excluded from Figure 6, D and E, for low exploration (n = 2 eYFP and 5 SwiChR++ mice) or having a preference during the training session (n = 1 eYFP and 2 SwiChR++ mice).
In Figures 8 and 9 and Extended Data Figure 9-1, n = 12 retro-eYFP-stress and 24 retro-SwiChR++-stress mice were used.
n = 1 retro-eYFP-stress mouse and 7 retro-SwiChR++-stress mice were excluded from Figure 8, J and K, for low exploration (n = 5 retro-SwiChR++-stress mice) or having a preference during the training session (n = 1 retro-eYFP-stress and 2 retro-SwiChR++-stress mice). These mice were not excluded from the sleep recording analysis.
Results
Activation of CRHPVN neurons impairs spatial object recognition memory and decreases sleep
By performing an SOR task combined with electroencephalogram (EEG) and electromyography (EMG) recordings, we tested whether activation of CRHPVN neurons following training is sufficient to impair spatial memory performance and reduce sleep (Fig. 1). To investigate this, we optogenetically activated CRHPVN neurons using the stabilized step-function opsin (SSFO), a double mutant excitatory channelrhodopsin. CRH-Cre mice were injected with an AAV encoding Cre-inducible SSFO (AAV2-Ef1ɑ-DIO-hChR2-eYFP) or eYFP (AAV2-Ef1ɑ-DIO-eYFP) in the PVN, followed by implantation of an optic fiber above the injection site and electrodes for EEG/EMG recordings (Fig. 1A). Two weeks later, we performed the SOR task (Fig. 1B). During habituation (ZT1–1.5, Days 1 and 2), the mice were familiarized with the open field where the training and testing would occur for two 5 min sessions per day. A visual cue was attached to one wall to help orient the mice within the field. On the training day (ZT1–1.5, Day 3), mice were placed in the open field with two identical objects (small glass bottles) for three 5 min training sessions. Immediately following training, optogenetic stimulation was performed for 1 h (1 s step pulses at 120–240 s intervals, 4 mW, ZT1.5–2.5) followed by EEG and EMG recordings (ZT2.5–7.5). Twenty-four hours later, during the test session (ZT1–1.5, Day 4), the mice were placed in the open field with the two familiar objects, one of which was moved to a new location, for one 5 min session. We found that eYFP control mice showed a significant increase in preference for the moved object and a positive discrimination ratio during the test session [Fig. 1C–E; mixed ANOVA: virus, p = 0.018; training vs test, p = 0.019; interaction, p = 0.014; t tests with Bonferroni’s correction: eYFP (training vs test), p = 0.004]. Conversely, SSFO mice exhibited no discernible preference between the moved and familiar objects, indicating impaired performance in this memory task [Fig. 1D; t tests with Bonferroni’s correction: SSFO (training vs test), p > 0.999]. Accordingly, the test day preference and discrimination ratio of the SSFO mice were significantly lower than the eYFP control mice [Fig. 1D,E; t tests with Bonferroni’s correction: test (eYFP vs SSFO), p = 0.002; t test: discrimination ratio, p = 0.007], suggesting that the activation of CRHPVN neurons after training impairs memory performance.
Activation of CRHPVN neurons impairs spatial object recognition memory. A, Top left, Schematic of optogenetic activation experiments with simultaneous EEG and EMG recordings. Mouse brain figure adapted from the Allen Reference Atlas—Mouse Brain (https://atlas.brain-map.org). Top right, Fluorescence image of the PVN in a CRH-Cre mouse injected with AAV2-Ef1α-DIO-hChR2-eYFP into the PVN. Scale bar, 300 µm. Bottom, Location of optic fiber tracts. B, Schematic indicating the timing of the spatial object recognition task in eYFP and SSFO mice. C, Representative locomotor trajectory line graphs, based on nose x- and y-coordinates, of eYFP and SSFO mice during the test session. D, Preference (%) for the moved object during training and testing sessions in eYFP (gray, left) and SSFO (pink, right) mice. Outline-only bars indicate the training day, while filled bars represent the test day. n = 9 eYFP mice and 10 SSFO mice. E, Discrimination ratio during the testing session for eYFP (gray, left) and SSFO (pink, right) mice. n = 9 eYFP mice and 10 SSFO mice. F, Quantification of behavioral activity during the first 15 min of optogenetic activation in eYFP (gray, left) and SSFO (pink, right) mice. Five behaviors were identified: grooming, walking, rearing, surveying, and inactive. n = 10 eYFP mice and 12 SSFO mice. G, Plasma corticosterone levels in SSFO mice before optogenetic activation (base, squares on the left) and immediately after 1 h of optogenetic activation of CRHPVN neurons (SSFO, circles on the right). n = 3 SSFO mice. Bars, averages across mice; dots and lines, individual mice; error bars, SEM. Two-way mixed model ANOVA followed by t tests with Bonferroni’s correction, unpaired t tests, and paired t tests, ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05. Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1.
Figure 1-1
Detailed statistical analyses of all datasets, related to Fig. 1-9 and Extended Data Fig. 2-1, 5-1, and 9-1. Download Figure 1-1, DOCX file.
Next, we examined whether the stimulation of CRHPVN neurons after training in the SOR task alters overall behavior. During the first 15 min of the SSFO stimulation period, we observed a significant increase in grooming behavior, consistent with previous studies, and a decrease in inactivity (Fig. 1F; t tests; grooming, p = 2.0 × 10−4, inactive, p < 1.0 × 10−4; Füzesi et al., 2016; Mitchell et al., 2024). Additionally, we confirmed that plasma corticosterone levels were significantly elevated following optogenetic stimulation of CRHPVN neurons (Fig. 1G; t test, p = 0.028; Füzesi et al., 2016).
We also examined how the activation of CRHPVN neurons after training impacts sleep. Several studies have highlighted critical windows for memory consolidation, suggesting that sleep disturbances or the manipulation of neural activity within the first 4–6 h after the final training session most severely impair memory consolidation (Smith and Rose, 1996; Smith et al., 1998; Graves et al., 2003; Palchykova et al., 2006; Prince et al., 2014; Bayer and Bertoglio, 2020; Hong et al., 2024). We found that the amount of wakefulness was significantly increased and non-rapid eye movement (NREM) sleep was reduced during the first 3 h of the sleep recording session following training (comprised of the 1 h optogenetic stimulation interval and the following 2 h sleep interval; Fig. 2A,B; t tests: NREM, p = 0.011; REM, p = 0.172; wake, p = 0.011), but not during the last 3 h (Hours 3–6; Fig. 2E). During the first hour, optogenetic activation of CRHPVN neurons significantly increased wakefulness, primarily by increasing the duration of wake episodes and reducing their frequency, concurrently decreasing NREM sleep amount and frequency and rapid eye movement (REM) sleep amount (Fig. 2C, t tests, percentages: NREM, p = 2.7 × 10−5; REM, p = 0.017; wake, p = 1.9 × 10−5; Extended Data Fig. 2-1B, t tests, frequency: NREM, p = 0.001; wake, p = 0.002; duration: wake, p = 0.009). During the following 2 h sleep interval (after laser stimulation ended), NREM sleep was increased (Fig. 2D, t test; NREM, p = 0.019). Next, we investigated whether changes in sleep–wake states in the first 3 h following training were correlated with memory performance. The correlation between the percentage of wakefulness and the discrimination ratio was marginally significant, suggesting that the overall increase in wakefulness and decrease in sleep may contribute to the impaired performance in the SOR task (Fig. 2F; linear mixed model, wake percentage: z = −1.716, p = 0.086; group comparison: z = −0.572, p = 0.567).
Activation of CRHPVN neurons disrupts sleep. A, Example recording of an eYFP (top) and SSFO mouse (bottom) during 1 h of laser stimulation, laser indicated by the light blue lines, and sleep for 2 h following laser stimulation. Shown are EEG power spectra, EMG amplitude, and color-coded brain states. B, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the combined 3 h period immediately following training on the memory task for eYFP (gray, left) and SSFO (pink, right) mice. n = 10 eYFP mice and 13 SSFO mice. C, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the 1 h optogenetic activation period immediately following training on the memory task for eYFP (gray, left) and SSFO (pink, right) mice. n = 10 eYFP mice and 13 SSFO mice. D, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the 2 h period immediately following, but not including, 1 h of optogenetic activation for eYFP (gray, left) and SSFO (pink, right) mice. n = 10 eYFP mice and 13 SSFO mice. E, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the final 3 h (Hours 3–6) of the recording period for eYFP (gray, left) and SSFO (pink, right) mice. n = 9 eYFP mice and 9 SSFO mice. F, Linear mixed model analysis illustrating the relationship between discrimination ratio (y-axis) and percentages of NREM, REM, and wake states (during the 0–3 h period; x-axis) in eYFP and SSFO mice. Each panel represents a different sleep state. Gray dots indicate individual data points for eYFP mice (n = 9) and pink dots for SSFO mice (n = 10). Lines represent fitted values from the linear mixed effects model. Error clouds represent the standard error of the mean from residuals. Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1. A detailed summary of the duration and frequency of each brain state for the specified time intervals is provided in Extended Data Figure 2-1. Bars, averages across mice; dots, individual mice; error bars, SEM. Unpaired t tests, ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05.
Figure 2-1
Effect of CRHPVN activation on sleep-wake states.
(A) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the combined 3-hour period immediately following training on the memory task for eYFP (gray, left) and SSFO (pink, right) mice. n = 10 eYFP mice and 13 SSFO mice.
(B) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the 1-hour optogenetic activation period immediately following training on the memory task for eYFP (gray, left) and SSFO (pink, right) mice. n = 10 eYFP mice and 13 SSFO mice.
(C) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the 2-hour period immediately following, but not including, 1 hour of optogenetic activation for eYFP (gray, left) and SSFO (pink, right) mice. n = 10 eYFP mice and 13 SSFO mice.
(D) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the final 3 hours (hours 3 to 6) of the recording period for eYFP (gray, left) and SSFO (pink, right) mice. n = 9 eYFP mice and 9 SSFO mice.
Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1.
Bars, averages across mice; dots, individual mice; error bars, s.e.m. Unpaired t-tests, ***P < 0.001; **P < 0.01; *P < 0.05. Download Figure 2-1, TIF file.
Inhibition of CRHPVN neurons during stress improves spatial object recognition memory and sleep
Next, we investigated whether the activity of CRHPVN neurons contributes to the memory impairments and disturbed sleep observed after stress. To first assess the impact of stress on the consolidation of spatial memory, we subjected mice to 1 h of acute restraint stress immediately after the final training session in the SOR task (Fig. 3A). Control mice, which received no manipulations, exhibited a significant increase in their preference for the moved object and a positive discrimination ratio [Fig. 3B–D; mixed ANOVA: group, p = 0.009; training vs test, p = 0.110; interaction, p = 0.043; t tests with Bonferroni’s correction: control (training vs test), p = 0.015]. In contrast, mice subjected to acute restraint stress showed no preference for the moved object and a discrimination ratio close to zero [Fig. 3B–D; t tests with Bonferroni’s correction: stress (training vs test), p > 0.999]. The test day preference and discrimination ratio were significantly lower in restraint stress mice compared with control mice [Fig. 3C,D; t tests with Bonferroni’s correction: test (control vs stress), p = 0.002; t test: discrimination ratio, p = 0.012).
Restraint stress impairs spatial object recognition memory. A, Schematic indicating the timing of the spatial object recognition task in no restraint stress (control) and restraint stress (stress) mice. The arrow indicates the timing of restraint stress. B, Representative locomotor trajectory line graphs, based on nose x- and y-coordinates, of control and stress mice during the test session. C, Preference (%) for the moved object during training and testing sessions in control (gray, left) and restraint stress (mint, right) mice. Outline-only bars indicate the training day, while filled bars represent the test day. n = 16 control mice and 12 restraint stress mice. D, Discrimination ratio during the testing session for control (gray, left) and restraint stress (mint, right) mice. n = 16 control mice and 12 restraint stress mice. Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1. Bars, averages across mice; dots and lines, individual mice; error bars, SEM. Two-way mixed model ANOVA followed by t tests with Bonferroni’s correction and unpaired t tests, **p < 0.01; *p < 0.05.
Subsequently, we tested whether optogenetic inhibition of CRHPVN neurons during restraint stress is sufficient to mitigate the memory performance deficits and sleep disturbances caused by acute restraint stress. CRH-Cre mice were bilaterally injected with an AAV encoding the Cre-inducible bistable chloride channel, SwiChR++ (AAV2-EF1ɑ-DIO-SwiChR++-eYFP) or eYFP (AAV2-EF1ɑ-DIO-eYFP) in the PVN, followed by optic fiber implantation above the injection sites (Fig. 4A; Berndt et al., 2016). Immediately following training, CRHPVN neurons in these mice were optogenetically inhibited during 1 h of restraint stress (1 s step pulses at 120–240 s intervals, 4 mW, ZT1.5–2.5), and their sleep–wake states were recorded afterward (ZT2.5–7.5; Fig. 4B). We found that eYFP-stress mice exhibited no discernible preference between the moved and familiar objects on the test day, indicating impaired performance in the memory task [Fig. 4C,D; mixed ANOVA: virus, p = 0.054; training vs test, p = 0.002; interaction, p = 0.005; t tests with Bonferroni’s correction: eYFP-stress (training vs test), p > 0.999]. In contrast, SwiChR++-mediated inhibition of CRHPVN neurons during stress resulted in a significant preference for the moved object [Fig. 4C,D; t tests with Bonferroni’s correction: SwiChR++-stress (training vs test), p = 3.0 × 10−4]. The test day preference and discrimination ratio of the SwiChR++-stress mice were significantly higher than the eYFP-stress mice, indicating improved memory task performance [Fig. 4D,E; t tests with Bonferroni’s correction: test (eYFP-stress vs SwiChR++-stress): p = 0.002; t test: discrimination ratio, p = 0.003]. This suggests that the activity of CRHPVN neurons during restraint stress following training contributes to the observed memory impairments.
Inhibition of CRHPVN neurons during restraint stress improves spatial object recognition memory. A, Top left, Schematic of optogenetic inhibition experiments with simultaneous EEG and EMG recordings. Top right, Fluorescence image of the PVN in a CRH-Cre mouse injected with AAV2-Ef1α-DIO-SwiChR++-eYFP into the PVN. Scale bar, 300 µm. Bottom, Location of optic fiber tracts. B, Schematic indicating the timing of the SOR task in eYFP-stress and SwiChR++-stress mice. C, Representative locomotor trajectory line graphs, based on nose x- and y-coordinates, of eYFP-stress and SwiChR++-stress mice during the test session. D, Preference (%) for the moved object during training and testing sessions in eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. Outline-only bars indicate the training day, while filled bars represent the test day. n = 11 eYFP-stress mice and 11 SwiChR++-stress mice. E, Discrimination ratio during the testing session for eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. n = 11 eYFP-stress mice and 11 SwiChR++-stress mice. F, Quantification of behavioral activity during the first 15 min immediately following optogenetic inhibition of CRHPVN neurons during restraint stress in eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. Five behaviors were identified: grooming, walking, rearing, surveying, and inactive. n = 13 eYFP-stress mice and 10 SwiChR++-stress mice. Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1. Bars, averages across mice; dots and lines, individual mice; error bars, SEM. Two-way mixed model ANOVA followed by t tests with Bonferroni’s correction and unpaired t tests, ***p < 0.001; **p < 0.01; *p < 0.05.
Next, we examined whether inhibiting CRHPVN neurons during restraint stress impacts stress-induced behaviors (Fig. 4F; Füzesi et al., 2016; Mitchell et al., 2024). Optogenetic inhibition of CRHPVN neurons during stress significantly reduced grooming and increased the time spent walking, suggesting that silencing CRHPVN neurons during stress promotes a shift in behavioral states, potentially reflecting an attenuated stress response (Fig. 4F; t tests: grooming, p = 0.041; walking, p = 0.015).
We then investigated whether inhibiting CRHPVN neurons during restraint stress impacts sleep–wake states (Fig. 5A–F). During restraint stress (Hours 0–1), both eYFP- and SwiChR++-stress mice were awake for the entire 1 h period (Fig. 5C). During the subsequent 2 h sleep interval (Hours 1–3), SwiChR++-stress mice exhibited slightly more NREM and REM sleep and reduced wakefulness compared with eYFP-stress mice, contributing to an overall tendency toward increased NREM and REM sleep across the entire 3 h recording period (Fig. 5B,D; t tests, p = 0.089, 0.222, and 0.080 for NREM, REM, and wake, respectively). In the subsequent sleep interval (Hours 3–6), we found that SwiChR++-stress mice had slightly, but significantly, more NREM and REM sleep and less wakefulness (Fig. 5E; t tests, p = 0.038, 0.030, and 0.009 for NREM, REM, and wake, respectively). The significant increase in REM sleep was likely due to an increased REM episode frequency (Extended Data Fig. 5-1C; t test, p = 0.017). The percentage of each brain state during the 3 h period following training and the discrimination ratio were not significantly correlated (Fig. 5F). These results indicate that although inhibiting CRHPVN neurons during stress improves memory and sleep, the increase in sleep may not directly contribute to the memory improvement. This suggests that CRHPVN neuron activity during stress may influence sleep and memory through independent processes.
Inhibition of CRHPVN neurons during restraint stress improves sleep. A, Example recording of an eYFP-stress (top) and SwiChR++-stress mouse (bottom) during 1 h laser stimulation, laser indicated by the light blue lines, during restraint stress and sleep for 2 h following laser stimulation. Shown are EEG power spectra, EMG amplitude, and color-coded brain states. B, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the combined 3 h period immediately following training on the SOR task for eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. n = 12 eYFP-stress mice and 14 SwiChR++-stress mice. C, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the 1 h period of CRHPVN inhibition during stress after training on the SOR memory task for eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. n = 12 eYFP-stress mice and 14 SwiChR++-stress mice. All mice in both groups spent the entire 1 h stress period awake. D, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the 2 h period immediately following, but not including, 1 h of CRHPVN optogenetic inhibition during restraint stress for eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. n = 12 eYFP-stress mice and 14 SwiChR++-stress mice. E, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the final 3 h (Hours 3–6) of the recording period for eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. n = 12 eYFP-stress mice and 14 SwiChR++-stress mice. F, Linear mixed model analysis illustrating the relationship between percentages of NREM, REM, and wake states (during the 0–3 h period; x-axis) and discrimination ratio (y-axis) in eYFP-stress and SwiChR++-stress mice. Each panel represents a different sleep state. Gray dots indicate individual data points for eYFP-stress mice (n = 10) and blue dots for SwiChR++-stress mice (n = 11). Lines represent fitted values from the linear mixed effects model. Error clouds represent the standard error of the mean from residuals. Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1. A detailed summary of the duration and frequency of each brain state for the specified time intervals is provided in Extended Data Figure 5-1. Bars, averages across mice; dots, individual mice; error bars, SEM. Unpaired t tests, **p < 0.01; *p < 0.05.
Figure 5-1
Effect of inhibition of CRHPVN neurons during restraint stress on sleep-wake states.
(A) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the combined 3-hour period immediately following training on the memory task for eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. n = 12 eYFP-stress mice and 14 SwiChR++-stress mice.
(B) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the 2-hour period immediately following, but not including, 1 hour of optogenetic inhibition for eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. n = 12 eYFP-stress mice and 14 SwiChR++-stress mice.
(C) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the final 3 hours (hours 3 to 6) of the recording period for eYFP-stress (gray, left) and SwiChR++-stress (blue, right) mice. n = 12 eYFP-stress mice and 14 SwiChR++-stress mice.
Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1.
Bars, averages across mice; dots, individual mice; error bars, s.e.m. Unpaired t-tests, *P < 0.05. Download Figure 5-1, TIF file.
Inhibition of CRHPVN neurons does not affect spatial object recognition memory and sleep
Next, we tested whether inhibiting CRHPVN neurons alone affects memory and sleep. CRH-Cre mice were bilaterally injected with AAV2-EF1ɑ-DIO-SwiChR++-eYFP or AAV2-EF1ɑ-DIO-eYFP in the PVN, followed by optic fiber implantation (Fig. 6A). Immediately following training, both eYFP and SwiChR++ mice received laser stimulation (1 s step pulses at 120–240 s intervals, 4 mW, ZT1.5–2.5; Fig. 6B). We found that both groups exhibited an increased preference for the moved object on the test day [Fig. 6C,D; mixed ANOVA: virus, p = 0.098; training vs test, p < 1.0 × 10−4; interaction, p = 0.974; t tests with Bonferroni’s correction: eYFP (training vs test), p = 0.004; SwiChR++ (training vs test), p = 0.009]. The test day preference and discrimination ratio of the SwiChR++ mice did not significantly differ from the eYFP mice [Fig. 6D,E; t tests with Bonferroni’s correction: test (eYFP vs SwiChR++), p = 0.437; t test: discrimination ratio, p = 0.332). Both groups of mice spent similar amounts of time in NREM sleep, REM sleep, and wakefulness during the 3 h period following training (comprised of the 1 h optogenetic stimulation interval and the following 2 h sleep interval; Fig. 7A,B; t tests: NREM, p = 0.434; REM, p = 0.298; wake, p = 0.351) as well as the other time intervals (Hours 0–1, 1–3, and 3–6; Fig. 7C–E). This indicates that optogenetic inhibition of CRHPVN neurons alone does not significantly alter spatial memory task performance or sleep–wake states (Figs. 6C–E and 7A–E).
Inhibition of CRHPVN neurons does not affect spatial object recognition memory. A, Top left, Schematic of optogenetic inhibition experiments with simultaneous EEG and EMG recordings. Top right, Fluorescence image of the PVN in a CRH-Cre mouse injected with AAV2-Ef1α-DIO-SwiChR++-eYFP into the PVN. Scale bar, 400 µm. Bottom, Location of optic fiber tracts. B, Schematic indicating the timing of the SOR task in eYFP and SwiChR++ mice. C, Representative locomotor trajectory line graphs, based on nose x- and y-coordinates, of eYFP and SwiChR++ mice during the test session. D, Preference (%) for the moved object during training and testing sessions in eYFP (gray, left) and SwiChR++ (green, right) mice. Outline-only bars indicate the training day, while filled bars represent the test day. n = 14 eYFP mice and 12 SwiChR++ mice. E, Discrimination ratio during the testing session for eYFP (gray, left) and SwiChR++ (green, right) mice. n = 14 eYFP mice and 12 SwiChR++ mice. Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1. Bars, averages across mice; dots and lines, individual mice; error bars, SEM. Two-way mixed model ANOVA followed by t tests with Bonferroni’s correction and unpaired t tests, **p < 0.01; *p < 0.05.
Inhibition of CRHPVN neurons does not affect sleep. A, Example recording of an eYFP (top) and SwiChR++ mouse (bottom) during 1 h laser stimulation, laser indicated by the light blue lines, and sleep for 2 h following laser stimulation. Shown are EEG power spectra, EMG amplitude, and color-coded brain states. B, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the combined 3 h period immediately following training on the SOR task for eYFP and SwiChR++ mice. n = 17 eYFP mice and 19 SwiChR++ mice. C, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the 1 h CRHPVN optogenetic inhibition period immediately following training on the memory task for eYFP (gray, left) and SwiChR++ (green, right) mice. n = 17 eYFP mice and 19 SwiChR++ mice. D, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the 2 h period immediately following, but not including, 1 h of CRHPVN optogenetic inhibition for eYFP (gray, left) and SwiChR++ (green, right) mice. n = 17 eYFP mice and 19 SwiChR++ mice. E, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the final 3 h (Hours 3–6) of the recording period for eYFP (gray, left) and SwiChR++ (green, right) mice. n = 17 eYFP mice and 19 SwiChR++ mice. Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1. Bars, averages across mice; dots, individual mice; error bars, SEM. Unpaired t tests.
Inhibition of CRHPVN→LH projections during restraint stress improves spatial object recognition memory and sleep
CRHPVN neurons project to the LH, a brain region implicated in the regulation of stress, sleep–wake states, and memory (Rho and Swanson, 1987; Graebner et al., 2015; Bonnavion et al., 2016; Füzesi et al., 2016; Li et al., 2020; Ono et al., 2020; Mitchell et al., 2024). We found that both CRHPVN activation and restraint stress increased the number of c-Fos+ cells in the LH, suggesting that the LH is a downstream target that could be involved in mediating the effects of CRHPVN neuron activation and stress on memory and sleep (Fig. 8A–F).
Inhibition of CRHPVN→LH projections during restraint stress improves spatial object recognition memory. A, Schematic indicating the timing of restraint stress and perfusion for c-Fos experiments. B, Representative images of c-Fos expression (green) in the lateral hypothalamus of no restraint stress (control) and restraint stress (stress) mice. Scale bar, 300 µm. C, Bar graphs indicating the number of c-Fos–positive cells, bilateral, counted in the LH averaged from three brain sections/mouse for control (green outline-only bars, left) and restraint stress (green filled bars, right) mice. n = 3 control mice and 3 restraint stress mice. D, Schematic indicating the timing of CRHPVN neuron activation (SSFO) and perfusion for c-Fos experiments. E, Representative images of c-Fos expression (green) in the LH of eYFP control and CRHPVN activation (SSFO) mice. Scale bar, 300 µm. F, Bar graphs indicating the number of c-Fos–positive cells, unilateral (due to unilateral virus expression), counted in the LH averaged from two brain sections/mouse for eYFP control (green outline-only bars, left) and SSFO (green filled bars, right) mice. n = 4 eYFP mice and 5 SSFO mice. G, Top left, Schematic of optogenetic inhibition experiments with simultaneous EEG and EMG recordings. Top right, Fluorescence image of the PVN in a CRH-Cre mouse injected with rAAV2-retro-Ef1α-DIO-SwiChR++-eYFP into the LH. Scale bar, 300 µm. Bottom, Location of optic fiber tracts. H, Schematic indicating the timing of the SOR task in retro-eYFP-stress and retro-SwiChR++-stress mice. I, Representative locomotor trajectory line graphs, based on nose x- and y-coordinates, of retro-eYFP-stress and retro-SwiChR++-stress mice during the test session. J, Preference (%) for the moved object during training and testing sessions in retro-eYFP-stress (gray, left) and retro-SwiChR++-stress (purple, right) mice. Outline-only bars indicate the training day, while filled bars represent the test day. n = 11 retro-eYFP-stress mice and 17 retro-SwiChR++-stress mice. K, Discrimination ratio during the testing session for retro-eYFP-stress (gray, left) and retro-SwiChR++-stress (purple, right) mice. n = 11 retro-eYFP-stress mice and 17 retro-SwiChR++-stress mice. Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1. Bars, averages across mice; dots and lines, individual mice; error bars, SEM. Two-way mixed model ANOVA followed by t tests with Bonferroni’s correction and unpaired t tests, ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05.
We tested whether inhibiting CRHPVN→LH projections during restraint stress was sufficient to mitigate the memory deficits and sleep disturbances observed after stress. CRH-Cre mice were bilaterally injected with retrograde AAVs encoding Cre-inducible SwiChR++ (rAAV2-retro-Ef1α-DIO-SwiChR++-eYFP) or eYFP (rAAV2-retro-Ef1α-DIO-eYFP) in the LH, followed by optic fiber implantation above the PVN (Fig. 8G). Immediately following training, we optogenetically inhibited CRHPVN→LH projections during 1 h of restraint stress (1 s step pulses at 120–240 s intervals, 4 mW, ZT1.5–2.5) and recorded sleep–wake states afterward (ZT2.5–7.5; Fig. 8H). Twenty-four hours later, we found that retro-eYFP-stress mice exhibited no discernible preference between the moved and familiar objects on the test day, indicating impaired memory performance after stress [Fig. 8I,J; mixed ANOVA: virus, p = 0.020; training vs test, p = 0.003; interaction, p = 0.005; t tests with Bonferroni’s correction: retro-eYFP-stress (training vs test), p > 0.999]. In contrast, inhibition of CRHPVN→LH projections during stress immediately following training resulted in a significant increase in the preference for the moved object [Fig. 8I,J; t tests with Bonferroni’s correction: retro-SwiChR++-stress (training vs test), p < 1.0 × 10−4]. The test day preference and discrimination ratio of the retro-SwiChR++-stress mice were significantly higher than the retro-eYFP-stress mice, indicating an improved memory task performance [Fig. 8J,K; t tests with Bonferroni’s correction: test (retro-eYFP-stress vs retro-SwiChR++-stress), p = 6.0 × 10−4; t test: discrimination ratio, p = 0.004]. This suggests that the activity of CRHPVN→LH projections during restraint stress contributes to the memory impairments observed after stress.
Next, we investigated whether inhibiting CRHPVN→LH projections during stress impacts sleep–wake states (Fig. 9A–F). During the restraint stress combined with optogenetic manipulation, both retro-eYFP-stress and retro-SwiChR++-stress mice were awake for the entire 1 h period (Fig. 9C). During the subsequent 2 h sleep interval (Hours 1–3), retro-SwiChR++-stress mice had significantly more NREM sleep and spent less time awake compared with retro-eYFP-stress mice, contributing to an overall increase in NREM sleep and decreased wakefulness during the entire 3 h recording period (Fig. 9A–D; t tests: NREM, p = 0.048; REM, p = 0.465; wake, p = 0.040). In the subsequent sleep interval (Hours 3–6), we did not find any differences in the amount of sleep and wakefulness between groups (Fig. 9E; t tests: NREM, p = 0.775; REM, p = 0.187; wake, p = 0.763). Our analyses did not reveal any significant differences in sleep architecture, including the frequency and duration of sleep and wake episodes, across any time intervals (Extended Data Fig. 9-1). The percentage of each brain state during the 3 h period following training and the discrimination ratio were not significantly correlated (Fig. 9F), suggesting that the increased sleep following inhibition of CRHPVN→LH projections during stress may not contribute to improved memory. Thus, similar to the effects observed with CRHPVN cell body inhibition during stress, CRHPVN projections to the LH may regulate sleep and memory through independent mechanisms. Taken together, these results suggest that the effects of CRHPVN neuronal stimulation on impairments in spatial memory and sleep are, at least in part, mediated by their projections to the LH.
Inhibition of CRHPVN→LH projections during restraint stress improves sleep. A, Example recording of a retro-eYFP-stress (top) and retro-SwiChR++-stress mouse (bottom) during 1 h laser stimulation, laser indicated by the light blue lines, during restraint stress and sleep for 2 h following laser stimulation. Shown are EEG power spectra, EMG amplitude, and color-coded brain states. B, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the combined 3 h period immediately following training on the SOR task for retro-eYFP-stress (gray, left) and retro-SwiChR++-stress (purple, right) mice. n = 12 retro-eYFP-stress mice and 24 retro-SwiChR++-stress mice. C, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the 1 h period of CRHPVN→LH projection inhibition during stress after training on the SOR memory task for retro-eYFP-stress (gray, left) and retro-SwiChR++-stress (purple, right) mice. n = 12 retro-eYFP-stress mice and 24 retro-SwiChR++-stress mice. All mice in both groups spent the entire 1 h stress period awake. D, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the 2 h period immediately following, but not including, 1 h of CRHPVN→LH projection optogenetic inhibition during restraint stress for retro-eYFP-stress (gray, left) and retro-SwiChR++-stress (purple, right) mice. n = 12 retro-eYFP-stress mice and 24 retro-SwiChR++-stress mice. E, Percentage of time spent in NREM sleep, REM sleep, and wakefulness during the final 3 h (Hours 3–6) of the recording period for retro-eYFP-stress (gray, left) and retro-SwiChR++-stress (purple, right) mice. n = 12 retro-eYFP-stress mice and 24 retro-SwiChR++-stress mice. F, Linear mixed model analysis illustrating the relationship between percentages of NREM, REM, and wake states (during the 0–3 h period; x-axis) and discrimination ratio (y-axis) in retro-eYFP-stress and retro-SwiChR++-stress mice. Each panel represents a different sleep state. Gray dots indicate individual data points for retro-eYFP-stress mice (n = 11) and purple dots for retro-SwiChR++-stress mice (n = 17). Lines represent fitted values from the linear mixed effects model. Error clouds represent the standard error of the mean from residuals. Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1. A detailed summary of the duration and frequency of each brain state for the specified time intervals is provided in Extended Data Figure 9-1. Bars, averages across mice; dots, individual mice; error bars, SEM. Unpaired t tests, *p < 0.05.
Figure 9-1
Effect of inhibition of CRHPVN→LH projections during restraint stress on sleep-wake states.
(A) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the combined 3-hour period immediately following training on the memory task for retro-eYFP-stress (gray, left) and retro-SwiChR++-stress (purple, right) mice. n = 12 retro-eYFP-stress mice and 24 retro-SwiChR++-stress mice.
(B) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the 2-hour period immediately following, but not including, 1 hour of optogenetic inhibition for retro-eYFP-stress (gray, left) and retro-SwiChR++-stress (purple, right) mice. n = 12 retro-eYFP-stress mice and 24 retro-SwiChR++-stress mice.
(C) Duration and frequency of NREM sleep, REM sleep, and wake episodes during the final 3 hours (hours 3 to 6) of the recording period for retro-eYFP-stress (gray, left) and retro-SwiChR++-stress (purple, right) mice. n = 12 retro-eYFP-stress mice and 24 retro-SwiChR++-stress mice.
Detailed statistical analyses for all datasets presented in this figure are provided in Extended Data Figure 1-1.
Bars, averages across mice; dots, individual mice; error bars, s.e.m. Unpaired t-tests. Download Figure 9-1, TIF file.
Mapping axon projections of CRHPVN→LH neurons
A previous study demonstrated that CRHPVN neurons send collateral projections to both the median eminence and the LH (Füzesi et al., 2016). To systematically map the collateral projections of CRHPVN→LH neurons, we expressed membrane-bound GFP (mGFP) and synaptophysin-mRuby (SYP-mRuby) selectively in this neuronal population (Fig. 10A). In addition to the LH, these neurons projected to several other brain regions implicated in regulating stress responses, sleep, and neuroendocrine functions, including the ventromedial hypothalamus (VMH), median eminence (ME), periaqueductal gray (PAG), and the locus coeruleus (LC), suggesting that multiple pathways may contribute to their effects on sleep and memory regulation after stress (Fig. 10B–F; Keay and Bandler, 2001; Yin and Gore, 2010; Poe et al., 2020; Khodai and Luckman, 2021; Shao et al., 2023).
Mapping axon projections of CRHPVN→LH neurons. A, Schematic of the axon projection mapping experiment. AAVrg-EF1α-DIO-FlpO was injected into the LH, and AAV8-hSyn-FLExFRT-mGFP-2A-synaptophysin-mRuby was injected into the PVN of CRH-Cre mice. B, Coronal section containing the lateral hypothalamus (LH). C, Coronal section containing the ventromedial hypothalamus (VMH). D, Coronal section containing the median eminence (ME). E, Coronal section containing the periaqueductal gray (PAG). F, Coronal section containing the locus coeruleus (LC). Larger image: scale bar, 200 µm. A region, marked by a white box, is magnified in the inset of each larger image. Scale bar, 10 µm. Red, synaptophysin-mRuby; green, mGFP; yellow, overlap; overlap marked with white triangles in inset.
Discussion
Our study provides insights into how CRHPVN neurons influence spatial memory and sleep–wake patterns in mice after stress. Using optogenetic manipulations, we demonstrated that activating CRHPVN neurons impaired performance on the SOR task, similar to acute restraint stress (Figs. 1–3), while optogenetic inhibition of CRHPVN neurons during restraint stress improved memory and slightly increased NREM and REM sleep compared with eYFP-stress mice (Figs. 4 and 5). Inhibition of CRHPVN neurons alone did not have a significant effect on SOR task performance or sleep–wake states (Figs. 6 and 7). Using c-Fos staining, we found that acute restraint stress and the stimulation of CRHPVN neurons activated neurons in the LH, suggesting that this region is a downstream target of CRHPVN neurons (Fig. 8). Optogenetically inhibiting CRHPVN→LH projections during restraint stress improved SOR task performance and significantly increased NREM sleep and reduced wakefulness compared with retro-eYFP-stress mice (Figs. 8 and 9). In addition to the LH, CRHPVN neurons projected to several other brain regions, suggesting that multiple pathways may be involved in regulating sleep and memory following stress (Fig. 10). Our results demonstrate that CRHPVN neurons contribute to stress-induced cognitive deficits and sleep disturbances, partially through their projections to the LH.
The activation of CRHPVN neurons is a key component of the body’s response to stress. Our study reveals that the activation of CRHPVN neurons significantly increases wakefulness and triggers stress-associated behaviors, like grooming, in the absence of stress, confirming earlier findings (Füzesi et al., 2016; Li et al., 2020; Ono et al., 2020; Mitchell et al., 2024). Unlike the sleep fragmentation seen after acute social defeat stress or LC stimulation (Antila et al., 2022), CRHPVN neuron activation strongly promoted wakefulness without inducing sleep fragmentation, suggesting that different neural circuits mediate these effects. For instance, noradrenergic neurons in the LC may influence sleep continuity, while CRHPVN neurons may regulate total sleep duration. Furthermore, we found that activation of CRHPVN neurons immediately following training impaired performance on the SOR task. CRHPVN neuron stimulation-induced wakefulness was correlated with impaired performance on the memory task, suggesting that an overall decrease in sleep may contribute to disrupted memory consolidation. Sleep deprivation impairs hippocampus-dependent memory (Walsh et al., 2011; Abel et al., 2013; Havekes et al., 2015; Kreutzmann et al., 2015; Klinzing et al., 2019; Heckman et al., 2020). Five or 6 h of total sleep deprivation is sufficient to impair the acquisition, consolidation, and retrieval of object-location memories, as well as the consolidation of contextual fear memory (Graves et al., 2003; Heckman et al., 2020).
Activation of CRHPVN neurons increases circulating glucocorticoid levels, which have been shown to play an important role in cognitive function, including memory consolidation processes, with effects varying depending on the timing and dose of glucocorticoids (Roozendaal, 2000, 2002). Studies have found that administration of corticosterone in rodents blocks the induction of long-term potentiation in the hippocampus (Dubrovsky et al., 1987; Filioini et al., 1991; Lupien and McEwen, 1997; Lupien and Lepage, 2001). Consistent with other studies, we showed that activation of CRHPVN neurons results in a significant increase in plasma corticosterone levels (Füzesi et al., 2016). Integrating these findings, memory impairments following stress or CRHPVN neuron activation might result from reduced sleep, elevated corticosterone levels, or a combination of these factors. Future studies could utilize sleep deprivation paradigms and glucocorticoid receptor antagonists to further clarify these interactions.
Furthermore, our results indicate that the effect of CRHPVN neuron activation on memory deficits may be partially mediated by their projections to the LH. CRHPVN neurons express mRNA for vesicular glutamate transporter 2 (VGluT2) as a substrate for fast synaptic transmission, which suggests that the CRHPVN→LH effects on memory and sleep might rely on excitatory glutamatergic signaling in the LH in addition to CRH peptide release (Hrabovszky et al., 2005). Moreover, genetic ablation of LH hypocretin/orexin neurons or crh gene knockdown significantly shortens the latency to NREM sleep onset following stress, suggesting a critical role for CRH and LH hypocretin/orexin neurons in mediating not only stress-induced sleep disturbances but also, potentially, memory deficits (Li et al., 2020).
The LH contains a variety of neurons expressing distinct molecular markers (Mickelsen et al., 2019). GABAergic neurons (Herrera et al., 2016; Venner et al., 2016), glutamatergic neurons (Wang et al., 2021; Smith et al., 2024), neurotensin (Naganuma et al., 2019), and CamKIIɑ-expressing neurons (Heiss et al., 2024) promote wakefulness. In the LH, activation of MCH neurons after the training period significantly impairs NOR memory (Izawa et al., 2019). Activation of hypocretin/orexin neurons in the LH promotes the release of plasma corticosterone (Bonnavion et al., 2015) and results in sleep fragmentation and impaired memory in the NOR task (Rolls et al., 2011). Given that hypocretin/orexin labels only a subset of LH neurons that become activated after CRHPVN stimulation, it remains to be studied which types of neurons in the LH are innervated by CRHPVN projections and activated by stress in order to mediate the stress-induced impairments in memory and sleep (Mitchell et al., 2024).
CRHPVN neurons, which are inhibited by rewards like sucrose, play an important role in motivational drive (Yuan et al., 2019). Optogenetic activation of these neurons and their projections to the LH increases acute stress-related behaviors and causes long-lasting deficits in sucrose preference (Mitchell et al., 2024). Thus, the activation of CRHPVN neurons might impair motivational drive and, subsequently, impair performance on the SOR task. Conversely, inhibition of these neurons and their projections to the LH during stress might have a protective effect, resulting in an improved memory task performance compared with mice that experienced stress without optogenetic inhibition.
Stress impairs hippocampal functions and circuits (Kim et al., 2007, 2012; Tomar et al., 2021; Tomar and McHugh, 2022). In particular, stress decreases the stability of firing rates of place cells in the hippocampus, accompanied by impairments of spatial memory consolidation (Kim et al., 2007). Similarly, amygdala stimulation in rats alters the spatial correlation of place maps and increases variability in the firing rate, whereas corticosterone injection does not (Kim et al., 2012). This suggests that the enhanced amygdalar activity, but not the elevated level of corticosterone, leads to a destabilization of spatial representations within the hippocampus by modifying the firing rates of place cells, mirroring the effects typically seen after behavioral stress (Kim et al., 2012). Hippocampal cells exhibit highly coordinated activity during sleep and resting states and sharp-wave ripples play a key role in memory consolidation (Buzsáki, 1989; Girardeau et al., 2009; Ramadan et al., 2009; Girardeau and Zugaro, 2011; Fernández-Ruiz et al., 2019). Neuromodulators such as acetylcholine and oxytocin are released in the hippocampus in a brain state-dependent manner, suggesting their vital role in regulating brain state-dependent memory functions (Zhang et al., 2024). We found that activation of CRHPVN neurons increased wakefulness, and this increase was correlated with impaired memory. However, this relationship weakened in the context of restraint stress. Inhibition of CRHPVN neurons and their projections to the LH during stress improved both memory and sleep, but these increases were not significantly correlated with one another, suggesting that these effects are mediated by separate pathways. Given that hippocampal pathways are crucial for regulating spatial memory, the memory improvements may be mediated by postsynaptic LH cells projecting to the hippocampal circuits, whereas sleep regulation might rely on sleep-regulatory neurons in the hypothalamus and brainstem, which are not directly connected to the hippocampal circuits (Bittencourt et al., 1992; Peyron et al., 1998; Marcus et al., 2001; Bittencourt, 2011; Izawa et al., 2019). Future studies could explore other factors disrupted by stress, such as sleep-related electrophysiological features and neurotransmitter dynamics, to better understand their contributions to sleep and memory impairments. This approach may offer deeper insights into the complex interplay between stress, memory, and sleep.
While we observed that 1 h of CRHPVN inhibition during the light phase after training in the SOR task did not significantly alter sleep–wake states, another study found that chemogenetic inhibition of these neurons during the dark phase led to a reduction in spontaneous wakefulness (Ono et al., 2020). Given that the activity of CRHPVN neurons is under circadian influence, it remains to be investigated whether their activity differentially modulates sleep and memory processes depending on the circadian cycle, especially after stress.
Together, our findings elucidate the role of CRHPVN neurons and their projections to the LH in modulating the impact of stress on memory and sleep–wake states. Cognitive impairments and sleep disturbances are hallmark features in psychiatric disorders such as post-traumatic stress disorder and major depressive disorder, often manifesting before clinical diagnosis (Ford and Kamerow, 1989; Chang et al., 1997; Koren et al., 2002; Neckelmann et al., 2007; Spoormaker and Montgomery, 2008; Meerlo et al., 2015; Pearson et al., 2023). By identifying the specific contributions of CRHPVN neurons to these dysfunctions, our study provides a foundation for targeted interventions that could mitigate such impairments. Therapeutic strategies that specifically modulate this pathway could not only improve cognitive and sleep outcomes but could also potentially delay the progression of related psychiatric conditions, offering advancements in treating stress-related disorders.
Data Availability
The code used for data analysis is publicly available at https://github.com/tortugar/Lab. All the data are available from the corresponding author upon reasonable request.
Footnotes
We thank the members of the Chung and Weber labs for their helpful discussion. This work was funded by the National Institute of Mental Health (R01-MH-136491), the Whitehall Foundation, the Alfred P. Sloan Foundation, the T32 Predoctoral Training Grant in Pharmacology (T32GM008076, A.W.), and the NIH Individual F31 Fellowship from the National Heart, Lung, and Blood Institute (F31HL160451, A.W.). We thank the Cell and Developmental Biology Microscopy Core (RRID:SCR_022373).
The authors declare no competing financial interests.
- Correspondence should be addressed to Shinjae Chung at shinjaec{at}pennmedicine.upenn.edu.
