A noradrenergic-hypothalamic neural substrate for stress-induced sleep disturbances

Abstract

In our daily life, we are exposed to uncontrollable and stressful events that disrupt our sleep. However, the underlying neural mechanisms deteriorating the quality of non-rapid eye movement sleep (NREMs) and REM sleep are largely unknown. Here, we show in mice that acute psychosocial stress disrupts sleep by increasing brief arousals (microarousals [MAs]), reducing sleep spindles, and impairing infraslow oscillations in the spindle band of the electroencephalogram during NREMs, while reducing REMs. This poor sleep quality was reflected in an increased number of calcium transients in the activity of noradrenergic (NE) neurons in the locus coeruleus (LC) during NREMs. Opto- and chemogenetic LC-NE activation in naïve mice is sufficient to change the sleep microarchitecture similar to stress. Conversely, chemogenetically inhibiting LC-NE neurons reduced MAs during NREMs and normalized their number after stress. Specifically inhibiting LC-NE neurons projecting to the preoptic area of the hypothalamus (POA) decreased MAs and enhanced spindles and REMs after stress. Optrode recordings revealed that stimulating LC-NE fibers in the POA indeed suppressed the spiking activity of POA neurons that are activated during sleep spindles and REMs and inactivated during MAs. Our findings reveal that changes in the dynamics of the stress-regulatory LC-NE neurons during sleep negatively affect sleep quality, partially through their interaction with the POA.

Keywords: microarousals; sleep; sleep spindles; stress.

Fig. 1.

LC-NE neurons are activated during spontaneous NREMs and after stress. (A) Top, Schematic of fiber photometry with simultaneous EEG and EMG recordings. Mouse brain figure adapted from the Allen Mouse Brain Atlas (© 2015 Allen Institute for Brain Science). Bottom, Fluorescence image of LC in a DBH-Cre mouse with AAV-FLEX-GCaMP6s injected into the LC. Scale bar, 250 μm. (B) Top, Example fiber photometry recording. Shown are parietal EEG spectrogram, EMG traces, color-coded brain states, and ΔF/F signal. Bottom, Parietal EEG spectrogram, brain states, EEG sigma power (10.5 to 16 Hz, green, parietal EEG), phase of EEG sigma power, and calcium signal during a selected interval (dashed box) at an expanded timescale. The filtered sigma power (dashed line) was used to determine the phase of the sigma power oscillation (middle). Freq, frequency. (C) Non-normalized and Z scored ΔF/F activity during REMs, wake, and NREMs. Bars, averages across mice; lines, individual mice; error bars, SEMs. One-way RM ANOVA followed by pairwise t tests with Bonferroni correction, ***P < 0.001. n = 22 mice. (D) Normalized PSD of sigma power in the parietal EEG and calcium activity during NREMs. The PSD for both signals was calculated only for consolidated bouts of NREMs (NREMs episodes ≧120 s, only interrupted by MAs [i.e., wake episodes ≦ 20 s]) and normalized for each animal by its mean power. Shadings, SEMs. (E) Average calcium activity during a single cycle of the sigma power oscillation in the parietal EEG. Each sigma power cycle was normalized in time, ranging from −π to π rad. Shadings, SEMs. (F) Cross-correlation between calcium activity and parietal EEG delta, theta, or sigma power during NREMs. Shadings, SEMs. (G) Calcium activity changes, parietal EEG and EMG traces, EEG spectrogram, and EMG amplitude at the transition to MAs. Shadings, SEMs. (H) Top, Schematic illustrating fiber photometry for baseline sleep recordings (Left) and the acute social defeat paradigm (Right). The experimental mouse (black mouse) was first exposed to a CD1 mouse (white mouse). Afterward, a fiber photometry recording was performed in the experimental mouse while staying in the cage of the CD1 mouse, with a perforated wall placed between them. Bottom, Fiber photometry recordings during baseline recordings (Left) and after acute social defeat stress (Right). (I) Number of MAs during NREMs for the 4-h recording. (J) Number of sleep spindles during NREMs. Spindles were detected in the frontal EEG. (K) Left, Normalized PSD of the parietal EEG sigma power during NREMs. Shadings, 95% confidence intervals (CIs). Right, Strength of infraslow sigma power oscillation. (L) Left, Number of calcium transients during NREMs. Right, Proportion of calcium transients coinciding with MAs. The algorithm for the detection of calcium transients is described in SI Appendix, Fig. S1D. (M) Left, Normalized PSD of the calcium activity during NREMs. Right, Strength of the infraslow calcium activity oscillation. (I–M) Bars, averages across mice; lines, individual mice; error bars, SEMs. Paired t tests; **P < 0.01; ***P < 0.001. n = 10 mice (SI Appendix, Figs. S1 and S2 and Table S1).

Fig. 2.

LC-NE neuron activation promotes MAs and suppresses spindles and REMs. (A) Left, Schematic of optogenetic experiment. Right, fluorescence image of LC in a DBH-Cre mouse, with AAV-DIO-ChR2-eYFP injected into the LC. Scale bar, 200 μm. (B) Example trial. Shown are parietal EEG spectrogram, EMG traces, color-coded brain states, and EEG and EMG raw traces during selected periods (dashed boxes) on an expanded timescale. Blue shading, laser stimulation interval (3 Hz for 20 s). (C) Percentage of wake, NREMs, REMs, or MAs before and during laser stimulation (3 Hz, 5-, 20-, or 120-s stimulation). The duration of the baseline interval without laser was equal to that of the following laser interval. n = 8 to 12 mice. (D) Percentage of time spent in wake, NREMs, REMs, or MAs before, during, and after laser stimulation (blue shading), averaged from 8 to 12 mice. Shadings, 95% CIs. (E) Effect of laser stimulation on the number of spindles during NREMs. Spindles were detected in the frontal EEG. Shadings, 95% CIs. (F) Normalized histogram of the duration of episodes scored as MA (≦20 s) or wake (>20 s). Shadings, 95% CIs. The area under the curve is 1. (G) Number of spindles at the MA onset during spontaneous NREMs. Spindles were detected in the frontal EEG. Shadings, 95% CIs. (H) Schematic of chemogenetic activation experiment. Left, Coronal diagram of mouse brain. Center, Fluorescence image of LC in a DBH-Cre mouse injected with AAV-DIO-mCherry or AAV-DIO-hM3DGq-mCherry (red) into the LC. Scale bar, 500 μm. Right, Intraperitoneal (IP) injection of saline (SAL) or CNO (1 mg/kg) followed by EEG and EMG recordings. (I) Example SAL (top) and CNO (bottom) session from one hM3DGq mouse. Shown are parietal EEG spectrogram, EMG traces, color-coded brain states, and parietal EEG and EMG traces during selected periods (dashed boxes) on an expanded timescale. (J) Number of MAs during NREMs for the 4-h recording following SAL or CNO injection in mCherry and hM3DGq mice. n = 8 to 9 mice. (K) Probability of NREMs to MA transitions during the peak and trough of the sigma power oscillation and difference of the probability between trough and peak following SAL or CNO injection. (L) Number of spindles during NREMs following SAL or CNO injection. Spindles were detected in the frontal EEG. (M) Left, PSDs of the sigma power in the parietal EEG during NREMs following SAL or CNO injection. For each mouse, the spectral density was normalized by its mean power. Shadings, 95% CIs. Right, Strength of the sigma power oscillation. (N) Percentage of time in wake, NREMs, or REMs following SAL or CNO injection. (C, J–N) Bars, averages across mice; lines, individual mice; error bars, SEMs. Paired t tests for (C) and mixed ANOVA with Bonferroni correction for (J–N); *P < 0.05; **P < 0.01; ***P < 0.001 (SI Appendix, Figs. S3–S5 and Table S1).

Fig. 3.

Inhibiting LC-NE neurons after stress reduces MAs and increases spindles. (A) Left, Schematic depicting injection of AAV-DIO-hM4DGi-mCherry or AAV-DIO-mCherry into the LC. Right, Schematic of pharmacogenetic inhibition experiment combined with acute social defeat stress. For baseline, mice were injected with CNO before the sleep recording (Top). On the stress day, mice were exposed to acute social defeat stress and then injected with CNO. Immediately afterward, their sleep was recorded, while they were separated from the CD1 mouse by a perforated wall (Bottom). (B) Example session from mCherry (Left) or hM4DGi mouse (Right) injected with CNO after stress. Shown are parietal EEG power spectra, EMG traces, and color-coded brain states. (C) Number of MAs during NREMs following CNO injection (2.5 or 5 mg/kg, IP) in mCherry and hM4DGi mice during the 4-h baseline recordings and after stress. n = 16 mCherry and 14 hM4DGi mice. (D) Probability of NREMs to MA transitions during the peak and trough of the sigma power oscillation and the difference in the probability between trough and peak in baseline recordings following CNO injection. Bars, averages across mice; error bars, SEMs. (E) Probability of NREMs to MA transitions, depending on the phase of the sigma power oscillation in mCherry and hM4DGi mice after stress following CNO injection. Bars, averages across mice; error bars, SEMs. t tests; *P < 0.05. (F) Number of spindles during NREMs of baseline and stress recordings in mCherry and hM4DGi mice following CNO injection. Spindles were detected in the frontal EEG. (G) Percentage of time in wake, NREMs, or REMs during baseline and stress recordings in mCherry and hM4DGi mice following CNO injection. (C, F, and G) Bars, averages across mice; dots, individual mice; error bars, SEMs. Mixed ANOVA followed by pairwise t tests with Bonferroni correction; *P < 0.05; **P < 0.01; ***P < 0.001 (SI Appendix, Fig. S6 and Table S1).

Fig. 4.

NE^LC→POA activation promotes MAs and suppresses spindles and REMs. (A) Left, Schematic of fiber photometry recordings to monitor NE levels in the POA using the GRAB_NE sensor. Right, Fluorescence image of POA in a mouse injected with AAV-GRAB_NE3 into the POA. Scale bar, 250 μm. (B) Top, Example fiber photometry recording. Shown are parietal EEG spectrogram, EMG traces, color-coded brain states, and ΔF/F signal. Bottom, Brain states, EEG sigma power (10.5 to 16 Hz, green, parietal EEG), phase of EEG sigma power, and calcium signal during a selected interval (dashed box) at an expanded timescale. The filtered sigma power (dashed line) was used to determine the phase of the sigma power oscillation (middle). (C) Non-normalized and Z scored ΔF/F activity during REMs, wake, and NREMs. Bars, averages across mice; lines, individual mice; error bars, SEMs. One-way RM ANOVA followed by pairwise t tests with Bonferroni correction; ***P < 0.001. n = 9 mice. (D) Average calcium activity during a single cycle of the sigma power oscillation. Each sigma power cycle was normalized in time, ranging from −π to π rad. Shadings, SEMs. (E) Cross-correlation between calcium activity and parietal EEG delta, theta, or sigma power during NREMs. Shadings, SEMs. (F) Left, Schematic of optogenetic experiment for stimulating axonal projections of LC-NE neurons in the POA. Right, Fluorescence image of POA in a DBH-Cre mouse injected with AAV-DIO-ChR2-eYFP into the LC and an optic fiber implanted into the POA. Scale bar, 200 μm. (G) Example trial. Shown are parietal EEG spectrogram, EMG traces, color-coded brain states, and EEG and EMG traces during selected periods (dashed boxes) on an expanded timescale. Blue shading, laser stimulation interval (3 Hz for 20 s). (H) Percentage of wake, NREMs, REMs, or MAs before and during laser stimulation (3 Hz for 20-s, 5 Hz for 20-s, or 5 Hz for 120-s stimulation). n = 7 mice. (I) Percentage of wake, NREMs, REMs, or MAs before, during, and after laser stimulation (blue shading), averaged across 7 mice. Shadings, 95% CIs. (J) Effect of laser stimulation on the number of spindles during NREMs. Shadings, 95% CIs. Spindles were detected in the frontal EEG. (K) Schematic of pharmacogenetic activation experiment to target NE neurons projecting to the POA. Left, Sagittal and coronal diagrams of mouse brain, fluorescence image of LC in a DBH-Cre mouse injected with AAV-retro-DIO-mCherry or AAV-retro-DIO-hM3DGq-mCherry (red) into the POA. 80.5 ± 2.5% of mCherry cells were found in the LC (186/233 cells, n = 3 mice). Scale bar, 250 μm. Right, Schematic depicting sleep recording following SAL or CNO (1 mg/kg) injection. (L) Example SAL (Left) and CNO (Right) session from one retro-hM3DGq mouse. (M) Number of MAs during NREMs for the 4-h recording following SAL or CNO injection in retro-mCherry and retro-hM3DGq mice. n = 6 mice. (N) Probability of NREMs to MA transitions during the peak and trough of the sigma power oscillation, and the difference in the probability between trough and peak following SAL or CNO injection. (O) Number of spindles during NREMs following SAL or CNO injection. Spindles were detected in the frontal EEG. (P) Percentage of time in wake, NREMs, or REMs following SAL or CNO injection. (H, M–P) Bars, averages across mice; lines, individual mice; error bars, SEMs. Paired t tests for (H) and mixed ANOVA with Bonferroni correction for (M-P); *P < 0.05; **P < 0.01; ***P < 0.001 (SI Appendix, Figs. S7 and S8 and Table S1).

Fig. 5.

Inhibiting NE^LC→POA neurons decreases MAs and increases REMs and sleep spindles after stress. (A) Top, Schematic depicting injection of AAV-retro-DIO-hM4DGi-mCherry or AAV-retro-DIO-mCherry into the LC. Bottom, Schematic of pharmacogenetic inhibition experiment combined with acute social defeat stress. (B) Example session from retro-mCherry (Left) or retro-hM4DGi (Right) expressing mouse injected with CNO after stress. Shown are parietal EEG power spectra, EMG traces, and color-coded brain states. (C) Number of MAs during NREMs following CNO injection (2.5 mg/kg, IP) in retro-mCherry and retro-hM4DGi mice during the 4-h baseline recordings and after stress. n = 6 retro-mCherry and 5 retro-hM4DGi mice. (D) Probability of NREMs to MA transitions during the peak and trough of the sigma power oscillation and the difference in the probability between trough and peak in baseline recordings of retro-mCherry and retro-hM4DGi mice following CNO injections. Bars, averages across mice; error bars, SEMs. (E) Probability of NREMs to MA transitions depending on the phase of the sigma power oscillation in retro-mCherry and retro-hM4DGi mice after stress following CNO injections. Bars, averages across mice; error bars, SEMs. t test; ***P < 0.001. (F) Number of spindles during NREMs of baseline and stress recordings in retro-mCherry and retro-hM4DGi mice following CNO injection. Spindles were detected in the frontal EEG. (G) Percentage of time in wake, NREMs, or REMs during baseline and stress recordings in retro-mCherry and retro-hM4DGi mice following CNO injection. (C, F, and G) Bars, averages across mice; dots, individual mice; error bars, SEMs. Mixed ANOVA followed by pairwise t tests with Bonferroni correction; *P < 0.05; **P < 0.01 (SI Appendix, Fig. S6 and Table S1).

Fig. 6.

Impact of NE^LC→POA axonal stimulation in the activity of POA neurons. (A) Schematic of optrode recordings to examine activity changes of POA neurons upon NE^LC→POA axonal stimulation. AAV-DIO-ChR2-eYFP was injected into the LC of a DBH-Cre mouse and an optrode was implanted into the POA, to record single POA units while stimulating LC-NE fibers. (B) Firing rate modulation of 45 inhibited (blue), 17 excited (orange), and 75 laser unresponsive (black) units. W, wake; R, REM; NR, NREM. (C) Left and Center, Spontaneous spike waveforms of example units that were inhibited (Top) or excited (Bottom) by NE^LC→POA laser stimulation. Right, Principal components (PC1 and PC2) of the spike waveforms. (D) Firing rates of 45 inhibited neurons during different brain states. Left, Pie chart showing the distribution of different subtypes within the inhibited units. Based on their brain state–dependent activity, units were subdivided into NR-max (NR), W-max (W), R-NR, R-W cells, and units not modulated by the brain state. Right, Average activity of inhibited units during REMs, NREMs, and wake; gray bars, averages across units; lines, individual units; error bars, SEMs. (E) Top, Spike raster showing multiple laser stimulation trials (10 Hz, 10 ms, 5 s). Bottom, Firing rates before, during, and after laser stimulation. Shadings, SEMs. (F) Firing rates of an example unit suppressed by NE^LC→POA laser stimulation. Shown are parietal EEG spectrogram, EMG traces, and color-coded brain states and the firing rates. (G) R-NR neurons that were significantly inhibited by NE^LC→POA stimulation. Left, Firing rates during different brain states. Gray bars, averages across units; lines, individual units; error bars, SEMs. Center Left, Cross-correlation between firing rates and parietal EEG delta, theta, or sigma power during NREMs. Shadings, 95% CIs. Center Right, Average firing rate during sleep spindles. 0 s corresponds to the sleep spindle onset. Spindles were detected in the parietal EEG. Shadings, SEMs. Right, Average firing rate during MAs. 0 s corresponds to the MA onset. Shadings, ±SEMs. (H) NR-max neurons that were significantly inhibited by NE^LC→POA stimulations. (I) Firing rates of 17 excited neurons during different brain states. Left, Pie chart showing the distribution of different subtypes within the excited units. Right, Average activity of excited units during REMs, NREMs, and wake. (J) Top, Spike raster showing multiple laser stimulation trials (10 Hz, 10 ms, 5 s). Bottom, Firing rates before, during, and after laser stimulation. (K) Firing rates of an example unit excited by NE^LC→POA laser stimulation. (L) W-max neurons that were significantly excited by NE^LC→POA stimulations. (M) R-W neurons that were significantly excited by NE^LC→POA stimulations. (D and I) Wilcoxon signed rank test with Bonferroni correction; *P < 0.05; ***P < 0.001 (SI Appendix, Fig. S9 and Table S1).