Oscillatory-Quality of sleep spindles links brain state with sleep regulation and function

Abstract

Here, we characterized the dynamics of sleep spindles, focusing on their damping, which we estimated using a metric called oscillatory-Quality (o-Quality), derived by fitting an autoregressive model to electrophysiological signals, recorded from the cortex in mice. The o-Quality of sleep spindles correlates weakly with their amplitude, shows marked laminar differences and regional topography across cortical regions, reflects the level of synchrony within and between cortical networks, is strongly modulated by sleep-wake history, reflects the degree of sensory disconnection, and correlates with the strength of coupling between spindles and slow waves. As most spindle events are highly localized and not detectable with conventional low-density recording approaches, o-Quality thus emerges as a valuable metric that allows us to infer the spread and dynamics of spindle activity across the brain and directly links their spatiotemporal dynamics with local and global regulation of brain states, sleep regulation, and function.

Fig. 1.. Spindles show a substantial variability in their oscillatory strength.

(A and B) Ten-second signal segments and respective spectrograms for corresponding signals recorded simultaneously from the frontal EEG [(A), top] and occipital EEG [(A), bottom] electrodes, or LFP recorded from layer 4 of S1 [(B), top] and layer 5 of S1 [(B), bottom]. Spectrograms are color-coded on a logarithmic scale. (C) Distribution of the maximum r value across poles with frequencies (𝑓_𝑘) between 10 and 15 Hz for an LFP signal recorded from layer 4 in S1. (D) Peak sigma (10 to 15 Hz) amplitude distribution for the same LFP signal used in (C) (layer 4 in S1). In (C) and (D), line = mean across seven mice. Shaded area = SEM. EEG, electroencephalogram; LFP, local field potential; S1, primary sensory cortex; SEM, standard error of the mean.

Fig. 2.. The o-Quality: a quantitative metric of spindle activity strength.

(A) Spectrogram of an 8-s segment of LFP recording from S1 in a representative mouse. Spectra are color-coded on a logarithmic scale. (B) Eight-second segment of LFP data during NREM sleep showing a sequence of detected spindle events, highlighted by shaded colored boxes. The yellow boxes indicate spindle events whose max 𝑟 values reached 0.94, while the purple box indicates a spindle event whose max 𝑟 value reached 0.95. (C) Absolute 𝑟 values for the four poles estimated by the AR(8) model. Each pole is represented with a different color. The black horizontal lines represent the upper threshold used for detection of oscillatory events (i.e., 𝑟_𝑏 = 0.92) and the lower threshold (𝑟_𝑎 = 0.90) used to merge or separate consecutive oscillatory events. (D) Frequencies 𝑓_𝑘 of the poles with lowest damping. (E) Examples of spindle events with different levels of damping (i.e., different maximum r values). The maximum 𝑟 value for each detected spindle was used to group spindles into four o-Quality groups (oQ1 to oQ4) such that strong-to-weak damping corresponds to low to high o-Quality. LFP, local field potential; S1, primary sensory cortex; AR, autoregressive.

Fig. 3.. The o-Quality reflects spatial dynamics of sleep spindles.

(A) Incidence (per minute of NREM sleep) of spindles detected in EEG (frontal, parietal, and occipital) and LFP (anterior S1, posterior S1, and M1) derivations as a function of spindle o-Quality. Dots = mean across mice; shaded areas = SEM. (B) Number of high–o-Quality (oQ4) spindles as a percent of total spindles detected in EEG (frontal, parietal, and occipital) and LFP (anterior S1, posterior S1, and M1) derivations. For boxplots: black lines = mean across mice, boxes = SEM, whiskers = 95% confidence intervals, and dots = individual values for each mouse. **P < 0.01, ***P < 0.001. (C) Frequency (in hertz) distribution for spindles detected in different EEG derivations (frontal, parietal, and occipital). Lines = mean across mice; shaded areas = SEM. (D) Histological verification of probe location across cortical layers in M1 and S1. Illustrations showing examples of spindle events detected in cortical layers 4 and 6 of S1 (right). (E) Mean spindle incidence per minute (left) and percentage of detected high–o-Quality (oQ4) spindles (percentage of total number of spindles; right) across different cortical layers of S1 and M1 cortices. (F) Example o-Quality 1 and o-Quality 4 spindles with LFP and corresponding current source density (CSD; red, current source; blue, current sink) signal of primary somatosensory cortex. Layer centroids are marked by roman numerals. EEG, electroencephalogram; LFP, local field potential; S1, primary sensory cortex; M1, primary motor cortex; SEM, standard error of the mean.

Fig. 4.. The spindle o-Quality reflects synchrony within local and global cortical networks.

(A) Diagram showing frontal EEG and LFP (S1 and M1) electrodes. (B) Representative S1 LFP and EEG traces with examples of local and global spindle events. (C) Spatial extent of LFP spindles as a function of spindle o-Quality (mean, SEM). (D) Mean maximum-envelope two-cycle average in a representative animal for oQ1 and oQ4 spindles. LFP traces are superimposed on the spatially smoothed CSD, averaged across all detected spindle events. (E) Spiking activity as a function of cycle phase obtained from the Hilbert transform of the LFP in layer 4. Compass plot: mean firing angle and resultant vector length for all spikes within 1 s of spindle midpoint by layer. Rayleigh’s test of circular uniformity confirmed significant phase coupling between MUA and LFP phase in every layer, o-Quality and mouse. (F) Mean frontal EEG power spectra during epochs with detected LFP spindle events in M1 (left) and S1 (right) as a percentage of epochs without spindles. (G) Mean EEG sigma power in the frontal derivation during epochs with detected spindles as a function of o-Quality of LFP spindles detected in S1 (dark blue, laminar probe; light blue, microwire array) and M1 (orange). Note: figures show mean and, where relevant, SEM across mice. EEG, electroencephalogram; S1, primary sensory cortex; M1, primary motor cortex; LFP, local field potential; SEM, standard error of the mean; CSD, current source density; MUA, multiunit activity. S1 laminar: n = 7; S1 micro-array: n = 7; M1 laminar = S1 laminar: n = 7.

Fig. 5.. Spindle o-Quality, slow waves, and sleep homeostasis.

(A) Representative examples of oQ1 and oQ4 spindles in S1. (B) Percentage of oQ1 and oQ4 spindles preceded by SW. (C) Relative LFP SWA during epochs with oQ1 and oQ4 spindles expressed as percentage of mean NREM SWA. (D) Time course of NREM SWA in S1 LFP, expressed as percentage of the 12-hour mean. (E) Incidence of oQ1 and oQ4 spindles as percentage of total across the 12-hour light period. (F) Time course of NREM SWA in the S1 LFP signal during baseline and after 6-hour SD, expressed as percentage of mean baseline SWA. (G) Mean spindle max r value during the first 2 hours after SD and corresponding baseline interval. (H) Spindle incidence ratio between sleep after SD and baseline as a function of spindle o-Quality. (I) EEG 10 to 15 Hz power during epochs with oQ4 LFP spindles during baseline sleep between ZT7-9 and corresponding interval after SD expressed as a percentage of epochs without spindles. (J) Incidence of SW during the first 2 hours after SD and corresponding baseline interval. (K) Percentage of oQ4 spindles preceded by SW during the first 2 hours after SD and the corresponding baseline interval. LFP, local field potential; SWA, slow wave activity (0.5 to 4 Hz); ZT, Zeitgeber time; SW, slow waves; SD, sleep deprivation. For (D) and (E), dots = mean across mice; shaded areas = SEM. For boxplots: black lines = mean across mice, boxes = SEM, whiskers = 95% confidence intervals, and dots = individual values for each mouse. Analyses were performed on one LFP (S1) channel per mouse that showed the highest spindle density. *P < 0.05, **P < 0.01, ***P < 0.001.

Fig. 6.. Glutamatergic neurotransmission is essential for large-scale but not local dynamics of spindles.

(A) Frontal EEG power spectra (mean, SEM) during NREM sleep. (B) Frontal EEG spindle incidence per minute of NREM sleep. (C) Maximum r value for frontal EEG spindles. (D) Average spectrograms centered on the midpoint of individual frontal EEG spindles. (E) Frontal EEG power spectra during epochs with spindle events, shown as percentage of epochs without spindles (mean, SEM). (F) Representative LFP traces (layer 4, S1) with spindles. (G) Mean layer 4 S1 spindle incidence per minute of NREM sleep. (H) Mean maximum r value for LFP spindles detected in layer 4 in S1. (I) Spectrograms centered on layer 4 S1 spindle midpoints. (J) Average NREM LFP power spectra (layer 4, S1). (K) Average LFP power spectra (layer 4, S1) during epochs with spindle events, shown as percentage of epochs without spindles. (L) Mean spatial extent of S1 LFP spindles recorded with 16-channel microwire arrays plotted as a function of o-Quality. (M) Mean frontal EEG sigma power during epochs with spindles detected in the LFP (layer 4, S1) as a function of their o-Quality. Mean values are shown as percentage of NREM sleep epochs without detected spindles. EEG, electroencephalogram; WT, wild type; SEM, standard error of the mean; LFP, local field potential; S1, primary sensory cortex. Dotted lines: significant differences between genotypes. Boxplots: black lines = mean, boxes = SEM, whiskers = 95% confidence intervals, and dots = individual values. GRIA1^−/− (n = 7) and WT (n = 5) for all panels showing averages across animals. ***P < 0.001; n.s., not significant.

Fig. 7.. The o-Quality of spindles is inversely related with the behavioral responsiveness to auditory stimulation during sleep.

(A) Examples of auditory stimulation during (left) and outside (right) spindle events, showing 3-s LFP and EMG segments for one mouse. (B) Spectrogram centered around the midpoint of individual spindle events detected in an LFP signal recorded from layer 4 of S1. LFP spectral power represents mean across mice (n = 7). Spectrograms are color-coded on a logarithmic scale (dB). (C) EMG response to sham stimulation (left) and sound stimulation (right) at time 0, showing stimulations delivered during spindles (n = 1400) and stimulations in nonspindle NREM sleep (n = 1700). (D) EMG variance during the 200-ms period of sham stimulation (left) and auditory stimulation (right) delivered outside or during spindle events. (E) EMG response to sham stimulation (left) and sound stimulation (right) at time 0, showing stimulations during spindles of high (oQ4) (n = 170) and low (oQ1) (n = 780) o-Quality. (F) EMG variance during the 200-ms period of auditory stimulation (left) or sham stimulation (right) delivered during spindle events of with high (oQ4) and low (oQ1) o-Quality. In (C) to (F), the EMG power (μV²) is normalized to the EMG power during NREM epochs with no stimulation. LFP, local field potential; EMG, electromyography. Lines = average across mice, shaded area = SEM. For boxplots: black lines = mean across mice, boxes = SEM, whiskers = 95% confidence intervals, and dots = individual values for each mouse. *P < 0.05, **P < 0.01.