Ongoing network state controls the length of sleep spindles via inhibitory activity

Abstract

Sleep spindles are major transient oscillations of the mammalian brain. Spindles are generated in the thalamus; however, what determines their duration is presently unclear. Here, we measured somatic activity of excitatory thalamocortical (TC) cells together with axonal activity of reciprocally coupled inhibitory reticular thalamic cells (nRTs) and quantified cycle-by-cycle alterations in their firing in vivo. We found that spindles with different durations were paralleled by distinct nRT activity, and nRT firing sharply dropped before the termination of all spindles. Both initial nRT and TC activity was correlated with spindle length, but nRT correlation was more robust. Analysis of spindles evoked by optogenetic activation of nRT showed that spindle probability, but not spindle length, was determined by the strength of the light stimulus. Our data indicate that during natural sleep a dynamically fluctuating thalamocortical network controls the duration of sleep spindles via the major inhibitory element of the circuits, the nRT.

Figure 1

Sleep Spindles in Ventrobasal Thalamus (A) Simultaneous cortical and thalamic recordings of a sleep spindle in a freely sleeping rat. The raw data show local field potentials in one cortical electrode and multiunit activity (MUA) in eight thalamic channels on one shank of a four shank electrode in the ventrobasal thalamus (VB) Inset: electrode tracks (white arrows) of the thalamic electrode. Note that all four tracks avoid nRT. (B) Smoothed MUA recorded by each of the four electrode shanks (shank distance: 200 μm) in natural sleep (B1) and under urethane anesthesia (B2). Spindles (blue lines) appear as rhythmic elevations of MUA synchronously on all shanks in natural sleep but as a spatially restricted signal under urethane. (C) Histogram of the spindle lengths from all unanesthetized (C1) and urethane anesthetized (C2) animals. Note the relative paucity of short spindles (five to six cycles) under urethane anesthesia. (D) Changes in cycle length during spindles with different durations. Under natural sleep cycle lengths first increase, followed by a decrease before the end in all spindles (D1). Under urethane spindles accelerate throughout the spindle (D2).

Figure 2

Two Types of Spike Waveforms in the Somatosensory Thalamus (A) Raw data from six channels of a single octrode in VB showing wide (black circle) and narrow spikes (magenta circle). (B) Bimodal distribution of action potential widths reveals two populations (black-wide spikes, magenta-narrow spikes). (C) (C1 and C2) Autocorrelograms of a wide spike unit (C1) measured by silicon probe in VB compared to a TC cell (C2) recorded and labeled by juxtacellular recording in VB. (C3 and C4) The same two histogram for and a narrow spike unit (C3) recorded in VB, compared to a nRT cell (C4) measured by juxtacellular recording in nRT. Note wider base (longer burst) and spindle modulation (insets) in case of the narrow spike unit and the nRT cell. The figures are based on recordings made under urethane anesthesia.

Figure 3

nRT Axonal Activity Recorded as Extracellular Signals in VB (A) Camera lucida drawing of a juxtacellularly recorded and labeled nRT cell with simultaneous somatic recording in the nRT axonal recording in the VB. Note the axonal arbor around the octrode. (B) Juxtacellular (bottom red) and silicon octrode (top black) traces of the recorded cell. Each somatic action potential is recorded as spikes in six out of the eight recording sites of the octrode, located 1 mm from the somatic electrode. (C) Spike triggered averages (STA) and cross correlogram (CCG) of the octrode recordings triggered by the somatic nRT action potentials (red dashed line 0 ms). (D) Photomicrograph of the recorded and labeled nRT cell.

Figure 4

Local Interactions between TC and nRT Cells during Spindles (A) A single spindle event displayed as smoothed multiunit activity in VB (red) together with TC cells (black) and nRT terminals (magenta) recorded by the same electrode shank under urethane anesthesia. The firing of both TC cells (black) and nRT axons (magenta) are locked to the local multiunit spindles. Note different firing frequency and burst length of the TC and nRT cells. (B) Polar plots showing the phase vectors of individual TC and nRT cells recorded on the same shank relative to the multiunit spindle collected during several recordings. Top: natural sleep, bottom: urethane anesthesia. One spindle cycle is 360°. TC cells consistently fire at an earlier phase of the oscillation compared to nRT cells in both conditions. (C) Cross correlograms between TC and nRT cells on the same electrode shank and between different shanks (200, 400, 600 μm apart) under urethane anesthesia. Robust correlation is evident only on the same shank (red), but modulation in the neighboring shank (green) also reaches significance. No correlation is apparent, however, on more distant shanks (blue, magenta).

Figure 5

Cycle-by-Cycle Dynamics of Excitatory and Inhibitory Activity during Spindles of Different Lengths during Natural Sleep (A) Perievent time histograms (top) and rasterplots (bottom) of TC (black) and nRT (magenta) units during spindles consisting of 6 (left) and 14 (right) cycles assembled from several representative sessions of three freely sleeping rats. Spindle peaks are aligned for better visibility. (B) Cycle-by-cycle cross-correlograms of TC and nRT units shows unchanged peak latency during spindles of six and 14 cycles. (C) Jitter (SD of spike distances from spindle peak) also remains stable during spindles. Each dot represents the mean data of a given cycle pooled across sessions and animals. (D) Cycle-by-cycle alteration in the mean number of spikes per cycle for nRT (magenta) and TC cells (black) for short (six cycles, left) and long (14 cycles right) long. Note different trajectories of nRT but similar trajectories of TC cells. Shading indicates SEM.

Figure 6

Probability of nRT Firing Displays Duration Specific Pattern during Natural Sleep (A) Cycle-by-cycle changes in the mean number of spikes/bursts for TC (black) and nRT (magenta) cells during spindles of six to 14 cycles. nRT units display a steady decrease in spike per burst during spindles for all spindle lengths, whereas values of TC cells remain stable. (B) Cycle-by-cycle changes in the probability of TC and nRT firing during spindles of different length. nRT cells display duration specific patterns. Each dot represents the mean data of a given cycle pooled across sessions and animals. Shading indicates SEM.

Figure 7

Initial Network State Correlates with Spindle Length during Natural Sleep (A) Participation probability of nRT cells (magenta) in the in the first cycle strongly correlates with length of the spindle, TC cells (black) display weaker but still significant interaction. (B) The initial number of spikes per burst in TC cells also correlates with the forthcoming spindle length. (C) Correlation between the participation probability in the first and last cycle for TC (black dots) and nRT (magenta dots) cells. Between the initial and final state, only the nRT participation probability shows significant correlation. (D) There is no correlation between the spikes/bursts in the initial and last cycle. In (A)–(D), each dot represents the mean value of spindles with given number of cycles pooled across sessions and animals. Only significant interactions are shown with numbers.

Figure 8

Durations of Optogenetically Induced Spindles Do Not Correlate with Stimulus Intensity (A) (A1) Vertically oriented traces of smoothed multiunit activity recorded by one electrode shank under urethane anesthesia in the VB of a mouse expressing channelrhodopsin-2 under parvalbumin promoter. Spindles were evoked by laser activation of nRT cells (red vertical lines) using three different laser intensities (0.12 mW, 1.6 mW, 10.6 mW) every 5 s. Note the state fluctuations between desynchronized (blue arrows), slow-wave sleep (green arrows), and lightly synchronized (red arrows) states with spontaneous spindles. Spindles can only be evoked (red ellipses) in the latter state. (A2) Dominant frequencies of the thalamic MUA activity on the rightmost traces in A1. Warm colors represent spindle frequencies. (B) Duration of spontaneous (black) and evoked spindles (colored according to laser intensity) during a long recording. Evoked and spontaneous spindles co-occur in epochs. One of the epochs (blue dotted line) is shown in expanded time scale (bottom). The length of neighboring spindles show great variability. (C) Probability of evoking a spindle increased with stimulus intensity (upper panel), but no significant difference (Kruskal-Wallis test) was found between the length of evoked spindles (lower panel). (D) Same as (C), but, instead of stimulus intensity, stimulus duration was varied. Data in (A)–(D) are from the same animal. (E) Distribution of all spontaneous and evoked spindle lengths summed from all animals and sessions. Note larger percentage of long (above 1,100 ms) and short (below 600 ms) spindles in spontaneous cases.