Enhanced thalamocortical functional connectivity during rapid-eye-movement sleep sawtooth waves

Abstract

Sawtooth waves (STWs) are present in electroencephalography (EEG) as bursts of slow 2–5 Hz oscillations contrasting with the low amplitude mixed frequency activity characterizing rapid-eye-movement (REM) sleep. We recently reported, using intracranial recordings, that they involve widespread cortical areas [1]. However, their generator(s) and functional role remain poorly understood. It has been suggested that STWs may be a cortical component of the ponto-geniculo-occipital waves (PGOs), which originate in the reticular pontine formation, propagate to the thalamic lateral geniculate nucleus and to the occipital cortex, and are associated with bursts of REMs [2, 3]. Although PGOs have been mostly described in cats, their intracranial correlates have been recorded in humans in the pontomesencephalic tegmentum and in the subthalamic nucleus in patients with Parkinson’s disease [4, 5], and in the primary visual cortex in patients with epilepsy [6]. Animal studies suggest that PGOs may extend beyond the visual cortex to several neocortical regions including sensory-motor and association areas [7]. They could facilitate synchronization of fast oscillations over widespread cortical areas, thus promoting REM sleep-dependent plasticity [7]. In non-human primates, Balzamo [3] reported PGO potentials resembling STWs in the thalamus and the cortex, providing the first direct evidence of a link between these activities. In humans, the situations in which both subcortical and cortical structures can be simultaneously accessed are very rare. Stereo-electroencephalography recordings (SEEG) performed for presurgical evaluation in drug-resistant focal epilepsy have allowed to demonstrate a peculiar slow activity in the thalamic medial pulvinar nucleus during REM sleep, contrasting with the fast desynchronized cortical activity [8]. However, the link between this slow thalamic activity and STWs remains unexplored. Here, we used combined polysomnography and SEEG recordings to investigate the dynamics of thalamic activity and thalamocortical functional connectivity at the time of scalp EEG STWs. We selected from the Epilepsy Department of Lyon University Hospital database consecutive patients with focal drug-resistant epilepsy who had undergone a SEEG exploration including the thalamus, with at least one-night recording with scalp EEG (scalp electrode positions depended on the extent and location of the SEEG implantation scheme: Fz, Cz in three patients; Cz, Pz in two patients; Fp2, F8, C4, T4, O2, Fp1, and C3 in one patient) and electro-oculography. Patients with abnormal thalamic activity including paroxysmal and non-paroxysmal activity as visually evaluated by epileptologists were excluded. Six patients were identified (Supplementary Table). All patients gave informed consent and the procedure was approved by the National Ethics Committee (CPP 09-CHUG-12, N°0907). The stereotactic electrode implantation procedure was performed using multicontact electrodes inserted into the brain perpendicular to the midsagittal plane. Anatomical localization of electrode contacts was verified using a post-implantation MRI. The placement of the contacts (one electrode, two to three contacts) within the thalamus was assessed using the automated anatomical labeling atlas three in the MNI space (Supplementary Figure 1). Night sleep recordings were conducted at least 5 days after surgery. SEEG and EEG signals were obtained using bipolar montages with the Micromed® software (Treviso, Italy) (sampling frequency: 256 Hz). Sleep scoring was visually performed according to standard guidelines adapted to SEEG, using video recordings and thalamic activity for REM sleep instead of using chin EMG [8]. STWs were manually marked on scalp EEG (Fz-Cz in three patients, Cz-Pz in two patients, and C4-T4 in one patient) independently by two neurophysiologists during all REM sleep episodes as previously described [1]. The overlap (≥1 second) of the two markings was selected for analyses, which were performed by comparing the STW segments with control REM sleep segments of the same duration but separated by at least 2 seconds from STW markings. The signal was filtered using a Butterworth band pass filter between 0.3 Hz and 115 Hz and a 50 Hz notch filter. The spectral content was estimated using the Fast Fourier Transform method. Spectrograms were generated using a Morlet wavelet transform with three cycles. The spectrogram was computed using a 1 Hz bin. Phase-based general connectivity was assessed by the phase locking value (PLV) between the scalp EEG channel used for STW detection and the thalamic SEEG channel. A cross-correlation analysis was used to determine the lag between the signals’ time series for each STW. The log-transformed spectral power and the PLV were weighted by the number of STWs per patient as follows: Cr=log(Pr)*1N*Rs where Cr = contribution of the STW, Pr = Power or PLV of a STW, N = amount of patients, Rs = amount of STWs for a specific patient. Thus, the contribution of each patient to the results was the same. This approach allowed us to deal with the imbalance in number of events per patient without sacrificing too much statistical power given the small dataset. A paired t-test using these weighted values was performed to compare the spectral power in the 2–5 Hz and 30–100 Hz bands (broadband and 10 Hz bin with Bonferroni correction for multiple comparisons) between STW and control segments, and the PLV between the thalamus and the cortex during STW vs control segment. The effect size was measured with Cohen’s d. A 2-sided Binomial test was used to assess the lead and lag between the thalamus and the cortex. A P-value below 0.05 was considered as statistically significant. A total of 176 STW segments were kept for analysis with a mean duration of 3.40 ± 1.4 s, 95% CI [3.23;3.58], and a mean frequency of 2.34 ± 0.74, 95% CI [2.32;2.36] Hz (Supplementary Table). Visual inspection of the SEEG raw signal showed that the thalamic activity was very similar to that exhibited on the cortex during scalp-detected STWs, suggesting that STWs also involve the thalamus (Figure 1A and Supplementary Figure 2). A peak around 2 Hz was clearly observed in the scalp EEG and thalamic activity spectra during STWs, which significantly differed from control segments (2–5 Hz band: p < 0.003, d = 0.31) (Figure 1B). Time-frequency maps confirmed this increase in the delta–theta band power, time-locked with the onset of STW markings. In addition, an increase in high frequencies between 30 and 100 Hz was observed in the thalamus (p < 0.001, d = 0.31), significant in all 10 Hz bands between 30–100 Hz bands except the 30–40 Hz and the 70–80 Hz bands. Thalamocortical activity during sawtooth waves, (A) Example of sawtooth waves recorded in the scalp EEG and thalamus SEEG during REM sleep (patient #2), EEG: electro-encephalogram; SEEG: stereo-electro-encephalogram; EOG: electro-oculogram The red box indicates the STW burst detected in the scalp EEG. The blue box indicates bursts of rapid-eye movements. The example is provided for patient #2. Note that the thalamic background slow activity is very similar to STW activity and that STWs precede bursts of rapid-eye movements. (B) Spectral analysis of scalp EEG and thalamic SEEG signal during sawtooth waves (left panels) Spectra computed during STW and control segments in the scalp EEG and in the thalamus in the six patients. The weighted mean ±SD values of spectral power in U2Hz units are presented. Both scalp and thalamic activity show a peak in the delta band between 2 and 2.5 Hz during STW segments (2–5 Hz band: p < 0.003, d = 0.31; 2Hz: p < 0.001, d = 0.36). The increase in the 30–100 Hz band is significant only in the thalamus (p < 0.001, d = 0.31). Significant results (p < 0.05) with effect size >0.2 for high-frequency sub-bands (10 Hz bins) are indicated with horizontal red lines. (right panels) Time-frequency maps of STW segments in the thalamus and in the scalp (t = 0 corresponds to the onset of the overlap in STW marking between the two readers, average of all events). A clear increase in the 2–5 Hz frequency band power is observed in both structures. (C) Functional connectivity between scalp and thalamus during sawtooth waves. (left panel) Distribution of the phase locking values (PLV, as assessed in Lachaux et al., Hum Brain Mapp. 1999) for STWs and control segments (median + quartiles and range). The PVL is calculated as PLV=E[ej∅(t)] where ∆∅(t) is the difference between the signals' instantaneous phases, and E is the average operator. It is significantly higher during STW than control segments (p < 0.001; d = 0.84). (right panel) Time lag between scalp and thalamic power increase in the 2–5 Hz band. The lag between the signals’ time series for each STW is calculated as Rxy(m) = E{xn+my*n}, where n is the signal length of the STW and control signal and m = 2n−1, the asterisk denotes complex conjugation, and E is the average operator. Then the maximum of R over m was taken and its corresponding lag index was used. The time lag for each individual STW segment (n = 176) is presented, showing that the thalamus increase in the 2–5 Hz band precedes that of the cortex for most STWs. The functional connectivity in the 2–5 Hz band between scalp and thalamus significantly increased during STWs as compared to control segments (p < 0.001; d = 0.84, Figure 1C). A consistent negative lag between scalp and thalamus was found, suggesting that the thalamus was involved in STWs before the cortex with a mean ±SD of 23 ± 22 ms (p < 0.001) (Figure 1C). No significant changes in functional connectivity in the high-frequency range (30–100 Hz broadband and 10 Hz bins from 30 to 100 Hz) were observed (p = 0.47). We found that STWs were associated with (1) an increase in low and fast activities in posterior thalamic nuclei, and (2) enhanced functional thalamocortical connectivity in the 2–5 Hz band with the thalamus driving the cortex. Our results are in line with findings of Balzamo [3] in non-human primates suggesting a link between STW and so-called “ponto-geniculo-cortical” activity. Although the pulvinar is an associative nucleus, it also receives inputs from cholinergic pontine nuclei. Species differences in the distribution of PGOs due to the variation in postsynaptic targets of the pontine generators have been reported [2]. Our observation of phasic thalamocortical potentials during REM sleep suggests that scalp-detected STWs may have a thalamic counterpart, whose link with potentials generated in the brain stem still needs to be demonstrated. The functional role of such potentials remains debated. We found an increased functional thalamocortical connectivity during STWs, which extends recent findings of Lambert et al. [9] and is in line with the report of enhanced synchronization between the anterior thalamus nucleus and the frontal cortex in delta and beta–gamma frequencies during phasic versus tonic REM sleep [10]. Although we observed an increase in thalamic fast activities during STWs, this was not the case in the scalp EEG. It can be hypothesized that the increase in cortical fast oscillations is brief and local which may limit its detectability in the scalp EEG [1]. Our study has several limitations. First, only six patients were included given the rarity of thalamic recordings. However, a large number of STWs were analyzed. Second, we focused on scalp EEG with a low number of available channels; we chose not to use SEEG to mark STWs given the heterogeneity of the explored cortices and the very local nature of STWs [1]. In conclusion, our study suggests that STWs are present in the thalamus and might be involved in phasic brainstem-elicited triggering of thalamocortical communication. This work was supported by the Natural Sciences and Engineering Research Council of Canada RGPIN2020-04127 and RGPAS-2020-00021 as well as by the Fonds de Recherche du Québec–Santé 2021-2025 Salary Award “Chercheur-boursier clinicien Senior” to B.F. T.A. was supported by the “Healthy Brains, Healthy Lives initiative at McGill University” awarded by the Canada First Research Excellence Fund and Fonds de recherche du Québec 2021-2022. Laure Peter-Derex, Tamir Avigdor, and Birgit Frauscher (Conception, design of the study). Laure Peter-Derex, Tamir Avigdor, Sylvain Rheims, Marc Guénot, Nicolas von Ellenrieder, Jean Gotman, and Birgit Frauscher (Acquisition and analysis of the data), Laure Peter-Derex, Tamir Avigdor, Sylvain Rheims, Marc Guénot, Nicolas von Ellenrieder, Jean Gotman, and Birgit Frauscher (Drafting of the manuscript). Financial disclosure: none. Non-financial disclosure: none. The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Topics & Concepts