Sleeping at the switch

Version of Record

Accepted for publication after peer review and revision.

Download
Cite
Share
CommentOpen annotations (there are currently 0 annotations on this page).

Version of Record published: September 20, 2021 (This version)
Accepted Manuscript published: August 27, 2021 (Go to version)
Accepted: August 26, 2021
Preprint posted: February 3, 2021 (Go to version)
Received: October 26, 2020

1. Of interest
ACC neural ensemble dynamics are structured by strategy prevalence

Mikhail Proskurin, Maxim Manakov, Alla Karpova

Research Article Dec 7, 2023
Further reading

Abstract
Introduction
Materials and methods
Results
Discussion
Data availability
References
Decision letter
Author response
Article and author information
Metrics

Abstract

Sleep slow waves are studied for their role in brain plasticity, homeostatic regulation, and their changes during aging. Here, we address the possibility that two types of slow waves co-exist in humans. Thirty young and 29 older adults underwent a night of polysomnographic recordings. Using the transition frequency, slow waves with a slow transition (slow switchers) and those with a fast transition (fast switchers) were discovered. Slow switchers had a high electroencephalography (EEG) connectivity along their depolarization transition while fast switchers had a lower connectivity dynamics and dissipated faster during the night. Aging was associated with lower temporal dissipation of sleep pressure in slow and fast switchers and lower EEG connectivity at the microscale of the oscillations, suggesting a decreased flexibility in the connectivity network of older individuals. Our findings show that two different types of slow waves with possible distinct underlying functions coexist in the slow wave spectrum.

Introduction

Sleep slow waves are non-rapid eye movement (NREM) sleep oscillations in the delta range (<4 Hz) reflecting a high neuronal synchronization (Diekelmann and Born, 2010). They generate sustained interest in neuroscience research for their role in sleep-dependent memory consolidation, synaptic plasticity, and as markers of homeostatic sleep pressure (Gais et al., 2002; Inostroza and Born, 2013; Steriade, 2006; Borbély, 2001; Diekelmann and Born, 2010). In human sleep studies, there is, however, a fundamental question as to whether EEG waves showing oscillations <4 Hz are a unique entity or rather hide two types of slow waves with specific functional roles. In humans, slow EEG frequencies are often divided into two components, slow waves (typically 1–4 Hz) vs slow oscillations (<1 Hz) (Achermann and Borbély, 1997; Mölle et al., 2002; Muehlroth et al., 2019). It has been hypothesized that these two components have different functional molecular regulation mechanisms (Lee et al., 2004) and responses to homeostatic pressure (Achermann and Borbély, 1997; Campbell et al., 2006). More recent studies in mice and humans showed that faster delta frequencies (2.5–4.5 Hz) react differently to sleep deprivation than lower delta frequencies (0.75–2 Hz), suggesting distinct neurophysiological substrates. Precisely, compared to lower delta frequencies, faster delta frequencies showed an increased incidence and power after a sleep deprivation protocol (Hubbard et al., 2020). In recent years, Siclari et al., 2014 showed that delta oscillations (1–4.5 Hz) with distinct cortical origins and distributions are sustained by different synchronization processes. Their team further identified two types of slow waves, widespread, steep (type I) and smaller, more circumscribed (type II) slow waves with only the second type showing homeostatic regulation (Bernardi et al., 2018). More recently, animal and human studies brought to light new evidence of two types of slow waves based on the positive and negative state duration of slow waves: one showing a positive correlation and another showing a negative correlation between the two (Nghiem et al., 2020). Here, we propose to describe the dichotomy in the delta frequency range based on a new parameter characterizing the time delay from the maximum negative point of the EEG slow wave to the maximum positive point: the transition frequency. Using this parameter, we show two types of slow waves driven by different pressures of homeostatic dissipation and endowed by specific EEG functional connectivity dynamics.

Sleep slow waves are characterized by a hyperpolarizing state (a negative phase in surface EEG/a down-state in animal literature) during which cortical neurons are synchronously silent, followed by a depolarizing state (a positive phase in surface EEG/an up-state in animal literature) during which cortical neurons fire intensively (Csercsa et al., 2010; Steriade, 2006; Chauvette et al., 2011). The transition from the negative to the positive phase is critical, as it is a strong marker of the ability of brain networks to efficiently switch from a state of hyperpolarization to a state of massive depolarization. The slope of the slow wave (the rate of amplitude change from the negative to the positive peak) is associated with the recruitment/decruitment of the neuronal population, with a steeper slope showing a quicker recruitment (Vyazovskiy et al., 2011). It is generally described as the best measure to assess the synaptic strength and sleep homeostasis compared to other classic parameters (Bersagliere and Achermann, 2010; Riedner et al., 2007). However, using the slope as a measure of transition speed also presents important limitations, as it is affected by slow wave amplitude: with similar positive and negative durations, higher slow waves will necessarily have steeper slopes (Bersagliere and Achermann, 2010). A novel metric that captures the transition speed and that is more independent of amplitude needs to be developed.

The study of slow waves necessarily involves the notion of age, as slow waves drastically change during adulthood. Compared to young adults, older individuals show lower slow wave density as well as reduced amplitude, smoother slope, and longer positive and negative phase durations of slow waves, possibly indicating that cortical neurons enter less synchronously into the hyperpolarization and depolarization phases (Carrier et al., 2011). Therefore, our ability to disentangle the influence of the slow wave amplitude over the slope is compromised in aging. A novel metric that captures the transition speed without being affected by amplitude needs to be developed, especially when studying older populations. Age-related EEG connectivity modification has also been recently described in the literature (Ujma et al., 2019; Bouchard et al., 2019) with major age-related differences in deeper NREM sleep, when slow waves are prominent. At the scale of sleep stages and cycles, our team reported that older individuals showed higher between-region EEG connectivity at the whole-brain scale in deep NREM sleep (stage N3) as compared to younger adults (Bouchard et al., 2019). These results support the notion that the brain of younger individuals during deeper NREM sleep stages operates with reduced long-range cortico-cortical connectivity (Spoormaker et al., 2011; Massimini et al., 2007). However, EEG connectivity at the scale of slow waves has yet to be studied in humans and during aging. Such information would allow a better understanding of the dynamic and distinct networks recruited during those oscillations in addition to providing functional clues to support the complementary phenomena happening in the delta frequency range as described in other studies. The goals of our study were thus to clearly identify the dichotomy in the slow wave’s spectrum and describe the EEG connectivity patterns and homeostatic decline of these two types of slow waves in young and older individuals.

Materials and methods

Participants and protocol

Request a detailed protocol

Fifty-nine participants, 30 young (14 women, 16 men; 20–30 years; mean = 23.49 ± 2.79 yo) and 29 older (18 women, 11 men; 50–70 years; mean = 59.6 ± 5.61 yo) adults in good physical and mental health, have completed the study protocol (demographic data for each group is presented in Supplementary file 1). Exclusion criteria were first investigated during a phone screening using a semi-structured interview. Smoking, a body mass index (BMI) over 27, the use of drugs and/or medication that could affect the sleep-wake cycle and/or the nervous system, complaints about the sleep-wake cycle and/or cognition, transmeridian travel within 3 months prior to the study, and night-shift work or night-shift work in the last 3 months all resulted in the exclusion of the participant. Participants included in the study were asked to maintain between 7 and 9 hr of sleep per night prior to the study. Participants with a score higher than 13 at the Beck Depression Inventory (Beck et al., 1988a) or a score higher than 7 at the Beck Anxiety Inventory (Beck et al., 1988b) were excluded from the study. Potential cognitive impairment and dementia were screened using a neuropsychological assessment in which intelligence quotient (IQ), memory, attention, processing speed, and executive functions were performed and ruled out for all participants. Premenopausal women had regular menstrual cycles (25–32 days), and menopausal women showed amenorrhea, for at least a year before the testing. They reported no night sweats or hot flashes. Perimenopausal women were excluded from the research. The protocol was approved by the ethics committee of the Hôpital du Sacré-Coeur de Montréal and was performed in accordance with the relevant guidelines and regulations (CMER-RNQ 08-08-002). Participants provided informed consent and received financial compensation for their participation.

Procedures

Request a detailed protocol

All participants underwent one screening and one experimental night of polysomnographic (PSG) recording at the Center for Advanced Research in Sleep Medicine at the Hôpital du Sacré-Coeur de Montréal. For the screening night only, PSG also included leg electromyogram (EMG), thoracoabdominal plethysmograph, oral/nasal canula as well as frontal, central, and parietal electrodes referred to linked earlobes (Iber et al., 2007). Participants with periodic leg movements or sleep apneas/hypopneas (index >10 per hr of sleep associated with a microarousal) were excluded from the study.

Polysomnographic recording for the experimental night

Request a detailed protocol

All participants filled out a sleep diary and followed a regular sleep-wake cycle for 7 days before the experimental night based on their individual habitual bedtimes and wake times (± 30 min). Bedtimes and wake times in the laboratory were also based on their own sleep schedules. On the experimental PSG night, 20 EEG derivations (Fp1, Fp2, Fz, F3, F4, F7, F8, Cz, C3, C4, Pz, P3, P4, Oz, O1, O2, T3, T4, T5, T6) referred to linked earlobes were recorded (10–20 international system; EEG: gain 10,000; band-pass 0.3–100 Hz; −6 dB), in addition to chin EMG, electrooculogram (EOG), and electrocardiogram (ECG). Signals were recorded using an amplifier system (grass model 15A54; Natus Neurology, Warwick, Rhode Island, USA) and digitized at a sampling rate of 256 Hz using a commercial software (Harmonie, Stellate Systems, Montreal, Quebec, Canada). Sleep stages (N1, N2, N3, and REM) were visually scored by an electrophysiology technician in 30 s epochs and according to the standard criteria of the American Academy of Sleep Medicine (AASM) (Iber et al., 2007), and sleep cycles were identified. Artifacts were first automatically detected (Brunner et al., 1996) and then visually inspected by a trained technician. PSG variables for each group for the experimental night are presented in Supplementary file 1.

Slow wave detection

Request a detailed protocol

Slow waves were detected automatically on artifact-free NREM (N2 and N3) epochs on all electrodes using previously published criteria (Dang-Vu et al., 2008; Dubé et al., 2015). Specifically, data was initially filtered between 0.3 and 4.0 Hz using a band-pass filter (- 3 dB at 0.3 and 4.0 Hz; −23 dB at 0.1 and 4.2 Hz), and slow waves were defined according to the following parameters: a negative peak below −40 µV, a peak-to-peak amplitude above 75 uV, the duration of negative deflection between 1500 and 125 ms, and the duration of positive deflection not exceeding 1000 ms.

Sleep spindle detection

Request a detailed protocol

Spindles were automatically detected on artifact-free NREM (N2 and N3) epochs on all electrodes using a previously published algorithm (Gaudreault et al., 2018; Lafortune et al., 2014; Martin et al., 2013). Specifically, the EEG signal was band-pass filtered between 10 and 16 Hz using a linear-phase finite impulse response filter (–3 dB at 10 and 16 Hz). The envelope amplitude of the Hilbert transform of this band-limited signal was smoothed and a threshold was set at the 75^th percentile. All events of duration between 0.5 and 3 s were then selected as a spindle. The overlap of a spindle oscillation with a slow wave, characterized by the onset of the spindle between - $π$ and $π$ /2 on the slow wave phase, was defined as a co-occurrence (see Figure 1A).

Figure 1

Download asset Open asset

Frequency and transition frequency of slow waves in young and older individuals.

Panel (A) represents the phase of a slow wave with the transition between the maximum negative phase (hyperpolarization (H)) and the maximum positive phase (depolarization (D)) in pink. Panel (B) represents a scatter plot of the exhaustive Fz inventory of the frequency ( $\frac{1}{T}$ ) and transition frequency ( $\frac{1}{2 τ}$ ) of each slow wave detected in N2 and N3 in young (light red) and older groups (dark red). The marginal distributions of the two frequencies show a similar distribution for the mean frequency, whereas the transition frequency shows distinct distributions with aging.

Slow wave characteristics

Request a detailed protocol

For each slow wave, we derived the map between the time and the phase obtained from the Hilbert transform of the filtered slow wave in the delta band ( $0.16 - 4 H z$ ). All slow waves were equally time referenced by choosing the zero phase at the maximum of the depolarization. Then, the temporal evolution during each slow wave was uniquely described with a phase ranging from $- 3 π / 2$ to $π / 2$ as illustrated in Figure 1A. In addition to general parameters like slow wave density (number per min) and frequency (inverse of the total duration $T$ ), we calculated the transition frequency extracted from the filtered slow wave in the delta band. For each slow wave, the transition frequency characterizes the half-wave associated with the depolarization transition. If $τ$ denotes the delay of the transition from the maximum negative point to the maximum positive point of the slow wave (see Figure 1A), then the transition frequency is defined as $f_{τ} = 1 / 2 τ$ . Figure 1B displays the scatter plot of the overall joint distribution of slow wave frequencies and transition frequencies for all slow waves detected on Fz in young and older individuals. The marginal distributions of the two frequencies clearly show an age difference in the distribution of the transition frequency $f_{τ}$ that is not observed for the frequency. We observed a critical value for the transition frequency around 1.2 Hz (dashed line) where the two distributions cross with aging. This change in the distribution suggests a model of mixture to reveal distinct modes that could be associated with different types of sleep slow waves that may evolve distinctively with aging.

Slow and fast switchers

Request a detailed protocol

As introduced in the previous section, we considered a mixture of Gaussians to modelize the distribution of the transition frequency of the slow wave. As seen in Figure 2A and B, distributions show two modes and any slow wave can then be labeled as slow switchers or fast switchers (cyan and dark blue distribution for young and older participants, respectively). More specifically, the probability distribution can be expanded as a sum of weighted Gaussians $p (f_{τ} | S w)$ ,

p (f_{τ}) = p (S l o w S w) p (f_{τ} | S l o w S w) + p (F a s t S w) p (f_{τ} | F a s t S w)

Figure 2 with 2 supplements see all

Download asset Open asset

Transition frequency and the slow and fast switcher definition.

Panels (A) and (B) are normalized histograms of the transition frequencies (in pink) for young (A) and older subjects (B), for slow waves detected in N2 and N3 frontal derivations (average of F3, F4, and Fz). The fit of the distributions can be decomposed as a mixture of Gaussians: one Gaussian (cyan) accounts for slow switchers, and the other (dark blue) accounts for the fast switchers. The separation line at f* = 1,2 Hz stands for the intersection between the two Gaussian distributions. Panel (C) shows the Akaike Information Criterion for the increasing number of Gaussians in the mixture. The lower the criterion with a sparse decomposition, the better the mixture fit (parietal and central derivations are presented in Figure 2—figure supplement 2).

where

p (S l o w S w) + p (F a s t S w) = 1

In this sum, $p (f_{τ} | S l o w S w)$ and $p (f_{τ} | F a s t S w)$ are Gaussian distributions that describe, depending on the class ‘slow switchers’ or ‘fast switchers’, the probability to transit with the frequency $f_{τ}$ . $p (S l o w S w)$ (resp. $p (F a s t S w)$ ) is the probability for the sleep slow wave to be a slow switcher (resp. a fast switcher). The reliability of the mixture model was further tested with the Akaike Information Criterion (Figure 2C) that assessed that a mixture with two Gaussian distributions is necessary and sufficient to fit the entire distribution of the transition frequency. This parametric model of $p (f_{τ})$ can be estimated using the EM (Expectation-Maximization) algorithm to fit the distribution for each individual. From this mixture of Gaussians, we can define the frequency $f^{*}$ where the two Gaussians intersect: a slow wave will be labeled as a ‘slow switcher’ if $f_{τ} < f^{*}$ , i.e. if $p (f_{τ} | S l o w S w) \geq p (f_{τ} | F a s t S w)$ and a fast switcher otherwise.

Slow and fast switcher modulation analysis

Request a detailed protocol

To evaluate the decline of slow and fast switchers throughout the night, we calculated the percentage of slow or fast switchers in each sleep cycle related to the respective total number of slow or fast switchers across the night. To statistically test the changes between the slow and fast switchers’ decline across sleep cycles, a three-way analysis of variance (ANOVA) with one factor [2 (Group: younger vs older)] and two repeated measures [2 (Switcher: slow vs fast)] × [3 (Cycle: cycle 1, 2, and 3)] was performed. p-values <0.05 were considered significant and simple effects were analyzed to follow up significant interactions.

Phase-locked connectivity analyses

Request a detailed protocol

The functional connectivity across the EEG derivations was assessed using a time-resolved phase lag index (PLI) calculated at six successive phases of the slow wave. Five phases were evenly spaced during the transition, whereas a sixth phase was defined after the depolarization maxima. Given a slow wave (further labeled by $k$ ) was detected on the derivation denoted by $n^{*}$ , we considered the internal phase of the detected oscillation and the simultaneous phase of the other EEG derivations, $φ_{n^{*}}^{(k)}$ and $φ_{m}^{(k)}$ , respectively (the * indicates the derivation on which the slow wave was detected). The slow wave PLI between $n^{*}$ and any other derivation $m$ is then defined by

p l i (n^{*}, m) = \frac{1}{N^{*}} \sum_{k k}^{} s i g n (s i n (φ_{n^{*}}^{(k)} - φ_{m}^{(k)}))

where the summation runs over the $N^{*}$ slow waves detected on $n^{*}$ . This quantity is calculated for six regularly spaced phases of the detected slow wave. Since this pairwise pli emphasizes the slow wave detected on $n^{*}$ , we further symmetrize the definition to account for all the slow waves detected over any pair (n, m):

P L I (n, m) = \frac{1}{2} (p l i (n^{*}, m) + p l i (n, m^{*}))

It is worth noting that slow waves that would truly propagate from $n$ to $m$ with a non-vanishing delay would contribute with $p l i (n^{*}, m) ≃ p l i (n, m^{*})$ since the non-vanishing delay will be of an opposite sign. The definition of this PLI thus emphasizes the connectivity due to cortico-cortical propagation of the slow waves, independently from the shape of the oscillation. For each of the six phases chosen along the slow waves, the statistical significance of the connections was assessed through a random resampling of the phase with a max-statistics over the full set of electrode pairs. This null-hypothesis modeling was used to define the threshold at each of the six-phase points, for a given p-value (0.01). Finally, to quantify the global strength of the connectivity, a global connectivity index (Bouchard et al., 2019) was then defined at each of the six slow wave phases as the sum of the PLI over the significant pairs of electrodes. An increasing value of this index qualitatively assesses a more interconnected network or a more significant phase-locked synchronization (with constant non-vanishing delay) among the EEG electrodes.

Results

Slow and fast switchers in the sleep slow wave inventory for young and older individuals

Figure 3 illustrates the slow wave density, distributions of transition frequencies, and the slow switcher probability for frontal (average of F3, F4, and Fz), central (average of C3, C4, and Cz), and parietal (average of P3, P4, and Pz) derivations. As expected, the slow wave density was significantly lower in older individuals as compared to younger participants for the three derivation clusters (Figure 3A). Our analyses demonstrated the existence of the ‘slow switchers’ and the ‘fast switchers’ represented by a bimodal distribution of the transition frequency, in each cluster and for both young and older individuals (Figure 3B and C). The cut-off frequency between the two Gaussian curves was statistically determined for each subject and then averaged over both groups (see ‘Slow and fast switchers’ in the ‘Materials and methods’ section). In young individuals, a cut-off frequency of 1.3 Hz was found for all derivations, whereas older individuals showed a frequency of 1.2 Hz in frontal and 1.1 Hz in central and parietal derivations. Since further analyses showed that a slow and fast switcher dichotomy exists with or without the concomitant occurrence of a spindle (see Figure 3—figure supplement 1), we present the analyses for all slow waves in Figure 3A,B,C and D. We also showed that older individuals had a higher probability of producing slow switchers than fast switchers when generating a slow wave compared to younger individuals (Figure 3D; p<0.0001 for all derivations). Figure 3—figure supplement 1 also shows the probabilities to make a slow switcher or a fast switcher slow wave depending on the presence of a co-occurring sleep spindle.

Figure 3 with 1 supplement see all

Download asset Open asset

Slow and fast switcher slow waves.

(A) Slow wave densities in clusters of frontal (F3, Fz, F4), central (C3, Cz, C4), and parietal (P3, Pz, P4) derivations in young and older individuals. Slow waves were detected in N2 and N3. (B and C) The distribution of probabilities of slow waves being slow (cyan) or fast switchers (dark blue) in younger and older individuals, respectively, with each curve representing one participant. We can observe the two distinct modes of sleep slow waves based on their transition frequency in each derivation. (D) Age-related differences in the probability of producing a slow switcher when generating a slow wave. Significant age differences in (A) and (D) were calculated using t-tests (***p<0.0001).

Fast switchers show a steeper decline than slow switchers at the beginning of night

Considering slow and fast switchers separately, Figure 4 displays the percentage of switchers in each cycle related to the total number of the same switchers across the whole night. Remarkably, the first three sleep cycles, which are present in almost all participants (cycles 4 and 5 mostly missing for the older individuals), showed an exponential decay that is significantly different for the two types of slow waves. In the younger group, the exponential $\sim e^{- r t}$ drives the fast and slow switchers’ decay with $r = 1.6 (R^{2} = 1)$ and $r = 1.3 (R^{2} = 1)$ , respectively. For older subjects, a much slower exponential decay also drives the fast switchers with $r = 0.6 (R^{2} = 1)$ whereas the slow switchers are rather evolving with an exponential reduction $\sim - e^{r t}$ with $r = 0.4 (R^{2} = 1)$ .

Figure 4

Download asset Open asset

Slow and fast switchers decline across sleep cycles.

For each participant in each group (Young individuals in Panel A and older individuals in Panel B), we clustered slow switchers and fast switchers with respect to the cycle during which they occurred. Considering slow and fast switchers separately, the percentage of switchers in each cycle related to the total number of switchers across the whole night is displayed. The curves are spline-interpolation of the overall average in each group.

The three-way ANOVA with repeated measures showed significant Group × Cycle [F(2,1) = 6.9, p = 0.001] and Cycle × Switcher [F(2,1) = 69.2, p<0.001] interactions as well as a specific Cycle effect [F(1.4, 80.1) = 69.7, p<0.001]. Simple effects analysis for the Group × Cycle interaction showed that younger individuals had, in general, more slow waves (averaged number of both types) in Cycle 1 (t(57) = 2.6, p<0.05) but less in Cycle 2 (t(57) = −2.2, p<0.05) when compared to older individuals, whereas no group differences were found for Cycle 3. These results highlight a stronger decrease of slow waves in young individuals as compared to the older individuals between Cycle 1 and Cycle 2, suggesting a steeper decline of homeostatic pressure. As for the Cycle × Switcher interaction, simple effects analysis showed a higher proportion of fast switchers than slow switchers for Cycle 1 (t(58) = −9.6, p<0.001), whereas an opposite effect was found for Cycle 2 (t(58) = 7.0, p<0.001) and Cycle 3 (t(57) = 6.0, p<0.001). When put together, these results suggest a steeper decline of fast switchers between Cycle 1 and Cycle 2 when compared to slow switchers, as demonstrated by the inversion of the slopes at the second time point.

EEG connectivity dynamics evolve differently within slow and fast switchers

PLI was used to investigate intrinsic slow wave EEG connectivity networks across different phases of the depolarization transition (Figure 5A). Since sleep spindles can involve dynamic changes in connectivity (Zerouali et al., 2014), analyses were performed separately for spindle-free slow waves and slow waves coupled to a spindle. Figure 5B and C show the connectivity graphs across consecutive phases in slow and fast switchers for younger individuals (Figure 5B) and older individuals (Figure 5C) for slow waves without sleep spindles. A similar analysis for slow waves with sleep spindles of young and older individuals is shown in Figure 6. Analysis of the global connectivity index, which quantifies the overall significance of the global connectivity of a graph, showed a distinct EEG connectivity strength for slow and fast switchers in both young (Figure 5D) and older (Figure 5E) adults. Our results also showed a higher overall EEG connectivity during slow switchers as compared to fast switchers and a higher connectivity in young individuals rather than in older individuals. More specifically, in younger individuals, we observed a global increase in EEG connectivity during slow switchers, which reached its highest connectivity strength at the maximum of depolarization. Whereas this scenario recruited connectivity patterns along the full slow wave depolarization in the absence of a spindle (Figure 5D), the connectivity involved in the slow wave with a spindle was concentrated later, around the maximum depolarization phase (Figure 6D). In older individuals, slow switchers showed a higher EEG connectivity along the depolarization transition while no significant link was found for the fast switchers. With aging, the fast switchers or the presence of a spindle drastically obliterated the EEG connectivity (Figures 5E and 6E).

Figure 5

Download asset Open asset

EEG connectivity signature of slow and fast switchers without sleep spindles in young and older individuals.

(A) Illustration of the six different phases along the depolarization transition (a, b, c, d, and e) and the hyperpolarization transition of the slow wave (f). (**B and C**) EEG connectivity graphs using phase-locked connectivity (phase lag index, PLI) metrics and statistically assessed by non-parametric statistics for the slow (upper level) and fast (lower level) switchers without sleep spindles, in young (B) and older (C) individuals. (**D and E**) Global connectivity index (GCI) values at each phase of the slow wave, obtained by the summation of the PLI values across the significant electrode pairs. The GCI thus shows the weight of significant links obtained through non-parametric analyses. Slow switchers are represented in cyan, whereas the fast switchers are represented in dark blue. Graphs have a common scale and can therefore be compared.

Figure 6

Download asset Open asset

EEG connectivity signature of slow and fast switchers with sleep spindles in young and older individuals.

(A) Illustration of the six different phases along the depolarization transition (a, b, c, d, and e) and the hyperpolarization transition of the slow wave (f). (**B and C**) EEG connectivity graphs, using phase-locked connectivity (phase lag index, PLI) metrics and statistically assessed by non-parametric statistics for the slow (upper level) and fast (lower level) switchers with sleep spindles, in young (B) and older (C) individuals. (**D and E**) Global connectivity index (GCI) values at each phase of the slow wave, obtained by the summation of the PLI values across the significant electrode pairs. Slow switchers are represented in cyan, whereas the fast switchers are represented in dark blue. Graphs have a common scale and can therefore be compared.

Discussion

In the present work, we have identified two types of slow waves: the slow switchers and the fast switchers. Slow and fast switchers showed distinct distributions of their transition frequency (the transition between the maximum negative point and the maximum positive point of the slow wave) and were detected in both age groups with the older participants showing a higher proportion of slow switchers. We demonstrated that slow and fast switchers are characterized by a specific EEG connectivity signature along the depolarization transition, with slow switchers presenting an overall higher EEG connectivity than fast switchers. Connectivity across slow waves was lower in older individuals as compared to younger ones. Most importantly, when looking at homeostatic regulation, fast switchers showed a steeper decline between the sleep cycles across the night as compared to slow switchers. Using a data-driven approach, the results of this study thus distinguish two types of slow waves present in younger and older individuals, with specific characteristics that could embody complementary functional roles.

Sleep slow waves are either slow or fast switchers

The frequencies related to half-wave components of filtered EEG were introduced decades ago in order to provide an alternative to the time-resolved spectral analysis of sleep (Geering et al., 1993). Such studies were mostly concerned with the half-waves defined by the zero-crossings of the entire high-pass-filtered EEG signals, although no consensus was reached. The present work introduces a new parameter in which half-waves and the associated frequency are defined from the depolarization transition of detected sleep slow waves. This intrinsic parameter, the transition frequency, objectively classifies sleep slow waves in humans into two categories: the slow and fast switchers. At the physiological level, this frequency is associated with the synchronized depolarization of neurons. In the last year, Nghiem et al., 2020 found two types of slow waves by analyzing temporal patterns of slow waves’ down-states and up-states. While this distinction is seen in sleep, it does not seem to be present in anesthesia, which could point to a specific implication of one type of slow wave in sleep-dependent memory consolidation mechanisms specifically during sleep . Our results add crucial information to the recent published articles, showing that the two types of slow waves are underpinned by a different EEG connectivity network dynamics across the depolarization transition itself and confirming the distinct homeostatic decline pattern. The use of a parameter more independent of the amplitude characteristic of the slow wave and associated specifically with the depolarization transition allows us to describe its intrinsic changes in aging. Taken altogether, these findings finally expand our understanding of the dichotomy described in the delta frequency of humans for years and how it is changing in the older population.

Slow and fast switchers show distinct EEG connectivity dynamics

Chauvette et al., 2010 described that in the cat’s cortex, active states begin with the firing of a single neuron, leading to another neuron firing, then leading to a cascade of firing neurons. Here, we observed at a much larger scale that the EEG connectivity also increases as the depolarization occurs and could be linked to the higher communication and increased synaptic activity during the depolarization phase (Chauvette et al., 2010). In particular, the slow switchers in young adults involved a significant increase in EEG connectivity to reach a maximum at the depolarization state of the slow waves. A study by Heib et al., 2013 showed that a longer duration of the depolarization phase of the slow wave was associated with better memory consolidation. Their hypothesis is that a longer depolarization phase could represent an increased possibility to effectuate an initial transfer of recent memory from the hippocampus to the cortex (Heib et al., 2013). Since our identified slow switchers show a slower transition frequency, their specific role in sleep-dependent cognitive processes should be investigated.

The increase in EEG connectivity at the scale of the oscillation seems to be complementary to the global disconnection we described in previous work in the delta band of N3 in the first sleep cycle of younger individuals (Bouchard et al., 2019). With the results of both studies, we can conjecture that this increase of connectivity during the depolarization transition of the slow switchers in young adults requires a global disconnection at a larger scale to make possible such a transient variation. The fast switchers, however, involved a lower connectivity index in young adults. Interestingly, if you look at Figure 1B, it seems that there exists a threshold (around 2.5 Hz) in the transition frequency, above which fast switchers are more difficult to produce. This critical frequency can be converted into a characteristic duration that corresponds to a period around 200 ms. Interestingly, this specific 200-ms duration was set up as the minimal time required for the establishment of connectivity networks, as recently measured by functional magnetic resonance imaging (fMRI) during resting states (Baker et al., 2014). We may hypothesize that if the connectivity of the slow wave is transiently associated with the establishment of a dynamic network, the depolarization of the slow wave can’t be faster than the temporal scale needed for the setting of the network. Additionally, slow waves with a faster transition (the fast switchers with a frequency higher than 2.5 Hz) would not allow enough time for the dynamic network to take place.

It is also worth noting the variability of the connectivity index involved at each phase of the slow switchers’ depolarization transition in younger individuals. This variability may reflect the diversity of networks recruited during such slow waves in those young adults. This idea of the transient reorganization of networks of a ‘flexible brain’ has been described in adults (Baker et al., 2014; Spielberg et al., 2015) and more recently in young children (Yin et al., 2020). Our results show that the flexibility of the slow switchers’ connectivity, with or without spindles, is reduced in aging. Aging also significantly impacts the overall connectivity involved with the switchers. Although minimal connectivity persists for slow switchers in older adults, it was completely abolished for the fast switchers. This reduction in connectivity at the scale of the slow waves’ depolarization transition in older individuals may be related to the general observation that the sleeping brain in aging remains functionally more connected at the scale of the sleep stages, namely N3 (Bouchard et al., 2019). It remains to be investigated if this change in connectivity dynamics could have precise functional consequences in aging, but the lack of EEG connectivity in older individuals suggests a decrease of flexibility in the ability to connect/disconnect and to mobilize the underlying network involved in slow waves. To our knowledge, our study is the first to provide a functional connectivity analysis at different phases along the depolarization transition of the slow wave in humans and it is the first to describe its changes in the context of aging.

The presence of spindles over a slow wave modifies EEG connectivity in both slow and fast switchers. For instance, our results show that the EEG connectivity seems delayed when there is a concomitant spindle. For the younger adults, the EEG connectivity increase observed during the depolarization transition of the slow wave happened to be concomitant with the beginning of the spindle, whereas this connectivity had already risen in the absence of a spindle. This observation is not without recalling recent findings regarding the relationship between neural oscillations and the dynamics of functional connectivity (Tewarie et al., 2019): the spindling oscillation emerging on the top of the slow wave requires a ‘static connectivity’ from the latter (Daffertshofer and van Wijk, 2011). This is especially true for the most represented slow waves produced by adults, that is, the fast switchers of the young adults and the slow switchers of the older individuals. In aging, the presence of the spindle is associated with no changes in EEG connectivity as measured by the global connectivity index. A more exhaustive investigation of the dynamics of EEG connectivity in the interaction between slow waves and spindles connectivity networks in aging could likely contribute to better explaining the changes in sleep-dependent memory consolidation observed in the older population.

Slow and fast switchers show distinct homeostatic responses

Our study shows that fast switchers undergo a steeper decline in the subsequent cycles, compared to slow switchers. When looking at the usual frequency of slow waves, Hubbard et al., 2020 showed that prolonged waking periods are followed by a higher prevalence of faster waves at the beginning of the sleep period. They also showed that fast delta frequencies in mice and humans showed a steeper decline than slow delta frequencies after sleep deprivation. Other studies describing slow oscillations and delta waves using the usual frequency argued that low frequencies (<1 Hz) are less modulated by homeostatic pressure (Achermann and Borbély, 1997; Campbell et al., 2006). A recent study by Kim et al., 2019 using a closed-loop optogenetic technique in rats was able to associate slow oscillations (<1 Hz) with consolidation of memory while slow waves (delta waves; <4 Hz) were involved in the forgetting process, showing dissociable and competing roles of the two rhythms in sleep-dependent memory consolidation. Kim et al., 2019 also argued that the brain could accelerate the up-state transition of slow waves to better dissipate homeostatic pressure. Although it is unknown whether slow or fast switchers respond differently to a homeostatic challenge, we can hypothesize that fast switchers would be more involved in the response to a sleep challenge such as sleep deprivation.

Slow and fast switchers evolve differently with aging

Older individuals in our study produced 60% of slow switchers compared to 40% for younger adults, which means that the prevalence of this type of oscillation significantly increases with advancing age. Compared to older participants, younger participants seem to have more efficient initiation and termination of slow waves’ transition as they generate slow waves with a steeper slope (Carrier et al., 2011; Ujma et al., 2019). This rationale could partially explain the higher prevalence of slow switchers observed in our aging population, namely, that the latter might need an overall longer delay in polarity reversal. Also, age-related changes in homeostatic response could be responsible for the changes in slow wave production (Tononi, 2009). Indeed, older subjects show a significant decrease in their ability to increase the characteristics of slow waves (density, amplitude, slope, and duration) after a sleep deprivation, and these effects are more prominent in prefrontal and frontal derivations (Lafortune et al., 2012). One could hypothesize that these specific effects of aging on homeostasis response in frontal areas would reflect the underlying changes specifically in fast switchers as compared to slow switchers.

We showed that there is an age-related reduction in homeostatic response for both slow and fast switchers. However, the relative ratio of slow and fast switchers across all sleep cycles was maintained with age. While fast switchers were predominant in the first cycle, slow switchers predominated in all the other cycles for both young and old subjects. One could thus hypothesize that these oscillations have a different functional role to play across the night, both for young and older subjects. Interestingly, the early night generally benefits verbal memory consolidation while subsequent sleep cycles could be more beneficial for procedural memory (Plihal and Born, 1997; Gais and Born, 2004). Future studies need to investigate the relative contribution of both fast and slow switchers to memory consolidation processes during sleep and overall sleep-dependent cognitive processes.

The animal literature on NREM slow wave parameters shows differences with humans. For instance, compared to humans, there is an age-related increase in frontal local field potentials (LFP) delta power (Soltani et al., 2019) and in slow waves’ amplitude and slope in older mice, suggesting higher neuronal synchronization (Panagiotou et al., 2017; McKillop et al., 2018). Sleep deprivation protocols also showed higher sleep pressure (Panagiotou et al., 2017) and a similar sleep pressure discharge between young and older mice (Wimmer et al., 2013). While sleep researchers are trying to understand and explain the differences (McKillop and Vyazovskiy, 2020), the new parameter, that is, the transition frequency, brings a new angle of analysis and could lead to interesting insights into this problem, for example, by looking at the proportion of slow and fast switchers, their proportion in sleep deprivation protocols, and their pattern of homeostatic decline and brain functional connectivity. While more sleep deprivation studies are needed to understand the functional role of slow and fast switchers and their value for the aging brain, looking into slow and fast switchers in animals would enhance our understanding of the sleeping brain.

Conclusion

This study is the first to use the transition frequency of slow waves to introduce and to study the slow and fast switchers in the slow wave spectrum, which were identified in both young and older adults. Slow and fast switchers present different connectivity dynamics along their depolarization transition, with slow switchers having a higher connectivity than fast switchers. They are also differently modulated during the night, with fast switchers showing steeper decreases at the beginning of the night. Aging was associated with a higher number of slow switchers than fast switchers, an overall lower EEG connectivity across the depolarization transition of slow waves, and a flatter homeostatic decline of both slow wave types across the night. Those results regarding slow waves likely imply different functional mechanisms associated with slow and fast switchers that could be modified in aging.

Data availability

All codes and transformed data used for all the analyses and most specifically to produce all of the figures of the paper can be freely accessible using this link : https://github.com/jmlina/Slow_Wave_Switchers (copy archived at https://archive.softwareheritage.org/swh:1:rev:3af1be579bd6d5aac1718d49c85e8af9c17541c1). Dataset cannot be shared as participants did not give consent for data sharing. For the raw data, a request needs to be formulated to the ethic committee of the Hôpital de Sacré-Coeur de Montréal, as rawdata of human participants cannot be made public under Québec’s law. The data provided will be anonymized. Researchers who request access to the data will need to provide their research protocol and their IRB approval for this protocol. The documents will be studied by the owner of the database (Julie Carrier) who will then also submit to her institution’s REB for authorization to share the data. Data requests should be addressed to: Julie Carrier (PI): julie.carrier.1@umontreal.ca Sonia Frenette (in cc) : sonia.frenette@umontreal.ca.

References

1. Achermann P
2. Borbély AA
(1997) Low-frequency (< 1 Hz) oscillations in the human sleep electroencephalogram
Neuroscience 81:213–222.
https://doi.org/10.1016/s0306-4522(97)00186-3
- PubMed
- Google Scholar
(2014) Fast transient networks in spontaneous human brain activity
eLife 3:e01867.
https://doi.org/10.7554/eLife.01867
- PubMed
- Google Scholar
1. Beck AT
2. Epstein N
3. Brown G
4. Steer RA
(1988a) An inventory for measuring clinical anxiety: psychometric properties
Journal of Consulting and Clinical Psychology 56:893–897.
https://doi.org/10.1037/0022-006X.56.6.893
- PubMed
- Google Scholar
(1988b) Psychometric properties of the beck depression inventory: twenty-five years of evaluation
Clinical Psychology Review 8:77–100.
https://doi.org/10.1016/0272-7358(88)90050-5
- Google Scholar
(2018) Local and widespread slow waves in stable NREM sleep: evidence for distinct regulation mechanisms
Frontiers in Human Neuroscience 12:1–13.
https://doi.org/10.3389/fnhum.2018.00248
- PubMed
- Google Scholar
1. Bersagliere A
2. Achermann P
(2010) Slow oscillations in human non-rapid eye movement sleep electroencephalogram: effects of increased sleep pressure
Journal of Sleep Research 19:228–237.
https://doi.org/10.1111/j.1365-2869.2009.00775.x
- PubMed
- Google Scholar
1. Borbély AA
(2001)
From slow waves to sleep homeostasis: new perspectives

Archives Italiennes De Biologie 139:53–61.
- PubMed
- Google Scholar
(2019) EEG connectivity across sleep cycles and age
Sleep 342:zsz236.
https://doi.org/10.1093/sleep/zsz236
- Google Scholar
1. Brunner DP
2. Vasko RC
3. Detka CS
4. Monahan JP
5. Reynolds CF
6. Kupfer DJ
(1996) Muscle artifacts in the sleep EEG: automated detection and effect on all-night EEG power spectra
Journal of Sleep Research 5:155–164.
https://doi.org/10.1046/j.1365-2869.1996.00009.x
- PubMed
- Google Scholar
(2006) Homeostatic behavior of fast Fourier transform power in very low frequency non-rapid eye movement human electroencephalogram
Neuroscience 140:1395–1399.
https://doi.org/10.1016/j.neuroscience.2006.03.005
- PubMed
- Google Scholar
1. Carrier J
2. Viens I
3. Poirier G
4. Robillard R
5. Lafortune M
6. Vandewalle G
7. Martin N
8. Barakat M
9. Paquet J
10. Filipini D
(2011) Sleep slow wave changes during the middle years of life
European Journal of Neuroscience 33:758–766.
https://doi.org/10.1111/j.1460-9568.2010.07543.x
- PubMed
- Google Scholar
(2010) Origin of active states in local neocortical networks during slow sleep oscillation
Cerebral Cortex 20:2660–2674.
https://doi.org/10.1093/cercor/bhq009
- PubMed
- Google Scholar
(2011) Properties of slow oscillation during slow-wave sleep and anesthesia in cats
Journal of Neuroscience 31:14998–15008.
https://doi.org/10.1523/JNEUROSCI.2339-11.2011
- PubMed
- Google Scholar
1. Csercsa R
2. Dombovári B
3. Fabó D
4. Wittner L
5. Eross L
6. Entz L
7. Sólyom A
8. Rásonyi G
9. Szucs A
10. Kelemen A
11. Jakus R
12. Juhos V
13. Grand L
14. Magony A
15. Halász P
16. Freund TF
17. Maglóczky Z
18. Cash SS
19. Papp L
20. Karmos G
21. Halgren E
22. Ulbert I
(2010) Laminar analysis of slow wave activity in humans
Brain 133:2814–2829.
https://doi.org/10.1093/brain/awq169
- PubMed
- Google Scholar
1. Daffertshofer A
2. van Wijk BC
(2011) On the influence of amplitude on the connectivity between phases
Frontiers in Neuroinformatics 5:1–12.
https://doi.org/10.3389/fninf.2011.00006
- PubMed
- Google Scholar
1. Dang-Vu TT
2. Schabus M
3. Desseilles M
4. Albouy G
5. Boly M
6. Darsaud A
7. Gais S
8. Rauchs G
9. Sterpenich V
10. Vandewalle G
11. Carrier J
12. Moonen G
13. Balteau E
14. Degueldre C
15. Luxen A
16. Phillips C
17. Maquet P
(2008) Spontaneous neural activity during human slow wave sleep
PNAS 105:15160–15165.
https://doi.org/10.1073/pnas.0801819105
- PubMed
- Google Scholar
1. Diekelmann S
2. Born J
(2010) The memory function of sleep
Nature reviews. Neuroscience 11:114–126.
https://doi.org/10.1038/nrn2762
- PubMed
- Google Scholar
1. Dubé J
2. Lafortune M
3. Bedetti C
4. Bouchard M
5. Gagnon JF
6. Doyon J
7. Evans AC
8. Lina JM
9. Carrier J
(2015) Cortical thinning explains changes in sleep slow waves during adulthood
Journal of Neuroscience 35:7795–7807.
https://doi.org/10.1523/JNEUROSCI.3956-14.2015
- PubMed
- Google Scholar
1. Gais S
2. Mölle M
3. Helms K
4. Born J
(2002) Learning-dependent increases in sleep spindle density
The Journal of Neuroscience 22:6830–6834.
https://doi.org/10.1523/JNEUROSCI.22-15-06830.2002
- PubMed
- Google Scholar
1. Gais S
2. Born J
(2004) Declarative memory consolidation: mechanisms acting during human sleep
Learning & Memory 11:679–685.
https://doi.org/10.1101/lm.80504
- PubMed
- Google Scholar
(2018) The association between white matter and sleep spindles differs in young and older individuals
Sleep 41:1–13.
https://doi.org/10.1093/sleep/zsy113
- Google Scholar
(1993) Period-amplitude analysis and power spectral analysis: a comparison based on all-night sleep EEG recordings
Journal of Sleep Research 2:121–129.
https://doi.org/10.1111/j.1365-2869.1993.tb00074.x
- PubMed
- Google Scholar
1. Heib DPJ
2. Hoedlmoser K
3. Anderer P
4. Zeitlhofer J
5. Gruber G
6. Klimesch W
7. Schabus M
(2013) Slow oscillation amplitudes and Up-State lengths relate to memory improvement
PLOS ONE 8:e82049-9.
https://doi.org/10.1371/journal.pone.0082049
- Google Scholar
(2020) Rapid fast-delta decay following prolonged wakefulness marks a phase of wake-inertia in NREM sleep
Nature Communications 11:1–16.
https://doi.org/10.1038/s41467-020-16915-0
- Google Scholar
Book
(2007)
The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology, and Technical Specifications

Westchester, IL: American Academy of Sleep Medicine.
- Google Scholar
1. Inostroza M
2. Born J
(2013) Sleep for preserving and transforming episodic memory
Annual Review of Neuroscience 36:79–102.
https://doi.org/10.1146/annurev-neuro-062012-170429
- PubMed
- Google Scholar
1. Kim J
2. Gulati T
3. Ganguly K
(2019) Competing roles of slow oscillations and Delta waves in memory consolidation versus forgetting
Cell 179:514–526.
https://doi.org/10.1016/j.cell.2019.08.040
- PubMed
- Google Scholar
1. Lafortune M
2. Gagnon JF
3. Latreille V
4. Vandewalle G
5. Martin N
6. Filipini D
7. Doyon J
8. Carrier J
(2012) Reduced slow-wave rebound during daytime recovery sleep in middle-aged subjects
PLOS ONE 7:e43224.
https://doi.org/10.1371/journal.pone.0043224
- PubMed
- Google Scholar
1. Lafortune M
2. Gagnon JF
3. Martin N
4. Latreille V
5. Dubé J
6. Bouchard M
7. Bastien C
8. Carrier J
(2014) Sleep spindles and rapid eye movement sleep as predictors of next morning cognitive performance in healthy middle-aged and older participants
Journal of Sleep Research 23:159–167.
https://doi.org/10.1111/jsr.12108
- PubMed
- Google Scholar
1. Lee J
2. Kim D
3. Shin HS
(2004) Lack of delta waves and sleep disturbances during non-rapid eye movement sleep in mice lacking alpha1G-subunit of T-type calcium channels
PNAS 101:18195–18199.
https://doi.org/10.1073/pnas.0408089101
- PubMed
- Google Scholar
Software
1. Lina JM
(2021a) Slow_Wave_Switchers, version 3 af1be5
Github.

https://github.com/jmlina/Slow_Wave_Switchers
Software
1. Lina J-M
(2021b)
SlowWaveSwitchers, version swh:1:rev:3af1be579bd6d5aac1718d49c85e8af9c17541c1 https://archive.softwareheritage.org/swh:1:rev:3af1be579bd6d5aac1718d49c85e8af9c17541c1

Software Heritage.
1. Martin N
2. Lafortune M
3. Godbout J
4. Barakat M
5. Robillard R
6. Poirier G
7. Bastien C
8. Carrier J
(2013) Topography of age-related changes in sleep spindles
Neurobiology of Aging 34:468–476.
https://doi.org/10.1016/j.neurobiolaging.2012.05.020
- PubMed
- Google Scholar
1. Massimini M
2. Ferrarelli F
3. Esser SK
4. Riedner BA
5. Huber R
6. Murphy M
7. Peterson MJ
8. Tononi G
(2007) Triggering sleep slow waves by transcranial magnetic stimulation
PNAS 104:8496–8501.
https://doi.org/10.1073/pnas.0702495104
- PubMed
- Google Scholar
1. McKillop LE
2. Fisher SP
3. Cui N
4. Peirson SN
5. Foster RG
6. Wafford KA
7. Vyazovskiy VV
(2018) Effects of aging on cortical neural dynamics and local sleep homeostasis in mice
The Journal of Neuroscience 38:3911–3928.
https://doi.org/10.1523/JNEUROSCI.2513-17.2018
- PubMed
- Google Scholar
1. McKillop LE
2. Vyazovskiy VV
(2020) Sleep and ageing: from human studies to rodent models
Current Opinion in Physiology 15:210–216.
https://doi.org/10.1016/j.cophys.2020.03.004
- PubMed
- Google Scholar
1. Mölle M
2. Marshall L
3. Gais S
4. Born J
(2002) Grouping of spindle activity during slow oscillations in human non-rapid eye movement sleep
The Journal of Neuroscience 22:10941–10947.
https://doi.org/10.1523/JNEUROSCI.22-24-10941.2002
- PubMed
- Google Scholar
(2019) Precise slow Oscillation-Spindle coupling promotes memory consolidation in younger and older adults
Scientific Reports 9:1–15.
https://doi.org/10.1038/s41598-018-36557-z
- PubMed
- Google Scholar
1. Nghiem TE
2. Tort-Colet N
3. Górski T
4. Ferrari U
5. Moghimyfiroozabad S
6. Goldman JS
7. Teleńczuk B
8. Capone C
9. Bal T
10. di Volo M
11. Destexhe A
(2020) Cholinergic switch between two types of slow waves in cerebral cortex
Cerebral Cortex 30:3451–3466.
https://doi.org/10.1093/cercor/bhz320
- PubMed
- Google Scholar
(2017) Differences in electroencephalographic non-rapid-eye movement sleep slow-wave characteristics between young and old mice
Scientific Reports 7:43656.
https://doi.org/10.1038/srep43656
- Google Scholar
1. Plihal W
2. Born J
(1997) Effects of early and late nocturnal sleep on declarative and procedural memory
Journal of Cognitive Neuroscience 9:534–547.
https://doi.org/10.1162/jocn.1997.9.4.534
- PubMed
- Google Scholar
1. Riedner BA
2. Vyazovskiy VV
3. Huber R
4. Massimini M
5. Esser S
6. Murphy M
7. Tononi G
(2007) Sleep homeostasis and cortical synchronization: III. A high-density EEG study of sleep slow waves in humans
Sleep 30:1643–1657.
https://doi.org/10.1093/sleep/30.12.1643
- PubMed
- Google Scholar
1. Siclari F
2. Bernardi G
3. Riedner BA
4. LaRocque JJ
5. Benca RM
6. Tononi G
(2014) Two distinct synchronization processes in the transition to sleep: a high-density electroencephalographic study
Sleep 37:1621–1637.
https://doi.org/10.5665/sleep.4070
- PubMed
- Google Scholar
1. Soltani S
2. Chauvette S
3. Bukhtiyarova O
4. Lina J-M
5. Dubé J
6. Seigneur J
7. Carrier J
8. Timofeev I
(2019) Sleep–Wake Cycle in Young and Older Mice
Frontiers in Systems Neuroscience 13:51.
https://doi.org/10.3389/fnsys.2019.00051
- Google Scholar
(2015) Flexible brain network reconfiguration supporting inhibitory control
PNAS 112:10020–10025.
https://doi.org/10.1073/pnas.1500048112
- PubMed
- Google Scholar
(2011) Large-scale functional brain networks in human non-rapid eye movement sleep: insights from combined electroencephalographic/functional magnetic resonance imaging studies
Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 369:3708–3729.
https://doi.org/10.1098/rsta.2011.0078
- PubMed
- Google Scholar
1. Steriade M
(2006) Grouping of brain rhythms in corticothalamic systems
Neuroscience 137:1087–1106.
https://doi.org/10.1016/j.neuroscience.2005.10.029
- PubMed
- Google Scholar
1. Tewarie P
2. Hunt BAE
3. O'Neill GC
4. Byrne A
5. Aquino K
6. Bauer M
7. Mullinger KJ
8. Coombes S
9. Brookes MJ
(2019) Relationships between neuronal oscillatory amplitude and dynamic functional connectivity
Cerebral Cortex 29:2668–2681.
https://doi.org/10.1093/cercor/bhy136
- PubMed
- Google Scholar
1. Tononi G
(2009) Slow wave homeostasis and synaptic plasticity
Journal of Clinical Sleep Medicine 5:S16–S19.
https://doi.org/10.5664/jcsm.5.2S.S16
- PubMed
- Google Scholar
1. Ujma PP
2. Konrad BN
3. Simor P
4. Gombos F
5. Körmendi J
6. Steiger A
7. Dresler M
8. Bódizs R
(2019) Sleep EEG functional connectivity varies with age and sex, but not general intelligence
Neurobiology of Aging 78:87–97.
https://doi.org/10.1016/j.neurobiolaging.2019.02.007
- PubMed
- Google Scholar
(2011) Electrophysiological correlates of sleep homeostasis in freely behaving rats
Progress in Brain Research 193:17–38.
https://doi.org/10.1016/B978-0-444-53839-0.00002-8
- PubMed
- Google Scholar
1. Wimmer ME
2. Rising J
3. Galante RJ
4. Wyner A
5. Pack AI
6. Abel T
(2013) Aging in mice reduces the ability to sustain sleep/wake states
PLOS ONE 8:e81880.
https://doi.org/10.1371/journal.pone.0081880
- PubMed
- Google Scholar
1. Yin W
2. Li T
3. Hung SC
4. Zhang H
5. Wang L
6. Shen D
7. Zhu H
8. Mucha PJ
9. Cohen JR
10. Lin W
(2020) The emergence of a functionally flexible brain during early infancy
PNAS 117:23904–23913.
https://doi.org/10.1073/pnas.2002645117
- PubMed
- Google Scholar
1. Zerouali Y
2. Lina JM
3. Sekerovic Z
4. Godbout J
5. Dube J
6. Jolicoeur P
7. Carrier J
(2014) A time-frequency analysis of the dynamics of cortical networks of sleep spindles from MEG-EEG recordings
Frontiers in Neuroscience 8:14–16.
https://doi.org/10.3389/fnins.2014.00310
- PubMed
- Google Scholar

Decision letter

Chris I Baker

Senior Editor; National Institute of Mental Health, National Institutes of Health, United States
Bryce A Mander

Reviewing Editor; University of California, Irvine, United States
Ksenia V Kastanenka

Reviewer; MassGeneral Institute for Neurodegenerative Disease, Massachusetts General Hospital, Harvard Medical School, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The authors characterize a specific novel electrophysiological property of slow waves during non-rapid eye movement (NREM ) sleep in humans related to the speed of transition (termed "transition frequency") between the two defining hyperpolarized and depolarized brain states associated with NREM slow waves. This study provides important new data to better understand the network dynamics giving rise to slow waves, the biological role of slow waves, how their expression changes with age, and offers critical implications for the biology of sleep-dependent cognition and cognitive impairment in later life.

Decision letter after peer review:

Thank you for submitting your article "Sleeping at the Switch" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Chris Baker as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Ksenia V. Kastanenka (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) The manuscript findings need to be more fully embedded into the existing literature, particularly with respect to non-human animal findings characterizing the physiological properties of slow waves and their distinctions along different dimensions, and further clarity regarding the novelty of this metric in relation to this literature is warranted.

2) The analysis methods implemented require greater detail and clarity.

3) Clear statistical demonstration of the independence of the transition frequency from slow wave amplitude and slope metrics is absent and was noted by multiple reviewers to be a critical concern.

Reviewer #1:

In this work, the authors propose a functional subdivision of EEG slow waves in humans based on "transition frequency" between putative UP and DOWN states. The data suggest an existence of two distinct subtypes of slow waves, called slow switchers and fast switchers. It appears that these two types differ with respect to the response to homeostatic sleep pressure, with respect to ageing and in relation to "connectivity". Overall, this is an interesting study, that provides potentially important new data that can help to understand the biological role of sleep slow waves.

I have specific comments as follows:

1. I would not assume that the polarity of slow waves recorded using scalp EEG can be easily related to neuronal UP and DOWN states or states of membrane depolarisation and hyperpolarisation. I urge the authors to discuss this more critically and amend terminology where relevant because this is misleading.

2. There is substantial animal literature, which investigated the relationship between EEG, LFP and neuronal population ON and OFF periods or UP/DOWN states. This earlier work in mice and rats addressed the effects of sleep-wake history and ageing on cortical slow waves and their neuronal correlates, including slow wave characteristics, such as slopes and the underlying neuronal synchrony; however, these studies are not mentioned.

3. What is the difference between the new metric proposed here and previously used slow wave slope, which was linked to synchronisation at the ON-OFF and OFF-ON transitions? The authors argue that their approach is based on "a parameter free of the amplitude characteristic of the slow wave", but this has not been demonstrated. It is likely that transition frequency is related to slow wave amplitude, even if the latter is not explicitly included in the calculation. Please provide further analyses to evaluate how transition frequency relates to slow wave amplitude, duration, and slopes to enable comparisons with earlier studies.

4. Figure 1 illustrates how the cut-off of 1.2 Hz was obtained from the data, but essential details are lacking. Do I understand correctly that the scatter plot shows all individual slow waves pooled across all subjects? Could you show individual subjects separately? Do you still see bimodality at the individual subject level? Does the difference between younger and older groups still manifest when you plot distributions of transition frequencies for individual younger and older subjects, and can this be tested statistically? I wondered if you were to replicate the analyses used by Hubbard et al., 2020 on your data, can you see bimodality they report in mice?

5. It is stated in Methods: "All slow waves were equally time referenced by choosing the zero phase at the maximum of the depolarization." It is not clear what is "maximum" here? I presume it is not uncommon that there are multiple positive peaks on the "depolarisation" part of the slow wave, which complicates this analysis. Was the number of peaks within a slow wave different between older and younger subjects? I wondered if the two categories of slow waves correspond to those with one peak and two peaks riding on the top of the "depolarisation/up" phase?

6. I wondered if the "critical frequency" or the dynamics of "fast" and "slow" switchers would change (or an additional category emerge) if you filter the signals at a higher frequency, such as at 6 Hz, rather than 4 Hz (e.g. see https://pubmed.ncbi.nlm.nih.gov/10607082/)

7. The analysis of EEG connectivity is interesting, but I wondered if there is a relationship between occurrence of multiple peaks on the depolarisation phase and connectivity? Riedner et al. study suggested that multipeak slow waves can reflect "collision" of slow waves originating from distant cortical areas: https://pubmed.ncbi.nlm.nih.gov/18246974/

Generally it is a work conducted up to a high standard, but the main issue that remains to be addressed is how robust the approach used here to changes in specific parameters, such as how the signals were filtered. I would also recommend providing some further analyses to assess how transition frequency relates to slow wave amplitude and slope.Reviewer #2:

Here, the authors present data regarding the occurrence of (relatively speaking) slow and fast onset slow waves in younger and older adults, and some phenomenology surrounding their occurrence. There is a reasonable argument to be made that this may represent an actual dichotomy – e.g. bimodal distribution of onset time (essentially the slope). The authors claim that this is a feature independent of wave amplitude, but no data is shown to this effect. This makes the interpretation of the significance of these data problematic, as many of the features about "fast switcher" waves are already well known regarding high-amplitude waves. Thus it is hard to know what is really new here – as much is already known about changes in amplitude with aging.

The authors have developed some new metrics to characterize features of slow wave waveforms during NREM sleep. They propose that there are two classes of slow waves and show some suggestive data that is indeed the case. For example, they show a clear bimodal distribution of transition frequencies (seemingly a measure of transition slope). It is not clear how this relates to wave amplitude- data addressing this are not shown – but it stands to reason that this feature maps onto wave amplitude very well. Because much is already known about amplitude of waves' relationship to synchrony across recording sites, and aging, and occurrence across the night, it is very unclear if this is just a retelling of the same story using a different metric.

An analysis of how the new metric relates to waves' amplitudes is absolutely necessary for the readers to know if there is anything clearly different between this study and what is already known.

A fair bit of the methodology around analyses data is not present, making it questionable what different figures are really showing – for example, it is unclear which stages of NREM sleep are used for various measures, and if these waves' occurrence differs between those stages.

Reviewer #3:

In this manuscript the authors aim to distinguish between two types of slow waves based on a new parameter characterizing the time delay from the up-to-down state, the transition frequency. Sleep and slow waves were monitored in young and older adults in good physical and mental health. The authors report distinction between two types of slow waves, slow switchers and fast switchers, which were detected in both age groups with the older individuals showing a higher proportion of slow switchers.

Strengths:

The manuscript is well-written and question under investigation is well-articulated.

The work is novel and significant to the field.

Conclusions are supported by the data.

Weaknesses: None

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Sleeping at the Switch" for further consideration by eLife. Your revised article has been evaluated by Chris Baker (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there was one remaining issue that needs to be addressed, as outlined below:

While the reviewers agreed that the manuscript was substantially improved through revisions, the consensus was that further consideration in the discussion of the study findings was needed in the context of the publication by Geering et al. (https://pubmed.ncbi.nlm.nih.gov/10607082/), which reports the differential temporal dynamics of slow waves at different frequencies and the impact of analysis methods on their metrics.

https://doi.org/10.7554/eLife.64337.sa1

Author response

Essential revisions:

1) The manuscript findings need to be more fully embedded into the existing literature, particularly with respect to non-human animal findings characterizing the physiological properties of slow waves and their distinctions along different dimensions, and further clarity regarding the novelty of this metric in relation to this literature is warranted.

2) The analysis methods implemented require greater detail and clarity.

3) Clear statistical demonstration of the independence of the transition frequency from slow wave amplitude and slope metrics is absent and was noted by multiple reviewers to be a critical concern.

Reviewer #1:

In this work, the authors propose a functional subdivision of EEG slow waves in humans based on "transition frequency" between putative UP and DOWN states. The data suggest an existence of two distinct subtypes of slow waves, called slow switchers and fast switchers. It appears that these two types differ with respect to the response to homeostatic sleep pressure, with respect to ageing and in relation to "connectivity". Overall, this is an interesting study, that provides potentially important new data that can help to understand the biological role of sleep slow waves.

I have specific comments as follows:

1. I would not assume that the polarity of slow waves recorded using scalp EEG can be easily related to neuronal UP and DOWN states or states of membrane depolarisation and hyperpolarisation. I urge the authors to discuss this more criticall, and amend terminology where relevant because this is misleading.

We agree with the reviewer and we clarified different sections in the article.

1) Modification in the Introduction section of the main manuscript.

1. “More recently, animal and human studies brought to light new evidence of two types of slow waves based on the up and down state duration: one showing positive correlation between consecutive up and down state duration, and another one showing negatively correlated up and down state durations.”

Was replaced by

“More recently, animal and human studies brought to light new evidence of two types of slow waves based on the positive and negative state duration: one showing positive correlation and another one showing a negative correlation between the two.” (p.3, lines 23-25)

2. “Here, we propose to describe the dichotomy in the δ frequency range based on a new parameter characterizing the time delay from the up-to-down state: the transition frequency.”

was replaced by

“Here, we propose to describe the dichotomy in the δ frequency range based on a new parameter characterizing the time delay from the maximum negative point of the EEG slow wave to the maximum positive point of the slow wave: the transition frequency.” (page 4, lines 2-3)

3. “Therefore, our ability to disentangle the influence of slow wave amplitude over our typical metric of the down-to-up state transition, e.g., the slope, is compromised in aging.”

Was replaced by

Therefore, our ability to disentangle the influence of slow wave amplitude over the slope, is compromised in aging. (page 5, lines 2-3)

4. “If $τ$ denotes the delay of the down-to-up-state transition (see Figure 1A)”.

was replaced by

“If $τ$ denotes the delay of the transition from the maximum negative point to the maximum positive point of the slow wave”. (page 8, lines 22-23)

2) Modification in the Discussion section of the main manuscript.

1. “of their transition frequency (down-to-up state transition)”

was replaced by

“(the transition between the maximum negative point and the maximum positive point of the slow wave)”. (page 19, lines 8-9)

2. “is the transition frequency which is strictly associated with the half-waves from the down-to-up state transition”.

was replaced by

“is the transition frequency which is strictly associated with the half-waves related to the depolarization transition of the slow wave.” (page 19, lines 23-24)

3. “Compared to older participants, younger participants seem to have more efficient initiation and termination of slow waves down-state and up-state as they are generating slow waves with a steeper slope”

was replaced by

“Compared to older participants, younger participants seem to have more efficient initiation and termination of slow waves’ transitions as they are generating slow waves with a steeper slope”. (page 23, lines 23-24-25)

2. There is substantial animal literature, which investigated the relationship between EEG, LFP and neuronal population ON and OFF periods or UP/DOWN states. This earlier work in mice and rats addressed the effects of sleep-wake history and ageing on cortical slow waves and their neuronal correlates, including slow wave characteristics, such as slopes and the underlying neuronal synchrony; however, these studies are not mentioned.

We thank the reviewer for his/her suggestion and modifications to the Discussion section of the main manuscript were made to include information about animal literature.

Changes made to the Introduction section of the main manuscript:

“The slope of the slow wave (the rate of amplitude change from the negative to the positive peak) is generally described as the best measure to assess synaptic strength and sleep homeostasis compared to other classic parameters”.

was replaced by

“The slope of the slow wave (the rate of amplitude change from the negative to the positive peak) is associated with the recruitment/decruitment of the neuronal population with a steeper slope showing a quicker recruitment (Vyazovskiy et al., 2011).” (p.4, line 12-15)

Added references:

Vyazovskiy, V. V., Cirelli, C., and Tononi, G. (2011). Electrophysiological correlates of sleep homeostasis in freely behaving rats. Progress in brain research, 193, 17–38.

Changes made to the Discussion section of the main manuscript:

“The animal literature on NREM slow waves parameters shows differences with humans. For instance, compared to humans, there is an age-related increase in frontal LFP δ power (Soltani et al., 2019) and in slow waves’ amplitude and slope in older mice suggesting higher neuronal synchronization (Panagiotou et al., 2017; McKillop et al., 2018). Sleep deprivation protocols also showed higher sleep pressure (Panagiotou et al., 2017) and similar sleep pressure discharge between young and older mice (Wimmer et al., 2013). While sleep researchers are trying to understand and explain the differences (McKillop et al., 2020), the new parameter i.e., transition frequency brings a new angle of analysis and could lead to interesting insight into this problem, for example, by looking at the proportion of slow and fast switchers, their proportion in sleep deprivation protocols and their pattern of homeostatic decline and brain functional connectivity. While more sleep deprivation studies are needed to understand the functional role of slow and fast switchers and their value for the aging brain, looking into slow and fast switchers in animals would enhance in our understanding of the sleeping brain.” (page 24, line 21-25 + p. 25, line 1-8)

Added references:

McKillop LE, Fisher SP, Cui N, Peirson SN, Foster RG, Wafford KA, Vyazovskiy VV.

(2018). Effects of Aging on Cortical Neural Dynamics and Local Sleep Homeostasis in Mice. J Neurosci. 18;38(16):3911-3928.

Panagiotou, M., Vyazovskiy, V.V., Meijer, J.H. and Deboer T. (2017) Differences in

electroencephalographic non-rapid-eye movement sleep slow-wave characteristics between young and old mice. Sci Rep. 3;7:43656.

Soltani, S., Chauvette, S., Bukhtiyarova, O., Lina, J.M., Dubé, J., Seigneur, J., Carrier, J.

and Timofeev, I. (2019). Sleep-Wake Cycle in Young an Older Mice. Front. Syst. Neurosci. 13(51).

Wimmer ME, Rising J, Galante RJ, Wyner A, Pack AI, et al. (2013). Aging in Mice

Reduces the Ability to Sustain Sleep/Wake States. PLOS ONE 8(12): e81880.

3. What is the difference between the new metric proposed here and previously used slow wave slope, which was linked to synchronisation at the ON-OFF and OFF-ON transitions? The authors argue that their approach is based on "a parameter free of the amplitude characteristic of the slow wave", but this has not been demonstrated. It is likely that transition frequency is related to slow wave amplitude, even if the latter is not explicitly included in the calculation. Please provide further analyses to evaluate how transition frequency relates to slow wave amplitude, duration and slopes to enable comparisons with earlier studies.

We thank the reviewer for giving us the opportunity to clarify this important point. The next figure and the following comments explain in more detail how the transition frequency is a slow wave intrinsic feature, giving a unique and non-arbitrary way of classifying slow waves. It also shows that the classification is mainly independent from its slope and amplitude.

Consider Figure 2—figure supplement 1:

We considered all the detected slow waves in the younger and older cohorts. The reasoning is explained with the older cohort, though it can be done with the younger individuals as well. The panel (A) shows the distribution of the transition frequency defined as $\frac{slope}{(2 \times a m p)}$ , $slope$ and $amp$ being the slope and the peak-to-peak amplitude respectively. As shown in the manuscript, this distribution can be accurately fitted with a mixture of two Gaussians: one Gaussian with a mean frequency of 0.93 Hz and the other with a mean frequency of 1.63 Hz. The dashed line in (A) is the crossing of the two Gaussians of the mixture, at 1.2 Hz and defines the threshold that allows separating the slow switchers from the fast switchers.

The graphics in panels (B) and (C) display the distribution of the slopes and the amplitudes. Both distributions are well fitted with a γ and an exponential distribution respectively, as shown with thicker lines. It is worth noting that the distributions of the slope and the amplitudes reflect a unique process whereas the bimodal mixture of the transition frequency distribution reveals the underlying distinction of two types of slow waves.

In order to complement this description, we provide the scatter plot of the panels (D) and (E) that displays the slow switchers (dark grey) and fast switchers (light grey) with respect to the slope or the amplitude and the transition frequency. In (E), the SWs scatter inside a cone defined by the lower and higher peak to peak amplitudes. Although the lower side of the cone is related to the minimal value of the peak-to-peak amplitude as defined in the detector ( $75 μ V$ ), the upper side of the cone indicates that the amplitude peak to peak is bounded and cannot exceed $\sim 433 μ V$ in the older individuals (this value is $\sim 515 μ V$ for the young). It is also worth noting that SWs with slope < 180 $μ V / s e c$ are essentially slow switchers. However, as the SWs’ slope increases, we obtain a mixture of slow and fast switchers with a growing proportion of fast switchers. For a given slope above this threshold (180 $μ V / s e c)$ , we can find both slow and fast switchers. There are a few slow waves with a very large slope ( $>\sim 834 μ V / s e c$ ) that are fast switchers. We interpret that in some circumstances, slow waves with a small or very large slope could have been singularly categorized. Nevertheless, an unambiguous classification of the slow waves can be done only from the transition frequency. Finally, the panel (F) displays the scatter plot of the slow waves with respect to the amplitude and the transition frequency. No correlation can relate those two features of the slow waves.

Changes made to the Introduction section of the main manuscript:

“A novel metric that captures the transition speed without being affected by amplitude needs to be developed.”

Was replaced by:

“A novel metric that captures the transition speed and that is more independent of amplitude needs to be developed.” (p.4, line 20-21)

Changes made to the Discussion section of the main manuscript:

“The use of a parameter free of the amplitude characteristic of the slow wave and associated specifically with the depolarization transition allows us to describe its intrinsic changes in aging.”

Was replaced by

“The use of a parameter more independent of the amplitude characteristic of the slow wave and associated specifically with the depolarization transition allows us to describe its intrinsic changes in aging.” (p.20, line 7-8)

4. Figure 1 illustrates how the cut-off of 1.2 Hz was obtained from the data, but essential details are lacking. Do I understand correctly that the scatter plot shows all individual slow waves pooled across all subjects? Could you show individual subjects separately? Do you still see bimodality at the individual subject level? Does the difference between younger and older groups still manifest when you plot distributions of transition frequencies for individual younger and older subjects, and can this be tested statistically? I wondered if you were to replicate the analyses used by Hubbard et al., 2020 on your data, can you see bimodality they report in mice?

The scatter plot indeed shows all individual slow waves pooled across all subjects (young and older adults). However, in the Results section, the Figure 3(B) and (C) show the Gaussian distributions of the mixture for each subject separately (see legend of figure 3: (B) and (C) The distribution of probabilities of slow waves being slow (cyan) or fast switchers (dark blue) in younger and older individuals, respectively, with each curve representing one participant.). The figure therefore shows this bimodality occurring at the individual level as well as for younger and older adults independently. Now, Figure 3—figure supplement 2 shows the same bimodality when we look at slow waves with and without concomitant sleep spindles.

Author response image 1

Download asset Open asset

In Hubbard et al. (2020), the authors describe two sets of slow waves defined when comparing sleep-deprived mice to control ones and using different parameters than the transition frequency. Although it is not possible to look at sleep deprivation in our current data as none of the participants underwent a sleep deprivation protocol, we do see one type of slow wave being more prevalent than the other at the beginning of the night (the fast switchers) when the homeostatic pressure is at its highest (See Figure 3). The fact that Hubbard et al. (2020) show that slow waves “accelerate” after a prolonged waking period tends to corroborate our findings. Figure 1 in our paper shows that we couldn’t have defined slow and fast switcher based on the frequency of the slow waves and Figure 2—figure supplement 1 shows that we couldn’t have used the slow waves’ slope nor the amplitude.The legend of Figure 3 was clarified to emphasize the mixture of Gaussians present at the individual level as displayed in Figure 3(B) and (C).

Changes made to the Discussion of the manuscript:

Addition: “When looking at the usual frequency of slow waves, Hubbard et al. (2020) show that prolonged waking periods are followed by a higher prevalence of faster waves at the beginning of the sleep period” (p.23, lines 5-7).

5. It is stated in Methods: "All slow waves were equally time referenced by choosing the zero phase at the maximum of the depolarization." It is not clear what is "maximum" here? I presume it is not uncommon that there are multiple positive peaks on the "depolarisation" part of the slow wave, which complicates this analysis. Was the number of peaks within a slow wave different between older and younger subjects? I wondered if the two categories of slow waves correspond to those with one peak and two peaks riding on the top of the "depolarisation/up" phase?

The maximum of the depolarization phase is illustrated in Figure 1A and was defined according to our detection criteria (i.e., defined as the absolute maximum of amplitude). The same points are usually used to define the slope and amplitude of slow waves in human studies. We therefore did not identify a multiple local maximum within the slow wave that could be interpreted in various ways, including the possible ringing effect of the filter. Our analyses considered the unambiguous global maximal peak for each phase of the slow wave. Moreover, visual inspection of the slow wave shows marginal occurrence of other local maximum peaks on the filtered signal. Although small wavelets might be occasionally identified in the main slow wave phenomenon when using surface EEG in humans, it can be very difficult to interpret the underlying neurophysiology because of the spatial resolution (the dense coverage of the HD-EEG may help to reproduce the local peaks across neighboring derivations).

6. I wondered if the "critical frequency" or the dynamics of "fast" and "slow" switchers would change (or an additional category emerge) if you filter the signals at a higher frequency, such as at 6 Hz, rather than 4 Hz (e.g. see https://pubmed.ncbi.nlm.nih.gov/10607082/)

This question raised by the reviewer is indeed interesting, but we didn’t analyze higher frequencies as our objective was to investigate slow waves in the usual frequency range as defined in the human literature. Even if we were to have used a 6Hz filter, other parameters used to identify SW (negative peak below -40 µV, a peak-to-peak amplitude above 75 uV, the duration of negative deflection between 1500 and 125 ms, and the duration of positive deflection not exceeding 1000 ms) would limit the number of waves identified between 4Hz and 6Hz and would thus unlikely move the transition frequency delimiting slow and fast switchers. Moreover, as seen in Figure 1 in the manuscript, the transition frequency is upper bounded and an investigation at higher frequency would likely not exhibit an other class of switchers.

7. The analysis of EEG connectivity is interesting, but I wondered if there is a relationship between occurrence of multiple peaks on the depolarisation phase and connectivity? Riedner et al. study suggested that multipeak slow waves can reflect "collision" of slow waves originating from distant cortical areas: https://pubmed.ncbi.nlm.nih.gov/18246974/

As mentioned previously, the occurrence of multiple local maxima on the filtered slow wave was marginal and difficult to interpret. The increase in functional connectivity happening in young adults during the depolarization phase of the slow wave is at its highest at the positive peak. Being able to identify “colliding” slow waves from distant cortical areas would definitely be interesting but is unfortunately out of the scope of this manuscript, especially with the EEG montage we used. This would be easier to assess with a high-density EEG that may be able to capture such interference with a higher spatial resolution.

Reviewer #2:

Here, the authors present data regarding the occurrence of (relatively speaking) slow and fast onset slow waves in younger and older adults, and some phenomenology surrounding their occurrence. There is a reasonable argument to be made that this may represent an actual dichotomy – e.g. bimodal distribution of onset time (essentially the slope). The authors claim that this is a feature independent of wave amplitude, but no data is shown to this effect. This makes the interpretation of the significance of these data problematic, as many of the features about "fast switcher" waves are already well known regarding high-amplitude waves. Thus it is hard to know what is really new here – as much is already known about changes in amplitude with aging.

The authors have developed some new metrics to characterize features of slow wave waveforms during NREM sleep. They propose that there are two classes of slow waves, and show some suggestive data that is indeed the case. For example, they show a clear bimodal distribution of transition frequencies (seemingly a measure of transition slope). It is not clear how this relates to wave amplitude- data addressing this are not shown – but it stands to reason that this feature maps onto wave amplitude very well. Because much is already known about amplitude of waves' relationship to synchrony across recording sites, and aging, and occurrence across the night, it is very unclear if this is just a retelling of the same story using a different metric.

An analysis of how the new metric relates to waves' amplitudes is absolutely necessary for the readers to know if there is anything clearly different between this study and what is already known.

We thank the reviewer for giving us the opportunity to clarify this important point. Figure 2—figure supplement 1 and the following comments explain in more detail how the transition frequency is a slow wave intrinsic feature, giving a unique and non-arbitrary way of classifying slow waves. It also shows the classification is mainly independent from its slope and amplitude.

The graphics in panels (B) and (C) display the distribution of the slopes and the amplitudes. Both distributions are well fitted with a gamma and an exponential distribution respectively, as shown with thicker lines. It is worth noting that the distributions of the slope and the amplitudes reflect a unique process whereas the bimodal mixture of the transition frequency distribution reveals the underlying distinction of two types of slow waves.

A fair bit of the methodology around analyses data is not present, making it questionable what different figures are really showing – for example, it is unclear which stages of NREM sleep are used for various measures, and if these waves' occurrence differs between those stages.

Slow waves and sleep spindles were detected on both N2 and N3 sleep stages. Both stages were thus used in our measures in order to include all slow waves and spindles.

See “Slow waves were detected automatically on artifact-free NREM (N2 and N3) epochs on all electrodes using previously published criteria” (page 7, line 20-21)

and

“Spindles were automatically detected on artifact-free NREM (N2 and N3) epochs on all electrodes using a previously published algorithm” (page 8, line 4-5)

The information about NREM sleep stages used was added to the legend of Figure 1, 2, and 3 of the main manuscript.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there was one remaining issue that needs to be addressed, as outlined below:

While the reviewers agreed that the manuscript was substantially improved through revisions, the consensus was that further consideration in the discussion of the study findings was needed in the context of the publication by Geering et al. (https://pubmed.ncbi.nlm.nih.gov/10607082/), which reports the differential temporal dynamics of slow waves at different frequencies and the impact of analysis methods on their metrics.

We want to thank the reviewers and editors for pointing us this article. The paper is interesting as it introduces the idea that half waves can bring complementary information to spectral analyses. We recognize there is similarities between this work and ours. However, there are also important methodological differences which prevent our results to be contaminated by some bias noted in the article.

The internal frequency we discuss in our paper is based on a different rationale than what proposed by Geering et al. We studied the frequency of the depolarization transition of detected sleep slow waves, whereas they analyzed the entire high passed EEG signal and they focused on half waves obtained from zero-crossing. Using zero crossings can introduce a bias as it can be arbitrarily manipulated through the choice of high pass filter, a notion the authors acknowledged themselves at the end of their discussion. It is worth noting that the authors also mentioned the ‘through to peak’ transition still using zero-crossing points but on the first order derivative of the signal. This is more in the spirit of our work. However, in our study, we only considered the detected slow waves from which we extracted ‘through to peak’ transition directly. Despite some correspondence in the kind of local characteristics, the two approaches are very different in terms of the signals under consideration. We argue that using the depolarization transition with the max and min peak of detected slow-waves allow us to evaluate a more physiological process, free of the methodological bias discussed in Geering et al., paper.

As suggested by the reviewer, we added a section at the beginning of the discussion to acknowledge their scientific contribution.

This paragraph was added to the Discussion section of the main manuscript.

The frequencies related to half-wave components of filtered EEG were introduced decades ago in order to provide an alternative to the time-resolved spectral analysis of sleep (Geering et al., 1993).

Such studies were mostly concerned with the half-waves defined by the zero-crossings of the entire high pass filtered EEG signals, although no consensus was reached. The present work introduces a new parameter in which half-waves and the associated frequency are defined from the depolarization transition of detected sleep slow waves. This intrinsic parameter, the transition frequency, objectively classifies sleep slow waves in humans into two categories: the slow and fast switchers. (p.15, Lines: 2-7).

https://doi.org/10.7554/eLife.64337.sa2

Article and author information

Author details

Maude Bouchard
1. Center for Advanced Research in Sleep Medicine, Hôpital du Sacré-Coeur de Montréal, Montreal, Canada
2. Department of Psychology, Université de Montréal, Montreal, Canada
Contribution
Conceptualization, Investigation, Methodology, Writing - original draft, Writing - review and editing

For correspondence
maude.bouchard.1@umontreal.ca

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-7858-4136
Jean-Marc Lina
1. Center for Advanced Research in Sleep Medicine, Hôpital du Sacré-Coeur de Montréal, Montreal, Canada
2. Department of Electrical Engineering, École de Technologie Supérieure, Montreal, Canada
3. Centre de Recherches Mathématiques, Université de Montréal, Montreal, Canada
Contribution
Conceptualization, Formal analysis, Supervision, Writing - review and editing

Competing interests
No competing interests declared
Pierre-Olivier Gaudreault

Center for Advanced Research in Sleep Medicine, Hôpital du Sacré-Coeur de Montréal, Montreal, Canada

Contribution
Formal analysis, Writing - review and editing

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0002-9863-4436
Alexandre Lafrenière

Center for Advanced Research in Sleep Medicine, Hôpital du Sacré-Coeur de Montréal, Montreal, Canada

Contribution
Writing - review and editing

Competing interests
No competing interests declared
Jonathan Dubé
1. Center for Advanced Research in Sleep Medicine, Hôpital du Sacré-Coeur de Montréal, Montreal, Canada
2. Department of Psychology, Université de Montréal, Montreal, Canada
Contribution
Writing - review and editing

Competing interests
No competing interests declared
Nadia Gosselin
1. Center for Advanced Research in Sleep Medicine, Hôpital du Sacré-Coeur de Montréal, Montreal, Canada
2. Department of Psychology, Université de Montréal, Montreal, Canada
Contribution
Writing - review and editing

Competing interests
No competing interests declared
Julie Carrier
1. Center for Advanced Research in Sleep Medicine, Hôpital du Sacré-Coeur de Montréal, Montreal, Canada
2. Department of Psychology, Université de Montréal, Montreal, Canada
Contribution
Resources, Supervision, Funding acquisition, Writing - review and editing

For correspondence
julie.carrier.1@umontreal.ca

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0001-5311-2370

Funding

Canadian Institutes of Health Research (Vanier scholarship)

Maude Bouchard

Canadian Institutes of Health Research (190750)

Julie Carrier

Natural Sciences and Engineering Research Council of Canada (NSERC-Discovery Program)

Jean-Marc Lina
Julie Carrier

Canadian Institutes of Health Research (Postdoctoral Fellowship)

Pierre-Olivier Gaudreault

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors would like to thank Sonia Frenette for her help with data collection and analysis and Carrie Schipper, for reviewing the manuscript. This work was supported by the Canadian Institutes of Health Research (CIHR), grant number 190750 (JC), the NSERC-Discovery programs (J-ML and JC), the CIHR Vanier scholarship (MB), and a CIHR postdoctoral award (P-OG).

All codes (Lina, 2021a) and transformed data used for all the analyses and most specifically to produce all of the figures of this article can be freely accessible using this link: https://github.com/jmlina/Slow_Wave_Switchers, (copy archived at swh:1:rev:3af1be579bd6d5aac1718d49c85e8af9c17541c1), Lina, 2021b.

Ethics

Human subjects: The protocol was approved by the ethics committee of the Hôpital du Sacré-Coeur de Montréal and performed in accordance with the relevant guidelines and regulations. Participants provided informed consent and received financial compensation for their participation. (CMER-RNQ 08-136 08-002).

Senior Editor

Chris I Baker, National Institute of Mental Health, National Institutes of Health, United States

Reviewing Editor

Bryce A Mander, University of California, Irvine, United States

Reviewer

Ksenia V Kastanenka, MassGeneral Institute for Neurodegenerative Disease, Massachusetts General Hospital, Harvard Medical School, United States

Version history

Received: October 26, 2020
Preprint posted: February 3, 2021 (view preprint)
Accepted: August 26, 2021
Accepted Manuscript published: August 27, 2021 (version 1)
Version of Record published: September 20, 2021 (version 2)

Copyright

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.