A preliminary investigation of the role of slow-wave activity in modulating waking EEG theta as a marker of sleep propensity in major depressive disorder

Abstract

Background:

Both EEG slow-wave activity (SWA) during sleep and EEG theta activity during waking have been shown to increase with extended waking, and decrease following sleep, suggesting that both are markers of sleep propensity. In individuals with major depressive disorder (MDD), however, altered patterns of SWA have been noted, suggesting that sleep homeostasis is dysregulated. This study aimed to examine if slow-wave disruption would alter sleep propensity differently in healthy controls (HC) and those with MDD.

Methods:

25 individuals (13 diagnosed with MDD and 12 HC) participated. Following one night of adaptation sleep, participants underwent one night of baseline sleep, and one night of selective slow-wave disruption by auditory stimuli. In the evening, before sleep, and in the morning following sleep, waking EEG was recorded from participants in an upright position, with eyes open.

Results:

Repeated measures ANOVA revealed a significant three-way interaction, such that AM theta activity was significantly lower following slow-wave disruption in those with MDD, but not in HC. Additionally, SWA was not correlated with theta activity in MDD.

Limitations:

These data are based on a relatively small sample size of unmedicated individuals with MDD. Conclusions: These data may suggest that SWA plays a differential role in the homeostatic regulation of sleep in HC, and in MDD, and provide additional evidence that the presence of SWA may be maladaptive in MDD.

Introduction

The process of sleep is under homeostatic regulation in humans, similar to internal temperature and glucose concentration within the blood. Behaviorally, an individual’s need for sleep increases over the course of continuous waking, as evidenced by increased sleepiness and fatigue, and biological indices such as increased slow-eye movements and delayed reaction time (Cajochen et al., 1999). Following a period of sleep, this drive for sleep, otherwise referred to as sleep propensity, decreases demonstrating the fundamental principle of homeostasis (Borbély, 1982). Experimental homeostatic sleep challenge protocols, including total sleep deprivation experiments, have allowed researchers to identify sleep slow-wave activity (SWA) as a physiological marker of sleep propensity (Borbély et al., 1981). This is because SWA (0.5-3.9Hz), derived from the quantitative analysis of sleep electroencephalography (EEG), has been shown to increase as a function of prior wakefulness (greater SWA power indicating longer periods spent awake), and decrease across a period of sleep (Dijk, 1995). In addition to the sleep EEG, there is evidence to suggest that the waking EEG may provide an additional biological marker of sleep propensity (Finelli et al., 2000; Aeschbach et al., 2001). During sleep deprivation protocols, theta activity (4-7 Hz) during waking increases, in a manner similar to SWA (Vyazovskiy & Tobler, 2005). Additionally, activity from a similar frequency range (3.25–6.25Hz) has been shown to decrease following a period of sleep (Plante et al., 2013).

In major depressive disorder (MDD), sleep architecture differs from that of healthy individuals. These differences include difficulties with sleep continuity, in addition to alterations in REM sleep such as decreased REM latency and increased REM density, and decreases in slow-wave sleep (SWS; Benca et al., 1992). These SWS differences have led some researchers to hypothesize that individuals with MDD have impairments in sleep homeostasis (Borbely et al., 1984; Plante et al., 2013; Goldstein et al., 2012). In fact, our group demonstrated that following a mild homeostatic sleep challenge, some individuals with MDD exhibit alterations in SWA such as decreased accumulation and slower dissipation of SWA that are not present in healthy individuals (Goldschmied et al., 2014). Similarly, Plante and colleagues (2013) demonstrated that individuals with MDD did not show a post-sleep reduction in theta-range activity that was evident in healthy individuals. Taken together, these studies provide evidence of impairments in the homeostatic regulation of sleep in some individuals with MDD.

Interestingly, in healthy individuals the reduction of theta-range activity following sleep is correlated with the amount of SWA during sleep (Plante et al., 2013), such that the more SWA an individual exhibits, the greater the reduction in theta-range activity the next morning. This relationship between SWA and measures of sleep propensity have prompted researchers to propose that in addition to serving as a marker of sleep propensity, SWA may actively play a role in reducing it (Tononi & Cirelli, 2003). In this way, the presence of SWA during sleep may be responsible for the subsequent reduction in theta activity during waking. In those with MDD, however, SWA and theta-range activity are not correlated (Plante et al., 2013), suggesting that sleep SWA may not serve the same function in those with MDD as it does in healthy individuals.

The aim of this study was to explore the changes in EEG theta activity in a sample of healthy (HC) and depressed individuals following one night of baseline, and one night of selective slow-wave disrupted sleep (SWD). Given that sleep including SWA was associated with reduction in low frequency activity in HC, and was not associated with changes in MDD, we hypothesize that SWD would lead to an attenuated decrease in theta activity in HC following SWD, and that SWD would not be associated with changes in theta in those with MDD.

Methods

Participants

The present sample included 15 individuals who met criteria for MDD (ages 18-48) based on the Structured Clinical Interview for DSM-IV, and who were medication-free for at least 6 weeks prior to study, in addition to 13 HC (ages 18-45) who did not meet criteria for current or past MDD. Inclusion criteria included two consecutive days of electroencephalographic recording without any difficulties or deviations from the protocol, English fluency, habitual and consistent sleep time between 6 and 8 h, with a habitual bed time between 10 pm and 12 am. Exclusion criteria included current use of medications (within 6 weeks) that are thought to impact sleep, including antidepressants, history of serious, unstable medical illnesses including but not limited to neurological and sleep disorders. The study was carried out in accordance with the latest version of the Declaration of Helsinki, was approved by the Institutional Review Board at the University of Michigan, and all participants provided written, informed consent.

Experimental design

Participants spent three consecutive nights in the sleep lab (adaptation, baseline, slow-wave disruption), with a nightly 7-h sleep opportunity. As this was a sleep challenge study, it was necessary to fix total sleep time across participants. We specifically recruited those who slept 6–8 h, habitually, guaranteeing that participants would have a sleep opportunity that would not differ more than 1 h from their habitual sleep schedule. Sleep diary and actigraphy were used to verify that participants maintained one week of consistent bedtimes and wake-times, and refrained from napping.

Measures

Beck Depression Inventory, 2^nd edition (BDI-II; Beck et al., 1996)

The BDI-II is a 21-item, self-report rating scale designed to assess the severity of depressive symptoms. Scores range from 0 to 63, with higher scores indicating greater severity of depressive symptomatology. BDI scores were also used as a diagnostic cutoff with a maximum score of 12 for the HC group and a minimum score of 14 for the MDD group.

Electroencephalography (EEG) Recording and Processing

Sleep and wake EEG, electromyogram (EMG), and electrooculogram (EOG) were collected in accordance with standardized techniques (see Goldschmied et al., 2014). All electrophysiological signals were acquired with a Vitaport™ III digital data acquisition system, with an equivalent sensitivity of 5 (50 μV, 0.5-s duration calibration) corresponding to a gain of 50,000. EEG filter settings were set at 0.3 and 70 Hz in order to attenuate electrical noise. All data were digitized at 256 Hz (Armitage et al., 2012).

The sleep electrode montage was composed of left and right frontal central, parietal, and occipital electrode sites (F3, F4, C3, C4, P3, P4, O1, O2), placed according to the International 10–20 System. EMG was recorded from a bipolar chin-cheek montage. EOG recordings were made from the supraorbital and infraorbital ridges of the eyes. The waking electrode montage was composed of nine electrodes (F3, Fz, F4, C3, Cz, C4, P3, Pz, P4) and EOG consisting of two electrodes placed lateral to the left canthus and above the right supraortibal ridge. For both sleep and wake EEG, a reference electrode, comprised of linked earlobes and passed through a 10 KΩ resistor, was utilized to minimize potential artifact. All EEG impedances remained below 2 kΩ. EEG data were visually scored following standard criteria in 30-sec epochs (Rechtschaffen & Kales, 1968). EEG was visually inspected epoch by epoch, and any epochs that contained movement, breathing muscle artifact, electrical artifact, baseline shift, or electrode problems were omitted from further analysis. In general, artifact rejection resulted in the exclusion of less than 5% of epochs.

Power spectral analysis (PSA) was performed on digitized EEG signals using an algorithm based on the fast Fourier transform (Press et al., 1989). Data were processed in 2-sec epochs (512 samples for each 2-sec) with a Hanning window taper, generating power (expressed as μV²) in the five frequency bands of the full EEG spectrum. For sleep EEG, spectral power was then averaged in 30 s epochs to correspond with stage scores (Armitage et al., 2000a; Armitage et al., 2000b). Analyses for the present paper were restricted to delta power (0.5-3.9 Hz). For waking EEG, analyses were based on theta power (4-7.9 Hz) averaged from frontal electrodes (F3, Fz, F4), based on the previous literature (Plante et al., 2013).

The delta power data was subsequently sorted by non-REM (NREM) period on baseline and SWD nights, using a definition of NREM period that matched previous studies (Armitage et al., 2000a; Armitage et al., 2000b). NREM period was defined as the succession of stages 2, 3, or 4 of ≥ 15-min duration and terminated by stage REM or a period of wakefulness of ≥ 5 minutes. Stage 1 sleep epochs were excluded. No minimum REM duration was required for the first or last REM period. For each subject, delta power was then summed and averaged relative to the number of epochs in each NREM period, henceforth, referred to as SWA.

Waking EEG activity was recorded with participants in an upright position, with eyes open for 6 minutes, in the evening prior to sleep, and in the morning following sleep. Participants were asked to look at a fixation cross for the duration of the experiment which was presented in white font on a uniform black background, centered on a 19” computer monitor, located approximately 80 cm from the participant, utilizing E-prime software. Participants were also given the following instructions, “During this experiment, try to relax and refrain from thinking about anything specific. You will also need to stay as still as possible during each 6 min condition, so please make sure you are comfortable.”

Selective slow-wave disruption procedure (SWD)

For the selective SWD procedure validated previously (Ferrara et al., 1999), EEG Channels C3 and C4 are continuously inspected during sleep. Acoustic tones of 1s duration (frequency = 1000 Hz; intensity = 20–100 dB) are delivered via earphone whenever two delta waves (1–4 Hz; >75 μV) are detected within 15 s of each other. The tone presentation begins with the lowest intensity (20 dB) and is increased by 5 dB intervals until a response occurs (sleep stage shift, K complex, EEG desynchronization, mixed and fast frequency, alpha burst, muscle tone increase, slow eye movements). The procedure is repeated at the next occurrence of 2 delta waves meeting criteria. See Cheng et al., 2015 for further detail.

Statistical analysis

One-way analysis of variance (ANOVA) was computed for each demographic and sleep variable to examine any differences between groups. Repeated measures ANOVA were computed for each macroarchitectural and microarchitectural variable to assess change between condition (baseline, SWD) for HC and MDD groups, separately. To test changes in theta activity, a three-way, repeated measures ANOVA was conducted with group (HC, MDD) as the between subjects factor, and time of day (AM, PM), and condition (baseline, SWD) as within subjects factors.

To determine if modulation of theta activity was directly associated with change in SWA, regression analyses were conducted for both baseline and SWD with change in waking theta activity from evening to morning serving as the dependent variable, and change in SWA from the first NREM period to the third NREM period as the predictor. Only the first three NREM periods were included for analysis, as not all subjects had four or more NREM periods across the night.

Exploratory regression analyses were also conducted to examine whether the sleep architectural changes resulting from SWD (Change in % Stage2, % REM, and % Awake and Movement) were associated with changes in waking theta activity in MDD.

Results

From the original sample of 28 participants (15 MDD), waking EEG data from 3 participants (MDD) were excluded from analysis due to identification of extreme outliers in waking theta power according to the Grubbs Test (Grubbs, 1950).

Following SWD, average SWA across the night, F(1, 11) = 7.44, p < 0.05; F(1, 9) = 7.78, p < 0.05, and SWA in the 1st NREM period, F(1, 11) = 17.98, p < 0.01; F(1, 9) = 17.98, p < 0.01, were reduced for both HC and those with MDD, respectively (see Table 1). In addition to reductions in SWA and SWS, HC also demonstrated an increase in stage 1, F(1,11) = 7.75, p<0.05, and stage 2 sleep, F(1,11) = 6.27, p<0.05, and those with MDD demonstrated increases in stage 2, F(1,10) = 18.51, p<0.01, and decreases in percent REM, F(1,10) = 7.90, p<0.05.

Table 1.

Means and standard deviations of polysomnographic variables, by group and condition.

		Healthy Controls (n=13)	Depressed (n=12)
Demographics
	Sex	7 F	6 F
	Age	25.69 (8.1)	26.25 (10.4)
	BDI	2.08 (3.0)	23.58 (6.9)†

Sleep Parameters	Baseline	SWD	Baseline	SWD
Total Sleep Period (min)	418.9 (24.0)	412.4 (26.0)	415.9 (43.6)	403.8 (43.1)
Sleep Latency (min)	8.2 (7.4)	11.3 (10.4)	10.4 (7.1)	19.3 (14.4)
% Stage 1	3.9 (2.2)	5.7 (2.6)*	4.3 (3.2)	5.6 (4.7)
% Stage 2	51.5 (10.1)	58.0 (8.3)*	53.6 (5.8)	61.6 (7.9)*
% Slow-wave Sleep	16.7 (6.5)	8.3 (8.4)*	15.1 (8.2)	6.0 (6.0)*
Awake & movement (%)	2.9 (1.6)	6.7 (6.1)*	3.1 (1.8)	6.9 (5.1)*
REM (%)	25.0 (9.8)	21.3 (6.6)	23.9 (5.0)	19.9 (7.8)*
REM Latency (min)	90.1 (55.8)	87.3 (50.1)	59.4 (38.6)	110.8 (83.9)
Average all-night SWA (μV2)	463.1 (99.3)	408.4 (68.0)*	426.1 (116.4)	386.4 (113.0)*
SWA (μV2) in 1st NREM	595.7 (161.5)	449.5 (90.7)*	572.8 (170.5)	417.8 (125.5)*

^†denotes significant between group difference

^*denotes significant within group difference

Regarding waking theta activity, results revealed a significant three-way interaction, Condition x Time x Group was significant, F(1,23)=5.33, p<.05, indicating that the pattern in which theta activity changed across time and condition was different between HC and those with MDD. No other main effects or interactions were detected (see Figure 1). Post-hoc contrasts revealed that in those with MDD, theta activity significantly decreased in the morning following SWD, F(1,11) = 8.9, p=0.01, whereas theta did not change in HC. No other pairwise comparisons were significant.

Figure 1.

Waking EEG Theta Activity by time of day, and condition

* p<.05

Regression analyses were conducted to examine whether the reduction in SWA during baseline and SWD was predictive of change in waking theta activity. Results revealed that in HC reduction in SWA from NREM 1 to NREM 3 was significantly associated with reduction in waking theta from evening to morning, β=0.612, t(9)=2.32, p<0.05, but only on the baseline night (see Figure 2). There was no significant association between reduction in SWA and waking theta activity on the SWD night. For the MDD group, reduction in SWA was not associated with reduction in theta activity on the baseline or SWD night.

Figure 2.

Scatterplot of change in waking theta activity with change in SWA activity on baseline night, HC and MDD

Exploratory regression analyses were also conducted to examine whether the decline in waking theta activity in MDD following SWD was significantly associated with changes to sleep architecture resulting from the SWD. Results indicated that change in waking theta activity was not directly related to changes in % Stage 2, % REM, or % Awake and movement.

Discussion

This study aimed to explore changes in waking EEG theta activity, theorized to be a measure of sleep drive, in healthy and depressed individuals following SWD. Our data demonstrated that SWA seems to play a fundamentally different role in the homeostatic regulation of sleep in healthy individuals than in those with MDD.

Previous studies have shown that sleep decreases waking EEG theta activity (i.e. sleep propensity) in HC (Plante et al., 2013); however, our data demonstrated that waking theta activity did not decrease following SWD, suggesting that SWA may be a necessary component for reducing waking theta activity and thus sleep propensity. This idea is in line with other work suggesting that SWA is functionally important. For example, the Synaptic Homeostasis Hypothesis (SHY; Tononi & Cirelli, 2003) postulates that SWA reflects two distinct phenomena. On one hand, SWA acts as a marker of sleep propensity and, by extension, of net synaptic strength (Tononi & Cirelli, 2006), a measure of neuronal connectivity that has been shown to increase with wakefulness (Liu et al., 2010). On the other hand, it is posited that SWA functionally facilitates the downscaling of synaptic strength. Because the presence of SWA is necessary for the reduction of synaptic strength, it follows that SWA would also be necessary for the reduction of sleep propensity.

In MDD, our data indicated that the presence of SWA was not associated with any modulation of theta activity. This result is also consistent with that found by Plante and colleagues (2013). One explanation is that, unlike in HC, SWA may actually prevent the reduction of sleep drive in MDD. The idea that SWA may be maladaptive in MDD has been postulated previously (Beersma & Van den Hoofdakker,1992), and is in line with two recent studies showing that reducing SWA improves mood (Cheng et al., 2015; Landsness et al., 2011). Furthermore, in contrast to our original hypotheses, SWD resulted in a reduction of waking theta activity in those with MDD, indicating that sleep propensity had decreased across the sleep period. These data suggest that sleep with disrupted SWA reduces waking theta activity and by extension, sleep propensity, further supporting the idea that SWA may be maladaptive in MDD. In this way, these data may establish that SWA plays a fundamentally different role in the modulation of sleep propensity, and in the homeostatic regulation of sleep between healthy and depressed individuals.

Plante and colleagues (2013) demonstrated that following baseline sleep, reductions in waking theta-range activity were correlated with sleep SWA in healthy individuals. Our results in HC replicated this finding, and showed that a greater reduction in SWA from NREM 1 to NREM 3 was predictive of a greater reduction in waking theta activity from evening to morning following baseline sleep; however, this pattern was not found in those with MDD, or following SWD. Additionally, because the reduction in waking theta activity in MDD following SWD was unexpected, we were interested to see if the reductions in theta activity in MDD were associated with objective changes in the sleep EEG. However, we were unable to identify a relationship between the change in waking theta with any sleep variables, including the degree of the reduction in SWA or reduction in REM. This may indicate that the decrease in waking theta activity may not be the direct result of any singular and specific sleep parameter change, but may be caused by more global changes in sleep dynamics occurring during SWD. Future work should attempt to identify the mechanism underlying the decrease in waking theta following SWD in order to better understand the impairments in sleep homeostasis in MDD.

Limitations

While the present results extend our understanding of SWA in depression, interpretations should be contextualized within study limitations. First, as this was a small-scale preliminary investigation into the role of SWA and its interplay with waking theta activity in depression, future studies should seek to confirm these results in larger samples. An increase in study power would be particularly important in order to examine interesting yet non-significant trends in the present data. For example, it appears that in a larger sample, those with MDD may have demonstrated overall decreased waking theta power as compared to the HC. As waking theta power is a proposed marker of sleep propensity, this may suggest that those with MDD have a lower sleep need than HC, which is in line with the S-deficiency hypothesis of depression (Borbely & Wirz-Justice, 1982). Second, this study relied on waking theta activity as a marker of sleep propensity. Future studies should also include additional objective measures of sleep need including the Multiple Sleep Latency Test, or momentary measures of sleepiness including the Karolinska Sleepiness Scale as convergent evidence. Lastly, this study used only a limited number of electrodes to quantify EEG changes which may have limited our ability to draw conclusions with regard to SWA dynamics. Future studies should aim to high-density EEG in order to examine the impact of topography on altered SWA in MDD.

In summary, this study investigated the effects of SWD on waking theta activity, a proxy measure of sleep propensity, in a sample of depressed and healthy individuals. Our results demonstrate that waking theta activity decreased in MDD, but not in HC, suggesting that SWA plays a differential role in the homeostatic regulation of sleep in those with MDD. Future work in this area will be needed to characterize the relationship between SWA, sleep homeostasis, and mood.

Highlights.

• Waking EEG theta activity and sleep slow-wave activity (SWA) are considered biomarkers of sleep propensity, increasing with wakefulness and decreasing with sleep.

• Following normal sleep, a greater reduction of SWA from the beginning to the end of the night was associated with a greater reduction in waking EEG theta activity from the evening to the morning in healthy controls, but not individuals with major depressive disorder (MDD), adding to the mounting evidence that sleep homeostasis is dysregulated in MDD.

• Following experimental slow-wave disruption, individuals with MDD, but not healthy controls, demonstrated a significant decrease in waking theta activity, which suggests that SWA may play a fundamentally different role in the homeostatic regulation of sleep in healthy individuals than in those with MDD, and may also suggest that SWA could be maladaptive in MDD.