Local sleep in an awake brain

Local sleep in an awake brain

Karlijn van Heijst

27 July 2018

Student ID: 10364196

MSc Brain and Cognitive Sciences, University of Amsterdam, track Behavioural Neuroscience

Internship period: 4 December 2017 – 27 July 2018 (42 EC) Brain & Cognition, University of Amsterdam

Daily supervisor: Esperanza Jubera-García Examiner 1: Filip van Opstal

Abstract

Though previously regarded only as a global whole brain phenomenon, more and more converging evidence supports a local regulation of sleep. Recent studies in both animal models and human participants have shown specific brain areas can go into a sleep-like activity state following prolonged performance of a task engaging these areas. What’s more, they also report a temporal relationship between occurrence of these sleep-like states and worsening performance. However, in these studies subjects were sleep-deprived and results were found in relatively large brain areas. Therefore, in the current project we aimed to find local sleep-like activity in the occipital cortex of healthy, awake and non-sleep-deprived human participants following performance of a visual task. In humans, local sleep-like activity can non-invasively be measured as a local increase in slow frequency oscillations in electroencephalogram (EEG) recordings, specifically of the delta and theta frequency range. Our results show local increases in both delta and theta activity in the occipital region after repeated within-day performance of the task. Further analyses will be carried out to investigate a possible temporal relationship. While our results do further support a possible localized mechanism behind sleep, more interesting might be the cognitive effect associated with the local sleep-like states now for the first time observed in non-sleep-deprived humans.

Contents

Introduction

4

Methods

6

Testing schedule and materials

6

Participants

7

EEG recordings

7

Texture discrimination task

7

Visual analogical scales and psychomotor vigilance task

8

Data analysis

10

Data elimination

11

Results

11

Texture discrimination task

11

Resting state EEG

12

Discussion

14

4

Introduction

Sleep is a physiological phenomenon occurring in all animals to a certain extent. Despite this wide prevalence, there is no clear consensus in the literature as to what the function of sleep is. Some suggest it only exists to kill time when animals are not engaged in other activities (Siegel, 2009), while others claim sleep does serve different functions, and specifically plays a role in brain function. One idea, for example, states that sleep is important for pruning of neuronal connections (Cirelli & Tononi, 2017). Without the influence of constant sensory input, according to this ‘synaptic

homeostasis hypothesis’, during sleep some connections are strengthened, while others are weakened, leading to a net downscaling of connections. This downscaling effect could be crucial for overall brain functioning (Cirelli & Tononi, 2017). Another theory suggesting sleep is important for brain functioning, is that of the ‘brain drain’, as proposed by Nedergaard and Goldman (2016). This theory states that sleep plays an important function in clearing toxic protein wastes and other biological debris that builds up in the brain during the day and that this ‘cleaning’ process is essential for cognitive functioning.

Apart from the function of sleep still being under debate, there are also different views on the mechanisms behind sleep, and more specifically sleep onset. Traditionally sleep onset was viewed as a global brain process initiated in subcortical structures (Zielinski, McKenna, & McCarley, 2016).However, an alternative view suggests sleep starts as a local phenomenon that can spread out to the whole brain if ‘enough’ local areas are included (Krueger, Frank, Wisor, & Roy, 2016; Krueger et al., 2008). This hypothesis provides a very different view on mechanisms behind whole brain sleep onset, but also links it to a brain connectivity function of sleep. More specifically, this view too suggests that specific groups of neurons or brain areas can transition into a sleep-like state following a period of activity (Krueger et al., 2016; Krueger, Huang, Rector, & Buysse, 2013; Krueger & Roy, 2016).

The first evidence indicating a more local regulation of sleep, was that some animal species can show sleep in only one hemisphere, while the other is awake, so called ‘unihemispheric’ sleep (Mascetti, 2016). For example, it has long been known that some species of dolphins show unihemispheric sleep, and a recent study showed evidence that migratory frigate birds can show unihemispheric sleep in flight (Rattenborg et al., 2016).

Beside this ‘half-brain’ sleep, additional studies have shown that even smaller sections of the cortex can individually go into sleep-like states. For example, Vyazovskiy and colleagues (2011) recorded individual neuronal activity and surrounding local field potentials (LFP’s) in the frontal motor cortex of sleep-deprived rats that were performing a demanding motor task. Their results show that populations of neurons that previously showed activity, could briefly go into states of decreased activity (or even complete inactivity) which they termed ‘off’ states. The occurrence of the short ‘off’ periods was often associated with local slow/theta waves (2-6 Hz) in the LFP. Moreover, the results of this study further suggest that occurrence of these local off states was temporally linked to an impairment in task performance. In another study, authors found local changes in neuronal activity indicative of local sleep in monkeys that were performing a visual task as they were falling asleep (Pigarev, Nothdurft, & Kastner, 1997). During performance of the task, single cell activity was recorded in visual area V4 and results showed that, like the Vyazovskiy et al. (2011) study, some neurons sometimes suddenly became less responsive to visual stimulation. Besides evidence for local sleep in animals, research has also shown that local ‘off’ periods of decreased activity can even occur in isolated cell cultures, without any stimulation (Hinard et al., 2012).

5 In addition to evidence from animal studies there is also increasing evidence from human participant studies in support of sleep being a more locally regulated process rather than ‘whole-brain’.

Electroencephalogram (EEG) recording has provided a non-invasive technique to study brain activity during sleep in humans. EEG measures activity of large groups of (pyramidal) neurons in the upper layers of the cortex by placing electrodes on the scalp. This combined activity is reflected as oscillations consisting of components of different wavelengths. Because brain activity has been extensively studied during sleep in humans using this technique, the activity patterns characteristic of sleep are well established. Different output characteristics of EEG recordings can be linked to different sleep states and intensity (Zielinski et al., 2016). Characteristic of sleep is an increase in oscillations of the lower frequency bands, specifically in the delta (1-4 Hz) and theta (4-9 Hz) range, while oscillations of higher frequencies are typical of wakefulness (Campbell, 2009; Zielinski et al., 2016). It is assumed that local ‘sleep-like’ activity in humans can be measured as a local increase in power of activity in these two frequency bands.

In an early study in humans supporting a more localized mechanism behind sleep, for which participants performed a task just before going to sleep, the researchers found a local increase in slow wave activity in areas involved in performing the task during 2 hours of subsequent sleep (Huber, Ghilardi, Massimini, & Tononi, 2004). Later, recordings of local sleep-like activity were made in sleep-deprived, but awake participants. In two studies, during which participants had to repeatedly perform different tasks, local increases in power of theta waves were found in the brain regions involved in these tasks (Bernardi et al., 2015; Hung et al., 2013). Specifically, Hung and colleagues let participants stay awake for 24 up to 36 hours while they were repeatedly performing a driving simulator or audiobook listening task. Their results show an increase in power of theta waves in posterior parietal and left frontal regions, respectively. Furthermore, they found an increase in slow wave activity in the same areas during the sleep following sleep deprivation (and task performance). Using a similar study design, Bernardi et al. (2015) later also showed a temporal correlation between the occurrence of local increases in the power of theta waves in task-related areas and deteriorating task performance. Lastly, as in the study conducted with rats by Vyazovskiy et al. (2011), Nir and colleagues (2017) recorded individual neuronal and LFP activity in the medial temporal lobe (MTL) in neurosurgical patients while they were performing a face/non-face categorization psychomotor vigilance task. Most of their subjects were sleep-deprived or awake for >12 hours when testing occurred. Their results show that temporary weakened and delayed neuronal spike discharges in the MTL were associated with ‘slow’ performance on the task. Furthermore, they also show a relative local increase (weaker decrease) in slow/theta activity (2-10 Hz) associated with ‘slow’ responses. What is of specific interest in these studies is the effect this temporary local sleep or attenuated neuronal activity seems to have on performance, both in humans and in rats (Bernardi et al., 2015: Nir et al., 2017; Vyazovskiy et al., 2011). These findings could be relevant for instances where it is known that when humans perform a certain demanding cognitive task, such as driving, for a prolonged period, performance decreases. However, as mentioned human participant studies showing these results to date are limited to only sleep-deprived participants and behavioral tasks involving large brain areas. Thus, evidence on the specifically localized occurrence of local ‘off’ states or sleep-like periods of activity in non-sleep-deprived humans is lacking. Such evidence would especially be interesting with respect to the observed cognitive effect of this phenomenon, possibly extending it to numerous other cognitive tasks.

Therefore, in the current experiment we aimed to find sleep-like activity in a more localized brain region than those targeted in previous studies in healthy, wake and non-sleep-deprived human

6 participants. To do this, we let participants perform a texture discrimination task (TDT) repeatedly over four within-day sessions. The TDT used in the present experiment, developed by Karni & Sagi (1991), involved detection of a centrally located fixation letter and a peripheral target stimulus of three diagonal bars against a background of horizontal bars. In previous studies this TDT has reliably shown a decrease in performance over four within-day sessions (Censor, Karni, & Sagi, 2006; Mednick et al., 2002; Pinchuk-Yacobi, Harris, & Sagi, 2016). It has also been shown that daytime napping and changing the stimulus location can relieve this performance deterioration (Censor & Sagi, 2009; Mednick, Drummond, Arman, & Boynton, _{2008; Mednick et al., 2002). Moreover,} performance on the TDT improves after a night’s sleep, with the extent of improvement depending on the amount of sleep received (Stickgold, Whidbee, Schirmer, Patel, & Hobson, 2000). Based on these findings, it has been suggested that the decrease in performance observed for multiple within-day TDT sessions could be due to saturation of connections between neurons in early visual

processing areas, located in the occipital region (Censor & Sagi, 2009). Pertinently, it has been shown that the decrement in performance on the TDT is related to a decrease in fMRI response in the primary visual cortex (Mednick et al., 2008).

As discussed earlier, it has been suggested that such localized saturation of connections could lead to local sleep-like activity patterns (Krueger et al., 2016; Krueger et al., 2013; Krueger & Roy, 2016). In this study we aim to test whether repeated performance of this demanding TDT could indeed lead to a local saturation of neurons in early visual areas as reflected in local sleep-like EEG activity patterns in the occipital region. A pilot study using a small number of participants has shown promising first results of a local increase in theta activity following TDT performance. In the current study we aim to replicate and strengthen these results.

Methods

Testing schedule and materials

The experimental sessions were spread out over two days (see figure 1). The first day consisted of only a baseline EEG recording in the evening. On the second day, participants performed the TDT four times in separate sessions with breaks in between. During the first and fourth TDT session EEG activity was also recorded. The second experimental day ended with another baseline EEG recording directly following the last TDT session around the same time in the evening as the baseline EEG recording on the previous day.

Testing took place in an isolated demi-lit room in the testing facility of the Psychology Department at the University of Amsterdam. Participants were seated in an immobile chair facing a computer screen placed on a table on which the stimuli were presented. The screen had a resolution of 1920 x 1080 pixels and a refresh rate of 60 Hz.A keyboard to record responses during

performance of the task, a chinrest to ensure participants’ heads would stay in the same position during the whole testing session and an eye-tracking camera (SR Research Eye Link) were also placed on the table. The distance from the participants head to the screen (when resting in the chinrest) was set at 60 cm. The eye-tracking system (sampling rate: 500 Hz) was used to ensure participants would fixate on the central point of the screen during performance of the TDT. The computer screen, EEG amplifier box and eye tracker camera were connected to computers in a control room directly adjacent to the testing room, from where the experimenters monitored the testing sessions. Experimenters could check on the participants’ well-being through a camera and an intercom.

7 Figure 1. Overview of the testing protocol spread out over two days. Around 18:45 on the first day baseline resting state EEG activity was recorded for 30 minutes. On the second day participants performed four 45-60 minute TDT sessions, starting at round 10:00, 12:15, 15:00 and 18:00, respectively. EEG activity was recorded during the first and fourth TDT session. Immediately after the fourth TDT session resting state EEG activity was again recorded.

Participants

33 participants (13 males, age 18 to 29, mean age ± SD: 21.97 ± 2.97), all university students, were recruited to take part in the study. All participants had normal vision (or corrected-to-normal by the use of contact lenses) and no history of psychological, psychiatric or neurological illness. Participants were asked to refrain from drinking coffee or tea on the days of the experiment, as well as to get enough sleep on the first day and not to take naps or engage in hard exercise in between sessions on the second day. The current study was approved by the University of Amsterdam Ethics Committee and all participants gave informed consent before starting the experiment. Participants were rewarded €80 or an equivalent amount of study credits for their time.

EEG recordings

For the EEG recordings during performance of the task and resting state recordings we used BioSemi EEG system (64 electrodes) at a sampling rate of 512 Hz. External electrodes were used for

background noise (left and right earlobes), blinks (under left eye and above left eyebrow) and eye movements (on the right/left side of both eyes). Electrode impedance was checked before each EEG recording and kept below 40 KΩ.

During resting state EEG recordings participants sat in the experimental room with their eyes closed and covered with cotton pads for 30 minutes, while EEG activity was recorded. At the halfway point of the session, the EEG recording was paused for approximately a minute during which the researchers talked to the participant to make sure he/she was still awake, any technical problems could be fixed and during which the participant was allowed to move. Participants were explicitly asked not to move during the rest of the baseline recording period.

Texture Discrimination Task (TDT)

On the second day of the experiment all 33 participants performed four testing sessions of the TDT, starting around 10:00, 12:15, 15:00, and 18:00, respectively. The TDT involved computer-generated textures as used in previous studies (Karni & Sagi, 1991). The target consisted of either a vertical or horizontal array of diagonally oriented bars, against a background of horizontally oriented bars (see

figure 2). To ensure fixation, participants were at the same time presented with a letter (either a ‘T’

or an ‘L’) at varying rotations in the center of the screen. At the start of each trial participants were presented with a white fixation cross in the middle of the screen. A trial started by pressing the space bar on the keyboard, after which the following sequence of screens would appear: blank screen (250 ms), stimulus screen (33 ms), blank screen interval (varying period of time), mask (100 ms), blank response screen (250 ms) (see figure 3). The time between stimulus presentation and the mask (thus the ‘blank screen interval’ in the sequence above), further referred to as ‘stimulus onset asynchrony’

8 Figure 2. Examples of the TDT stimulus screen, showing the target stimulus of a vertical or horizontal array of three diagonally oriented bars against a background of horizontal bars, and the fixation letter (‘T’ or ‘L’) in the center of the screen in different orientations. Target stimulus location was varied randomly across trials, but always within the same quadrant of the screen over all four sessions: either bottom left (left panel) or top right (right panel).

(SOA), was decreased across blocks from 470 ms to 30 ms (470 ms, 370 ms, 270 ms, 230 ms, 220 ms, 200 ms, 180 ms, 170 ms, 130 ms, 120 ms, 100 ms, 80 ms, 70 ms, 50 ms, 30 ms, with SOA’s between 200 ms and 30 ms repeated twice for two consecutive blocks). For each trial, participants were asked to firstly report the fixation letter (T or L) and then to report whether the target was horizontal or vertical when presented with the response screen.

At the start of the first TDT session participants were randomly allocated to experimental group right or left, meaning the target stimulus would be presented in either the top right or bottom left

quadrant of the screen (see figure 2). Per participant, the target was always presented in the same quadrant of the screen throughout all testing sessions in randomly allocated locations, thus targeting the left or right visual hemifield (see figure 4).

Before the first TDT session participants had to complete a training session consisting of a couple of trials under supervision of the researcher to assure they understood the task. Furthermore, before each TDT session participants were encouraged to stay focused and motivated until the end of the session. One session typically lasted for approximately 45-60 minutes and participants were notified at the halfway point. If activity was recorded during the TDT session (sessions 1 and 4), participants were explicitly asked not to move while performing the task.

Visual Analogical Scales (VAS) and Psychomotor Vigilance Task (PVT)

As a subjective measure of their own level of fatigue, participants were asked to fill out VAS both for mental fatigue (VASf) and sleepiness (VASs) before and after each TDT and resting state EEG session (Lee, Hicks, & Nino-Murcia, 1991). The difference between the two different scales was explained to participants as mental fatigue describing the fatigue one can feel after ‘prolonged engagement in a cognitively demanding task’, such as reading a scientific paper, and sleepiness as the feeling of ‘the

9 Figure 3. Sequence of screens presented during one complete trial of the TDT. At the beginning of a trial participants were presented with a fixation cross. Upon starting the trail by pressing the space bar, a blank screen appeared (250 ms), followed by the stimulus screen (33 ms), blank screen (varying time period), mask (100 ms) and a blank answer screen (250 ms).

need of closing one’s eyes and wanting to go to sleep.’ Both scales consisted of a horizontal black line of 150 mm in length along which participants could mark their level of fatigue/sleepiness. For mental fatigue a mark at 0 mm indicated that participants were feeling very fatigued and a mark at 150 mm indicated they were not mentally fatigued at all. Conversely, for sleepiness a mark at 0 mm indicated they were not feeling very sleepy and a mark at 150 mm indicated they were feeling very sleepy.

Beside the VAS’, participants also performed a PVT directly before each TDT session as an objective measure of vigilance (Dorrian, Rogers, & Dinges, 2005). The PVT involved appearance of a counter in the middle of the screen to which participants had to respond. Specifically, participants were instructed to ‘stop’ the counter as fast as possible when it appeared by pressing the space bar, upon which the counter stopped for a short amount of time and then disappeared. This happened a number of times for a duration of 5 minutes. The counted appeared at varying time intervals, so participants could not anticipate when it would appear. During the PVT the refresh rate of the screen was set to 100 Hz.

Figure 4. Scalp map showing electrode locations and the areas of interest for the different target stimulus locations of the TDT. Circled in blue is the cluster of electrodes of the area of interest in the left hemisphere. Circled in red is the cluster of electrodes of the area of interest in the right hemisphere.

10

Data analysis

TDT data

Consistent with previous studies using the same TDT, we assessed performance by determining the SOA at which participants reached 80% target detection accuracy (Censor et al., 2006; Mednick et al., 2008; Mednick et al., 2002; Stickgold et al., 2000). In order to do this, for each session of each

individual participant the accuracy was calculated for each SOA. These values were then plotted against the SOA’s and the following logistic model was fitted to the data:

𝑦𝐸𝑥𝑝𝑜 = 0.5 + 0.5 1 + 𝑒−(𝑏1+𝑏2𝑥1)

The plots (for each participant and session separately) were visually inspected for fit of the data points to the model and whether the shape of the model followed a logistic curve.

The function of this logistic model was then rewritten to calculate the SOA at 80% accuracy as follows:

𝑆𝑂𝐴 = −ln (( 0.5

0.8 − 0.5− 1) + 𝑏1) 𝑏2

To assess whether overall performance on the TDT decreased over sessions, as reflected in a higher SOA to reach 80% accuracy in target detection, we performed a mixed linear effects analysis of the relationship between SOA at 80% accuracy and session using R (R Core Team, 2017) and lme4 (Bates, Maechler, Bolker, & Walker, 2015). In this analysis, session was included as fixed effect and subjects as random effect. Residual plots were visually inspected and deviations from normality and

homoscedasticity were not found. Likelihood ratio tests were performed to yield p-values by comparing the full model with the effect of session to a simpler model in which this effect was not included.

EEG data

The EEG data was preprocessed using the Matlab toolbox EEGLAB (Delorme & Makeig, 2004). After loading channel locations and removing unused channels, the signals for the remaining channels were re-referenced to the external electrodes on the earlobes. A high-pass filter (Basic FIR, 0-0.5 Hz) was applied to remove linear trends from the data. Independent components analysis (ICA) was carried out, after which the EEGLAB plugin ADJUST was used to remove artifacts. ADJUST is an automated algorithm optimized to capture components that include blinks, eye movements and generic discontinuities (Mognon, Jovicich, Bruzzone, & Buiatti, _{2011). These components were then} manually removed from the data.

To get a general idea of the effect of the TDT performance on EEG activity, we were interested in the differences in power of the oscillations in the delta and theta frequency band between the two resting state sessions (day before, and after performance of the TDT). For both resting state EEG sessions, power spectral density estimates of the delta and theta frequency bands were computed for all 64 electrodes by applying the pwelch function from the Matlab signal

processing toolbox. Average differences in power were calculated for each electrode separately and paired t-tests were carried out in R (R Core Team, 2017) to compare the power values for each electrode between the two resting state sessions.

11 To investigate a possible localized effect of TDT performance on EEG activity patterns more

specifically, we next compared power of the delta and theta frequency within two areas of interest between the two experimental groups. Because during the TDT the stimulus was presented in either the right upper or left lower quadrant of the screen, the two areas of interest were located in the left and right hemisphere, in the occipital region (where early visual processing areas are located) (see

figure 4). For each area of interest the mean power of the three electrodes in the cluster was

calculated per resting state session. Two-way mixed analysis of variance (ANOVA), including resting state session as a within groups and experimental group as a between groups factor, was used in R (R Core Team, 2017) to compare the power between the experimental groups for each area of interest separately.

Data elimination

Participants were eliminated from analysis based on performance on the TDT and quality of the EEG data. Requirement for performance on the TDT, was that each participant reached the threshold of 80% correct at least once during each session. Failure to reach 80% accuracy would imply that the logistic regression model used to fit the data and to calculate the SOA at 80% correct would not provide a reliable outcome to assess performance (or be impossible to calculate). Moreover, previous studies consistently report using 80% target detection accuracy as a measure of performance, implying that failure to reach this 80% correct rate indicates poor overall

performance/motivation of the subject (Censor et al., 2006; Mednick et al., 2008; Mednick et al., 2002). Conversely, for the current project it was essential that subjects were motivated/focused to perform the TDT task during all sessions. Therefore, participants that did not reach 80% target detection accuracy at all during one or more sessions were excluded from all further analysis of the TDT and EEG data.

Upon inspection of the individual EEG data we excluded one resting state recording each for two different participants (for one the baseline resting state recording, for the other the second resting state recording). For both these sessions there was large amount of noise in the data that, because of time constraints, could not be eliminated. Therefore these two participants were also excluded from all further analysis of the TDT and EEG data. Eventually, these criteria led to the exclusion of 8 participants, resulting in a final dataset of 26 participants (9 males, age 18 to 28, mean age ± SD: 21.88 ± 2.72) that was used for analysis.

Results

Texture discrimination task (TDT)

For the TDT we were interested in performance over the four sessions on the second day of the experiment. As explained above, performance was measured as the SOA at which participants still reached 80% accuracy in target stimulus detection. In a linear mixed effects analysis, we found that session significantly affected SOA at 80% accuracy (χ2_{(3) = 38.001, p = 2.825e-08), increasing it for}

TDT session 2, session 3, and session 4 by 31.7 ms ± 8.4 (standard errors), 43.2 ms ± 8.4 and 55.6 ms ± 8.3, respectively, compared to the value for session 1 (see figure 5).

Upon visual inspection of the data one participant showed outlying values for the SOA at 80% accuracy for all four TDT sessions. To investigate whether there was a significant effect of these outlying values on the results of the analysis, the above described analysis was repeated excluding this participant. This exclusion did not lead to any major changes in the overall effect of session on

12 the SOA at 80% accuracy, nor on the effect sizes of the different sessions. Because there was no effect of the outlying performance of this individual participant on the TDT results, it was chosen to keep this participant in the analysis.

Resting state EEG

Global differences after performance of the TDT

As a first analysis, we wanted to know whether across the whole scalp EEG recordings would show sleep-like activity after repeated performance of the TDT. In order to do this, we focused on power differences in low frequency oscillations, specifically in the delta (1-4 Hz) and theta (4-8 Hz) frequency band. For these two frequency bands, we compared the power of the signal between the baseline resting state EEG recording and the resting state EEG recording after performance of the TDT at each electrode individually. Figure 6 shows the average differences at each electrode for the delta and theta rhythms, plotted onto scalp maps. For the delta rhythms (see figure 6A) we found a clear increase in power during the second resting state recording (after TDT performance) compared to the baseline recording of electrodes in the occipital region, specifically centering in either the left or right hemisphere. These local increases were mostly highly statistically significant (p<0.01), as shown in

figure 6B. Furthermore, we also found a (larger) increase in power of oscillations in the delta

frequency in the medial-frontal region around the central electrode (Cz). The increase in power for electrodes in this region was again highly statistically significant (p<0.02; figure 6B). Besides local

increases in delta rhythm power, figure 6A also shows significant local decreases in power, primarily in electrodes in the left lateral-frontal region.

For the theta rhythms we also found an increase in power in the occipital region between the two resting state recordings, but centering around electrodes in the medial occipital area (see figure

6C). Again, these differences were highly statistically significant as shown in figure 6D (p<0.01).

Additionally, we found the same statistically significant increase in theta power in the medial-frontal region as observed for the delta rhythms (figures 6C and 6D). What can also be seen in figure 6C is that we did not find local decreases in theta rhythm power, as we did for the delta waves. Moreover, as shown in figure 6D, we found a statistically significant increase in power for the theta waves at many electrode locations outside the occipital and medial-frontal regions. However, the local increases in power in both the occipital and medial-frontal regions appear to be larger compared to the differences for the rest of the scalp (figure 6C).

P-values shown in figures 6B and 6D are not corrected for multiple testing.

Hemispheric differences for different target stimulus locations

Beside the effect of repeated performance of the TDT on the delta and theta waves across the whole scalp, we were specifically interested in the local effect on activity in the occipital region. Therefore,

100 120 140 160 180 200 220 1 2 3 4

SO

A

at

80

%

acc

u

racy

(ms

)

TDT session

Figure 5. Average performance on the TDT over four same-day sessions (N = 26). Performance is shown as stimulus onset asynchrony (SOA) in ms (mean ± SEM) at which participants still reached a target detection accuracy of 80%.

13 we next examined whether there was a specific effect of experimental group (target stimulus

presented in either bottom left or top right part of the screen) on the changes in EEG activity in these two frequency bands in either the right or left hemisphere, respectively. We calculated the mean power of the cluster of three electrodes in each area of interest (see figure 4) and compared these between resting state sessions and experimental groups for the clusters separately. A two-way mixed ANOVA revealed that, as expected from the above described results (see figure 6), for the delta waves there was a significant effect of session for both areas of interest (left hemisphere cluster: F(1,24) = 5.562, p = 0.027; right hemisphere cluster: F(1,24) = 9.152, p < 0.01). The same result was found for the theta power (left cluster: F(1,24) = 13.884, p < 0.01; right cluster: F(1,24) = 15.215, p < 0.001). However, for both areas of interest and both frequency ranges no statistically significant differences in power were found between different experimental groups and also no significant interaction effect between group and session. Mean power values of the electrode clusters in the

A

B

C

D

Figure 6. Scalp maps showing average differences in EEG power density (dB) between the resting state EEG recordings before and after performance of the TDT, and accompanying p-values (N = 26). A, Power differences in dB for oscillations in the delta frequency (1-4 Hz). Colors toward the red end of the spectrum indicate an increase in power during the second resting state EEG recording (after performance of the TDT). B, P-values for the power differences between resting state sessions for oscillations in the delta frequency. Blue indicates a significant difference (p~0.01). C and D, Power differences in dB for oscillations in the theta frequency (4-8 Hz) and p-values for these differences, respectively. Plotted p-values are not corrected for multiple testing.

14 occipital areas of interest for both resting state sessions are shown in figure 7 for the two

experimental groups separately.

Discussion

In the current project we aimed to measure local sleep like activity in occipital brain regions following repeated performance of a visual TDT. We have found that performance on the TDT decreased over sessions and that after four same-day sessions activity in the delta and theta frequency band was increased compared to baseline activity. These findings replicate and strengthen previous results of the same experiment in a pilot sample of 10 participants (also included in this study). Our results show that specific brain areas can go into a state of ‘sleep-like’ activity following prolonged

3 5 7 9 11 13

baseline EEG post TDT EEG

d elta p o wer (d B)

A

baseline EEG post TDT EEG

d elta p o we r (dB)

B

baseline EEG post TDT EEG

th eta p o wer (d B)

C

baseline EEG post TDT EEG

theta p o we r (dB)

D

Figure 7. Differences in average EEG power density (dB) for the left and right hemispheric occipital clusters between the resting state EEG recordings before and after performance of the TDT (mean ± SEM, N = 26). These differences are plotted separately for the two experimental groups with the TDT target stimulus presentation in either the lower left (‘left group’) or top right (‘right group’) quadrant of the screen. A and B, Power differences in dB for oscillations in the delta frequency (1-4 Hz) for the left and right hemispheric occipital cluster, respectively. C and D, Power differences in dB for oscillations in the theta frequency (4-8 Hz) for the left and right hemispheric occipital cluster, respectively. No statistically significant effect of experimental group for was found for either of the frequency bands or hemispheric clusters.

15 engagement in a task. This is the first study to date that has shown these results for a specific brain region, that is the occipital region where early visual processing areas are located, in awake, non-sleep-deprived human participants.

Decreasing TDT performance may be due to saturation of early visual areas

Participants showed an average decrease in performance on the TDT, as reflected in a higher SOA to still reach 80% target detection accuracy over four same-day sessions. With these results we

replicate a deteriorating performance effect seen in previous studies (Censor et al., 2006; Mednick et al., 2002; Pinchuk-Yacobi et al., 2016). Furthermore, the magnitude of the observed decrease (an SOA of around 50 ms higher needed to reach 80% accuracy in session 4 compared to session 1) is comparable to previously reported results (Mednick et al., 2002).

More specifically, earlier studies have shown that repeated exposure, or ‘over-exposure’, to the target stimulus in the same location of the screen leads to deteriorating performance. However, performance does not deteriorate if the target location is varied, or if participants are presented with less trials in total during a TDT session, resulting in less exposure to the stimulus (Censor et al., 2006; Censor & Sagi, 2009; Mednick et al., 2002). Based on these results, it has been suggested that the deterioration in performance, such as also seen in our experiment, is not a result of global fatigue, but is due to saturation of early visual processing areas (Censor & Sagi, 2009).

Local sleep-like activity in the occipital region

Our results show a local increase in theta band activity in the occipital region following performance of four same-day TDT sessions, compared to baseline activity around the same hour of the day. This finding is consistent with previous studies showing an association between repeated performance of a task and an increase in theta activity in associated cortical areas in both humans and rats (Bernardi et al., 2015; Hung et al., 2013; Nir et al., 2017; Vyazovskiy et al., 2011). The important difference, however, is that in previous studies subjects were sleep-deprived when these local increases in theta activity were recorded while in the current project participants were not sleep-deprived. Theta band activity (approximately 4-9 Hz) is especially prominent during rapid eye movement sleep (REMS), but also occurs during non-rapid eye movement sleep (NREMS) and the most important generator of these rhythms is the hippocampus (Zielinski et al., 2016).

Besides this local increase in theta waves, we have also found a local increase in delta band activity in the occipital region following TDT performance. Contrary to local increases in theta waves, previous studies do not report finding a local increase in power in the delta range following repeated performance of a task (though of note: Vyazovskiy and colleagues (2011) report a local increase in the 2-6 Hz range, and Nir et al. (2017) in the 2-10 Hz range, overlapping with the defined ranges for both the delta and theta bands). Delta rhythms are most prominently found during NREMS (then often referred to as ‘slow wave activity’), although they also occur during REMS (Campbell, 2009; Zielinksi et al., 2016). The highest amount of delta activity occurs at the beginning of sleep. Evidence suggests that rhythms of the delta frequency are partly produced by hyperpolarization of the membrane potential of cortical pyramidal neurons, which would be an indication of decreased activity (Zielinski et al., 2016).

Beside the local increase in delta and theta power in the occipital region, we also found (significant) increases in activity of both frequency bands in a large medial-frontal region following TDT performance. Cortical areas in this region are involved in sustained attention and our findings suggest this region also transitioned into a sleep-like state following performance of the TDT (Berry, Sarter, & Lustig, 2017; Sarter, Gehring, & Kozak, 2006). While not reported in previous studies with

16 a similar design, this finding is not surprising as participants had to perform a demanding visual task for four 45-60 minute same-day sessions.

While our findings show a general task-related increase in delta and theta activity localized to the occipital region, further analyses comparing experimental groups with different target stimulus locations failed to discriminate between relative increases in the different visual hemispheres. Had we found a local difference between the two hemispheres for the experimental groups, we would have provided evidence for local sleep-like activity that would be far more localized than evidence to date shows. However, more refined analysis methods might still shed more light on hemispheric differences.

Saturation of early visual areas leads to local sleep-like activity

Combined, our findings show that repeated same-day performance of the TDT leads to deteriorating behavioral outputs, which is associated with a local increase in ‘sleep-like’ activity in the occipital region. Building on previous studies using the same TDT, these results support the idea that the decrease in performance we see in our participants could be due to saturation of early visual

processing areas located in the occipital region (according to literature, the used experimental design would definitely be viewed as ‘over-exposure’ to the stimulus). This local saturation could have led to the observed increases in delta and theta activity. Previously reported findings of a beneficial effect of whole-brain sleep on TDT performance further support the hypothesis that the observed local sleep-like patterns may play a role in performance deterioration observed in this TDT: whole brain sleep might provide these areas with the accompanying hypothesized benefits relieving the local saturation of the neurons (Censor et al., 2006; Cirelli & Tononi, 2017; Mednick et al., 2002; Nedergaard & Goldman, 2016; Stickgold et al., 2000).

Although the current findings provide promising first evidence of local sleep-like states occurring in awake, non-sleep-deprived humans following performance of a task, the low spatial resolution of EEG does not give opportunity to measure a very specific localized effect (Burle et al., 2015). Indeed, in the present study the observed significant changes in theta activity were relatively high in the occipital region, but were found to be significant for a large number of electrodes. In the future it would be interesting to repeat the current study engaging early visual areas while using more specific measurements for neuronal activity in the early visual cortex of animal models or perhaps, following Nir et al. (2017), in human neurosurgical patients. Given the numerous developed visual tasks, such as the TDT used in the current experiment, early visual areas can be reliably

targeted, thus providing opportunity for measuring very specifically localized consequences for neuronal activity. Like Bernardi and colleagues (2015) have recently shown using EEG, and Nir et al. (2017) recording single neurons in neurosurgical patients, this could also provide an opportunity to further study the temporal relationship between decreases in performance and very localized sleep-like activity in humans.

Even in the current project, however, further analyses could still provide more robust evidence for the association between decreased performance and local sleep-like activity in the EEG. Firstly, although ICA was used to remove noisy components, the EEG data could still be more thoroughly cleaned by removing individual periods of noise in the data that were possibly not detected by ICA. Moreover, current reported differences in EEG delta and theta activity were not corrected for multiple testing, which can still give a more precise indication of the observed differences. Lastly, source localization of the EEG signal can provide a more specific indication of the localization of the delta and theta changes that were observed.

17 Beside these adjustments to the current analyses, still more work needs to be done to fully analyze all the data that has been collected in this project. The EEG recordings during TDT session 1 and 4 could provide further information on the temporal relationship between the occurrence of brief periods of increased slow wave activity and task performance, by directly linking trials with false response to the EEG signal at the time of those trials. It would also be interesting to investigate a correlation between a decrease in task performance and an increase in delta and theta power, because this will further support a causal effect of local ‘sleep-like activity’ and the observed behavioral decrement. Analyzing the eyetracker data will provide a control measure of focus of the participants on the fixation cross or letter during the experiment. This is of importance for drawing conclusions about the (peripheral) location the target stimulus was presented in and thus the expected scalp region of the change in EEG activity. The VASf, VASs and PVT can finally be used to provide insight into subjective and objective levels of fatigue, sleepiness and vigilance, that can possibly have affected TDT performance.

Local to global sleep

By showing that specific brain areas can go into a ‘sleep-like’ state following a period of activity, the current findings support the view of a localized regulation of sleep within individual brain areas. This theory proposes that multiple areas independently going into an ‘off-line’ state could lead to whole brain sleep (Krueger et al., 2016; Krueger et al., 2013). Our results add to growing evidence

supporting the related theory that following a period of activity cortical areas become ‘saturated’ and go into a ‘sleep-like’ state of decreased activity to renormalize synaptic connectivity and clear toxic waste (Krueger et al., 2016; Krueger et al., 2013; Krueger & Roy, 2016). However, how this process relates to actual whole brain sleep, and whether using the term sleep-like activity is even fitting, remains to be investigated. Importantly, whole-brain sleep proposedly serves a multitude of other functions in the body, therefore not only making it a hypothesized essential physiological process for brain functioning (Krueger et al., 2016). Moreover, some proposed brain functions of sleep, such as the globally sweeping away of toxic protein and other biological debris through the glymphatic system, would be difficult to achieve for only locally occurring ‘sleep’ (Nedergaard & Goldman, 2016).

All in all, in light of the current evidence, it seems difficult to directly connect local sleep-like activity states to global whole-brain sleep. More interesting might be the implication of local sleep for cognitive functioning.

Possible cognitive effect of ‘local sleep’

Of special interest in previous studies, is the finding in both human participants and rats that occurrence of local (relative) increases in theta activity show a temporal relation with decreases in performance (Bernardi et al., 2015; Nir et al., 2017; Vyazovskiy et al., 2011). Though recorded in sleep-deprived subjects, these findings still provide a valuable perspective on possible mechanisms behind cognitive fatigue in general. It is well known that performance on many different tasks, such as driving, decreases with increasing time spent on the activity. Local sleep-like activity could possibly play a role in these instances of decreasing performance. The current project provides a first step by showing that local increases in sleep-like activity can be associated with decreasing task performance in awake, non-sleep-deprived humans. As mentioned before, extending the current project by investigating the temporal relationship between local increases in delta and theta power and deteriorating performance could provide even more promising evidence.

18

Conclusion

This is the first study to date that shows local sleep-like activity can occur in specific brain areas following repeated performance of a task targeting those areas in awake and non-sleep-deprived humans. Thereby it supports the hypothesis that a transition into a sleep-like state can be locally regulated in the brain following a period of high activity. This finding could have implications for how we deal with cognitive fatigue and how we view sleep in general. What exactly causes these

instances of local sleep-like activity, how it relates to whole brain sleep and whether it plays a role in cognitive fatigue still remains to be investigated.

19

References

Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1), 1-48. doi:10.18637/jss.v067.i01

Bernardi, G., Siclari, F., Yu, X., Zennig, C., Bellesi, M., Ricciardi, E., . . . Tononi, G. (2015). Neural and behavioral correlates of extended training during sleep deprivation in humans: Evidence for local, task-specific effects. The Journal of Neuroscience : The Official Journal of the Society for

Neuroscience, 35(11), 4487-4500. doi:10.1523/JNEUROSCI.4567-14.2015 [doi]

Berry, A. S., Sarter, M., & Lustig, C. (2017). Distinct frontoparietal networks underlying attentional effort and cognitive control. Journal of Cognitive Neuroscience, 29(7), 1212-1225.

doi:10.1162/jocn_a_01112 [doi]

Burle, B., Spieser, L., Roger, C., Casini, L., Hasbroucq, T., & Vidal, F. (2015). Spatial and temporal resolutions of EEG: Is it really black and white? A scalp current density view. International

Journal of Psychophysiology : Official Journal of the International Organization of Psychophysiology, 97(3), 210-220. doi:10.1016/j.ijpsycho.2015.05.004 [doi]

Campbell, I. G. (2009). EEG recording and analysis for sleep research. Current Protocols in

Neuroscience, Chapter 10, Unit10.2. doi:10.1002/0471142301.ns1002s49 [doi]

Censor, N., Karni, A., & Sagi, D. (2006). A link between perceptual learning, adaptation and sleep. Vision Research, 46(23), 4071-4074. doi:S0042-6989(06)00333-6 [pii]

Censor, N., & Sagi, D. (2009). Global resistance to local perceptual adaptation in texture discrimination. Vision Research, 49(21), 2550-2556. doi:10.1016/j.visres.2009.03.018 [doi]

Cirelli, C., & Tononi, G. (2017). The sleeping brain. Cerebrum : The Dana Forum on Brain

Science, 2017 doi:cer-07-17 [pii]

Delorme, A., & Makeig, S. (2004). EEGLAB: An open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of Neuroscience Methods, 134(1), 9-21. doi:10.1016/j.jneumeth.2003.10.009 [doi]

20 Dorrian, J., Rogers, N. L., & Dinges, D. F. (2005). Psychomotor vigilance performance: Neurocognitive

assay sensitive to sleep loss. Sleep Deprivation: Clinical Issues, Pharmacology and Sleep Loss

Effects, , 39-70.

Hinard, V., Mikhail, C., Pradervand, S., Curie, T., Houtkooper, R. H., Auwerx, J., . . . Tafti, M. (2012). Key electrophysiological, molecular, and metabolic signatures of sleep and wakefulness revealed in primary cortical cultures. The Journal of Neuroscience : The Official Journal of the

Society for Neuroscience, 32(36), 12506-12517. doi:10.1523/JNEUROSCI.2306-12.2012 [doi]

Huber, R., Ghilardi, M. F., Massimini, M., & Tononi, G. (2004). Local sleep and learning. Nature, 430(6995), 78-81. doi:10.1038/nature02663 [doi]

Hung, C. S., Sarasso, S., Ferrarelli, F., Riedner, B., Ghilardi, M. F., Cirelli, C., & Tononi, G. (2013). Local experience-dependent changes in the wake EEG after prolonged wakefulness. Sleep, 36(1), 59-72. doi:10.5665/sleep.2302 [doi]

Karni, A., & Sagi, D. (1991). Where practice makes perfect in texture discrimination: Evidence for primary visual cortex plasticity. Proceedings of the National Academy of Sciences of the United

States of America, 88(11), 4966-4970.

Krueger, J. M., Frank, M. G., Wisor, J. P., & Roy, S. (2016). Sleep function: Toward elucidating an enigma. Sleep Medicine Reviews, 28, 46-54. doi:10.1016/j.smrv.2015.08.005 [doi]

Krueger, J. M., Huang, Y. H., Rector, D. M., & Buysse, D. J. (2013). Sleep: A synchrony of cell activity-driven small network states. The European Journal of Neuroscience, 38(2), 2199-2209.

doi:10.1111/ejn.12238 [doi]

Krueger, J. M., Rector, D. M., Roy, S., Van Dongen, H. P., Belenky, G., & Panksepp, J. (2008). Sleep as a fundamental property of neuronal assemblies. Nature Reviews.Neuroscience, 9(12), 910-919. doi:10.1038/nrn2521 [doi]

Krueger, J. M., & Roy, S. (2016). Sleep's kernel: Surprisingly small sections of brain, and even neuronal and glial networks in a dish, display many electrical indicators of sleep. Scientist (Philadelphia,

Pa.), 30(3), 36-41.

Lee, K. A., Hicks, G., & Nino-Murcia, G. (1991). Validity and reliability of a scale to assess fatigue. Psychiatry Research, 36(3), 291-298. doi:0165-1781(91)90027-M [pii]

21 Mascetti, G. G. (2016). Unihemispheric sleep and asymmetrical sleep: Behavioral, neurophysiological,

and functional perspectives. Nature and Science of Sleep, 8, 221-238. doi:10.2147/NSS.S71970 [doi]

Mednick, S. C., Drummond, S. P., Arman, A. C., & Boynton, G. M. (2008). Perceptual deterioration is reflected in the neural response: FMRI study of nappers and non-nappers. Perception, 37(7), 1086-1097. doi:10.1068/p5998 [doi]

Mednick, S. C., Nakayama, K., Cantero, J. L., Atienza, M., Levin, A. A., Pathak, N., & Stickgold, R. (2002). The restorative effect of naps on perceptual deterioration. Nature Neuroscience, 5(7), 677-681. doi:10.1038/nn864 [doi]

Mognon, A., Jovicich, J., Bruzzone, L., & Buiatti, M. (2011). ADJUST: An automatic EEG artifact detector based on the joint use of spatial and temporal features. Psychophysiology, 48(2), 229-240. doi:10.1111/j.1469-8986.2010.01061.x [doi]

Nedergaard, M., & Goldman, S. A. (2016). Brain drain. Scientific American, 314(3), 44-49.

Nir, Y., Andrillon, T., Marmelshtein, A., Suthana, N., Cirelli, C., Tononi, G., & Fried, I. (2017). Selective neuronal lapses precede human cognitive lapses following sleep deprivation. Nature

Medicine, 23(12), 1474-1480. doi:10.1038/nm.4433 [doi]

Pigarev, I. N., Nothdurft, H. C., & Kastner, S. (1997). Evidence for asynchronous development of sleep in cortical areas. Neuroreport, 8(11), 2557-2560.

Pinchuk-Yacobi, N., Harris, H., & Sagi, D. (2016). Target-selective tilt aftereffect during texture learning. Vision Research, 124, 44-51. doi:10.1016/j.visres.2016.06.008 [doi]

Rattenborg, N. C., Voirin, B., Cruz, S. M., Tisdale, R., Dell'Omo, G., Lipp, H. P., . . . Vyssotski, A. L. (2016). Evidence that birds sleep in mid-flight. Nature Communications, 7, 12468.

doi:10.1038/ncomms12468 [doi]

Sarter, M., Gehring, W. J., & Kozak, R. (2006). More attention must be paid: The neurobiology of attentional effort. Brain Research Reviews, 51(2), 145-160. doi:S0165-0173(05)00166-9 [pii]

Siegel, J. M. (2009). Sleep viewed as a state of adaptive inactivity. Nature

22 Stickgold, R., Whidbee, D., Schirmer, B., Patel, V., & Hobson, J. A. (2000). Visual discrimination task

improvement: A multi-step process occurring during sleep. Journal of Cognitive

Neuroscience, 12(2), 246-254.

Vyazovskiy, V. V., Olcese, U., Hanlon, E. C., Nir, Y., Cirelli, C., & Tononi, G. (2011). Local sleep in awake rats. Nature, 472(7344), 443-447. doi:10.1038/nature10009 [doi]

Zielinski, M. R., McKenna, J. T., & McCarley, R. W. (2016). Functions and mechanisms of sleep. AIMS