ARC genotype modulates slow wave sleep and spectral power in the sleep EEG after total sleep deprivation

Abstract

Study Objectives

There are large individual differences in the homeostatic response to sleep deprivation, as reflected in slow wave sleep (SWS) and electroencephalogram (EEG) spectral power, which have largely been left unexplained. Recent evidence suggests the possible involvement of the activity-regulated cytoskeleton-associated protein (ARC) gene. Here we assessed the effects of the “c.*742 + 58C > T non-coding single nucleotide polymorphism” of the human ARC gene (rs35900184) on sleep-physiological and waking-neurobehavioral responses to total sleep deprivation (TSD).

Methods

N = 50 healthy, young adults participated in a 4-day/3-night in-laboratory study with a 38-h TSD period, flanked by 10-h baseline and recovery sleep opportunities. Sleep was recorded polysomnographically and the EEG of non-rapid eye movement (NREM) and rapid eye movement (REM) sleep was subjected to spectral analysis. Waking neurobehavioral functioning was measured with the psychomotor vigilance test (PVT) and the Karolinska Sleepiness Scale (KSS).

Results

ARC C/C homozygotes, compared to T allele carriers, showed a greater SWS rebound during recovery sleep after TSD relative to baseline. ARC T/T homozygotes showed increased EEG spectral power in the NREM theta and alpha bands and in the REM delta, theta, alpha, and beta bands, but there was no significant genotype difference in the NREM delta power response to TSD. There were also no significant genotype differences in the impact of TSD on PVT performance and KSS sleepiness.

Conclusions

Individual differences in the sleep physiological rebound after TSD were influenced by ARC genotype. However, our findings were only partially consistent with ARC mediating the sleep homeostatic response to sleep deprivation.

This article is part of the Genetic and Other Molecular Underpinnings of Sleep, Sleep Disorders, and Circadian Rhythms Including Translational Approaches Collection.

Statement of Significance

This study showed that individual differences in the human sleep physiological response to total sleep deprivation are partially predicted by cytoskeleton-associated protein (ARC) genotype, and that ARC may play a role in the homeostatic regulation of sleep. However, ARC also affected the non-rapid eye movement and rapid eye movement sleep electroencephalogram in ways that may not pertain to homeostatic sleep regulation through mechanisms that have yet to be identified.

Introduction

Sleep and waking neurobehavioral functioning are regulated by a homeostatic process interacting with a circadian process [1,2]. The homeostatic process entails accumulation of sleep pressure with time spent awake, which is subsequently dissipated with time spent asleep. The time course of the homeostatic process can be observed in the electroencephalogram (EEG) of non-rapid eye movement (NREM) sleep stages N2 and N3 [3]. Specifically, EEG delta power (energy in the 0.8–4.0 Hz frequency band), also known as slow wave activity (SWA), reflects the intensity of sleep [4] and the rate of homeostatic pressure dissipation [5].

Acute total sleep deprivation (TSD) leads to a homeostatic rebound during subsequent recovery sleep, with consolidation of sleep, increases in the expression of sleep stage N3 (also known as slow wave sleep, SWS), and enhancement of delta power in the NREM sleep EEG [6,7]. The manifestation of SWS and SWA, which is most pronounced during the first few hours of the sleep period [6,8], is also influenced by the circadian process [9], which generates variation in the propensity for rapid eye movement (REM) sleep over time of day [10]. NREM and REM sleep alternate across sleep cycles of  ~ 90–110 min duration through a reciprocal interaction mechanism [11,12]. It is believed that SWS and NREM delta power are correlates of the recuperative potential of sleep [13], although that interpretation appears to break down under conditions other than acute TSD [14], including chronic sleep restriction [15]. Regardless, SWS and SWA are considered reliable indices of sleep homeostasis [16].

Remarkably, there are large, trait-like individual differences in the expression of SWS and the magnitude of SWA, which persist from baseline sleep to recovery sleep after TSD [17,18]. These trait individual differences pose a significant challenge for understanding the recuperative potential and function of SWS [18,19], but offer opportunities for investigating the mechanisms underlying sleep homeostatic regulation, which have hitherto not been fully identified [20,21]. Recent evidence from rodent studies [22] suggests that the activity-regulated cytoskeleton-associated protein (Arc) gene may be involved. Arc is a neural immediate early gene, the expression of which is upregulated during extended wakefulness and downregulated during sleep, potentially tracking the sleep homeostatic process. Arc knockout mice, in comparison with wild-type mice, display behavioral and molecular phenotypes that implicate Arc in the homeostatic regulation of sleep. For example, following 4 h of TSD, Arc knockout mice display blunted NREM sleep and reduced spectral power in the lower frequencies (delta band) of the NREM sleep EEG [22].

In humans, there is a variant of the ARC gene, the NM_015193.4:c.*742 + 58C > T non-coding single nucleotide polymorphism (SNP), which could potentially explain some of the inter-individual variability in SWS and SWA. The variant allele (T) of this SNP appears in approximately 28% of the global population, either as heterozygotes or homozygotes, whereas the ancestral allele occurs more frequently, in ~ 72% of the population (1000 Genomes Project). The human ARC gene is involved in dopaminergic and glutamatergic signaling [23], and ARC’s impact on these or other systems may have implications for sleep regulation that may be exposed by investigating the effects of the ARC SNP on sleep physiological and waking neurobehavioral markers of sleep homeostasis. While the molecular and functional consequences of the ARC SNP in question have yet to be characterized, we hypothesized that this SNP is associated with the individual differences in SWS and SWA homeostatic responses to TSD, in a similar manner to what has been shown in the rodent literature with the Arc knockout model described above [22]. Given the role of sleep homeostasis in waking neurobehavioral functioning [2], we also hypothesized that any differences in sleep homeostatic regulation attributable to the ARC SNP may have consequences for waking neurobehavioral functioning in response to TSD. Here we tested these hypotheses in a laboratory study of TSD in healthy young adults.

Methods

Subjects

N = 50 healthy, young adults (27.3 ± 4.9 years old; 54% women) participated in one of two in-laboratory studies conducted in the Sleep and Performance Research Center at Washington State University Health Sciences Spokane. Study eligibility was determined by means of a telephone interview, two in-laboratory screening sessions which included a physical examination and multiple questionnaires, blood and urine testing, and evaluation of baseline polysomnography. Subjects met the following inclusion criteria: age 22–40 years; physically and psychologically healthy; no clinically relevant abnormalities in blood and urine; no current medical or drug treatments (except contraceptives); free from traces of drugs and alcohol; not a current smoker; no history of moderate to severe brain injury; no history of a learning disability; no previous adverse reaction to acute sleep deprivation; no vision or hearing impairment unless corrected to normal; not pregnant; no sleep or circadian disorders; no travel across time zones within 1 month prior to study enrollment; no shift work within 3 months prior to study enrollment; and proficient English speaker. In addition, subjects reported a regular wake time between 06:00 and 09:00 and a habitual sleep duration between 6 and 10 h per night.

The studies were approved by Washington State University’s Institutional Review Board. All subjects gave written informed consent and were financially compensated for their time.

Experimental design and procedures

The two studies from which the data are drawn were equivalent in design. They were first described by Satterfield et al. (2015) [24] and Chavali et al. [25].

For the week prior to entering the laboratory, subjects were instructed to maintain their habitual sleep and wake times at home, to avoid napping, and to refrain from caffeine, alcohol, tobacco, and drug use. Sleep and wake times were verified by means of wrist actigraphy (Actiwatch-2, Philips Respironics, Bend, OR, USA), sleep diary, and called-in bedtimes and waketimes. A urine drug test and alcohol breathalyzer test were performed before admission into the laboratory.

Each of the two studies included a 4-day/3-night in-laboratory experiment with a 38-h period of TSD. The TSD period was preceded by a baseline day, including a 10-h time in bed (TIB) baseline sleep opportunity from 22:00 to 08:00; and followed by a recovery day, including a 10-h TIB recovery sleep opportunity from 22:00 to 08:00. Both baseline and recovery sleep periods were recorded with polysomnography (PSG). Waking neurobehavioral functioning was measured approximately every 2 h during scheduled wakefulness. Subjects received training on neurobehavioral assessments on day 1 of the laboratory study (see Figure 1).

Figure 1.

Schematic of the design of the two studies from which the data were drawn. Black bars denote sleep periods; light bars denote wake periods. Diamonds (baseline) and squares (TSD) show the timing of neurobehavioral test bouts used for data analyses, at well-rested baseline and 24 h later during TSD, respectively. “T” indicates task training on day 1.

Subjects stayed inside the laboratory through the duration of the experiment and did not have any contact with individuals outside of the laboratory. They did not have access to live radio or television, phones, personal computers, video games, or the internet. They spent scheduled sleep periods and performed neurobehavioral testing in their own individual room within the laboratory, and they were behaviorally monitored by trained research assistants continuously. The laboratory conditions were strictly controlled, with ambient temperature maintained at 21 ± 1°C. Light levels were fixed below 100 lux during scheduled wakefulness and < 1 lux (lights off) during scheduled sleep periods.

Nighttime polysomnography and spectral analysis

The baseline and recovery sleep periods were recorded polysomnographically using a digital system (Nihon Kohden, Foothill Ranch, CA, USA). The 10-20 system [26] was used to place EEG electrodes at frontal (F3, F4), central (C3, C4), and occipital (O1, O2) locations. Electrooculography electrodes were placed on the outer canthus of the left and right eye, and EMG electrodes were placed on either side of the mandible midline. Reference electrodes were placed at the mastoids (M1, M2). Electrode impedances were checked and a biocalibration was performed at the start of each PSG recording.

The PSG records were visually scored in 30-s epochs by a registered polysomnographic technologist using standardized criteria [27]. Sleep architecture was described using the following sleep variables: time spent in sleep stages N1, N2, and N3 (i.e. SWS) and REM sleep, total sleep time (TST), sleep latency (SL), wake after sleep onset (WASO), and sleep efficiency (SE).

In addition, spectral power in the EEG of NREM sleep (stages N2 and N3) and REM sleep was assessed (C3-M3 derivation). Artifact detection (including cross-talk from eye movements during REM sleep) was performed visually in 5-s subepochs. After artifact removal, spectral analysis was performed with fast Fourier transformation for each 5-s subepoch [17]. Spectral power in 0.2 Hz frequency bins was averaged over the artifact-free subepochs in each 30-s epoch of NREM (stage N2 or N3) sleep and in each 30-s epoch of REM sleep. Subsequently, spectral power was calculated by NREM or REM epoch for the frequency bins in each of four frequency bands: delta (0.8–4.0 Hz), theta (4.2–8.0 Hz), alpha (8.2–12.0 Hz), and beta (12.2–16.0 Hz). For each frequency bin, power was subsequently summed across all epochs of NREM sleep and across all epochs of REM sleep.

Neurobehavioral functioning

At approximately 2-h intervals during scheduled wakefulness (see Figure 1), a test battery consisting of the psychomotor vigilance test (PVT) [28,29] and a computerized Karolinska Sleepiness Scale (KSS) [30], among other tests, was administered on a desktop computer in each subject’s individual bedroom. The PVT is a 10-min serial reaction time task considered to be a gold standard measure of behavioral alertness [28]. Subjects were asked to respond to a visual stimulus, in the form of a millisecond counter, by pressing a button on a response box as quickly as possible without making false starts. The stimulus was presented at random intervals between 2 and 10 s. Log-transformed signal-to-noise ratio (LSNR), a measure of the fidelity of information processing, was used to quantify PVT performance [25]. The KSS is a nine-point Likert scale on which subjects rated their current level of sleepiness ranging from 1 (“extremely alert”) to 9 (“extremely sleepy/fighting sleep”) [30].

Genotyping

A venous whole blood sample was collected from each subject prior to the experiment. Blood was collected in a vacutainer tube coated with K₂-EDTA (dipotassium-ethylenediaminetetraacetic acid), aliquoted, and immediately stored at −80°C until time of analysis. Samples were red-cell-depleted and genomic DNA (gDNA) was extracted. The gDNA samples were assayed for the ARC NM_015193.4:c.*742 + 58C > T non-coding SNP (rs35900184, chromosome 8) using the Functionally Tested TaqMan Assay (Assay ID: C____256014_20; ThermoFisher Scientific, Waltham, MA, USA). Real-time polymerase chain reaction (PCR) procedures were performed according to the manufacturer’s protocol using VIC/FAM context sequence AGGGTGTCTGGGGTCTGGGGAGGAG[C/T]TCATTTGGCCGGGACAGCTGATGGT. Reactions contained 20 ng wet gDNA in a final volume of 25 μL. Samples were assayed in duplicate. All plates included a no-DNA negative control. Allelic discrimination analysis was performed using MJ Opticon Monitor Analysis Software v3.1 (Bio-Rad Laboratories, Hercules, CA, USA).

Data analyses

The genotypes in our sample were tested for deviation from Hardy–Weinberg equilibrium using a χ² test [31]. Logistic regression was used to test for differences in sex and race/ethnicity distributions by ARC genotype. One-way analysis of variance (ANOVA) was used to test for differences in age as a function of ARC genotype.

Habitual sleep duration and intermittent wakefulness, as well as habitual bedtimes and waketimes, were estimated from actigraphically measured sleep (Actiwatch 2; Phillips Respironics, Murrysville, PA) in the 7 days prior to the laboratory experiment. Actigraphic records were scored using Actiware software (Version 6.0.7; Phillips Respironics, Murrysville, PA), guided by sleep diary [32]. Data were averaged across the 7 days for each subject, then compared between ARC genotypes (C/C, C/T, and T/T) using one-way ANOVA.

Sleep variables were analyzed using mixed-effects ANOVA with fixed effects for ARC genotype (C/C, C/T, and T/T), night (baseline, recovery), and their interaction, with study as a covariate. As age affects SWS [33], a sleep variable of primary interest, age was added as a covariate as well. A random effect over subjects was placed on the intercept. ARC genotype by night interaction was the primary effect of interest and used as a measure of sleep homeostatic response to TSD.

For each spectral band (delta, theta, alpha, and beta) of the NREM and REM sleep EEG, log-transformed power was analyzed using mixed-effects ANOVA with fixed effects for ARC genotype, night, frequency bin, and their two- and three-way interactions. Study and age were included as covariates, and a random effect over subjects was placed on the intercept. Reported results are for total band power and the effects thereon of ARC genotype, night, and their interaction, estimated as part of the mixed-effects ANOVA.

PVT LSNR data and KSS scores were averaged over the first 12 h of the TSD period after a 1-h period immediately after awakening (i.e. from 09:00 until 21:00) to determine neurobehavioral functioning at baseline controlling for any sleep inertia [34], and over the last 12 h of the TSD period prior to a 1-h period immediately before bedtime (i.e. from 09:00 until 21:00 on the second day of TSD) to determine waking neurobehavioral functioning during TSD at the same times of day as baseline. The PVT and KSS data were separately analyzed using a mixed-effects ANOVA with fixed effects for ARC genotype, waking period (baseline, TSD), and their interaction. Study and age were included as covariates, and a random effect over subjects was placed on the intercept. ARC genotype by night interaction was the primary effect of interest and was used as a measure of the sleep homeostatic response to TSD.

Planned contrasts were included for pairwise comparisons between ARC genotypes for each variable of interest. Analyses were repeated with sex and race added as additional covariates. There were no substantive changes in any of the results; therefore, these latter analyses are not reported.

Data analyses on sleep architecture, NREM EEG spectral power, PVT, and KSS variables were performed on the full subject sample (N = 50). Actigraphy data for one subject were missing due to equipment failure; therefore, this subject was not included in the analyses of at-home sleep but was included in all other analyses reported. In addition, data from seven subjects were excluded from the REM EEG spectral power analyses due to poor EEG quality and/or insufficient number of remaining epochs for accurate assessment. Therefore, statistical analyses for the REM EEG spectrum were based on 43 of the 50 subjects.

Results

Study sample and genotypes

Subject characteristics are summarized in Table 1. The ARC genotype distribution was 33 C/C homozygotes, 11 C/T heterozygotes, and 6 T/T homozygotes. The ancestral allele (C) was present at a frequency of 0.77 and the variant allele (T) was present at a frequency of 0.23 in the study population, similar to that reported for healthy subjects in the published literature [35] and that of the global population. The genotype distribution of our sample deviated from Hardy–Weinberg equilibrium (χ²₁ = 7.18, p = .007). The ARC genotypes did not vary significantly by age (F_2,47 = 0.51, p = .61), sex (χ²₂ = 1.19, p = .55), or race/ethnicity (χ²₂ = 0.005, p > .99).

Table 1.

Subject characteristics

	ARC genotype			All
	C/C	C/T	T/T
Subjects	33	11	6	50
Age (mean ± SD)	27.8 ± 4.9	26.1 ± 4.7	27.0 ± 5.4	27.3 ± 4.9
Sex
Female	16	7	4	27
Male	17	4	2	23
Race/Ethnicity
American Indian/Alaska Native	0	0	0	0
Asian	0	0	0	0
Black	0	2	0	2
Native Hawaiian/Pacific Islander	0	0	0	0
White	30	9	6	45
Mixed	2	0	0	2
Undisclosed	1	0	0	1
Habitual sleep (mean ± SD)
Sleep duration (h)	7.0 ± 1.2	7.4 ± 1.0	7.1 ± 0.8	7.1 ± 1.1
Intermittent wakefulness (min)	57.0 ± 28.5	57.1 ± 21.1	43.5 ± 19.7	55.4 ± 26.4
Bedtime (HH:MM ± min)	23:15 ± 59.7	23:18 ± 75.6	23:50 ± 63.1	23:20 ± 68.1
Waketime (HH:MM ± min)	07:09 ± 71.9	07:41 ± 80.9	07:38 ± 70.2	07:19 ± 74.9

Subjects’ habitual sleep duration, estimated as the average actigraphically measured sleep duration in the 7 days before the in-laboratory experiment, was 7.1 ± 1.1 h (mean ± SD). Intermittent wakefulness was 0.9 ± 0.4 h. Habitual bedtime was 23:20 ± 1.1 h and habitual waketime was 07:19 ± 1.2 h. Table 1 shows that these data were stratified by ARC genotype. There were no significant differences between genotypes for sleep duration (F_2,44 = 0.63, p = .54), intermittent wakefulness (F_2,44 = 2.06, p = .14), bedtime (F_2,44 = 0.54, p = .59), or waketime (F_2,44 = 1.02, p = .37).

Baseline and recovery sleep architecture

Figure 2 shows time spent in each sleep stage (N1, N2, SWS, and REM), as well as TST, SL, WASO, and SE, stratified by ARC genotype, for the baseline and recovery nights. As expected, there were significant main effects of night on all these sleep variables, with recovery sleep after TSD showing increased N1, SWS, REM, TST, and SE, and decreased N2, SL, and WASO. There were no significant main effects of ARC genotype, and no significant differences between ARC genotypes at baseline, for any of the sleep variables. See Table 2 for the statistical test results. The distribution of sleep stages across each sleep period for each ARC genotype is shown in the supplementary material (Supplementary Figures S1 and S2).

Figure 2.

Effects of ARC genotype on sleep architecture. Polysomnography-measured sleep variables are shown (means ± SEM) for baseline and recovery sleep, stratified by ARC genotype. **p < .01 for ARC genotype by night interaction.

Table 2.

Main statistical results for sleep architecture, spectral power in the NREM and REM sleep EEG, and waking neurobehavioral outcomes. Bolded table entries indicate statistical significance (p < .05)

	ARC genotype	Study night	Interaction
Sleep architecture
Stage N1	F 2,47 = 1.03 p = .37	F 1,47  = 32.76, p < .001	F 2,47 = 1.06, p = .36
Stage N2	F 2,47 = 0.35 p = .71	F 1,47  = 13.53, p < .001	F 2,47 = 0.97, p = .39
Stage N3 (SWS)	F 2,47 = 0.36, p = .70	F 1,47  = 257.09, p < .001	F 2,47  = 4.84, p = .012
REM	F 2,47 = 0.06 p = .94	F 1,47  = 9.74, p = .003	F 2,47 = 0.51, p = .60
Total sleep time	F 2,47 = 1.17, p = .32	F 1,47  = 37.05 p < .001	F 2,47 = 37.05, p = .44
WASO	F 2,47 = 1.21 p = .31	F 1,47  = 23.38, p < .001	F 2,47 = 1.17 p = .32
Sleep latency	F 2,47 = 1.03 p = .37	F 1,47  = 32.76, p < .001	F 2,47 = 1.06 p = .36
Sleep efficiency	F 2,47 = 1.24, p = .30	F 1,47  = 35.30, p < .001	F 2,47 = 0.92, p = .41
NREM sleep EEG
Delta (0.8–4.0 Hz)	F 2,1551 = 0.07, p = .93	F 1,1551  = 790.69, p < .001	F 2,1551 = 1.13, p = .32
Theta (4.2–8.0 Hz)	F 2,1833 = 0.05, p = .95	F 1,1833  = 503.01, p < .001	F 2,1833  = 5.94, p = .003
Alpha (8.2–12.0 Hz)	F 2,1833 = 1.15, p = .32	F 1,1833  = 62.90, p < .001	F 2,1833  = 8.58, p < .001
Beta (12.2–16.0 Hz)	F 2,1833 = 0.52, p = .59	F 1,1833  = 8.66, p = .003	F 2,1833 = 1.64, p = .19
REM sleep EEG
Delta (0.8–4.0 Hz)	F 2,1320 = 0.93, p = .36	F 1,1320  = 159.08, p < .001	F 2,1320  = 14.76, p < .001
Theta (4.2–8.0 Hz)	F 2,1560 = 0.29, p = .75	F 1,1560  = 86.93, p < .001	F 2,1560  = 28.55, p < .001
Alpha (8.2–12.0 Hz)	F 2,1560 = 0.59, p = .55	F 1,1560 = 0.04, p = .81	F 2,1560  = 22.53, p < .001
Beta (12.2–16.0 Hz)	F 2,1560 = 0.15, p = .86	F 1,1560  = 136.43, p < .001	F 2,1560  = 33.19, p < .001
Waking neurobehavioral
PVT LSNR	F 2,47 = 0.63, p = .54	F 1,47  = 64.32, p < .001	F 2,47 = 0.22, p = .80
KSS	F 2,47 = 2.00, p = .15	F 1,47  = 108.62, p < .001	F 2,47 = 1.32, p = .28

There was, however, a significant interaction between ARC genotype and night for SWS (F_2,47 = 4.84, p = .012), indicating a differential response to TSD across ARC genotypes (see Figure 3, top left). During baseline sleep, time spent in SWS was 109.7 ± 40.9 min (mean ± SEM). During recovery sleep after 38 h of TSD, time spent in SWS was 165.0 ± 44.5 min, corresponding to a 55.3 min increase relative to baseline. There were large individual differences in this SWS rebound effect, with the SWS increase ranging from 16.0 to 99.0 min among subjects. ARC genotype explained a portion of these differences. Planned comparisons showed that C/C homozygotes had a larger SWS rebound than C/T heterozygotes (F_1,47 = 5.91, p = .019) and T/T homozygotes (F_1,47 = 5.66, p = .022). During recovery sleep, C/C homozygotes exhibited 61.0 ± 3.0 min more SWS, whereas C/T heterozygotes and T/T homozygotes exhibited only 46.0 ± 5.2 min and 42.4 ± 7.1 min more SWS, respectively, compared to baseline (see Figure 3).

Figure 3.

Effects of ARC genotype on the sleep response to total sleep deprivation. Differences (Δ) between recovery and baseline sleep are shown (means ± SEM) for SWS (stage N3) and for total power in the spectral bands of the NREM and REM sleep EEG for which a significant ARC genotype by night interaction was found. *p < .05, **p < .01, ***p < .001 for pairwise comparisons between ARC genotypes.

Spectral power in the NREM and REM sleep EEG

Figure 4 shows log-transformed total power in each of the spectral bands for NREM sleep (top panels) and REM sleep (bottom panels). There were no main effects of ARC genotype for any of these, but there were significant main effects of night in all cases except alpha power during REM sleep. For the NREM sleep EEG, there were increases in delta, theta, and alpha power, and a decrease in beta power, during recovery sleep after 38 h of TSD relative to baseline. For the REM sleep EEG, on the other hand, there were increases in delta, theta, and beta power during recovery sleep relative to baseline, with no significant overall change in alpha power. See Table 2 for the statistical test results. The distribution of spectral power across each sleep period for each ARC genotype is shown in the supplementary material (Supplementary Figures S1 and S2).

Figure 4.

Effects of ARC genotype on spectral power in the sleep EEG. Total power in the delta (0.8–4.0 Hz), theta (4.2–8.0 Hz), alpha (8.2–12.0 Hz), and beta (12.2–16.0 Hz) bands of NREM (stages N2 and N3) sleep (top panels) and REM sleep (bottom panels) is shown (means ± SEM, log₁₀ transformed) for baseline and recovery sleep, stratified by ARC genotype. **p < .01 for ARC genotype by night interaction.

There were significant interactions between ARC genotype and night in the NREM sleep EEG for the theta band (F_2,1833 = 5.94, p = .003) and for the alpha band (F_2,1833 = 8.58, p < .001), and in the REM sleep EEG for the delta band (F_2,1320 = 14.76, p < .001), theta band (F_2,1560 = 28.55, p < .001), alpha band (F_2,1560 = 22.53, p < .001), and beta band (F_2,1560 = 33.19, p < .001). Planned comparisons showed that T/T homozygotes exhibited larger increases after 38 h of TSD, compared to C/C homozygotes and C/T heterozygotes, for NREM power in the theta and alpha bands and for REM power in all four bands. Furthermore, C/C homozygotes exhibited larger increases compared to C/T heterozygotes for REM power in the theta and alpha bands (see Figure 4).

Waking neurobehavioral outcomes

Figure 5 shows PVT performance (LSNR) and KSS subjective sleepiness, stratified by ARC genotype, for the 12-h well-rested baseline waking period and for the circadian-matched 12-h period at the end of 38 h TSD. As expected, there were significant main effects of waking period, with poorer performance on the PVT (F_1,47 = 64.32, p < .001) and increased subjective sleepiness on the KSS (F_1,47 = 108.62, p < .001) during TSD compared to well-rested baseline. There were no significant main effects of ARC genotype and no genotype by waking period interactions.¹ See Table 2 for the statistical test results.

Figure 5.

Effects of ARC genotype on waking neurobehavioral outcomes. Psychomotor vigilance test (PVT) LSNR and Karolinska Sleepiness Scale (KSS) scores are shown (means ± SEM) at well-rested baseline and during total sleep deprivation, stratified by ARC genotype.

Covariates for age, sex, and race

There were significant effects of age on N3 (F_1,47 = 6.48, p = .014), with older individuals showing less N3; on WASO (F_1,47 = 4.98, p = .032), with older adults experiencing more WASO; and on NREM delta power (F_1,1551 = 4.07, p = .044), with older adults exhibiting less delta power. In secondary analyses, there were significant effects of sex on N3 (F_1,47 = 5.77, p = .020), with women having more N3; on SE (F_1,47 = 4.07, p = .049), with women experiencing less SE; on NREM delta power (F_1,1551 = 12.33, p < .001), theta power (F_1,1833 = 9.26, p = .002), alpha power (F_1,1833 = 5.23, p = .022), and beta power (F₁,₁₈₃₃ = 15.55, p < .001), with women experiencing more NREM power across all four bands; and on REM delta power (F_1,1320 = 6.81, p = .009), theta power (F_1,1560 = 9.31, p = .002), alpha power (F_1,1560 = 15.88, p < .001), and beta power (F_1,1560 = 5.79, p = .016), with women experiencing more REM power across all four bands. There were significant effects of race on REM delta power (F_1,1320 = 2.91, p = .032), with individuals identifying as White exhibiting more delta power than individuals identifying as Black or mixed race. Controlling for sex and race in the secondary analyses did not alter any of the main findings.

Discussion

We investigated the role of the human ARC SNP (rs35900184) in the sleep physiological and waking neurobehavioral responses to TSD. We found that the ARC T allele is associated with an attenuated SWS response to TSD, while simultaneously displaying intensified NREM and REM EEG spectral power responses to TSD, with no apparent genotype effects on waking neurobehavioral functioning. These seemingly incongruent findings suggest that ARC may play a role in homeostatic sleep regulation, as we hypothesized, but its impact is more complex than anticipated.

As expected, 38 h of TSD resulted in distinct changes in overall sleep architecture, as observed in recovery sleep compared to baseline sleep. Given 10-h TIB for both baseline and recovery sleep, there were characteristic increases in TST, SE, N1, SWS, and REM in response to TSD, whereas SL, N2, and WASO decreased (Figure 2) [36]. Overall, spectral power in the sleep EEG increased after TSD in the NREM delta, theta, and alpha bands and in the REM delta, theta, and beta bands, while NREM beta power was reduced, regardless of ARC genotype (Figure 4), as previously observed [5, 9, 37–40]. TSD also degraded performance on the PVT and increased subjective sleepiness on the KSS (Figure 5), indicating that the TSD intervention of the study produced the intended sleep physiological and waking neurobehavioral effects. Expected age-related changes in WASO, SWS, and NREM delta power were observed [41, 42], as well as sex differences in multiple sleep parameters [42]—in addition to an effect of race on REM delta power, which appears to be a novel observation.

From the outset, we hypothesized that the ARC SNP investigated here would be associated with differences in the SWS and SWA (NREM delta power) homeostatic responses to TSD. We did find that individuals homozygous for the C allele had a greater SWS rebound during recovery sleep, relative to baseline, compared to carriers of the T allele (Figure 3, top left). There was no significant effect of ARC genotype on REM sleep, indicating that the effect of the C allele on SWS was not mediated by a difference in REM rebound. The ARC genotype effect on sleep homeostasis as measured by SWS is consistent with evidence demonstrating that Arc knockout mice exhibit an altered sleep rebound in NREM sleep in response to 4 h of sleep deprivation [22].

Yet, individuals homozygous for the C allele did not show a more intense rebound during recovery sleep after TSD in NREM delta power, as would have been expected to be associated with the greater SWS rebound [43]. Rather, individuals homozygous for the T allele expressed more NREM power in the theta and alpha bands and more REM power across the EEG spectrum (Figure 3). Theta and alpha power in the NREM sleep EEG have been found to correlate with homeostatic sleep pressure [44], suggesting that T/T homozygotes might have a higher sleep need. However, such an interpretation would be inconsistent with the higher SWS rebound in the C/C homozygotes, and with the absence of any genotype differences in estimated habitual sleep duration. As such, the significance of our findings on EEG spectral power with regard to sleep homeostasis, if any, is unclear. Structural brain differences mediated by ARC genotype may underlie the pervasive differences across multiple NREM and REM EEG spectral bands; this remains to be investigated.

No effect of ARC genotype on waking neurobehavioral functioning was observed. This result does not support our hypothesis that any differences in sleep homeostatic regulation attributable to ARC would have consequences for the waking neurobehavioral response to TSD. While individual differences in the sleep-related and wake-related responses to sleep deprivation do not necessarily align [18,45–51], the lack of an ARC genotype effect on such sensitive measures as PVT performance impairment and KSS subjective sleepiness during TSD would be unexpected if ARC regulates sleep homeostasis. On the other hand, the ARC gene has been shown to be involved in synaptic plasticity in the brain [52]—a process critical for optimal next-day functioning that has previously been linked to SWS [53]. ARC gene expression is induced in response to extended wakefulness [22], and is produced in neurons specifically involved in learning and memory [54, 55]. Investigating the effects of ARC genotype on the response to TSD in learning and memory tasks might have yielded more insight into ARC’s role in waking neurobehavioral functioning while sleep deprived.

An important limitation of our study is that the sample was relatively small, with a particularly small representation of T allele homozygotes, and the prevalence of the ARC genotypes was not in Hardy–Weinberg equilibrium, so our results may not generalize to the larger population of healthy young adults from which the sample was drawn. That does not, however, change the remarkable diversity of ARC genotypic influences across the sleep physiological and waking neurobehavioral outcomes [56]. Our REM EEG spectral power data should be interpreted with care, however, as considerable portions of the REM sleep EEG were discarded during artifact rejection due to cross-talk from REM-associated eye movements. In addition, while the functional significance of the ARC SNP is not fully understood, our findings suggest that the T allele assessed here behaves similarly to mice lacking the Arc gene altogether [22], and therefore likely alters gene function. However, additional research is needed to fully elucidate the functional significance of this ARC SNP at a molecular level. Lastly, it should be noted that while polymorphisms of a single gene may provide insight into underlying mechanisms, phenotype–genotype relationships are usually multi-dimensional and other genes and factors are likely to be involved in shaping our study results [57]. As genes and their associated variants rarely work in isolation, it is possible that the ARC SNP assessed here may be inherited, or closely linked, to other SNPs of the same or nearby genes that may influence ARC’s effects on the homeostatic regulation of sleep. Therefore, we cannot be entirely certain that the effects seen here can be attributed to the ARC SNP alone.

In conclusion, our data show that human ARC genotype plays a role in inter-individual variability in the sleep physiological response to sleep deprivation. Carriers of the T allele show a dampening of the SWS rebound characteristic of recovery sleep after TSD, but without a commensurate change of spectral power in the NREM sleep EEG. On the other hand, T/T homozygotes show increases in NREM theta and alpha power, as well as power across all four REM spectral power bands examined here. No significant effect of ARC genotype on waking neurobehavioral functioning during TSD was found. Collectively, these findings are only partially consistent with ARC genotype mediating the sleep homeostatic response to sleep deprivation. ARC’s role in the sleep physiological and waking neurobehavioral responses to sleep loss may be more specifically related to ARC’s role in synaptic downscaling during sleep [55]. This might also explain the ARC genotype-related differences we observed in spectral power in the sleep EEG in frequency bands other than delta; at least for theta power a relationship with synaptic downscaling has been noticed [58, 59]. Finally, we note that ARC might be a target of interest for pharmacological interventions aiming to moderate the sleep physiological rebound after sleep deprivation, which could be of relevance in sleep disorders and other medical conditions involving hypersomnolence.

Author contributions

Brieann C. Satterfield (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review & editing [equal]), Myles G. Finlay (Data curation, Methodology, Writing—review & editing [equal]), Sofia K. Fluke (Data curation, Writing—review & editing [supporting]), Lillian Skeiky (Data curation, Writing—review & editing [supporting]), Michelle A. Schmidt (Methodology, Writing—review & editing [supporting]), Jonathan Wisor (Conceptualization, Writing—review & editing [supporting]), and Hans P. A. Van Dongen (Formal analysis, Funding acquisition, Investigation, Methodology, Writing—original draft, Writing—review & editing [equal])

Disclosure statement

Financial disclosure: This research was funded by the Office of Naval Research grant N00014-13-1-0302, the National Institutes of Health grant R21CA167691, and the Army Research Office Multidisciplinary University Research Initiative grant W911NF2210223. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Non-financial disclosure: None.

Data availability

The datasets generated and analyzed as part of this manuscript are available from the corresponding author upon reasonable request.