Acute partial sleep restriction attenuates working memory performance and does not affect BDNF in young adults

Introduction

Sleep plays a critical role in modulating cognitive function, and short sleep durations impose a profound physiological stress that adversely impacts cognition in humans. Despite the importance of continuous and adequate sleep, contemporary societal demands shorten sleep duration [1] such that almost 10 % of the global population reports sleep durations of less than 6 h per night [2]. The impact of shortened sleep on cognitive function depends on numerous factors, including sleep quality, sleep duration, and frequent insomnia symptoms [3], [4], [5], [6]. Chronically shortened sleep associates with decreases in working memory (WM) performance, decreases in vigilance [7], 8], and increases in the risk of neurological disease [9]. Acute reductions in sleep durations, such as observed during total sleep deprivation (TSD) where sleep is eliminated for 24 h or more, also impairs both motor performance [10] and neurobehavioral performance which may increase rates of serious workplace errors [11].

Indeed, laboratory evidence demonstrates that acute TSD or partial sleep restriction (PSR) transiently impairs many aspects of cognition including WM, evidenced by decreases in accuracy on an N-back test of WM in healthy humans [12], [13], [14]. Not surprisingly, the magnitude of cognitive dysfunction increases as a function of greater accumulated sleep debt [13], such that total sleep deprivation results in exacerbated cognitive dysfunction compared to PSR across longer periods of wakefulness [14]. Indeed, total sleep deprivation represents an excellent model for investigating the deleterious physiological effects of prolonged wakefulness; however, humans may rarely miss an entire night of sleep. Conversely, PSR due to poor sleep hygiene and rigid lifestyle and occupational schedules is likely more common in society across longer timescales [15]. Although the negative impacts of chronic PSR on cognition are well described, the effects of acute PSR (e.g., one night of 3 h of sleep) on WM in males and females are comparatively less researched.

Factors that underlie the impaired cognition that is associated with shortened sleep may be integrative in nature. One such factor may involve changes in concentrations and signalling of the neurotrophin brain-derived neurotrophic factor (BDNF). BDNF is critical for cognitive processes including learning and memory. Further, BDNF orchestrates activity-dependent synaptic plasticity and is implicated in the regulation of sleep by increasing regional sleep pressure in activity-dependent manner during waking hours [16]. Circulating BDNF follows a biphasic response to sleep stressors, where chronic perturbations to sleep decrease concentrations of circulating BDNF, and acute total sleep deprivation increases BDNF levels [17], 18]. Increases in circulating BDNF, without the expected impairments to cognitive performance following acute total sleep deprivation, may reflect synaptic potentiation due to prolonged waking without appropriate synaptic rescaling during sleep and thereby uncouple the protective relationship between elevated circulating BDNF and cognitive performance [17]. The impact of acute PSR on circulating BDNF in healthy humans, however, remains unexplored. It is plausible that the early stress-associated elevations in circulating BDNF following sleep restriction may contribute to the preservation working memory, in accordance with the proposed adaptive response. However, domain-specific attenuations in attentional measures, which are sensitive to acute instances of PSR [13], may be impaired due to the increases in pressure to sleep prompted by BDNF [16] and decreases in vigilance characteristic of acute sleep restriction. Furthermore, adequate vascular function such as vasomotion of cerebral arterioles penetrating the brain parenchyma [19] – which may be in part facilitated by BDNF signaling [20] – is integral for sleep-dependent metabolic flux in the brin and the overall maintenance of cerebral health and cognitive function [21]. Research further indicates that maladaptive sleep causes significant perturbations to vascular function [22]. Examining associations between vascular function and WM performance during instances of restricted sleep presents an opportunity to highlight cognitive performance as the output of interactions between multiple physiological systems, all reliant on adequate sleep duration.

The objective of this study was to investigate the effect of PSR on circulating plasma BDNF and WM performance in young and healthy adults. We hypothesized that WM performance would decrease following a single night of PSR, and that this impairment would be associated with an increase in circulating plasma BDNF compared to when measured after one night of normal sleep. The summary of this article is presented in Figure 1.

Figure 1:

Graphical representation of this study. Key points: (1) working memory assessments were conducted using an N-back task and plasma BDNF was measured after normal sleep and after acute partial sleep restriction in order to examine the relationships between acute perturbations to sleep and circulating BDNF dynamics in humans. (2) Sleep restriction slightly reduced reaction time during the N-back task relative to a normal sleeping control group. (3) Changes in reaction time were independent of changes in plasma BDNF.

Materials and methods

This paper presents an analysis of secondary outcomes from an experiment that was designed to examine the effects of PSR on vascular function [23]. Twenty-two male and female adults between the ages of 18–35 were recruited to participate in this experiment involving one night of PSR. A parallel-arm and normal sleeping time-control group, consisting of an additional sex-matched 22 participants, was used to account for the passage of time and potential learning effects due to repeated cognitive testing. Inclusion criteria for participation were the absence of any diagnosed sleeping disorders (e.g., obstructive sleep apnea, insomnia, narcolepsy), or a history of diagnosed medical disorders including any cardiovascular, musculoskeletal, and metabolic diseases. Data from female participants in the experimental group were collected during the early follicular phase of the menstrual cycle as determined by self-report, except for one participant who reported use of an intrauterine device. All participants provided written informed consent and all experimental procedures that were approved by the Hamilton Integrated Research Ethics Board (Project ID: 8270) and conformed to the Declaration of Helsinki, with the exception of preregistration.

Results

Working memory performance

Working memory performance scores assessed using the N-back test are presented in Table 2. We found a significant group by time interaction for the 3-back mean hit RT (p=0.038; β=−100.46; 95 % CI [−192.20, −8.73]), where the mean absolute magnitude of change in RT from Visit 1 to Visit 2 in the experimental group recorded a +36.13 ms difference (95 % CI [−33.87, 106.13], p=0.697), and the control group recorded a mean RT −64.34 ms difference (95 % CI [−132.15, 3.47], p=0.227; Figure 2). There was, however, no group by time interaction in the overall mean hit RT (p=0.070; β=−41.20; 95 % CI [−84.49, 2.09]), nor in the 1-back mean hit RT (p=0.972; β=0.61; 95 % CI [−33.41, 34.63]). We evaluated the overall proportion of correct responses and found no group by time interaction (p=0.534; β=0.606; 95 % CI [−1.29, 2.50]). The statistically insignificant group by time interaction persisted when analyzing the hit rate in the 1-back condition (p=0.542; β=−0.01; 95 % CI [−0.05, 0.02]) and 3-back condition (p=0.756; β=0.01; 95 % CI [−0.06, 0.09]) separately. There were no significant group by time interactions for overall d-prime scores (p=0.666; β=0.09; 95 % CI [−0.30, 0.48]), 1-back d-prime scores (p=0.731; β=−0.08; 95 % CI [−0.53, 0.37]), or 3-back d’ scores (p=0.614; β=0.14; 95 % CI [−0.39, 0.66]).

Table 2:

Working memory performance as evaluated by the N-back experimental paradigm.

	Time-control group (n=22)		Experimental group (n=22)		p-Value
	Visit 1	Visit 2	NS	PSR	Condition	Group	Interaction
1-Back hit rate	98.8 (3.3)	96.4 (7.3)	97.6 (6.4)	96.4 (8.2)	0.072	0.725	0.542
3-Back hit rate	79.1 (14.3)	83.0 (13.6)	72.7 (16.1)	75.5 (15.9)	0.093	0.095	0.756
1-Back d-prime	4.95 (0.52)	4.67 (0.75)	4.77 (0.77)	4.56 (0.91)	0.038a	0.456	0.731
3-Back d-prime	2.71 (1.10)	3.04 (1.15)	2.40 (0.94)	2.60 (1.00)	0.058	0.193	0.614
1-Back RT	579 (121)	587 (140)	630 (135)	638 (130)	0.347	0.195	0.972
3-Back RT	818 (191)	754 (133)	822 (247)	858 (249)	0.550	0.364	0.038a

1. RT, reaction time (ms); LMEM, linear mixed effects model. Data from n=22 represent means (SD); ^aSignificantly different at p<0.05 using a linear mixed effects model.

Figure 2:

Mean hit reaction time (milliseconds; ms) after visit #1 and after visit #2 in the time-control group, and after normal sleep and partial sleep restriction in the experimental group. Grey lines represent individual responses, and the dot and err or bars represent mean±SD (n=22 per group).

Circulating plasma BDNF

Blood samples were obtained from fourteen participants in the experimental group who completed the WM tests, and blood samples were also obtained from an additional two participants who completed the sleep restriction intervention but without WM performance metrics. Some participants had completed the trial before blood samples were included in the protocol, explaining the discrepancy in the total number of participants vs. the number of participants with blood samples. Plasma BDNF failed tests of normality and were right-skewed based on positive Fisher skewness coefficients. We therefore log-transformed the plasma BDNF data which then subsequently passed normality testing. There was no change in plasma BDNF levels following PSR compared to NS (presented on the log scale; NS: 489.01±590.94 pg/mL vs. PSR: 516.58±691.00 pg/mL; p=0.720; Cohen’s d=−0.04; Figure 3A). Contrary to our hypothesis, changes in the impaired 3-back RT response did not relate to changes in circulating plasma BDNF (R²=−0.07; p=0.769; Figure 3B). Further, exploratory correlation analyses revealed that there were no associations between plasma BDNF and measures of WM N-back performance metrics (Figure 3B; all p>0.05).

Figure 3:

Plasma BDNF following normal sleep (NS) versus partial sleep restriction (PSR), and its relation to changes in working memory performance. (A) Log plasma BDNF (pg/mL) measured after NS and after acute PSR. Grey lines represent individual participant responses (n=16). Data represents mean±SD. (B) Relationship between the change in 3-back (N3) mean hit reaction time (RT) and the change in BDNF (pg/mL) between normal sleep and acute PSR (n=14).

Exploratory associations between working memory performance, cardiorespiratory fitness, and vascular measures

We were interested to see if cardiorespiratory fitness, estimated by VO₂ peak, moderated the change in RT between conditions given prior research suggesting protective effects of cardiorespiratory fitness on WM performance during poor sleep [26]. The standardized change in RT during the 3-back condition was regressed against VO₂ peak and, contrary to our exploratory hypothesis, we found a negative linear relationship between the standardized change in RT and VO₂ peak (Pearson’s r=−0.48, p=0.025; Figure 4). This negative relationship persisted after adjusting for sex and the perceived impact of sleep restriction. Another exploratory analysis was conducted to examine associations between indices of peripheral vascular function and working memory between NS and acute PSR. No associations were detected between changes in any resting hemodynamic parameter (SBP, DBP, and HR) and measures of working memory performance (all p>0.05) between sleep conditions. Furthermore, no associations were found between changes in arterial stiffness estimated by pulse wave velocity and measures of working memory performance (all p>0.05). Interestingly, there was a significant negative correlation between changes in %FMD and changes in overall RT (Pearson’s r=−0.47, 95 % CI [−38.3, −2.3], p=0.028; Figure 5A), and changes in RT in the 3-back condition, such that larger %FMD values correlated with faster RT (Pearson’s r=−0.43, 95 % CI [−79.2, −1.3], p=0.044; Figure 5B).

Figure 4:

Correlation analysis depicting the standardized change in mean hit reaction time in the 3-back (N3) condition relative to a time-control as a function of cardiorespiratory fitness (VO₂ peak) in the experimental group (n=22).

Figure 5:

Correlation analysis of the association between the change in 3-back (N3) reaction time (A) and overall reaction time (B) as a function of the change in % flow-mediated dilation measured at the brachial artery (BA FMD) from the experimental group (n=22).

Discussion

The purpose of this study was to investigate the impact of acute PSR on circulating plasma BDNF and WM performance in young adults. A single night of PSR acutely worsened WM performance compared to a normal sleeping time-control group, such that there was no observed improvement due to repeated testing in the PSR group. This effect was mild in magnitude given the opposing directionality of the mean RT scores between conditions across the experimental and control groups, and did not persist following post-hoc testing. However, contrary to our hypothesis there was no effect of PSR on plasma BDNF. Interestingly, exploratory analysis revealed that higher cardiorespiratory fitness was associated with a greater decrement in WM after acute PSR. These data add to emerging findings that acute PSR may alter WM performance independent of plasma BDNF concentrations. These findings provide important insights into the integrated physiological and neurobiological consequences of acute sleep restriction.

The observed acute and mild impairment in RT during the 3-back condition is consistent with work showing slower processing speed in females after acute PSR [36]. We add to this burgeoning body of literature by showing that acute PSR mildly impacts reaction time in both males and females. We present novel findings that changes in WM occur without changes in plasma BDNF. A shortened sleep opportunity may limit the processes underlying memory consolidation and synaptic rescaling [37]. Reduced opportunities for memory consolidation due to acute PSR may, in part, have limited the learning effects due to repeated N-back testing that ostensibly accrued overnight in the time-control group. Consistent with the active systems consolidation theory [37], synaptic rescaling during sleep reorganizes synaptic connections within and between the hippocampus and neocortex. Accordingly, acute PSR may impair consolidation of synapses created during the preceding waking period. Our data suggest, however, a mild interaction between acute PSR and WM performance given a small RT increase during the high-cognitive load 3-back task in the experimental group relative to the control group. This observation aligns with the notion that WM decrements are exacerbated as a function of accruing short sleep [14], and highlights acute PSR as a relevant, but mild cognitive stressor. Future research using an acute PSR model may therefore be guided by our findings to include larger sample sizes that are statistically powered to detect the small effects that we report herein.

The neurotrophin BDNF is critical for orchestrating synaptic plasticity and memory formation and plays an important role in regulating sleep pressure [16]. Accordingly, we assessed changes in plasma BDNF in the morning after a single night of PSR to examine the extent to which changes in circulating BDNF, particularly in a context of elevated sleep pressure, related to changes in WM. Contrary to our hypothesis, there was no change in plasma BDNF following acute PSR compared to NS. Reducing sleep to 3 h on a single night may not be a sufficient stimulus to evoke a stress-induced adaptive response. Indeed, circulating BDNF increases following acute total sleep restriction and may represent an adaptive response to protect against cognitive impairment during waking hours and increase non-rapid eye movement sleep during subsequent convalescent sleep [38]. In agreement with others [13], 36], the current experiment shows mild effects of acute PSR on WM evidenced by a slowing of response time in the 3-back condition only, which represents a higher imposed WM processing challenge. Indeed, the 1-back performance was not different from baseline, nor from the time-control group, suggesting that the effects of acute sleep loss are limited to tasks with higher WM challenge. These results substantiate the speculation that sleep loss may first impact cellular neuronal signalling before progressing to affect broader physiological and cerebral processes as the duration of sleep restriction increases. Therefore, the impact of acute PSR on BDNF and other physiological markers may not be as pronounced as observed in more severe forms of acute sleep loss such as TSD.

Prior research indicates that higher levels of cardiorespiratory fitness may protect cognition following acute sleep deprivation [26]. Accordingly, we explored cardiorespiratory fitness as a candidate physiological factor that may buffer against impaired WM following PSR. However, we observed the opposite relationship, as higher VO₂ peak was associated with slowed RT acutely following a single night of PSR. These data suggest that individuals with higher aerobic fitness in our sample may have placed a greater emphasis on lifestyle factors, such as exercise and sleep. The disturbance to sleep caused by our intervention may have been perceived as more disruptive for these individuals, which may have influenced working memory performance via expectancy biases. Indeed, individual perceptions of sleep duration have been shown to modulate RT, independent of objectively measured sleep duration [39]. However, the negative relationship between VO₂ peak and WM RT was not impacted when we controlled for sleep perceptions in our sample. Future research may build on this exploratory finding to determine the extent to which the perceptions of sleep loss and associated expectancy biases modulate WM performance, particularly in individuals with high aerobic capacities.

We also explored the relationship between peripheral vascular function and cognition, as previous research indicates that vasomotion in the cerebral vasculature affects metabolic clearance from the brain during sleep [40]. We regressed changes in vascular endothelial cell function against changes in 3-back RT between sleep conditions and found that increased vascular function was associated with faster 3-back RT. Although exploratory, this preliminary result is conceptually consistent with importance of vascular dynamics for facilitating the flow and clearance of metabolic waste throughout the brain parenchyma during sleep [19]. A responsive endothelium, as inferred by improved conduit artery vasodilation following hyperemia, may confer a more robust cerebrovascular oscillation and therefore enhance clearance. This result, however, is exploratory and the associations between peripheral and cerebral endothelial function that are required to support this hypothesis are not yet sufficiently established [41].

Among the strengths of this study include the ecological relevance of the study design wherein sleep restriction occurred in free-living conditions familiar to participants. This enabled participants to partake in behaviours intrinsic to their daily routines within conditions, in contrast to the constrained array of activities inherent in a supervised sleep restriction setting. Furthermore, the N-back task was used to assess a spectrum of imposed cognitive processing demands that spanned from easier processing loads (1-back condition) to comparatively difficult processing loads (3-back condition) in a randomized order. That a discernable impairment in RT was observed in the 3-back condition underscores the discriminative capacity of the N-back task in delineating nuanced cognitive impairment.

Limitations of this experiment also warrant acknowledgment. First, BDNF has been proposed to follow a biphasic pattern such that circulating concentrations of BDNF increase concomitantly with the acute stressor, and then dissipate during the sleep opportunity [17], 18]. A nighttime measure of circulating BDNF during the sleep restriction period, while sleep pressure is building, would increase the temporal resolution of circulating BDNF levels and therefore provide a better index of the integrative stress that acute PSR imposed on the body. Second, we reconciled the observed impairment in RT by deferring to reductions in opportunities for active systems consolidation and synaptic rescaling. That our current study had no direct measures of either encephalography or synaptic scaling as a proxy for consolidation limits our mechanistic speculation and presents an intriguing opportunity for further research. Objective measures of sleep quality, such as polysomnography, may have also provided insight into sleep architecture and learning that was not captured with the subjective and objective methods used to measure sleep duration herein. Measures of sleep architecture during partial sleep would provide powerful evidence to substantiate the observed effects of acute PSR on WM to extend our findings. Outcomes involving the proportion of time spent in REM sleep, and the slow-wave amplitudes characteristic of NREM, should be described in subsequent PSR interventions. Sex differences in WM performance following acute PSR have not been investigated. Our research was underpowered to detect potential sex differences, however equal representation of sexes was included in each group and sensitivity analyses with participants separated by sex did not indicate that results would differ between sexes.

Conclusions

Acute PSR yielded a mild impairment to RT relative to a time-matched and normal sleeping control on the N-back test of WM independent of BDNF changes between NS and acute PSR. The subtle effects of PSR on RT observed in our data have broad implications for public health messaging given the prevalence of PSR, and highlights sleep duration as a relevant consideration prior to the assessment of WM in future investigations particularly where anticipated effects may be small. Furthermore, the slower RT response occurred in the absence of changes in plasma BDNF. Our exploratory analyses warrant follow-up experiments that measure integrative mechanisms, such as cardiorespiratory fitness and hemodynamic function, which may contribute to changes in cognitive processing during perturbed sleep. Our research characterizing the effects of acute PSR on WM performance and plasma BDNF levels in young adults contributes valuable insight into our understanding about the immediate consequences of short-term sleep insufficiency.