Sleep-dependent upscaled excitability and saturated neuroplasticity in the human brain: From brain physiology to cognition

word count: 156 Main text word count (excluding the Materials and methods, References, and Figure legends): 5130 Display items (Figures + Tables): 7 Supplementary material display items: 3 figure, 5 table


18
Over the past decade, a strong link has been established between sleep and cognition 19 (Deak and Stickgold, 2010;Lim and Dinges, 2010). Adequate sleep is critical for optimal 20 cognitive functions across the lifespan (Carskadon, 2011;Lo et al., 2016) and findings from 21 experimental settings support this critical role of sleep for cognition in animals (Walker and 22 Stickgold, 2006), and humans (Krause et al., 2017) especially for memory consolidation and 23 sequence learning (Stickgold, 2005;Chouhan et al., 2021). As a ubiquitous physiological 24 phenomenon, sleep has extensive impacts on brain physiology and especially on parameters 25 relevant for cognition such as brain excitability and plasticity. 26 Previous experimental studies, mostly in nonhuman animals, have linked sleep and synaptic 27 homeostasis. Specifically, extended wakefulness (or sleep deprivation) is associated with the 28 expression of long-term potentiation (LTP)-related molecular changes and plasticity-related 29 genes (e.g. BDNF, CREB) in the brain, leading to saturation of synaptic potentiation (Tononi 30 and Cirelli, 2003). Sleep, on the other hand, desaturates synapses that have been potentiated 31 during wakefulness, resulting in an improved signal-to-noise ratio and a renewed capacity for 32 encoding new information and cognitive processing (Kuhn et al., 2016). A recent study 33 confirmed this sleep-dependent synaptic downscaling by showing reduced or weakened 34 synaptic connections in the primary motor and somatosensory cortex during sleep (de Vivo et 35 al., 2017). This demonstrates that sleep is required for preparing the brain for proper 36 cognitive, motor, and physiological functioning, however, the effect of sleep on specific 37 parameters of human brain physiology and their association with cognition and behaviour 38 remains to be further determined. 39 In humans, molecular mechanisms of synaptic homeostasis cannot be directly studied, 40 however, non-invasive (indirect) markers of brain physiology can be used for studying the 41 impact of sleep and extended wakefulness on synaptic potentiation and cortical excitability 42 (Kuhn et al., 2016). Non-invasive brain stimulation (NIBS) techniques are safe methods for 43 directly monitoring, and modifying brain functions in humans providing a means for studying 44 the causality of brain-behaviour relationships (Polanía et al., 2018). Several NIBS techniques,45 including transcranial magnetic stimulation (TMS) and transcranial electrical stimulation 46 (tES), are widely used to non-invasively monitor and induce changes in cortical excitability, 47 and neuroplasticity (Nitsche and Paulus, 2000;Huang et al., 2017;Polanía et al., 2018). 48 Recently, these techniques have been used for studying the contribution of sleep to brain 49 physiological and cognitive functions in humans. It has been shown that cortical excitability 50 increases after sleep deprivation (Kuhn et al., 2016;Ly et al., 2016). This increase of cortical 51 excitability after sleep deprivation is linked to increased net synaptic strength in humans (i.e., 52 TMS intensity required to elicit a predefined motor-evoked potential, EEG activity) (Kuhn et 53 al., 2016), animals (i.e, synaptic potentiation) (Vyazovskiy et al., 2008), and reduced 54 inhibitory intracortical mechanisms (Kreuzer et al., 2011;Placidi et al., 2013). This can 55 reduce the ability of the brain to induce neuroplasticity (as a result of synaptic saturation). In 56 this line, decreased LTP has been shown in rats after sleep deprivation shown in both, in vivo 57 and in vitro (McDermott et al., 2003;Kopp et al., 2006;Vyazovskiy et al., 2008), and a recent 58 human study also showed decreased LTP-like plasticity, induced by paired associative 59 stimulation (Kuhn et al., 2016). 60 While these studies added novel insight into sleep-dependent effects on human brain 61 physiology, a comprehensive investigation of how brain physiology and cognition interact 62 under sleep deprivation is missing. Different excitatory and inhibitory neurotransmitter 63 systems (e.g. glutamatergic, dopaminergic, GABAergic, cholinergic) are involved in cortico-64 cortical and corticospinal excitability which are also closely related to neuroplasticity and 65 cognition. These specific parameters of brain excitability can be monitored with different 66 TMS protocols. Furthermore, the suggested role of the sleep-wake cycle for the inducibility of 67 synaptic plasticity, which appears to be dependent on an optimal sleep-dependent temporal 68 window (Kuhn et al., 2016), can be investigated via induction of both LTP-and/or LTD-like 69 plasticity. Very few studies in humans have investigated induced neuroplasticity under sleep 70 deprivation conditions so far, and these studies are mostly limited to LTP-like plasticity 71 (Kuhn et al., 2016), leaving LTD-like plasticity untouched. Gaining knowledge about LTP-72 and LTD-like plasticity mechanisms of action can extend the sleep synaptic hypothesis 73 (Tononi and Cirelli, 2014) for the human brain. Moreover, whether and how these 74 physiological parameters are associated with cognitive functions remains to be investigated.

75
A common research paradigm for studying the effect of sleep on these parameters of brain 76 physiology and cognition is sleep deprivation. In this context, a strong and strict control is 77 exerted over all possible exogenous (body posture, light, temperature) and endogenous (stress 78 level, digestion, motivation) factors that might affect the sleep-wake cycle, avoiding a 79 confounding influence of these parameters on the factors of interest (Schmidt et al., 2007). In 80 this paradigm, participants are kept in an extended wakefulness condition and are deprived of 81 sleep for a certain amount of time. Using this paradigm, we systematically investigated the 82 impact of one-night sleep deprivation, compared to one-night sufficient sleep on converging 83 measures of human brain physiology and cognition. Specifically, we monitored cortical 84 excitability of the brain via TMS protocols that measure cortical inhibition and facilitation (i), 85 induced both LTP-like and LTD-like plasticity (ii), and measured synaptic strength via the 86 resting-EEG theta/alpha pattern (iii). We also examined learning and memory formation, 87 behavioral indices of brain plasticity (iv), and higher-order cognitive functions (attention, 88 working memory) which depend on cortical excitability (v) and their electrophysiological 89 correlates. Cortisol and melatonin levels were also assessed (vi). All measurements were 90 conducted under sufficient sleep vs one-night sleep deprivation at fixed times ( Figure 1). 91 Briefly, we demonstrate that after sleep deprivation, the human brain displays a hyperexcited  Neuroplasticity was induced with anodal and cathodal stimulation after sufficient sleep (SS) vs sleep deprivation (SD). SI1 mv=Stimulation intensity to elicit an MEP amplitude of 1 mV, M1=primary motor cortex, Fp2=right supraorbital area. c, Saliva samples were taken at 8:45 in each session. Following the resting-EEG acquisition, participants performed motor learning, working memory, and attention tasks at the beginning of each experimental session (sufficient sleep vs sleep deprivation) while their EEG was recorded. SRTT=serial reaction time task, AX-CPT=AX continuous performance task.  Corticospinal and corticocortical excitability after sufficient sleep vs sleep deprivation. a, There is a trend of higher corticospinal excitability after the sleep deprivation session compared to sufficient sleep especially at 150% of RMT intensity. b, Cortical inhibition significantly decreased after sleep deprivation as compared with sufficient sleep (tISI2=4.24,p<0.001;tISI3=4.50,p<0.001). In contrast, cortical facilitation is significantly upscaled after sleep deprivation compared with sufficient sleep (tISI10=7.69, p<0.001; tISI15= 6.66, p<0.001). c, I-wave peaks were significantly facilitated for early and middle ISIs after both, sufficient sleep and sleep deprivation, and for late ISIs only after sleep deprivation. For all ISIs, I-wave peaks were significantly more upscaled after sleep deprivation vs sufficient sleep (tearly=3.90, p<0.001; tmiddle=3.91, p<0.001; tlate=4.40, p<0.001), indicative of less cortical inhibition. d, Cortical inhibitory effect of peripheral nerve stimulation on motor cortical excitability was observed only after sufficient sleep (tISI20=4.53,p<0.001;tISI40=4.25,p<0.001), whereas the inhibitory effect of peripheral stimulation turned to excitatory effects after sleep deprivation (tISI20=2.54,p=0.035;tISI40=4.55,p<0.001). MEP amplitude were significantly upscaled after sleep deprivation vs sufficient sleep (tISI20=7. 08,p<0.001;tISI40=8.83,p<0.001). All pairwise comparisons were calculated using the Bonferroni correction for multiple comparisons (n=30). All error bars represent the standard error of means (s.e.m.). Filled symbols represent a significant difference in MEP amplitudes compared to the respective test pulses (for SICI-ICF, I-Wave, SAI) or MEP at RMT intensity (for I-O curve). Asterisks represent statistically significant comparisons between sleep conditions. Note: MEP = motor evoked potential; RMT = resting motor threshold; ms = milliseconds.

163
Taken together, our results demonstrate that corticocortical and corticospinal excitability are 164 upscaled after sleep deprivation. Interestingly, cortical inhibition was decreased or turned into 165 facilitation which is again indicative of higher cortical excitability. Cortical excitability is 166 closely related to LTP/LTD plasticity in the brain and is expected to be related to changes of 167 synaptic potentiation after lack of sleep (Tononi and Cirelli, 2003;Kuhn et al., 2016). 168 Accordingly, in the next step, we investigated the impact of non-invasively inducing Here, we were interested in determining how sleep deprivation, and the resultant upscaled 174 cortical excitability, affect LTP-and-LTD-like plasticity in the brain. The sleep synaptic 175 hypothesis proposes that synaptic strength is saturated during long awake times and restored 176 after sleep (Tononi and Cirelli, 2014). Saturation can lead to decreased LTP-like plasticity in 177 humans (Kuhn et al., 2016). Accordingly, we expected that induction of LTP-like plasticity is 178 decreased due to saturated synaptic strength and hyperexcited brain state identified in the 179 previous section. We were also interested in determining how induction of LTD-like plasticity 180 is affected under this brain state, which has not been investigated so far. To this end, 181 participants received "anodal vs sham" and "cathodal vs sham" transcranial direct current  increased cortical excitability due to sleep pressure (Vyazovskiy et al., 2008;Kuhn et al., 227 2016). In line with this, we investigated how sleep deprivation affects resting-state brain 228 oscillations at the theta band (4-7 Hz), beta band (15-30 Hz) as another marker for cortical 229 excitability, vigilance and arousal (Eoh et al., 2005;Fischer et al., 2008) and alpha band (8-14 230 Hz) which is important for cognition (e.g. memory, attention) (Klimesch, 2012 difference at blocks 6-5 and blocks 6-7 only after sufficient sleep and lower committed errors 266 ( Fig. 5a,b). Absolute RT, error rate, and RT variability were similarly affected by sleep 267 deprivation (Fig. S1). Next, we explored electrophysiological correlates of motor learning.

268
The P300 component is evoked in response to stimuli of low probability and stimulus 269 sequence (Squires et al., 1976 analyses indicated that sleep deprivation was related to a significantly smaller P300 amplitude 320 in these (Fig 5o,p), and other electrodes of interest ( Fig S3c).  (n=1) and ERP analyses (n=2) and AX-CPT ERP analyses (n=3), the data of some participants were excluded from the analysis for excessive noise. BL = block, RT = reaction time. See also Figure S1-S3. The mean age of participants was 24.62±4.16 years. 50 percent of the participants were male.

330
Age and gender did not correlate with the dependent variables discussed in previous sections.

331
Ratings of sleepiness and alertness at 9:00 AM showed a significantly higher sleep pressure, Proposed mechanism for plasticity induction. The intracellular calcium concentration (x-axis) determines directionality of plasticity (Lisman, 2001). It can be assumed that intracellular calcium concentration under no sleep pressure (sufficient sleep) is at an optimal level leading to stronger tDCS-induced LTP/D-like plasticity. Under sleep deprivation, LTP-like plasticity induction via anodal tDCS is prevented due to calcium overflow, and LTD-like plasticity via cathodal tDCS is converted to LTP-like plasticity possibly via (a) enhanced baseline calcium (due to upscaled excitability) which makes minor calcium increase obtained from cathodal stimulation to be sufficient to induce LTP-like plasticity and (b) gradual downregulating upscaled cortical excitability and opening some synaptic space for LTP-related plasticity induction.  In the "sufficient sleep" condition, participants had to go to bed at around 23:00 and have at 495 least 8 hours of uninterrupted sleep. The experiment was scheduled to start at 9:00. with the Karolinska sleepiness scale (KSS) (Akerstedt and Gillberg, 1990) and the Stanford 514 Sleepiness Scale (SSS) (Hoddes et al., 1972). All external factors that could affect circadian 515 rhythmicity such as light and food intake were controlled during the experiment.  2003) and was determined as the lowest stimulator intensity required to evoke a peak-to-peak 544 MEP of 50 µV in the relaxed ADM muscle in at least five out of ten consecutive trials. The

545
AMT was determined as the lowest stimulator intensity required to elicit MEP response of 546 ∼200-300 μV during moderate tonic contraction of the right ADM muscle (∼20% of the 547 maximum muscle strength) (Rothwell et al., 1999) in at least three of six consecutive trials. amplitudes with increasing TMS intensity (Chen, 2000). The slope of the recruitment curve 552 increases at higher TMS intensities with higher glutamatergic and adrenergic transmission and 553 decreases by drugs that enhance effects of GABA (Chen, 2000;Paulus et al., 2008

559
The SICI-ICF is a TMS paired-pulse protocol for monitoring of GABAergic-mediated cortical 560 inhibition and the glutamate-mediated cortical facilitation (Chen, 2000). In this protocol, a 561 subthreshold conditioning stimulus (determined as 70% of AMT) is followed by a 562 suprathreshold test stimulus which was adjusted to evoke a baseline MEP of ∼1 mV. The amplitudes (Kujirai et al., 1993;Lazzaro et al., 1998;Lazzaro et al., 2003). The stimuli 568 (subthreshold and suprathreshold stimuli) were organized in blocks in which each ISI and one 569 single test stimulus were applied once in pseudorandomized order. Each block was repeated 570 15 times, which resulted in a total of 90 single-pulse or paired-pulse MEP per session. The 571 exact interval between the paired pulses was randomized (4 ± 0.4 s).  (Ziemann et al., 1998;Di Lazzaro et al., 576 2012)). In this protocol, two successive stimuli (supra-and subthreshold) are separated by 577 short ISIs, but this protocol involves a suprathreshold first stimulus and a subthreshold second 578 stimulus (Ziemann et al., 1998 peaks (Ziemann et al., 1998)   given medium TMS intensity) was identified with TMS and marked with a water-proof pen.

617
The stimulation intensity was then adjusted to evoke MEPs with a peak-to-peak amplitude of sham tDCS (Ambrus et al., 2012). TMS intensity was set to evoke MEPs of approximately 1-644 mV peak-to-peak amplitude and single-pulse MEPs were then obtained.  instructed to press a button with the right index finger whenever the letter A (correct cue) was 733 followed by the letter X (correct target) as quickly and accurately as possible. All other 734 sequences were to be ignored, including sequences in which an incorrect cue (designated 'B', 735 but comprising all letters other than A or X) was followed by the target letter (X), or 736 sequences in which a correct cue (A) was followed by an incorrect target (designated 'Y', but 737 comprising all letters other than A or X). The AX sequences are presented with a high 738 probability, to guarantee a strong response bias. The tasks consisted of 240 pairs of letters 739 (480 trials) with 40% "AX", 40% "BY", 10% "BX" and 10% of "AY". Accuracy and RT 740 were recorded for the target trials. to the head using high-viscosity electrolyte gel (SuperVisc, Easycap, Herrsching, Germany).

768
All impedances were kept below 10 kΩ throughout the experimental sessions. EEG data were 769 collected in a shielded room, and no spectral peaks at 50 Hz were observed. Raw EEG data   Means of RT, RT variability, and accuracy for SRTT blocks 5, 6, and 7 were calculated.

838
Trials with wrong responses, as well as those with RTs of less than 150 ms (Collins and Long,839 1996; Mella et al., 2015) or more than 3000 ms, and trials which deviated by 3 standard

2.2.Reported tDCS side effects 945
The reported side effects during each tDCS session (average ± SD) after sufficient sleep and 946 sleep deprivation are summarized in Table S4. The results of the 2 (group: anodal, cathodal) × 947 2 (sleep condition) 2 × (tDCS state: active, sham) factorial ANOVA conducted for each side 948 effect showed no interaction or main effects, except for a significant main effect of tDCS state 949 for itching, and tingling (Table S5)  To explore blinding efficacy we asked participants to guess whether they received real tDCS (1 971 mA) or sham tDCS (0 mA) after each stimulation condition across sleep conditions. Using the 972 Chi-square Test for Associations, we explored whether participants in each group (anodal,    (Fig. S1b). Furthermore, when every single block was compared across sleep 999 conditions, the number of committed errors was significantly higher at BL 4, 6, 7, and 8 after 1000 sleep deprivation (Fig. S1c). 1001 Fig. S1. The impact of sleep deprivation on motor learning performance. a, BL 5-6 absolute RT difference represents sequence learning and was significant only after sufficient sleep (t=2.78, p=0.005) but not sleep deprivation (t=1.47, p=0.141). The BL 6-7 RT difference represents learning retention and was again significant only after sufficient sleep (t=2.16, p=0.031) but not sleep deprivation (t=0.86, p=0.392). Asterisks [*] represent statistically significant differences between learning blocks RT (BL 6-5, BL 6-7]. The brackets refer to RT difference between blocks 6 vs 5 and 6 vs 7. b, After sleep deprivation, participants committed more errors at block 6 compared to block 5 (t=2.38, p=0.024) but not 7 (t=0.70, p=0.489). c, Block-specific error rate was however, significantly higher after sleep deprivation in BL 6 (t=3. 80,p<0.001),7 (t=3.12,p=0.004),and also BL 4 (t=2.41,p=0.022) and 8 (t=3.09,p=0.004), as compared to the sufficient sleep condition. d, Participants showed higher RT variability after sleep deprivation, in all learning block including block 5 (t=3.17, p=0.004), block 6 (t = 4.43, p<0.001), and block 7 (t=3.89, p<0.001). All pairwise comparisons are calculated using post hoc Student's t-tests (paired, p<0.05 Fig. S2. P300 amplitudes of electrodes C1, C2, P1, and P2 during motor sequence learning across sleep conditions. For electrode C1, there was no significant difference between learning blocks across sleep conditions. However, between-condition comparisons show significantly larger P300 amplitudes in all learning blocks after sufficient sleep compared to sleep deprivation. For electrode C2, Pairwise comparisons show a significantly larger P300 amplitude in block 6 compared to blocks 5 and 7 only after sufficient sleep as compared to sleep deprivation. Betweencondition comparisons of respective blocks show a significantly higher P300 amplitude at blocks 6 and 7 in the sufficient sleep condition. For electrode P1, there was no significant difference between individual learning blocks across sleep conditions (i.e., sufficient sleep vs sleep deprivation). However, within-condition comparisons show a significantly higher P300 amplitude in block 6 vs 5 and block 6 vs 7 only after sufficient sleep. The same pattern of response was found for electrode P2. All pairwise comparisons are calculated using post hoc t-tests (paired, p <0.05). n=30. [*/ns] indicates significant/non-significant differences between each block across sleep conditions. Filled symbols represent significant differences between BL 6-5 and BL 6-7. ns = nonsignificant. and congruent trials (t=2.18, p=0.037) after sleep deprivation compared to sufficient sleep. n=29. b, N200 and N450 ERP components of Stroop task performance across sleep conditions for electrode Cz. The N200 component was significantly larger for the incongruent trials, but not congruent trials, after sufficient sleep vs sleep deprivation for the electrodes Cz (t=3.51, p=0.002). The N450 did not significantly differ during incongruent vs congruent trials across sleep sessions. n=28. c, The P300 amplitude was significantly larger after sufficient sleep at electrodes F3 (t=2.18, p=0.038), F4 (t=2.66, p=0.013), C3 (t=4.93, p<.001), and C4 (t=2.75, p=0.011). The temporal window of 250-650 ms including the P300 amplitude (300-600 ms). n=27. Error bars represent s.e.m. ns = nonsignificant; Asterisks [*] indicate significant differences. All pairwise comparisons were calculated via post hoc Student's t-tests (paired, p<0.05).

2.2.Reported tDCS side effects
The reported side effects during each tDCS session (average ± SD) after sufficient sleep and sleep deprivation are summarized in Table S4. The results of the 2 (group: anodal, cathodal) × 2 (sleep condition) 2 × (tDCS state: active, sham) factorial ANOVA conducted for each side effect showed no interaction or main effects, except for a significant main effect of tDCS state for itching, and tingling (Table S5). Pairwise comparisons of itching and tingling ratings with post hoc t-tests revealed a significantly higher rating for itching sensation during anodal tDCS compared to the sham condition only after sleep deprivation (t = 2.41, p = 0.030). When the ratings were compared regardless of stimulation polarity (i.e., active tDCS vs sham tDCS), a significantly higher rating of the itching sensation was observed between active and sham tDCS after both sufficient sleep (t = 2.85, p = 0.008) and sleep deprivation (t = 3.06, p = 0.005). The intensity of the reported side effects was in general low. The presence and intensity of the side-effects were rated on a numerical scale ranging from zero to five, zero representing no and five extremely strong sensations. Data are presented as mean ± SD.

1.1.Error rate
The number of errors in the learning blocks in the respective sleep conditions was analyzed as well. The results of the 2×3 ANOVA showed a significant interaction of sleep condition×block (F1.96 = 8.49, p = 0.001, ηp 2 = 0.23) and a significant main effect of sleep condition (F1 = 6.49, p =0.016, ηp 2 = 0.18), but not block (F1.73 = 1.02, p = 0.365). Post hoc t-tests showed a significantly higher number of committed errors at block 6 compared to block 5 only after sleep deprivation (Fig. S1b). Furthermore, when every single block was compared across sleep conditions, the number of committed errors was significantly higher at BL 4, 6, 7, and 8 after sleep deprivation (Fig. S1c).
Furthermore, when every single block was compared across sleep conditions, RT variability was significantly higher at each block (BL 5, 6, 7) after sleep deprivation compared to sufficient sleep ( Fig. S1d).
For the electrode C2, the results of the ANOVA showed a significant interaction of sleep condition × block (F2=3.32, p=0.043; ηp 2 =0.10), and main effects of sleep condition (F1=5.02, p=0.033, ηp 2 =0.15) and block (F1.42=4.76, p=0.023, ηp 2 =0.14). Post hoc comparisons of learning block 8 showed a significantly higher P300 amplitude at block 6 compared to block 5 and 7 only after sufficient sleep (t6-5=2.74, p=0.010, t6-7=2.64, p=0.013) but not sleep deprivation (t6-5=0. 70, p=0.485, t6-7=1.92, p=0.064) . The P300 amplitudes were also significantly larger at blocks 6 and 7 after sufficient sleep compared to the sleep deprivation (tBL6=2. 23,p=0.034,tBL7=2.21,p=0.035) (Fig. S2). Finally, the results of the ANOVA conducted for the electrode P1 showed a significant main effect of learning blocks (F1.40=5.68, p=0.013, ηp 2 =0.16) but not sleep condition (F1=1.54, amplitude within and between conditions showed that the P300 amplitude was significantly larger at block 6 vs 5 (t=4.27, p<0.001) and 6 vs 7 (t=5.17, p<0.001) only after sufficient sleep (Fig S2).  S2. P300 amplitudes of electrodes C1, C2, P1, and P2 during motor sequence learning across sleep conditions. For electrode C1, there was no significant difference between learning blocks across sleep conditions. However, between-condition comparisons show significantly larger P300 amplitudes in all learning blocks after sufficient sleep compared to the sleep deprivation. For electrode C2, Pairwise comparisons show a significantly larger P300 amplitude in block 6 compared to blocks 5 and 7 only after sufficient sleep as compared to sleep deprivation. Between-condition comparisons of respective blocks show a significantly higher P300 amplitude at blocks 6 and 7 in the sufficient sleep condition. For electrode P1, there was no significant difference between individual learning blocks across sleep conditions (i.e., sufficient sleep vs sleep deprivation). However, within-condition comparisons show a significantly higher P300 amplitude in block 6 vs 5 and block 6 vs 7 only after sufficient sleep. The same pattern of response was found for electrode P2. All pairwise comparisons are calculated using post hoc t-tests (paired, p <0.05). n=30. [*/ns] indicates significant/non-significant differences between each block across sleep conditions. Filled symbols represent significant differences between BL 6-5 and BL 6-7. ns = nonsignificant.

Working memory and attention tasks
For working memory performance, we also calculated variability of RT at a secondary outcome measure. The result of the within-subject design ANOVA revealed a significant main effect of sleep conditions RT variability of hits (F1=4.78, p=0.037). Post hoc Student's t-tests showed a significantly enhanced WM performance with significantly more RT variability after sufficient sleep, which could be due to an accuracy-RT trade-off. In the Stroop task, we investigated performance accuracy and ERP components at electrode Cz as well. The results of respective ANOVAs showed a significant main effect of sleep condition on the overall accuracy of the Stroop stage (F1=6.32, p=0.018; ηp 2 =0.18), accuracy of congruent trials (F1=4.77, p=0.037; ηp 2 =0.14), and accuracy of incongruent trials (F1=5.03, p=0.029; ηp 2 =0.16). Post hoc comparisons of accuracy rate revealed that participants had a significantly higher number of accurate responses to trials in the Stroop stage as well as incongruent and congruent trials (Fig. S3a). For the electrode Cz, the results of the 2 (congruency)×2 (sleep condition) ANOVA showed only a significant main effect of sleep condition on the N200 (F1=9.03, p=0.006; ηp 2 =0.25) but not N450 component (Fig.   S3b). Similarly, post hoc Student´s t-tests indicated a significantly smaller N200 amplitude, for the incongruent trials only, for the Cz electrode after sleep deprivation as compared to sufficient sleep ( Fig. S3b). Finally for the AX-CPT task, we also analyzed ERP components at other potentially relevant electrodes (F3, F4, C3, C4), and a comparable main effect of sleep condition was found on P300 amplitude for electrodes F3 (F1 = 4.77, p = 0.038; ηp 2 = 0.15), F4 (F1 = 7.82, p = 0.011; ηp 2 = 0.21), C3 (F1 = 24.31, p < 0.001; ηp 2 = 0.48), and C4 (F1 = 7.60, p = 0.011; ηp 2 = 0.22). Post hoc Student´s t-tests indicated that sleep deprivation was related to a significantly smaller P300 amplitude in the F3, F4, C3 and C4 electrodes (Fig. S3c). .018), incongruent trials (t=2.30, p=0.029) and congruent trials (t=2.18, p=0.037) after sleep deprivation compared to sufficient sleep. n=29. b, N200 and N450 ERP components of Stroop task performance across sleep conditions for electrode Cz. The N200 component was significantly larger for the incongruent trials, but not congruent trials, after sufficient sleep vs sleep deprivation for the electrodes Cz (t=3.51, p=0.002). The N450 did not significantly differ during incongruent vs congruent trials across sleep session. n=28. c, The P300 amplitude was significantly larger after sufficient sleep at electrodes F3 (t=2.18, p=0.038), F4 (t=2.66, p=0.013), C3 (t=4.93, p<.001), and C4 (t=2.75, p=0.011). The temporal window of 250-650 ms including the P300 amplitude (300-600 ms). n=27. Error bars represent s.e.m. ns = nonsignificant; Asterisks [*] indicate significant differences. All pairwise comparisons were calculated via post hoc Student's t-tests (paired, p<0.05).

3.1.Correlation between sequence learning and plasticity induction
To explore the association between motor learning and plasticity, we calculated the correlation between the respective parameters (Pearson's correlation, two-tailed). We found a significant negative correlation between enhanced anodal LTP-like plasticity after sufficient sleep and enhanced motor learning (indicated by reduced RT at learning blocks). Specifically, MEP amplitude enhancement after anodal tDCS was negatively correlated with both sequence learning acquisition (block 6 -5 RT difference) (r=-0.558, p=0.031) and sequence learning retention (block 6 -7 RT difference) (r=-0.734, p=0.002). This indicated that LTP-like plasticity effects after sufficient sleep were associated with better sequence learning. No correlation was found between cathodal LTD-like plasticity and sequence learning.

3.2.Correlation between cortical excitability, working memory, and attention 11
To explore the association between physiological parameters of cortical excitability, and cognitive performance, we correlated performance in the 3-back letter task, Stroop test and AX-CPT with the respective cortical excitability results. In the 3-back letter task, enhanced d prime index (a measure of performance accuracy) was positively correlated with cortical facilitation measured by ICF at ISI of 15 ms (i.e., larger MEP at ICF) after having sufficient sleep (r= 0.425, p=0.019). Conversely, lower accurate response during sleep deprivation was negatively correlated with converted intracortical inhibition to facilitation (i.e., larger MEP amplitude) measured by SAI at ISI of 40 ms (r= -0.386, p=0.035). This indicates that upscaled cortical facilitation was associated with poor working memory performance.
No correlation was observed between Stropp task outcome measures and cortical excitability measures. For AX-CPT task performance, there was only a significant negative correlation between enhanced performance accuracy after sufficient sleep and reduced intracortical inhibition (measured by averaged MEPs of SICI) at the same time (r= -0.372, p=0.043). This indicates that improved task performance (i.e., higher accuracy) were associated with decreased intracortical inhibition after having sufficient sleep. In the sleep deprivation condition, lower performance accuracy and negatively correlated with higher corticospinal excitability (i.e., enhanced MEP at 150% of RMT intensity) (r= -0.429, p=0.018).