Cognitive functions and underlying parameters of human brain physiology are associated with chronotype

Circadian rhythms have natural relative variations among humans known as chronotype. Chronotype or being a morning or evening person, has a specific physiological, behavioural, and also genetic manifestation. Whether and how chronotype modulates human brain physiology and cognition is, however, not well understood. Here we examine how cortical excitability, neuroplasticity, and cognition are associated with chronotype in early and late chronotype individuals. We monitor motor cortical excitability, brain stimulation-induced neuroplasticity, and examine motor learning and cognitive functions at circadian-preferred and non-preferred times of day in 32 individuals. Motor learning and cognitive performance (working memory, and attention) along with their electrophysiological components are significantly enhanced at the circadian-preferred, compared to the non-preferred time. This outperformance is associated with enhanced cortical excitability (prominent cortical facilitation, diminished cortical inhibition), and long-term potentiation/depression-like plasticity. Our data show convergent findings of how chronotype can modulate human brain functions from basic physiological mechanisms to behaviour and higher-order cognition.

The reported side effects during each tDCS session for both ECs and LCs are summarized in Supplementary Table 4. The results of the ANOVA conducted for each side effect show a significant effect of tDCS on tingling and burning sensation during stimulation, but no significant effects for visual phenomena, itching, and pain (Supplementary Table 5). Pairwise comparisons of tingling ratings with post hoc t-tests revealed significant differences between morning anodal vs morning sham in ECs (t= 3.56, p= 0.003). In LCs, the tingling rating was significantly different only during evening cathodal vs evening sham (t= 2.17, p= 0.046). For the "burning" rating, a significant difference was found between evening anodal vs evening sham (t= 2.65, p= 0.018) in ECs, and a significant difference was found between evening cathodal vs evening sham (t= 2.36, p= 0.032) in LCs. The intensity of the reported side effects was in general low.

2.3.tDCS Blinding efficacy
To explore blinding efficacy, we asked participants to guess whether they received real tDCS (1 mA) or sham tDCS (0 mA) after each stimulation condition. Using the Chi-square Test for Associations, we explored whether participants could correctly discern each real stimulation condition (morning anodal, morning cathodal, evening anodal, evening cathodal) from its respective sham condition (morning sham, evening sham). The results of the respective Chi-square tests show no significant differences of participants' guesses between each real stimulation vs sham stimulation in both, ECs and LCs, and the whole group of participants (All) (Supplementary Table 6).

2.4.3D Modeling of the electrical current flow
Three-dimensional models of electrical current flow in the head induced by tDCS protocol (anodal M1, cathodal Fp2, 1 mA) applied to an adult head (New York (ICBM-NY) head 1 ). The MR images was first segmented into 6 tissue types: gray matter (GM), white matter (WM), CSF, skull, scalp, and air cavities using the SPM8 software package (Welcome Trust Center for Neuroimaging, London, UK) with an improved tissue probability map. A custom MATLAB script (MATLAB R2016b, MathWork Inc., Natick, MA, USA) was then used to correct for segmentation errors made by SPM 2 . Then, a 3D model of the segmented images, with addition of the electrodes and saline-soaked sponges, was designed using the Simpleware software package version 5 (Synopsys, Mountain View, CA). Finally, the current flow distribution inside the head was calculated based on the finite element method using COMSOL Multiphysics software package version 5.2 (COMSOL Inc., Burlington, MA). The conductivity values used for each tissue type were as follows (in S/m): GM: 0.276; WM: 0.126; CSF: 1.65; skull: 0.015; scalp: 0.465; air: 2.5 9 10-14; saline-soaked sponge: 1.5; electrode rubber: 29 3, 4 .

Implicit motor learning
To test if the learning sequence was preserved after the presentation of random stimuli in block 6, we analyzed RT differences of block 6 vs 7 too.  Fig. 1).

3.2.Error rate
We analyzed the number of errors in the learning blocks (block 5-7) to see if the error rate was affected by circadian preferred vs non-preferred times. The results of the 2 (daytime) × 2 (chronotype) × 3 (block 5-7) ANOVA show a significant interaction of chronotype×daytime  Fig. 2. Error rate and RT variability in the learning blocks for ECs and LCs. Mean error and RT variability were analyzed using a mixed-factorial (daytime×chronotype) ANOVA. a, In both ECs and LCs, participants significantly committed more errors in block 6 vs block 5 only at the circadian non-preferred time (ECsevening: p=0.032; LCsmorning: p=0.023). The error rate difference from block 6 to 7 was not significant in both groups. b, ECs and LCs displayed numerically higher RT variability in learning blocks at the circadian non-preferred time. The difference was, however, significant only for LCs in the morning (p=0.047). All pairwise comparisons are calculated by post hoc t-tests (paired, two-sided, p<0.05). n = 31 (15 ECs, 16 LCs). Data are presented as mean values±SEM. The horizontal bar shows the median, the + shows the mean, the upper and lower boundaries show the 25th and 75th percentiles, respectively and the whiskers show the 5-95 percentile. BL = block; ECs = early chronotypes; LCs = late chronotypes; ns = nonsignificant; RT = reaction time; SDRT = reaction time variability; ms = millisecond. Asterisks [*] indicate a significant difference.

1.1.RT variability
In addition to the RT, which is the primary outcome variable of interest in this task, we also analyzed RT variability of learning blocks. The results of the 2 (daytime) × 2 (chronotype) ×

SRTT
In addition to the Pz electrode, which is a major electrode of interest for the analysis of the P300 component, we analyzed other electrodes at temporal-parietal regions relevant for P300 activity, and identified a similar trend for chronotype-, and daytime-dependent alterations of the P300 amplitude at the P3 electrode. To test for statistical significance, we analyzed P-300 amplitudes (250-500 ms) of blocks (blocks 5-7), and amplitude differences at block 5 vs 6 (learning acquisition), and block 6 vs 7 (learning retention  Fig. 3a). Analysis of the P-300 amplitude differences (P300block6-P300block5) show a significant interaction of chronotype×daytime (F1=4.70, p=0.038; ηp 2 =0.14) in the P3 electrode, and no significant main effects of chronotype (F1=0.33, p=0.567) and daytime (F1=1.85, p=0.184). The P300 amplitude difference between Block 5 and 6 was significant only in ECs in the morning compared to the respective difference in the evening (t=2.63, p=0.009) (mean±SEMmorning, 0.73±0.23µV; mean±SEMevening, 0.06±0.19µV), but not in LCs (mean±SEMmorning, 0.20±0.09µV; mean±SEMevening, 0.36±0.23µV) (Supplementary Fig. 3b). The ANOVA results for Block 6-7 amplitude differences show no significant interaction of chronotype × daytime or main effects of chronotype and daytime. Supplementary Fig. 3: P300 amplitudes of electrode P3 during motor sequence learning for ECs and LCs. ERP amplitudes were analyzed using a mixed-factorial (daytime×chronotype) ANOVA. a P300 component was calculated per block in both groups. Pairwise comparisons show that only ECs displayed a significantly larger P300 at block 6 vs block 5 in the morning (p=0.039). b The P300 amplitude difference of learning blocks was calculated for each group at circadian preferred vs non-preferred times. Only ECs displayed a significantly larger P300 amplitude difference in the morning compared to the respective value in the evening (p=0.009). All pairwise comparisons are calculated by post hoc t-tests (paired, two-sided, p<0.05). n = 31 (15 ECs, 16 LCs). Data are presented as mean values±SEM. The horizontal bar shows the median, the + shows the mean, the upper and lower boundaries show the 25th and 75th percentiles, respectively and the whiskers show the 5-95 percentile. BL = block; ECs = early chronotypes; LCs = late chronotypes; ns = nonsignificant. [*] indicates a significant difference.

Working memory and attention tasks
Beyond Pz, also the contribution of Cz to the P300 and N200 components was evaluated  Fig. 4a,b). indicates higher negativity of N200 at circadian-preferred times in both groups. Supplementary Fig. 4: ERP components of working memory and attention task performance at electrode Cz across time of day for ECs and LCs. Behavioral and electrophysiological data were analyzed using a mixedfactorial (daytime×chronotype) ANOVA. a, b The P300 component was calculated for both groups at electrode Cz. ECs displayed a significantly larger P300 in the morning compared to the evening at electrode Cz (p=0.003). c, d, e The N200 component at electrode Cz was calculated for congruent (d) and incongruent (e) trials in each group at circadian preferred vs non-preferred times. Pairwise comparisons show that only ECs displayed a significantly larger N200 amplitude in the morning compared to the respective value in the evening for both congruent (p=0.002) and incongruent (p=0.021) trials. The amplitude difference of the N200 component (N200 morning -N200 evening) is, however, significant for both groups. f The N450 component was calculated for incongruent trials at electrode Fz in both groups. The N450 component is larger only for ECs in the morning. The amplitude difference of the N450 component (N450 morning -N450 evening) is however significant for both groups. All pairwise comparisons are calculated by post hoc t-tests (paired, two-sided, p<0.05). n=31 (15 ECs, 16 LCs). Data are presented as mean values±SEM. In a and c, the horizontal bar shows the median, the + shows the mean, the upper and lower boundaries show the 25th and 75th percentiles, respectively and the whiskers show the 5-95 percentile. BL = block; ECs = early chronotypes; LCs = late chronotypes; ns = nonsignificant; eve = evening; mo = morning; P3 = P300 component; N2 = N200; ms = millisecond. [*] indicates a significant difference.
In addition to N200, we analyzed the N450 which is a prominent ERP marker related to Stroop conflict, especially for incongruent trials and is usually observed at frontocentral and centroparietal electrode positions. The results of the mixed-model ANOVA ( Fig. 4f). The results of the separate 2 × 2 ANOVAs conducted for the congruent and incongruent trials showed a significant interaction of chronotype×daytime (F1=4.41, p=0.044; ηp 2 =0.16) only for incongruent trials. However, when we compared the amplitude difference values between morning to evening in a one-way ANOVA, chronotype had a significant effect on both, incongruent (F1=4.41, p=0.044; ηp 2 =0.13), and congruent (F1=4.80, p=0.037; ηp 2 =0.14) trials, indicating higher negativity of the N450 at the circadian-preferred times of both groups.
Analysis of the N200 amplitude of the electrode Cz showed no significant interaction of chronotype×daytime or main effects of chronotype and daytime on the N200 amplitude difference.

Correlation between sequence learning and plasticity induction
To explore the association between motor learning and plasticity, we calculated the correlation between the respective parameters. In LCs, anodal tDCS effects (MEP amplitude enhancement) were positively correlated with sequence learning (block 6 -5 RT difference) in the evening (r=0.543, p=0.017), indicating that larger LTP-like plasticity effects were associated with enhanced sequence learning. Furthermore, cathodal tDCS-induced LTD-like plasticity (MEP amplitude reduction) in the morning for LCs was positively correlated with sequence learning in the morning (r=0.441, p=0.043), which means that reduced LTD-like plasticity was associated with poor sequence learning at the circadian non-preferred time. In ECs, we did not see a significant correlation between sequence learning and tDCS-induced plasticity in the morning and evening.

Correlation between cortical excitability, working memory, and attention
The correlation between performance in the 3-back letter task, Stroop test and AX-CPT with the cortical excitability results was investigated. For ECs, RT of working memory performance in the evening was negatively correlated with cortical inhibition measured by I-wave facilitation (rISI-late= -0.453, p=0.039), indicating that performance was slower with increased cortical inhibition in the evening. Accuracy of working memory in the evening was positively correlated with cortical inhibition measured by I-wave facilitation protocol at early ISI (rISI-early= 0.500, p=0.024) which means that also accuracy decreased with increased inhibition (marked by larger MEP) in the evening. Also in ECs, accuracy of working memory in the morning was negatively correlated with cortical inhibition, as measured by I-wave facilitation at late ISI (rISI-late= -0.519, p=0.020), which means improved accuracy was associated with increased inhibition in the morning. In LCs, we found a significant positive correlation between working memory accuracy and cortical facilitation measured by ICF at ISI 10 ms in the evening (rISI10=0.561, p<0.012), which means that enhanced accuracy was associated with larger cortical facilitation.
In the Stroop test, which is a measure of selective attention, intracortical facilitation (ICF, ISI-15), was positively correlated with response accuracy in the evening for LCs (r=0.743, in the evening, which is indicative of less inhibition, and Stroop test RT were negatively correlated only for LCs in the evening (roverall trials= -0.487, p=0.028; rincongruent trials= -0.465, p=0.035; rcongruent trials= -0.507, p=0.023). This indicates that selective attention improved with reduced inhibition. In sum, better task performance in the evening for LCs was significantly associated with higher cortical facilitation and lower cortical inhibition. We found no correlation between corticocortical/corticospinal excitability and selective attention in ECs.
For AX-CPT task performance, accuracy was negatively correlated with cortical inhibition measured by the SAI protocol at ISI 20 ms (r= -0.452, p=0.039) in the morning for ECs, which indicates that improved accuracy was associated with increased cortical inhibition at the circadianpreferred time in this chronotype. In LCs, accuracy in the evening was positively (raccuracy= 0.521, p=0.019) correlated with cortico-spinal facilitation (I-O curve) at 110% RMT intensity, whereas RT was negatively (rRT= -0.467, p=0.031) correlated with the same protocol at 150% of RMT intensity. This indicates that improved task performance (i.e., higher accuracy and faster RT) were associated with increased cortico-spinal facilitation. Taken together, improved sustained attention at the circadian-preferred time, measured by higher accuracy of or shorter RT in AX-CPT performance, was associated with higher cortical inhibition in ECs and higher cortico-spinal facilitation in LCs.

Correlation between chronotype score, physiological parameters, and behavior
We also explored potential correlations between the participants' scores on DMEQ (i.e., circadian preference) and major outcome measures. No significant correlations were observed between DMEQ scores and cortical excitability except for intracortical inhibition (measured by SICI) of ECs obtained in the morning, which was positively correlated with DMEQ scores (r= 0.658, p=0.006). No significant correlations were neither found between DMEQ scores and neuroplasticity, or major outcome measures of the behavioral tasks.

Correlation between cognitive tasks
We also explored correlations of performance between the cognitive tasks. No significant correlations were observed between working memory and motor learning at the group level and in ECs alone. In LCs, however, a significant correlation was found between working memory RT in the evening and learning block of SRTT in the evening (rBL5= -0.513, p=0.042; rBL6= -0.614, p=0.011) indicating faster RT in both tasks in the evening were associated. For other tasks, no significant correlations were found.
Overall, these data show no robust correlations of performance between the different tasks.

Supplementary Tables
Supplementary Table 1    Note: The presence and intensity of tDCS 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. ECs = early chronotypes; LCs = late chronotypes.
Supplementary The presence and intensity of reported side-effects during tDCS were analyzed by repeated-measures mixedmodel ANOVAs with tDCS (6 values) as the within-subject and chronotype (ECs, LCs) as the between-subject factors. Significant effects of tingling and burning, but no significant effects of other side-effects, were revealed. Significant effects are marked in bold (where P < 0.05). Pairwise comparisons are calculated using Student's t-test (two-sided). n =32 (16 per group). Supplementary