Advertisement

Increased aperiodic neural activity during sleep in major depressive disorder

  • Yevgenia Rosenblum
    Correspondence
    Corresponding author: Yevgenia Rosenblum. Donders Institute for Brain, Cognition and Behavior. Kapittelweg 29, 6525 EN Nijmegen, Netherlands.
    Affiliations
    Radboud University, Donders Institute for Brain, Cognition and Behavior, Nijmegen, Netherlands

    Radboud University Medical Centre, Department of Cognitive Neuroscience, Nijmegen, Netherlands
    Search for articles by this author
  • Leonore Bovy
    Affiliations
    Radboud University, Donders Institute for Brain, Cognition and Behavior, Nijmegen, Netherlands

    Radboud University Medical Centre, Department of Cognitive Neuroscience, Nijmegen, Netherlands
    Search for articles by this author
  • Frederik D. Weber
    Affiliations
    Radboud University, Donders Institute for Brain, Cognition and Behavior, Nijmegen, Netherlands

    Radboud University Medical Centre, Department of Cognitive Neuroscience, Nijmegen, Netherlands

    Netherlands Institute for Neuroscience, Department of Sleep and Cognition, Amsterdam, Netherlands
    Search for articles by this author
  • Axel Steiger
    Affiliations
    Max Planck Institute of Psychiatry, Munich, Germany
    Search for articles by this author
  • Marcel Zeising
    Affiliations
    Klinikum Ingolstadt, Centre of Mental Health, Ingolstadt, Germany
    Search for articles by this author
  • Martin Dresler
    Affiliations
    Radboud University, Donders Institute for Brain, Cognition and Behavior, Nijmegen, Netherlands

    Radboud University Medical Centre, Department of Cognitive Neuroscience, Nijmegen, Netherlands
    Search for articles by this author
Open AccessPublished:October 25, 2022DOI:https://doi.org/10.1016/j.bpsgos.2022.10.001

      Abstract

      Background

      In major depressive disorder (MDD), patients often express subjective sleep complaints while polysomnographic studies report only subtle alterations of the electroencephalographic (EEG) signal. We hypothesize that differentiating the signal into its oscillatory and aperiodic components may bring new insights into our understanding of sleep abnormalities in MDD. Specifically, we investigate aperiodic neural activity during sleep and its relationships with the sleep architecture, depression severity, and responsivity to antidepressant treatment.

      Methods

      Polysomnography was recorded in 38 MDD patients (in unmedicated and 7-day medicated states) and 38 age-matched healthy controls (n=76). Aperiodic power component was calculated using the Irregularly Resampled Auto-Spectral Analysis. Depression severity was assessed with the Hamilton Depression Rating Scale. We replicated the analysis using two independently collected datasets of medicated patients and controls (n=60 and n=80).

      Results

      Unmedicated patients showed flatter aperiodic slopes compared to controls during non-REM 2 sleep (p=0.009). Medicated patients showed flatter aperiodic slopes compared to their later medicated state (p-values<0.001) and controls during all sleep stages (p-values<0.03). In medicated patients, flatter aperiodic slopes during non-REM sleep were linked to the higher proportion of non-REM 1, lower proportion of REM, delayed onset of non-REM-3 and REM, and shorter total sleep time.

      Conclusion

      Flatter slopes of aperiodic EEG power may reflect noisier neural activity due to increased excitation-to-inhibition balance, representing a new disease-relevant feature of sleep in MDD.

      Keywords

      Abbreviations:

      AUC (area under the curve), E/I (excitation-to-inhibition), HAM-D (Hamilton depression rating scale), MDD (major depressive disorder), NaSSA (noradrenergic and specific serotonergic antidepressants), NDRI (norepinephrine-dopamine reuptake inhibitor), REM (rapid eye movement), SNRI (serotonin-norepinephrine reuptake inhibitors), SSRI (selective serotonin reuptake inhibitors), TCA (tricyclic antidepressants), WASO (wakefulness after sleep onset)

      1. Introduction

      Major depressive disorder (MDD) is a common psychiatric disorder characterized by at least two weeks of pervasive low mood, anhedonia, inappropriate guilt, and feelings of worthlessness (

      American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.).

      ). In 2017, MDD affected ∼2% of the world population (

      World Health Organisation, WHO (2018) Fact sheet No 369: Depression. Available at: http://www.who.int/mediacentre/factsheets/fs369/en/. Accessed 21 August 2021

      ). The percentage of people who are affected at one point in their life varies from 7% to 21%, reflecting the fact that MDD is a serious public health problem (

      World Health Organisation, WHO (2018) Fact sheet No 369: Depression. Available at: http://www.who.int/mediacentre/factsheets/fs369/en/. Accessed 21 August 2021

      ). Besides abnormalities of mood and affect, MDD patients often have sleep complaints, including insomnia (in ∼60%) or hypersomnia (in ∼15%), as well as fatigue, excessive daytime sleepiness, and lack of concentration while awake (
      • Staner L.
      Comorbidity of insomnia and depression.
      ). Broad evidence suggests that disturbances of sleep-wake rhythms and circadian time-keeping system underlie the pathophysiology of depression (

      Courtet P, Olié E. Circadian dimension and severity of depression. European Neuropsychopharmacology. 2012 Jan 1;22:S476-S481. https://doi.org/10.1016/j.euroneuro.2012.07.009

      ). Understanding the mechanisms of these alterations might bring new insights into the understanding of MDD.
      Intriguingly, whereas some polysomnographic studies confirm subjective sleep complaints of the patients by reporting decreased slow-wave and delta amplitudes, higher spindle amplitude, lower spindle density, and a more dispersed slow-wave-spindle coupling, others suggest that oscillatory changes in MDD might be more subtle (
      • Armitage R.
      • Hoffmann R.
      • Trivedi M.
      • Rush A.J.
      Slow-wave activity in NREM sleep: sex and age effects in depressed outpatients and healthy controls.
      ,

      Bovy L, Weber FD, Tendolkar I, Fernández G, Czisch M, Steiger A, et al. Non-REM sleep in major depressive disorder. bioRxiv. 2021 Jan 1. https://doi.org/10.1101/2021.03.19.436132

      ). One of the possible explanations for the divergent oscillatory findings is the confounding effect of aperiodic (i.e., non-oscillatory, scale-free) activity (
      • Gerster M.
      • Waterstraat G.
      • Litvak V.
      • Lehnertz K.
      • Schnitzler A.
      • Florin E.
      • et al.
      Separating neural oscillations from aperiodic 1/f activity: challenges and recommendations.
      ,
      • Donoghue T.
      • Schaworonkow N.
      • Voytek B.
      Methodological considerations for studying neural oscillations.
      ,
      • Ouyang G.
      • Hildebrandt A.
      • Schmitz F.
      • Herrmann C.S.
      Decomposing alpha and 1/f brain activities reveals their differential associations with cognitive processing speed.
      ). For that reason, recently, it has been recommended to differentiate the total electroencephalographic (EEG) spectral power into its oscillatory and aperiodic components in order to "avoid misrepresentation and misinterpretation of the data" while studying oscillations (
      • Gerster M.
      • Waterstraat G.
      • Litvak V.
      • Lehnertz K.
      • Schnitzler A.
      • Florin E.
      • et al.
      Separating neural oscillations from aperiodic 1/f activity: challenges and recommendations.
      ,
      • Donoghue T.
      • Schaworonkow N.
      • Voytek B.
      Methodological considerations for studying neural oscillations.
      ,
      • Ouyang G.
      • Hildebrandt A.
      • Schmitz F.
      • Herrmann C.S.
      Decomposing alpha and 1/f brain activities reveals their differential associations with cognitive processing speed.
      ). In addition, exploring aperiodic activity is important per se as it is a distinct type of brain dynamics with its own functional significance and rich information content able to provide a window into diverse neural processes (
      • Donoghue T.
      • Schaworonkow N.
      • Voytek B.
      Methodological considerations for studying neural oscillations.
      ,
      • Ouyang G.
      • Hildebrandt A.
      • Schmitz F.
      • Herrmann C.S.
      Decomposing alpha and 1/f brain activities reveals their differential associations with cognitive processing speed.
      ,
      • Waschke L.
      • Donoghue T.
      • Fiedler L.
      • Smith S.
      • Garrett D.D.
      • Voytek B.
      • et al.
      Modality-specific tracking of attention and sensory statistics in the human electrophysiological spectral exponent.
      ).
      Currently, aperiodic activity receives increasing attention with reports on aperiodic changes associated with sleep phases, tasks, age, and disease (
      • Gerster M.
      • Waterstraat G.
      • Litvak V.
      • Lehnertz K.
      • Schnitzler A.
      • Florin E.
      • et al.
      Separating neural oscillations from aperiodic 1/f activity: challenges and recommendations.
      ,
      • Donoghue T.
      • Schaworonkow N.
      • Voytek B.
      Methodological considerations for studying neural oscillations.
      ,
      • Ouyang G.
      • Hildebrandt A.
      • Schmitz F.
      • Herrmann C.S.
      Decomposing alpha and 1/f brain activities reveals their differential associations with cognitive processing speed.
      ,
      • Waschke L.
      • Donoghue T.
      • Fiedler L.
      • Smith S.
      • Garrett D.D.
      • Voytek B.
      • et al.
      Modality-specific tracking of attention and sensory statistics in the human electrophysiological spectral exponent.
      ,
      • He B.J.
      Scale-free brain activity: past, present, and future.
      ,
      • Miller K.J.
      • Sorensen L.B.
      • Ojemann J.G.
      • Den Nijs M.
      Power-law scaling in the brain surface electric potential.
      ,
      • He B.J.
      • Zempel J.M.
      • Snyder A.Z.
      • Raichle M.E.
      The temporal structures and functional significance of scale-free brain activity.
      ,
      • Voytek B.
      • Knight R.T.
      Dynamic network communication as a unifying neural basis for cognition, development, aging, and disease.
      ,
      • Gao R.
      • Peterson E.J.
      • Voytek B.
      Inferring synaptic excitation/inhibition balance from field potentials.
      ,
      • Lendner J.D.
      • Helfrich R.F.
      • Mander B.A.
      • Romundstad L.
      • Lin J.J.
      • Walker M.P.
      • et al.
      An electrophysiological marker of arousal level in humans.
      ). Notably, it has been shown that the slope of the aperiodic component reflects the ratio between excitatory and inhibitory currents in the brain (
      • Waschke L.
      • Donoghue T.
      • Fiedler L.
      • Smith S.
      • Garrett D.D.
      • Voytek B.
      • et al.
      Modality-specific tracking of attention and sensory statistics in the human electrophysiological spectral exponent.
      ,
      • Gao R.
      • Peterson E.J.
      • Voytek B.
      Inferring synaptic excitation/inhibition balance from field potentials.
      ,
      • Lendner J.D.
      • Helfrich R.F.
      • Mander B.A.
      • Romundstad L.
      • Lin J.J.
      • Walker M.P.
      • et al.
      An electrophysiological marker of arousal level in humans.
      ), while the height of the spectra is related to neural spiking rates (
      • Donoghue T.
      • Schaworonkow N.
      • Voytek B.
      Methodological considerations for studying neural oscillations.
      ,
      • Waschke L.
      • Donoghue T.
      • Fiedler L.
      • Smith S.
      • Garrett D.D.
      • Voytek B.
      • et al.
      Modality-specific tracking of attention and sensory statistics in the human electrophysiological spectral exponent.
      ). Besides this, a steeper aperiodic spectrum can also reflect greater synchronization while a flatter spectrum can indicate reduced synchronization (i.e., greater neural noise; 14). In view of the crucial role of the proper balance between neural excitation and inhibition (E/I) for healthy cognition, behavior (
      • Foss-Feig J.H.
      • Adkinson B.D.
      • Ji J.L.
      • Yang G.
      • Srihari V.H.
      • McPartland J.C.
      • et al.
      Searching for cross-diagnostic convergence: neural mechanisms governing excitation and inhibition balance in schizophrenia and autism spectrum disorders.
      ), and sleep, aperiodic activity seems to be a promising tool for investigating MDD with its cholinergic, monoaminergic (
      • Janowsky D.
      • Davis J.
      • El-Yousef M.K.
      • Sekerke H.J.
      A cholinergic-adrenergic hypothesis of mania and depression.
      ,
      • Wichniak A.
      • Wierzbicka A.
      • Jernajczyk W.
      Sleep as a biomarker for depression.
      ), glutamatergic (
      • Moriguchi S.
      • Takamiya A.
      • Noda Y.
      • Horita N.
      • Wada M.
      • Tsugawa S.
      • et al.
      Glutamatergic neurometabolite levels in major depressive disorder: a systematic review and meta-analysis of proton magnetic resonance spectroscopy studies.
      ), and GABAergic imbalance (
      • Luscher B.
      • Shen Q.
      • Sahir N.
      The GABAergic deficit hypothesis of major depressive disorder.
      ,

      Mazza F, Griffiths J, Hay E. Biomarkers of reduced inhibition in human cortical microcircuit signals in depression. In JOURNAL OF COMPUTATIONAL NEUROSCIENCE 2021 Dec 1 (Vol. 49, No. SUPPL 1, pp. S86-S86). VAN GODEWIJCKSTRAAT 30, 3311 GZ DORDRECHT, NETHERLANDS: SPRINGER.

      ). In MDD, the E/I ratio could be further affected by prescribed antidepressants.
      In view of this background, here, we explore aperiodic activity during sleep in MDD and its relationships with the sleep architecture, depression severity, and responsivity to antidepressant treatment. This study has an exploratory nature with no a priori hypothesis on the direction of aperiodic changes.

      2. Methods and Materials

      Participants

      We retrospectively analyzed polysomnographic recordings from a previous study conducted at the Max Planck Institute of Psychiatry, Munich, Germany (

      Bovy L, Weber FD, Tendolkar I, Fernández G, Czisch M, Steiger A, et al. Non-REM sleep in major depressive disorder. bioRxiv. 2021 Jan 1. https://doi.org/10.1101/2021.03.19.436132

      ). The sample consisted of 40 patients with MDD and 40 healthy controls individually matched by age (±2 years of tolerance) and gender (Table 1). None of the patients was treated with sedative antidepressants.
      Table 1Demographic, clinical, and sleep characteristics of the participants (mean ± SD)
      GroupPatientsControls
      CharacteristicUnmedicated7-day medicated
      No. of participants38---38
      Age, years31.3±10.2---31.6±10.4
      Gender ratio, F/M18/20---21/17
      HAM-D19.9±3.815.2±4.8---
      No. of previous episodes1.76±3.0------
      Non-REM-1, min (%)52.38 (12.4%)62.53 (15.0%)52.34 (12.4%)
      Non-REM-2, min (%)186.51 (44.1%)201.03 (48.0%)195.11 (46.2%)
      Non-REM-3, min (%)76.64 (18.4%)70.74 (17.2%)82.96 (19.9%)
      REM sleep, min (%)68.93 (16.3%)49.54 (11.8%)bc69.34 (16.3%)
      WASO37.83 (8.8%)a33.28 (8.0%)b21.82 (5.2%)
      Total non-REM time, min263.16 (68.2%)271.76 (70.6%)278.07 (69.7%)
      Total sleep time, min384.47±40.0383.83±38.8399.75±39.4
      Sleep onset, min24.1±30.820.9±13.815.7±9.6
      Non-REM-3 onset, min24.91±24.024.14±19.818.21±8.4
      REM onset, min98.61±52.6157.63±79.6bc85.41±34.8
      Sleep stage percentages are given with respect to total sleep time. Non-REM sleep was defined as the combination of non-REM-2 and non-REM-3 without non-REM-1 sleep. SD – standard deviation, REM – rapid eye movement sleep, WASO – wakefulness after sleep onset, HAM-D – Hamilton Depression Rating Scale, a – significant difference between controls and unmedicated patients, b – significant difference between controls and medicated patients, c – significant difference between unmedicated and medicated states of the patients.
      Exclusion criteria included suicidality, shift working, transmeridian flights in the preceding three months, drug or alcohol dependence, professional piano skills, professional typewriting skills, sleep disorders, pregnancy, and a history of severe physical disorders. Subjects who received long-acting medication before the beginning of the experiment were excluded if the treatment was not stopped in time to ensure a complete wash-out (e.g., antipsychotics, fluoxetine). Due to technical failure in the EEG data of two medicated patients, all paired analyses were matched on the remaining full datasets (n=38 per group).
      To confirm the results, we replicated the analyses using two independently collected datasets of short and long-term medicated MDD patients (Supplementary Material-5). All studies were approved by the Ethics committee of the University of Munich. All participants gave written informed consent.

      Questionnaires

      Depression severity of patients was measured with the Hamilton Depression Rating Scale (HAM-D) at baseline ("unmedicated") and 7 days after the commencement of antidepressant treatment ("medicated"). A higher score reflects higher depression severity. In Supplementary Material-4, we also report the Pittsburgh Sleep Quality Index, which was available in a subset of the patients.

      Polysomnography

      All participants slept in the sleep laboratory, and all had an adaptation night before the examination night. For the EEG of the examination night, 118 Ag/AgCl electrodes were applied using an Easycap 128Ch-BrainCap (EasyCap GmbH, Herrsching, Germany). Polysomnography was recorded (sampling rate of 200 Hz), stored, and digitized following the 10-5 system (
      • Oostenveld R.
      • Praamstra P.
      The five percent electrode system for high-resolution EEG and ERP measurements.
      ) with a JE-209A amplifier (Neurofax Software, Nihon Kohden Europe GmbH, Rosbach, Germany) with a common-mode rejection ratio of ≥ 110 dB and with impedances below 10 kOhm, including EEG (filtered at 0.016 Hz high pass only, -6 dB/octave), electrooculography, mental/submental electromyography with a ground electrode attached at the forehead. During the recording, the EEG was referenced to the average of the AFF5H and AFF1H, which were predefined by the hardware setup. For the offline analysis, the data was re-referenced to the average of all electrodes.
      Polysomnography of the patients was recorded at two timepoints: when unmedicated and when 7-day medicated. Sleep was scored by independent experts according to the AASM standards (

      Iber C. The AASM manual for the scoring of sleep and associated events: Rules. Terminology and technical specification. Published online 2007.

      ). We analyzed separately all sleep stages and the wakefulness occurring after sleep onset (WASO). The epochs scored as the “wake” before the sleep onset and after morning awakening were excluded from the analysis as they were not available for all participants. Epochs with EMG and EEG artifacts and channels with more than 20% artifacts during non-REM sleep were manually excluded by an experienced scorer before all automatic analyses. The rejection percentage is reported in Supplementary Table S6.1.
      In Supplementary Material S-3, we also report morning resting-state EEG measured in a subset of participants to explore whether the observed effects are specific to sleep.

      Spectral power

      Total EEG power was differentiated into its aperiodic (fractal) and oscillatory components using the Irregularly Resampled Auto-Spectral Analysis (
      • Wen H.
      • Liu Z.
      Separating fractal and oscillatory components in the power spectrum of neurophysiological signal.
      ). A MATLAB implementation of the algorithm was adapted from the Fieldtrip website (http://www.fieldtriptoolbox.org/example/irasa). Specifically, we used the ft_freqanalysis function with the cfg.method=’irasa’ for each 30 s of sleep, corresponding to the conventionally defined sleep epochs. The function was called twice, with the cfg.output=’fractal’ and cfg.output=’original’ for the total power and its aperiodic component, respectively. The aperiodic component was transformed to log-log coordinates by standard least squares regression, where the slope of the line was calculated as the power-law exponent estimation.
      Power was averaged over each sleep stage as defined by the hypnogram over five topographical areas: 1) frontal (Fz, F1, F2, F3, F4, F5, F6, F7, F8, F9, F10); 2) central (Cz, C1, C2, C3, C4, C5, C6); 3) parietal (Pz, P1, P2, P3, P4, P5, P6, P7, P8, P9, P10); 4) occipital (Oz, O1, O2); 5) temporal (T7, T8).
      The signal was filtered in the 0.2–48Hz frequency band. In Supplementary Material-1, we also analyze low (2–20Hz) and high (30–48Hz) bands to control for a possible distortion of the linear fit by excluding low frequencies with strong oscillatory activity (
      • Gao R.
      • Peterson E.J.
      • Voytek B.
      Inferring synaptic excitation/inhibition balance from field potentials.
      ) and for the reliable discrimination between wakefulness and REM sleep, respectively (
      • Lendner J.D.
      • Helfrich R.F.
      • Mander B.A.
      • Romundstad L.
      • Lin J.J.
      • Walker M.P.
      • et al.
      An electrophysiological marker of arousal level in humans.
      ).
      In Supplementary Material-2, we report the analysis of the oscillatory component to explore whether the effect is specific to aperiodic activity.

      Statistical analysis

      To analyze aperiodic slopes, we used five ANCOVAs for each sleep stage separately with the five-level "brain area" as within-subject factor and two-level "study group" as between-subjects factor to compare 1) unmedicated patients and controls; 2) the same patients when 7-day medicated and controls. Even though we matched the participants' age individually, given that at the group level the age ranged from 19 to 54 years, we added to the analysis the "age" factor as a covariate. In view of between-group differences in the proportions of sleep stages (Table 1), when appropriate (namely, for WASO and REM), we added to the ANCOVAs the proportion of a given sleep stage in each study group as an additional covariate. We performed five additional ANOVAs for each sleep stage separately to compare unmedicated and 7-day medicated states of the patients using the "state" as the within-subject factor.
      The Benjamini-Hochberg's adjustment was applied to control for multiple comparisons (5 tests reflecting the number of sleep stages) with a false discovery rate set at 0.05 and the α-level set in the 0.01–0.05 range. For all ANOVAs/ANCOVAs we applied Greenhouse-Geisser correction since Mauchly's test revealed that the sphericity assumption was violated (ε<0.75, p<0.05). The assumptions of normality and homogeneity of variance were tested using the Q-Q plot and Levene's homogeneity test, respectively.
      Then, we performed post hoc analysis to compare each pair of groups for each area and sleep stage separately. We used the two-tailed Student's unpaired t-test to compare patients to controls and paired t-test to compare the unmedicated and medicated states of the patients. Effect sizes were calculated with Cohen's d.
      To study the effect of the antidepressant treatment on aperiodic activity we stratified the patients by 1) antidepressant class; 2) REM-suppressive vs REM-non-suppressive antidepressants as reported in Table 2. Then, we performed 25 non-parametric Mann-Whitney U two-tailed tests for each sleep stage and area separately, since after this stratification the samples were too small to perform ANOVA. Benjamini-Hochberg's adjustment for 25 tests (5 stages by 5 areas) was applied with the α-level set in the 0.002–0.050 range.
      Table 2Demographic and clinical characteristics of the subgroups of patients by medication class (mean ± SD)
      Medication classSample sizeAgeGenderNo. of previous episodesHAM-D baselineHAM-D 7-day
      SSRI (citalopram, escitalopram, paroxetine, sertraline)1329.9±10.08 F0.6±0.819.7±4.213.9±4.6
      TCA (trimipramine, amitriptyline, amitriptylinoxide)836.6±11.94 F2.1±1.122.1±3.416.6±5.5
      NDRI (bupropion)630.7±10.53 F0.7±0.518.5±3. 517.8±3.2
      SNRI (venlafaxine, duloxetine)631.7±10.92 F1.7±0.8318.3±2.514.7±5.5
      NaSSA (mirtazapine)526.8±6.13 F2.6±3.220.2±4.813.8±4.7
      REM suppressive (SSRI, SNRI, amitriptylin, amitriptylinoxide)2131.1±10.311 F1.1±1.019.2±3.614.1±4.5
      REM non-suppressive1731.6±10.47 F1.7±2.020.7±4.116.6±4.9
      The NaSSA subgroup was not analyzed separately due to a small sample size. HAM-D – Hamilton Depression Rating Scale, REM – rapid eye movement sleep, SD – standard deviation, NaSSA – noradrenergic and specific serotonergic antidepressants, NDRI – norepinephrine-dopamine reuptake inhibitor, SNRI – serotonin-norepinephrine reuptake inhibitors, SSRI – selective serotonin reuptake inhibitors, TCA – tricyclic antidepressants.
      The diagnostic accuracy of the frontal aperiodic slopes was defined using the area under the Receiver Operating Characteristic curve (AUC). Pearson’s correlations were used to assess associations between the frontal aperiodic slopes on one side and 1) features of sleep architecture; 2) HAM-D scores of the participants at baseline and 7-day on the other side. SPSS software (version 25; SPSS, Inc) was used for all statistical analyses.

      3. Results

      The demographic, clinical, and sleep characteristics of the participants are reported in Table 1. Patients in both the unmedicated and medicated states showed increased WASO compared to controls. Medicated patients further showed decreased REM sleep proportion and prolonged REM sleep onset compared to the controls and own unmedicated state.

      Aperiodic slopes

      Frontal total spectral power and its components for all sleep stages for all study groups are shown in Fig.1. The results are presented in Table 3 and Fig.2.
      Figure thumbnail gr1
      Figure 1EEG power components. Total EEG spectral power (left), its aperiodic (right), and oscillatory (middle) components averaged over frontal electrodes are plotted in the log-log space as a function of frequency for non-REM (N2 + N3, first row) and REM (second row) sleep for each study group (different lines). Patients in both unmedicated (red lines) and medicated states (blue lines) show decreased oscillatory activity and steeper decay of the aperiodic component compared to controls (black lines). The total spectral power is comparable in all groups (coinciding lines of the left subgraphs).
      Table 3Aperiodic slopes
      GroupUnmedicated MDD7-day medicated MDDHC
      MeanArea/StageFCPOTFCPOTFCPOT
      Wake-2.43-2.20-2.46-2.50-1.89-2.44-2.14-2.45-2.45-1.77-2.49-2.20-2.44-2.50-1.89
      N1-2.84-2.83-2.84-2.95-2.65-2.79-2.75-2.81-2.91-2.50-2.98-2.88-2.92-3.06-2.78
      N2-3.22-3.23-3.19-3.28-3.12-3.17-3.17-3.16-3.25-2.95-3.35-3.31-3.28-3.38-3.25
      N3-3.63-3.63-3.62-3.67-3.55-3.57-3.55-3.57-3.63-3.35-3.70-3.67-3.66-3.71-3.64
      REM-2.84-2.86-2.88-2.95-2.88-2.76-2.77-2.79-2.87-2.79-2.96-2.88-2.93-3.05-2.97
      ComparisonUnmedicated MDD-HCMedicated MDD-HCUnmedicated-medicated MDD
      Effect sizeArea/StageFCPOTFCPOTFCPOT
      Wake0.14-0.01-0.040.020.000.110.14-0.030.120.20-0.030.140.010.130.24
      N10.460.290.440.400.370.620.710.640.500.740.390.480.310.290.47
      N20.640.620.660.460.500.790.860.810.580.890.350.420.270.290.51
      N30.580.310.310.340.500.860.730.610.590.840.370.490.400.290.45
      REM0.370.100.310.310.260.570.780.780.540.500.841.181.041.150.57
      ComparisonUnmedicated MDD-HCMedicated MDD-HCUnmedicated-medicated MDD
      FactorGroupAreaInteractionGroupAreaInteractionGroupAreaInteraction
      FPFpFpFpFpFpFpFpFp
      ANOVAWakeˆ0.20.6728.60.000.30.780.90.3425.70.001.80.141.00.32131.20.002.50.07
      N13.40.075.70.011.30.309.20.004.70.013.80.037.90.0118.10.006.10.01
      N27.20.013.80.031.20.3014.40.003.50.036.50.006.60.0254.90.008.00.00
      N34.30.041.40.252.30.1014.90.000.50.608.80.007.90.0018.10.006.10.01
      REMˆ1.80.180.90.501.50.235.10.031.00.401.80.1838.90.0031.60.0020.60.00
      Bold font indicates statistically significant p-values after the correction for multiple comparisons, effect sizes are interpreted as small (0.2–0.5), medium (0.5–0.8), and large (0.8–1.2). ˆ – ANCOVAs adjusted for the proportion of the corresponding sleep stage, MDD – major depressive disorder, HC – healthy controls, F – frontal, C – central, P – parietal, O – occipital, T – temporal electrodes, REM – rapid eye movement sleep, N – non–rapid eye movement sleep.
      Figure thumbnail gr2
      Figure 2Aperiodic slopes. Slopes of the broadband (0.2–48 Hz) aperiodic power component over each sleep stage and area for each study group. Unmedicated patients (red) show flatter (more positive values) slopes during N2 compared to controls (black). 7-day medicated patients (blue) show flatter slopes compared to their own unmedicated state (red) and controls (black) during all sleep stages – but not the wakefulness after sleep onset. MDD – 38 major depressive disorder patients, unmed. – unmedicated, med. – 7-day medicated MDD patients, HC – 38 healthy controls, F – frontal, C – central, P – parietal, O – occipital, T – temporal electrodes.
      Unmedicated patients showed flatter slopes compared to controls during N2 (p=0.009) in all areas and during N3 sleep (p=0.04) in the frontal and temporal areas with medium effect sizes.
      Medicated patients showed flatter slopes compared to controls during N1, N2, N3, and REM sleep in all areas with medium effect sizes. These findings were replicated in two independent datasets (Supplementary Material-5).
      Patients in the medicated state showed flatter slopes compared to the own unmedicated state with large effect size during REM sleep and with small effect size during N3. Likewise, the medicated state showed flatter slopes during N1 and N2 in the frontal, central, and temporal areas with small effect sizes compared to the unmedicated state (Fig.2, Table 3).
      Aperiodic slopes did not correlate with depression severity (HAM-D scores) neither at baseline nor at 7-day assessments.

      ROC analysis

      Frontal slopes measured during N1, N2, N3, and REM sleep discriminated both between the unmedicated state of the patients and controls (AUC=0.66–0.74, all p-values<0.02) as well as between the medicated state of the patients and controls (AUC=0.74–0.76, all p-values<0.001).

      Medication effect

      The demographic and clinical characteristics of the subgroups of patients stratified by the medication class are reported in Table 2. The results are presented in Fig.3 and Supplementary Material-5 (Table S5.5).
      Figure thumbnail gr3
      Figure 3Effect of REM-suppressive medications. Slopes of the aperiodic power component in the 0.2-48Hz frequency band were averaged over each sleep stage over each area. Patients who take REM-suppressive antidepressants for 7-day (red, n=21) show flatter slopes (higher values) than patients who take REM non-suppressive antidepressants for 7-day (black, n=17) during all sleep stages – but not the wakefulness after sleep onset. F – frontal, C – central, P – parietal, O – occipital, T – temporal electrodes.
      The patients who took REM-suppressive antidepressants showed flatter slopes compared to patients who took REM-non-suppressive antidepressants during N1 (F=13.9, p=0.001), N2 (F=42.8, p<0.001), N3 (F=21.9, p<0.001), and REM sleep (F=10.4, p=0.003) with large effect sizes (d-values=0.6–2.2).
      The patients who took REM-suppressive SNRIs showed flatter slopes during N2, N3, and REM sleep with large to huge effect sizes compared to the patients who took non-SNRIs (p-values<0.001–0.026, d-values=0.9–2.8). These patients also showed flatter slopes compared to the patients who took non-SNRIs REM-suppressive antidepressants (p-values=0.005–0.019, d-values=1.5–1.7), such as SSRIs (p-values=0.009–0.028, d-values=1.3–1.8) during REM sleep with very large effect sizes. However, these findings did not pass the correction for multiple comparisons. Aperiodic activity was comparable among patients who took SSRIs, TCAs, or NDRIs compared to their pooled controls.

      Aperiodic slopes and sleep architecture

      Correlations between aperiodic slopes and sleep architecture are reported in Table 4. In unmedicated patients, flatter aperiodic slopes during N1 were associated with the delayed onset of N3. In medicated patients, flatter aperiodic slopes during non-REM sleep were linked to the higher proportion of N1, lower proportion of REM, delayed onset of N3 and REM, and shorter total sleep time. In controls, flatter aperiodic slopes during N3 sleep were associated with the higher proportion of WASO and lower proportion of REM sleep. Some of these findings were replicated in an independent dataset (Supplementary Material-5, Table S5.2).
      Table 4Correlations between aperiodic slopes and sleep architecture (r)
      GroupPatientsControls
      Pair for correlationUnmedicatedMedicated
      N1 slope – N1%n.s.0.347*n.s.
      N1 slope – SWS onset0.335*0.351*n.s.
      N2 slope – N1%n.s.0.479**n.s.
      N2 slope – REM%n.s.-0.548**n.s.
      N2 slope – REM onsetn.s.0.391*n.s.
      N2 slope – TSTn.s.-0.328*n.s.
      N3 slope – WASOn.s.n.s.0.328*
      N3 slope – N1%n.s.0.455**n.s.
      N3 slope – REM%n.s.-0.357*-0.591**
      Pearson correlations coefficients between aperiodic slopes measured during a particular sleep stage and features of sleep architecture are presented for each group separately. Only the r’s associated with the statistically significant p-values are presented, i.e., the rest of the possible combinations between aperiodic slopes and sleep architecture features were statistically non-significant, * – 0.05 > p > 0.01, ** – p<0.01. Sleep stage percentages were calculated with respect to TST, TST – total sleep time, REM – rapid eye movement sleep, N – non-rapid eye movement sleep, WASO – wakefulness after sleep onset, r – Pearson correlation coefficient.
      The reported in the Results alterations were specific to sleep and were not observed during the morning resting state (Supplementary Material-3).

      4. Discussion

      To the best of our knowledge, this is the first study that examines sleep-related aperiodic activity in MDD and its relationships with the sleep architecture, depression severity, and responsivity to antidepressant treatment. We found that unmedicated patients show flatter aperiodic slopes during non-REM sleep compared to controls. Patients in the medicated state show flatter aperiodic slopes compared to the own unmedicated state and healthy controls during both non-REM and REM sleep. In medicated patients, flatter aperiodic slopes during non-REM sleep were linked to the lower proportion and delayed onset of REM sleep. We replicated several of our findings in two independently collected datasets of medicated patients. Below we aim at an interpretation of these findings.
      The functional significance of aperiodic dynamics is still a mystery with several interpretations suggested so far. For example, aperiodic activity can manifest in the overall firing rate of cortical neurons (
      • Ouyang G.
      • Hildebrandt A.
      • Schmitz F.
      • Herrmann C.S.
      Decomposing alpha and 1/f brain activities reveals their differential associations with cognitive processing speed.
      ,
      • Gao R.
      • Peterson E.J.
      • Voytek B.
      Inferring synaptic excitation/inhibition balance from field potentials.
      ,

      Buzsáki G, Anastassiou CA, Koch C. The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nature reviews neuroscience. 2012 Jun;13(6):407-420.

      ) as measured by local field potentials or EEG. When many neurons fire relatively simultaneously, the power spectrum will decay faster, being relatively stronger in low frequencies and relatively weaker in the higher ones. Mathematically, this will be expressed by a more negative (steeper) slope, which, in turn, reflects a higher power-law exponent. A steeper slope can signify redundancy (
      • He B.J.
      Scale-free brain activity: past, present, and future.
      ), excessive or insufficient propagation of the signal (
      • Tagliazucchi E.
      • van Someren E.J.
      The large-scale functional connectivity correlates of consciousness and arousal during the healthy and pathological human sleep cycle.
      ), or increased dendritic filtering (

      Buzsáki G, Anastassiou CA, Koch C. The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nature reviews neuroscience. 2012 Jun;13(6):407-420.

      ). When neurons fire relatively asynchronously, the spectral power is shifted towards higher frequencies and its slope is flatter, reflecting reduced temporal autocorrelations (
      • Donoghue T.
      • Schaworonkow N.
      • Voytek B.
      Methodological considerations for studying neural oscillations.
      ), high entropy rate of cortical systems (

      Medel V, Irani M, Ossandon T, Boncompte G. Complexity and 1/f slope jointly reflect cortical states across different E/I balances. bioRxiv. 2020 Jan 1. https://doi.org/10.1101/2020.09.15.298497

      ), or a noisier neural background (
      • Li S.C.
      • Sikström S.
      Integrative neurocomputational perspectives on cognitive aging, neuromodulation, and representation.
      ,
      • Voytek B.
      • Kramer M.A.
      • Case J.
      • Lepage K.Q.
      • Tempesta Z.R.
      • Knight R.T.
      • et al.
      Age-related changes in 1/f neural electrophysiological noise.
      ,

      Miskovic V, MacDonald KJ, Rhodes LJ, Cote KA. Changes in EEG multiscale entropy and power‐law frequency scaling during the human sleep cycle. Human Brain Mapp. 2019 Feb 1;40(2):538-551. https://doi.org/10.1002/hbm.24393

      ). Following this, flatter aperiodic slopes observed here in unmedicated and medicated MDD patients may reflect noisier neural background activity (

      Miskovic V, MacDonald KJ, Rhodes LJ, Cote KA. Changes in EEG multiscale entropy and power‐law frequency scaling during the human sleep cycle. Human Brain Mapp. 2019 Feb 1;40(2):538-551. https://doi.org/10.1002/hbm.24393

      ), which in turn can adversely affect sleep and its restorative function.
      Adding to this, pharmacological, physiological, and computational studies linked aperiodic activity to the balance between excitatory and inhibitory currents in the brain (
      • Gao R.
      • Peterson E.J.
      • Voytek B.
      Inferring synaptic excitation/inhibition balance from field potentials.
      ,
      • Lendner J.D.
      • Helfrich R.F.
      • Mander B.A.
      • Romundstad L.
      • Lin J.J.
      • Walker M.P.
      • et al.
      An electrophysiological marker of arousal level in humans.
      ). Specifically, Gao et al. (
      • Gao R.
      • Peterson E.J.
      • Voytek B.
      Inferring synaptic excitation/inhibition balance from field potentials.
      ) showed that the aperiodic slope in the 30-50Hz range reliably tracked the induction and the recovery from propofol-induced anesthesia in rats and macaques. Similarly, in humans, inhibition was boosted by the propofol administration, and the slope became steeper when inhibition increased (
      • Lendner J.D.
      • Helfrich R.F.
      • Mander B.A.
      • Romundstad L.
      • Lin J.J.
      • Walker M.P.
      • et al.
      An electrophysiological marker of arousal level in humans.
      ). Subsequent studies interpreted the 1/f exponent as an indicator of E/I balance also for other frequency bands, e.g., 3-55Hz (
      • Waschke L.
      • Donoghue T.
      • Fiedler L.
      • Smith S.
      • Garrett D.D.
      • Voytek B.
      • et al.
      Modality-specific tracking of attention and sensory statistics in the human electrophysiological spectral exponent.
      ), 0.5-35Hz (

      Miskovic V, MacDonald KJ, Rhodes LJ, Cote KA. Changes in EEG multiscale entropy and power‐law frequency scaling during the human sleep cycle. Human Brain Mapp. 2019 Feb 1;40(2):538-551. https://doi.org/10.1002/hbm.24393

      ), 1-40Hz, 1-20Hz, 20-40Hz (
      • Colombo M.A.
      • Napolitani M.
      • Boly M.
      • Gosseries O.
      • Casarotto S.
      • Rosanova M.
      • et al.
      The spectral exponent of the resting EEG indexes the presence of consciousness during unresponsiveness induced by propofol, xenon, and ketamine.
      ). In Supplementary Material-1, we analyzed aperiodic activity in the 2-20Hz and 30-48Hz bands and confirmed the broadband (0.2-48Hz) analysis reported in the main text with the exception of the high-band activity, which was comparable in unmedicated patients and controls.
      The right balance between neural excitation and inhibition is crucial for optimal signal formation and transmission, synaptic plasticity, neuronal growth, and pruning and, thus, enables flexible behavior and cognition (
      • Foss-Feig J.H.
      • Adkinson B.D.
      • Ji J.L.
      • Yang G.
      • Srihari V.H.
      • McPartland J.C.
      • et al.
      Searching for cross-diagnostic convergence: neural mechanisms governing excitation and inhibition balance in schizophrenia and autism spectrum disorders.
      ). Correspondingly, any perturbations in the E/I balance may lead to brain disease (
      • Foss-Feig J.H.
      • Adkinson B.D.
      • Ji J.L.
      • Yang G.
      • Srihari V.H.
      • McPartland J.C.
      • et al.
      Searching for cross-diagnostic convergence: neural mechanisms governing excitation and inhibition balance in schizophrenia and autism spectrum disorders.
      ). For example, schizophrenia has been associated with a low E/I ratio caused by hypoactive receptors for the excitatory neurotransmitter glutamate (
      • Foss-Feig J.H.
      • Adkinson B.D.
      • Ji J.L.
      • Yang G.
      • Srihari V.H.
      • McPartland J.C.
      • et al.
      Searching for cross-diagnostic convergence: neural mechanisms governing excitation and inhibition balance in schizophrenia and autism spectrum disorders.
      ) and steeper aperiodic slopes compared to controls during rest (
      • Ramsay I.
      • Lynn P.
      • Lee E.
      • Schermitzler B.
      • Leipold D.
      • Sponheim S.
      Disturbances in Aperiodic Neural Activity During Resting State in Patients With Schizophrenia.
      ). Analogously, depression might be associated with E/I perturbations due to its cholinergic-monoaminergic (
      • Janowsky D.
      • Davis J.
      • El-Yousef M.K.
      • Sekerke H.J.
      A cholinergic-adrenergic hypothesis of mania and depression.
      ,
      • Wichniak A.
      • Wierzbicka A.
      • Jernajczyk W.
      Sleep as a biomarker for depression.
      ), glutamatergic (
      • Moriguchi S.
      • Takamiya A.
      • Noda Y.
      • Horita N.
      • Wada M.
      • Tsugawa S.
      • et al.
      Glutamatergic neurometabolite levels in major depressive disorder: a systematic review and meta-analysis of proton magnetic resonance spectroscopy studies.
      ), and/or GABAergic imbalance.
      Thus, it has been suggested that GABAergic deficit may play a central role in the etiology of MDD, especially in melancholic (
      • Luscher B.
      • Shen Q.
      • Sahir N.
      The GABAergic deficit hypothesis of major depressive disorder.
      ) and treatment-resistant types of depression (

      Mazza F, Griffiths J, Hay E. Biomarkers of reduced inhibition in human cortical microcircuit signals in depression. In JOURNAL OF COMPUTATIONAL NEUROSCIENCE 2021 Dec 1 (Vol. 49, No. SUPPL 1, pp. S86-S86). VAN GODEWIJCKSTRAAT 30, 3311 GZ DORDRECHT, NETHERLANDS: SPRINGER.

      ), while targeting the E/I imbalance in depression via enhancing the GABAergic system with antidepressant therapies may contribute to a greater remission rate and reduce the risk of relapse (
      • Prévot T.
      • Sibille E.
      Altered GABA-mediated information processing and cognitive dysfunctions in depression and other brain disorders.
      ). At the cellular level, changes in GABAergic interneurons affect the regulation of excitatory signals from and onto pyramidal neurons (
      • Northoff G.
      • Sibille E.
      Why are cortical GABA neurons relevant to internal focus in depression? A cross-level model linking cellular, biochemical and neural network findings.
      ), the primary contributors to the EEG signal. Following this literature, flatter aperiodic slopes observed here during sleep in unmedicated and medicated patients may reflect a shift in the E/I ratio in favor of excitation due to cellular alterations of the GABAergic, glutamatergic, and cholinergic-monoaminergic systems. Nevertheless, it should be stressed that thus far, there is no evidence that altered aperiodic dynamics can be used as readouts of cholinergic, monoaminergic, glutamatergic, and GABAergic imbalance. Moreover, whereas some authors suggest that the aperiodic slope is an indicator of E/I balance (
      • Waschke L.
      • Donoghue T.
      • Fiedler L.
      • Smith S.
      • Garrett D.D.
      • Voytek B.
      • et al.
      Modality-specific tracking of attention and sensory statistics in the human electrophysiological spectral exponent.
      ,
      • Gao R.
      • Peterson E.J.
      • Voytek B.
      Inferring synaptic excitation/inhibition balance from field potentials.
      ,
      • Lendner J.D.
      • Helfrich R.F.
      • Mander B.A.
      • Romundstad L.
      • Lin J.J.
      • Walker M.P.
      • et al.
      An electrophysiological marker of arousal level in humans.
      ), others state that currently, the relationship between aperiodic slopes and E/I balance remains a hypothesis to be further validated (
      • Gerster M.
      • Waterstraat G.
      • Litvak V.
      • Lehnertz K.
      • Schnitzler A.
      • Florin E.
      • et al.
      Separating neural oscillations from aperiodic 1/f activity: challenges and recommendations.
      ). In addition, when aiming to link aperiodic slopes and E/I balance, one should keep in mind that aperiodic 1/f-like processes are very ubiquitous in nature and are not limited to neural activity (
      • Gerster M.
      • Waterstraat G.
      • Litvak V.
      • Lehnertz K.
      • Schnitzler A.
      • Florin E.
      • et al.
      Separating neural oscillations from aperiodic 1/f activity: challenges and recommendations.
      ,
      • Waschke L.
      • Donoghue T.
      • Fiedler L.
      • Smith S.
      • Garrett D.D.
      • Voytek B.
      • et al.
      Modality-specific tracking of attention and sensory statistics in the human electrophysiological spectral exponent.
      ,
      • He B.J.
      Scale-free brain activity: past, present, and future.
      ).
      Of special interest was the effect of antidepressants: we found that during all sleep stages, medicated patients showed flatter slopes compared to controls. We replicated this association using two independently collected datasets of short and long-term medicated patients (Supplementary Material-5). In addition, we found that the medicated state showed flatter slopes compared to the own unmedicated state. In line with our findings, a recent study in healthy females has reported that one week of intake of the SSRI escitalopram induces a flattening of aperiodic slopes during rest in favor of excitation (

      Zsido RG, Molloy EN, Cesnaite E, Zheleva G, Beinhölzl N, Scharrer U, et al. One‐week escitalopram intake alters the excitation–inhibition balance in the healthy female brain. Human brain mapping. 2021 Jan 1. https://doi.org/10.1101/2021.07.09.451806

      ).
      Furthermore, we found that in 7-day medicated patients, flatter aperiodic slopes during non-REM sleep correlated with such alterations in sleep architecture as a higher proportion of non-REM stage 1, a lower proportion of REM, delayed onset of non-REM stage 3 and REM sleep, and shorter total sleep time. In the literature, these alterations in sleep architecture are often interpreted as impaired sleep (
      • Steiger A.
      • Pawlowski M.
      Depression and sleep.
      ). Specifically, MDD patients have been reported to show prolonged sleep latency, increased WASO, early morning awakening, reduced slow-wave sleep, shortened latency and increased amount of REM sleep (
      • Palagini L.
      • Baglioni C.
      • Ciapparelli A.
      • Gemignani A.
      • Riemann D.
      REM sleep dysregulation in depression: state of the art.
      ). Here, unmedicated MDD patients have shown increased WASO while other sleep architecture features were comparable to those measured in controls. The same patients in the medicated state showed increased WASO, decreased REM sleep proportion and prolonged REM sleep onset.
      Delayed onset and reduced amount of REM sleep are well-known aftereffects of almost all antidepressants (
      • Palagini L.
      • Baglioni C.
      • Ciapparelli A.
      • Gemignani A.
      • Riemann D.
      REM sleep dysregulation in depression: state of the art.
      ). Notably, we replicated the association between flatter aperiodic slopes during non-REM sleep and a decreased proportion of REM sleep and delayed REM sleep onset using an independent dataset of 7-day medicated patients (Supplementary Material 5, Table S5.2). This link is in line with the previous proposition that some antidepressants (for example, such SNRI as venlafaxine) may impair sleep due to their activating effects (
      • Wichniak A.
      • Wierzbicka A.
      • Walęcka M.
      • Jernajczyk W.
      Effects of antidepressants on sleep.
      ). Furthermore, the observed association suggests that aperiodic slopes flattering seen in medicated MDD patients is a potential readout of altered sleep architecture and impaired sleep known in this disorder. Nevertheless, further studies are needed to test this possibility.
      Whereas our findings bring new insights about the association between aperiodic activity, sleep architecture, and antidepressants, they do not advance the current understanding of the treatment response (or lack thereof) as changes in aperiodic activity did not correlate with clinical improvement (as assessed by the HAM-D). It is possible that other depression scales (that were not available in this study) would be more sensitive in detecting the hypothesized association between depression severity and aperiodic activity. Given that a deeper understanding of the antidepressants' effects on sleep is crucial for successful treatment, future large-scale longitudinal research is required to reveal whether aperiodic activity can serve as a marker for predicting individual cortical responsivity to different antidepressants.
      Besides their clinical importance, our findings are also essential from the methodological point of view as they confirm the importance of a recent recommendation to differentiate the total spectral power to its components in order "to avoid misrepresentation and misinterpretation of the data" (
      • Donoghue T.
      • Schaworonkow N.
      • Voytek B.
      Methodological considerations for studying neural oscillations.
      ,
      • Ouyang G.
      • Hildebrandt A.
      • Schmitz F.
      • Herrmann C.S.
      Decomposing alpha and 1/f brain activities reveals their differential associations with cognitive processing speed.
      ). Namely, we observed comparable total (i.e., non-differentiated to its components) spectral power (analyzed in 6) but different oscillatory and aperiodic components in unmedicated MDD patients and controls (Fig.1).
      Finally, here, aperiodic alterations were observed during sleep but not during WASO suggesting their specificity to sleep. Nevertheless, it should be kept in mind that the wake EEG is more variable and prone to artifacts than sleep EEG; therefore, the performed analysis might have not enough statistical power to detect between-group differences. Interestingly, in unmedicated patients, the morning resting state EEG showed steeper low-band and flatter high-band slopes compared to controls, while broadband slopes were comparable in both groups. This preliminary finding, however, requires further validation due to the small size of the tested sample (16 patients vs 16 controls, Supplementary Material-3).
      This study is not without limitations. First, one should keep in mind that the diagnosis of depression is subjective, and subtypes of depression likely exist even though they have not been systematically distinguished. Likewise, the stratification of the patients by antidepressant classes performed here was not clean enough as the patients used different antidepressants. Second, this research is correlational and precludes causal relations between the neurobiology of MDD and aperiodic activity.
      In conclusion, our findings suggest that flatter aperiodic slopes represent a new disease-relevant feature of sleep in MDD, which may reflect unstable, noisy neural activity due to a shift of the excitation-to-inhibition ratio in favor of excitation. In the future, these findings may lead to the development of a biomarker for personalized disease monitoring and therapy.

      Acknowledgments

      The first draft of this paper has been posted on medRxiv.
      We would like to thank all the participants for participating in this study. We would like to thank Sofia Tzioridou for her helpful suggestions.

      References

      1. American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.).

      2. World Health Organisation, WHO (2018) Fact sheet No 369: Depression. Available at: http://www.who.int/mediacentre/factsheets/fs369/en/. Accessed 21 August 2021

        • Staner L.
        Comorbidity of insomnia and depression.
        Sleep medicine review. 2010; 14: 35-46
      3. Courtet P, Olié E. Circadian dimension and severity of depression. European Neuropsychopharmacology. 2012 Jan 1;22:S476-S481. https://doi.org/10.1016/j.euroneuro.2012.07.009

        • Armitage R.
        • Hoffmann R.
        • Trivedi M.
        • Rush A.J.
        Slow-wave activity in NREM sleep: sex and age effects in depressed outpatients and healthy controls.
        Psychiatry research. 2000; 95: 201-213https://doi.org/10.1016/S0165-1781(00)00178-5
      4. Bovy L, Weber FD, Tendolkar I, Fernández G, Czisch M, Steiger A, et al. Non-REM sleep in major depressive disorder. bioRxiv. 2021 Jan 1. https://doi.org/10.1101/2021.03.19.436132

        • Gerster M.
        • Waterstraat G.
        • Litvak V.
        • Lehnertz K.
        • Schnitzler A.
        • Florin E.
        • et al.
        Separating neural oscillations from aperiodic 1/f activity: challenges and recommendations.
        Neuroinformatics. 2022 Apr 7; : 1-22https://doi.org/10.1007/s12021-022-09581-8
        • Donoghue T.
        • Schaworonkow N.
        • Voytek B.
        Methodological considerations for studying neural oscillations.
        European journal of neuroscience. 2021 Jul 16;
        • Ouyang G.
        • Hildebrandt A.
        • Schmitz F.
        • Herrmann C.S.
        Decomposing alpha and 1/f brain activities reveals their differential associations with cognitive processing speed.
        NeuroImage. 2020 Jan 15; 205116304https://doi.org/10.1016/j.neuroimage.2019.116304
        • Waschke L.
        • Donoghue T.
        • Fiedler L.
        • Smith S.
        • Garrett D.D.
        • Voytek B.
        • et al.
        Modality-specific tracking of attention and sensory statistics in the human electrophysiological spectral exponent.
        Elife. 2021; 10https://doi.org/10.7554/eLife.70068
        • He B.J.
        Scale-free brain activity: past, present, and future.
        Trends in cognitive sciences. 2014; 18: 480-487
        • Miller K.J.
        • Sorensen L.B.
        • Ojemann J.G.
        • Den Nijs M.
        Power-law scaling in the brain surface electric potential.
        PLoS computational biology. 2009 Dec 18; 5e1000609https://doi.org/10.1371/journal.pcbi.1000609
        • He B.J.
        • Zempel J.M.
        • Snyder A.Z.
        • Raichle M.E.
        The temporal structures and functional significance of scale-free brain activity.
        Neuron. 2010 May 13; 66: 353-369
        • Voytek B.
        • Knight R.T.
        Dynamic network communication as a unifying neural basis for cognition, development, aging, and disease.
        Biological psychiatry. 2015 Jun 15; 77: 1089-1097https://doi.org/10.1016/j.biopsych.2015.04.016
        • Gao R.
        • Peterson E.J.
        • Voytek B.
        Inferring synaptic excitation/inhibition balance from field potentials.
        NeuroImage. 2017; 158: 70-78https://doi.org/10.1016/j.neuroimage.2017.06.078
        • Lendner J.D.
        • Helfrich R.F.
        • Mander B.A.
        • Romundstad L.
        • Lin J.J.
        • Walker M.P.
        • et al.
        An electrophysiological marker of arousal level in humans.
        Elife. 2020 Jul 28; 9e55092https://doi.org/10.7554/eLife.55092
        • Foss-Feig J.H.
        • Adkinson B.D.
        • Ji J.L.
        • Yang G.
        • Srihari V.H.
        • McPartland J.C.
        • et al.
        Searching for cross-diagnostic convergence: neural mechanisms governing excitation and inhibition balance in schizophrenia and autism spectrum disorders.
        Biological psychiatry. 2017 May 15; 81: 848-861https://doi.org/10.1016/j.biopsych.2017.03.005
        • Janowsky D.
        • Davis J.
        • El-Yousef M.K.
        • Sekerke H.J.
        A cholinergic-adrenergic hypothesis of mania and depression.
        The Lancet. 1972 Sep 23; 300: 632-635
        • Wichniak A.
        • Wierzbicka A.
        • Jernajczyk W.
        Sleep as a biomarker for depression.
        International review of psychiatry. 2013 Oct 1; 25: 632-645https://doi.org/10.3109/09540261.2013.812067
        • Moriguchi S.
        • Takamiya A.
        • Noda Y.
        • Horita N.
        • Wada M.
        • Tsugawa S.
        • et al.
        Glutamatergic neurometabolite levels in major depressive disorder: a systematic review and meta-analysis of proton magnetic resonance spectroscopy studies.
        Molecular psychiatry. 2019 Jul; 24: 952-964https://doi.org/10.1038/s41380-018-0252-9
        • Luscher B.
        • Shen Q.
        • Sahir N.
        The GABAergic deficit hypothesis of major depressive disorder.
        Molecular psychiatry. 2011 Apr; 16: 383-406https://doi.org/10.1038/mp.2010.120
      5. Mazza F, Griffiths J, Hay E. Biomarkers of reduced inhibition in human cortical microcircuit signals in depression. In JOURNAL OF COMPUTATIONAL NEUROSCIENCE 2021 Dec 1 (Vol. 49, No. SUPPL 1, pp. S86-S86). VAN GODEWIJCKSTRAAT 30, 3311 GZ DORDRECHT, NETHERLANDS: SPRINGER.

        • Oostenveld R.
        • Praamstra P.
        The five percent electrode system for high-resolution EEG and ERP measurements.
        Clinical neurophysiology. 2001 Apr 1; 112: 713-719https://doi.org/10.1016/S1388-2457(00)00527-7
      6. Iber C. The AASM manual for the scoring of sleep and associated events: Rules. Terminology and technical specification. Published online 2007.

        • Wen H.
        • Liu Z.
        Separating fractal and oscillatory components in the power spectrum of neurophysiological signal.
        Brain topography. 2016 Jan; 29: 13-26
      7. Buzsáki G, Anastassiou CA, Koch C. The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nature reviews neuroscience. 2012 Jun;13(6):407-420.

        • Tagliazucchi E.
        • van Someren E.J.
        The large-scale functional connectivity correlates of consciousness and arousal during the healthy and pathological human sleep cycle.
        NeuroImage. 2017 Oct 15; 160: 55-72https://doi.org/10.1016/j.neuroimage.2017.06.026
      8. Medel V, Irani M, Ossandon T, Boncompte G. Complexity and 1/f slope jointly reflect cortical states across different E/I balances. bioRxiv. 2020 Jan 1. https://doi.org/10.1101/2020.09.15.298497

        • Li S.C.
        • Sikström S.
        Integrative neurocomputational perspectives on cognitive aging, neuromodulation, and representation.
        Neuroscience & biobehavioral reviews. 2002 Nov 1; 26: 795-808
        • Voytek B.
        • Kramer M.A.
        • Case J.
        • Lepage K.Q.
        • Tempesta Z.R.
        • Knight R.T.
        • et al.
        Age-related changes in 1/f neural electrophysiological noise.
        Journal of neuroscience. 2015 Sep 23; 35: 13257-13265https://doi.org/10.1523/JNEUROSCI.2332-14.2015
      9. Miskovic V, MacDonald KJ, Rhodes LJ, Cote KA. Changes in EEG multiscale entropy and power‐law frequency scaling during the human sleep cycle. Human Brain Mapp. 2019 Feb 1;40(2):538-551. https://doi.org/10.1002/hbm.24393

        • Colombo M.A.
        • Napolitani M.
        • Boly M.
        • Gosseries O.
        • Casarotto S.
        • Rosanova M.
        • et al.
        The spectral exponent of the resting EEG indexes the presence of consciousness during unresponsiveness induced by propofol, xenon, and ketamine.
        NeuroImage. 2019; 189: 631-644https://doi.org/10.1016/j.neuroimage.2019.01.024
        • Ramsay I.
        • Lynn P.
        • Lee E.
        • Schermitzler B.
        • Leipold D.
        • Sponheim S.
        Disturbances in Aperiodic Neural Activity During Resting State in Patients With Schizophrenia.
        Biological psychiatry. 2021 May 1; 89: S254-S255https://doi.org/10.1016/j.biopsych.2021.02.637
        • Prévot T.
        • Sibille E.
        Altered GABA-mediated information processing and cognitive dysfunctions in depression and other brain disorders.
        Molecular psychiatry. 2021 Jan; 26: 151-167https://doi.org/10.1038/s41380-020-0727-3
        • Northoff G.
        • Sibille E.
        Why are cortical GABA neurons relevant to internal focus in depression? A cross-level model linking cellular, biochemical and neural network findings.
        Molecular psychiatry. 2014 Sep; 19: 966-977https://doi.org/10.1038/mp.2014.68
      10. Zsido RG, Molloy EN, Cesnaite E, Zheleva G, Beinhölzl N, Scharrer U, et al. One‐week escitalopram intake alters the excitation–inhibition balance in the healthy female brain. Human brain mapping. 2021 Jan 1. https://doi.org/10.1101/2021.07.09.451806

        • Steiger A.
        • Pawlowski M.
        Depression and sleep.
        International journal of molecular sciences. 2019 Jan; 20: 607https://doi.org/10.3390/ijms20030607
        • Palagini L.
        • Baglioni C.
        • Ciapparelli A.
        • Gemignani A.
        • Riemann D.
        REM sleep dysregulation in depression: state of the art.
        Sleep medicine reviews. 2013 Oct 1; 17: 377-390https://doi.org/10.1016/j.smrv.2012.11.001
        • Wichniak A.
        • Wierzbicka A.
        • Walęcka M.
        • Jernajczyk W.
        Effects of antidepressants on sleep.
        Current psychiatry reports. 2017 Sep; 19: 1-7https://doi.org/10.1007/s11920-017-0816-4