Scholarly article on topic 'Alpha Power Increase After Transcranial Alternating Current Stimulation at Alpha Frequency (α-tACS) Reflects Plastic Changes Rather Than Entrainment'

Alpha Power Increase After Transcranial Alternating Current Stimulation at Alpha Frequency (α-tACS) Reflects Plastic Changes Rather Than Entrainment Academic research paper on "Biological sciences"

CC BY
0
0
Share paper
Academic journal
Brain Stimulation
OECD Field of science
Keywords
{"Transcranial alternating current stimulation" / "Alpha oscillations" / Entrainment / "Spike-timing dependent plasticity" / Electroencephalogram / Synchronization}

Abstract of research paper on Biological sciences, author of scientific article — Alexandra Vossen, Joachim Gross, Gregor Thut

Abstract Background Periodic stimulation of occipital areas using transcranial alternating current stimulation (tACS) at alpha (α) frequency (8–12 Hz) enhances electroencephalographic (EEG) α-oscillation long after tACS-offset. Two mechanisms have been suggested to underlie these changes in oscillatory EEG activity: tACS-induced entrainment of brain oscillations and/or tACS-induced changes in oscillatory circuits by spike-timing dependent plasticity. Objective We tested to what extent plasticity can account for tACS-aftereffects when controlling for entrainment “echoes.” To this end, we used a novel, intermittent tACS protocol and investigated the strength of the aftereffect as a function of phase continuity between successive tACS episodes, as well as the match between stimulation frequency and endogenous α-frequency. Methods 12 healthy participants were stimulated at around individual α-frequency for 11–15 min in four sessions using intermittent tACS or sham. Successive tACS events were either phase-continuous or phase-discontinuous, and either 3 or 8 s long. EEG α-phase and power changes were compared after and between episodes of α-tACS across conditions and against sham. Results α-aftereffects were successfully replicated after intermittent stimulation using 8-s but not 3-s trains. These aftereffects did not reveal any of the characteristics of entrainment echoes in that they were independent of tACS phase-continuity and showed neither prolonged phase alignment nor frequency synchronization to the exact stimulation frequency. Conclusion Our results indicate that plasticity mechanisms are sufficient to explain α-aftereffects in response to α-tACS, and inform models of tACS-induced plasticity in oscillatory circuits. Modifying brain oscillations with tACS holds promise for clinical applications in disorders involving abnormal neural synchrony.

Academic research paper on topic "Alpha Power Increase After Transcranial Alternating Current Stimulation at Alpha Frequency (α-tACS) Reflects Plastic Changes Rather Than Entrainment"

Accepted Manuscript

STIMULATION

Alpha power increase after transcranial alternating current stimulation at alpha-frequency (a-tACS) reflects plastic changes rather than entrainment

Alexandra Vossen, Joachim Gross, Gregor Thut

S^MAiftfii« «AÜAAAAAA iJIHAJMAMA

-mmmmmmmmrn

PII: S1935-861X(14)00436-7

DOI: 10.1016/j.brs.2014.12.004

Reference: BRS 654

To appear in: Brain Stimulation

Received Date: 4 September 2014 Revised Date: 3 December 2014 Accepted Date: 14 December 2014

Please cite this article as: Vossen A, Gross J, Thut G, Alpha power increase after transcranial alternating current stimulation at alpha-frequency (a-tACS) reflects plastic changes rather than entrainment, Brain Stimulation (2015), doi: 10.1016/j.brs.2014.12.004.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Alpha power increase after transcranial alternating current stimulation at alpha-frequency (a-tACS) reflects plastic changes rather than entrainment

Alexandra Vossena, Joachim Grossb, Gregor Thutb

aSchool of Psychology & bInstitute of Neuroscience and Psychology

University of Glasgow

58 Hillhead Street

Glasgow G12 8QB

United Kingdom

a.vossen.1@research.gla.ac.uk (A. Vossen) Joachim.Gross@glasgow.ac.uk (J. Gross) Gregor.Thut@glasgow.ac.uk (G. Thut)

Corresponding authors: A. Vossen (Tel: +44 (0)141 330 5151), G. Thut (Tel: +44(0)141 330 3395)

Abstract

Background: Periodic stimulation of occipital areas using transcranial alternating current stimulation (tACS) at alpha (a) frequency (8-12Hz) enhances electroencephalographic (EEG) a-oscillation long after tACS-offset. Two mechanisms have been suggested to underlie these changes in oscillatory EEG activity: tACS-induced entrainment of brain oscillations and/or tACS-induced changes in oscillatory circuits by spike-timing dependent plasticity. Objective: We tested to what extent plasticity can account for tACS-aftereffects when controlling for entrainment "echoes". To this end, we used a novel, intermittent tACS protocol and investigated the strength of the aftereffect as a function of phase continuity between successive tACS episodes, as well as the match between stimulation frequency and endogenous a-frequency.

Methods: 12 healthy participants were stimulated at around individual a-frequency for 15-20min in four sessions using intermittent tACS or sham. Successive tACS events were either phase-continuous or phase-discontinuous, and either 3 or 8 sec long. EEG a- phase and power changes were compared after and between episodes of a-tACS across conditions and against sham.

Results: a-aftereffects were successfully replicated after intermittent stimulation using 8-second but not 3-second trains. These aftereffects did not reveal any of the characteristics of entrainment echoes in that they were independent of tACS phase-continuity and showed neither prolonged phase alignment nor frequency synchronization to the exact stimulation frequency.

Conclusion: Our results indicate that plasticity mechanisms are sufficient to explain a-aftereffects in response to a-tACS, and inform models of tACS induced plasticity in oscillatory circuits. Modifying brain oscillations with tACS holds promise for clinical applications in disorders involving abnormal neural synchrony.

Keywords: transcranial alternating current stimulation, alpha oscillations, entrainment, spike-timing dependent plasticity, electroencephalogram, synchronisation

Many human electrophysiological studies have mapped specific aspects of perception, memory, and cognition onto specific features of oscillatory brain activity, including phase and frequency [see e.g. 1-10]. Conventionally, such functional maps are established via noninvasive recording techniques such as electro/ magnetoencephalography (EEG/MEG) by examining task-related modulation of oscillatory brain activity or its covariation with behavioural performance measures. However, these maps are correlational by nature and do not permit the distinction between epiphenomenal and causal functional accounts. Recent attempts to demonstrate causal roles of oscillatory brain activity in implementing function have used non-invasive brain stimulation (NIBS, [11-14]) to promote natural neural frequencies [15]. To this end, an external, periodic electromagnetic force is applied over the appropriate (potentially task-relevant) brain area at the area's preferred oscillatory frequency with the aim to synchronise the intrinsic oscillations to the external force, and to assess the associated behavioural consequences (reviewed in [9,16,17]). A promising NIBS-approach for such controlled intervention is transcranial alternating current stimulation (tACS). tACS involves the induction of a weak sinusoidal electric current between two or more scalp electrodes [17,18] which can be applied at biologically relevant frequencies (i.e. those frequencies spontaneously exhibited by neural networks). In line with the view that tACS selectively interacts with underlying brain oscillations and functions, increasing numbers of behavioural studies show that frequency-tuned tACS affects specific aspects of perception [19-25], memory [26-28], motor function [29-34] and higher-order cognition [35-37], in many instances matching the known correlational links between EEG/MEG-frequency and function [19-36] (but see [38-41]). However, little is known about the electrophysiological underpinnings and the precise mechanism through which tACS procures its effects.

Two main hypotheses have been suggested: that tACS directly entrains underlying brain oscillations [17,18,42,43] and/or that tACS leads to synaptic changes via spike-timing dependent plasticity mechanisms [28,44]. Entrainment of brain oscillations refers to the temporal alignment of intrinsic brain activity to periodic (e.g. sensory, electrical, or magnetic) stimulation [16]. It involves i) a neural population capable of producing rhythmic activity at the desired frequency, and ii) phase alignment of this intrinsic activity to the phase of the external driving source. tACS-induced entrainment has been demonstrated during (online to) tACS both behaviourally [24,30,37] and electrophysiologically in humans [20,37], as well as in animal studies both in vitro and in vivo [45-49]. The latter work, as well as research on photic driving in humans [50,51], indicate that entrainment is strongest when stimulation frequency is at or close to the network's preferred frequency (Eigenfrequency). Specifically, the stimulated system is then expected to respond at the driving frequency rather than its Eigenfrequency [45,52]. Spike-timing dependent plasticity, on the other hand, has been suggested to underlie the enhancement of oscillatory brain activity at tACS-frequency beyond stimulation (i.e. offline to tACSj [28,44]. Such aftereffects have been reported in the form of enhanced posterior a-power after prolonged (ca. 10-20min) occipito-parietal a-tACS (or a-tACS with a DC-offset) [20,24,44] that lasted for at least 30 minutes [53]. In the present study, we tested to what extent plasticity can account for tACS-aftereffects when controlling for entrainment "echoes", i.e. entrained activity that remains stable after the end of rhythmic stimulation. To this end, we employed an intermittent tACS-protocol and applied short parieto-occipital a-tACS trains interrupted by breaks of equal duration. Total tACS-duration was comparable to the continuous a-tACS-protocols previously reported to lead to offline a-enhancement [20,24,44,53]. In order to assess the contribution of entrainment echoes to the a-aftereffect, we manipulated phase-continuity (continuous versus discontinuous) between successive a-tACS trains. Based on observations online to tACS (see

[20]) as well as theoretical groundwork [44,52], we reasoned that if entrainment echoes come into play, a-enhancement should be 1) stronger when intermittent a-tACS trains are applied in phase-continuous versus phase-discontinuous regimes, 2) centered at stimulation frequency rather than intrinsic Eigenfrequency, and 3) stronger when the stimulation frequency matches the spontaneous a-frequency, while 4) EEG phase-locking to the phase of the tACS-train should outlast tACS-offset as a minimum requirement for stable entrainment over minutes. Our EEG results confirmed enhanced a-power after a-tACS compared to sham stimulation, but did not reveal any of the hypothesised offline entrainment characteristics. Consistent with plasticity as the predominant cause for aftereffects, a-enhancement 1) occurred irrespective of phase-continuity between trains, 2) was observed at spontaneous a-peak frequency, and was 3) neither stronger with tACS at intrinsic a-frequency, nor 4) associated with prolonged phase-locking beyond tACS.

Materials and methods

Participants

12 healthy volunteers (6 male, age 27±5 years) completed this study. All volunteers gave written informed consent and received monetary compensation for their participation. The study was approved by the local ethics committee of the College of Science and Engineering, University of Glasgow. No participants reported a history of neurological/ psychiatric disorders or any other contraindication to tACS (current use of psychoactive medication/drugs, metal implants, pregnancy).

tACS was administered through a battery driven constant current stimulator (DC Stimulator Plus, NeuroConn, Ilmenau/Germany) controlled through Spike2 software via a Power1401

mkII microcomputer (both Cambridge Electronic Design, Cambridge/UK). 5x7cm rubber electrodes in saline-soaked sponges (0.9%-NaCl) with a thin layer of electrode gel were attached to the scalp with rubber bands. Electrodes were placed bilaterally over PO7/PO9 and P08/P010 of the 10/10-system (Fig.1A; cf. [44]).

Individual stimulation frequency (ISF) and intensity were determined once, in the first session, for all four sessions. ISF was determined from resting EEG with eyes open by identifying each individual peak frequency in the a-range (8-12Hz) at electrode POz using Fast Fourier Transforms (FFTs, frequency resolution .5Hz) and ranged from 8-11Hz across participants. tACS-intensity was adjusted below individual phosphene- and discomfort threshold using a staircase procedure (see Supplementary information) and ranged from 1.35-2mA (peak-to-peak).

tACS was administered in a within-subject design with three active conditions and one sham condition (Fig.1B) on four different days. In all active conditions a-tACS at ISF was applied in an intermittent on/off pattern. Total stimulation duration (amount of on-time) in each active condition was constant for any particular participant (7200 a-cycles at ISF) but varied across participants due to the variability in individual posterior a-frequency (i.e. from ~11min for an ISF of 11Hz to ~15min for an ISF of 8Hz). Total session duration was twice the length of total stimulation time (or equivalent for sham). Across conditions, we varied the length of single tACS-epochs (on-period) as well as phase-consistency across epochs. In the short, phase-continuous condition (ShortCo) (Fig.1B.1), tACS was switched on for 30 cycles (i.e. on-periods of 3 s in participants with a 10Hz-ISF) followed by an off-period of the same duration. This was repeated 240 times with phase continuity between successive on-states (i.e. by adjusting amplitude, but not phase, of a virtual sine-wave spanning the whole stimulation session). In the long, phase-continuous condition (LongCo)(Fig.1B.2), tACS was

switched on/off with phase continuity (as above) for 80 cycles (i.e. on/off for 8s-epochs in participants with a 10Hz-ISF) in 90 repetitions. The long, _phase-discontinuous condition (LongDis)(F\ g.1B.3) was identical to LongCo, except that phase-continuity was disrupted across single tACS-epochs by introducing a phase shift of 0° , 90°, 180°, or 270° to the virtual sine wave during off-periods (approximately equal probability) with respect to the previous on-period, thus initiating tACS at a different phase angle. In all active conditions, tACS-intensity was ramped up over the first 10 cycles to minimise unpleasant sensations under the electrodes. Finally, in the sham condition, only one short tACS-train (10 cycles ramp-up, 10 cycles ramp-down) was administered at the beginning of the session. This condition was included to control for tACS-unspecific effects (e.g. fatigue).

EEG recordings

EEG was recorded at the midline sites Fpz, Fz, Cz, CPz, Pz, and POz (referenced to AFz) (Fig.1A; cf.[44]) using a BrainAmpMRPlus amplifier (BrainProducts, Munich/Germany). Vertical eye movements were recorded from two additional electrodes above and below the right eye. The signal was bandpass-filtered online between .1-1000Hz and digitised at a sampling rate of 5kHz.

Procedure

Each participant underwent four sessions of maximally two hours each. Sessions were at least 3 days apart. Preparation of tACS- and EEG-electrodes took ~45min. Data acquisition started with 2 minutes of resting EEG with eyes open (pre-test) (Fig.1C). Participants then underwent one of the stimulation protocols in counterbalanced order while EEG was continuously measured. For the duration of each protocol, participants performed a visual

vigilance task to maintain alertness (see Supplemental Material). Finally, an additional 2min of resting EEG with eyes open was recorded (post-test).

EEG Analysis

Analyses were conducted in BrainVision Analyzer 2.0 (BrainProducts) and Matlab (MathWorks, Natick/USA) using the Fieldtrip Toolbox (Donders Centre for Cognitive Neuroimaging, Nijmegen/NL). For all statistical analyses, non-parametric tests were used (IBM SPSS Statistics, version 19.0, IBM Corp, Armonk/US; see Supplemental Material). The reported results refer to the signal recorded at electrode POz except for one subject, where due to excessive noise in one condition we chose to analyse Pz instead.

Analysis of aftereffects in a-power (pre vs post-test)

The analysis of the pre- and post-tACS EEG measurements largely followed [24,44]. The "eyes open" resting EEGs were segmented into 1s-epochs. Epochs containing eye movement and muscle contraction artefacts were discarded. A fast Fourier transform (FFT) for frequencies between 1 and 20Hz (.5Hz resolution) was calculated for individual epochs using a Hanning window and 2s zero-padding. The resulting spectra of each condition were averaged across epochs as well as across the individually determined a-bands (ISF±2Hz) per tACS-condition. Normalised relative changes of mean a-power from pre-test to post-test were calculated in decibel (change=10*log10(post-test/pre-test)).

Analysis of offline changes in a-activity in the intermittent, tACS-free intervals Pre-processing: Epochs of 2.3s duration were extracted from the EEG between successive tACS-trains starting 100ms after tACS-offset (due to residual tACS-artefact in the first 100ms of EEG). Very noisy epochs and epochs with eye blinks at trial-onset were removed

after visual inspection of the data. Remaining eye blink contaminations were then eliminated (1) using a principal component denoising approach (implemented in Fieldtrip) with the bipolar EOG-derivation as reference signal (using 1-8Hz bandpass-filtered data to optimize blink detection, and applying the respective PCA-weights to the original data), and (2) by discarding the epoch if elimination was not successful. Because both long conditions had significantly lower trial numbers than sham and ShortCo, we randomly sampled (without replacement), for each participant and each condition, as many trials as available in the condition with the lowest trial number. All subsequent analyses were conducted on these subsamples of equal size.

Analysis of relative change in induced a-power: We followed a similar pipeline as for the analysis of the pre- and post-tACS data. From the pre-processed data, two 1s-epochs were cut at the beginning of each 2.3s-interval. These were divided into blocks of early and late epochs, respectively (i.e. first and second half of the experimental session). FFT-spectra were calculated for each 1s-epoch separately, and subsequently averaged per block and tACS-condition. Average power in the individual stimulation band (ISF±2Hz) for each block was again log-normalised to pre-test power.

Analysis of a _ phase-locking: To obtain phase information, preprocessed data were bandpass-filtered in individual a-bands (ISF±2Hz) and Hilbert-transformed. The resulting complex values were normalised to unit amplitude. The phase locking value (PLV) was computed for each time point as the absolute value of the mean of these normalised complex values across trials. PLVs were averaged across the first 200ms of the 2.3s-epoch (i.e. from 100-300ms post artefact) and then across epochs within early and late blocks in each tACS-condition.

Results

a-Aftereffect replicated with intermittent a-tACS

We found a-power (ISF±2) to be enhanced after intermittent a-tACS (pre versus post-test), with participants showing on average stronger a-enhancement after active tACS as compared to sham (see Fig.2A for group-averages, Fig.2B for individual data). Specifically, in both long conditions individual responses were highly consistent across participants, with 11 out of 12 participants showing stronger a-enhancement to a-tACS in the long phase-continuous and 10 out of 12 in the long phase-discontinuous condition as compared to sham (Fig.2B, middle and right panel: LongCo vs. Sham and LongDis vs Sham; Suppl.Fig. 1, right). Statistically, a main effect of condition was confirmed by a Friedman Test (X 3 =11.1, P=0.011). Breaking down this effect using the Wilcoxon Signed Rank Tests (2-tailed) indeed revealed significant a-enhancement only for both long tACS conditions compared to sham (LongCo vs. Sham: Z=2.82, P=0.005; LongDis vs. Sham: Z=2.04, P =0.041; ShortCo vs. Sham: Z=1.26, P=0.21), replicating the a-aftereffect previously reported for continuous a-tACS-protocols [20,24,44,53].

a-Aftereffect does not differ between phase-continuous and phase-discontinuous protocols

a-enhancement after active tACS (LongCo> LongDis> ShortCo) did not significantly differ between conditions (all P>0.05, Fig.2A). While long intermittent tACS significantly enhanced a-power (relative to sham), this enhancement was observed irrespective of phase-continuity between tACS-trains. Hence, introducing phase jitter during tACS did not disrupt the a-aftereffect, which speaks against prolonged entrainment echoes contributing to the aftereffects.

a-Aftereffects do not peak at stimulation frequency, but at preferred cortical frequency While we stimulated at a fixed frequency (ISF= individual a frequency (IAF) at day 1), several participants showed variable IAF across sessions. This was established by randomly

sampling (1000 repetitions with replacement) and averaging subsets of spectra from 1s epochs in pretest-EEG within each session to extract peak-frequency in the 8-12Hz-range. IAF on a given day was defined as the mode of these peaks. As a consequence, ISF deviated from IAF between sessions for several participants (range: -1.5Hz to +3.0Hz). This allowed us to assess whether aftereffects peaked at ISF or spontaneous IAF. Note that ISF was in most cases slightly below the IAF of a given session (Fig.3A). Breaking down the a-band into nine bins (IAF-2 to IAF+2, in 0.5Hz steps) (Fig.3B), we found that tACS-aftereffects (LongCo > LongDis > ShortCo) peaked at IAF and IAF+0.5Hz (rather than ISF), i.e. not showing the left-skew of the ISF histogram (see Fig.3B). Separate Friedman Tests on the relative a-increase in the IAF-centred a-band (IAF-0.5Hz to IAF+0.5Hz) and the two flanker a-bands (IAF-2Hz to IAF-1Hz/ IAF+1Hz to ISF+2Hz) revealed significant aftereffects in the IAF-centred band (X23=8.1, P=0.044) and the higher a-band (X23=9.0, P=0.029). At the IAF-centred band, the contrasts of both LongCo- and LongDis-conditions against Sham were significant (Wilcoxon Signed Rank Test; LongCo: Z=2.90, P =0.004; LongDis: Z=1.96, P=0.05; all other P>0.05). In the higher a-band, only LongCo was different from Sham (Z=2.51, P=0.012). Importantly, repeating the same analysis but now centred on ISF (instead of IAF) did not reveal significant tACS-related a-aftereffects at ISF (ISF-0.5Hz to ISF+0.5Hz, Friedman P>0.05). Hence, tACS-induced aftereffects were observed at or above the preferred cortical frequency but not at stimulation frequency, which again is inconsistent with prolonged entrainment echoes contributing to the aftereffect.

No enhancement of a-aftereffects when stimulation and preferred frequency match

In addition, we took advantage of the variability of IAF relative to ISF to assess the dependence of a-enhancement on the ISF-to-IAF match in any given session. To this end, we correlated the difference in a-enhancement during tACS relative to sham against the

deviation of ISF from actual IAF (i.e. tACS minus sham vs. ISF minus IAF). We found that no active tACS-condition showed stronger a-enhancement with better match between ISF and IAF (Fig.4). Instead, we found a significant inverse relationship in the most effective condition (LongCo), with stronger tACS-induced a-enhancement for greater deviations between ISF and IAF (Fig.4, green rectangles, Spearman's rho=-0.90, P<0.001). This association remained strong even with the most extreme case removed (Spearman's rho=-0.87, P< 0.001). A correlation derived from a small sample must be considered with caution but the data show that a-enhancement does not depend on a perfect match between ISF and IAF, contrary to what would be expected from entrainment echoes, and in favour of plasticity effects.

Analysis of offline a-changes in intermittent, tACS-free intervals

The pattern of tACS-induced a-power changes in the intermittent intervals during stimulation (Fig.5A) was suggestive of a progressive build-up of the a-aftereffects shown in Fig.2A (but not significant in either early/late block (X 3=6.0/4.7, P=0.112/0.195). Critically, we found no evidence of induced phase-locking (versus sham) in these intervals (i.e. after ~8 sec stimulation with individual tACS trains; Fig.5B) (early: X23=2.5, P=48; late X23= 0.7, P=0.87), again disagreeing with entrainment echoes contributing to the tACS-aftereffects. The absence of phase-locking immediately after tACS offset shows that online entrainment (if present) does not outlast the tACS trains even between individual trials, and rules out the survival of entrainment echoes for several minutes.

Discussion

This study tested in a novel intermittent tACS paradigm whether plasticity mechanisms are sufficient to explain a-aftereffects in response to a-tACS. To this end, we manipulated phase

continuity and train duration in three discontinuous tACS-protocols with constant total stimulation time and compared tACS-induced offline a-changes against sham. While the previously reported offline a-enhancement [20,44,53,54] was replicated, our data rule out entrainment echoes as a possible explanation of the a-aftereffect in our intermittent protocol, and support the plasticity model as the underlying mechanism. Despite growing evidence for entrainment during tACS [20,25,30,32,37,54], our findings indicate that online tACS-entrainment effects may not be strong enough to outlast stimulation, while offline tACS plasticity effects may be present in the absence of entrainment echoes. A similar distinction between online and offline effects has been made for transcranial magnetic stimulation (TMS): Short bursts of rhythmic TMS enhance brain oscillations at TMS-frequency during (i.e. online to) TMS by immediate entrainment [55-57], but prolonged TMS leads to longer-lasting effects on brain oscillations that have been attributed to other mechanisms (i.e. long term potentiation or -depression) ([43,58], see also [59]). An open question is to what extent online entrainment effects and offline plasticity effects are independent. Below we discuss, in light of our and related recent findings, two plasticity models, which assume dependence versus independence of online entrainment and offline plasticity effects, respectively.

tACS-induced plasticity: The Spike-timing dependent plasticity account

As introduced above, one mechanism that has been proposed to explain tACS-induced a -aftereffects [28,44] is spike-timing dependent plasticity (STDP). In STDP, the order and timing of pre-and postsynaptic potentials determine the magnitude, and direction, of changes in synaptic strength [60-62]. Zaehle et al [44] used a neural network model incorporating STDP-rules to show that periodic 10Hz-stimulation can strengthen or weaken the synaptic weights of neuronal circuits (recurrent loops) depending on their reverberation frequency. In this model, online entrainment is the window into longer lasting synaptic plasticity effects

that translate into frequency-specific changes in oscillatory activity. The model is illustrated in Figure 6, which is adapted from [44] with a slight modification: We assume higher weights for selective circuits (here with a periodicity of 100ms or 10Hz,see Fig6A) to accommodate physiological constraints (here the presence of an individual's dominant a-frequency). This slightly deviates from the model of Zaehle et al. [44], which presumes uniform distribution of starting weights across all loops, i.e. does not explicitly take into account the existence of intrinsic resonance frequencies (although motivated by them). Specifically, with this new assumption, the model predicts synaptic strengthening in dominant (a-)loops when the stimulation frequency falls into a narrow range of frequencies slightly lower than the spontaneous a-peak (Fig 6B), which is in line with our present results. Under these conditions, post-synaptic events (S1, see Fig 6A) driven by tACS are generated at a slightly slower pace (< IAF) than the time required for the feedback through the recurrent dominant (a-)loops (resonating at IAF). As a consequence, pre-synaptic (feedback) events (S2, see Fig 6A) have a higher likelihood to slightly precede the post-synaptic (tACS) events in these loops (see Fig 6B, bottom), leading to strengthening of their associated synapses. We emphasise that this model is based on a number of assumptions (see also [44]), including that 10Hz spike burst result from a 10Hz alternating current, and that the synaptic strengthening of the effective recurrent loops leads to an increase in natural a-activity. If these assumptions hold, this model matches our data, which show that slower stimulation (relative to IAF) enhances oscillations in the individual a - (here: faster) frequency.

It is important to note that assuming higher weights predicts greater effects at the resonance frequency of a person's dominant circuit when stimulated at nearby frequencies, but lesser or no effects at non-dominant frequencies. In other words, in a participant with a 10Hz a-peak, aftereffects would predominantly be observed at this intrinsic 10Hz frequency after stimulation at a nearby frequency (~10Hz), but no aftereffects should be observed at non-

intrinsic frequencies (e.g. 7Hz) with stimulation near these frequencies (~7Hz) (nor should there be 10Hz aftereffects after 7Hz stimulation). In addition, we point out that the assumption of higher weights for dominant oscillations also adds a factor of state-dependency to the model, which is in line with observations from NIBS studies using tACS [17,53,63,64], transcranial random noise stimulation (tRNS) [65], direct current stimulation (tDCS) [64,6669], and TMS [70,71], showing that the stimulation outcome is often dependent on the concurrent brain state or the task being executed.

Importantly this model not only predicts a-enhancement, but also a-suppression as a consequence of synaptic weakening in dominant a-loops when stimulation is applied at slightly faster frequencies relative to the spontaneous a-peak frequency (Fig 6C). This parallels classical STDP models in which synapses are strengthened when the postsynaptic potential (here: spiking of the driving neuron at tACS frequency) follows the presynaptic potential (here: the feedback to the driving neuron via the recurrent loop), and weakened when the order is reversed. This prediction needs to be verified experimentally.

tACS-induced plasticity: Patterned brain stimulation inducing long-term potentiation or depression

Long-term plasticity and associated effects on brain oscillations have been observed without fine-tuning the stimulation frequency to specific neuronal circuits. For instance, prolonged transcranial direct current stimulation (tDCS), which has no oscillatory component and whose effects have been associated with changes in excitability and synaptic efficacy [72-75], may also lead to enhanced a-activity [76-78]. Hence, other mechanisms than long-lasting STDP in specific reverberating circuits could explain the tACS-aftereffects observed here. For instance, aftereffects of both TMS and tDCS have been related to long term depression (LTD) and potentiation (LTP) [11,12,14,38,73,79,80] depending on parameters which do not

show any obvious link to intrinsic brain oscillations. These effects often manifest in cortical excitability changes. As posterior a-activity is taken to be an indicator of cortical excitability [81-83], offline a-changes could reflect these forms of LTD and LTP (but see [59]). In addition, it should be noted that overall metabolic or perfusion changes might be correlated with, and could possibly explain, excitability/a-changes [84-86]. Predictions derived from such periodicity-independent mechanisms would differ from the STDP account. Unlike with STDP, LTD or LTP should then occur to a similar extent for a broad range of stimulation protocols, such as reported for instance with repetitive TMS where LTD is associated with continuous low-frequency stimulation up to 1Hz and LTP with interleaved or patterned high-frequency stimulation across many frequencies (5-20Hz and iTBS) [87]. While our data provide evidence for the plasticity account, it cannot disambiguate between the above SDTP model and alternative mechanisms. Both computational and empirical research is needed to establish the existence and width of specific tACS-frequency windows that give rise to aftereffects and their relation to intrinsic brain oscillations.

Limitations of our study

Firstly, our design did not entail a condition with continuous stimulation, precluding a direct comparison between continuous and intermittent tACS aftereffects. It is therefore conceivable that continuous, but not intermittent, tACS leads to lasting entrainment given that in a typical tACS-protocol the brain oscillators are subjected to prolonged phase alignment over thousands of cycles. However, oscillatory phase in EEG recordings is generally instable over time, and as our data show, does not outlive tACS offset for more than 100ms, thus strengthening our conclusion that the aftereffect is predominantly a consequence of plastic changes. Secondly, we have no information about processes online to tACS. Nonetheless, there is growing evidence that entrainment during tACS is likely, and may even be a prerequisite (though not the underlying process) for plasticity effects (see models above). In

line with this view, Helfrich et al. [20] found that participants with greater a-power during tACS -- i.e. stronger entrainment -- also tended to show greater aftereffects. Thirdly, as in previous studies [20,24,44,53] comparisons to control frequencies are missing. Accordingly, it is unclear how frequency-specific the aftereffects are, although some insight on frequency-specificity can be derived from the observed variability in individual a-frequency with respect to a constant tACS frequency, with aftereffect magnitude being relatively unaffected by frequency mismatch. However, here the size of the mismatch was overall relatively small, and future studies need to clarify whether deviations (small or large) make a difference to outcomes. Moreover, we stimulated below, rather than above IAF. In the light of the STDP model, it will be interesting to determine if the direction of a (small) mismatch has a qualitative influence on the direction of the induced changes. Lastly, there is no data available whether the observed quantitative change in a-power has any functional significance. This needs to be tested through additional behavioural manipulations pre versus post-tACS.

Conclusion

Offline a-enhancement after a-tACS reflects short-term neural plasticity rather than entrained activity, although it is likely that mechanisms set in motion by online entrainment are prerequisite to such effects. This underlines the potential of tACS as a therapeutic tool. In addition, our findings may be informative for study-designs. Given that a-aftereffects were negligible with short trains (3 s) and participants overall tolerated the discontinuous stimulation well, intermittent event-related tACS paradigms with short trains could be viable tools in cognitive research on online tACS effects when potential confounds from aftereffects must be minimised.

Acknowledgment

This work was supported by a PhD Studentship from the College of Science and Engineering, University of Glasgow, to Alexandra Vossen, and a Wellcome Trust Award to Gregor Thut and Joachim Gross [grant number 098434, 098433].

References

[1] Arnal LH, Giraud A-L. Cortical oscillations and sensory predictions. Trends Cogn Sci 2012;16:390-8.

[2] Buzsaki G. Rhythms of the brain. New York: Oxford University Press; 2006.

[3] Giraud A-L, Poeppel D. Cortical oscillations and speech processing: emerging computational principles and operations. Nat Neurosci 2012;15:511-7.

[4] Jerbi K, Ossandon T, Hamame CM, Senova S, Dalal SS, Jung J, et al. Task-related gamma-band dynamics from an intracerebral perspective: review and implications for surface EEG and MEG. Hum Brain Mapp 2009;30:1758-71.

[5] Lopes da Silva F. EEG and MEG: relevance to neuroscience. Neuron 2013;80:1112-28.

[6] Schnitzler A, Gross J. Normal and pathological oscillatory communication in the brain. Nat Rev Neurosci 2005;6:285-96.

[7] Siegel M, Donner TH, Engel AK. Spectral fingerprints of large-scale neuronal interactions. Nat Rev Neurosci 2012;13:121-34.

[8] Singer W. Cortical dynamics revisited. Trends Cogn Sci 2013;17:616-26.

[9] Thut G, Miniussi C, Gross J. The functional importance of rhythmic activity in the brain. Curr Biol 2012;22:R658-63.

[10] Wang X. Neurophysiological and computational principles of cortical rhythms in cognition. Physiol Rev 2010;90:1195-268.

[11] Miniussi C, Ambrus GG, Walsh V, Antal A. Transcranial Magnetic and Electric Stimulation in Perception and Cognition Research. In: Miniussi C, Paulus W, Rossini PM, editors. Transcranial Brain Stimul. 1st ed., CRC Press; 2012, p. 337-58.

[12] Kuo M-F, Nitsche MA. Effects of transcranial electrical stimulation on cognition. Clin EEG Neurosci 2012;43:192-9.

[13] Miniussi C, Harris JA, Ruzzoli M. Modelling non-invasive brain stimulation in cognitive neuroscience. Neurosci Biobehav Rev 2013;37:1702-12.

[14] Dayan E, Censor N, Buch ER, Sandrini M, Cohen LG. Noninvasive brain stimulation: from physiology to network dynamics and back. Nat Neurosci 2013;16:838-44.

[15] Rosanova M, Casali A, Bellina V, Resta F, Mariotti M, Massimini M. Natural frequencies of human corticothalamic circuits. J Neurosci 2009;29:7679-85.

[16] Thut G, Schyns PG, Gross J. Entrainment of perceptually relevant brain oscillations by non-invasive rhythmic stimulation of the human brain. Front Psychol 2011;2:1-10.

[17] Herrmann CS, Rach S, Neuling T, Strüber D. Transcranial alternating current stimulation: a review of the underlying mechanisms and modulation of cognitive processes. Front Hum Neurosci 2013;7:279.

[18] Antal A, Paulus W. Transcranial alternating current stimulation (tACS). Front Hum Neurosci 2013;7:317.

[19] Feurra M, Paulus W, Walsh V, Kanai R. Frequency specific modulation of human somatosensory cortex. Front Psychol 2011;2:13.

[20] Helfrich RF, Schneider TR, Rach S, Trautmann-Lengsfeld SA, Engel AK, Herrmann CS. Entrainment of brain oscillations by transcranial alternating current stimulation. Curr Biol 2014;24:333-9.

[21] Kanai R, Chaieb L, Antal A, Walsh V, Paulus W. Frequency-dependent electrical stimulation of the visual cortex. Curr Biol 2008;18:1839-43.

[22] Kanai R, Paulus W, Walsh V. Transcranial alternating current stimulation (tACS) modulates cortical excitability as assessed by TMS-induced phosphene thresholds. Clin Neurophysiol 2010;121:1551-4.

[23] Laczo B, Antal A, Niebergall R, Treue S, Paulus W. Transcranial alternating stimulation in a high gamma frequency range applied over V1 improves contrast perception but does not modulate spatial attention. Brain Stimul 2012;5:484-91.

[24] Neuling T, Rach S, Wagner S, Wolters CH, Herrmann CS. Good vibrations: oscillatory phase shapes perception. Neuroimage 2012;63:771-8.

[25] Strüber D, Rach S, Trautmann-Lengsfeld S, Engel AK, Herrmann CS. Antiphasic 40 Hz oscillatory current stimulation affects bistable motion perception. Brain Topogr 2014;27:158-71.

[26] Marshall L, Helgadottir H, Mölle M, Born J. Boosting slow oscillations during sleep potentiates memory. Nature 2006;444:610-3.

[27] Meiron O, Lavidor M. Prefrontal oscillatory stimulation modulates access to cognitive control references in retrospective metacognitive commentary. Clin Neurophysiol 2014;125:77-82.

[28] Polania R, Nitsche MA, Korman C, Batsikadze G, Paulus W. The importance of timing in segregated theta phase-coupling for cognitive performance. Curr Biol 2012;22:1314-8.

[29] Brittain J-S, Probert-Smith P, Aziz TZ, Brown P. Tremor suppression by rhythmic transcranial current stimulation. Curr Biol 2013;23:436-40.

[30] Feurra M, Bianco G, Santarnecchi E, Del Testa M, Rossi A, Rossi S. Frequency-dependent tuning of the human motor system induced by transcranial oscillatory potentials. J Neurosci 2011;31:12165-70.

[31] Joundi RA, Jenkinson N, Brittain J-S, Aziz TZ, Brown P. Driving oscillatory activity in the human cortex enhances motor performance. Curr Biol 2012;22:403-7.

[32] Pogosyan A, Gaynor LD, Eusebio A, Brown P. Boosting cortical activity at Beta-band frequencies slows movement in humans. Curr Biol 2009;19:1637-41.

[33] Schutter DJLG, Hortensius R. Brain oscillations and frequency-dependent modulation of cortical excitability. Brain Stimul 2011;4:97-103.

[34] Wach C, Krause V, Moliadze V, Paulus W, Schnitzler A, Pollok B. Effects of 10 Hz and 20 Hz transcranial alternating current stimulation (tACS) on motor functions and motor cortical excitability. Behav Brain Res 2013;241:1-6.

[35] Sela T, Kilim A, Lavidor M. Transcranial alternating current stimulation increases risk-taking behavior in the balloon analog risk task. Front Neurosci 2012;6:22.

[36] Santarnecchi E, Polizzotto NR, Godone M, Giovannelli F, Feurra M, Matzen L, et al. Frequency-dependent enhancement of fluid intelligence induced by transcranial oscillatory potentials. Curr Biol 2013;23:1449-53.

[37] Voss U, Holzmann R, Hobson A, Paulus W, Koppehele-Gossel J, Klimke A, et al. Induction of self awareness in dreams through frontal low current stimulation of gamma activity. Nat Neurosci 2014:1-5.

[38] Brignani D, Ruzzoli M, Mauri P, Miniussi C. Is transcranial alternating current stimulation effective in modulating brain oscillations? PLoS One 2013;8:e56589.

[39] Kar K, Krekelberg B. Transcranial electrical stimulation over visual cortex evokes phosphenes with a retinal origin. J Neurophysiol 2012;108:2173-8.

[40] Vanneste S, Walsh V, Van De Heyning P, De Ridder D. Comparing immediate transient tinnitus suppression using tACS and tDCS: a placebo-controlled study. Exp Brain Res 2013;226:25-31.

[41] Wach C, Krause V, Moliadze V, Paulus W, Schnitzler A, Pollok B. The effect of 10 Hz transcranial alternating current stimulation (tACS) on corticomuscular coherence. Front Hum Neurosci 2013;7:511.

[42] Reato D, Rahman A, Bikson M, Parra LC. Effects of weak transcranial alternating current stimulation on brain activity-a review of known mechanisms from animal studies. Front Hum Neurosci 2013;7:687.

[43] Thut G, Miniussi C. New insights into rhythmic brain activity from TMS-EEG studies. Trends Cogn Sci 2009;13:182-9.

[44] Zaehle T, Rach S, Herrmann CS. Transcranial alternating current stimulation enhances individual alpha activity in human EEG. PLoS One 2010;5:e13766.

[45] Ali MM, Sellers KK, Fröhlich F. Transcranial alternating current stimulation modulates large-scale cortical network activity by network resonance. J Neurosci 2013;33:11262-75.

[46] Fröhlich F, McCormick DA. Endogenous electric fields may guide neocortical network activity. Neuron 2010;67:129-43.

[47] Ozen S, Sirota A, Belluscio MA, Anastassiou CA, Stark E, Koch C, et al. Transcranial electric stimulation entrains cortical neuronal populations in rats. J Neurosci 2010;30:11476-85.

[48] Deans JK, Powell AD, Jefferys JGR. Sensitivity of coherent oscillations in rat hippocampus to AC electric fields. J Physiol 2007;583:555-65.

[49] Reato D, Rahman A, Bikson M, Parra LC. Low-intensity electrical stimulation affects network dynamics by modulating population rate and spike timing. J Neurosci 2010;30:15067-79.

[50] Halbleib A, Gratkowski M, Schwab K, Ligges C, Witte H, Haueisen J. Topographic analysis of engagement and disengagement of neural oscillators in photic driving: a combined electroencephalogram/magnetoencephalogram study. J Clin Neurophysiol 2012;29:33-41.

[51] Herrmann CS. Human EEG responses to 1-100 Hz flicker: resonance phenomena in visual cortex and their potential correlation to cognitive phenomena. Exp Brain Res 2001;137:346-53.

[52] Pikovsky A, Rosenblum M, Kurths J. Synchronization: A universal concept in nonlinear sciences. 1st ed. New York, NY: Cambridge University Press; 2001.

[53] Neuling T, Rach S, Herrmann CS. Orchestrating neuronal networks: sustained aftereffects of transcranial alternating current stimulation depend upon brain states. Front Hum Neurosci 2013;7:161.

[54] Neuling T, Rach S, Wagner S, Wolters CH, Herrmann CS. Good vibrations: oscillatory phase shapes perception. Neuroimage 2012;63:771-8.

[55] Hanslmayr S, Matuschek J, Fellner M-C. Entrainment of Prefrontal Beta Oscillations Induces an Endogenous Echo and Impairs Memory Formation. Curr Biol 2014:1-6.

[56] Jaegle A, Ro T. Direct Control of Visual Perception with Direct Phase-specific Modulation of Posterior Parietal Cortex. J Cogn Neurosci 2013;26:422-32.

[57] Thut G, Veniero D, Romei V, Miniussi C, Schyns P, Gross J. Rhythmic TMS causes local entrainment of natural oscillatory signatures. Curr Biol 2011;21:1176-85.

[58] Thut G, Pascual-Leone A. A review of combined TMS-EEG studies to characterize lasting effects of repetitive TMS and assess their usefulness in cognitive and clinical neuroscience. Brain Topogr 2010;22:219-32.

[59] Veniero D, Brignani D, Thut G, Miniussi C. Alpha-generation as basic response-signature to transcranial magnetic stimulation (TMS) targeting the human resting motor cortex: A TMS/EEG co-registration study. Psychophysiology 2011;48:1381-9.

[60] Feldman DE. The spike-timing dependence of plasticity. Neuron 2012;75:556-71.

[61] Dan Y, Poo M. Spike Timing-Dependent Plasticity : From Synapse to Perception 2006:1033-48.

[62] Caporale N, Dan Y. Spike-timingdependent plasticity: A Hebbian learning rule. Annu Rev Neurosci 2008;31:25-46.

[63] Feurra M, Pasqualetti P, Bianco G, Santarnecchi E, Rossi A, Rossi S. State-dependent effects of transcranial oscillatory currents on the motor system: what you think matters. J Neurosci 2013;33:17483-9.

[64] Bortoletto M, Pellicciari MC, Rodella C, Miniussi C. The interaction with task-induced activity is more important than polarization: a tDCS study. Brain Stimul 2014.

[65] Snowball A, Tachtsidis I, Popescu T, Thompson J, Delazer M, Zamarian L, et al. Long-term enhancement of brain function and cognition using cognitive training and brain stimulation. Curr Biol 2013;23:987-92.

[66] Bradnam L V, Stinear CM, Lewis GN, Byblow WD. Task-dependent modulation of inputs to proximal upper limb following transcranial direct current stimulation of primary motor cortex. J Neurophysiol 2010;103:2382-9.

[67] Karok S, Witney AG. Enhanced motor learning following task-concurrent dual transcranial direct current stimulation. PLoS One 2013;8.

[68] Jones KT, Gozenman F, Berryhill ME. The strategy and motivational influences on the beneficial effect of neurostimulation: A tDCS and fNIRS study. Neuroimage 2015;105:238-47.

[69] Gill J, Shah PP, Hamilton R. It's the thought that counts: Examining the task-dependent effects of transcranial direct current stimulation on executive function. Brain Stimul 2014.

[70] Silvanto J. State-dependency of transcranial magnetic stimulation. Brain Topogr 2008;21:1-10.

[71] Bestmann S, Swayne O, Blankenburg F, Ruff CC, Haggard P, Weiskopf N, et al. Dorsal premotor cortex exerts state-dependent causal influences on activity in contralateral primary motor and dorsal premotor cortex. Cereb Cortex 2008;18:1281-91.

[72] Nitsche MA, Paulus W. Excitability changes induced in the human motor cortex by weak transcranial direct current stimulation. J Physiol 2000;527:633-9.

[73] Antal A, Paulus W, Nitsche MA. Electrical stimulation and visual network plasticity. Restor Neurol Neurosci 2011;29:365-74.

[74] Rahman A, Reato D, Arlotti M, Gasca F, Datta A, Parra LC, et al. Cellular effects of acute direct current stimulation: somatic and synaptic terminal effects. J Physiol 2013;591:2563-78.

[75] Nitsche MA, Antal A, Liebetanz D, Lang N, Tergau F, Paulus W. Neuroplasticity induced by transcranial direct current stimulation. In: Epstein CM, Wassermann EM, Ziemann U, editors. Oxford Handb. transcranial direct Curr. Stimul., New York, NY: Oxford University Press; 2008, p. 201-18.

[76] Spitoni GF, Cimmino RL, Bozzacchi C, Pizzamiglio L, Di Russo F. Modulation of spontaneous alpha brain rhythms using low-intensity transcranial direct-current stimulation. Front Hum Neurosci 2013;7:529.

[77] Hsu T-Y, Tseng P, Liang W-K, Cheng S-K, Juan C-H. Transcranial direct current stimulation over right posterior parietal cortex changes prestimulus alpha oscillation in visual short-term memory task. Neuroimage 2014.

[78] Puanhvuan D, Nojima K, Wongsawat Y, Iramina K. Effects of repetitive transcranial magnetic stimulation and transcranial direct current stimulation on posterior alpha wave. IEEJ Trans Electr Electron Eng 2013;8:263-8.

[79] Stagg CJ, Nitsche MA. Physiological basis of transcranial direct current stimulation. Neuroscienti st 2011;17:37-53.

[80] Ziemann U. TMS induced plasticity in human cortex. Rev Neurosci 2004;15:253-66.

[81] Romei V, Brodbeck V, Michel C, Amedi A, Pascual-Leone A, Thut G. Spontaneous fluctuations in posterior alpha-band EEG activity reflect variability in excitability of human visual areas. Cereb Cortex 2008;18:2010-8.

[82] Romei V, Rihs T, Brodbeck V, Thut G. Resting electroencephalogram alpha-power over posterior sites indexes baseline visual cortex excitability. Neuroreport 2008;19:203-8.

[83] Lange J, Oostenveld R, Fries P. Reduced occipital alpha power indexes enhanced excitability rather than improved visual perception. J Neurosci 2013;33:3212-20.

[84] Stagg CJ, Lin RL, Mezue M, Segerdahl A, Kong Y, Xie J, et al. Widespread modulation of cerebral perfusion induced during and after transcranial direct current stimulation applied to the left dorsolateral prefrontal cortex. J Neurosci 2013;33:11425-31.

[85] Antal A, Polania R, Schmidt-Samoa C, Dechent P, Paulus W. Transcranial direct current stimulation over the primary motor cortex during fMRI. Neuroimage 2011;55:590-6.

[86] Laufs H, Kleinschmidt A, Beyerle A, Eger E, Salek-Haddadi A, Preibisch C, et al. EEG-correlated fMRI of human alpha activity. Neuroimage 2003;19:1463-76.

[87] Rossi S, Hallett M, Rossini PM, Pascual-Leone A. Safety, ethical considerations, and application guidelines for the use of transcranial magnetic stimulation in clinical practice and research. Clin Neurophysiol 2009;120:2008-39.

[88] McDonnell M, Abbott D. What is stochastic resonance? Definitions, misconceptions, debates, and its relevance to biology. PLoS Comput Biol 2009;5.

Figure legend

Figure 1. Experimental design. A) Experimental setup and B) procedure. For details refer to section Procedure. C) Examples for the different tACS protocols. For details refer to section

Figure 2. Alpha-aftereffects across protocols. A) Mean relative increase (dB) in individual alpha band power from pre-test to post-test. Both long protocols are followed by a significantly higher alpha-increase compared to sham. Asterisks reflect significant pairwise comparisons using Wilcoxon Signed Rank Tests (a=.05). Only the respective comparisons between Sham and LongCo (lower brace), and Sham and LongDis (upper brace), were significant. B) Relative increase in mean power in the individual alpha band (individual stimulation frequency (ISF) ± 2Hz) from pre-test to post-test per participant. Each active stimulation condition is compared to Sham. Black lines represent individual differences between sham and active conditions, red line represents the mean difference. Most volunteers show a greater increase after stimulation with long (80 cycles at ISF) trains compared to sham.

Figure 3. Alpha-aftereffects relative to IAF and ISF. A) Individual stimulation frequency (ISF) relative to IAF. The distribution shows that there was a tendency to stimulate at a lower frequency than the "optimal" alpha frequency. B) IAF-aligned alpha aftereffects (difference between active protocols and sham) in mean relative power increase from pre-test to post-test (dB). Frequencies within the individual alpha band are defined by the individual alpha frequency (IAF) measured on the day of each session. The average increase tended to be stronger at IAF and above, i.e. slightly higher than at ISF. Error bars represent standard error of the mean.

Figure 4. Correlations between relative alpha increase and extent of the mismatch between individual stimulation frequency (ISF) and individual alpha frequency (IAF).

Data points to the left of the origin show sessions during which stimulation frequency was lower than the actual peak (established before each session). At least for the most effective protocol (LongCo), greater mismatch is associated with stronger alpha increase.

Figure 5. Alpha-effects in intermittent, tACS-free intervals. A) Mean relative increase (dB) in individual alpha band power for early (left) vs late (right) trials during silent periods between stimulation trains compared to pre-test. Grey outline shows mean increase between pre-and post-test for each condition (as shown in Figure 2A). B) Mean phase locking value across trials for early (left) vs late (right) trials. A value of 0 means no phase locking, a value of 1 means perfect phase locking. There is no evidence for enhanced phase locking. Error bars represent standard error of the mean.

Figure 6. Simplified STDP model of alpha aftereffects by tACS (adapted from [47]). A)

A population of neurons oscillating at alpha-frequencies. Recurrent loops within this population reverberate at different delays, leading to a net oscillatory frequency depending on which connection dominates. Dominant frequency can slowly fluctuate over time/days. In this example, delays of 100ms dominate, leading to a dominant 10Hz oscillation. B) and C). Stimulation by tACS. Some neurons are modulated by tACS (grey circles) while others are not (blue circles) (tACS effect is unlikely homogeneous across neuronal tissues and locations). Consider the synapse on the grey neurons. Events are triggered rhythmically by tACS (postsynaptic S1, assuming action potential generation shaped by stochastic resonance [88]). These events are then followed by presynaptic events (S2) generated through recurrent

loops at the delay of the dominant cycle (here 100ms). When neurons are stimulated at a frequency slightly slower than the dominant frequency of the loop (IAF) (B), presynaptic events slightly precede postsynaptic events of the next cycle, leading to strengthening of the synapse (LTP). Conversely, when neurons are stimulated at a frequency slightly faster than the dominant frequency (C), presynaptic events slightly follow postsynaptic events of the next cycle, leading to weakening of the synapse (LTD). Note that this model is speculative. While taking into account preferred frequency, and accounting for our data (B), it requires testing for situation (C).

A. Electrode positions

Change detection task

C. Procedure

Pretest rest EEG ISF determination tACS intensity (phosphenes) tACS/sham + visual task .^mi'illll Posttest rest EEG

"'WJil M

22-30 (depending on ISF)

t(min)

B. Intermittent tACS-protocols

1. ShortCo: Short/phase continuous

(30 ISF cycles on/off, 240 trains at continuous phase angle)

Virtual sine wave Actual tACS current

3. LongDis: Short/phase discontinuous

(80 ISF cycles on/off, 90 trains with change of phase angle)

ot 1.5 o

■X 1 JO

CD 0.5 it= CD

ShortCo-Sham LongCo-Sham LongDis-Sham

Stimulation frequency (ISF) mismatch relative to individual alpha frequency (IAF)

IAF-2 IAF-1.5 IAF-1 IAF-.5 IAF IAF+.5

Frequency relative to IAF

IAF+1.5

CO ■o

O CD it= CD

ShortCo-Sham LongCo-Sham LongDis-Sham

o 2 -1.5 -1 -0.5 ( --1 ♦ ♦ -2 -2 - i i i i i i 0.5 1 1.5 2 2.5 3 ♦

-3 > A

Stimulation frequency (ISF) mismatch relative to individual alpha frequency (IAF)

early intermittent Sham late intermittent

* i ShortCo

■ LongCo

■ LongDis Post

B tACS < IAF

S1: postsynaptic

S2: presynaptic

C tACS > IAF

iii iii

AAA/ AAA/V

Highlights

■ Intermittent periodic stimulation of occipital areas with alpha-tACS enhances offline EEG alpha power.

■ Alpha-aftereffects cannot be explained by neuronal entrainment but are more likely due to plastic changes.

■ We propose a physiological constraint to a recent model of tACS-induced spike-timing dependent plasticity.

Alpha power increase after transcranial alternating current stimulation at alpha-frequency (a-tACS) reflects plastic changes rather than entrainment (Supplementary methods)

Alexandra Vossen3, Joachim Grossb, Gregor Thutb

aSchool of Psychology & bInstitute of Neuroscience and Psychology

University of Glasgow

58 Hillhead Street

Glasgow G12 8QB

United Kingdom

a.vossen.1@research.gla.ac.uk (A. Vossen) Joachim.Gross@glasgow.ac.uk (J. Gross) Gregor.Thut@glasgow.ac.uk (G. Thut)

Corresponding authors: A. Vossen (Tel: +44 (0)141 330 5151), G. Thut (Tel: +44(0)141 330 3395)

Supplementary Methods

Staircase procedure to determine individual tACS intensity

Intensity was determined by administering 80 tACS cycles at ISF with increasing intensity from .75mA peak-to-peak (pp) (at which all volunteers reported no or very weak sensations) in steps of .25mA up to 2mA/pp (maximum current density .002857mA/cm2, [1]) or until the person reported phosphenes or perceived the stimulation as too uncomfortable. In this case intensity was decreased by .1mA/pp until no phosphenes were detected and until the stimulation was acceptable to the participant.

Visual Change Detection Task

To ensure that participants remain vigilant they performed a slow change detection task of the same duration as the stimulation protocol. Stimuli were presented at low frequency and low saliency to minimise interference of visual processing with induced alpha activity. Stimuli were also temporally uncorrelated with tACS on/off-periods. Volunteers were asked to maintain fixation on a white cross centrally presented on a grey background. A red circle of similar luminance as the background (diameter 30pxl) was presented in the lower central visual field (Presentation software, version 16.3, Neurobehavioral Systems, Albany, US). Participants were asked to respond to a colour change from red to green (duration 150ms) by mouse click as quickly as possible. Target events occurred after intervals of between 2.5-4.5min length. After each third of trials a break of 45s was inserted to allow participants to move and blink. While the stimulation protocol was continued during this pause, these trials, and trials containing a target event, were not included in the analysis.

Choice of non-parametric statistics

While the increase data were fairly normally distributed, there were two outliers in the LongDis condition and one in the sham condition (criterion: greater or smaller than 1.5 times the interquartile range), two of which remained outliers when ignoring stimulation condition (see Supplemental Fig. S1, left part). The data points moreover belonged to three different participants. In addition, verifying non-normality (for instance by using Kolmogorov-Smirnov or Shapiro-Wilk's W tests for normality)

in a small sample is not very reliable because of low power, which is why non-parametric test are usually considered the "safer" option in small samples. We therefore used non-parametric tests, which do not require that distributions meet stringent criteria.

log-normalised alpha power increase

Difference tACS vs Sham

ShortCo LongCo LongDis: Sham

ShortCo-Sham LongCo-Sham LongDis - Sham

Supplemental Figure SI. Individual results. Left: Alpha increase from pre- to post-test in dB per participant. Each participant is represented by the same symbol in all conditions. Outliers are marked with a red plus sign, horizontal line at zero dB reflects no change. Right: Difference score (alpha increase in active condition minus increase after sham) in dB. Note that most participants showed relative greater alpha activity post stimulation in both long (LongCo, LongDis) conditions compared to sham.