Content uploaded by Ivan Alekseichuk
Author content
All content in this area was uploaded by Ivan Alekseichuk on Oct 06, 2017
Content may be subject to copyright.
1
Intrahemispheric Theta Rhythm Desynchronization Impairs Working
Memory
1
Ivan Alekseichuk1,*, Stefanie Corinna Pabel1, Andrea Antal1, Walter Paulus1
1Department of Clinical Neurophysiology, University Medical Center Göttingen, Georg-August
University Göttingen, Göttingen, Germany
*Correspondence: ivan.alekseichuk@med.uni-goettingen.de
Abstract
Background: There is a growing interest in large-scale connectivity as one of the crucial factors
in working memory. Correlative evidence has revealed the anatomical and
electrophysiological players in the working memory network, but understanding of the
effective role of their connectivity remains elusive. Objective: In this double-blind, placebo-
controlled study we aimed to identify the causal role of theta phase connectivity in visual-
spatial working memory. Methods: The frontoparietal network was over- or de-synchronized
in the anterior-posterior direction by multi-electrode, 6 Hz transcranial alternating current
stimulation (tACS). Results: A decrease in memory performance and increase in reaction time
was caused by frontoparietal intrahemispheric desynchronization. According to the diffusion
drift model, this originated in a lower signal-to-noise ratio, known as the drift rate index, in
the memory system. The EEG analysis revealed a corresponding decrease in phase
connectivity between prefrontal and parietal areas after tACS-driven desynchronization. The
over-synchronization did not result in any changes in either the behavioral or
electrophysiological levels in healthy participants. Conclusion: Taken together, we
demonstrate the feasibility of manipulating multi-site large-scale networks in humans, and
the disruptive effect of frontoparietal desynchronization on theta phase connectivity and
visual-spatial working memory.
The final publication is available at IOS Press through http://dx.doi.org/10.3233/RNN-160714
2
1. Introduction
Successful voluntary behavior demands timely activation and cooperation of multiple brain
regions that together form the large-scale networks. Inherent precision and coordination
within the neural networks are essential for the compound cognitive processes, such as
working memory. The human working memory is subdivided into visual-spatial and
phonological domains (Baddeley, 2012). Functional neuroanatomical studies show that
visual-spatial working memory, of interest in this present study, arises in the bilateral parietal
and prefrontal areas (Barbey, Colom, Paul, & Grafman, 2014; Müller & Knight, 2006). There is
discussion regarding the neurophysiological organization. One view allocates working
memory-related computations to the prefrontal cortex, responsible for the operations with
items and cognitive control, and memory storage area to the posterior parietal cortex (Nee
et al., 2013; Todd & Marois, 2004). The other view is that both cortexes are self-coordinating
and interacting during working memory processes, functioning as a single frontoparietal
network and sharing the responsibility (Constantinidis & Klingberg, 2016; Ester, Sprague, &
Serences, 2015). Electrophysiological measurements demonstrated involvement of
oscillatory activity in the theta and gamma bands (Hsieh & Ranganath, 2014; Roux, Wibral,
Mohr, Singer, & Uhlhaas, 2012) and theta-gamma coupling (Alekseichuk, Turi, Amador de
Lara, Antal, & Paulus, 2016; Canolty & Knight, 2010; Lisman & Jensen, 2013) during the
memory task. Commonly, weak gamma rhythms are interpreted as the local activity of fast
spiking interneurons (Buzsáki & Wang, 2012; Cardin et al., 2009), while the theta oscillations
are driven by the distant pacemakers, reflecting remote communications (Colgin, 2013;
Fujisawa & Buzsáki, 2011). This theta activity was also found to be the primary coordinator
between the areas in the frontoparietal network, mainly through the mechanism of phase
synchronization (Fell & Axmacher, 2011; Salazar, Dotson, Bressler, & Gray, 2012). The multi-
scale graph analysis of phase connectivity during the visual cognitive task showed
reorganization within the frontoparietal and occipital networks, including increase in
clustering and rich club coefficients, predominantly in the theta band (Bola & Sabel, 2015).
Nevertheless, the causal role of theta phase connectivity and the degree to which it is driving
visual working memory function is not yet clear. Active manipulation by transcranial
alternating current stimulation (tACS) makes it possible to enhance the power of brain
oscillations and entrain their phase, thus, allowing us to observe the consequences of
3
alternative brain states (Antal et al., 2008; Antal & Herrmann, 2016; Thut, Miniussi, & Gross,
2012). Both power modulation and phase entrainment are frequency-specific and spatially
restricted in their primary target by the shape of the applied electrical field. The modification
of local electrical landscape with tACS affects the neuronal communications through the
mechanism of ephaptic coupling (Anastassiou & Koch, 2015; Anastassiou, Perin, Markram, &
Koch, 2011). The feasibility of such manipulations on the neural level was demonstrated by
using the in silica and in vitro models (Deans, Powell, & Jefferys, 2007; Ozen et al., 2010;
Reato, Rahman, Bikson, & Parra, 2010). It was shown that an electrical field as weak as 0.2
mV/mm can influence local neural processes. Furthermore the pioneering
electrophysiological in vivo recordings in primates and humans demonstrated the possibility
of generating such electrical field in the brain tissues during tACS (Opitz et al., 2016). From
the biophysical perspective, due to the phase entrainment, injection of in-phase alternating
current in two or more oscillating systems should lead to their synchronization, and injection
of anti-phase current – to desynchronization. The synchronization between two brain regions
can be achieved, if biologically plausible, by at least three electrodes, whereby two electrodes
would act in-phase and the third – in anti-phase as the “return” electrode. On the other hand,
the power modulation depends only on the electrical field density and the neurophysiological
state before and during the stimulation.
In the present study we aimed to investigate the causal role of theta connectivity for working
memory by utilizing tACS. Since the theta rhythm arises from the subcortico-cortical
interactions, and subcortical structures cannot be selectively stimulated with existing non-
invasive techniques, this study focused on the neocortical, frontoparietal theta connectivity.
It has already been shown that desynchronization of the left prefrontal cortex versus the left
parietal cortex by tACS negatively impacted reaction time in the verbal working memory task,
while their synchronization versus the central medial brain area improved the reaction time
(Polania, Nitsche, Korman, Batsikadze, & Paulus, 2012). Other researchers reported that
desynchronization of the medial prefrontal versus medial parietal cortexes led to a poorer
verbal working memory performance in one case (Chander et al., 2016), and to a higher score
in the digital span test in another (Vosskuhl, Huster, & Herrmann, 2015). The latter study,
however, utilized the individual stimulation protocols, where the frequency was tailored to
the subject specific theta peak. Thus, manipulations using theta oscillations can affect working
4
memory, and the outcomes depend on the anatomical target, stimulation parameters and
the cognitive task. Nevertheless, two conceptual limitations became clear from the preceding
experiments. First, in every study only one or two regions were taken into account, leaving
questions open regarding possible phase interactions within a multisite brain network.
Second, previous studies utilized a two-electrode montage or used different electrode
montages between the conditions, thereby undermining definitive differentiation between
the power modulation and phase de/synchronization effects. Here, for the first time, we
entrained the theta phase connectivity between four neocortical sites that are hosting visual-
spatial working memory. That is, the left and right prefrontal and posterior parietal areas. We
hypothesized that (i) frontoparietal synchronization – simultaneous synchronization of left
prefrontal with left parietal cortexes and right prefrontal with right parietal cortexes, will
optimize the cognitive processes during the working memory task, while (ii) the frontoparietal
desynchronization will negatively affect performance. Further we introduced the
computational behavioral model – the diffusion drift model, in order to evaluate the signal-
to-noise ratio in the memory system and response conservativeness due to the frontoparietal
de/synchronization (Ratcliff, 1978; Ratcliff, Smith, Brown, & McKoon, 2016). The diffusion
drift model assumes that a simple forced-choice decision, such as recognition of a memory
item, is made within a noisy neural process that accumulates information over time from a
starting point toward one of the response boundaries. The deterministic part of this noisy
process, the drift rate, can be defined as the speed of evidence accumulation or as an index
for the signal-to-noise ratio in the system. For the memory task, the drift rate represents the
quality of match between the stimulus and probe. The response boundaries are characterize
the conservativeness or amount of evidences needed for the decision on the recognition of
memory item and therefore reflect the cognitive strategy regarding the speed-accuracy
preference. Also the model allows to extract the part of reaction time that do not relate to
the evidence accumulation process – the non-decision time. We expected that the
manipulation with frontoparietal phase connectivity will affect the informational transfer
and, as the result, the signal-to-noise ratio within the working memory network, but not the
response conservativeness or non-decision time.
5
2. Material and Methods
2.1. Participants
Thirty-five healthy, adult volunteers with normal or corrected-to-normal vision, without a
history of neurological or psychiatric diseases took part in the study. Ten volunteers (6
females, 24 ± 2 years old, 16.7 ± 1.8 years of education) were assigned to the natural history
group and twenty-five participants (13 females, 23.5 ± 2.9 years old, 16.6 ± 2.1 years of
education) formed the experimental group. All volunteers gave informed consent in
accordance with the Declaration of Helsinki and regulation of the Ethics Committee of Georg-
August University Göttingen.
2.2. Experimental Procedure
The volunteers participated in three sessions, performing visual-spatial working memory
tests. They were naive regarding the study a priori and were familiarized with the laboratory
and setting during the standardized introductory meeting. Participants in the natural history
group (Natural) received no intervention during any of the three sessions. Their performance
was used to assess the cognitive responses on the working memory task in the absent of real
or placebo effects and to investigate the conceivable learning effect over three sessions. The
experimental group took part in the placebo-controlled, randomized, crossover, double-blind
study which included three conditions: frontoparietal synchronization (Sync), frontoparietal
desynchronization (Desync) and sham control (Placebo). TACS was applied for the whole test
period (17-19 minutes) and was framed with the EEG recording.
2.3. Working Memory Task
The participants performed the two-back visual-spatial working memory task during each
session (Fig. 1). The task comprised two blocks of trials with a one-minute break between the
blocks. Each block contained 100 trials and required 8-9 minutes to complete. Within every
trial there were stimulus and response periods. The stimulus consisted from four dots of the
same shape and size randomly distributed across the visual field. The color altered between
purple and blue from trial to trial to facilitate the memorization in two-back fashion.
6
Fig. 1. Working memory task. During the session, the volunteer performed the two-back visual-spatial working
memory test. The test comprised two 100-trials blocks. Each trial consisted of the stimulus (0.6 sec) and the
response period (up to 2.5 sec), which were separated by the variable intervals (1-2 sec). The participants
indicated whether the recent stimulus matched the two-back item by pressing the “yes” or “no” button on the
response pad.
In the response period, volunteers were requested to press the “yes” or “no” button,
indicating whether the stimulus was in the same position as two trials before. The order of
trials was pseudo-randomized across the sessions. The amount of matching and non-
matching trials per session was equal. The test was generated in PsychoPy (Peirce, 2008).
Responses were acquired by using the response pad RB-740 (Cedrus™). The test produced
two primary types of data: the responses and the reaction times. All responses across the
single session were analyzed to obtain the individual condition-specific correct hit and false
alarm rates. Further, these two parameters were combined in the working memory
performance score by subtracting the false alarm rate from the correct hit rate. The reaction
time data were log-transformed to correct for the right skewed distribution before the
statistical analysis.
To further investigate the behavioral differences between the conditions that reflected in
both the memory performance and in the reaction time, we implemented the diffusion drift
model. The fundamental idea behind the model is that prior to the response in an alternative
task the participant is accumulating evidence in favor of one or another decision. This
accumulation process has a time cost which contributes to the total reaction time. When the
amount of accumulated evidence reaches a certain critical threshold, the answer is given.
There are multiple mathematical approaches to model such a decision-making process which
vary in complexity and number of free parameters. Here we implemented the Robust-EZ drift
model (Wagenmakers, van der Maas, Dolan, & Grasman, 2008; Wagenmakers, van der Maas,
& Grasman, 2007). It is the analytical solution that combines relative simplicity and high
7
precision, as compared to more parameter-rich methods (van Ravenzwaaij & Oberauer,
2009). Three main parameters were estimated in our case. The drift rate (v) indicates the
speed of evidence accumulation or, in other words, information uptake per time. The
boundary separation (a) describes the amount of evidence needed for the decision. And the
non-decision time (Ter) specifies the part of reaction time in which no evidence accumulation
is occurring. Mathematically, the Robust-EZ procedure begins with fitting the reaction time
data to the mixed model with two components: the uniform distribution of response
contaminants and exponentially modified Gaussian distribution that captures the process of
interest. The latter is the linear combination of exponential distribution with rate τ and normal
distribution with mean µ and standard deviation σ. The mixed distribution is fitted by using
the maximum likelihood procedure with quasi-Newton routine for optimization and method-
of-moments estimation of starting values:
where M1, M2, and M3 are the moments of first, second, and third order, respectively. Then,
the reaction time mean (MRTEG) and variance (VRTEG) can be recovered as the functions of ex-
Gaussian distribution:
Finally, the Robust-EZ procedure introduces the proportion of correct responses (Pc) and
scaling parameter s (by default, s = 0.1) to determine the drift rate (v), boundary separation
(a), and non-decision time (Ter) as follows:
8
Statistical differences at the group level were assessed with the bootstrap t-test. This test
estimates the probability that the observed numerical difference between two groups comes
from the null hypothesis distribution, which is obtained by bootstrapping (resampling with
replacement) the dataset. Bootstrapping overcomes many limitations inherent in the classical
t-test, such as assumption regarding the data distribution, yet keeps high statistical power
and intuitive interpretability. The effect size was estimated in terms of Cohen’s d and reported
together with 95% confidence intervals. The order effect was tested by fitting the linear
regression to the data with “session order” as a categorical predictor.
2.4. Transcranial Alternating Current Stimulation (tACS)
Transcranial electrical stimulation and electrophysiological recording were performed using
the combined device StarStim (Neuroelectrics™). Four stimulation electrodes were positioned
over the AF3, AF4, P3, and P4 points according to the international 10-10 system (Fig. 2A, B).
The stimulation was delivered via Ag/AgCl electrodes (r = 1 cm) with electroconductive gel
during the test procedure, which lasted about 18 minutes, including 10 s fade-in and fade-out
periods. Small stimulation electrodes provided higher precision of electric field, while the skin
sensations were not inflated in comparison to larger stimulation electrodes (Turi et al., 2014).
The alternating current was applied with the frequency of 6 Hz and the intensity of 1 mA peak-
to-baseline. Impedance was kept below 10 kOhm.
Frontoparietal synchronization (Sync): The electrodes AF3 and P3 from one side, and AF4 and
P4 from the contra-lateral side were in-phase. Thus, the prefrontal and parietal electrodes
were synchronized in pairs, while left and right hemisphere electrodes were in anti-phase to
ensure the same amount of inward and outward current to the brain.
Frontoparietal desynchronization (Desync): The prefrontal electrodes AF3 and AF4, and the
parietal electrodes P3 and P4, were in phase. In this configuration, the prefrontal and parietal
electrodes were desynchronized in pairs.
Sham control (Placebo): Control intervention was delivered with the same intensity and via
the same electrode montage (AF3, AF4, P3, and P4) as for the real stimulation conditions, but
only for 10 s at the beginning and end of the session, according to fade-in/fade-out placebo
9
protocol (Ambrus et al., 2012). The phase relationships between the stimulation electrodes
were randomized.
Fig. 2. Transcranial alternating current stimulation. (A-B) Multi-electrode montage for the conditions of
frontoparietal desynchronization (Desync) and frontoparietal synchronization (Sync). The electrodes were
positioned over the points AF3, AF4, P3, and P4 according to the international 10-10 EEG system. Alternating
current was delivered during the working memory task via all four electrodes at a frequency of 6 Hz and an
intensity of 1 mA peak-to-baseline. The phase of the current is color-coded. (C-D) Realistic, anisotropic, finite
element models of the current flow. The field strength is color-coded from 0 (dark blue) to 0.55 mV/mm (dark
red).
To estimate the electrical field distribution in the brain, we calculated the realistic,
anisotropic, finite element head models (Fig. 2C, D). The following compartments were
modeled by using SimNIBS 2 (Thielscher, Antunes, & Saturnino, 2015): scalp (σ = 0.465 S/m),
bone (σ = 0.010 S/m), cerebrospinal fluid (σ = 1.654 S/m), gray matter (σ = 0.275 S/m), and
white matter (σ = 0.126 S/m). Post-processing and visualization were done in Gmsh (Geuzaine
& Remacle, 2009).
2.5. Electroencephalography (EEG)
Eyes-open, resting state EEG was recorded before and after each stimulation and test
procedure for three minutes against the reference electrode on the right earlobe. Eight
Ag/AgCl electrodes were positioned over the Fpz, AF3, AF4, C3, C4, P3, P4, and Oz points
10
according to the international 10-10 system. The sampling rate was equal to 500 Hz at an
analog-digital precision of 24 bits. Impedance was kept below 10 kOhm.
EEG was analyzed in MATLAB 2016a with the Fieldtrip toolbox (Oostenveld, Fries, Maris, &
Schoffelen, 2011). First, the recordings were high-pass filtered (> 1 Hz) with an 8th-order
forward-backward Butterworth filter. The line noise (50 and 100 Hz) was suppressed by using
the discrete Fourier transform filter. The data were then divided into 2-s-long trials with 50%
overlap. The trials with abnormal within-channel variance were visually inspected and
removed if a technical or muscle artifact was confirmed. On average 13.6% of trials, between
6.5% and 17.2%, were excluded. After that, we utilized the independent component analysis.
One component, related to eye movements, was rejected. The preprocessing was finished
after the low-pass filtering (< 30 Hz) with the 8th-order forward-backward Butterworth filter.
Phase connectivity was estimated for every pair of electrodes before and after the stimulation
and test procedure. Changes between pre- and post-stimulation EEG connectivity are known
as reliable and characteristic reflection of neurophysiological brain state in healthy humans
and patients (Alekseichuk et al., 2016; Bola et al., 2014; Gall et al., 2016). First, the frequency
analysis was performed using the Hanning tapers in the range from 1 to 30 Hz. The data were
separated on the frequency bands 1-4 Hz (delta), 4-8 Hz (theta), 8-12 Hz (alpha), 12-20 Hz
(low beta), and 20-30 Hz (high beta). Then, the weighted phase lag indexes (wPLI) were
calculated for every trial and frequency band, and averaged to obtain the individual,
condition-specific, pre- and post-stimulation connectivity values. This metric is known as
robust, volume conduction and noise insensitive estimator of functional connectivity in the
human brain (Vinck, Oostenveld, Van Wingerden, Battaglia, & Pennartz, 2011). It depends
only on the imaginary part of coherence and debiased with respect to the sample size:
Where average weight
and the weight .
The pairwise product of thresholding . Xj and Xk is
complex-valued cross-spectrums of i-th and k-th trial between 1 and N.
11
Any increase or decrease in phase connectivity after the stimulation, as compared to the pre-
stimulation state, was assessed for every condition by using the bootstrap t-test (10,000
iterations). To compensate for multiple comparisons, all tests were corrected within each
condition according to the false discovery rate procedure (Benjamini & Yekutieli, 2001).
Significant changes (pcorr ≤ 0.05) in phase connectivity were visualized in BrainNet Viewer (Xia,
Wang, & He, 2013).
3. Results
First, we analyzed the working memory performance and reaction time changes (Fig. 3A, B).
Interestingly, both control stimulation (Placebo) and frontoparietal synchronization (Sync)
resulted in the similar performance score (72.8 ± 2.3% and 71.9 ± 2.0%, accordingly) and
reaction time score (5.881 ± 0.030 and 5.890 ± 0.033). Same values were observed for the
natural history group (Natural; performance: 72.1 ± 1.5% and reaction time: 5.887 ± 0.031).
However, the frontoparietal desynchronization (Desync) led to the decreased memory
performance (69.0 ± 2.4%) and slower reaction time (5.930 ± 0.035). Significant differences
were observed between Desync and Placebo conditions (performance: bootstrap t-test p =
0.018, Cohen’s d = 0.41 [0.05 – 0.78]; reaction time: p = 0.015, d = 0.35 [0.01 – 0.80]), and
between Desync and Sync conditions (performance: p = 0.019, d = 0.37 [0.03 – 0.89]; reaction
time: p = 0.049, d = 0.27 [-0.03 – 0.75]), but not between Synch and Placebo conditions
(performance: p = 0.282, d = 0.10 [-0.30 – 0.53]; reaction time: p = 0.339, d = 0.06 [-0.40 –
0.42]). Taking the crossover study design, we utilized the linear model to control for the
learning effect within the natural history group (performance: R2 = 0.05, p = 0.24; reaction
time: R2 = 0.09, p = 0.25) and session order effect within the experimental group
(performance: R2 = 0.01, p = 0.76; reaction time: R2 = 0.01, p = 0.71). Hence, learning and
session order had no significant impact on the group level statistics.
Further, we introduced the diffusion drift model to account for the simultaneous changes in
working memory performance and reaction time, and investigate what behavioral aspects
were modified (Fig. 3C, D, E).
12
Fig. 3. Behavioral results. The outcomes of the working memory tests for the natural history group (Natural, n =
10) and experimental group (n = 25). The latter experienced the following conditions in a randomized order:
frontoparietal synchronization (Synch), frontoparietal desynchronization (Desync), and sham control (Placebo).
All statistical comparisons were conducted within the experimental group by using the bootstrap t-test and
depicted as horizontal brackets (* – p ≤ 0.05, n.s. – p > 0.05). All values are presented as mean ± SEM. (A) Working
memory performance score. (B) Log-normalized reaction time. (C-E) Main parameters of the diffusion drift
model. According to the model, prior to the participant’s response to the question, he/she is accumulating
evidence in favor of one or another alternative answer. The drift rate (v) shows the deterministic component of
a noisy evidence accumulation process, and can be treated as the speed of accumulation or signal-to-noise ratio.
The boundary separation (a) indicates the conservativeness, i.e. the amount of accumulated evidence needed
to make a response. Finally, the non-decision time (Ter) denotes the reaction time excluding the evidence-
accumulation process.
13
The drift rate (v) was impaired due to Desync stimulation (v = 0.249 ± 0.015) in statistical
comparison with Placebo condition (v = 0.273 ± 0.014; bootstrap t-test p = 0.031, Cohen’s d
= 0.47 [0.09 – 0.97]) as well as Sync condition (v = 0.275 ± 0.013; p = 0.017, d = 0.34 [0.04 –
0.78]), while two latter conditions demonstrated no differences (p = 0.44, d = 0.02 [-0.39 –
0.43]) and were numerically similar to Natural group (v = 0.270 ± 0.011). At the same time,
the boundary separation (a) did not differ between Placebo (a = 0.075 ± 0.004) and other
conditions: Desync (a = 0.077 ± 0.004; p = 0.16, d = 0.18 [-0.19 – 0.73]), Sync (a = 0.073 ±
0.004; p = 0.19, d = 0.19 [-0.22 – 0.62]), and Natural (a = 0.076 ± 0.003). The nondecision time
(Ter) also was not modulated by any stimulation protocols in numerical comparison to Natural
group (Ter = 0.250 ± 0.009) and statistical comparison to Placebo condition (Ter = 0.242 ±
0.006): both Desync (Ter = 0.245 ± 0.009; p = 0.29, d = 0.18 [-0.19 – 0.72]) and Sync
interventions (Ter = 0.247 ± 0.011, p = 0.20, d = 0.19 [-0.21 – 0.61]) are characterized by the
same values. Once again, the session order had no detectable impact on any parameter (v: R2
= 0.03, p = 0.39; a: R2 = 0.01, p = 0.68; Ter: R2 = 0.01, p = 0.87).
We estimated the phase connectivity difference maps in the theta range between the post-
and pre-stimulation states for every condition (Fig. 4). Significant decrease in theta
connectivity between the prefrontal (AF3, Fpz, and AF4) and parietal (P3 and P4) areas was
observed after Desync stimulation.
Fig. 4. Phase connectivity difference maps. The
phase connectivity was estimated in the
frequency band between 4 Hz and 8 Hz for the
electroencephalographic data that were
recorded before and after every experimental
session. In the figure, all electrodes that showed
a significant decrease in phase coupling are
linked with blue lines, and an increase – with
green lines (bootstrap t-test, p ≤ 0.05, FDR
corrected). Orange spheres depict the
stimulation and recording electrodes, and yellow
spheres depict recording-only electrodes.
14
Altogether, 21.4% of electrode-pairs (6 out of 28) indicated the lower weighted phase lag
index (bootstrap t-test p ≤ 0.05, FDR corrected). At the same time, no modification of phase
connectivity was introduced by Placebo, and only minor changes accompanied Sync
stimulation, where one pair of electrodes (C4-P4) showed an increase in theta connectivity.
No significant post-stimulation changes in phase connectivity (bootstrap t-test p ≤ 0.05, FDR
corrected) were observed for other brain rhythms, namely, the delta, alpha, low beta and
high beta frequencies.
4. Discussion
In this study we explore the effects of frontoparietal intrahemispheric synchronization and
desynchronization on visual-spatial working memory. Contrary to expectations, we did not
find a significant cognitive improvement during the synchronization between the left
prefrontal with left parietal cortexes, and right prefrontal with right parietal cortexes
(condition Sync). However, the second hypothesis could be proven: desynchronization
between prefrontal and posterior parietal cortexes (Desync) has a negative impact on both
working memory performance and reaction time. In agreement with the behavioral data, the
EEG analysis showed a decrease in theta phase connectivity between the anterior and
posterior electrodes only after Desync stimulation, and no aftereffects in any other cases.
Previously, Polania and colleagues (Polania et al., 2012) reported a faster or slower reaction
time in the verbal working memory task due to synchronizing or desynchronizing tACS
between the left hemispheric prefrontal and parietal cortexes. We not only reproduced the
impairment in reaction time, but also found a lower performance score due to our
bihemispheric desynchronization. The positive effect of frontoparietal synchronization was
not reproduced, possibly because of a different electrode montage in the previous study that
targeted not just frontal-parietal, but frontal-parietal-central areas (electrodes F3-P3 and
return over Cz, according to the international 10-10 system), leading to a differently shaped
electrical field. Here we avoided this problem by using four stimulation electrodes and
switching the current between them, which allowed to keep the electrode montage same for
every condition. Chander and colleagues (Chander et al., 2016) applied tACS to target the
frontal midline theta activity via the electrodes over the medial prefrontal versus medial
parietal areas (Fpz versus Pz). As a result, the stimulation produced a frontoparietal
15
desynchronization, which negatively affected verbal working memory performance and,
according to the ongoing magnetoencephalography, reduced midline theta activity. However,
they did not examined the effect of frontoparietal synchronization.
Applying the diffusion drift modeling approach, in agreement with the prediction we
observed a lower drift rate index under the desynchronizing tACS with unaltered non-decision
time and boundary separation. The latter is interesting as it is the main mechanism by which
the speed-accuracy trade-off can arise (Wagenmakers et al., 2007). Thereby, we can
hypothesize that mild behavioral impairment in our study was not the consequence of
different response strategy, but the result of a lower quality of information, possibly due to
the disturbed encoding or comparison between the items. From the other side, no changes
in the non-decision time together with longer overall reaction time during Desync stimulation
argue in favor of weakened working memory computations rather than delays in other
physiological processes. The specific decrease in the drift rate index was also reported by
Philiastides et al. in connection with a poorer performance in a visual discriminative task due
to low-frequency repetitive transcranial magnetic stimulation (TMS) over the left prefrontal
cortex (Philiastides, Auksztulewicz, Heekeren, & Blankenburg, 2011). In line with modern
assumptions that TMS disturbs the connectivity of large scale networks, one can speculate
about the link between the stable large-scale neural networks and the drift rate index in the
working memory processes.
Taking together, our analysis shows that desynchronizing theta tACS over the prefrontal
versus parietal cortexes (Desync) worsens working memory performance and increases
reaction time due to the decline in the information uptake (i.e. drift rate), which is also
reflected in the outlasting theta rhythm desynchronization in the neocortex. At the same
time, the sham stimulation (Placebo) and frontoparietal synchronization (Sync) induced no
significant changes on the performance or electrophysiological levels. It should be
emphasized that the conditions Sync and Desync led to these distinct results by utilizing
exactly the same power, latency, and frequency of tACS with the only difference in the AC
phase. Therefore, possible online effect of power modulation per se was either not present
or overtaken by the effect of phase entrainment; in any case, it can be ruled out as a
mechanism behind the observed cognitive effect. Furthermore, the double-blind nature of
the experimental procedure ensured that the impact of subjective perception and
16
expectations by the study volunteers and investigator was kept at the minimum. Together
these points exclude any alternative explanation for the observed results and reinforce the
hypothesis of phase entrainment within the frontoparietal network.
Further interpreting the modification of theta phase connectivity, one limitation of the
present study should be mentioned. By concentrating on multiple brain regions and their
connectivity, we tackle the intricate functional relationships, because the frontoparietal
synchronization in our study can be seen as interhemispheric desynchronization, and vice
versa (Fig. 2A, B). That is why the interpretation of stimulation protocols and their outcomes
sought and found extra support in a priori expectation that theta connectivity in the anterior-
posterior direction is more prominent and thus more susceptible to the manipulation than
the between-hemispheric connectivity (Sauseng, Klimesch, Schabus, & Doppelmayr, 2005;
Summerfield & Mangels, 2005). This was further confirmed by the present EEG data, which
indicates an impact on the phase connectivity between the prefrontal and parietal electrodes
in Desync condition, but not between the left and right electrodes, supporting the anterior-
posterior view of this intervention. Our assumptions and results are also in line with the
influential hypothesis regarding visual working memory architecture that suggests a degree
of independence for informational processing and capacity limits between left and right
hemispheres (Franconeri, Alvarez, & Cavanagh, 2013; Umemoto, Drew, Ester, & Awh, 2010).
This originates in the unilateral activation of left or right occipital cortex by right- or left-sided
visual information.
The present findings offer new insights on the role of theta phase connectivity as the naturally
well-adjusted functional substrate for visual-spatial working memory. While the long-range
theta coherence has already been identified as the key phenomenon for working memory
function (Fell & Axmacher, 2011; Lisman, 2010), new results suggest that the frontoparietal
theta oscillations in healthy humans are an optimized system. This optimized system
preferentially reacts to disturbances, leading to impairment of cognitive function. However,
having our data in mind, we can speculate that the performance of the system cannot be that
easily improved if it is already in the optimal state (Furuya, Klaus, Nitsche, Paulus, & Altenmu,
2014). It still remains to be seen whether humans with minor cognitive impairment will
improve in response to synchronizing tACS. It may be interesting to consider the present
results in a broader sense as a piece in the framework of “communication through coherence”
17
in the brain (Fries, 2015). Phase connectivity may have variable roles depending on the
network state − hence continuous enhancement of frontoparietal synchronization may not
be the best way to facilitate working memory processes. Future investigations will have an
opportunity to clarify the impact of state-specific synchronization, e.g. by using closed-loop
stimulation during the encoding, maintenance, or retrieval memory stages. With this first
attempt to coordinate as many as four brain areas with tACS, we provide – in addition to new
data on the working memory processes − good experimental evidence supporting the
feasibility of multi-site phase entrainment. Further research may overcome the necessity to
have anti-phase electrodes/areas and improve the current setting by introducing multiple
Laplacian electrode arrays (Helfrich et al. 2014).
From the clinical perspective, both de- and over-synchronization within the large-scale brain
networks are associated with numerous neuropsychiatric disorders, such as schizophrenia,
anxiety, depression, and autism (reviewed in Menon 2011; Voytek and Knight 2015). While it
is straightforward that the lack of between-area coordination can lead to impairment of
network-based neural functions, over-synchronization can do no less harm by diminishing the
informational transfer due to the lower computational dimensionality (i.e. lower “number of
channels” for communication). Therefore the method of soft, large-scale desynchronization
holds out definite promise for clinical practice.
In conclusion, we demonstrate that manipulation with theta phase connectivity between the
bihemispheric prefrontal and parietal areas is feasible, that tACS-induced frontoparietal
desynchronization in healthy humans decreases phase connectivity, and that a deficit in
frontoparietal theta phase connectivity impairs visual-spatial working memory.
Acknowledgments
IA conceptualized the study. IA, AA, and WP designed the experiments. SCP collected the data.
IA analyzed the data. AA and WP supervised the study. IA, SCP, AA, and WP wrote the paper.
We thank Dr. Zsolt Turi for inspiring discussions regarding the present study and Christine
Crozier for proofreading the manuscript. This work was supported by the Deutsche
Forschungsgemeinschaft (SPP 1665 to WP).
18
References
Alekseichuk, I., Turi, Z., Amador de Lara, G., Antal, A., & Paulus, W. (2016). Spatial Working
Memory in Humans Depends on Theta and High Gamma Synchronization in the Prefrontal
Cortex. Current Biology, 26(12), 1513–1521. https://doi.org/10.1016/j.cub.2016.04.035
Ambrus, G. G., Al-Moyed, H., Chaieb, L., Sarp, L., Antal, A., & Paulus, W. (2012). The fade-in –
Short stimulation – Fade out approach to sham tDCS – Reliable at 1 mA for naïve and
experienced subjects, but not investigators. Brain Stimulation, 5(4), 499–504.
https://doi.org/10.1016/j.brs.2011.12.001
Anastassiou, C. A., & Koch, C. (2015). Ephaptic coupling to endogenous electric field activity:
Why bother? Current Opinion in Neurobiology, 31, 95–103.
https://doi.org/10.1016/j.conb.2014.09.002
Anastassiou, C. A., Perin, R., Markram, H., & Koch, C. (2011). Ephaptic coupling of cortical
neurons. Nature Neuroscience, 14(2), 217–223. https://doi.org/10.1038/nn.2727
Antal, A., Boros, K., Poreisz, C., Chaieb, L., Terney, D., & Paulus, W. (2008). Comparatively
weak after-effects of transcranial alternating current stimulation (tACS) on cortical excitability
in humans. Brain Stimulation, 1(2), 97–105. https://doi.org/10.1016/j.brs.2007.10.001
Antal, A., & Herrmann, C. S. (2016). Transcranial Alternating Current and Random Noise
Stimulation: Possible Mechanisms. Neural Plasticity, 2016, 1–12.
https://doi.org/10.1155/2016/3616807
Azevedo, F. A. C., Carvalho, L. R. B., Grinberg, L. T., Farfel, J. M., Ferretti, R. E. L., Leite, R. E. P.,
… Herculano-Houzel, S. (2009). Equal numbers of neuronal and nonneuronal cells make the
human brain an isometrically scaled-up primate brain. The Journal of Comparative Neurology,
513(5), 532–541. https://doi.org/10.1002/cne.21974
Baddeley, A. (2012). Working Memory: Theories, Models, and Controversies. Annual Review
of Psychology, 63(1), 1–29. https://doi.org/10.1146/annurev-psych-120710-100422
Barbey, A., Colom, R., Paul, E., & Grafman, J. (2014). Architecture of fluid intelligence and
working memory revealed by lesion mapping. Brain Structure & Function, 219(2), 485–94.
https://doi.org/10.1007/s00429-013-0512-z
Bélanger, M., Allaman, I., & Magistretti, P. J. (2011). Brain Energy Metabolism: Focus on
19
Astrocyte-Neuron Metabolic Cooperation. Cell Metabolism, 14(6), 724–738.
https://doi.org/10.1016/j.cmet.2011.08.016
Benjamini, Y., & Yekutieli, D. (2001). The control of the false discovery rate in multiple testing
under depencency. The Annals of Statistics, 29(4), 1165–1188.
https://doi.org/10.1214/aos/1013699998
Bola, M., Gall, C., Moewes, C., Fedorov, A., Hinrichs, H., & Sabel, B. A. (2014). Brain functional
connectivity network breakdown and restoration in blindness. Neurology, 83(6), 542–551.
https://doi.org/10.1212/WNL.0000000000000672
Bola, M., & Sabel, B. A. (2015). Dynamic reorganization of brain functional networks during
cognition. NeuroImage, 114, 398–413. https://doi.org/10.1016/j.neuroimage.2015.03.057
Bullmore, E., & Sporns, O. (2012). The economy of brain network organization. Nature
Reviews Neuroscience, 13, 336–349. https://doi.org/10.1038/nrn3214
Buzsáki, G., Anastassiou, C. A., & Koch, C. (2012). The origin of extracellular fields and currents
— EEG, ECoG, LFP and spikes. Nature Reviews Neuroscience, 13, 407–420.
https://doi.org/10.1038/nrn3241
Buzsáki, G., & Wang, X.-J. (2012). Mechanisms of Gamma Oscillations. Annual Review of
Neuroscience, 35(1), 203–225. https://doi.org/10.1146/annurev-neuro-062111-150444
Canolty, R., & Knight, R. (2010). The functional role of cross-frequency coupling. Trends in
Cognitive Sciences, 14(11), 506–15. https://doi.org/10.1016/j.tics.2010.09.001
Caporale, N., & Dan, Y. (2008). Spike Timing–Dependent Plasticity: A Hebbian Learning Rule.
Annual Review of Neuroscience, 31(1), 25–46.
https://doi.org/10.1146/annurev.neuro.31.060407.125639
Cardin, J. A., Carlén, M., Meletis, K., Knoblich, U., Zhang, F., Deisseroth, K., … Moore, C. I.
(2009). Driving fast-spiking cells induces gamma rhythm and controls sensory responses.
Nature, 459(7247), 663–667. https://doi.org/10.1038/nature08002
Chander, B. S., Witkowski, M., Braun, C., Robinson, S. E., Born, J., Cohen, L. G., … Soekadar, S.
R. (2016). tACS Phase Locking of Frontal Midline Theta Oscillations Disrupts Working Memory
Performance. Frontiers in Cellular Neuroscience, 10(May), 1–10.
https://doi.org/10.3389/fncel.2016.00120
20
Colgin, L. (2013). Mechanisms and Functions of Theta Rhythms. Annual Review of
Neuroscience, 36(May), 295–312. https://doi.org/10.1146/annurev-neuro-062012-170330
Constantinidis, C., & Klingberg, T. (2016). The neuroscience of working memory capacity and
training. Nature Reviews Neuroscience, 17(7), 438–449. https://doi.org/10.1038/nrn.2016.43
Deans, J. K., Powell, A. D., & Jefferys, J. G. R. (2007). Sensitivity of coherent oscillations in rat
hippocampus to AC electric fields. The Journal of Physiology, 583(Pt 2), 555–565.
https://doi.org/10.1113/jphysiol.2007.137711
Ester, E. F., Sprague, T. C., & Serences, J. T. (2015). Parietal and Frontal Cortex Encode
Stimulus- Specific Mnemonic Representations during Visual Working Memory. Neuron, 87(4),
893–905. https://doi.org/10.1016/j.neuron.2015.07.013
Fell, J., & Axmacher, N. (2011). The role of phase synchronization in memory processes.
Nature Reviews Neuroscience, 12(2), 105–118. https://doi.org/10.1038/nrn2979
Franconeri, S. L., Alvarez, G. A., & Cavanagh, P. (2013). Flexible cognitive resources:
competitive content maps for attention and memory. Trends in Cognitive Sciences, 17(3),
134–141. https://doi.org/10.1016/j.tics.2013.01.010
Fries, P. (2015). Rhythms for Cognition: Communication through Coherence. Neuron, 88(1),
220–235. https://doi.org/10.1016/j.neuron.2015.09.034
Fujisawa, S., & Buzsáki, G. (2011). A 4 Hz Oscillation Adaptively Synchronizes Prefrontal, VTA,
and Hippocampal Activities. Neuron, 72(1), 153–165.
https://doi.org/10.1016/j.neuron.2011.08.018
Furuya, S., Klaus, M., Nitsche, M. A., Paulus, W., & Altenmu, E. (2014). Ceiling Effects Prevent
Further Improvement of Transcranial Stimulation in Skilled Musicians. Journal of
Neuroscience, 34(41), 13834–13839. https://doi.org/10.1523/JNEUROSCI.1170-14.2014
Gall, C., Schmidt, S., Schittkowski, M. P., Antal, A., Ambrus, G. G., Paulus, W., … Sabel, B. A.
(2016). Alternating Current Stimulation for Vision Restoration after Optic Nerve Damage: A
Randomized Clinical Trial. PLOS ONE, 11(6), e0156134.
https://doi.org/10.1371/journal.pone.0156134
Geuzaine, C., & Remacle, J.-F. (2009). Gmsh: A 3-D finite element mesh generator with built-
in pre- and post-processing facilities. International Journal for Numerical Methods in
21
Engineering, 79(11), 1309–1331. https://doi.org/10.1002/nme.2579
Helfrich, R. F., Knepper, H., Nolte, G., Strüber, D., Rach, S., Herrmann, C. S., … Engel, A. K.
(2014). Selective Modulation of Interhemispheric Functional Connectivity by HD-tACS Shapes
Perception. PLoS Biology, 12(12), e1002031. https://doi.org/10.1371/journal.pbio.1002031
Hsieh, L.-T., & Ranganath, C. (2014). Frontal midline theta oscillations during working memory
maintenance and episodic encoding and retrieval. NeuroImage, 85, 721–729.
https://doi.org/10.1016/j.neuroimage.2013.08.003
Lisman, J. (2010). Working Memory: The Importance of Theta and Gamma Oscillations.
Current Biology, 20(11), R490–R492. https://doi.org/10.1016/j.cub.2010.04.011
Lisman, J., & Jensen, O. (2013). The Theta-Gamma Neural Code. Neuron, 77(6), 1002–1016.
https://doi.org/10.1016/j.neuron.2013.03.007
Menon, V. (2011). Large-scale brain networks and psychopathology: a unifying triple network
model. Trends in Cognitive Sciences, 15(10), 483–506.
https://doi.org/10.1016/j.tics.2011.08.003
Müller, N. G., & Knight, R. T. (2006). The functional neuroanatomy of working memory:
Contributions of human brain lesion studies. Neuroscience, 139(1), 51–58.
https://doi.org/10.1016/j.neuroscience.2005.09.018
Nee, D., Brown, J., Askren, M., Berman, M., Demiralp, E., Krawitz, A., & Jonides, J. (2013). A
meta-analysis of executive components of working memory. Cerebral Cortex, 23(2), 264–82.
https://doi.org/10.1093/cercor/bhs007
Oostenveld, R., Fries, P., Maris, E., & Schoffelen, J.-M. (2011). FieldTrip: Open source software
for advanced analysis of MEG, EEG, and invasive electrophysiological data. Computational
Intelligence and Neuroscience, 2011, 1–9. https://doi.org/10.1155/2011/156869
Opitz, A., Falchier, A., Yan, C., Yeagle, E. M., Linn, G. S., Megevand, P., … Schroeder, C. E.
(2016). Spatiotemporal structure of intracranial electric fields induced by transcranial electric
stimulation in humans and nonhuman primates. Scientific Reports, 6, 31236.
https://doi.org/10.1038/srep31236
Ozen, S., Sirota, A., Belluscio, M., Anastassiou, C. A., Stark, E., & Koch, C. (2010). Transcranial
Electric Stimulation Entrains Cortical Neuronal Populations in Rats. The Journal of
22
Neuroscience, 30(34), 11476–11485. https://doi.org/10.1523/JNEUROSCI.5252-09.2010
Peirce, J. W. (2008). Generating stimuli for neuroscience using PsychoPy. Frontiers in
Neuroinformatics, 2(January), 1–8. https://doi.org/10.3389/neuro.11.010.2008
Philiastides, M. G., Auksztulewicz, R., Heekeren, H. R., & Blankenburg, F. (2011). Causal role
of dorsolateral prefrontal cortex in human perceptual decision making. Current Biology,
21(11), 980–983. https://doi.org/10.1016/j.cub.2011.04.034
Polania, R., Nitsche, M., Korman, C., Batsikadze, G., & Paulus, W. (2012). The importance of
timing in segregated theta phase-coupling for cognitive performance. Current Biology, 22(14),
1314–8. https://doi.org/10.1016/j.cub.2012.05.021
Ratcliff, R. (1978). A theory of memory retrieval. Psychological Review, 85(2), 59–108.
https://doi.org/10.1037/0033-295X.85.2.59
Ratcliff, R., Smith, P. L., Brown, S. D., & McKoon, G. (2016). Diffusion Decision Model: Current
Issues and History. Trends in Cognitive Sciences, 20(4), 260–281.
https://doi.org/10.1016/j.tics.2016.01.007
Reato, D., Rahman, A., Bikson, M., & Parra, L. (2010). Low-Intensity Electrical Stimulation
Affects Network Dynamics by Modulating Population Rate and Spike Timing. The Journal of
Neuroscience, 30(45), 15067–15079. https://doi.org/10.1523/JNEUROSCI.2059-10.2010
Roux, F., Wibral, M., Mohr, H., Singer, W., & Uhlhaas, P. (2012). Gamma-Band Activity in
Human Prefrontal Cortex Codes for the Number of Relevant Items Maintained in Working
Memory. The Journal of Neuroscience, 32(36), 12411–12420.
https://doi.org/10.1523/JNEUROSCI.0421-12.2012
Salazar, R. F., Dotson, N. M., Bressler, S. L., & Gray, C. M. (2012). Content-Specific Fronto-
Parietal Synchronization During Visual Working Memory. Science, 338(6110), 1097–1100.
https://doi.org/10.1126/science.1224000
Sauseng, P., Klimesch, W., Schabus, M., & Doppelmayr, M. (2005). Fronto-parietal EEG
coherence in theta and upper alpha reflect central executive functions of working memory.
International Journal of Psychophysiology, 57(2), 97–103.
https://doi.org/10.1016/j.ijpsycho.2005.03.018
Summerfield, C., & Mangels, J. A. (2005). Coherent theta-band EEG activity predicts item-
23
context binding during encoding. NeuroImage, 24(3), 692–703.
https://doi.org/10.1016/j.neuroimage.2004.09.012
Thielscher, A., Antunes, A., & Saturnino, G. (2015). Field modeling for transcranial magnetic
stimulation: A useful tool to understand the physiological effects of TMS? In IEEE EMBS 2015,
Milano, Italy. IEEE. https://doi.org/10.1109/EMBC.2015.7318340
Thut, G., Miniussi, C., & Gross, J. (2012). The functional importance of rhythmic activity in the
brain. Current Biology, 22(16), R658–R663. https://doi.org/10.1016/j.cub.2012.06.061
Todd, J. J., & Marois, R. (2004). Capacity limit of visual short-term memory in human posterior
parietal cortex. Nature, 428(6984), 751–754. https://doi.org/10.1038/nature02466
Turi, Z., Ambrus, G. G., Ho, K. A., Sengupta, T., Paulus, W., & Antal, A. (2014). When size
matters: Large electrodes induce greater stimulation-related cutaneous discomfort than
smaller electrodes at equivalent current density. Brain Stimulation, 7(3), 460–467.
https://doi.org/10.1016/j.brs.2014.01.059
Umemoto, A., Drew, T., Ester, E. F., & Awh, E. (2010). A bilateral advantage for storage in
visual working memory. Cognition, 117(1), 69–79.
https://doi.org/10.1016/j.cognition.2010.07.001
van Ravenzwaaij, D., & Oberauer, K. (2009). How to use the diffusion model: Parameter
recovery of three methods: EZ, fast-dm, and DMAT. Journal of Mathematical Psychology,
53(6), 463–473. https://doi.org/10.1016/j.jmp.2009.09.004
Vinck, M., Oostenveld, R., Van Wingerden, M., Battaglia, F., & Pennartz, C. M. A. (2011). An
improved index of phase-synchronization for electrophysiological data in the presence of
volume-conduction, noise and sample-size bias. NeuroImage, 55(4), 1548–1565.
https://doi.org/10.1016/j.neuroimage.2011.01.055
Vosskuhl, J., Huster, R. J., & Herrmann, C. S. (2015). Increase in short-term memory capacity
induced by down-regulating individual theta frequency via transcranial alternating current
stimulation. Frontiers in Human Neuroscience, 9(May), 257.
https://doi.org/10.3389/fnhum.2015.00257
Voytek, B., & Knight, R. T. (2015). Dynamic network communication as a unifying neural basis
for cognition, development, aging, and disease. Biological Psychiatry, 77(12), 1089–1097.
24
https://doi.org/10.1016/j.biopsych.2015.04.016
Wagenmakers, E.-J., van der Maas, H. L. J., Dolan, C. V, & Grasman, R. P. P. P. (2008). EZ does
it! Extensions of the EZ-diffusion model. Psychonomic Bulletin & Review, 15(6), 1229–1235.
https://doi.org/10.3758/PBR.15.6.1229
Wagenmakers, E.-J., van der Maas, H. L. J., & Grasman, R. P. P. P. (2007). An EZ-diffusion model
for response time and accuracy. Psychonomic Bulletin & Review, 14(1), 3–22.
https://doi.org/10.3758/BF03194023
Xia, M., Wang, J., & He, Y. (2013). BrainNet Viewer: A Network Visualization Tool for Human
Brain Connectomics. PLoS ONE, 8(7). https://doi.org/10.1371/journal.pone.0068910
