ArticlePDF Available

Low frequency transcranial electrical stimulation does not entrain sleep rhythms measured by human intracranial recordings


Abstract and Figures

Transcranial electrical stimulation has widespread clinical and research applications, yet its effect on ongoing neural activity in humans is not well established. Previous reports argue that transcranial alternating current stimulation (tACS) can entrain and enhance neural rhythms related to memory, but the evidence from non-invasive recordings has remained inconclusive. Here, we measure endogenous spindle and theta activity intracranially in humans during low-frequency tACS and find no stable entrainment of spindle power during non-REM sleep, nor of theta power during resting wakefulness. As positive controls, we find robust entrainment of spindle activity to endogenous slow-wave activity in 66% of electrodes as well as entrainment to rhythmic noise-burst acoustic stimulation in 14% of electrodes. We conclude that low-frequency tACS at common stimulation intensities neither acutely modulates spindle activity during sleep nor theta activity during waking rest, likely because of the attenuated electrical fields reaching the cortical surface.
No caption available
This content is subject to copyright. Terms and conditions apply.
Low frequency transcranial electrical stimulation
does not entrain sleep rhythms measured by
human intracranial recordings
Belen Lafon 1, Simon Henin 2,3, Yu Huang 1, Daniel Friedman 2,3, Lucia Melloni 2,3,4,
Thomas Thesen 3,5, Werner Doyle2,6, György Buzsáki 3,7, Orrin Devinsky 2,3, Lucas C. Parra 1&
Anli A. Liu 2,3
Transcranial electrical stimulation has widespread clinical and research applications, yet its
effect on ongoing neural activity in humans is not well established. Previous reports argue
that transcranial alternating current stimulation (tACS) can entrain and enhance neural
rhythms related to memory, but the evidence from non-invasive recordings has remained
inconclusive. Here, we measure endogenous spindle and theta activity intracranially in
humans during low-frequency tACS and nd no stable entrainment of spindle power during
non-REM sleep, nor of theta power during resting wakefulness. As positive controls, we nd
robust entrainment of spindle activity to endogenous slow-wave activity in 66% of electrodes
as well as entrainment to rhythmic noise-burst acoustic stimulation in 14% of electrodes. We
conclude that low-frequency tACS at common stimulation intensities neither acutely mod-
ulates spindle activity during sleep nor theta activity during waking rest, likely because of the
attenuated electrical elds reaching the cortical surface.
Corrected: Author correction
DOI: 10.1038/s41467-017-01045-x OPEN
1Department of Biomedical Engineering, City College of New York, 160 Convent Ave, New York, NY 10031, USA. 2New York University Comprehensive
Epilepsy Center, 223 East 34th Street, New York, NY 10016, USA. 3Department of Neurology, New York University School of Medicine, 240 East 38th St,
20th Floor, New York, NY 10016, USA. 4Department of Neuroscience, Max Planck Institute for Empirical Aesthetics, Gruneburgweg 14, 60322 Frankfurt am
Main, Germany. 5Department of Physiology and Neuroscience, St. Georges University, St. Georges, Grenada. 6Department of Neurosurgery NYU School of
Medicine, 530 1st Avenue, Suite 7W, New York, NY 10016, USA. 7New York University Neuroscience Institute, 450 East 29th St, New York, NY 10016, USA.
Belen Lafon and Simon Henin contributed equally to this work. Lucas C. Parra and Anli Liu jointly supervised this work. Correspondence and requests for
materials should be addressed to A.Liu. (email:
NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x | 1
Content courtesy of Springer Nature, terms of use apply. Rights reserved
The therapeutic potential of transcranial electrical stimula-
tion (TES) has been examined in over 70 diverse condi-
tions, including major depression, epilepsy, pain, stroke
rehabilitation, Parkinsons disease, and tinnitus1. TES is motivated
by the well-established biophysical observation that externally
applied electric elds can affect neuronal excitability24. However,
behavioral effects in human are often weak and difcult to repli-
cate, which underscores a lack of basic mechanistic understanding
of how applied electrical elds interact with brain activity59.
Here, we focus on transcranial alternating current stimulation
(tACS)10, which has potential to affect neural oscillations relevant
to normal cognition and neurological disease11. Specic oscilla-
tion frequencies characterize different arousal states and cognitive
processes, and oscillations coordinate activity between local and
distant brain regions12,13. For example, in non-REM (NREM)
sleep, neocortical slow-wave activity coordinates thalamocortical
sleep spindles14,15. It has been argued that low-frequency tACS,
applied during NREM in healthy human subjects, can boost
associative memory, presumably by entraining these nested
rhythms1621. Furthermore, low-frequency tACS applied during
waking rest has resulted in widespread increases in theta activity,
associated with improved encoding22. It has been proposed that
tACS applied at the dominant network oscillations may entrain
these endogenous rhythms in the gamma2325, beta26, alpha27,28,
theta29, and slow frequencies16,30,31.
In this study, we tested a widely cited protocol16, which showed
that declarative memory performance could be enhanced by
applying low-frequency tACS during NREM sleep. This work has
since inspired a number of replication studies, which have shown
either positive results18,19,21 or no effects5,7,32,33. To validate the
physiological underpinnings of this promising memory effect, we
applied low-frequency tACS during sleep while measuring neural
responses intracranially in surgical epilepsy patients. We test, in
particular, the widely held belief that brain rhythms are entrained
by tACS applied at the frequency that matches the endogenous
rhythm. In the case of NREM sleep this is particularly advanta-
geous as one can measure spindle and gamma activity, which is
strongly coupled to the slow wave, yet is uncontaminated by the
low-frequency stimulation artifact34,35.
Thirteen patients undergoing invasive monitoring for epilepsy
surgery were tested with sinusoidal tACS (0.75 and 1 Hz) during a
period of NREM sleep and/or waking rest. The primary outcome
measure during sleep is phase-amplitude coupling (PAC), which
captures the modulation of spindle amplitude (at 10 and 14 Hz)
with the phase of tACS. This measure is motivated by the relevance
of spindles to sleep-dependent cognitive processes, the strong
entrainment of spindles to native slow oscillations36,andthepre-
vious reports of tACS enhancement of endogenous slow-wave and
spindle activity16,17,20,37,38. Our primary outcome measure during
wakefulness is PAC of theta frequency (7 Hz) amplitude to tACS
phase, as theta power has previously been implicated in tACS22.
Secondary outcome measures include gamma (70110 Hz) mod-
ulation during NREM sleep39 and alpha (10 Hz) and gamma
modulation during waking rest. We also tested on an additional
subject the conguration of Marshall et al.16 with identical electrode
location, charge density, waveform, frequency, and intensity.
Finally, we measured the intensity of induced electric elds at the
cortical surface, and leveraged validated computational models to
estimate eld strengths across the brain40.Wedonotnd an acute
effect of the applied tACS on brain rhythms and attribute the
outcome to the weak-induced elds.
As a positive control, we conrm that spindle power is
modulated with the endogenous slow-wave rhythm during sleep
in the same subjects across a majority of electrodes. Additionally,
we nd that acoustic stimulation with brief noise bursts reliably
evokes slow-wave and related spindle activity comparable to
effects found in healthy subjects using scalp electro-
encephalography (EEG)41. The null ndings on entrainment
together with these positive controls rule out the hypothesis that
low-frequency tACS applied at conventional current intensities
can acutely entrain slow-wave, spindle, gamma, or theta activity.
We conclude that previous reports of behavioral effects of slow
oscillating tACS applied during NREM sleep and waking rest on
memory consolidation are not the result of direct neuronal
entrainment. We discuss alternative mechanisms and propose
new directions for research.
Intracranial measurements of induced electric elds.We
applied low-frequency sinusoidal tACS (0.75, 1 Hz), on 13
patients with medication-refractory epilepsy with implanted
RH S14
Grid Strip Depth
RH S12
Fig. 1 Electrode coverage for 18 patients who underwent invasive monitoring for epilepsy surgery and received either tACS or acoustic stimulation. Thirteen
patients had tACS applied at a 1 or 0.75 Hz frequency, at stimulation intensities ranging from 0.5 to 2 mA, during waking rest (S1S6) and daytime NREM
sleep (S7S13), one patient had trapezoidal tACS (S14), and four subjects received 0.75 and 1 Hz acoustic stimulation (A1A4). Electrode placement varied
by clinical indication, and consisted of a combination of strips, grids, and depth electrodes. Seven subjects had bilateral coverage (S1, S2, S3, S6, S8, S10,
S13, S14). A total of 2156 electrodes total were utilized (1700 tACS; 113 Trapezoidal tACS; 343 acoustic), or an average of 120 electrodes per subject.
Further demographic and clinical characteristics, electrode coverage, and stimulation protocols are summarized in Supplementary Table 1
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x
2NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
subdural and depth electrodes (Fig. 1; Methods; Supplementary
Table 1). Overall, we recorded and analyzed signals from 1700
electrodes (mean 131 electrodes per subject) from the 13 patients
receiving TES using this protocol. The positions of the recording
electrodes sites in each patient are shown in Fig. 1. For stimula-
tion, two scalp electrodes (2 cm × 2 cm rubber electrodes) were
placed over the frontopolar and occipital poles, accessible under
the surgical bandage. The aim was to generate the strongest
possible elds in the brain by keeping electrodes at the furthest
distance on the skull42, while remaining apart from temporal
craniotomies so as to replicate the current ow of normal head
anatomy43. We conrmed that the craniotomy defect had only
minor effects on eld intensities as models of current ow with
and without the defect yield comparable eld intensities (0.05 ±
0.02 V/m and 0.04 ±0.02 V/m, respectively, mean ±Std across
subjects). In one patient (S10), the stimulation electrodes were
placed at three additional locations to examine the effects of
electrode placement on extracerebral and intracerebral current
spread. For three subjects (subjects S11S13), frontal electrodes
were offset to the left or right (Fp1 or Fp2) so that intracranial
electrode coverage, which was largely constrained to one hemi-
sphere, covered areas of maximal electric elds and thereby
optimize recording of potential entrainment effects. Maximal
stimulation intensity of tACS was adjusted for each subject based
on amplier saturation and patient sensation. Intensity ranged
from 0.5 to 2.5 mA (current density 0.1250.625 mA mA/cm2),
which is stronger than in previous reports5,7,16,18,19,32.
Since the electrical eld, or the voltage gradient (V/m), induced
in cortex critically determines neuronal polarization44 and thus
neuronal activity4,45, we measured the local gradients in
electrodes across all subdural and depth electrodes. We calculated
the electric elds in the direction of neighboring electrodes
(i.e., the projected eld; Fig. 2a) scaled to correspond to 1 mA of
stimulation (voltages scale linearly with applied currents up to
saturation levels of the ampliers40). The median projected eld
measured across all electrode locations and subjects is 0.02 V/m.
At the highest current intensity tested (2.5 mA), peak intensity
reaches only 0.16 V/m across recording electrodes (S10).
To assess the induced electric elds over brain areas not
sampled by electrode recordings and in the direction of maximal
eld intensity, we used previously established modeling techni-
ques40. Briey, computational models were individualized for
each patient based on their magnetic resonance imaging (MRI)
images and calibrated with the measured electric elds in each
patient. The models were then used to predict eld intensities
throughout the brain. For one patient (subject S13), electric eld
distribution for four different stimulation electrode congura-
tions were estimated. The simulated electric elds predict the
measured elds well (r=0.81 ±0.12, mean and standard
deviation across N=10 subjects modeled, all p<105). The
modeling results show larger elds for some locations of the brain
as compared to the recorded elds (Fig. 2b, c), as expected given
the limited electrode coverage. Maximal eld magnitudes at
hippocampal depth electrodes have a median of 0.05 mV/mm
and a maximum of 0.11 V/m, for 2 mA stimulation intensity
(N=8 subjects). Median value across all the brains and electrode
locations is 0.08 V/m (Fig. 2).
Spindles entrain to the phase of endogenous slow waves. Of the
13 subjects with tACS, 7 were stimulated with several 5 or 10-min
blocks at 0.75 and 1 Hz during either daytime or nocturnal
NREM sleep (Supplementary Table 1). Four subjects were
stimulated during daytime NREM sleep; three were stimulated
during nocturnal NREM sleep; and six were stimulated during
waking rest. Duration and conditions of stimulation for each
subject varied based on clinical constraints and the total time in
NREM sleep (sleep stages N2 and N3, American Academy of
Sleep Medicine 2012 convention46).
Sleep data from the seven subjects (subjects S7S13) were
analyzed over two separate nights without stimulation. We rst
visually identied periods of NREM sleep in the intracranial
recordings. The average duration of the NREM recorded was
30.0 ±10.5 min. Slow-wave activity spans broad frequency
spectrum between 0.5 and 4 Hz with no clear peak in the
frequency spectrum within these three octaves (Supplementary
Fig. 6and Supplementary Note 1) consistent with scalp EEG
recordings in normal subjects4749. Therefore, we analyzed
activity and stimulated at both 0.75 and 1 Hz, values previously
used in the literature16,22,31. As a primary outcome measure, we
tested the modulation of spindle amplitude (for fast spindles: 14
Hz bandpass lter with 7 Hz bandwidth; for slow spindles: 10 Hz
bandpass with 5 Hz bandwidth) by the phase of the endogenous
slow oscillations (1 Hz bandpass lter with 1 Hz bandwidth).
PAC was measured with the modulation index for individual
electrodes and statistical signicance was established using
surrogate data with randomized phase (see Methods),
here and in the remainder of the text. Recording sites that
Field proj. (recordings)
Subject Subject Subject
Field mag. (models, at electrodes) Field mag. (models, entire brain)
Fig. 2 Measured and estimated electric eld magnitudes. aField projections calculated as the difference in recorded voltages between neighboring
electrodes divided by electrode distance for each subject (with four montage orientations shown for S10), beld magnitudes at electrode locations
predicted by calibrated current-ow models, and cmodel-predicted eld magnitude across the entire brain. Red lines indicate the medians, and boxes span
from 5 to 95% of the data, with whiskers extending to the minima and maxima. All values shown here correspond to the maximal current intensity applied
for each subject during stimulation (S1: 1 mA; S2: 0.75 mA; S3: 1 mA; S4: 1 mA; S5: 1 mA; S6: 1 mA; S7: 1.5 mA; S8: 2 mA; S9: 1.5 mA; S10A: 0.3 mA; S10B: 1
mA; S10C: 1 mA; S10D: 0.3 mA; S11: 0.3 mA; S12: 1 mA; S13: 1 mA). The difference in magnitude across subjects is primarily due to these varying stimulation
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x ARTICLE
NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x | 3
Content courtesy of Springer Nature, terms of use apply. Rights reserved
exhibited PAC are widespread over the cortex, representing
most recorded regions (Fig. 3c). Of the 753 electrodes tested
across subjects, 497 exhibit signicant PAC (electrodes with
q<0.05 after FDR correction, electrode count summarized in
Table 1and Fig. 4b). Similar results are obtained with a 0.75 Hz
bandpass lter (509 of 753 electrodes, p=0.55, Fisher exact
ratio test, Supplementary Note 2). Endogenous spindle
power amplitude is highly modulated by the phase of endogenous
slow waves (example electrodes with corresponding p-values in
Fig. 4a). Seventy-seven percent of the channels showing
PAC during night 1 also demonstrate PAC during night 2
(Figs. 3a, 4b, Supplementary Table 2, and Supplementary Fig. 6
for 0.75 Hz bandpass). The differences from one night
to the other may be due to the differing depth of NREM sleep
in the 30-min window tested, in addition to inter-individual
We also computed the preferred phase of the spindle power
within the endogenous slow-wave cycle for both nights and nd
that it is stable across nights (angle difference averaged across
electrodes: 9.7°±13.6°mean and Std across subjects, Fig. 3a,
and Supplementary Fig. 7). 75 ±16% (mean across all subjects) of
the entrained electrodes have a preferred phase closer to 0°, but
25 ±16% align closer to 180°. Slow waves reect changes from
cortical up and down states, with high and low neuronal ring
activity, respectively49. For most electrodes, 0° corresponds to the
upstate of the slow oscillation (Fig. 3b), associated with increased
broadband gamma activity (70110 Hz), which is a measure of
neuronal ring35 (Fig. 3a). Electrodes with 180° reect a sign
reversal of slow waves often observed in intracranial recordings35.
As a secondary outcome measure, we computed the same PAC
for gamma activity and found 533 of 753 electrodes entrained to
the phase of the slow oscillation (Fig. 3a). When we calculated the
difference of their preferred phase (Fig. 3a, Supplementary
Table 2), we nd that spindle and gamma activity cluster around
a phase difference of 0° and coincide with the upstate of the slow
100 Night 1
Night 2
–180 0 180
Phase (deg) Phase (deg)
1 s
Superior Rostral Right hemisphere
Phase (deg) Phase (deg) Phase (deg) Phase (deg) Phase (deg)
–180 0 180 –180 0 180 –180 0 180 –180 0 180 –180 0 180 –180 0 180
S8 S9 S10 S11 S12 S13
Fig. 3 PAC of fast spindle and gamma activity to the phase of endogenous slow oscillation. aHistogram of the mean preferred phase (of 1 Hz bandpassed
signal) across electrodes with signicant PAC for two different nights (Night 1, green; Night 2, orange). Each histogram corresponds to a different subject.
0° corresponds to the positive phase of the raw iEEG traces, +/180° corresponds to the negative phase. The upstate is characterized by increased fast
spindle (14 Hz) and increased gamma (70110 Hz) activity. Electrodes with increased spindle/gamma activity in the positive phase (0°) are mostly cortical,
and those with increase in the negative phase (180°) are mostly depth electrodes, which is consistent with previous reports35,49.bExamples with spindle
activity occurring with the positive phase of the slow-wave cycle as seen in the raw iEEG traces for two representative cortical electrodes (subject S7,
electrode G06l; subject S8, electrodes RIP08). cLocations of cortical electrodes with signicantly PAC of fast spindle activity for S8. Night 1 (top row,
green) and night 2 (bottom row, orange) show widespread and consistent entrainment across nights. Left to right: top view, frontal view, right view. Black
dots indicate the locations of the subdural grid electrodes
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x
4NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
In addition to measuring PAC, we implemented a spindle
detection algorithm50 and computed their preferred phase within
the slow-wave cycle. We obtain similar results with this spindle
detection algorithm albeit with a smaller number of signicantly
entrained electrodes (187 instead of 497 of 753 electrodes;
Supplementary Table 2).
Low-frequency tACS does not entrain sleep spindles. In four
patients during an afternoon nap, and three patients during
night-time sleep, we applied several 510-min blocks of con-
tinuous stimulation during NREM sleep (Stages 2 and 3). We
tested tACS at 0.75 and 1 Hz consistent with previous
studies16,17. We reasoned that if the applied low-frequency tACS
entrains endogenous slow-wave activity, then spindle power,
which is coupled to slow-wave activity, should occur in phase
with the applied elds. Because slow-wave activity cannot be
directly assessed during tACS due to the stimulation artifact,
spindle power (14, 10 Hz) was used as an index of the slow-wave
rhythm. We used identical PAC processing and statistical tests as
with endogenous sleep (modulation index r, assessed for sig-
nicance with phase-randomized surrogate data), with the
exception that phase could now be inferred exactly from the
stimulation artifact, readily observed in the raw recordings
(Supplementary Fig. 4). In subject S8, we nd signicant PAC of
fast spindle power (14 Hz band) in one 1 of 116 electrodes after
correcting for multiple comparisons across electrodes (FDR, q<
0.05; Fig. 4b, for this single electrode N=292 cycles,
p=0.0003 prior for FDR correction). However, PAC in this
electrode does not persist in following stimulation blocks sepa-
rated by only 5-min intervals. The strength of evidence for PAC is
compared across blocks in Supplementary Fig. 8. Other subjects
(S7, S9, S10) do not demonstrate signicant fast spindle
entrainment (at q<0.05) during any stimulation blocks.
Figure 4b shows a summary of all electrodes tested at 1 Hz. For
slow spindle power (10 Hz), only 1 of 98 electrodes demonstrates
signicant entrainment after correcting for multiple comparisons
(FDR, q<0.05, r=7.48, N=216 cycles, p=0.0005 prior to FDR
correction), but this was not maintained across stimulation blocks
(Table 1).
To further increase the power of the statistical test we
combined the data across tACS blocks. The total stimulation
time tested after concatenating all blocks was 30 min (subject S7),
30 min (subject S8), 20 min (subject S9, 1 Hz), 10 min (subject S9,
0.75 Hz), 10 min (subject S10), 20 min (subject S11), 7.5 min
(subject S12), 20 min (subject S13, 1 Hz), and 20 min (subject S13,
0.75 Hz). After concatenating all blocks and FDR correction, we
nd no electrodes demonstrating signicant spindle entrainment
at 14 or 10 Hz.
In three subjects (S9, S11, S13) we also tested entrainment of
spindle power during tACS at 0.75 Hz as this frequency is a
commonly used parameter in previously published studies16,17.
Figure 4b shows a summary of all electrodes tested at 0.75 and 1
Hz. Only one electrode (out of 93 total electrodes) in one of the
three subjects (S9) tested demonstrates fast spindle entrainment
with 0.75 Hz tACS comparisons (FDR, q<0.05, r=1.18, N=216
cycles, p=0.0001 prior to FRD correction), and only in one out of
two stimulation blocks. In summary, no stable entrainment of
spindle activity to the applied electric elds was detected for low-
frequency tACS applied for intensities of up to 2.5 mA during
NREM sleep.
Low-frequency tACS does not alter spindle-gamma alignment.
Unlike applied stimulation waveforms, which remain steady
during application, the endogenous slow waves remain periodic
for at most two cycles (Supplementary Fig. 5), and are more often
isolated events. This lack of temporal coherence is the source of
the broadness of the slow-wave activity spectrum and lack of a
well-dened peak in most electrodes (Supplementary Fig. 6).
Because the endogenous slow-wave oscillation is not reliably
discerned during stimulation due to the overwhelming stimula-
tion artifact, we used gamma activity to infer the phase of the
endogenous slow-wave rhythm. It is possible that tACS did not
entrain spindle activity, but nonetheless interferes with the slow-
wave rhythm, which is thought to coordinate activity in higher
frequency bands. We therefore tested whether low-frequency
tACS disrupts the temporal alignment of gamma and spindle
activity, as measured by the cross-correlation of the instantaneous
amplitude of these two rhythms (Fig. 5). As expected from the
Table 1 Summary of PAC results across all subjects and conditions tested
(subject ID)
Frequency Arousal
Subjects Total #
# electr.
Freq. band
# electr.
Consistent across
blocks (%)
None (S7S13) 1 Hz
NREM 7 860 753 14 Hz (fast
497 77
None (S7S13) 1 Hz
NREM 7 860 753 10 Hz (slow
361 57
tACS (S7S13) 1 Hz (.75 Hz) NREM 7 860 598 14 Hz (fast
2 (1) 0
tACS (S7S13) 1 Hz (.75 Hz) NREM 7 860 598 10 Hz (slow
tACS (S1S6) 1 Hz Wake 5 716 584 7 Hz (theta) 0 0
tACS (S1S6) 1 Hz Wake 5 714 578 90 Hz (high
tACS (S1S6) 1 Hz Wake 6 840 695 10 Hz (alpha) 8 0
Acoustic (A1A4) 1 Hz (0.75
NREM 4 448 343 14 Hz (fast
20 (51) 4
Acoustic (A1A4) 1 Hz (0.75
NREM 4 448 343 10 Hz (slow
16 (33) 4
Trapezoidal (S14) 0.75 Hz NREM 1 126 113 1 Hz (slow-
Columns indicate (from left to right): the type of stimulation, if any, along with the subject ID; the frequency of stimulation or analysis that provided the phase for the PAC measure; the arousal state of
the subject; number of subjects tested; total number of electrodes tested across all subjects; frequency band that provided the amplitude for the PAC measure; total number of electrodes that showed
signicant entrainment in at least one block of stimulation/night; fraction of electrodes that entrained consistently across two or more blocks/nights of all electrodes tested
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x ARTICLE
NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x | 5
Content courtesy of Springer Nature, terms of use apply. Rights reserved
alignment of the two rhythms to the 0° phase or cortical upstate
during endogenous sleep (Fig. 3a), the cross-correlation of the
gamma and spindle power is maximal at zero-time delay (the
median across electrodes and subjects S7S13 ±the interquartile
range is 0.0020 ±0.16 s) for endogenous NREM sleep (Fig. 5,
Nights 1 and 2). During stimulation, the temporal relationship of
spindle with gamma activity (with a time lag of zero. Night 1:
0.0156 ±0.020 s; Night 2: 0.0078 ±0.023 s) is maintained
(Fig. 5). This suggests that tACS does not disrupt the temporal
coordination of spindle and gamma activity.
Low-frequency trapezoidal tACS does not entrain slow waves.
To further test our null ndings, we performed a close replication
of the original study in which tACS was found to enhance slow-
wave activity16. This was performed in one additional surgical
patient (subject S14), during night-time sleep. We used the same
montage, electrodes, waveform, and (lower) stimulation intensity
of that earlier study16 (bilateral ring electrodes on F3/F4 and
mastoids with 0.75 Hz on/off trapezoid at 0.52 mA) and stimu-
lated during NREM sleep periods. This electrode placement was
facilitated in this one subject by a custom surgical bandage. The
trapezoidal waveform does not permit analysis during stimulation
due to the broad band artifacts they generate (for this reason we
used sinusoidal stimulation for the majority of the experiments).
Thus, here we analyzed whether the slow-wave activity immedi-
ately after stimulation remained time-aligned across trials fol-
lowing previous methods16,37. To achieve comparable statistical
power to these earlier studies in a single subject we stimulated in
80 short blocks (8 cycles, or 10 s duration each; see also discussion
on statistical power in Supplementary Note 3). After correcting
for multiple comparisons (FDR, q<0.05), none of the 113 elec-
trodes exhibit phase-locked slow-wave oscillations (0.51.5 Hz) in
the subsequent stimulation-free intervals (using Rayleigh test for
non-uniformity, see Methods; uncorrected p-values reported in
Supplementary Fig. 12). Remarkably, subsequent clinical evalua-
tion did not reveal any electrical after-discharge or interictal
abnormalities; the patient was diagnosed as non-epileptic.
Low-frequency tACS does not modulate theta during wake-
fulness. During waking rest, we also tested whether the applied
eld modulates theta activity (7 Hz bandpass, 3.5 Hz bandwidth)
as a primary outcome measure, and alpha (10 Hz bandpass, 5 Hz
bandwidth) or gamma activity (90 Hz bandpass, 40 Hz band-
width) as exploratory outcome measures. We applied 1 Hz tACS
to six patients during wakeful rest. After correcting for multiple
comparisons (FDR, q<0.05), none of the 584 electrodes tested
for subjects S1S4 and S6 exhibit signicant PAC. Subject S5 was
not analyzed for theta entrainment due to the presence of
p=0.1506 p=0.0100 p=0.8540 p=0.0474 p=0.2342 p=0.4517 p=0.9433 p=0.0005 p=0.0007 p=0.0006 p<0.0001
p=0.0002 p=0.0003 p=0.0006 p=0.0006 p=0.0046 p<0.0001 p<0.0001 p<0.0001 p<0.0001p<0.0001
Night 1
Night 1 Night 2 tACS Acoustic
Stimulation Night 2
Spindle amp. (a.u)
–180 0 180
–180 0 180
–180 0 180
–180 0 180
–180 0 180
–180 0 180
–180 0180
–180 0180
–180 0 180
–180 0 180
–180 0180
–180 0 180 –180 0 180 –180 0 180 –180 0 180 –180 0 180 –180 0 180 –180 0 180 –180 0 180 –180 0 180 –180 0 180 –180 0 180
–180 0180 –180 0180 –180 0180 –180 0180 –180 0180 –180 0180 –180 0 180 –180 0 180 –180 0180 –180 0180 –180 0 180
Phase (deg) Phase (deg) Phase (deg) Phase (deg) Phase (deg) Phase (deg) Phase (deg) Phase (deg) Phase (deg) Phase (deg) Phase (deg)
Night Night Night Night Night Night Night Night Night Night TAC S
(1 Hz)
(1 Hz)
(1 Hz | 0.75 Hz)
(1 Hz | 0.75 Hz)
(0.75|1 Hz)
(0.75|1 Hz)
(0.75|1 Hz)
(0.75|1 Hz)
(1 Hz)
(0.75 Hz)
(1 Hz)
Electrodes (#)
S7 S8 S9 S10 S11 S12 S13 A1 A2 A3 A4
Fig. 4 Modulation of fast spindle power with phase of the slow oscillation. Different conditions are indicated in color: endogenous sleep (green, orange),
tACS (black), acoustic stimulation (AS, red). aFast spindle power (14 Hz band) relative to the phase of slow oscillation shown in one representative
electrode for each subject. Here, 0° phase corresponds to the physiological upstate of slow-wave activity. For endogenous slow-wave oscillation sleep
periods (green and orange) phase is determined from 1 Hz bandpassed signal. For tACS (black traces) and acoustic stimulation (AS, red traces) blocks
during nap (S7S10, A2A3) and during night-time sleep (S11S13, A1, A4), phase is determined from the stimulation artifact or acoustic trigger pulses in
the recordings, where 0° corresponds to the peak in anodal stimulation (e.g., peak positive stimulation relative to frontal/anodal electrode) for tACS, or the
time of sound delivery in the case of AS. Note the modulation of spindle power with the phase of the endogenous slow oscillation in each subject during
night-time sleep and during acoustic stimulation, and the lack of consistent modulation with the phase of tACS. Both nights have similar preferred phase.
Each column represents the same electrode per subject. bFraction of recordings sites with signicantly entrained spindle activity during endogenous sleep,
tACS, and acoustic stimulation. Each bar represents a block of data analyzed. Bar height indicates the number of electrodes available, with a fractionof
electrodes discarded due to poor data quality or excessive interictal activity (white), a fraction of electrodes with non-signicant entrainment (gray), and a
fraction of electrodes with signicant spindle entrainment (night 1, green; night 2, orange; tACS, black; AS, red) after correction for multiple comparisons.
Each bar during tACS/AS represents a stimulation block of 5 min. During tACS stimulation only two recording sites (in S8, rst block, red bar; and S09,
second block of 0.75 Hz, red bar) showed signicant spindle entrainment, although this was not a stable nding across stimulation blocks. With acoustic
stimulation, many more electrodes were found to be reliably entrained across both 0.75 and 1 Hz stimulation rates
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x
6NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
signicant environmental artifact. For alpha activity, signicant
PAC is present in 1 of 122 electrodes tested in subject 1 (FDR,
q<0.05, N=594 cycles, p=0.0004 uncorrected), 2/78 electrodes
in subject S4 (FDR, q<0.05, N=594 cycles, uncorrected
p-values =0.007, 0.003), and 5/155 electrodes in subject S6 (FDR,
q<0.05, N=594 cycles, p=0.0004, 0.0001, 0.0015, 0.0018,
0.0001, uncorrected). None of the 340 electrodes recorded for
subjects S2, S3, and S5 demonstrate signicant modulation of
alpha activity with tACS phase. Signicant coupling between
stimulation phase and gamma activity is detected in 1/155 (FDR,
q<0.05, N=592 cycles, p=0.0001 uncorrected) recording sites
for subject 6, but not in any of the 423 electrodes tested for
subjects S1, S2, S4, and S5. Subject S3 was not analyzed for
gamma power entrainment due to the presence of signicant 60
Hz artifact. However, none of the recording sites with signicant
entrainment effects for theta, alpha, or gamma activity in one
stimulation block show consistent effects across repeated stimu-
lation blocks. We conclude that 1 Hz tACS does not entrain theta,
alpha, or gamma activity during waking rest.
Pulsed auditory stimulation induces slow-wave/spindle
entrainment. It could be argued that slow-wave activity cannot
be reliably entrained in epilepsy patients. To directly address this
possibility, we stimulated an additional four patients (subjects
A1A4) with low-frequency acoustic stimulation (50 ms, pink-
noise bursts repeated at 0.75 and 1 Hz) delivered through ear-
phones during nocturnal NREM sleep. Using the same analysis as
described earlier (modulation index, assessed for signicance with
phase-randomized surrogate data), we nd a number of electro-
des that showed signicant PAC entrainment during both 0.75
and 1 Hz auditory stimulation after FDR correction (q<0.05; p-
values prior to FDR correction are shown in Supplementary
Fig. 10, Fig. 4; number of electrodes entrained for patient at 0.75
Hz, at 1 Hz, and consistently entrained across stimulation blocks
as follows A1: 38, 11, 11; A2: 7, 1, 1; A3: 0, 2, 0; A4: 6, 6, 1). The
lower numbers in A3 may be due to the lower sound volume used
for sleep comfort. In total, we found 51/343 (14.8%) entrained
electrodes during 0.75 Hz, and 20/343 (5.8%) entrained electrodes
during 1 Hz acoustic stimulation, when pooled across all subjects.
In addition, time-locked analysis of the individual electrode
waveforms relative to stimulation onset reveals robust slow-wave
events and an increase in spindle power during the slow-wave
upstate, when compared to the sham stimulation delivered during
a baseline period of sleep (χ2(99) =200.23, p<0.0001, Fig. 6).
These ndings suggest that low-frequency auditory stimulation
delivered during NREM can reliably entrain slow-spindle activity
in epilepsy patients.
Finally, to exclude the possibility that we missed a genuine effect,
we performed a number of additional post hoc analyses with less
conservative thresholds, and limiting analysis to electrodes with
highest eld intensity, or those with marked spindle activity.
Additionally, we tested for changes in spindle, slow-wave and theta
power before and after stimulation. These analyses are detailed in
Supplementary Note 4, and show no consistent effects.
We set out to test the hypothesis that low-frequency tACS can
acutely modulate endogenous slow-wave rhythms during NREM
sleep or theta activity during wakefulness. We nd that low-
frequency tACS, at 1 or 0.75 Hz, applied during NREM sleep at
common stimulation intensities does not reliably entrain spindle
oscillations during NREM sleep. This contrasts with spindle
activity that is strongly entrained by endogenous slow oscillations
in almost two-thirds of the electrodes in depth and cortical sur-
face electrodes recorded during two nights of sleep. Gamma
activity also reliably entrains to the upstate of slow-wave activity,
at a similar phase as spindles. Furthermore, time alignment
between the spindle and gamma rhythms is not altered by tACS,
suggesting that co-occurrence of these rhythms to the cortical
upstate was not disrupted by stimulation51. Despite medications,
sleep physiology (i.e., coupling of spindle to slow activity during
NREM sleep) in this patient population resembles that of healthy
subjects49. In contrast, low-frequency auditory stimulation during
NREM sleep evokes reliable slow-spindle events as seen in healthy
subjects41,52. Furthermore, low-frequency tACS during waking
rest does not modulate theta, alpha, or gamma frequency activity
with the phase of stimulation. Thus, low-frequency sinusoidal
tACS does not entrain dominant rhythms during NREM sleep or
waking rest.
20 20
20 20 0.2
Night 1
Night 2
–1 0 1 –1 0 1 –1 0 1 –1 01–1 01–1 0 1 –1 01
–1 01–1 0 1 –1 01–1 01–1 0 1 –1 01–1 0 1
–1 0
Time lag (s) Time lag (s) Time lag (s) Time lag (s) Time lag (s) Time lag (s) Time lag (s)
1–1 0 1 –1 01–1 01–1 0 1 –1 01–1 0 1
S8 S9 S10 S11 S12 S13
Fig. 5 Cross-correlation between the amplitudes of spindle and gamma oscillations. Panel shows cross correlation for all electrodes for seven subjects
(S7S13) during two nights of sleep without stimulation and one with tACS. False color indicates correlation values. Vertical axis indicates electrode
number and horizontal axis indicates time lag between the two rhythms. The peak at zero time-lag indicates that spindle and gamma occur at the same
time, negative lag indicates that spindle activity precedes gamma activity
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x ARTICLE
NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x | 7
Content courtesy of Springer Nature, terms of use apply. Rights reserved
It is tacitly assumed that tACS can entrain ongoing brain
rhythms to boost memory performance16,18,19. However, pre-
vious studies have not directly measured entrainment of these
latter rhythms. The present study is the rst human experiment
to assess the direct physiological effects of tACS by measuring
cortical oscillations via invasive EEG. Here, we demonstrate that
tACS does not necessarily induce coupled slow-spindle rhythms
in human sleep. This specic result is important because of the
exciting but mixed results reported with tACS and sleep-
dependent memory enhancement16,18,19, which highlights the
lack of mechanistic understanding of this technique.
The strengths of our study depend on the simultaneous mea-
surement of brain activity directly from the cortical surface
during stimulation. Intracranial EEG (iEEG) provides superior
signal quality compared to scalp EEG, allowing the direct mea-
surement of induced electrical elds and their acute effect on
mesoscopic brain activity. We collected a large data set with
18 subjects, 2156 electrodes (mean 120 electrodes per subject),
and 9003600 trials (oscillation cycles) per subject. Furthermore,
to ensure that our null ndings are not due to limitations in
analytical technique, we included two positive control conditions,
including endogenous sleep and acoustic stimulation, which
demonstrate that slow-spindle rhythms can be entrained in these
surgical patients and that the analytical methods can detect such
Previous estimates of entrainment of slow-wave activity with
applied slow tACS relied on averaging the periods immediately
following stimulation, which gives a much smaller number of
trials. In other words, measurement during stimulation with iEEG
permits more sensitive detection of PAC as compared to mea-
surement after stimulation with scalp EEG. Finally, we applied
stronger stimulation intensities compared to previously published
reports16,17. In summary, our experimental and analytical
approach permits detection of small changes in modulation,
which we do not nd.
Previously proposed explanations for how TES may affect
neuronal activity rely mainly on in vitro slice preparations and
in vivo experiments in rodents53. In vitro studies demonstrate
that a weak uniform (DC) electric eld can alter the resting
transmembrane potential, which increases or decreases average
ring rates of the affected neurons23,44. In vitro studies using AC
stimulation demonstrate single neuron entrainment by mod-
ulating the rate and timing of neural ring23,30,54. It is assumed
that small changes in the ring rate and coherence of the affected
neuronal populations may be amplied by network-level
activity23,54. However, the weak effects of the applied eld also
compete with the endogenous neuronal drive, which controls the
instantaneous phase and amplitude of the rhythmic activity. For
stimulation to entrain neural activity that does not have a steady
rhythmic behavior, such as slow-wave activity, one would need
fairly strong effects that reset phase in a single cycle. This may
explain why several-fold stronger currents are needed to entrain
some network oscillations compared to the lower intensities
required to alter ring rates of several neurons30,31.
Number of slow-wave down-states
Frequency (Hz)
000.1 0.2 0.3 0.4 0.5
Latency from audio stimulus onset (s)
0.6 0.7 0.8
Stimulus onset
0.2 s
20 µV
0.9 1
00.1 0.2 0.3 0.4 0.5
Time (s)
0.6 0.7 0.8 0.9 1
Fig. 6 Slow-wave entrainment to 1 Hz acoustic stimulation. aNumber of detected slow-wave oscillations relative to stimulus onset, summed over the 99
electrodes analyzed, in a 5-min interval for subject A4 during stimulation (red) and an equivalent baseline sleep period (black). Inset shows the averaged
evoked response in both conditions in one representative electrode (depth electrode DPMT 03), time-locked to stimulus onset (note the positive peak a
cortical down-state in this depth electrode, which is opposite from the majority of cortical electrodes shown as examples in Fig. 3b). bTFR of time-locked
epochs (e.g., relative to stimulus onset) in one representative electrode (DPMT 03), showing an increase in slow spindle power (812 Hz) during slow-
wave down-statesconsistent with previous literature80 (0.40.8 s RE: stimulation onset) and fast spindle power (1218 Hz) during up-states(0.150.4
s, RE: stimulation onset) as found for endogenous slow-wave oscillations (see Figs. 3,4). TFR is computed relative to baseline sleep period with sham
stimulation. Solid innermost curve represents signicant increases relative to baseline after FDR correction (one-tailed t-test, p<0.05)
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x
8NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Our null ndings likely result from the weak-induced electrical
elds. At the highest current intensity applied (2.5 mA), peak
intensities reach only 0.16 V/m across recording electrodes.
Although the recording electrodes may not capture the maximal
electric elds induced by a given stimulation montage, our
model-based extrapolations suggest that peak elds do not exceed
0.43 V/m intracranially when stimulating at 2 mA40, which is
comparable to a recent report of 0.5 V/m as maximal intensities55.
Thus, eld magnitudes may be an order of magnitude smaller
than required to entrain slow-wave rhythms, which may explain
why experimental attempts to entrain sleep rhythms for memory
enhancement have been inconsistent5,7. To place these weak-
induced elds into perspective, synchronous neuronal activity
under physiological conditions can generate approximately 12
V/m gradient across the CA1 pyramidal layer during theta
oscillations in the hippocampus56 and during slow oscillations in
the neocortex30. Under ideal circumstances when the electric
elds are applied in a phase-locked manner coupling directly with
the cellular membrane in vitro, 0.25 V/m can bias a single neu-
rons spike threshold to induce cumulative effects and entrain
network oscillations23,57. In behaving rats, hippocampal ring
rate, but not local eld potential oscillations, was modulated at a
threshold of 1 V/m31. In other scenarios, 5 V/m or more may be
needed to entrain network oscillations30,53.
The weak electric elds measured in our study suggest that
higher intensity stimulation may be required to instantaneously
affect network-level neuronal activity. The practical implication
of our ndings is that new methods of neuromodulation, which
increase the amount of current reaching the brain, are needed to
advance the eld. However, transcranial stimulation intensity is
limited by clinical considerations such as sensory perception.
These constraints are especially relevant for sleep studies (one of
our subjects was wakened by >2 mA stimulation and we had to
decrease the intensity), subject blinding58, and patient safety59.
Another possible explanation for our null ndings is the
observation that endogenous slow-wave activity does not have a
steady rhythm, but rather spans wide frequency range (0.54 Hz).
In human sleep, slow waves appear to be unitary events, and a
regular rhythm is preserved for at most two cycles. Thus, the
continuous application of a weak, low-frequency stimulation
waveform aligns imprecisely to these irregular events. In order for
spindle activity to align to the stimulation phase, the effect of
stimulation would have to be nearly instantaneous. This would
require pacing of the endogenous rhythm on a nearly cycle-by-
cycle basis. By contrast, the most sensitive effects of stimulation23
(at weak elds of 0.2 V/m) occur when the endogenous rhythm is
narrow-band such that stimulation effects can accumulate over
many cycles, leading to resonant phenomena. Our results there-
fore do not rule out more sensitive resonant effects of tACS for
more rhythmic narrow-band activity such as alpha and spindle
Finally, these null effects may result from the anatomical dif-
ferences seen when scaling from in vitro to rodent to human
studies. Due to complex cortical folding in humans, portions of
the network that are differently oriented relative to the induced
potential eld may demonstrate both excitatory and inhibitory
We acknowledge the difference in stimulating electrode pla-
cement between our protocol and previously published protocols,
yet do not conclude that the null effect is because of this
experimental difference. Because of the clinical limitations of the
surgical bandage, we were able to place electrodes in the fron-
topolar and occipital regions, directly below the surgical dressing.
The advantage of maximizing electrode distance as in our pro-
tocol is to decrease the amount of current shunted by the scalp
(compared to electrode placement in the bilateral frontal regions
(F3/F4) and mastoids as prior). This generally should lead to
stronger electric elds in the brain as compared to the
F3/F4 montage42, as supported by our recently validated current
ow models40.
The positive control of acoustic stimulation demonstrates that
slow-wave and spindle activity can be modulated during NREM
sleep in our patients. Acoustic stimulation has previously been
shown to enhance sleep oscillations in healthy subjects52 with an
increase in sleep-dependent memory consolidation41. Using a
low-frequency acoustic analog of tACS during NREM, we observe
sleep spindles are modulated by the rhythmic stimulation in
814.3% of the electrodes analyzed. Additionally, the timing of
spindle activity coincides with the upstate for both acoustic sti-
mulation and endogenous slow waves. Together this suggests that
endogenous slow wave may have been entrained by the rhythmic
stimulation. Regardless of the mechanism (entrained vs. induced
spindle), this indicates that modulation of sleep oscillations is
possible, and that the analytic techniques used to assess
entrainment are sensitive to the PAC of slow-wave/spindle events.
There are, however, limitations of our study. First, because the
recording electrodes are placed according to clinical indication,
they do not necessarily capture the maximal electric elds
induced by a given stimulation montage. Despite broad coverage
of the cortical surface, peak electric elds may occur outside the
recorded areas (but should not exceed 0.43 V/m for 2 mA).
Nonetheless, spindle entrainment during endogenous sleep and
acoustic stimulation show the widespread nature of these cortical
rhythms. Even if our electrodes did not capture the maximal
electric elds, we would have expected to see a small number of
electrodes entrained from tACS. Second, while we did examine
changes in power before and after stimulation, the experiments
were not optimally designed for this purpose, and thus uctua-
tions that were found cannot be attributed to tACS, but are likely
the result for natural uctuations of power within and across
nights of sleep. Third, we only considered low-frequency stimu-
lation. Given the dependence of oscillatory stimulation of the
specic of network rhythms53, these ndings may not generalize
to other tACS frequencies. In summary, as with all null results,
one cannot rule out effects outside the parameters tested here. In
particular, it is possible that there were lasting effects on spindle
power, which we did not resolve here because of their natural
uctuations during sleep on the time scales of minutes. Such
changes were reported in past studies16,21,60 by averaging over
many subjects and may have resulted from the net-DC currents
used there, which we did not test here.
Patients with refractory focal-onset epilepsy undergoing sur-
gical evaluation represent a unique opportunity to record directly
from the cortical surface during non-invasive stimulation. While
aberrant local networks involved in seizure propagation are
observed, other cortical and subcortical functions are often con-
sidered normal. The potential effects of antiepileptic medications
may decrease overall excitability, although some patients were
tapered off all medications during stimulation. Widespread and
strong spindle entrainment by native slow oscillations during
sleep demonstrates that our methods detected physiologically
meaningful changes despite medication use, similar to what has
been reported previously regarding sleep rhythms recorded in
epilepsy surgical patients61. We demonstrate that entrainment of
spindle activity in epilepsy patients is possible with acoustic sti-
mulation. Finally, an exact replication experiment of previously
published protocols, delivered to a healthy brain, conrmed the
absence of acute entrainment of slow waves and further
strengthens the generalizability of our results.
We emphasize that our null results for tACS do not contradict
the reported behavioral effects. While positive behavioral results
have been found in rodents20, a meta-analysis on memory effects
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x ARTICLE
NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x | 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved
reports mixed results in humans33. There may be other means by
which non-invasive stimulation affects brain activity in multiple
indirect ways, including activation of afferent nerves58,62, retina
and the vestibular apparatus63, astrocytes and perivascular
elements64,65, glial activation65, synaptic plasticity66, as well as
through placebo effects67, which merit further exploration.
Our ndings suggest that investigation of novel methods of
stimulation delivery are needed. This may include methods that
induce stronger electric elds at the cortical surface (while
minding patient safety and sensation), or may be based on
acoustic and other modes of sensory stimulation. Furthermore,
applying current at the optimal phase of endogenous rhythms in
a closed-loop manner may be more effective68. However, such
responsive methods can be therapeutic only if the impaired net-
work pattern is identied and continually monitored69,70.
In some medically justied cases, chronic closed-loop feedback
stimulation may be benecial. Electrode plates may be placed
directly onto the skull or on the cortical surface to bypass the
skin, as illustrated by an FDA-approved therapeutic device, used
to detect and disrupt electrographic seizures71. Further progress
requires the investigation of novel electrode arrangements and
stimulation delivery to produce meaningful and reproducible
physiological and behavioral effects.
At a minimum, the present data suggests that tACS at stimu-
lation intensities of up to 2.5 mA does not entrain slow-wave
activity during NREM sleep. More generally, our result challenges
the common assumption that tACS entrains and enhance endo-
genous rhythms. Thus, future studies will now have the burden of
proof when such claims are made.
Human subjects. This study was performed in epilepsy patients undergoing sur-
gical evaluation with iEEG monitoring at New York University Medical Center
(NYUMC). The protocol was approved by the NYUMC Institutional Review Board
and the Clinical Trials Registration number was NCT02263274 (www.clinicaltrials.
gov). Subjects were eligible according to pre-established criteria, including: (1) age
over 18 years old; (2) undergoing invasive monitoring for seizure localization for
epilepsy surgery; and (3) ability to provide informed consent or have a legal
guardian who could consent. Exclusion criteria included (1) signicant cognitive
impairment (IQ <70), (2) facial or forehead skin breakdown that would interfere
with surface electrode placement, (3) contraindication to MRI, (4) known adhesive
allergy, (5) space occupying lesion, and (6) subjects who had an electrographic
seizure for 1 h prior to the stimu lation procedure. All patients (n=17) or their
caregivers provided informed consent. Subjects were enrolled between December
2013 and May 2017. A table listing subject characteristics is included in Supple-
mentary Table 1.
Sleep staging. Stimulation was performed after patients had entered at least 5 min
of continuous NREM sleep, during a daytime nap or nocturnal sleep. Initial sleep
staging was performed by visual online analysis, for the presence of a slow-wave
and spindle activity as detected in the real-time iEEG seen at the bedside by a
physician board-certied in clinical neurophysiology, as well as by direct clinical
observation. As part of standard practice at NYU, an extradural lead is customarily
placed near the vertex of the craniotomy to aid with spindle detection. When the
patient aroused or drifted into a lighter stage of sleep, stimulation was stopped. The
iEEG segments were later conrmed by a second board-certied physician to be
consistent with N2 and N3 sleep. During ofine processing, we selected segments
of NREM sleep for analysis by comparing raw spectrograms of sleep (after artifact
subtraction), to demonstrate that the depth of NREM sleep is similar across testing
conditions (Supplementary Fig. 6).
iEEG recordings. iEEG was recorded from implanted subdural platinum-iridium
electrodes embedded in silastic sheets (2.3 mm diameter contacts, 10 mm
centercenter spacing, Ad-Tech Medical Instrument, Racine, WI) or depth elec-
trodes (1.1 mm diameter, 510 mm centercenter spacing). The decision to
implant, placement or recording electrodes, and the duration of invasive mon-
itoring were determined solely on clinical grounds and without reference to this
study. Electrodes were arranged as grid arrays (8 ×8 contacts, 10 or 5 mm center-
to-center spacing), linear strips (1 × 4 to 12 contacts) , or depth electrodes (1 × 8 or
12 contacts), or some combination thereof. Subdural electrodes covered extensive
portions of lateral and medial frontal, parietal, occipital, and temporal cortex of the
left and/or right hemisphere.
Within 24 h after surgical implantation of electrodes, patients underwent a
post-operative brain MRI to conrm subdural electrode placement. Electrode were
localized and mapped onto the pre-implant and post-implant MRI using geometric
models of the electrode strips/girds and the cortical surface72.
Here, we present an efcient method to accurately localize intracranial electrode
arrays based on pre-implantation and post-implantation MR images that
incorporates array geometry and the individuals cortical surface.
Clinical (macroelectrode) recording equipment. Recordings from grid, strip, and
depth electrode arrays were made using a NicoletOne C64 clinical amplier (Natus
Neurologics, Middleton, WI), bandpass ltered from 0.16250 Hz and digitized at
512 Hz. ECoG signals were referenced to a two-contact subdural strip facing
toward the skull near the craniotomy site. A similar two-contact strip screwed to
the skull was used for the instrument ground.
NeuroConn DC stimulator. The DC-STIMULATOR PLUS (NeuroConn, Ger-
many) is a CE-certied medical device for conducting noninvasive TES in humans.
The stimulator is a micro-processor-controlled constant current source, which
continuously monitors electrode impedance, and detect insufcient contact with
the skin. The device is battery powered, and therefore electrically isolated from the
clinical recording electrodes and equipment.
Low-frequency tACS. We performed 0.75 and 1 Hz sinusoidal tACS on 13 epi-
leptic patients with implanted subdural and depth electrodes. Seven subjects were
stimulated during NREM sleep (four daytime nap; three nocturnal sleep) and six
subjects were stimulated during waking rest, eyes closed. Patients were over 18
years old and uent in English. Subjects were excluded if they had frequent (>2)
electroclinical seizures in the 24 h preceding stimulation . Patient characteristics and
electrode coverage are summarized in Supplementary Table 1and Fig. 1.
All subjects tolerated scalp stimulation. All subjects who were stimulated during
night-time sleep (N=3, subjects 1113) and most subjects during an afternoon nap
(N=4, subjects 710) were able to sleep through trials at stimulation intensities
between 0.5 and 2 mA. One subject (subject 13) woke from sleep and reported an
itching sensation during one stimu lation block with 2.5 mA current intensity. For
the tACS experiments, we recorded and analyzed from 1700 electrodes without
artifacts (mean of 131 electrodes per subject, example electrode in Fig. 3b). There
were no complications from stimulation, and no induced electrographic seizures.
One patient (subject 7) had a typical electroclinical seizure during stimulation.
Because this patient had frequent spontaneous seizures, it was determined by the
patients epileptologist that stimulation was unlikely to have caused the seizure.
Furthermore, we enrolled one subject (S14) who had a bilateral subdural strip
and depth survey, to perform a precise replication experiment of prior protocols.
This patient had multiple target clinical events captured, which were non-epileptic
in nature. He did not have any interictal or ictal activity captured during 1 week of
monitoring, even while medications were being withdrawn. In other words, the
patient did not demonstrate any epilepsy-related pathophysiology.
We reviewed the hour of iEEG recording prior to stimulation to exclude recent
seizures. We performed a pre-stimulation clinical assessment (including
assessment of the stimulation skin site and neurologic examination). A physician
(AL) was present at the bedside during the entire procedure to monitor for safety.
The patients iEEG recording was monitored in real time at the bedside during
stimulation for seizures.
For patients S1S13, two stimulating electrodes were placed medially over the
frontal and occipital poles (2 cm × 2 cm rubber electrodes) for patients S1S10. In
patients S10S13 the frontal electrode was offset from midl ine by 3 cm (S10 and
S13 left frontal; S11 and S12 right frontal) to minimize distance from stimulating
electrodes to recording electrodes. In one patient S10, the stimulation electrodes
were placed at three additional locations to examine the effects of electrode
placement on extracerebral and intracerebral current spread. Subjects were covered
with a nickel-cadmium shroud to reduce environmental artifact during recording,
and other sources of environmental noise (60 Hz) were minimized in the patient
area. The stimulation protocol used the NeuroConn DC Stimulator Plus
(NeuroConn, Germany), with a biphasic sinusoi dal waveform at 0.75 and 1 Hz, at
variable intensities between 0.3 and 2 mA, for 10 s (cycles) to determine the peak
intensity at which amplier saturation occurred. Thereafter, subjects were
stimulated with TES at 0.75 and 1 Hz, at variable intensities up to the peak
intensity, for a duration between 5 and 10 min (10 min for A1A6 and 5 min for
S1S4). Up to four blocks of stimulation were applied, until subjects woke up. The
more than ten-fold increase of the subdurally recorded iEEG amplitude, compared
to the EEG signal73, allowed for simultaneous recording and stimulation (up to
saturation levels of the ampliers). Stimulation was immediately stopped in the
event of an electrographic seizure (S7). A repeat clinical assessment (including
assessment of stimulation skin site and neurologic examination) was performed
after stimulation.
For S14, who was enrolled to perform a replication experiment of prior
protocols, we selected a surgical patient who had a bilateral strip and depth survey.
There were two windows that were cut into the patients surgical bandage to allow
electrode placement at the F3/F4 positions. Stimulation electrodes (8 mm Ag/Cl
ring type) were applied bilaterally, with anodes at F3/F4 and cathodes on each
mastoid. To test for acute effects, we utilized a stimulation protocol using
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x
10 NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
trapezoidal waveform (0.33 s ramp up/0.33 s steady state/0.33 s ramp down/0.33 s
zero current), 0.75 Hz, 00.26 mA, for 8 cycles on/8 cycles off (10.66 s ON/10.66 s
OFF), for 80 cycles. Stimulation was started after the rst 5 min of NREM sleep.
Low-frequency noise-burst auditory stimulation. Auditory stimulation consisted
of 50 ms pink-noise bursts (1/f spectrum) repeated at a rate of 1 of 0.75 Hz to four
patients (subjects A1A4). Sound was delivered via at-prole headphones (Bed-
phones, Millwood, NY), which were placed on the patients ears. Placement of the
headphones required access under the surgical bandage and was administered by
an epilepsy physician (AL), with patient verication of correct positioning.
Acoustic pulses (50 ms, pink noise, 5 ms on/off ramps) were digitally generated and
delivered via a laptop placed at the bedside. The sound level of the stimulation was
manually adjusted for each patient to maximize comfort (e.g., ability to sleep with
sound playing in the background). The volume was recorded and estimates of the
sound level presented to each subject were assessed via an ear simulator (KEMAR
Head and Torso simulator, Knowles Research, coupled to a B&K type 3134
Pressure microphone and B&K type 2230 Sound Level Meter, Bruel & Kjaer,
Denmark). Resultant peak sound pressure level estimates for each subject are 72 dB
SPL (subject A1), 68 dB SPL (subject A2), 46 dB SPL (subject A3), 70 dB SPL
(subject A4).
Similar to the procedure used during tACS stimulation, acoustic stimulation
was presented at repetition rates of 0.75 and 1 Hz, in blocks of 5 min during NREM
sleep. For each subject, we collected a block at each stimulation rate during NREM
sleep, which was visually conrmed ofine. The order of the presentation blocks
was randomized across subjects. An awake control condition was performed for
each patient, to verify the presence of acoustic ERPs. Additional time-locked TTL
triggers were generated for each stimulus presentation and recorded by the EEG
ampliers DC input to aid in ofine analysis.
iEEG data preprocessing. All electrodes were inspected for signal quality by
plotting spectrogram, raw voltage, and the power spectrum. We recorded from a
total of 1700 electrodes for this analysis. Electrodes were discarded based on high
60 Hz noise (likely due to poor contact impedance), amplier saturation (clipping),
or poor removal of the tACS artifact (due to non-stationarity of stimulation
artefact, typically resulting from patient movement). Example of artefact-free
recording electrodes are shown in Supplementary Figs. 2and 3. The electrodes
remaining for each subject during tACS were: 122/126 (S1), 112/126 (S2), 117/126
(S3), 78/84 (S4), 111/124 (S5), 155/254 (S6), 80/128 (S7), 116/122 (S8), 89/100 (S9,
1 Hz tACS), 93/100 (S9, 0.75 Hz), 83/124 (S10), 49/98 (S11), 79/102 (S12), 88/188
(S13, 1 Hz tACS), 93/188 (S13, 0.75 Hz tACS). During endogenous sleep the
electrodes remaining for each subject during night 1 were: 123/126 (S7), 103/122
(S8), 96/100 (S9), 91/124 (S10), 70/98 (S11), 94/102 (S12), 170/188 (S13). During
night 2: 121/126 (S7), 113/122 (S8), 89/100 (S9), 96/124 (S10), 82/98 (S11), 93/102
(S12), 164/188 (S13).
Measurement and modeling of electric elds. During tACS, the current alter-
nates in directionality between two stimulating electrodes. This alternation results
in a sinusoidal signal that can be used to determine the magnitude of the stimu-
lation voltages. Magnitude was estimated by tting a sinusoid to the signal at each
electrode location and estimating the magnitude of the tted signal. The output of
this processing was plotted and manually inspected electrode by electrode.
The measured voltage in each location is then used to derive the projected
electric eld, by subtracting potential values between adjacent electrode pairs and
dividing by their distance, resulting in V/m. The adjacent electrode was dened as
the closest electrode within a 10 mm vicinity for cortical electrodes on the same
grid array and linear strip, and 5 mm vicinity for depth electrodes on the same
strip, to reect the different inter-electrode distance. It is important to realize that
this only captures a fraction of the eld magnitude at any given location as the eld
orientation may not be parallel to the direction of two neighboring electrodes. The
distant stimulating electrodes were expected to generate the strongest eld
intensities on the cortical surface directly under the scalp electrodes40. However,
recording electrodes lay predominantly orthogonal to eld direction (parallel to the
cortical surface). Thus, the measured eld projections will not capture maximal
intensities, except in the rare circumstance that a depth array is precisely
underneath one of the stimulating electrode and oriented toward a second, distant
stimulating electrode.
The computational models were built following our previous work74. Briey,
the MRI for each patient was automatically segmented by the New Segment
toolbox75 in Statistical Parametric Mapping 8 (Wellcome Trust Centre for
Neuroimaging, London, UK) in Matlab (R2013a, MathWorks, Natick, MA).
Segmentation errors were corrected rst by a customized Matlab script74 and then
by hand in ScanIP software (v4.2, Simpleware Ltd., Exeter, UK). The eld of view
of the clinical MRI scans was extended down to the neck by co-registering a
standard head74, and pasting the lower portion of the standard head to the model.
The 2 × 2 cm stimulation electrodes were positioned on the model using CAD
software. For each patient, a nite element model was generated from the
segmentation data and then the electric potential distribution was computed
assuming 1 mA current through the stimulation electrodes. Tissue conductiv ities
were adjusted to minimize the mean-square difference between predicted and
measured potentials. With these calibrated models, we then computed electric
elds throughout the brain. Electric potentials of model and measurements
corresponded closely, with correlation values of r=0.95 ±0.04 (mean ±Std across
patients, N=1545 electrodes across 10 subjects). Electric eld is the spatial
derivative of these potentials. They are estimated as the difference in electric
potential between neighboring electrodes, divided by the distance. This is the
electric eld projected on the orientation of the electrode pair40.
Phase-amplitude coupling. PAC measures the degree to which the amplitude of a
high-frequency oscillations, A
(t), is aligned with the phase of a lower frequency,
(t). We were interested in the interaction between the amplitude of spindle
activity band at 14 Hz with the phase of endogenous slow oscillations at 1 Hz (or
0.75 Hz) during sleep as well as entrainment to the applied stimulation (tACS and
acoustic). This section refers to 14 Hz activity, but the identical analysis was done
for power amplitudes in the theta, alpha, and gamma bands (see next paragraph).
To measure entrainment of spindle oscillations during endogenous sleep, we
compared spindle power against the phase of the endogenous slow oscillation
activity (ϕ
(t), low-frequency endogenous). To measure entrainment to tACS, we
used the phase of the electrical stimulation artifact (ϕ
low-frequency stimulation; Supplementary Fig. 1). To obtain the phase during
acoustic stimulation we used the delay from the onset of each noise burst. To
obtain the phase during tACS and remove the stimulation artifact, we rst modeled
the 1 Hz artifact as a linear superposition of sines and cosines at multiples of a base
frequency (harmonics up to 40 Hz) by tting the amplitude of each sine/cosine and
the base frequency. An example of this tting procedure is shown in Supple-
mentary Fig. 4with the top row indicating the raw signals and the bottom row
showing the signals after the tted harmonic artifact has been subtracted. The
resulting harmonic t captures the 1 Hz stimulation artifact including any har-
monic distortion that may have resulted from amplier nonlinearities. We calcu-
lated the stimulation phase from the harmonic t (Supplementary Fig. 1D). To
obtain the phase during endogenous sleep, rst we applied a complex-valued
Morlet wavelet lter centered at 1 Hz (or 0.75 Hz) with a bandwidth of 1 Hz (or
0.75 Hz; in humans, the center frequency of slow-wave activity is often assumed to
be <1 Hz)48. The instantaneous phase can be directly extracted from the complex-
valued ltered signal. Peak and trough of the slow oscillation are indicated by 0°
and 180°, respectively, which represent the cortical upstate and downstate as dis-
cussed in the main text.
To obtain the instantaneous amplitude of the high-frequency rhythm during
tACS we subtract the harmonic t (Supplementary Fig. 4) and ltered the residual
iEEG signal with a complex-valued Morlet bandpass for spindle activity (fc =14
Hz, bandwidth =7 Hz), for alpha activity (fc =10 Hz, 5 Hz bandwidth) and for
gamma activity (fc =90 Hz, bandwid th =40 Hz, Supplementary Fig. 1C). The
instantaneous HF power was obtained by taking the absolute value of the ltered
signal (Supplementary Fig. 1E). Patient movement and other artifacts resulted in
outliers in the HF amplitude estimates. We removed these by removing cycles with
excessive power as follows. For each electrode and each LF cycle we compute the
mean power of the HF band (square amplitude averaged over one LF cycle). Cycles
are removed as outliers if their mean power exceeds two times the interquartile
range across all cycles in an electrode. To obtain the amplitude of the high-
frequency rhythm during endogenous sleep and auditory stimulation we used the
same procedures starting with the raw iEEG signal (no harmonic t is needed).
Outlier rejection was done as before based on the mean power in the high-
frequency band. In addition, for endogenous slow wave we excluded cycles for
which the slow oscillation amplitude was below 50 µV. For the analysis of the
acoustic stimulation we used the same procedures as with the endogenous sleep
except that the phase was dened based on the time since the onset of the noise
burst, and cycle duration determined from the average inter-stimulus interval (TTL
pulse) for each subject.
PAC is measured here using the modulation index r, which is dened as the
absolute value of the time average, r=|<z(t)>|, of the complex-valued quantity,
zðtÞ¼AHFðtÞeiϕLF ðtÞ(Supplementary Fig. 1F). Time average <z(t)>is computed
over cycles and time within a cycle. If z(t) does not have a radially symmetric
distribution this will cause the time average r(modulation index) to be different
from zero. This can be a result of (1) the amplitude A
(t) is consistently higher at
a certain phase, or (2) ϕ
(t) is not uniformly distributed in time. The phase is
uniformly distributed in the case of tACS and acoustic stimulation. However,
during endogenous sleep, phase is extracted from the slow oscillations, which is not
sinusoidal resulting in non-uniform phase distributions. Consequently, we applied
a histogram equalization to the phase distribution, ensuring that non-zero
modulation index is only a result of modulated amplitude coupling, which we
conrmed by testing for signicance using surrogate data with constant HF
Randomized surrogate data to estimate statistical signicance of PAC. Sig-
nicance was determined by randomizing the phase of each LF cycle and thus
creating the distribution of the modulation index r, under the null hypothesis of no
PAC. Phase randomization makes no assumptions on the distribution of the data,
except that the phase is uniform, which has been addressed in the paragraph above.
For each slow oscillation cycle the low-frequency phase, ϕ
(t), was incremented by
a random value uniformly distributed between 0° and 180°. The phase was shifted
by the same random phase for all electrodes but independently for different cycles.
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x ARTICLE
NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x | 11
Content courtesy of Springer Nature, terms of use apply. Rights reserved
The randomizing procedure was repeated 10,000 times and the p-value was
measured as the rate of a random phase having a modulation index higher than the
original data. The minimum numerical p-value possible, given the number of
randomizations, was 104(1/number of shufes). All the p-values were corrected
for multiple comparisons across electrodes using FDR correction76 with q<0.05.
No correction was performed across segments (as we use segments to determine
how reliable a potential entrainment is over longer periods of time) or frequency
bands (as these were planned comparison). Uncorrected p-values for electrodes
that had p<0.05 in at least one stimulation block are shown in Supplementary
Figs. 79for tACS, endogenous slow-wave (no stimulation), and acoustic stimu-
lation, respectively.
Note that we examine the acute effects of modulation within a single cycle, thus
permitting many single trials(N=300 cycles for 1 Hz and N=225 for 0.75 Hz
within each 5-min block, repeated 26 times for each stimulation intensity; see
number of blocks tested for each subject in Supplementary Table 1). This large
number of trials permits, in principle, detection of small changes in power within a
cycle, on the order of a few percent (e.g., assuming independent noise: sqrt(1/300)
=5.6%). See Supplementary Note 3for an extended discussion on statistical power.
To aid comparison with endogenous sleep and tACS stimulation condition, we
analyzed entrainment for different sleep duration segments. The longer the
recordings (approximately 20 min), the more accurate the mean vector strength
technique is at measuring entrainment (approximately 50% of electrodes
entrained). For shorter durations of endogenous sleep (5 min of iEEG) comparable
to the duration of tACS blocks, there is still modulation of spindle activity but in
only 10% of the electrodes.
Spindle event detection. Spindle detection follows existing methods50. Briey, the
signal is bandpass ltered in the spindle band as above, and instantaneous power is
obtained as the square amplitude of the ltered signal. For each channel a detection
threshold is dened as six times the median of the instantaneous power. For
segments that cross this threshold a lower threshold (1 standard deviation of the
power) is applied to detect onset and offset of the spindle event. Only spindles
whose duration was between 0.2 and 2 s were considered for further analysis50,77.
To avoid false positive detections due to patient movements or high-frequency
artifacts spindle events that coincide with increases in broadband power were
discarded. Events exhibiting broadband power increases (p>0.10, comparing the
maximum broadband power in that event vs. the distribution of broadband power
during the whole recording) were excluded.
Slow-wave detection during auditory stimulation. Additional analyses were
performed on the acoustic stimulation data to assess the physiological origin of the
PAC entrainment. First, a slow-wave detection algorithm was applied during sti-
mulation segments to assess whether underlying slow-wave activity was altered
during stimulation. In contrast to tACS, this is possible for auditory stimulation as
there are no electric stimulation artefacts in the iEEG signal. Following previous
literature61,78, slow-wave detection consisted of bandpass ltering the waveform
(fc =0.75 Hz, bandwidth =0.75 Hz) and using a zero-crossing algorithm to identify
events where two subsequent negative zero-crossing (e.g., from positive to negative)
were within the range of 0.52s (20.5 Hz, bandwidth =1.5 Hz). Subsequently,
once a slow-wave event was identied, the down-state was identied as the
minimum voltage within this event. Second, time-frequency response (TFR)
functions were calculated to assess whether stimulation trials consisted of phy-
siological sleep spindles. TFRs were computed using a 6-cycle Morlet wavelets
between 5 and 25 Hz and averaging the resulting spectrograms (locked to stimu-
lation onset) across all noise bursts (Fig. 6). Signicance between conditions (sleep
vs. baseline) was assessed via paired sample t-test (one-tailed, p<0.05) compared
across all stimulation trials (n=300).
Analysis of changes in power before and after stimulation. Qualitative com-
parison between the NREM during the two nights without stimulation and the
stimulation period show no evident changes in the power spectrum (Supplemen-
tary Fig. 6). To determine if there were signicant changes in power at the different
frequency bands before vs. after stimulation we used the Chronux toolbox79 (http://; version 2.12). Briey, for each electrode, 30 s immediately after each
5-min stimulation block were compared against 30 s preceding the rst block.
Differences in power in the slow oscillation (0.51 Hz) and spindle bands (10 and
14 Hz) are shown in Supplementary Fig. 12. Statistical signicance of power
changes is computed with the Chronux toolbox for the fast spindle band and slow
oscillation band and FDR corrected (q<0.05).
Entrainment with electric stimulation assessed after stimulation block.In
subject S14 we collected enough stimulation blocks to evaluate slow-wave
entrainment in the intervals immediately following tACS. For this, slow-wave
entrainment was tested following previous reports37. Briey, the 10-s stimulation-
free interval immediately after stimulation was t with a sinusoid (0.51.5 Hz) and
phase coherence was calculated across all trials. Statistical signicance was calcu-
lated using a circular test statistic (Rayleigh test for non-uniformity). In addition, to
test for after-effects, we utilized the same trapezoidal waveform, frequency, and
intensity in ve separate 5-min blocks during NREM sleep (5 min ON/5 min OFF).
We compared these sessions against two control periods of endogenous NREM
sleep for spindle (10, 14 Hz) entrainment. The power in the stimulation-free
intervals was calculated for slow oscillations (0.51 Hz), slow spindle activity (812
Hz), and fast spindle activity (1215 Hz). This quantity was then compared to the
power during a different night, when the subject was in an equal sleep state. Slow
oscillations and slow spindle activity were averaged across locations close to Fz.
And fast spindle activity was determined across parietal locations.
Code availability. The code used to generate the main ndings of the current study
are available from the corresponding author upon reasonable request.
Data availability. The data sets generated during and/or analyzed during the
current study are available from the corresponding author upon reasonable request.
Received: 3 June 2017 Accepted: 15 August 2017
1. Liu A., Fregni F., Hummel F. & Pascual-Leone A. in Transcranial Brain
Stimulation Eds. Miniussi, C., Paulus, W. & Rossini, P. M. (CRC Press, 2012).
2. Bindman, L. J., Lippold, O. C. & Redfearn, J. W. Long-lasting changes in the
level of the electrical activity of the cerebral cortex produced bypolarizing
currents. Nature 196, 584585 (1962).
3. Purpura, D. P. & Mcmurtry, J. G. Intracellular activities and evoked potential
changes during polarization of motor cortex. J. Neurophysiol. 28, 166185
4. Jefferys, J. G. Nonsynaptic modulation of neuronal activity in the brain: electric
currents and extracellular ions. Physiol. Rev. 75, 689723
5. Eggert, T. et al. No effects of slow oscillatory transcranial direct current
stimulation (tDCS) on sleep-dependent memory consolidation in healthy
elderly subjects. Brain Stimulation 6, 938945 (2013).
6. Brignani, D., Ruzzoli, M., Mauri, P. & Miniussi, C. Is transcranial alternating
current stimulation effective in modulating brain oscillations? PLoS ONE 8,
e56589 (2013).
7. Sahlem, G. L. et al. Oscillating square wave transcranial direct current
stimulation (tDCS) delivered during slow wave sleep does not improve
declarative memory more than sham: a randomized sham controlled crossover
study. Brain Stimulation 8, 528534 (2015).
8. Horvath, J. C., Forte, J. D. & Carter, O. Quantitative review nds no evidence of
cognitive effects in healthy populations from single-session transcranial direct
current stimulation (tDCS). Brain Stimulation 8, 535550 (2015).
9. Horvath, J. C., Forte, J. D. & Carter, O. Evidence that transcranial direct current
stimulation (tDCS) generates little-to-no reliable neurophysiologic effect
beyond MEP amplitude modulation in healthy human subjects: a systematic
review. Neuropsychologia 66, 213236 (2015).
10. Merton, P. A. & Morton, H. B. Stimulation of the cerebral cortex in the intact
human subject. Nature 285, 227227 (1980).
11. Antal, A. & Paulus, W. Transcranial alternating current stimulation (tACS).
Front. Hum. Neurosci. 7, 317 (2013).
12. Buzsáki, G. & Draguhn, A. Neuronal oscillations in cortical networks. Science
304, 19261929 (2004).
13. Maris, E., van Vugt, M. & Kahana, M. Spatially distributed patterns of
oscillatory coupling between high-frequency amplitudes and low-frequency
phases in human iEEG. Neuroimage 54, 836850 (2011).
14. Steriade, M. & Timofeev, I. Neuronal plasticity in thalamocortical networks
during sleep and waking oscillations. Neuron 37, 563576 (2003).
15. Steriade, M., Contreras, D., Curro Dossi, R. & Nunez, A. The slow (<1 Hz)
oscillation in reticular thalamic and thalamocortical neurons: scenario of sleep
rhythm generation in interacting thalamic and neocortical networks. J.
Neurosci. 13, 3284 (1993).
16. Marshall, L., Helgadóttir, H., Mölle, M. & Born, J. Boosting slow oscillations
during sleep potentiates memory. Nature 444, 610613 (2006).
17. Antonenko, D., Diekelmann, S., Olsen, C., Born, J. & Molle, M. Napping to
renew learning capacity: enhanced encoding after stimulation of sleep slow
oscillations. Eur. J. Neurosci. 37, 11421151 (2013).
18. Westerberg, C. E. et al. Memory improvement via slow-oscillatory stimulation
during sleep in older adults. Neurobiol. Aging 36, 25772586 (2015).
19. Prehn-Kristensen, A. et al. Transcranial oscillatory direct current stimulation
during sleep improves declarative memory consolidation in children with
attention-decit/hyperactivity disorder to a level comparable to healthy
controls. Brain Stimulation 7, 793799 (2014).
20. Binder, S. et al. Transcranial slow oscillation stimulation during sleep enhances
memory consolidation in rats. Brain Stimulation 7, 508515 (2014).
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x
12 NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
21. Ladenbauer, J. et al. Promoting sleep oscillations and their functional coupling
by transcranial stimulation enhances memory consolidation in mild cognitive
impairment. J. Neurosci. 37, 71117124 (2017).
22. Kirov, R., Weiss, C., Siebner, H. R., Born, J. & Marshall, L. Slow oscillation
electrical brain stimulation during waking promotes EEG theta activity
and memory encoding. Proc. Natl. Acad. Sci. USA 106, 1546015465
23. Reato, D., Rahman, A., Bikson, M. & Parra, L. C. Low-intensity electrical
stimulation affects network dynamics by modulating population rate and spike
timing. J. Neurosci. 30, 1506715079 (2010).
24. Strüber, D., Rach, S., Trautmann-Lengsfeld, S. A., Engel, A. K. & Herrmann, C.
S. Antiphasic 40 Hz oscillatory current stimulation affects bistable motion
perception. Brain Topogr. 27, 158171 (2014).
25. Voss, U. et al. Induction of self awareness in dreams through frontal low
current stimulation of gamma activity. Nat. Neurosci. 17, 810812
26. Pogosyan, A., Gaynor, L. D., Eusebio, A. & Brown, P. Boosting cortical activity
at beta-band frequencies slows movement in humans. Curr. Biol. 19, 16371641
27. Zaehle, T., Rach, S. & Herrmann, C. S. Transcranial alternating current
stimulation enhances individual alpha activity in human EEG. PLoS ONE 5,
e13766 (2010).
28. Merlet, I. et al. From oscillatory transcranial current stimulation to scalp EEG
changes: a biophysical and physiological modeling study. PLoS ONE 8, e57330
29. Marshall, L., Kirov, R., Brade, J., Molle, M. & Born, J. Transcranial electrical
currents to probe EEG brain rhythms and memory consolidation during sleep
in humans. PLoS ONE 6, e16905 (2011).
30. Fröhlich, F. & McCormick, D. A. Endogenous electric elds may guide
neocortical network activity. Neuron 67, 129143 (2010).
31. Ozen, S. et al. Transcranial electric stimulation entrains cortical neuronal
populations in rats. J. Neurosci. 30, 1147611485 (2010).
32. Garside, P., Arizpe, J., Lau, C. I., Goh, C. & Walsh, V. Cross-hemispheric
alternating current stimulation during a nap disrupts slow wave
activity and associated memory consolidation. Brain Stimulation 8, 520527
33. Barham, M. P., Enticott, P. G., Conduit, R. & Lum, J. A. Transcranial electrical
stimulation during sleep enhances declarative (but not procedural) memory
consolidation: evidence from a meta-analysis. Neurosci. Biobehav. Rev. 63,
6577 (2016).
34. Rasch, B. & Born, J. About sleeps role in memory. Physiol. Rev. 93, 681766
35. Cash, S. S. et al. The human K-complex represents an isolated cortical down-
state. Science 324, 1084 (2009).
36. Dudai, Y., Karni, A. & Born, J. The consolidation and transformation of
memory. Neuron 88,2032 (2015).
37. Reato, D. et al. Transcranial electrical stimulation accelerates human sleep
homeostasis. PLoS Comput. Biol. 9, e1002898 (2013).
38. Lustenberger, C. et al. Feedback-controlled transcranial alternating current
stimulation reveals a functional role of sleep spindles in motor memory
consolidation. Curr. Biol. 26, 21272136 (2016).
39. Le Van Quyen, M. et al. Large-scale microelectrode recordings of high-
frequency gamma oscillations in human cortex during sleep. J. Neurosci. 30,
77707782 (2010).
40. Huang Y. et al. Measurements and models of electric elds in the in vivo
human brain during transcranial electric stimulation. Elife 6, e18834 (2017).
41. Ngo, H.-V. V., Martinetz, T., Born, J. & Mölle, M. Auditory closed-loop
stimulation of the sleep slow oscillation enhances memory. Neuron 78, 545553
42. Dmochowski, J. P., Datta, A., Bikson, M., Su, Y. & Parra, L. C. Optimized multi-
electrode stimulation increases focality and intensity at target. J. Neural. Eng. 8,
046011 (2011).
43. Datta, A., Bikson, M. & Fregni, F. Transcranial direct current stimulation in
patients with skull defects and skull plates: high-resolution computational FEM
study of factors altering cortical current ow. Neuroimage 52, 12681278
44. Bikson, M. et al. Effects of uniform extracellular DC electric elds on
excitability in rat hippocampal slices in vitro. J. Physiol. 557, 175190
45. Ranck, J. B. Jr. Which elements are excited in electrical stimulation of
mammalian central nervous system: a review. Brain Res. 98, 417440
46. Berry, R. B. et al. The AASM Manual for the Scoring of Sleep and Associated
Events. Rules, Terminology and Technical Specications (American Academy of
Sleep Medicine, Darien, IL, 2012).
47. Finelli, L. A., Borbely, A. A. & Achermann, P. Functional topography of the
human nonREM sleep electroencephalogram. Eur. J. Neurosci. 13, 22822290
48. Achermann, P. & Borbely, A. A. Low-frequency (<1 Hz) oscillations in the
human sleep electroencephalogram. Neuroscience 81, 213222
49. Nir, Y. et al. Regional slow waves and spindles in human sleep. Neuron 70,
153169 (2011).
50. Andrillon, T. et al. Sleep spindles in humans: insights from intracranial EEG
and unit recordings. J. Neurosci. 31, 1782117834 (2011).
51. Valencia, M., Artieda, J., Bolam, J. P. & Mena-Segovia, J. Dynamic
interaction of spindles and gamma activity during cortical slow oscillations
and its modulation by subcortical afferents. PLoS ONE 8, e67540
52. Ngo, H. V., Claussen, J. C., Born, J. & Molle, M. Induction of slow oscillations
by rhythmic acoustic stimulation. J. Sleep Res. 22,2231 (2013).
53. Reato, D., Rahman, A., Bikson, M. & Parra, L. C. Effects of weak
transcranial alternating current stimulation on brain activitya review
of known mechanisms from animal studies. Front. Hum. Neurosci. 7, 687
54. Radman, T., Su, Y., An, J. H., Parra, L. C. & Bikson, M. Spike timing amplies
the effect of electric elds on neurons: implications for endogenous eld effects.
J. Neurosci. 27, 30303036 (2007).
55. Opitz, A. et al. Spatiotemporal structure of intracranial electric elds induced
by transcranial electric stimulation in humans and nonhuman primates. Sci.
Rep. 6, 31236 (2016).
56. Buzsaki, G. Theta oscillations in the hippocampus. Neuron 33, 325340
57. Francis, J. T., Gluckman, B. J. & Schiff, S. J. Sensitivity of neurons to weak
electric elds. J. Neurosci. 23, 72557261 (2003).
58. Fertonani, A., Ferrari, C. & Miniussi, C. What do you feel if I apply transcranial
electric stimulation? Safety, sensations and secondary induced effects. Clin.
Neurophysiol. 126, 21812188 (2015).
59. Shiozawa, P. et al. Safety of repeated transcranial direct current stimulation in
impaired skin: a case report. J. ECT 29, 147148 (2013).
60. Marshall, L., Molle, M., Hallschmid, M. & Born, J. Transcranial direct current
stimulation during sleep improves declarative memory. J. Neurosci. 24,
99859992 (2004).
61. Staresina, B. P. et al. Hierarchical nesting of slow oscillations, spindles and
ripples in the human hippocampus during sleep. Nat. Neurosci. 18, 16791686
62. Tyler, W. J. et al. Transdermal neuromodulation of noradrenergic activity
suppresses psychophysiological and biochemical stress responses in humans.
Sci. Rep. 5, 13865 (2015).
63. Schutter, D. J. L. G. & Hortensius, R. Retinal origin of phosphenes to
transcranial alternating current stimulation. Clin. Neurophysiol. 121,
10801084 (2010).
64. Ruohonen, J. & Karhu, J. tDCS possibly stimulates glial cells. Clin.
Neurophysiol. 123, 20062009 (2012).
65. Monai, H. et al. Calcium imaging reveals glial involvement in transcranial
direct current stimulation-induced plasticity in mouse brain. Nat. Commun. 7,
11100 (2016).
66. Kronberg, G., Bridi, M., Abel, T., Bikson, M. & Parra, L. C. Direct current
stimulation modulates LTP and LTD: activity dependence and dendritic effects.
Brain Stimulation 10,5158 (2017).
67. Horvath, J. C. Are current blinding methods for transcranial direct current
stimulation (tDCS) effective in healthy populations? Clin. Neurophysiol. 126,
20452046 (2015).
68. Berényi, A., Belluscio, M., Mao, D. & Buzsáki, G. Closed-loop control of
epilepsy by transcranial electrical stimulation. Science 337, 735 (2012).
69. Krook-Magnuson, E., Gelinas, J. N., Soltesz, I. & Buzsáki, G. Neuroelectronics
and biooptics: closed-loop technologies in neurological disorders. JAMA
Neurol. 72, 823829 (2015).
70. Krook-Magnuson, E. & Soltesz, I. Beyond the hammer and the scalpel:
selective circuit control for the epilepsies. Nat. Neurosci. 18, 331338
71. Morrell, M. J. Responsive cortical stimulation for the treatment of medically
intractable partial epilepsy. Neurology 77, 12951304 (2011).
72. Yang, A. I. et al. Localization of dense intracranial electrode arrays using
magnetic resonance imaging. Neuroimage 63, 157165 (2012).
73. Engel, A. K., Moll, C. K. E., Fried, I. & Ojemann, G. A. Invasive recordings from
the human brain: clinical insights and beyond. Nat. Rev. Neurosci. 6,3547
74. Huang, Y. et al. Automated MRI segmentation for individualized
modeling of current ow in the human head. J. Neural. Eng. 10, 066004
75. Ashburner, J. & Friston, K. J. Unied segmentation. Neuroimage 26, 839851
76. Benjamini Y. & Hochberg Y. Controlling the false discovery rate: a practical
and powerful approach to multiple testing. J. R. Stat. Soc. B Stat. Methodol. 57,
289300 (1995).
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x ARTICLE
NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x | 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved
77. Warby, S. C. et al. Sleep-spindle detection: crowdsourcing and evaluating
performance of experts, non-experts and automated methods. Nat. Methods 11,
385392 (2014).
78. Molle, M., Marshall, L., Gais, S. & Born, J. Grouping of spindle activity during
slow oscillations in human non-rapid eye movement sleep. J. Neurosci. 22,
1094110947 (2002).
79. Bokil, H., Purpura, K., Schoffelen, J. M., Thomson, D. & Mitra, P. Comparing
spectra and coherences for groups of unequal size. J. Neurosci. Methods 159,
337345 (2007).
80. Klinzing, J. G. et al. Spindle activity phase-locked to sleep slow oscillations.
Neuroimage 134, 607616 (2016).
This work was supported by R01 MH107396-01, R44 NS092144, R01 MH-092926,
NS095123, R41 NS076123, NYU Program Project Grant Development Initiative
(PPG-DI), NYU Clinical and Translational Science Institute (CTSI), NYU Finding a
Cure for Epilepsy (FACES), and the Zimin Foundation. The authors would also like to
thank Xiuyuan (Hugh) Wang for his work on the electrode reconstruction, as well as
Preet Minhas for her early work in data collection and patient coordination.
Author contributions
Conception and design: G.B., O.D., A.A.L., D.F., L.C.P.; institutional review board
approval: A.A.L., T.T., L.M.; acquisition of data: A.A.L., B.L., W.D., D.F., S.H.; analysis
and interpretation of data: B.L., Y.H., L.C.P., A.A.L., D.F., G.B., O.D., S.H.; gure pre-
paration: B.L., Y.H., L.C.P., S.H.; drafting the manuscript: A.A.L., L.P., G.B., O.D.; critical
revisions: all authors.
Additional information
Supplementary Information accompanies this paper at
Competing interests: L.P. has shares in Soterix Medical Devices. The remaining authors
declare no competing nancial interests.
Reprints and permission information is available online at
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional afliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the articles Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
articles Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this license, visit
© The Author(s) 2017
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01045-x
14 NATURE COMMUNICATIONS |8: 1199 |DOI: 10.1038/s41467-017-01045-x |
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

Supplementary resources (2)

... In violin plots representing estimated distributions of data (generated with violinplot, FieldTrip toolbox 92 , MATLAB, MathWorks), lines represent 5, 50 and 95 percentiles. No statistical methods were used to predetermine sample sizes but our sample sizes are similar to those generally used in previous publications 44, 73,95 . Data collection and analysis were not performed blind to the conditions of the experiments. ...
Full-text available
Memory consolidation during sleep is thought to depend on the coordinated interplay between cortical slow waves, thalamocortical sleep spindles and hippocampal ripples, but direct evidence is lacking. Here, we implemented real-time closed-loop deep brain stimulation in human prefrontal cortex during sleep and tested its effects on sleep electrophysiology and on overnight consolidation of declarative memory. Synchronizing the stimulation to the active phases of endogenous slow waves in the medial temporal lobe (MTL) enhanced sleep spindles, boosted locking of brain-wide neural spiking activity to MTL slow waves, and improved coupling between MTL ripples and thalamocortical oscillations. Furthermore, synchronized stimulation enhanced the accuracy of recognition memory. By contrast, identical stimulation without this precise time-locking was not associated with, and sometimes even degraded, these electrophysiological and behavioral effects. Notably, individual changes in memory accuracy were highly correlated with electrophysiological effects. Our results indicate that hippocampo–thalamocortical synchronization during sleep causally supports human memory consolidation.
... Transcranial alternating current stimulation (tACS) is a noninvasive technique that mediates frequency-specific entrainment of cortical activity (3). The rapidly growing number of publications assessing the state of the field (4-6), emergence of new rhythmic neuromodulation protocols (7)(8)(9), and conflicting evidence about the effectiveness of tACS (10) suggest the need for a systematic and quantitative examination of the existing literature. ...
Transcranial alternating current stimulation (tACS) has attracted interest as a technique for causal investigations into how rhythmic fluctuations in brain neural activity influence cognition and for promoting cognitive rehabilitation. We conducted a systematic review and meta-analysis of the effects of tACS on cognitive function across 102 published studies, which included 2893 individuals in healthy, aging, and neuropsychiatric populations. A total of 304 effects were extracted from these 102 studies. We found modest to moderate improvements in cognitive function with tACS treatment that were evident in several cognitive domains, including working memory, long-term memory, attention, executive control, and fluid intelligence. Improvements in cognitive function were generally stronger after completion of tACS ("offline" effects) than during tACS treatment ("online" effects). Improvements in cognitive function were greater in studies that used current flow models to optimize or confirm neuromodulation targets by stimulating electric fields generated in the brain by tACS protocols. In studies targeting multiple brain regions concurrently, cognitive function changed bidirectionally (improved or decreased) according to the relative phase, or alignment, of the alternating current in the two brain regions (in phase versus antiphase). We also noted improvements in cognitive function separately in older adults and in individuals with neuropsychiatric illnesses. Overall, our findings contribute to the debate surrounding the effectiveness of tACS for cognitive rehabilitation, quantitatively demonstrate its potential, and indicate further directions for optimal tACS clinical study design.
... This development could prove very beneficial for treating long-term mental health diseases. Grossman et al., (2017), in their study, observed that researchers had used the combination of electrode groups and EEG multimodal stimulation to greatly improve the stimulation accuracy [62][63][64][65][66][67][68][69][70] (by about 2-3 cm 2 ) and the stimulation depth (by about 3-4 cm) (Figures 4 and 5). Electric stimulation technology, compared to the drug treatment, has similar effects on some functional neurological diseases such as depression and epilepsy. ...
Full-text available
We review the research progress on noninvasive neural regulatory systems through system design and theoretical guidance. We provide an overview of the development history of noninvasive neuromodulation technology, focusing on system design. We also discuss typical cases of neuromodulation that use modern noninvasive electrical stimulation and the main limitations associated with this technology. In addition, we propose a closed-loop system design solution of the “time domain”, “space domain”, and “multi-electrode combination”. For theoretical guidance, this paper provides an overview of the “digital brain” development process used for noninvasive electrical-stimulation-targeted modeling and the development of “digital human” programs in various countries. We also summarize the core problems of the existing “digital brain” used for noninvasive electrical-stimulation-targeted modeling according to the existing achievements and propose segmenting the tissue. For this, the tissue parameters of a multimodal image obtained from a fresh cadaver were considered as an index. The digital projection of the multimodal image of the brain of a living individual was implemented, following which the segmented tissues could be reconstructed to obtain a “digital twin brain” model with personalized tissue structure differences. The “closed-loop system” and “personalized digital twin brain” not only enable the noninvasive electrical stimulation of neuromodulation to achieve the visualization of the results and adaptive regulation of the stimulation parameters but also enable the system to have individual differences and more accurate stimulation.
... The online version of this article includes the following figure supplement(s) for figure 2: imitates the characteristics of tACS, considering that both approaches are based on an external periodical driver applied to the brain. The specific application of the nodal natural oscillatory frequency is based on reports that suggest electrophysiological oscillations can be synchronized by in-phase tACS stimulation (Helfrich et al., 2014), even though this mechanism has been recently disputed (Lafon et al., 2017) and further research is required for its validation. On the other hand, the simulated Sync/ Noise stimulation increases/decreases the overall value of the bifurcation parameter underlying the switching of the dynamical regime of a specific brain region. ...
Full-text available
The treatment of neurodegenerative diseases is hindered by lack of interventions capable of steering multimodal whole-brain dynamics towards patterns indicative of preserved brain health. To address this problem, we combined deep learning with a model capable of reproducing whole-brain functional connectivity in patients diagnosed with Alzheimer's disease (AD) and behavioral variant frontotemporal dementia (bvFTD). These models included disease-specific atrophy maps as priors to modulate local parameters, revealing increased stability of hippocampal and insular dynamics as signatures of brain atrophy in AD and bvFTD, respectively. Using variational autoencoders, we visualized different pathologies and their severity as the evolution of trajectories in a low-dimensional latent space. Finally, we perturbed the model to reveal key AD- and bvFTD-specific regions to induce transitions from pathological to healthy brain states. Overall, we obtained novel insights on disease progression and control by means of external stimulation, while identifying dynamical mechanisms that underlie functional alterations in neurodegeneration.
... Additionally, tACS during sleep may not always show a neural entrainment effect. For example, [67] found no entrainment effect of low amplitude tACS (2.5 mA) on slow oscillations during SWS using implanted recording electrodes in epileptic patients. For an excellent review on this body of work, see [68]. ...
Full-text available
Previous studies have found a benefit of closed-loop transcranial alternating current stimulation (CL-tACS) matched to ongoing slow-wave oscillations (SWO) during sleep on memory consolidation for words in a paired associates task (PAT). Here, we examined the effects of CL-tACS in a retroactive interference PAT (ri-PAT) paradigm, where additional stimuli were presented to increase interference and reduce memory performance. Thirty-one participants were tested on a PAT before sleep, and CL-tACS was applied over the right and left DLPFC (F3 and F4) vs. mastoids for five cycles after detection of the onset of each discrete event of SWO during sleep. Participants were awoken the following morning, learned a new PAT list, and then were tested on the original list. There was a significant effect of stimulation condition (p = 0.04297; Cohen’s d = 0.768), where verum stimulation resulted in reduced retroactive interference compared with sham and a significant interaction of encoding strength and stimulation condition (p = 0.03591). Planned simple effects testing within levels of encoding revealed a significant effect of stimulation only for low-encoders (p = 0.0066; Cohen’s d = 1.075) but not high-encoders. We demonstrate here for the first time that CL-tACS during sleep can enhance the protective benefits on retroactive interference in participants who have lower encoding aptitude.
Intracranial electroencephalography (iEEG) enables us to record and modulate neuronal responses from the level of macroscopic networks to small assemblies and single cells in both cortical and subcortical regions of the human brain with high spatial and temporal precision, offering significant methodological advantages over other non-invasive imaging and stimulating technologies. Leveraging these technical strengths of iEEG, in combination with sophisticated multivariate analytical approaches, researchers have obtained unprecedented insights into many long-standing problems in cognitive neuroscience. This chapter aims to illustrate these contributions, focusing on human memory. In particular, we describe how iEEG could advance our understanding of (1) the dynamic and transformative nature of short-term and long-term memory representations; (2) the role of hippocampal high-frequency neural activities, especially ripple activities, in memory formation, consolidation, and retrieval; (3) the information coding scheme of single-neuronal activity in the hippocampus and other brain regions; and (4) the common and different neural mechanisms between humans, primates and rodents. Moreover, we briefly discuss how iEEG studies can contribute to developing the state-of-the-art brain-computer interface and closed-loop brain stimulation. We conclude by summarizing the strengths and limitations of the iEEG method and providing practical guidance on how to choose between iEEG and other methods.
Transcranial magnetic stimulation (TMS) is widely employed as a tool to investigate and treat brain diseases. However, little is known about the direct effects of TMS on the brain. Non-human primates (NHPs) are a valuable translational model to investigate how TMS affects brain circuits given their neurophysiological similarity with humans and their capacity to perform complex tasks that approach human behavior. This systematic review aimed to identify studies using TMS in NHPs as well as to assess their methodological quality through a modified reference checklist. The results show high heterogeneity and superficiality in the studies regarding the report of the TMS parameters, which have not improved over the years. This checklist can be used for future TMS studies with NHPs to ensure transparency and critical appraisal. The use of the checklist would improve methodological soundness and interpretation of the studies, facilitating the translation of the findings to humans. The review also discusses how advancements in the field can elucidate the effects of TMS in the brain.
Creativity is a crucial element regarding the adaption of many animals to their environment. However, the exact neurobiological underpinnings of creativity are elusive. However, recent research provides more and more evidence that creativity is based on an optimal balance of flexibility and stability of brain network interaction. One theory is that there exists a dopamine-dependent balance between distinct brain regions modulating cognitive focus and flexibility. An imbalance between these areas may have significant effects on creativity as has been demonstrated in patients with Parkinson’s disease, for example, who have shown changes in creativity. To further understand these relationships, it is essential to implement causative approaches to validate neurobiological-based theories. However, the current established approaches lack the ability to specifically target mechanisms that are the key players in creativity. A novel brain stimulation method, low-intensity focused ultrasound, may be a promising technique to overcome these shortcomings and provide unique insights into the neuronal mechanisms underlying creativity and its alterations in neurodegenerative diseases.
Full-text available
Alzheimer's disease (AD) not only involves loss of memory functions but also prominent deterioration of sleep physiology, already evident in the stage of mild cognitive impairment (MCI). Cortical slow oscillations (SO, 0.5-1 Hz) and thalamo-cortical spindle activity (12-15 Hz) during sleep, and their temporal coordination, are considered critical for memory formation. We investigated the potential of slow oscillatory transcranial direct current stimulation (so-tDCS), applied during a daytime nap in a sleep state-dependent manner, to modulate these activity patterns and sleep-related memory consolidation in 9 male and 7 female human patients with MCI.Stimulation significantly increased overall SO and spindle power, amplified spindle power during SO up-phases, and led to stronger synchronization between SO and spindle power fluctuations in electroencephalographic recordings. Moreover, visual declarative memory was improved by so-tDCS compared to sham stimulation, associated with stronger synchronization. These findings indicate a well-tolerated therapeutic approach for disordered sleep physiology and memory deficits in MCI patients and advance our understanding of offline memory consolidation.SIGNIFICANCE STATEMENTIn the light of increasing evidence that sleep disruption is crucially involved in the progression of Alzheimer's disease (AD) sleep appears as a promising treatment target in this pathology - in particular, to counteract memory decline. This study demonstrates the potential of a noninvasive brain stimulation method during sleep in patients with mild cognitive impairment (MCI), a precursor of AD, and advances our understanding of its mechanism. We provide first time evidence that slow oscillatory transcranial stimulation amplifies the functional cross-frequency coupling between memory-relevant brain oscillations, and improves visual memory consolidation in patients with MCI.
Full-text available
Transcranial electric stimulation (TES) is an emerging technique, developed to non-invasively modulate brain function. However, the spatiotemporal distribution of the intracranial electric fields induced by TES remains poorly understood. In particular, it is unclear how much current actually reaches the brain, and how it distributes across the brain. Lack of this basic information precludes a firm mechanistic understanding of TES effects. In this study we directly measure the spatial and temporal characteristics of the electric field generated by TES using stereotactic EEG (s-EEG) electrode arrays implanted in cebus monkeys and surgical epilepsy patients. We found a small frequency dependent decrease (10%) in magnitudes of TES induced potentials and negligible phase shifts over space. Electric field strengths were strongest in superficial brain regions with maximum values of about 0.5 mV/mm. Our results provide crucial information of the underlying biophysics in TES applications in humans and the optimization and design of TES stimulation protocols. In addition, our findings have broad implications concerning electric field propagation in non-invasive recording techniques such as EEG/MEG.
Full-text available
The <1Hz slow oscillation (SO) and spindles are hallmarks of mammalian non-rapid eye movement and slow wave sleep. Spindle activity occurring phase-locked to the SO is considered a candidate mediator of memory consolidation during sleep. We used source localization of magnetoencephalographic (MEG) and electroencephalographic (EEG) recordings from 11 sleeping human subjects for an in-depth analysis of the temporal and spatial properties of sleep spindles co-occurring with SOs. Slow oscillations and spindles were identified in the EEG and related to the MEG signal, providing enhanced spatial resolution. In the temporal domain, we confirmed a phase-locking of classical 12-15Hz fast spindle activity to the depolarizing SO up-state and of 9-12Hz slow spindle activity to the up-to-down-state transition of the SO. In the spatial domain, we show a broad spread of spindle activity, with less distinct anterior-posterior separation of fast and slow spindles than commonly seen in the EEG. We further tested a prediction of current memory consolidation models, namely the existence of a spatial bias of SOs over sleep spindles as a mechanism to promote localized neuronal synchronization and plasticity. In contrast to that prediction, a comparison of SOs dominating over the left vs. right hemisphere did not reveal any signs of a concurrent lateralization of spindle activity co-occurring with these SOs. Our data are consistent with the concept of the neocortical SO exerting a top-down control over thalamic spindle generation. However, they call into question the notion that SOs locally coordinate spindles and thereby inform spindle-related memory processing.
Transcranial electric stimulation aims to stimulate the brain by applying weak electrical currents at the scalp. However, the magnitude and spatial distribution of electric fields in the human brain are unknown. Here we measure electric potentials intracranially in ten epilepsy patients and estimate electric fields across the entire brain by leveraging calibrated current-flow models. Electric field magnitudes at the cortical surface reach values of 0.4 V/m, which is at the lower limit of effectiveness in animal studies. When individual anatomy is taken into account, the predicted electric field magnitudes correlate with the recorded values (r=0.89 and r=0.84 in cortical and depth electrodes, respectively). Modeling white matter anisotropy and different skull compartments does not improve accuracy, but correct magnitude estimates require an adjustment of conductivity values used in the literature. This is the first study to validate and calibrate current-flow models with in vivo intracranial recordings in humans, providing a solid foundation for targeting of stimulation and interpretation of clinical trials.
Background: Transcranial direct current stimulation (tDCS) has been reported to improve various forms of learning in humans. Stimulation is often applied during training, producing lasting enhancements that are specific to the learned task. These learning effects are thought to be mediated by altered synaptic plasticity. However, the effects of DCS during the induction of endogenous synaptic plasticity remain largely unexplored. Objective/Hypothesis: Here we are interested in the effects of DCS applied during synaptic plasticity induction. Methods: To model endogenous plasticity we induced long-term potentiation (LTP) and depression (LTD) at Schaffer collateral synapses in CA1 of rat hippocampal slices. Anodal and cathodal DCS at 20 V/m were applied throughout plasticity induction in both apical and basal dendritic compartments. Results: When DCS was paired with concurrent plasticity induction, the resulting plasticity was biased towards potentiation, such that LTP was enhanced and LTD was reduced. Remarkably, both anodal and cathodal stimulation can produce this bias, depending on the dendritic location and type of plasticity induction. Cathodal DCS enhanced LTP in apical dendrites while anodal DCS enhanced LTP in basal dendrites. Both anodal and cathodal DCS reduced LTD in apical dendrites. DCS did not affect synapses that were weakly active or when NMDA receptors were blocked. Conclusions: These results highlight the role of DCS as a modulator, rather than inducer of synaptic plasticity, as well as the dependence of DCS effects on the spatial and temporal properties of endogenous synaptic activity. The relevance of the present results to human tDCS should be validated in future studies.
Transcranical direct current stimulation (tDCS) is a treatment known to ameliorate various neurological conditions and enhance memory and cognition in humans. tDCS has gained traction for its potential therapeutic value; however, little is known about its mechanism of action. Using a transgenic mouse expressing G-CaMP7 in astrocytes and a subpopulation of excitatory neurons, we find that tDCS induces large-amplitude astrocytic Ca²⁺ surges across the entire cortex with no obvious changes in the local field potential. Moreover, sensory evoked cortical responses are enhanced after tDCS. These enhancements are dependent on the alpha-1 adrenergic receptor (A1AR) and are not observed in IP3R2 (inositol trisphosphate receptor type 2) knockout mice, in which astrocytic Ca2+ surges are absent. Together, we propose that tDCS changes the metaplasticity of the cortex through astrocytic Ca2⁺/IP3 signalling. Moreover, the stimulation parameters were found to be sufficient to alleviate a mouse model of depression by chronic restraint stress. • Download high-res image (285KB) • Download full-size image
Background: A 2006 trial in healthy medical students found that anodal slow oscillating tDCS delivered bi-frontally during slow wave sleep had an enhancing effect in declarative, but not procedural memory. Although there have been supporting animal studies, and similar findings in pathological groups, this study has not been replicated, or refuted, in the intervening years. We therefore tested these earlier results for replication using similar methods with the exception of current waveform (square in our study, nearly sinusoidal in the original). Objective/hypothesis: Our objective was to test the findings of a 2006 trial suggesting bi-frontal anodal tDCS during slow wave sleep enhances declarative memory. Methods: Twelve students (mean age 25, 9 women) free of medical problems underwent two testing conditions (active, sham) in a randomized counterbalanced fashion. Active stimulation consisted of oscillating square wave tDCS delivered during early Non-Rapid Eye Movement (NREM) sleep. The sham condition consisted of setting-up the tDCS device and electrodes, but not turning it on during sleep. tDCS was delivered bi-frontally with anodes placed at F3/F4, and cathodes placed at mastoids. Current density was 0.517 mA/cm2, and oscillated between zero and maximal current at a frequency of 0.75 Hz. Stimulation occurred during five–five minute blocks with 1-min inter-block intervals (25 min total stimulation). The primary outcomes were both declarative memory consolidation measured by a paired word association test (PWA), and non-declarative memory, measured by a non-dominant finger-tapping test (FTT). We also recorded and analyzed sleep EEG. Results: There was no difference in the number of paired word associations remembered before compared to after sleep [(active = 3.1 ± 3.0 SD more associations) (sham = 3.8 ± 3.1 SD more associations)]. Finger tapping improved, (non-significantly) following active stimulation [(3.6 ± 2.7 SD correctly typed sequences) compared to sham stimulation (2.3 ± 2.2 SD correctly typed sequences)]. Conclusion: In this study, we failed to find improvements in declarative or performance memory and could not replicate an earlier study using nearly identical settings. Specifically we failed to find a beneficial effect on either overnight declarative or non-declarative memory consolidation via square-wave oscillating tDCS intervention applied bi-frontally during early NREM sleep. It is unclear if the morphology of the tDCS pulse is critical in any memory related improvements.
Transient episodes of brain oscillations are a common feature of both the waking and the sleeping brain. Sleep spindles represent a prominent example of a poorly understood transient brain oscillation that is impaired in disorders such as Alzheimer’s disease and schizophrenia. However, the causal role of these bouts of thalamo-cortical oscillations remains unknown. Demonstrating a functional role of sleep spindles in cognitive processes has, so far, been hindered by the lack of a tool to target transient brain oscillations in real time. Here, we show, for the first time, selective enhancement of sleep spindles with non-invasive brain stimulation in humans. We developed a system that detects sleep spindles in real time and applies oscillatory stimulation. Our stimulation selectively enhanced spindle activity as determined by increased sigma activity after transcranial alternating current stimulation (tACS) application. This targeted modulation caused significant enhancement of motor memory consolidation that correlated with the stimulation-induced change in fast spindle activity. Strikingly, we found a similar correlation between motor memory and spindle characteristics during the sham night for the same spindle frequencies and electrode locations. Therefore, our results directly demonstrate a functional relationship between oscillatory spindle activity and cognition.