www.sciencedirect.com

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 7

Available online at

ScienceDirect

Journal homepage: www.elsevier.com/locate/cortex

Research report

Cortical inhibition of distinct mechanisms in thedorsolateral prefrontal cortex is related to workingmemory performance: A TMSeEEG study

Nigel C. Rogasch a,b,*, Zafiris J. Daskalakis c and Paul B. Fitzgerald a

a Monash Alfred Psychiatry Research Centre, Central Clinical School, The Alfred and Monash University, Melbourne,

Australiab Monash Clinical and Imaging Neuroscience, School of Psychological Science and Monash Biomedical Imaging,

Monash University, Melbourne, Australiac Temerty Centre for Therapeutic Brain Intervention, Centre for Addiction and Mental Health, University of Toronto,

Toronto, Canada

a r t i c l e i n f o

Article history:

Received 19 May 2014

Reviewed 28 July 2014

Revised 23 August 2014

Accepted 7 October 2014

Action editor Jacinta O’Shea

Published online 22 October 2014

Keywords:

Transcranial magnetic stimulation

Electroencephalography

Cortical inhibition

Dorsolateral prefrontal cortex

Working memory

Abbreviations: GABA, ɣ-amino butyric acidevoked potential; TEP, TMS-evoked corticaabductor pollicus brevis; RMT, resting motor* Corresponding author. Monash Clinical a

Australia.E-mail address: nigel.rogasch@monash.e

http://dx.doi.org/10.1016/j.cortex.2014.10.0030010-9452/© 2014 Elsevier Ltd. All rights rese

a b s t r a c t

Paired-pulse transcranialmagnetic stimulationcombinedwithelectroencephalography (TMS

eEEG) is a method for studying cortical inhibition from the dorsolateral prefrontal cortex

(DLPFC). However, little is known about the mechanisms underlying TMS-evoked cortical

potentials (TEPs) from this region, let alone inhibition of these components. The aim of this

studywas toassess cortical inhibitionofdistinctTEPsandoscillations in theDLPFCusingTMS

eEEGand to investigate the relationship of thesemechanisms toworkingmemory. 30healthy

volunteers received single and paired (interstimulus interval ¼ 100 msec) TMS to the left

DLPFC. Variations in long-interval cortical inhibition (LICI) of different TEP peaks (N40, P60,

N100) and different TMS-evoked oscillations (alpha, lower beta, upper beta, gamma) were

compared between individuals. Variation in N100 slope following single pulse TMS, another

putative marker of inhibition, was also compared with LICI of each measure. Finally, these

measures were correlated with performance of a working memory task. LICI resulted in sig-

nificant suppression of all TEP peaks and TMS-evoked oscillations (all p < .05). There were no

significant correlations between LICI of different TEP peaks or TMS-evoked oscillations with

the exception of P60 and N100. Variation in N100 slope correlated with LICI of N40 and beta

oscillations. In addition, LICI of P60 and N100 were differentially correlated with working

memory performance. The results suggest that both the LICI paradigm and N100 following

single pulse TMS reflect complementary methods for assessing GABAB-mediated cortical

inhibition in the DLPFC. Furthermore, these measures demonstrate the importance of pre-

frontal GABAB-mediated inhibitory control for working memory performance.

© 2014 Elsevier Ltd. All rights reserved.

; TMS, transcranial magnetic stimulation; LICI, long-interval cortical inhibition; MEP, motorl potentials; EEG, electroencephalography; DLPFC, dorsolateral prefrontal cortex; APB,threshold.

nd Imaging Neuroscience, Building 220, Monash University, Melbourne, Victoria, 3800,

du (N.C. Rogasch).

rved.

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 7 69

1. Introduction

Cortical inhibition refers to suppression of neuronal activity

and is a fundamental mechanism for both the generation and

control of coordinated cortical network activity (Isaacson &

Scanziani, 2011). In the mature cortex, cortical inhibition is

largely governed by the neurotransmitter ɣ-amino butyric acid

(GABA), which alters polarization of neuronal membranes via

fast acting GABAA receptors and slower acting GABAB-re-

ceptors (Krnjevi�c, 1997). The dynamic properties of these re-

ceptor sub-types appear to serve different functional roles.

GABAA receptors are fundamental for generating fast, coor-

dinated network activity such as gamma oscillations

(30e80 Hz) (Cardin et al., 2009; Sohal, Zhang, Yizhar, &

Deisseroth, 2009; Whittington, Traub, & Jefferys, 1995), how-

ever the functional role of GABAB-mediated inhibition is less

clear. Recent work has suggested that GABAB-mediated

cortical inhibition plays an important role in modulating

cortical network activity (Kohl & Paulsen, 2010). Importantly,

dysfunction of GABAB-mediated inhibition may play a crucial

role in neurological and psychiatric conditions that are

thought to result from impaired control of network activity,

such as epilepsy (Schuler et al., 2001) and schizophrenia

(Daskalakis & George, 2009; Rogasch, Daskalakis, & Fitzgerald,

2014).

In humans, GABAB-mediated cortical inhibition can be

assessed using transcranial magnetic stimulation (TMS).

TMS utilizes electromagnetic induction to non-invasively

depolarize excitatory and inhibitory cortical neurons across

the scalp (Barker, Jalinous, & Freeston, 1985). When a

suprathreshold TMS pulse is preceded by a suprathreshold

conditioning pulse at intervals of 50e200 msec (i.e., paired-

pulse TMS), TMS-evoked neuronal activity is suppressed

through a process known as long-interval cortical inhibition

(LICI). This can be measured as either a decrease in motor

cortical output via motor evoked potentials (MEPs) in pe-

ripheral muscles (Nakamura, Kitagawa, Kawaguchi, & Tsuji,

1997; Valls-Sol�e, Pascual-Leone, Wassermann, & Hallett,

1992) or modulation of TMS-evoked cortical potentials (TEPs)

assessed directly from the cortex using electroencephalog-

raphy (EEG) (Daskalakis, Farzan, Barr, Maller, et al., 2008;

Fitzgerald et al., 2008). In addition to paired-pulse para-

digms, a growing body of evidence suggests that the N100, a

negative TEP, also represents GABAB-mediated inhibitory

function. For instance, different motor tasks modulate N100

amplitude in a way consistent with cortical inhibition

(Bonnard, Spieser, Meziane, de Graaf, & Pailhous, 2009;

Bruckmann et al., 2012; Kici�c, Lioumis, Ilmoniemi, &

Nikulin, 2008; Nikulin, Kicic, Kahkonen, & Ilmoniemi, 2003;

Spieser, Meziane, & Bonnard, 2010), N100 amplitude corre-

lates with motor measures of inhibition including LICI

(Rogasch, Daskalakis, & Fitzgerald, 2013) and the silent

period (Farzan et al., 2013) and the N100 over motor cortex is

specifically increased by a GABAB-receptor agonist (Premoli

et al., 2014).

Although useful for studying motor physiology, the real

strength of combined TMSeEEG is in studying mechanisms

outside the motor cortex. LICI of both TMS-evoked activity

(Fitzgerald et al., 2008; Fitzgerald, Maller, Hoy, Farzan, &

Daskalakis, 2009; Daskalakis, Farzan, Barr, Maller, et al.,

2008) and TMS-evoked oscillations (Farzan et al., 2010a, 2009)

has been demonstrated from the dorsolateral prefrontal cor-

tex (DLPFC) using TMSeEEG. In addition, prefrontal LICI

strength correlates with individual performance on a working

memory task (Daskalakis, Farzan, Barr, Rusjan, et al., 2008;

Hoppenbrouwers et al., 2013), providing preliminary evi-

dence for a role of GABAB-mediated inhibition in cognition.

However, little is known about the mechanisms that underlie

TEPs or TMS-evoked oscillations from the DLPFC, let alone

inhibition of these measures. In addition, it remains unclear

whether LICI suppresses TMS-evoked outputs to other cortical

regions as well as local activity.

Themajority of studies assessing LICI from the DLPFC have

collapsed analysis across time, removing information on

distinct mechanisms made possible by analysing specific TEP

peaks. Therefore, the aim of this study was to compare LICI of

distinct TEP peaks and TMS-evoked oscillatory bands in the

DLPFC and to assess the physiological and functional rele-

vance of these measures. We assessed natural variation in

LICI strength across a population of healthy volunteers using

single and paired-pulse TMSeEEG. First, we assessed whether

variation in LICI of different TEP peaks and different TMS-

evoked oscillations from DLPFC were related or independent.

Second, we evaluated whether the N100 slope following

single-pulse TMS was associated with LICI of TEPs and LICI of

TMS-evoked oscillations. Third, we assessed whether LICI

suppressed TMS-evoked activity and oscillations across the

scalp as well as at the site of stimulation. Finally, to assess the

functional relevance of these measures, we investigated the

relationship between inhibition in DLPFC and working mem-

ory performance.

2. Materials and methods

2.1. Participants

30 volunteers participated in the current study (32.2 ± 11

years, 8 female). Volunteers had no history of neurological or

psychiatric illnesses and provided informed written consent

before commencement of the study. All experimental pro-

cedures were approved by the Monash University, Alfred

Hospital and Centre for Addiction and Mental Health Human

Research Ethics Committees in accordance with the declara-

tion of Helsinki.

2.2. Procedures

Participants were seated comfortably with their hands resting

in their lap. An EEG cap was fitted to their head and electrodes

were placed over the right abductor pollicus brevis (APB)

muscle for electromyographic recordings. Resting motor

threshold (RMT) and the TMS intensity required to evoke an

MEP of ~1 mV were then determined over the motor cortical

region that produced the largest responses in APB. Following

motor measures, the coil was positioned so the centre rested

between the F3 and F5 electrode and the handlewas rotated to

a 45� angle relative to midline, producing a posterioreanterior

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 770

current flow in the underlying cortex. This position provides

the most accurate estimation of left DLPFC (border of BA9 and

BA46) in the absence of neuronavigational equipment

(Fitzgerald, Maller, Hoy, Thomson, & Daskalakis, 2009; Rusjan

et al., 2010). The coil borderwasmarked using a felt tipped pen

to allow repositioning. Participants then received 75 single

and 75 paired (interstimulus interval ¼ 100 msec) TMS pulses

to the left DLPFC while they sat quietly and looked directly

ahead with their eyes open. Both single and paired pulses

were delivered at the intensity required for 1 mV (maximum

stimulator output ¼ 68.2 ± 11%) and were randomly inter-

leaved at a rate of .2 Hz (10% jitter). Nine participants received

50 single and 50 paired sham TMS pulses prior to real TMS to

assess auditory-evoked potentials resulting from paired pulse

TMS (see Supplementary materials and methods). Twenty

participants also performed a Sternbergworkingmemory task

with both a low (5 letters) and high (7 letters) load condition

(Bailey, Segrave, Hoy, Maller, & Fitzgerald, 2014). The order of

TMS and working memory were counterbalanced across par-

ticipants to account for any potential order effects. A detailed

description of the procedures is provided in the

Supplementary Materials and Methods.

2.3. Analysis

Analysis of EEG data were performed using EEGLAB (Delorme

& Makeig, 2004), fieldtrip (Oostenveld, Fries, Maris, &

Schoffelen, 2011) and custom scripts on the MATLAB plat-

form (R2013a, The Mathworks, USA). Data were epoched

around the TMS pulse, baseline corrected, the TMS artifact

removed, downsampled and filtered (Rogasch, Thomson,

Daskalakis, & Fitzgerald, 2013). Independent component

analysis was used to remove additional artifacts (Rogasch,

Thomson, et al., 2014). A detailed description of the EEG

analysis and artifact removal is provided in the

Supplementary Materials and Methods.

2.4. TEPs

The effect of paired pulse TMS on TEPs was analysed in two

ways; a region of interest (ROI) analysis to evaluate the local

effects of paired pulse TMS and global scalp analysis to assess

the effect of paired pulse TMS on outputs from the DLPFC

across the cortex. For all analyses, TEPs following the single

test pulsewere comparedwith TEPs following the conditioned

test pulse (i.e., the second pulse in the paired condition). For

the ROI analysis, three peakswere analysed at the F3 electrode

for both single and paired pulses consistent with our previous

studies; N40, P60 and N100 (Rogasch, Thomson, et al., 2014,

2013). The N40 peak latency was calculated at the negative

peak closest to 40 msec (25e55 msec), the P60 as the positive

peak closest to 60 msec (45e75 msec) and the N100 the

negative peak closest to 100 msec (85e145 msec). The peak

amplitude was calculated as the average signal between

±5 msec of the single pulse peak latency for N40 and P60 and

±10 msec for N100. LICI of TEPs was quantified by normalising

the difference between unconditioned (single test pulse) and

conditioned (paired test pulse) TEPs to the overall size of the

TEP (25e150 msec):

EquationA : LICIN40� � � �

¼ N40single�N40paired = Minsingle�Maxsingle �100

Equation B : LICIP60

¼ �P60single � P60paired

�=�Maxsingle �Minsingle

�� 100

EquationC : LICIN100

¼ �N100single�N100paired

�=�Minsingle�Maxsingle

��100

TEPs last for up to 300 msec following stimulation and

represent activity from multiple sources across the brain.

Therefore, TEPs resulting from the conditioning pulse are

likely still present following the test pulse in the paired con-

dition. As the main comparison of interest in this study is size

of the TEPs generated by the test pulse following both single

and paired conditions, the ongoing TEPs generated by the

conditioning pulse could ‘contaminate’ these measures. In-

spection of the trace following the test pulse in the paired

condition (i.e., the period of interest to quantify LICI) revealed

several peaks that were consistent with ongoing TEPs gener-

ated by the conditioning pulse (Fig. S2A). To further demon-

strate this point, the single-pulse trace was time-shifted

100 msec to match the conditioning pulse of the paired-pulse

condition (Fig. S2B). In order to recover the true TEP activity

generated by the test pulse in the paired condition uncon-

taminated by activity from the conditioning pulse, activity

from the time-shifted single pulse trace was subtracted from

activity following the conditioning pulse (Fig. S2C). Theoreti-

cally, this should leave only TEP activity resulting from the test

pulse in the paired condition. The ‘corrected’ TEPs following

the test pulse in the paired pulse condition were then

compared with the TEPs from the single pulse condition and

LICI was calculated again using the above formulas.

Slope analysis was use to compare the inhibitorymeasures

of LICI and the N100 following single pulse TMS (Huber et al.,

2013; Rogasch, Daskalakis, et al., 2013). N100 slope was

determined by calculating the mean first derivative between

90 and 98 msec following single pulse TMS, a period that

corresponds to the inhibitory period of LICI. The effect of LICI

on output from the DLPFC was assessed by comparing TEPs

evoked by the unconditioned and conditioned TMS pulse over

all electrodes between 25 and 300 msec.

2.5. TMS-evoked oscillations

The effect of paired pulseTMSonTMS-evoked oscillationswas

also assessedusing an ROI analysis and a global scalp analysis.

For paired-pulse conditions, this analysis was performed on

the corrected test pulsedata (seeSection 2.4). As forTEPs, TMS-

evoked oscillations following the single test pulse were

compared with TMS-evoked oscillations following the condi-

tioned test pulse.TMS-evokedoscillationswereobtainedusing

wavelet decomposition (3.5 oscillation cycles, steps of 1 Hz

between 8 and 45 Hz) on averaged trials for each individual

electrode.Oscillationswerenormalisedbydividing the relative

postTMSpowerbymeanbaselinepower (�600 to�100) in each

frequency bin. For ROI analysis at the F3 electrode, oscillations

were averaged across frequency bands (alpha: 8e12 Hz, lower

beta: 13e20 Hz, upper beta: 12e30 Hz, gamma: 31e45 Hz) and

Fig. 1 e LICI of TEPs following paired-pulse TMS at the F3

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 7 71

across time (25e200 msec for alpha and lower beta,

25e100 msec for upper beta and gamma). LICI of TMS-evoked

oscillations was quantified by normalising the difference be-

tween unconditioned and conditioned oscillations in each

frequency band to the mean power across all frequencies:

Equation D : LICIfreq ¼�Singlefreq � Pairedfreq

�=Singleall � 100

The effect of LICI on oscillations evoked in distant cortical

regions was assessed by comparing the difference between un-

conditioned and conditioned TMS-evoked oscillations over time

(25e250 msec), frequency (8e45 Hz) and space (all electrodes).

2.6. Working memory performance

To quantify working memory performance, both accuracy (%

correct) and reaction time to correct response were calculated

for low and high load conditions.

2.7. Statistics

Comparisons of single and paired pulses on ROI analyses (peak

amplitudes, TMS-evoked frequency bands) were made be-

tween conditioned and unconditioned responses indepen-

dently for each peak and frequency using paired t-tests. This

approach was chosen as comparisons between peaks or fre-

quencies were not of interest at the group level and the

hypothesised directional effect of paired TMS was different

between peaks (paired TMS was expected to result in a more

positive N40 and N100 and more negative P60 compared with

single TMS). Correlations between LICImeasures (different TEP

peaks and different TMS-evoked oscillation frequencies) and

between N100 slope and LICI measures were performed using

Pearson's correlations (with and without outliers removed;

outlier >3 � SD). Comparisons of single and paired pulses on

global scalpTEPs (space and time) and global scalp TMS-evoked

oscillations (space, time and frequency) were performed using

non-parametric, cluster-based permutation statistics (Maris &

Oostenveld, 2007). This approach controls for multiple-

comparisons across any combination of space, time and fre-

quency. Clusters were defined as two or more neighboring

electrodes in which the t-statistic at a given time or frequency

point exceeded a threshold of p < .05 (dependent t-test). Elec-

trodes, timepoints and/or frequencybinswithabove-threshold

values were used for subsequent cluster-based permutation

analysis.MonteCarlo p-valueswere calculated on2000 random

permutations and a value of p < .05 was used as the cluster-

statistic significance threshold for all analyses. Pearson's cor-

relationswereused tocompareLICI ofTEPs, LICI ofTMS-evoked

oscillationsandN100slopewithworkingmemoryaccuracyand

reaction times across low and high loads separately.

electrode. A) Grand averaged TEPs from the F3 electrode

following single, paired and paired (corrected) TMS. Times

used for analysis of each peak are indicated by the grey

bars. B) N40 peak amplitude following single and paired

TMS. C) P60 peak amplitude following single and paired

TMS. D) N100 peak amplitude following single and paired

TMS. E) Correlation between N100 slope and LICI of N40

peak. *indicates p < .05 compared with single pulse

amplitude.

3. Results

3.1. TEPs

3.1.1. ROI analysisSingle-pulse TMS over DLPFC resulted in three consistent

peaks measured at the F3 electrode (directly under the coil); a

negative peak at 40.3 ± 7 msec (N40; distinguishable in 29

participants), a positive peak at 61.9 ± 11 msec (P60; distin-

guishable in 29 participants) and a second negative peak at

113.2 ± 14 msec (N100; distinguishable in 29 participants).

Paired-pulse TMS resulted in TEPs after both the conditioning

and test pulses (Fig. S2). Inspection of the corrected paired-

pulse trace revealed improved matching of shape between

the single and paired test-pulses, suggesting the subtraction

approach was successful (Fig. 1A, Fig. S2).

To quantify LICI, differences in N40, P60 and N100 ampli-

tude following the test stimulus were compared between the

single-pulse condition and both the uncorrected and

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 772

corrected paired-pulse condition at the F3 electrode. N40 was

significantly reduced following paired-pulse TMS in both

conditions [uncorrected, t(28) ¼ �2.2, p ¼ .019; t(28) ¼ �2.0,

corrected, p ¼ .027], P60 was significantly reduced in the cor-

rected [t(28) ¼ 2.4, p ¼ .011] and trended toward significance in

the uncorrected condition [t(28) ¼ 1.3, p ¼ .098] and N100 was

significantly reduced in both conditions [uncorrected,

t(28) ¼ �5.4, p < .001; corrected, t(28) ¼ �2.8, p ¼ .005;

Fig. 1BeD].

To assess whether LICI measured from different peaks

reflected inhibition of related or independent mechanisms,

correlation analyseswere performed betweennormalised LICI

scores from each peak. There was no significant relationship

between LICI measuredwith N40 and P60 (uncorrected p¼ .32,

corrected p ¼ .36) or N40 and N100 (uncorrected p ¼ .67, cor-

rected p ¼ .37), however, LICI measured with P60 significantly

correlated with N100 for the corrected condition (r ¼ �.539,

p ¼ .003) and trended toward significance in the uncorrected

condition (p ¼ .096).

To assess whether the N100 following single pulse TMS

conveys similar inhibitory information to that as LICI

measured with paired-pulses, the slope of the N100

(90e98msec, analogous to the timing of LICI) from single pulse

TMSwas comparedwith normalised LICImeasured from each

peak. The N100 slope significantly correlated with LICI from

corrected N40 (Fig. 1E) and trended towards significance with

uncorrected N40, but did not correlate with any other peak

(Table 1). To assess whether the association with LICI

measured from N40 was specific to the N100 slope, correla-

tions between the preceding negative slope (60e70 msec) and

positive slope (80e90msec) were also assessed. Therewere no

significant correlations between LICI of any peak and either

slope (all p > .05). Finally, the N100 slope was also significantly

correlated with the N100 slope of the conditioning pulse (�10

to �2 msec, i.e., the slope immediately preceding the test

pulse in the paired condition; r ¼ .383, p ¼ .040). Taken

together, these findings suggest that the slope of the N100

represents a similar mechanism to LICI of N40. As correcting

or not correcting the paired-pulse trace gave similar results

and correcting resulted in a more plausible shape of the test

TEP, the remaining analysis was completed on corrected

paired-pulse traces.

3.1.2. Global scalp analysisTo assess the effect of LICI across the scalp, TEPs were

compared between single and corrected paired conditions

across space and time using cluster-based statistics. One

Table 1 e Correlations between N100 slope gradient andLICI of TEP peaks.

Correlation coefficient (r) p-Value

N40, uncorrected �.35 .06

N40, corrected �.47 .01*

P60, uncorrected �.08 .69

P60, corrected �.15 .43

N100, uncorrected �.07 .71

N100 corrected .07 .73

*p < .05.

significant negative cluster (109e283 msec, p < .001; paired

signal more negative than single) and one significant positive

cluster (138e295 msec, p < .001; paired signal more positive

than single) was observed indicating reduced overall EEG

signal following paired-pulse TMS compared with single-

pulse TMS (Fig. 2). Inspection of topographic plots revealed

that inhibition began in a cluster of electrodes close to the site

of stimulation and spread across the centre of the scalp to

contralateral frontal/temporal regions. Over the next 30 msec,

inhibition spread posteriorly and extended over bilateral pa-

rietal and occipital regions while also developing over a

fronto-central cluster. Inhibition of early latency TEPs did not

survive correction for multiple comparisons at the global

level.

3.2. TMS-evoked oscillations

3.2.1. ROI analysisTo quantify LICI of TMS-evoked oscillations, differences in

TMS-evoked alpha, lower beta, upper beta and gamma oscil-

lations were compared between single and corrected paired-

pulse conditions at the F3 electrode. Evoked oscillations

were significantly reduced in lower beta [t(29) ¼ 2.3, p ¼ .028],

upper beta [t(29) ¼ 3.0, p ¼ .005] and gamma [t(29) ¼ 2.9,

p ¼ .008] frequency bands and trended toward significance in

the alpha band [t(29) ¼ 1.9, p ¼ .067] following paired-pulse

Fig. 2 e Butterfly plots averaged across all individuals

demonstrating the difference in TEP across all electrodes

following single- (blue lines) and paired-pulse (red lines)

TMS over DLPFC. Topoplots represent t-statistic values

across the scalp at time points marked by arrows

(blue ¼ paired more negative than single, red ¼ paired

more positive than single). White crosses represents

significantly different electrodes making up the cluster at

these time points (p < .05, cluster based statistics). The blue

bar represents timing of the significant negative cluster,

the red bar represents timing of the significant positive

cluster.

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 7 73

TMS (Fig. 3A, B). To assess whether inhibition of different

frequency bands represented similar or independent mecha-

nisms, LICI was correlated between oscillatory bands. Data

from one participant represented an extreme outlier and was

excluded from all correlation analyses. LICI of alpha trended

towards a significant correlation with LICI of gamma

(r ¼ �.361, p ¼ .054), however, there was no other significant

correlation between LICI of any frequency band (all p > .15).

To assess whether N100 following single pulse TMS was

associated with inhibition of any particular frequency, N100

slope (90e98 msec) was correlated with LICI strength of each

frequency band. N100 slope trended towards significant cor-

relations with lower beta (r ¼ �.264, p ¼ .17) and upper beta

inhibition (r ¼ �.291, p ¼ .11), but not alpha (p ¼ .68) or gamma

inhibition (p¼ .94). To further explore this relationship, LICI of

individual frequencies in the beta band (13e30 Hz) were ana-

lysed. Significant correlations were found between N100 slope

and LICI of oscillations from 16 to 23 Hz (p < .05), suggesting

individuals with steeper N100 slope gradients displayed

increased LICI of mid-beta oscillations (Fig. 3C). This rela-

tionship remained significant for oscillations between 17 and

22 Hz following removal of a second outlier (p < .05). There

were also no significant correlations between inhibition of any

frequency band and slopes at either 60e70 msec or

80e90 msec (all p > .05).

Fig. 3 e LICI of TEPs following paired-pulse TMS at the F3 elect

individuals following single and paired TMS (0 msec ¼ timing

frequency bands following single and paired TMS (alpha and lo

gamma ¼ 25e100 msec). D) Correlation between N100 slope an

remained significant with removal of the outlier (r ¼ ¡.328, p ¼

3.2.2. Global scalp analysisTo assess the effect of LICI across the scalp in the frequency

domain, TMS-evoked oscillations were compared between

single and corrected paired conditions across space, time and

frequency using cluster-based statistics. One significant pos-

itive cluster (p < .001) was observed spanning all electrodes

and multiple time points (25e185 msec; Fig. 4). TMS-evoked

alpha, lower beta, upper beta and gamma oscillations were

inhibited over bilateral frontal electrodes up to 150 msec.

Upper beta, gamma and particularly lower beta oscillations

were also inhibited between 50 and 160 msec over right

parieto-occiptal electrodes. Inhibition was strongest at the

site of stimulation and evolved over time to both frontal and

parieto-occipital regions over the contralateral hemisphere.

3.3. Relationship between prefrontal cortical inhibitionand working memory

LICI of P60 negatively correlated with low load accuracy

(r ¼ �.58, p < .01; Fig. 5A), LICI of N100 negatively correlated

with low load reaction time (r ¼ �.48, p¼ .04; Fig. 5B) and N100

slope trended towards a positive correlation with low load

accuracy (r ¼ .41, p ¼ .07). There were no significant relation-

ships between LICI of N40 or LICI of TMS-evoked oscillations

and working memory performance (all p > .05).

rode. A) Time-frequency plots from the averaged across all

of test pulse). B) TMS-evoked oscillations from different

wer beta ¼ 25e200 msec, upper beta and

d LICI of TMS-evoked oscillations at 20 Hz. The correlation

.045). *p < .05 compared with single.

Fig. 4 e Topoplots averaged across all individual

demonstrating the difference in TMS-evoked oscillations

across the scalp following single and paired TMS. Red

colours indicate a decrease in TMS-evoked oscillation

frequency power (relative to baseline) following paired-

pulse TMS. A) Inhibition of alpha and lower beta

frequencies between 25 and 200 msec. B) Inhibition of

upper beta and gamma oscillations between 25 and

125 msec.

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 774

4. Discussion

There are four novel findings to the current study. First, LICI of

early and late TEP peaks and LICI of different TMS-evoked

oscillations varied independently over DLPFC, suggesting

that each of these peaks/processes represents a different

underlying physiological mechanism. Second, the N100 slope

following single pulse TMS correlated with LICI of N40 and

mid-beta oscillations, suggesting that N100 over DLPFC also

Fig. 5 e The relationship between LICI of TEPs at the F3

electrode and working memory performance. A)

Correlation between LICI of P60 and low load working

memory accuracy. B) Correlation between LICI of N100 and

low load working memory reaction time.

directly represents cortical inhibition. Third, LICI suppressed

TMS-evoked activity and oscillations across the scalp as well

as at the site of stimulation, suggesting TMS-evoked outputs

from the DLPFC were also inhibited following paired-pulse

TMS. Finally, LICI of P60 and N100 were differentially related

to working memory performance.

4.1. Mechanisms of TEPs over DLPFC

TEPs reflect the summation of a complex interplay between

excitatory and inhibitory activity resulting from depolarisa-

tion of neuronal populations in the cortex following TMS

(Huber et al., 2013). However, characteristics of TEP peaks such

as amplitude, latency and polarity appear to reflect properties

of the dominant mechanism during a given period (Rogasch &

Fitzgerald, 2013). A growing body of evidence from the motor

cortex suggests that earlier TEP peaks (10e30 msec) reflect

excitatory neurotransmission, whereas later peaks

(40e50 msec, 100e200 msec) reflect inhibitory neurotrans-

mission mediated by GABAA and GABAB receptors, respec-

tively (Premoli et al., 2014). In support, we recently

demonstrated that LICI of early and late TEP peaks were

independently modulated following systematic changes in

conditioning and test intensity over motor cortex (Rogasch,

Daskalakis, et al., 2013). In the current study we have shown

that LICI of early and late TEPs in the DLPFC also vary inde-

pendently across individuals. In addition, we found that LICI

of different TMS-evoked oscillation frequencies vary inde-

pendently from each other. These findings support the notion

that independent mechanisms generate different peaks and

frequencies following TMS over DLPFC.

Despite the independence of TMS-evoked activity in the

DLPFC, little is known about the underlying mechanisms of

these peaks and frequencies. We found that the N100 slope

was associated with LICI of the N40 and mid-beta oscillations

over left DLPFC. Importantly, this relationshipwas not present

for slopes from other peaks and also correlated with the N100

slope generated by the conditioning pulse. The link between

N100 and LICI in DLPFC mirrors our recent finding of a similar

relationship in the motor cortex (Rogasch, Daskalakis, et al.,

2013). Although good evidence exists that the motor cortex

N100 is related to cortical inhibition (Bender et al., 2005;

Bonnard et al., 2009; Bruckmann et al., 2012; Farzan et al.,

2013; Kici�c et al., 2008; Komssi, K€ahk€onen, & Ilmoniemi,

2004; Nikulin et al., 2003; Premoli et al., 2014; Spieser et al.,

2010), this is the first evidence we are aware of directly link-

ing the DLPFC N100 with cortical inhibition. To confirm this

relationship over DLPFC, pharmacological interventions using

selective GABAB-receptor agonists are required.

We have also demonstrated that LICI over DLPFC results in

suppression of all TEP peaks (N40, P60 andN100). Although the

mechanistic origin of the N40 and P60 over DLPFC remain

unclear, we have provided evidence in this study that the

N100 is consistent with a GABAB-mediated inhibitory poten-

tial. Therefore, suppression of the N100 with LICI may repre-

sent suppression of inhibitory activity in the DLPFC. Both rat

and human brain slice experiments have demonstrated that

paired-pulse depression of GABAB-mediated postsynaptic

potentials results from reduced inhibitory output following

presynaptic inhibition of inhibitory interneurons (Davies,

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 7 75

Davies,& Collingridge, 1990; Deisz, 1999; Deisz& Prince, 1989).

TMS studies in motor cortex have also provided evidence for

presynaptic inhibition, with LICI suppressing activity of other

inhibitory networks such as those mediating short-interval

cortical inhibition (Cash, Ziemann, Murray, & Thickbroom,

2010; Sanger, Garg, & Chen, 2001). In addition, we have also

found that LICI results in N100 suppression in a paired-pulse

TMSeEEG study over motor cortex (Rogasch, Daskalakis,

et al., 2013). Taken together, these findings suggest that LICI

of N100 may represent a novel method for measuring pre-

synaptic inhibition in the DLPFC. Further paired-pulse exper-

iments are required to strengthen this finding.

4.2. Suppression of DLPFC outputs by LICI

A single TMS pulse to the DLPFC not only activates local intra-

cortical networks, but also results in activation of (presum-

ably) connected cortical regions. For example, we recently

used source localisation to demonstrate that time-dependent

changes in global scalp activity following TMS over DLPFC

represented activity from multiple sources across the brain

(Rogasch, Thomson, et al., 2014). In the current study, we have

demonstrated that global scalp activity is reduced by LICI in a

time dependent manner, suggesting output from the DLPFC is

suppressed alongside local cortical activity by GABAB-medi-

ated inhibition. This finding agrees with studies in the motor

cortex in which both the early and late phases of interhemi-

spheric inhibition between motor cortices are reduced in the

presence of LICI (Lee, Gunraj, & Chen, 2007). In the time

domain, suppression was particularly evident from 100 msec

onwards and evolved from the site of stimulation to contra-

lateral and posterior regions. Suppression of earlier clusters

(~40 and 60 msec) evident in the ROI analysis did not survive

correction for multiple comparisons. This could reflect a

known limitation of cluster based statistics in which large

clusters (i.e., 100e300 msec) bias against smaller clusters

(Maris & Oostenveld, 2007). Alternatively, inter-individual

variation in latency of these early peaks could limit the

sensitivity of this analysis in the time domain. In the fre-

quency domain, inhibition of beta and gamma oscillations at

the site of stimulation were evident within the first 100 msec.

In addition, generation of later alpha (100e200) and earlier

gamma (25e50 msec) oscillations in contralateral and poste-

rior sites were also suppressed. This finding agrees with

another recent study, which also demonstrated LICI of oscil-

lations distant from the stimulated region following both

motor and prefrontal TMS (Garcia Dominguez, Radhu, Farzan,

& Daskalakis, 2014). One proposed role of the DLPFC is to

selectively activate specific cortical networks required during

a task (Benchenane, Tiesinga, & Battaglia, 2011; Fell &

Axmacher, 2011; Zanto, Rubens, Thangavel, & Gazzaley,

2011). GABAB-dependent inhibition may sub-serve this role

by mediating the output of the DLPFC during such events.

4.3. Relationship between DLPFC inhibition and workingmemory

Previous studies have reported that individuals with higher

LICI over DLPFC between 50 and 150 msec demonstrated

enhanced working memory ability on the n-back task

(Daskalakis, Farzan, Barr, Rusjan, et al., 2008) and letter-

number sequencing (Hoppenbrouwers et al., 2013). We have

partially replicated this finding in a third working memory

task, the Sternberg letter recognition task. Importantly, we

observed a temporal delineation within this relationship, with

lower LICI of P60 and higher LICI of N100 correlating with

higher working memory ability. The dichotomous nature of

this finding further supports independent neuralmechanisms

underlying the P60 and N100. As the N100 likely represents a

measure of inhibitory function, this finding also suggests that

adequate suppression of prefrontal inhibitory circuits (i.e.,

LICI of N100) plays an important role in shaping cognitive

performance (Kohl & Paulsen, 2010). Following from this,

disrupted prefrontal inhibitory control such as that observed

in schizophrenia (Farzan et al., 2010b) may also contribute to

cognitive dysfunction.

4.4. Limitations

There are several limitations to the current study. First, we did

not use neuronavigation to target the DLPFC. However, we did

use a site based on the 10/20 EEG system which provides the

closest estimate to this region without using neuronavigation

(Fitzgerald, Maller, Hoy, Farzan, et al., 2009; Fitzgerald, Maller,

Hoy, Thomson, et al., 2009; Rusjan et al., 2010). Second, stim-

ulation parameters were based on the motor cortex, which

does not necessarily translate to the DLPFC (Pell, Roth, &

Zangen, 2011). Currently, there is no consensus on a method

for deciding TMS intensities outside of motor cortex. Motor

cortex parameters were chosen as they produce consistent

LICI in the DLPFC across sessions (Farzan et al., 2010a) and

therefore are adequate for the purpose of this study. Finally,

the TMS-evoked EEG signal was recovered from highly arti-

factual recordings using ICA. Certainty that TMS-evoked

neural activity was not altered by removing these artifacts is

therefore not possible.

5. Conclusions

LICI over DLPFC suppresses both local TMS-evoked activity

and TMS-evoked output to distant cortical regions. At the site

of stimulation, different TEP peaks and different TMS-evoked

oscillations reflect independent mechanisms. The N100 over

DLPFC is consistent with the mechanism responsible for LICI,

most likely GABAB-mediated inhibition. The LICI paradigm

and the N100 are different methods which provide comple-

mentary information on the generation and inhibitory role of

GABAB-mediated potentials. These methods will prove useful

for investigating the role of prefrontal GABAB-mediated inhi-

bition in cognition and pathological conditions such as

schizophrenia.

Disclosures and conflict of interests

PBF has received equipment for research fromMagVenture A/

S, Medtronic Ltd and Brainsway Ltd and funding for research

from Cervel Neurotech. In the last 5 years, ZJD received

external funding through Brainsway Inc and a travel

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 776

allowance through Pfizer and Merck. ZJD has also received

speaker funding through Sepracor Inc, AstraZeneca and

served on the advisory board for Hoffmann-La Roche Limited.

NCR reports no conflicts of interest.

Acknowledgements

The authors wish to thank all volunteers for their participa-

tion. The following work contributed to the doctoral thesis of

NCR. NCR was supported by a postgraduate biomedical

research scholarship (607223) and currently holds a research

fellowship (1072057) from the National Health and Medical

Research Council (NHMRC) of Australia. PBF is supported by

an NHMRC Practitioner Fellowship.

Supplementary data

Supplementary data related to this article can be found at

http://dx.doi.org/10.1016/j.cortex.2014.10.003.

r e f e r e n c e s

Bailey, N. W., Segrave, R. a, Hoy, K. E., Maller, J. J., &Fitzgerald, P. B. (2014). Impaired upper alpha synchronisationduring working memory retention in depression anddepression following traumatic brain injury. BiologicalPsychology, 99C, 115e124. http://dx.doi.org/10.1016/j.biopsycho.2014.03.008.

Barker, A. T., Jalinous, R., & Freeston, I. L. (1985). Non-invasivemagnetic stimulation of human motor cortex. Lancet, 1(8437),1106e1107.

Benchenane, K., Tiesinga, P. H., & Battaglia, F. P. (2011).Oscillations in the prefrontal cortex: a gateway tomemory andattention. Current Opinion in Neurobiology, 21(3), 475e485.http://dx.doi.org/10.1016/j.conb.2011.01.004.

Bender, S., Basseler, K., Sebastian, I., Resch, F., Kammer, T.,Oelkers-Ax, R., et al. (2005). Electroencephalographic responseto transcranial magnetic stimulation in children: evidence forgiant inhibitory potentials. Annals of Neurology, 58(1), 58e67.http://dx.doi.org/10.1002/ana.20521.

Bonnard, M., Spieser, L., Meziane, H. B., de Graaf, J. B., &Pailhous, J. (2009). Prior intention can locally tune inhibitoryprocesses in the primary motor cortex: direct evidence fromcombined TMS-EEG. European Journal of Neuroscience, 30(5),913e923. http://dx.doi.org/10.1111/j.1460-9568.2009.06864.x.

Bruckmann, S., Hauk, D., Roessner, V., Resch, F., Freitag, C. M.,Kammer, T., et al. (2012). Cortical inhibition in attention deficithyperactivity disorder: new insights from theelectroencephalographic response to transcranial magneticstimulation. Brain, 135(Pt 7), 2215e2230. http://dx.doi.org/10.1093/brain/aws071.

Cardin, J. a, Carl�en, M., Meletis, K., Knoblich, U., Zhang, F.,Deisseroth, K., et al. (2009). Driving fast-spiking cells inducesgamma rhythm and controls sensory responses. Nature,459(7247), 663e667. http://dx.doi.org/10.1038/nature08002.

Cash, R. F. H., Ziemann, U., Murray, K., & Thickbroom, G. W.(2010). Late cortical disinhibition in human motor cortex: atriple-pulse transcranial magnetic stimulation study. Journal ofNeurophysiology, 103(1), 511e518. http://dx.doi.org/10.1152/jn.00782.2009.

Daskalakis, Z. J., Farzan, F., Barr, M. S., Maller, J. J., Chen, R., &Fitzgerald, P. B. (2008). Long-interval cortical inhibition fromthe dorsolateral prefrontal cortex: a TMS-EEG study.Neuropsychopharmacology, 33(12), 2860e2869. http://dx.doi.org/10.1038/npp.2008.22.

Daskalakis, Z. J., Farzan, F., Barr, M. S., Rusjan, P. M., Favalli, G.,Levinson, A. J., et al. (2008). Evaluating the relationshipbetween long interval cortical inhibition, working memoryand gamma band activity in the dorsolateral prefrontal cortex.Clinical EEG and Neuroscience, 39(3), 150e155.

Daskalakis, Z. J., & George, T. P. (2009). Clozapine, GABA(B), andthe treatment of resistant schizophrenia. Clinical Pharmacologyand Therapeutics, 86(4), 442e446. http://dx.doi.org/10.1038/clpt.2009.115.

Davies, C. H., Davies, S. N., & Collingridge, G. L. (1990). Paired-pulse depression of monosynaptic GABA-mediated inhibitorypostsynaptic responses in rat hippocampus. Journal ofPhysiology, 424, 513e531.

Deisz, R. A. (1999). GABA(B) receptor-mediated effects in humanand rat neocortical neurones in vitro. Neuropharmacology,38(11), 1755e1766.

Deisz, R. A., & Prince, D. A. (1989). Frequency-dependentdepression of inhibition in guinea-pig neocortex in vitro byGABAB receptor feed-back on GABA release. Journal ofPhysiology, 412, 513e541.

Delorme, A., & Makeig, S. (2004). EEGLAB: an open source toolboxfor analysis of single-trial EEG dynamics includingindependent component analysis. Journal of NeuroscienceMethods, 134(1), 9e21. http://dx.doi.org/10.1016/j.jneumeth.2003.10.009.

Farzan, F., Barr, M. S., Hoppenbrouwers, S. S., Fitzgerald, P. B.,Chen, R., Pascual-Leone, A., et al. (2013). The EEG correlates ofthe TMS-induced EMG silent period in humans. NeuroImage,83C, 120e134. http://dx.doi.org/10.1016/j.neuroimage.2013.06.059.

Farzan, F., Barr, M. S., Levinson, a. J., Chen, R., Wong, W.,Fitzgerald, P. B., et al. (2010a). Reliability of long-intervalcortical inhibition in healthy human subjects: a TMS-EEGstudy. Journal of Neurophysiology, 104(3), 1339e1346. http://dx.doi.org/10.1152/jn.00279.2010.

Farzan, F., Barr, M. S., Levinson, A. J., Chen, R., Wong, W.,Fitzgerald, P. B., et al. (2010b). Evidence for gamma inhibitiondeficits in the dorsolateral prefrontal cortex of patients withschizophrenia. Brain, 133(Pt 5), 1505e1514. http://dx.doi.org/10.1093/brain/awq046.

Farzan, F., Barr, M. S., Wong, W., Chen, R., Fitzgerald, P. B., &Daskalakis, Z. J. (2009). Suppression of gamma-oscillations inthe dorsolateral prefrontal cortex following long intervalcortical inhibition: a TMS-EEG study.Neuropsychopharmacology, 34(6), 1543e1551. http://dx.doi.org/10.1038/npp.2008.211.

Fell, J., & Axmacher, N. (2011). The role of phase synchronizationin memory processes. Nature Reviews Neuroscience, 12(2),105e118. http://dx.doi.org/10.1038/nrn2979.

Fitzgerald, P. B., Daskalakis, Z. J., Hoy, K., Farzan, F., Upton, D. J.,Cooper, N. R., et al. (2008). Cortical inhibition in motor andnon-motor regions: a combined TMS-EEG study. Clinical EEGand Neuroscience, 39(3), 112e117.

Fitzgerald, P. B., Maller, J. J., Hoy, K., Farzan, F., & Daskalakis, Z. J.(2009). GABA and cortical inhibition in motor and non-motorregions using combined TMS-EEG: a time analysis. ClinicalNeurophysiology, 120(9), 1706e1710. http://dx.doi.org/10.1016/j.clinph.2009.06.019.

Fitzgerald, P. B., Maller, J. J., Hoy, K. E., Thomson, R., &Daskalakis, Z. J. (2009). Exploring the optimal site for thelocalization of dorsolateral prefrontal cortex in brainstimulation experiments. Brain Stimulation, 2(4), 234e237.http://dx.doi.org/10.1016/j.brs.2009.03.002.

c o r t e x 6 4 ( 2 0 1 5 ) 6 8e7 7 77

Garcia Dominguez, L., Radhu, N., Farzan, F., & Daskalakis, Z. J.(2014). Characterizing long interval cortical inhibition over thetime-frequency domain. PloS One, 9(3), e92354. http://dx.doi.org/10.1371/journal.pone.0092354.

Hoppenbrouwers, S. S., De Jesus, D. R., Stirpe, T., Fitzgerald, P. B.,Voineskos, A. N., Schutter, D. J. L. G., et al. (2013). Inhibitorydeficits in the dorsolateral prefrontal cortex in psychopathicoffenders. Cortex, 49(5), 1377e1385. http://dx.doi.org/10.1016/j.cortex.2012.06.003.

Huber, R., M€aki, H., Rosanova, M., Casarotto, S., Canali, P.,Casali, A. G., et al. (2013). Human cortical excitability increaseswith time awake. Cerebral Cortex, 23(2), 332e338. http://dx.doi.org/10.1093/cercor/bhs014.

Isaacson, J. S., & Scanziani, M. (2011). How inhibition shapescortical activity. Neuron, 72(2), 231e243. http://dx.doi.org/10.1016/j.neuron.2011.09.027.

Kici�c, D., Lioumis, P., Ilmoniemi, R. J., & Nikulin, V. V. (2008).Bilateral changes in excitability of sensorimotor corticesduring unilateral movement: combinedelectroencephalographic and transcranial magneticstimulation study. Neuroscience, 152(4), 1119e1129. http://dx.doi.org/10.1016/j.neuroscience.2008.01.043.

Kohl, M. M., & Paulsen, O. (2010). The roles of GABABreceptors in cortical network activity. Advances inPharmacology, 58(10), 205e229. http://dx.doi.org/10.1016/S1054-3589(10)58009-8.

Komssi, S., K€ahk€onen, S., & Ilmoniemi, R. J. (2004). The effect ofstimulus intensity on brain responses evoked by transcranialmagnetic stimulation. Human Brain Mapping, 21(3), 154e164.http://dx.doi.org/10.1002/hbm.10159.

Krnjevi�c, K. (1997). Role of GABA in cerebral cortex. CanadianJournal of Physiology and Pharmacology, 75(5), 439e451.

Lee, H., Gunraj, C., & Chen, R. (2007). The effects of inhibitory andfacilitatory intracortical circuits on interhemisphericinhibition in the human motor cortex. Journal of Physiology,580(Pt.3), 1021e1032. http://dx.doi.org/10.1113/jphysiol.2006.126011.

Maris, E., & Oostenveld, R. (2007). Nonparametric statisticaltesting of EEG- and MEG-data. Journal of Neuroscience Methods,164(1), 177e190. http://dx.doi.org/10.1016/j.jneumeth.2007.03.024.

Nakamura, H., Kitagawa, H., Kawaguchi, Y., & Tsuji, H. (1997).Intracortical facilitation and inhibition after transcranialmagnetic stimulation in conscious humans. Journal ofPhysiology, 498(Pt 3), 817e823.

Nikulin, V. V., Kicic, D., Kahkonen, S., & Ilmoniemi, R. J. (2003).Modulation of electroencephalographic responses totranscranial magnetic stimulation: evidence for changes incortical excitability related to movement. European Journal ofNeuroscience, 18(5), 1206e1212. http://dx.doi.org/10.1046/j.1460-9568.2003.02858.x.

Oostenveld, R., Fries, P., Maris, E., & Schoffelen, J.-M. (2011).FieldTrip: open source software for advanced analysis of MEG,EEG, and invasive electrophysiological data. ComputationalIntelligence and Neuroscience, 2011, 156869. http://dx.doi.org/10.1155/2011/156869.

Pell, G. S., Roth, Y., & Zangen, A. (2011). Modulation of corticalexcitability induced by repetitive transcranial magneticstimulation: influence of timing and geometrical parametersand underlying mechanisms. Progress in Neurobiology, 93(1),59e98. http://dx.doi.org/10.1016/j.pneurobio.2010.10.003.

Premoli, I., Castellanos, N., Rivolta, D., Belardinelli, P., Bajo, R.,Zipser, C., et al. (2014). TMS-eeg signatures of GABAergicneurotransmission in the human cortex. Journal ofNeuroscience, 34(16), 5603e5612. http://dx.doi.org/10.1523/JNEUROSCI.5089-13.2014.

Rogasch, N. C., Daskalakis, Z. J., & Fitzgerald, P. B. (2013).Mechanisms underlying long-interval cortical inhibition in thehuman motor cortex: a TMS-EEG study. Journal ofNeurophysiology, 109(1), 89e98. http://dx.doi.org/10.1152/jn.00762.2012.

Rogasch, N. C., Daskalakis, Z. J., & Fitzgerald, P. B. (2014). Corticalinhibition, excitation, and connectivity in schizophrenia: areview of insights from transcranial magnetic stimulation.Schizophrenia Bulletin, 40(3), 685e696. http://dx.doi.org/10.1093/schbul/sbt078.

Rogasch, N. C., & Fitzgerald, P. B. (2013). Assessing corticalnetwork properties using TMS-EEG. Human Brain Mapping,34(7), 1652e1669. http://dx.doi.org/10.1002/hbm.22016.

Rogasch, N. C., Thomson, R. H., Daskalakis, Z. J., & Fitzgerald, P. B.(2013). Short-latency artifacts associated with concurrentTMS-EEG. Brain Stimulation, 6(6), 868e876. http://dx.doi.org/10.1016/j.brs.2013.04.004.

Rogasch, N. C., Thomson, R. H., Farzan, F., Fitzgibbon, B. M.,Bailey, N. W., Hernandez-Pavon, J. C., et al. (2014). Removingartefacts from TMS-EEG recordings using independentcomponent analysis: Importance for assessing prefrontal andmotor cortex network properties. NeuroImage, 101, 425e439.http://dx.doi.org/10.1016/j.neuroimage.2014.07.037.

Rusjan, P. M., Barr, M. S., Farzan, F., Arenovich, T., Maller, J. J.,Fitzgerald, P. B., et al. (2010). Optimal transcranial magneticstimulation coil placement for targeting the dorsolateralprefrontal cortex using novel magnetic resonance image-guided neuronavigation. Human Brain Mapping, 31(11),1643e1652. http://dx.doi.org/10.1002/hbm.20964.

Sanger, T. D., Garg, R. R., & Chen, R. (2001). Interactions betweentwo different inhibitory systems in the human motor cortex.Journal of Physiology, 530(Pt 2), 307e317.

Schuler, V., Luscher, C., Blanchet, C., Klix, N., Sansig, G., Klebs, K.,et al. (2001). Epilepsy, hyperalgesia, impaired memory, andloss of pre- and postsynaptic GABA(B) responses in micelacking GABA(B(1)). Neuron, 31(1), 47e58.

Sohal, V. S., Zhang, F., Yizhar, O., & Deisseroth, K. (2009).Parvalbumin neurons and gamma rhythms enhance corticalcircuit performance. Nature, 459(7247), 698e702. http://dx.doi.org/10.1038/nature07991.

Spieser, L., Meziane, H. B., & Bonnard, M. (2010). Corticalmechanisms underlying stretch reflex adaptation to intention:a combined EEG-TMS study. NeuroImage, 52(1), 316e325. http://dx.doi.org/10.1016/j.neuroimage.2010.04.020.

Valls-Sol�e, J., Pascual-Leone, a, Wassermann, E. M., & Hallett, M.(1992). Human motor evoked responses to paired transcranialmagnetic stimuli. Electroencephalography and ClinicalNeurophysiology, 85(6), 355e364.

Whittington, M. A., Traub, R. D., & Jefferys, J. G. (1995).Synchronized oscillations in interneuron networks driven bymetabotropic glutamate receptor activation. Nature, 373(6515),612e615. http://dx.doi.org/10.1038/373612a0.

Zanto, T. P., Rubens, M. T., Thangavel, A., & Gazzaley, A. (2011).Causal role of the prefrontal cortex in top-downmodulation ofvisual processing and working memory. Nature Neuroscience,14(5), 656e661. http://dx.doi.org/10.1038/nn.2773.