Journal of Neurophysiology

Reliability of Long-Interval Cortical Inhibition in Healthy Human Subjects: A TMS–EEG Study

Faranak Farzan, Mera S. Barr, Andrea J. Levinson, Robert Chen, Willy Wong, Paul B. Fitzgerald, Zafiris J. Daskalakis

Abstract

Cortical inhibition (CI) is measured by transcranial magnetic stimulation (TMS) combined with electromyography (EMG) through long-interval CI (LICI) and cortical silent period (CSP) paradigms. Recently, we illustrated that LICI can be measured from the dorsolateral prefrontal cortex (DLPFC) through combined TMS with electroencephalography (EEG). We further demonstrated that LICI had different effects on cortical oscillations in the DLPFC compared with motor cortex. The purpose of this study was to establish the validity and reliability of TMS–EEG indices of CI and to replicate our previous findings in an extended sample. The validity of TMS–EEG was examined by evaluating its relationship to standard EMG measures of LICI and the CSP in the left motor cortex in 36 and 16 subjects, respectively. Test–retest reliability was examined in 14 subjects who returned for a repeat session within 7 days of the first session. LICI was applied to the left DLPFC in 30 subjects to compare LICI in the DLPFC with that in the motor cortex. In the motor cortex, EEG measures of LICI correlated with EMG measures of LICI and CSP. All indices of LICI showed high test–retest reliability in motor cortex and DLPFC. Gamma and beta oscillations were significantly inhibited in the DLPFC but not in the motor cortex, confirming previous findings in an extended sample. These findings demonstrate that indexing LICI through TMS combined with EEG is a valid and reliable method to evaluate inhibition from motor and prefrontal regions.

INTRODUCTION

Cortical Inhibition (CI) refers to the neurophysiological process whereby cortical γ-aminobutyric acid (GABA) inhibitory interneurons suppress the activity of other neurons in the cortex. Traditionally, CI can be measured through the application of paired-pulse transcranial magnetic stimulation (TMS) to the motor cortex and measured in the target muscle as a suppression of motor-evoked potentials (MEPs) recorded by electromyography (EMG). Long-interval cortical inhibition (LICI) is a paired-pulse TMS paradigm used to measure CI (Valls-Sole et al. 1992). In LICI, the application of a conditioning stimulus within an interval of 50–200 ms prior to a test stimulus suppresses the amplitude of the MEP evoked by the test stimulus by about 50% (Daskalakis et al. 2002), compared with test stimulus delivered alone. Previous reports have associated LICI with GABAB receptor-mediated inhibitory neurotransmission. For example, administration of baclofen, a GABAB receptor agonist, was shown to enhance LICI (Florian et al. 2008; McDonnell et al. 2006). The cortical silent period (CSP) represents another TMS paradigm that involves a transient suppression of EMG activity following TMS to the motor cortex during voluntary contraction of the target muscle. The CSP is also associated with GABAB receptor-mediated inhibitory neurotransmission (Chen et al. 1999; Nakamura et al. 1997; Siebner et al. 1998; Werhahn et al. 1999). For example, Siebner et al. (1998) reported a significant prolongation of the CSP duration following a continuous intrathecal administration of baclofen in a patient with generalized dystonia. Also, tiagabine, a GABA uptake inhibitor, was shown to enhance LICI and prolong CSP duration (Werhahn et al. 1999). Furthermore, administration of vigabatrin, a selective GABAergic drug that increases the availability of GABA in the brain, has also been shown to enhance LICI and CSP (Pierantozzi et al. 2004).

Recording CI (e.g., CSP and LICI) through TMS–EMG, however, is limited to the motor cortex. Measuring CI from other cortical regions such as the prefrontal cortex may provide an important neurophysiological index of GABAB receptor inhibition and provide more insights into pathophysiology of disorders closely associated with prefrontal deficits such as schizophrenia. In this regard, we recently reported methods by which LICI can be measured from non-motor cortical regions through a combination of TMS and electroencephalography (EEG) (Daskalakis et al. 2008; Farzan et al. 2009). Three major findings were reported. First, EMG measures of LICI strongly correlated with the EEG measures of LICI (Daskalakis et al. 2008). Second, in nine healthy subjects, LICI was produced in the dorsolateral prefrontal cortex (DLPFC), a region strongly associated with the pathophysiology of schizophrenia (Daskalakis et al. 2008) and other neuropsychiatric disorders. Third, we studied the modulation of cortical frequency bands including delta (1–3.5 Hz), theta (4–7 Hz), alpha (8–12 Hz), beta (12.5–28 Hz), and gamma (30–50 Hz) oscillations in response to application of LICI in the DLPFC and the motor cortex. Cortical gamma oscillations, which are associated with higher-order cognitive processes (e.g., working memory) (Cho et al. 2006), were inhibited in response to LICI applied to the DLPFC but not to the motor cortex (Farzan et al. 2009). Combined TMS–EEG thus permits recording of LICI from non-motor cortical regions at selected cortical frequencies.

However, there are some limitations in these studies. The first limitation is the validity of TMS–EEG as a neurophysiological tool for measuring GABAB receptor inhibition. In our previous report, we evaluated the relationship between EEG measures of LICI and EMG measures of LICI. To further validate TMS–EEG measures of CI, its relationship to other measures of CI such as the CSP duration should be established. In addition, indexing the test–retest reliability of such methods is also necessary. High test–retest reliability and strong validity fulfill two criteria toward the development of this neurophysiological measure as a candidate endophenotype for various neurological and psychiatric disorders (Braff et al. 2008; Farzan et al. 2010). Finally, an important limitation is a relatively small sample size. Replicating our previous findings in a larger subject sample should minimize type I error, stabilize statistical parameter estimates (Norman and Streiner 2000), and validate this measure as an index of CI directly from the cortex.

This study thus has three separate experimental objectives. The first objective was to establish the validity of the TMS–EEG measure of LICI as a neurophysiological index of GABAB receptor inhibition by comparing EEG measures of LICI to more conventional EMG measures of LICI and CSP in the motor cortex. The second objective was to examine the test–retest reliability of EEG measures of LICI in both motor cortex and DLPFC. The third objective was to replicate our previous findings comparing the inhibitory effect of LICI on all five cortical oscillations—including delta (1–3.5 Hz), theta (4–7 Hz), alpha (8–12 Hz), beta (12.5–28 Hz), and gamma (30–50 Hz) oscillations—in an expanded subject sample because our previous studies demonstrated that higher frequency oscillations (i.e., beta and gamma bands) were significantly inhibited in the DLPFC compared with the motor cortex, where lower frequency oscillations (i.e., delta, theta, and alpha) were predominantly inhibited.

METHODS

Subjects

We studied 36 right-handed healthy subjects (mean age = 33.7 yr, SD = 7.7 yr, range = 20–48 yr; 20 males, 16 females), data from 15 of whom (mean age = 34.7 yr, SD = 8.1 yr, range = 23–47 yr; 5 males, 10 females) had been included in our previous reports (Daskalakis et al. 2008; Farzan et al. 2009). Subjects were recruited through advertisement and psychopathology was ruled out through the personality assessment screener (Psychological Assessment Resources). Exclusion criteria also included a self-reported medical illness or a history of drug or alcohol abuse. In all subjects, handedness was confirmed using the Oldfield Handedness Inventory (Oldfield 1971). All participants gave their written informed consent and the protocol was approved by the local ethics committee at the Centre for Addiction and Mental Health in accordance with the declaration of Helsinki.

Study design

Three experiments were conducted.

Study 1: validity of TMS–EEG.

In our previous report (Daskalakis et al. 2008) 15 healthy subjects had undergone LICI in the motor cortex. In this study, 21 new subjects underwent LICI to the motor cortex and, as such, a total of 36 subjects received LICI to the left motor cortex. A subset of 16 subjects (mean age = 31.4 yr, SD = 6.4 yr, range = 20–40 yr; 11 males, 5 females) also received the CSP paradigm. CSP was administered in the same session as LICI recording.

Study 2: test–retest reliability of EEG and EMG indices of LICI.

In this study, a subset of 14 subjects (mean age = 30.8 yr, SD = 6.7 yr, range = 20–40 yr; 10 males, 4 females), selected randomly, returned within 7 days of the original testing session for a second session. In the second session, all conditions administered in the first testing session were applied in the same order.

Study 3: effect of LICI in the motor cortex and DLPFC.

In this experiment, we evaluated the modulatory effect of LICI across cortical oscillations in an extended sample size of 36 healthy subjects in the motor cortex and in a subset of 30 subjects in the DLPFC.

Data recording

Transcranial magnetic stimulation.

Monophasic TMS pulses were administered using a 7 cm figure-of-eight coil and two Magstim 200 stimulators (Magstim, Royston, UK), connected via a Bistim module. At the beginning of each experiment, resting motor threshold was determined by applying single pulses of TMS to the motor cortex while the coil was placed at the optimal position for eliciting MEPs from the right abductor pollicis brevis (APB) muscle. Resting motor threshold was defined as the minimum stimulus intensity that elicited an MEP of >50 μV in ≥5 of 10 trials (Rossini et al. 1994). We determined resting motor threshold once prior to positioning the EEG cap on the head and once after. This was done to compare stimulus intensity with other studies that recorded LICI through EMG only. This intensity corresponded to an average of 42.4 ± 7.7% of maximum stimulator output without the EEG cap and to 54.1 ± 10.6% with the EEG cap across 36 subjects. Furthermore, for each subject, we determined the stimulus intensity that elicited mean peak-to-peak MEP amplitude of 1 mV in 20 trials. In LICI, both test stimulus and conditioning stimulus were administered at this intensity in both motor cortex and DLPFC. This intensity corresponded to 67.7 ± 13.3% of maximum stimulator output across 36 subjects and to 66.0 ± 12.8% in a subset of 30 subjects who also received LICI to the DLPFC. In both regions, the optimal position was marked on the EEG cap with a felt pen, to ensure identical placement of the coil throughout the experiment, and the handle of the coil pointed backward, perpendicularly to the presumed direction of the central sulcus, about 45° to the midsagittal line.

Sham stimulation.

To control for the effect of TMS click-induced auditory activation on the cortical evoked potentials and single- and paired-pulse sham stimulations were administered at the same intensity as that used for active stimulation, but with the coil angled at 90° from the scalp resting on one wing of the coil. Throughout this study, the term “cortical evoked potential” is conventionally used to refer to the cortical activity recorded following TMS application. Sham stimulation was administered in a subset of 26 subjects (mean age = 33.7 yr, SD = 7.1 yr, range = 20–47 yr; 18 males, 8 females) in both the left motor cortex and DLPFC.

Electromyography.

Throughout the experiments, subjects were seated in a comfortable armchair with their hands positioned on a pillow placed over their laps and they were instructed to maintain relaxation as EMG was monitored on a computer screen. Two disposable surface disc electrodes were placed in a tendon-belly arrangement over the right APB muscle, a ground electrode was placed over the right elbow, and EMG activity was acquired through Signal software (Cambridge Electronics Design [CED], Cambridge, UK). The EMG signals were amplified (Model 2024F; Intronix Technologies, Bolton, Ontario, Canada), filtered (band-pass 2 Hz to 5 kHz), digitized at 5 kHz (Micro 1401, CED), and stored in a laboratory computer for off-line analysis.

Electroencephalography.

Cortical evoked potentials were acquired through a 64-channel Synamps2 EEG system. A 64-channel EEG cap was positioned on each subject's head and for all electrodes impedance was lowered to ≤5 kΩ. Four additional electrodes were placed on the outer side of each eye and above and below the left eye to monitor eye movement artifacts. All electrodes were referenced to an electrode placed posterior to the Cz electrode. EEG signals were recorded with filters at DC to 200 Hz at a 20 kHz sampling rate, which was shown to avoid saturation of amplifiers and minimize the TMS-related artifact (Daskalakis et al. 2008). Throughout the experiments, EEG and EMG were recorded simultaneously.

Long-interval cortical inhibition.

The LICI paradigm involves the pairing of a suprathreshold conditioning stimulus followed by a suprathreshold test stimulus at long interstimulus intervals (ISIs; e.g., 100 ms), which inhibits the MEP produced by test stimulus (Valls-Sole et al. 1992). In LICI, activation of GABAB receptors by the conditioning stimulus inhibits the excitation of the cortex to a second pulse (test stimulus) that would otherwise excite the cortex if delivered alone. It is suggested that the GABAB receptor activation peaks around 150 to 200 ms poststimulus (McCormick 1989) and, as such, LICI is optimal when conditioning stimulus precedes the test stimulus by about 100–150 ms (Sanger et al. 2001). An ISI of 100 ms was also used in our original report (Daskalakis et al. 2008) and, as such, in this experiment we evaluated LICI at this interval (i.e., LICI100). Both conditioning stimulus and test stimulus were suprathreshold and adjusted to produce a mean peak-to-peak MEP amplitude of 1 mV (Valls-Sole et al. 1992). One hundred TMS stimuli were delivered per condition (i.e., paired conditioning stimulus–test stimulus and test stimulus alone) every 5 s. This interval is commonly used in TMS studies evaluating CI because it does not result in habituation with repeated stimulation (Kujirai et al. 1993; Nakamura et al. 1997; Sanger et al. 2001; Ziemann et al. 1996a). All conditions (i.e., active/sham, motor/DLPFC) were randomized to avoid order effects.

Localization of DLPFC.

DLPFC was localized as previously described (Daskalakis et al. 2008; Farzan et al. 2009). In short, a T1-weighted MRI scan was obtained for each subject with seven fiducial markers in place. Through MATLAB software (The MathWorks, Natick, MA), DLPFC was marked as a bright spot on each subject's MRI at the junction of the middle and anterior one third of the middle frontal gyrus [Talairach coordinates (x, y, z) = (50, 30, 36)] corresponding with posterior regions of Brodmann area 9 (BA9), which overlaps with the superior section of BA46. Prior to the experiment, the DLPFC was marked on each subject's scalp through neuronavigation techniques using the MRIco/reg and MINIBIRD system (Ascension Technologies, Burlington, VT).

Cortical silent period.

The CSP was recorded during voluntary contraction of the right APB muscle by stimulation of the left motor cortex at 140% of resting motor threshold. This corresponded to 76.1 ± 15.4% of stimulator output in a subset of 16 subjects who participated in this experiment. Fifty stimuli were delivered every 5 s, whereas the APB muscle was contracted at 20% of maximum muscle contraction measured by a strain gauge meter.

Data analysis

EEG and EMG measures of LICI were obtained in a manner similar to that in our previous reports (Daskalakis et al. 2008; Farzan et al. 2009) as follows.

EMG measures of inhibition.

For each subject, the EMG measures of LICI100 were indexed by comparing the area under the curve of the rectified averaged MEP following the single pulse of TMS (unconditioned) with the area under curve following the paired pulses of TMS (conditioned) and the inhibition was indexed as follows [1Areaunderrectifiedcurve(conditioned)Areaunderrectifiedcurve(unconditioned)]×100 (1)

EEG measures of inhibition.

The EEG recordings were first processed off-line by commercially available software (Neuroscan; Compumedics, Charlotte, NC). The EEG data were down-sampled to 1 kHz sampling frequency and segmented with respect to the TMS test stimulus such that each epoch included 1,000 ms prestimulus baseline and a 1,000 ms poststimulus activity. Epochs were baseline corrected with respect to the TMS-free prestimulus interval (1,000 to 110 ms prior to the test stimulus). Baseline correction is a process of estimating an average value for the DC offset in the prestimulus interval that is then subtracted from the poststimulus interval. The baseline corrected post-test stimulus intervals (25–1,000 ms), which were not contaminated by TMS artifact, were extracted and digitally filtered by using a zero phase shift 1 to 100 Hz band-pass filter (48 dB/octave). At this stage, epochs were manually reviewed and trials contaminated with muscle activity, movement, and TMS artifacts (6.0 ± 0.7% of trials in each condition) were excluded from further analysis. Finally, the averaged signals at each recording site were computed from the movement-free epochs (∼94 trials per condition) and were fed into an automated eyeblink correction algorithm (Croft et al. 2005). The eyeblink corrected averaged EEG waveforms were then imported into MATLAB and further analyses were carried out by means of EEGLAB toolbox (Delorme and Makeig 2004). To obtain the total amount of EEG inhibition, for each subject the TMS-evoked cortical potentials following the single and paired pulse of TMS were band-pass filtered (1–50 Hz) and CItotal was calculated through Eq. 1. To quantify the EEG measures of LICI100 in all five frequency bands, for each subject, the average TMS-evoked cortical potentials following single and paired pulses of TMS were decomposed into the delta (δ: 1–3.5 Hz), theta (θ: 4–7 Hz), alpha (α: 8–12 Hz), beta (β: 12.5–28 Hz), and gamma (γ: 30–50 Hz) frequency components by means of a Hamming-based zero phase-shift finite impulse response filter and, for each frequency band, inhibition was obtained through Eq. 1 (i.e., area under rectified curve for averaged EEG recordings between 50 and 150 ms post-test stimulus). The first interval (i.e., 50 ms poststimulus) was chosen because it represents the earliest artifact-free data that can be recorded poststimulus; the second interval (i.e., 150 ms poststimulus) was chosen because it represents the duration of GABAB receptor-mediated inhibitory postsynaptic potentials (IPSPs) (Deisz 1999) (i.e., 250 ms) elicited by the conditioning stimulus (Deisz 1999). We referred to these measures as CIδ, CIθ, CIα, CIβ, and CIγ to represent the extent of inhibition in each frequency band (Fig. 1). Finally, to evaluate LICI directly from the motor cortex, the C3 electrode was used because it has been shown to be the electrode that best represents evoked activity in the hand area of motor cortex and is closest to the optimal site of APB activation through TMS (Cui et al. 1999). To capture LICI in the DLPFC, the recording electrode of interest was AF3, which optimally represents the overlap of BA9 and BA46 of the DLPFC (Fig. 2).

Fig. 1.

Topographic illustration of modulation of cortical oscillations following application of long-interval cortical inhibition (LICI) to the left motor cortex and dorsolateral prefrontal cortex (DLPFC). Each topographic plot illustrates the average suppression of cortical activity in response to application of LICI paradigm to the left motor cortex (left) and left DLPFC (right) across 30 subjects who received transcranial magnetic stimulation (TMS) in both cortical regions. Inhibition is obtained by Eq. 1 (see methods) and hot colors represent the area of maximum inhibition. These plots suggest that application of LICI to the motor cortex results in attenuation of delta, theta, and alpha oscillations, whereas in the DLPFC, beta and gamma oscillations are also inhibited. Topographic head plots were obtained by EEGLAB toolbox (Delorme and Makeig 2004).

Fig. 2.

TMS evoked cortical oscillations following the application of LICI to the motor cortex and DLPFC. The waveforms represents mean rectified cortical oscillations in delta (1–3.5 Hz), theta (4–7 Hz), alpha (8–12 Hz), beta (12.5–28 Hz), and gamma (30–50 Hz) frequency bands, from top to bottom, respectively. The waveforms are recorded following the delivery of single-pulse (solid waveforms) and paired-pulse (dash waveforms) TMS to the left motor cortex (left) and the left DLPFC (right). The data are averaged across 30 subjects who received TMS in both cortical regions. In all figures, the x-axis represents the time, 50 to 150 ms, after the delivery of test stimulus. In DLPFC, the y-axis represents the electroencephalographic (EEG) potential recorded from the AF3 electrode, which optimally represents the overlap of Brodmann area 9 (BA9) and BA46 of the DLPFC. In motor cortex, the y-axis represents the EEG potential recorded from the C3 electrode, which was closest to the optimal site of abductor pollicis brevis (APB) muscle activation following application of TMS.

Cortical silent period.

To measure mean CSP duration, trials were averaged and rectified. The CSP duration was then measured from the onset of MEP to the reoccurrence of any background EMG activity.

Statistics.

We used two-tailed paired t-tests to compare the mean area under the rectified curve in single pulse with that following the paired pulse. For each variable, descriptive values are reported as mean ± SD and values of P < 0.05 were considered significant. The Pearson correlation coefficient was used to investigate the relationship between 1) EEG and EMG measures of LICI, 2) EEG measures of LICI in the motor cortex and LICI in the DLPFC, 3) EMG measures of LICI and CSP, and 4) EEG measures of LICI in the motor cortex and CSP. We used the methods suggested by Meng et al. (1992) to compare the size of correlation between EMG measures of LICI and CSP with the correlation between EEG measures of LICI and CSP. The method proposed by Meng and colleagues is based on the Fisher z transformation and provides a test and confidence interval for comparing two correlated correlations (Meng et al. 1992). Furthermore, Cronbach's alpha was used to assess the test–retest reliability of 1) EMG measures of LICI, 2) EEG measures of LICI in the motor cortex, 3) LICI in the DLPFC, and 4) LICI across cortical oscillations in both motor cortex and DLPFC. Finally, repeated-measures ANOVA was used to examine the inhibition across the five cortical oscillations and two cortical regions. For these analyses, post hoc multiple comparisons were performed when applicable and significance levels were Bonferroni adjusted. Most statistical analyses were performed in SPSS 15.0 (SPSS, Chicago, IL).

RESULTS

Study 1: validity of TMS–EEG

Study 1.1: the relationship between EEG and EMG measures of LICI.

Similar to our previous reports, we found a significant suppression of the mean area under the rectified EMG curve (74.5 ± 20.9%) following LICI100 compared with the test stimulus delivered alone (t = 7.3, degrees of freedom [df] = 35, P < 0.001). We also found a significant suppression (CItotal = 39.0 ± 31.2%) of the mean area under the rectified cortical evoked potentials following LICI100 compared with the test stimulus delivered alone (t = 5.8, df = 35, P < 0.001). Furthermore, consistent with our first study, there was a significant correlation between EEG and EMG measures of LICI (n = 36, r = 0.85, P < 0.001; Fig. 3).

Fig. 3.

EEG measures of LICI correlate with electromyographic (EMG) measures of LICI. Data obtained from 36 healthy subjects. The x-axis represents the EMG measures of LICI100 measured from the right APB muscle, and y-axis represents EEG measures of LICI100 recorded from the C3 electrode.

Study 1.2: the relationship between EEG measures of LICI in the motor cortex and CSP.

The average duration of CSP was 140.3 ± 38.0 ms. In subjects that underwent CSP, the average EMG and EEG measures of LICI were 74.0 ± 20.5 and 34.4 ± 30.1%, respectively. We found a significant correlation between EMG measures of LICI and duration of CSP (r = 0.61, P = 0.01; Fig. 4A). We also found a strong correlation between EEG measures of LICI and CSP (r = 0.8, P < 0.001; Fig. 4B). Finally, using the methods suggested by Meng et al. (1992) we found that the correlation between EEG measures of LICI and CSP duration was slightly stronger (trending toward significance) than the correlation between EMG measures of LICI and CSP duration (P = 0.09; one-tailed).

Fig. 4.

Relationship between LICI and cortical silent period (CSP), measures of γ-aminobutyric acid type B (GABAB) receptor-mediated inhibition. Data obtained from 16 subjects who received both LICI and CSP to the left motor cortex. In both figures, x-axes represent the duration of CSP. A: the y-axis represents the EMG measures of LICI as measured from the right APB muscle. There is a significant correlation between duration of CSP and EMG measures of LICI (r = 0.61, P = 0.01). B: the y-axis represents the EEG measures of LICI recorded from the C3 electrode, which is closest to the stimulation site. There is a significant correlation between EEG measures of LICI and duration of CSP (r = 0.80, P < 0.001).

Study 2: test–retest reliability of EEG and EMG indices of LICI

Study 2.1: test–retest reliability in the motor cortex.

The reliability analysis of EMG and EEG measures of LICI revealed Cronbach's alpha values of 0.88 and 0.93, respectively. Furthermore, the test–retest reliability revealed a Cronbach's alpha of 0.92 for CIδ, 0.85 for CIθ, 0.86 for CIα, 0.66 for CIβ, and 0.76 for CIγ (Fig. 5A). In EEG studies (Neuper et al. 2005), a Cronbach's alpha of ≥0.7 indicates high reproducibility and consistency. Therefore our findings show high test–retest reliability across all indices of CI in the motor cortex, except for CIβ, which has moderate test–retest reliability.

Fig. 5.

Test–retest reliability of measures of cortical inhibition (CI) following application of LICI to the left motor cortex and DLPFC. Data obtained from 14 subjects in whom LICI100 was applied to both the motor cortex and DLPFC on 2 sessions separated by ≤7 days. Histograms illustrate the between sessions consistency for 7 measures of CI in the motor cortex (i.e., CIδ, CIθ, CIα, CIβ, CIγ, CItotal, and EMG measures of LICI100), and for 6 measures of CI in the DLPFC (i.e., CIδ, CIθ, CIα, CIβ, CIγ, CItotal). The y-axes represent Cronbach's alpha, a measure of reliability and internal consistency. In the motor cortex, CIδ, CIθ, CIα, CIγ, CItotal, and EMG measures of LICI had Cronbach's alpha >0.70 and CIβ had Cronbach's alpha of 0.66. In the DLPFC, all indices of CI had Cronbach's alpha >0.70.

Study 2.2: test–retest reliability in the DLPFC.

The reliability analysis for CItotal revealed a Cronbach's alpha of 0.97. Furthermore, the test–retest reliability revealed a Cronbach's alpha of 0.88 for CIδ, 0.93 for CIθ, 0.95 for CIα, 0.71 for CIβ, and 0.92 for CIγ (Fig. 5B). These data suggest that all indices of CI have high test–retest reliability and internal consistency in the DLPFC.

Study 3: the effect of LICI in the motor cortex and DLPFC

Study 3.1: LICI in the motor cortex.

Repeated-measures ANOVA was performed to examine the effect of LICI across five cortical oscillations (CIδ, CIθ, CIα, CIβ, and CIγ), with frequency as the main effect. This analysis revealed a significant effect of frequency (F = 7.3, df = 4, P < 0.001). We performed post hoc analyses and found a significant inhibition across cortical oscillations of delta (CIδ = 37.3 ± 51.8%; t = 4.8, df = 35, P < 0.001), theta (CIθ = 46.1 ± 40.6%; t = 6.0, df = 35, P < 0.001), and alpha oscillations (CIα = 26.8 ± 36.7%; t = 4.3, df = 35, P < 0.001), whereas beta (CIβ = 6.5 ± 45.4%; t = 1.7, df = 35, P = 0.1) and gamma oscillations (CIγ = 8.1 ± 45.1%; t = 1.5, df = 35, P = 0.1) were not inhibited (Fig. 6). This finding is also consistent with the results of our previous report (Farzan et al. 2009).

Fig. 6.

Differential modulatory effect of LICI applied to the left motor cortex and left DLPFC. Data obtained from 36 subjects in the motor cortex and 30 subjects in the DLPFC. Histograms represent the average inhibition of 5 cortical oscillations in response to the application of LICI100 to the left motor cortex (left) and left DLPFC (right). Error bars indicate SE. In the motor cortex (left), application of LICI results in significant suppression of delta, theta, and alpha oscillations only. In the DLPFC, LICI results in inhibition of all cortical oscillations.

Study 3.2: LICI in the DLPFC.

We found a significant suppression (CItotal = 31.9 ± 49.5%) of the mean area under the rectified cortical evoked potentials following LICI100 compared with test stimulus alone (t = 5.1, df = 29, P < 0.001). Repeated-measures ANOVA was performed to examine the effect of LICI across five cortical oscillations (CIδ, CIθ, CIα, CIβ, and CIγ), with frequency as the main effect. Consistent with our previous report (Farzan et al. 2009), this analysis revealed no significant effect of frequency (F = 1.3, df = 4, P = 0.3). Post hoc analysis confirmed that both low-frequency (CIδ = 29.5 ± 41.7%, t = 4.6, df = 29, P < 0.001; CIθ = 25.9 ± 58.6%, t = 4.6, df = 29, P < 0.001; CIα = 33.0 ± 36.7%, t = 5.8, df = 29, P < 0.001) and high-frequency oscillations (CIβ = 37.6 ± 21.1%, t = 6.3, df = 29, P < 0.001; CIγ = 40.0 ± 21.2%, t = 6.0, df = 29, P < 0.001) were inhibited following LICI100 compared with the test stimulus delivered alone (Fig. 6).

Study 3.3: comparison between motor cortex and DLPFC.

A repeated-measures ANOVA revealed a significant region (i.e., motor cortex and DLPFC) by frequency (i.e., CIδ, CIθ, CIα, CIβ, and CIγ) interaction effect (F = 8.6, df = 4, P < 0.001). Post hoc analysis revealed that this interaction was due to a more pronounced suppression of higher frequency oscillations (CIβ, CIγ) in the DLPFC compared with motor cortex (Figs. 1, 2, and 6). That is CIβ and CIγ were significantly higher in the DLPFC than in the motor cortex (P < 0.001 and P = 0.001, respectively). This finding is consistent with our previous report (Farzan et al. 2009). However, we did not find a correlation between CItotal in the motor cortex and CItotal in the DLPFC in this extended sample (n = 30, r = −0.2, P = 0.2). This is in contrast to our previous study in which a significant correlation was observed in a small sample of nine subjects (Daskalakis et al. 2008).

Study 3.4: effect of TMS-induced auditory activation.

To rule out the effect of TMS-induced auditory potentials contributing to the above-mentioned findings, activity of the cortical evoked potentials following application of sham stimulation was subtracted from the cortical evoked potentials following active stimulation in both motor cortex and DLPFC in 26 subjects who had received both active and sham stimulation. Following this subtraction, the correlation between EMG measures of LICI and EEG measures of LICI remained significant (r = 0.59, P = 0.001). Furthermore, similar to the above-cited findings, delta, theta, and alpha oscillations were significantly inhibited in the motor cortex and DLPFC (P < 0.01 for all conditions), whereas beta and gamma oscillations were significantly inhibited only in the DLPFC (P = 0.001).

DISCUSSION

This study had three main findings. First, we established the validity of TMS–EEG measure of LICI by evaluating its relationship to conventional EMG measures of GABAB receptor inhibition in the motor cortex. We showed that 1) following application of LICI to the left motor cortex, EEG measures of LICI correlated significantly with EMG measures of LICI in an extended sample size; 2) EEG measures of LICI correlated significantly with duration of CSP; and 3) EMG measures of LICI were also significantly correlated with duration of CSP. Second, we demonstrated that all indices of LICI had high test–retest reliability in both motor cortex and DLPFC. Finally, in an extended sample, we replicated previous findings demonstrating that LICI resulted in significant inhibition of beta and gamma oscillations in the DLPFC but not in the motor cortex, in which only lower frequencies (i.e., delta, theta, and alpha oscillations) were inhibited. We further demonstrated that LICI measured from DLPFC did not correlate with LICI measured from the motor cortex.

The results of this study further confirm that TMS–EEG is a valid neurophysiological technique that allows for recording CI from motor and non-motor regions of the cortex. The finding that EEG measures of LICI correlate with EMG measures of LICI suggests that LICI has a cortical origin, as previously suggested (Chen et al. 1999; Di Lazzaro et al. 2002; Nakamura et al. 1997). Importantly, the finding that EEG and EMG measures of LICI correlate with CSP duration suggests that all these measures are likely related to similar neurophysiological mechanisms. Although both LICI and CSP have been separately associated with GABAB receptor-mediated inhibitory neurotransmission, previous reports did not find a direct association between the two measures. For example, CSP was not prolonged following oral administration of baclofen (Inghilleri et al. 1996; McDonnell et al. 2006), whereas LICI was enhanced (McDonnell et al. 2006). Furthermore, following fatiguing hand exercise, LICI was enhanced whereas CSP remained unchanged (Benwell et al. 2007). There are several reasons why previous studies might not have found a correlation between these measures. First, in this study, we delivered ≤100 trials per condition (i.e., 100 trials in LICI and 50 trials in CSP), whereas previous studies (Inghilleri et al. 1996; McDonnell et al. 2006) administered fewer stimuli (e.g., 10 trials). There is a high degree of variability in amplitude or area of MEPs elicited by identical consecutive stimuli (Kiers et al. 1993). Thus to obtain a true mean value, an adequate number of stimuli should be delivered (Kiers et al. 1993). In this regard, a progressive decrease in MEP area variability was shown as the number of stimuli increased (Brasil-Neto et al. 1992; Kiers et al. 1993). Therefore the above-mentioned inconsistency in the LICI–CSP relationship may be related to a high degree of intrasubject variability due to insufficient number of trials delivered per condition. Finally, previous studies suggest that the early part (i.e., first 50 ms) of CSP is due to spinal mechanism, whereas the later part is mediated by cortical inhibition (Chen et al. 1999; Inghilleri et al. 1993). In this study, we found a stronger correlation between EEG, rather than EMG, measures of LICI and CSP. It may be that the early part of CSP and, to some extent, the EMG measures of LICI are mediated by a different spinal mechanism and recording LICI directly from the motor cortex minimizes the variability that has non-cortical origins.

Reliability of TMS-induced MEPs have been previously studied in healthy subjects. For example, it has been shown that TMS motor maps for the APB muscle were reliable between two sessions (Corneal et al. 2005). A few studies have also investigated the reliability of short-interval paired-pulse TMS in the motor cortex. For example, short-interval cortical inhibition and facilitation were shown to be reliable (Boroojerdi et al. 2000; Maeda et al. 2002). Boroojerdi and colleagues further demonstrated that intersession variability could be reduced by increasing the number of stimuli delivered. More recently, Lioumis et al. (2009) demonstrated the reproducibility of TMS-evoked EEG responses in motor cortex and DLPFC. It was demonstrated that the amplitudes of TMS-evoked cortical potentials were highly reliable in both the DLPFC and motor cortex, at various stimulus intensities, within a 1-wk interval. Our study extends the previous reports by demonstrating the reliability of LICI in the motor cortex measured through both EMG and EEG and in the DLPFC measured through EEG. Our results further demonstrate that combined TMS–EEG allows for reliable measurement of LICI across cortical oscillations.

Our findings also demonstrate that LICI of the DLPFC and motor cortex are likely mediated by different neural circuits. This is supported by two observations. First, the overall inhibition (CItotal) in the DLPFC did not correlate with the CItotal in the motor cortex. Second, gamma and beta oscillations were inhibited in the DLPFC but not in the motor cortex. As previously suggested (Farzan et al. 2009), this selective inhibition of gamma oscillations in the DLPFC, compared with that in the motor cortex, may be related to the functional role of this cortical region in mediating higher-order cortical processes. That is, the inhibition of gamma oscillations may be necessary for optimal functioning of DLPFC in healthy subjects. Finally, the fact that we did not find a correlation between LICI in the DLPFC and motor cortex emphasizes the importance of recording CI directly from DLPFC, which is made possible through a combination of TMS with EEG.

Some limitations of our study should be noted. First, although we found a high test–retest reliability, interrater reliability needs to be examined in future studies. Second, in this study we tested subjects within 1 wk of the first session, whereas reliability should be examined for longer retest intervals to further investigate the long-term reliability of CI measured through combined TMS–EEG. Furthermore, future studies should examine the effect of trigeminal activation on LICI measured from motor cortex and DLPFC. It has been demonstrated that a conditioning stimulus, applied over the left supraorbital foramen, activates the trigeminal nerve and inhibits the peripheral MEP elicited by a test stimulus applied to the left motor cortex (Siebner et al. 1999). Although trigeminomotor inhibition has been shown to occur at ISIs of 30–65 ms, and it was not observed at an ISI of 100 ms, the interference of trigeminomotor inhibition on EEG measures of LICI cannot be completely ruled out. In addition, the modulatory effect of single-pulse TMS on alpha and beta oscillations was previously demonstrated at the subcortical level and through recordings from electrodes implanted in subthalamic nucleus in patients with Parkinson's disease (Gaynor et al. 2008). In our study, TMS modulation of oscillations is investigated through EEG at the scalp surface. To enhance our understanding of the origin and pathways of TMS-induced modulation of brain oscillations, future studies should use source localization techniques or attempt to record brain oscillations simultaneously from both cortical (through EEG) and subcortical regions (through subcortical electrodes). Finally, it has been recently shown that voluntary contraction of target muscle decreases LICI compared with LICI measured at rest (Hammond and Vallence 2007). In our study LICI was indexed at rest only and, as such, future studies should further examine the relationship between CSP duration and magnitude of LICI measured at various levels of voluntary contraction of target muscle.

In summary, combining TMS with EEG is a valid and reliable technique for recording of LICI from the motor cortex and DLPFC. It is also possible to reliably record the effect of LICI on cortical oscillations in different cortical regions. The finding that EEG measures of LICI correlate with the duration of CSP further suggests that LICI and CSP are both related to similar neurotransmitter mechanisms such as postsynaptic GABAB receptor-mediated inhibition. Finally, the differing effects of LICI on cortical oscillations in the DLPFC compared with those in the motor cortex emphasize the importance of measuring CI directly from the cortical regions where these impairments may underlie the pathophysiology of various neuropsychiatric disorders.

GRANTS

This work was funded, in part, by Canadian Institutes of Health Research (CIHR) Grant MOP 62917 to R. Chen; CIHR Doctoral Award to F. Farzan; CIHR Clinician Scientist Award to Z. J. Daskalakis; Operating and Studentship Awards from the Ontario Mental Health Foundation to Z. J. Daskalakis and M. S. Barr, respectively; a National Health and Medical Research Council Practitioner Fellowship to P. B. Fitzgerald; and Constance and Stephen Lieber through a National Alliance for Research on Schizophrenia and Depression Lieber Young Investigator awards to Z. J. Daskalakis, P. B. Fitzgerald, and A. J. Levinson.

DISCLOSURES

Z. J. Daskalakis and P. B. Fitzgerald both received external funding through Neuronetics Inc. (Malvern, PA). Z. J. Daskalakis received external funding through Aspect Medical Systems, Inc. (Norwood, MA) and travel support through Pfizer Inc. (New York, NY). P. B. Fitzgerald has received equipment for research studies from MagVenture A/S (Farum, Denmark) and Brainsway Ltd (Jerusalem, Israel).

ACKNOWLEDGMENTS

We thank all persons and volunteers for participation in the successful completion of the study.

REFERENCES

View Abstract