Abstract

Transcranial magnetic stimulation (TMS) of the primary motor cortex (M1) evokes several volleys of corticospinal activity. While the earliest wave (D-wave) originates from axonal activation of cortico-spinal neurons (CSN), later waves (I-waves) result from activation of mono- and polysynaptic inputs to CSNs. Different coil orientations preferentially stimulate cortical elements evoking different outputs: latero-medial-induced current (LM) elicits D-waves and short-latency electromyographic responses (MEPs); posterior–anterior current (PA) evokes early I-waves. Anterior–posterior current (AP) is more variable and tends to recruit later I-waves, featuring longer onset latencies compared with PA-TMS. We tested whether the variability in response to AP-TMS was related to functional connectivity of the stimulated M1 in 20 right-handed healthy subjects who underwent functional magnetic resonance imaging while performing an isometric contraction task. The MEP-latency after AP-TMS (relative to LM-TMS) was strongly correlated with functional connectivity between the stimulated M1 and a network involving cortical premotor areas. This indicates that stronger premotor–M1 connectivity increases the probability that AP-TMS recruits shorter latency input to CSNs. In conclusion, our data strongly support the hypothesis that TMS of M1 activates distinct neuronal pathways depending on the orientation of the stimulation coil. Particularly, AP currents seem to recruit short latency cortico-cortical projections from premotor areas.

Introduction

Most of our knowledge about the actions of transcranial magnetic stimulation (TMS) derives from studies of the motor cortex, where a single TMS pulse evokes a simple and quantifiable output: the contralateral muscle twitch which is readily measured by electromyography (motor evoked potential, (MEP)). Although it remains unknown what and where TMS stimulates, central recordings of the descending activity set up in the corticospinal tract by single-pulse TMS over the primary motor cortex (M1) reveal a series of volleys of activity which have a frequency of ∼600 Hz at spinal level (Berardelli et al. 1990; Burke et al. 1993; Di Lazzaro et al. 2012). The earliest wave (D-wave) originates from direct activation of axons of corticospinal neurons (CSN). The later waves (I-waves) are thought to result from mono- and polysynaptic inputs to CSNs and may interact with intrinsic membrane properties of the cells resulting in a pattern of synchronized activity (Day et al. 1989; Di Lazzaro et al. 1998a).

It is also well known that monophasic single-pulse TMS of M1 with different directions of induced electrical current evokes different combinations of D- and I-waves that recruit MEPs in hand muscles with characteristic onset latencies (Day et al. 1989; Hamada et al. 2013). These effects are best observed when TMS is applied during a weak voluntary contraction, which raises the excitability of synaptic connections in the corticospinal pathway. Latero-medial (LM)-induced current activates corticospinal axons directly in the subcortical white matter and produces MEPs with the shortest latencies. Stimulation with posterior–anterior (PA) current activates presumed monosynaptic inputs to CSNs; as a consequence MEPs occur 1–2 ms later compared with LM-TMS. Anterior–posterior (AP) TMS tends to recruit more indirect inputs to CSNs, which results in MEP-onset latencies that are longer compared with PA-TMS (Day et al. 1989; Sakai et al. 1997; Di Lazzaro et al. 2012). Therefore, it is likely that TMS activates different cortical elements in the brain depending on the direction of the induced current.

However, TMS may be more complicated than we thought. Although the difference in MEP onsets between LM- and PA-TMS is relatively constant between individuals, suggesting that similar input pathways are activated in all participants, we have shown recently that there is much more variation in the individual response to AP-TMS: in some people the MEP latency can be the same as for PA-TMS whereas in other subjects it may be up to 4 ms longer (Hamada et al. 2013). One reason for this could be anatomical: AP currents might activate anatomically different inputs to CSNs in different individuals, depending perhaps on their orientation with respect to the induced current. In contrast, the reason could also be functional: AP currents could activate similar inputs to CSNs in all individuals, however, the excitability of each pathway may vary between individuals depending on its functional state. Therefore, the pathway causing CSN activation might be determined by the individual pattern of preactivation (Fig. 1). The present study sets out to test this latter hypothesis.

Figure 1.

(A) Activation of a direct (e.g., monosynaptic) pathway projecting from premotor cortex (PMC) onto CSN situated in primary motor cortex (M1) by AP-TMS, resulting in short onset latency MEPs. In contrast, short latency MEPs induced by PA-TMS are thought to result from trans-synaptic activation of CSN within M1 (modified from Lemon 2008). (B) Longer onset latencies following AP-TMS might be due to the stimulation of indirect (e.g., polysynaptic) pathways from PMC onto CSNs. Individual recruitment may arise from preactivation-dependent differences in excitability between direct and indirect pathways.

Figure 1.

(A) Activation of a direct (e.g., monosynaptic) pathway projecting from premotor cortex (PMC) onto CSN situated in primary motor cortex (M1) by AP-TMS, resulting in short onset latency MEPs. In contrast, short latency MEPs induced by PA-TMS are thought to result from trans-synaptic activation of CSN within M1 (modified from Lemon 2008). (B) Longer onset latencies following AP-TMS might be due to the stimulation of indirect (e.g., polysynaptic) pathways from PMC onto CSNs. Individual recruitment may arise from preactivation-dependent differences in excitability between direct and indirect pathways.

The variation in responses to AP-TMS is only observed at low intensities of stimulation when MEPs are evoked while volunteers are performing a low level voluntary contraction. We reasoned that if a person's response to AP-TMS depends on the relative excitability of particular input pathways to CSNs, then it should reflect levels of activity in those pathways during the contraction. To gain some insight into the latter we used a connectivity analysis on data from functional magnetic resonance imaging (fMRI), again recorded during low intensity contraction of the TMS target muscles, in the same individuals.

Functional connectivity estimated via fMRI assesses the temporal correlation between blood oxygen level–dependent (BOLD) activity-changes in remote regions of brain (Friston 1994). Areas that are more directly connected should yield higher functional connectivity estimates compared with those featuring more nodes in-between (introducing more noise and, thereby, uncorrelated activity). Accordingly, we calculated the pattern of functional connectivity of the motor cortical hand representation targeted by TMS (“TMS hotspot”) during weak voluntary contraction. As mentioned above, our hypothesis was that AP currents could activate the same inputs to CSNs in all individuals, but the excitability of each pathway may vary due to the functional state of each pathway (Fig. 1). Indeed the fMRI analysis should also reveal the origin of the shared pathways activated by AP-TMS since they will be represented by regions, which feature strong connectivity with the stimulation hotspot and correlate with MEP latency.

Finally, it is possible to make predictions about the response to PA-TMS. Since the onset latency varies little between people, we presume its pathway is readily excited in all individuals, and therefore there will be no correlation between MEP latencies from PA-TMS and the functional connectivity analysis. The primary goal of this study was to identify whether inter-individual patterns of preactivation within the motor system might underlie the variance of AP-TMS-induced MEP-onset latency, thereby furthering our insights into the mechanisms of action of TMS.

Materials and Methods

Subjects

Twenty healthy subjects (12 females, all right handed, mean age 28.7 ± 6.81 standard deviation [SD]) without any history of neurological or psychiatric disease or contraindication to TMS (Rossi et al. 2009) participated in our study after giving written informed consent in accordance with the Declaration of Helsinki. The study was approved by the local ethics committee of the medical faculty of the University of Cologne.

Recordings

Subjects were seated in a comfortable chair. EMG activity was recorded from the right first dorsal interosseous (FDI) muscle using Ag/AgCl surface electrodes (Tyco Healthcare, Neustadt, Germany) placed in a belly-to-tendon montage. The EMG signal was amplified, filtered (0.5 Hz high-pass and 30–300 Hz band-pass) and digitized using a PowerLab 26T and LabChart software package version 6.0 (ADInstruments Ltd, Dunedin, New Zealand).

Transcranial Magnetic Stimulation

TMS was performed using a Magstim 2002 stimulator (The Magstim Co. Ltd) equipped with a 70 mm figure-of-eight coil. Coil positions were monitored and recorded throughout the TMS session with a Brainsight2 computerized frameless stereotaxic system (Rogue Research Inc., Montreal, Canada). By applying single monophasic TMS pulses over the cortical hand area, different descending volleys can be elicited depending on coil orientation: (1) posterior–anterior (PA) directed currents evoke I-waves which have a short onset latency, (2) anterior–posterior (AP) directed currents preferentially recruit I-waves with longer onset latencies, although difference varies between subjects (Hamada et al. 2013), and (3) latero-medial (LM) directed currents at high intensities evoke direct waves (D-waves; Day et al. 1989; Di Lazzaro et al. 1998a; Hamada et al. 2013). The following coil orientations were used in the present study to induce the 3 current directions mentioned above: (1) PA directed current: coil held postero-laterally at an angle of ∼45° to midline, (2) AP directed current: coil positioned at 180° relative to (1), and (3) LM directed current: coil 90° to midline with the handle pointing to the left (for further details see Hamada et al. 2013). The “TMS hotspot” was defined as coil position eliciting maximal FDI-response to PA currents with minimal stimulator output intensity. For both other coil orientations the same hotspot was used (Sakai et al. 1997; Arai et al. 2005; Diekhoff et al. 2011). Coil positions were logged into the neuronavigation software and maintained throughout the experiment. The resting motor threshold (RMT) was defined as the minimum stimulator intensity to evoke an MEP of at least 50 µV in >5 out of 10 trials, applying a PA directed current (Rossi et al. 2009). The active motor threshold (AMT) was defined as the minimal intensity needed to elicit an MEP of 200 µV in >5 out of 10 trials during ∼10% contraction of the FDI muscle (monitored by a force transducer).

Experimental Parameters

The latency between TMS pulse application and MEP onset was measured for PA-, AP-, and LM-directed currents during constant contraction of the FDI muscle (∼10% of the maximum contraction, with online feedback presented on a monitor). Stimulus intensity was set to 110% AMTpa, 110% AMTap, and 150% AMTlm (or 50% of maximum stimulator outputs [MSOs] in subjects whose 150% AMTlm did not reach 50% MSO). This constraint was used in order to assure that MEP elicited in LM orientation really reflected D-waves (Werhahn et al. 1994). To confirm the reproducibility of the data, 20 MEPs were obtained for PA and AP currents and 10 for LM currents. Following 10 stimulations subjects were asked to relax their hand to avoid fatigue. These measurements were taken over 10–15 min. Of note, all subjects successfully maintained the required force during MEP recordings without significant fluctuations. The shortest latency was measured by an automated method to minimize any observer bias, using a custom-made program. In each trial, the onset was defined as the time point where rectified EMG signals exceed an average plus 2 SDs of the prestimulus EMG level (−100 to 0 ms of TMS).

Previous studies reported that the onset latencies of MEPs depended on coil orientation with those evoked by LM currents having the shortest latency, followed by MEPs induced by PA current and finally MEPs elicited by AP current (Hamada et al. 2013). These findings suggest that PA currents preferentially activate early I-waves, and AP currents mainly I-waves with longer onset latencies (Day et al. 1989; Rothwell 1997; Sakai et al. 1997). Although the latencies evoked by PA currents are relatively consistent, the MEP onset following AP-TMS was reported to be more variable within a relatively large cohort of healthy subjects (Hamada et al. 2013), indicating that I-waves recruited by AP currents are not necessarily late I-waves of the PA cortical circuit, but may represent I-waves produced by a different cortical circuit. Given that PA and AP currents activate different neuronal pathway (Di Lazzaro et al. 2012), AP currents evoked I-waves are likely to be different from PA evoked I-waves (Di Lazzaro and Ziemann 2013). Therefore in this study, we decided to use the latency difference between LM and AP or PA evoked MEP onsets as a measure of I-wave recruitment, with longer latency differences reflecting more efficient late I-wave recruitment (Hamada et al. 2013). It should be noted that the amplitudes of the MEPs were similar for AP-TMS and PA-TMS since they employed stimuli of the same relative intensity (i.e., 110% AMT). MEPs following LM-TMS were larger since a higher intensity was used in order to make sure the shortest latency D-wave volley was evoked.

Functional Magnetic Resonance Imaging

We probed motor system connectivity during constant contraction of the FDI muscle. FMRI data were recorded on a Siemens Trio 3.0 T scanner (Siemens Medical Solutions, Erlangen, Germany) using a gradient echo planar imaging (EPI) sequence with the following parameters: time repetition (TR) = 2070 ms, time echo (TE) = 30 ms, field of view (FoV) = 200 mm, 32 slices, voxel size: 3.1 × 3.1 × 3.1 mm3, 1 mm gap, flip angle = 90°, EPIs: 225 volumes, 3 dummy images, duration: 7.5 min. The whole brain was covered by slices extending from the vertex to lower parts of the cerebellum. Instructions were presented to the subject in the scanner by a shielded TFT-screen, visible via a mirror mounted to the MR head-coil. Additionally, a functional localizer task was performed involving rhythmic index abductions, to identify individual motor areas for control purposes. In a separate session, high-resolution T1-weighted structural images were acquired (TR = 2250 ms, TE = 3.93 ms, FoV = 256 mm, 176 sagittal slices, voxel size = 1.0 × 1.0 × 1.0 mm3).

The active motor task consisted of a constant contraction against an fMRI compatible grip force sensor (Grip Force Bimanual Fiber Optic Response Pad; Current Designs, Inc., Philadelphia PA, USA) placed between the bases of the thumb and index finger of the right hand in pronation position of the hand. This setup allowed a rather selective recruitment of the FDI. Note that positioning of the grip-force sensor within the subject's right hand was identical for both the TMS session and fMRI recordings to achieve similar muscle-activation during constant thumb–index abduction. After measuring maximal contraction force, subjects were instructed to maintain a constant pressure of 4–6% of their maximal contraction force for the entire scanning session. Online feedback of the actual pressure exerted on the grip force sensor was provided via the TFT screen. The actual force was displayed in Newton, and the 4–6% range was given in green font color, coding successful task performance. Font color switched to red for grip forces outside the 4–6% range. Our preexperimental testing ensured that this force level could be maintained thought the entire 7.5-min fMRI scan without fatigue (in contrast to higher contraction forces like 10%), but is still sufficient to elicit significant connectivity estimates on a single-subject level. All 20 participants were capable of maintaining the required force without measurable deviations.

Image Preprocessing

Statistical Parametric Mapping (SPM8, http://www.fil.ion.ucl.ac.uk/) was used to preprocess the fMRI data. Spatial realignment to the mean image was performed for all EPI volumes followed by coregistration with the structural T1-weighted image. Spatial normalization was achieved via the unified segmentation approach (Ashburner and Friston 2005) using the standard template of the Montreal Neurological Institute (MNI, Canada). The T1-weighted image was segmented into tissue probability maps, with the gray matter map serving as a mask for all connectivity analyses. Finally, all images were smoothed with an isotropic Gaussian kernel of 6 mm full-width-at-half-maximum.

Functional Connectivity Analysis

We pursued a seed-voxel approach to compute functional connectivity of M1 (as defined by TMS) and any other voxel in the brain, representing the coherence in BOLD-fluctuations between the TMS hotspot (obtained from the neuronavigation software) and the rest of the brain (zu Eulenburg et al. 2012). To correct for confounds due to physiological processes and motion-related effects, variance explained by (1) motion parameters derived from image realignment, (2) their first derivative, (3) mean gray, white matter and CSF signal intensity, and (4) coherent signal changes across the whole brain were removed from the data (Bzdok et al. 2011; zu Eulenburg et al. 2012). Then, data were band pass filtered preserving frequencies between 0.01 and 0.08 Hz (Biswal et al. 1995; Fox and Raichle 2007). The first eigenvariate of the adjusted time-series was extracted for all voxels within 10 mm spheres around the individual TMS-hotspot position from the normalized single-subject data. Principle component analysis was performed using all time-series of voxels within the 10 mm sphere surrounding the seed. Linear (Pearson's) correlation coefficients were calculated between the first principle component representing BOLD-fluctuations at the TMS hotspot and every other voxel of the brain. Group analysis was performed using mass-univariate analysis (one sample t-tests) on Fisher's z-scores of the correlation coefficients across subjects. Effects passing a threshold of P < 0.05 (FWE cluster-level corrected for multiple comparisons, cluster-forming threshold: P < 0.05 uncorrected) were regarded as significant. In addition, for control purposes, functional connectivity was assessed for seed-voxels representing other key areas of the motor system, such as dorsal and ventral premotor cortex (dPMC, vPMC) and supplementary motor area (SMA). Individual seed coordinates (Table 2) were derived from the functional localizer task which successfully activated dPMC, vPMC, and SMA. Again, the adjusted time-series were extracted for all voxels within 10 mm spheres around the individual seed-voxels. Linear Pearson's correlation coefficients were calculated between the first principle component and the extracted time-series of every other voxel of the brain, computing whole-brain functional connectivity maps for each seed.

Comparison of TMS latency and Functional Connectivity

To assess whether individual TMS latency was related to functional connectivity within the active motor system (i.e., during constant FDI contraction) multiple regression analyses were performed including functional connectivity of left M1 and AP-LM latencies using SPM8. Of note, latency measurements and estimation of functional connectivity was blinded (M.H. assessed TMS latencies, L.J.V. computed functional connectivity). To assess the specificity of our results, regression analyses were also performed for PA-LM latencies. In order to control for putative confounds, individual AMTs and AMT differences were included as covariates in the regression analyses. Furthermore, multiple regression analyses were computed between TMS latency measurements and functional connectivity maps of other key motor areas not targeted by the stimulation (dPMC, vPMC, SMA; MNI coordinates are given in Table 2).

Motor Control Experiment

In order to test whether differences in cortical pathway recruitment following AP-TMS (Fig. 1) relate to behavioral parameters, we conducted a control experiment with a subsample of subjects (n = 12). Previous studies suggest premotor areas (e.g., the dorsal premotor cortex) to be involved in the generation of movements following external cues as implemented in reaction time tasks (Passingham et al. 2004; Moisa et al. 2012). Therefore, faster pathways between premotor areas and M1 (i.e., featured by subjects with small AP-LM latencies) might lead to faster motor responses in reaction time tasks. Hence, subjects were instructed to look at a centrally presented fixation cross (white cross on black screen), and to press a computer keyboard button with the right index finger as soon as a visual cue (white arrow pointing to the right; duration: 400 ms) appeared. The inter-stimulus intervals randomly varied between 900 and 2600 ms, thereby minimizing stimulus anticipation. After a 1-min training, subjects performed 3 blocks of trials, each consisting of 36 button presses. Hence, 108 index finger responses were recorded. Of note, the task relied on movements of the right index finger including activation of the TMS target muscle (i.e., right FDI). Finally, linear correlations were computed between individual MEP-onset latencies and RTs, as well as individual motor system connectivity and RTs.

Results

Individual TMS thresholds and latencies are given in Table 1. MNI coordinates of the TMS hotspot and other seed-voxels for functional connectivity estimation are presented in Table 2.

Table 1

Individual AMTs and coil orientation-dependent latency measures

Subject AMTpa AMTap AMTlm PA AP LM PA-LM AP-LM 
34 45 27 23.0 25.1 20.4 2.6 4.7 
24 37 29 25.9 25.9 21.1 4.8 4.8 
26 43 32 23.0 23.3 21.9 1.1 1.4 
30 49 43 22.6 23.3 20.8 1.9 2.5 
30 51 27 20.7 18.8 18.2 2.4 0.5 
33 30 31 19.1 23.3 17.7 1.4 5.6 
22 35 28 18.6 21.7 17.6 0.9 4.0 
61 78 54 19.7 20.8 17.7 2.0 3.1 
34 52 34 20.7 21.8 20.0 0.7 1.8 
10 26 32 31 22.0 24.7 20.2 1.8 4.5 
11 27 42 39 21.3 26.1 20.8 0.5 5.2 
12 27 39 29 20.6 22.2 17.7 2.9 4.5 
13 25 47 31 22.9 23.1 22.0 1.0 1.2 
14 28 34 34 20.9 22.4 19.7 1.3 2.8 
15 28 47 29 21.5 22.2 21.1 0.4 1.1 
16 21 28 22 22.8 24.1 22.1 0.7 2.1 
17 27 40 28 21.0 21.3 19.8 1.2 1.5 
18 24 31 41 21.4 22.3 18.5 2.9 3.7 
19 40 49 56 21.1 23.6 19.4 1.7 4.2 
20 30 42 30 23.5 26.0 20.8 2.7 5.2 
Mean 29.85 42.55 33.75 21.6 23.1 19.9 1.7 3.2 
SD 8.40 10.82 8.58 1.7 1.9 1.5 1.1 1.6 
Subject AMTpa AMTap AMTlm PA AP LM PA-LM AP-LM 
34 45 27 23.0 25.1 20.4 2.6 4.7 
24 37 29 25.9 25.9 21.1 4.8 4.8 
26 43 32 23.0 23.3 21.9 1.1 1.4 
30 49 43 22.6 23.3 20.8 1.9 2.5 
30 51 27 20.7 18.8 18.2 2.4 0.5 
33 30 31 19.1 23.3 17.7 1.4 5.6 
22 35 28 18.6 21.7 17.6 0.9 4.0 
61 78 54 19.7 20.8 17.7 2.0 3.1 
34 52 34 20.7 21.8 20.0 0.7 1.8 
10 26 32 31 22.0 24.7 20.2 1.8 4.5 
11 27 42 39 21.3 26.1 20.8 0.5 5.2 
12 27 39 29 20.6 22.2 17.7 2.9 4.5 
13 25 47 31 22.9 23.1 22.0 1.0 1.2 
14 28 34 34 20.9 22.4 19.7 1.3 2.8 
15 28 47 29 21.5 22.2 21.1 0.4 1.1 
16 21 28 22 22.8 24.1 22.1 0.7 2.1 
17 27 40 28 21.0 21.3 19.8 1.2 1.5 
18 24 31 41 21.4 22.3 18.5 2.9 3.7 
19 40 49 56 21.1 23.6 19.4 1.7 4.2 
20 30 42 30 23.5 26.0 20.8 2.7 5.2 
Mean 29.85 42.55 33.75 21.6 23.1 19.9 1.7 3.2 
SD 8.40 10.82 8.58 1.7 1.9 1.5 1.1 1.6 

Coil orientation represented as PA, posterior–anterior; AP, anterior–posterior, and LM, latero-medial.

Table 2

Individual coordinates of seed-voxels used for the computation of seed-based whole-brain functional connectivity (MNI coordinates)

Subject TMS Hotspot (M1) SMA vPMC dPMC 
−33 −11 69 −5 −5 65 −53 2 44 −35 −11 68 
−36 −27 71 −9 −11 57 −50 −5 44 −35 −11 68 
−36 −18 71 −8 −12 69 −54 −8 29 −35 −18 72 
−47 −11 63 −5 −5 57 −51 −2 33 −39 −6 62 
−42 −21 69 −5 −12 53 −57 −11 47 −44 −12 62 
−50 −17 53 −9 −9 63 −56 0 32 −30 −23 72 
−41 −17 66 −5 −14 53 −63 5 23 −39 −14 65 
−33 0 66 −5 −11 56 −56 −9 47 −33 −17 69 
−51 −9 59 −5 −9 65 −59 2 29 −35 −17 69 
10 −36 −5 60 −8 −11 62 −53 −9 47 −36 −8 63 
11 −44 −5 53 −9 0 51 −56 −6 44 −41 −9 60 
12 −42 −11 63 −5 −11 60 −62 9 23 −41 −12 63 
13 −45 −2 59 −5 0 59 −54 −2 42 −47 −6 55 
14 −38 −2 65 −8 −5 74 −56 −2 41 −33 −9 56 
15 −47 6 54 −5 −12 60 −60 8 32 −42 −9 60 
16 −32 −5 66 −6 −5 68 −58 0 23 −39 −9 65 
17 −54 −17 51 −6 −11 62 −56 5 39 −29 −14 74 
18 −53 −9 57 −5 −5 62 −57 9 39 −32 −17 71 
19 −48 −8 59 −5 −15 59 −57 6 39 −32 −6 66 
20 −47 −5 54 −5 −11 57 −54 −6 44 −36 −14 65 
Mean −43 −10 61 −6 −8 60 −56 −1 37 −36 −12 65 
SD 7 8 6 2 4 6 3 6 8 5 4 5 
Subject TMS Hotspot (M1) SMA vPMC dPMC 
−33 −11 69 −5 −5 65 −53 2 44 −35 −11 68 
−36 −27 71 −9 −11 57 −50 −5 44 −35 −11 68 
−36 −18 71 −8 −12 69 −54 −8 29 −35 −18 72 
−47 −11 63 −5 −5 57 −51 −2 33 −39 −6 62 
−42 −21 69 −5 −12 53 −57 −11 47 −44 −12 62 
−50 −17 53 −9 −9 63 −56 0 32 −30 −23 72 
−41 −17 66 −5 −14 53 −63 5 23 −39 −14 65 
−33 0 66 −5 −11 56 −56 −9 47 −33 −17 69 
−51 −9 59 −5 −9 65 −59 2 29 −35 −17 69 
10 −36 −5 60 −8 −11 62 −53 −9 47 −36 −8 63 
11 −44 −5 53 −9 0 51 −56 −6 44 −41 −9 60 
12 −42 −11 63 −5 −11 60 −62 9 23 −41 −12 63 
13 −45 −2 59 −5 0 59 −54 −2 42 −47 −6 55 
14 −38 −2 65 −8 −5 74 −56 −2 41 −33 −9 56 
15 −47 6 54 −5 −12 60 −60 8 32 −42 −9 60 
16 −32 −5 66 −6 −5 68 −58 0 23 −39 −9 65 
17 −54 −17 51 −6 −11 62 −56 5 39 −29 −14 74 
18 −53 −9 57 −5 −5 62 −57 9 39 −32 −17 71 
19 −48 −8 59 −5 −15 59 −57 6 39 −32 −6 66 
20 −47 −5 54 −5 −11 57 −54 −6 44 −36 −14 65 
Mean −43 −10 61 −6 −8 60 −56 −1 37 −36 −12 65 
SD 7 8 6 2 4 6 3 6 8 5 4 5 

SMA, supplementary motor area; vPMC, ventral premotor cortex; dPMC, dorsal premotor cortex.

Latency Differences and Coil Orientation

MEPs evoked by LM-TMS at high intensities (150% AMTlm) were used to estimate the individual D-wave onset. In fact, in all subjects, the onset latency of MEP evoked by LM current with high stimulus intensity was shortest compared with those elicited by PA or AP current (Table 1). Therefore, the latency of near-threshold MEPs elicited by AP stimulation relative to individual D-wave activation (AP-LM) was used to evaluate presumed I-wave recruitment. Accordingly, the onset latency of near-threshold MEPs elicited by the more standard PA stimulation was also measured relative to individual D-wave activation (PA-LM). The group results of all 20 subjects are shown in Figure 2. Consistent with previous results (Hamada et al. 2013), PA-LM latency was significantly shorter compared with AP-LM latency (2 sided paired t-test, P = 0.0003).

Figure 2.

(A) Schematic representation of coil orientations and typical example of MEPs during constant contraction of the TMS target muscle (FDI). Arrow indicates the timing of TMS and arrow head indicates the onset of MEPs. Calibration bars, 1 mV, 20 ms. (B) Group results of TMS PA-LM and AP-LM latency differences from 20 subjects.

Figure 2.

(A) Schematic representation of coil orientations and typical example of MEPs during constant contraction of the TMS target muscle (FDI). Arrow indicates the timing of TMS and arrow head indicates the onset of MEPs. Calibration bars, 1 mV, 20 ms. (B) Group results of TMS PA-LM and AP-LM latency differences from 20 subjects.

Functional Connectivity

Functional connectivity maps seeded from the stimulated left M1 (i.e., TMS hotspot) revealed correlated activity during “contraction state” in a lateralized network comprising fronto-parietal sensorimotor regions (Fig. 3A). Strongest connectivity was found in the vicinity of the seed voxel in left M1. When correlating these connectivity estimates with individual AP-LM latencies, significant negative correlations were found for M1 seeds in a motor cortical network including left M1, left ventral and dorsal premotor cortex and bilateral supplementary motor areas (P < 0.05 FWE cluster-level corrected; Fig. 3B). Hence, the stronger the connectivity of these motor areas with the stimulated M1, the shorter was the AP-LM latency. In other words, it took less time to activate the right FDI-muscle using AP oriented TMS over left M1 in subjects who featured stronger functional connectivity of M1 within the cortical motor network. On the other hand, there were no significant positive correlations between M1-connectivity and AP-LM latency. Thus, in subjects featuring longer AP-LM latencies no brain region showed higher synchronicity in BOLD-fluctuations with left M1 during contraction of the right hand.

Figure 3.

(A) Group functional connectivity of M1 (TMS hotspot) during constant thumb–index-contraction (at 5% maximal force involving activation of the FDI muscle; P < 0.05, FWE cluster-level corrected). (B) Negative correlation of AP-LM latency and M1-connectivity. Marked areas featured strong M1-connectivity in subjects with short AP-LM latencies (P < 0.05, FWE cluster-level corrected). The scatterplot shows the significant correlation between M1-connectivity and AP-LM latency obtained from the cluster in (B).

Figure 3.

(A) Group functional connectivity of M1 (TMS hotspot) during constant thumb–index-contraction (at 5% maximal force involving activation of the FDI muscle; P < 0.05, FWE cluster-level corrected). (B) Negative correlation of AP-LM latency and M1-connectivity. Marked areas featured strong M1-connectivity in subjects with short AP-LM latencies (P < 0.05, FWE cluster-level corrected). The scatterplot shows the significant correlation between M1-connectivity and AP-LM latency obtained from the cluster in (B).

One might argue that MEP-onset latencies partly depend on the stimulation intensity, with inter-individual differences in coil orientation-specific AMTs possibly biasing the significant correlation between M1-connectivity and AP-LM latency. However, no significant correlations were evident between coil-dependent AMTs and latency onsets (AMTap∼AP: r = −0.364, P = 0.114; AMTpa∼PA: r = −0.317, P = 0.174; AMTlm∼LM: r = −0.264, P = 0.262). Furthermore, AP-LM latency differences did not correlate with MEP amplitudes evoked by AP currents (r = −0.337, P = 0.158). These findings are in accordance with previous data (Hamada et al. 2013). Furthermore, we included AMTap, AMTlm, and the AMT difference (AMTap-lm) as covariates into our regression analysis, which yielded very similar correlation results. In contrast, no significant correlations were evident between M1-connectivity and PA-LM latency (even if excluding all covariates).

Correlation with Connectivity Maps of Other Motor Areas

Finally, we tested for correlations between MEP-onset latencies with functional connectivity maps of other cortical motor areas, i.e., vPMC, dPMC, and SMA (MNI coordinates of seed-voxels are given in Table 2). However, functional connectivity of none of these areas did correlate with AP-LM nor PA-LM latencies, thus corroborating the specificity of the correlations found between functional connectivity of the stimulated M1 and AP-LM latencies.

Motor Performance

A simple RT task was performed for a subpopulation of our cohort (n = 12) resulting in a group mean RT of 287.7 ms ± 17.1 SD (time between visual cue and motor response). Linear Pearson's correlation analysis including MEP-onset latencies did not reveal a significant relation between RT and AP-LM (r = 0.282, P = 0.375) or between RT and PA-LM (r = −0.311, P = 0.325). Thus, individual differences in MEP-onset latencies following TMS with different current orientations do not seem to impact on motor performance in a simple RT task. Furthermore, no significant correlation was evident between motor system functional connectivity and RT (P > 0.3, FWE cluster-level corrected).

Discussion

The present data confirm earlier findings that different coil orientations at the identical stimulation spot produce MEPs with different onset latencies: PA stimulation recruits MEPs that have an earlier onset compared with AP stimulation. In addition, there is a particularly large variation in latencies for AP-TMS, expressed relative to the shortest D-wave latency seen with LM-TMS. Parallel fMRI experiments on the same group of subjects showed that there was high functional connectivity between the hand area of left M1 and a widespread network of frontal and parietal sites, during constant hand-muscle contraction. The novel finding was that the AP latency correlated with functional connectivity of left M1 with ipsilateral premotor cortex and bilateral SMA. Thus, people who had the earliest onset MEPs to AP stimulation (AP-LM difference) featured the strongest functional connectivity in this frontal motor network. Of note, functional connectivity estimates were assessed in the same state as MEP latencies, i.e., during constant contraction of the FDI muscle. Importantly, PA latency did not correlate with functional connectivity of the stimulation hotspot.

We argue below that functional connectivity during constant contraction is likely to be strongest between cortical sites that have direct, short latency synaptic connections. We hypothesize that: (1) PA stimulation activates fibers projecting onto CSNs within M1, (2) AP stimulation activates a range of synaptic inputs from frontal motor areas onto M1, and (3) the inputs that are activated are the ones that are most excitable at the time of the TMS pulse. Thus, people in whom voluntary contraction most strongly increases the excitability of direct inputs to M1 will be more likely to have these inputs activated by a TMS pulse.

Coil Orientation and Descending Activity

We still have rather limited knowledge about how TMS-induced currents interact with the complex circuitry of the human cortex (Di Lazzaro and Ziemann 2013). Estimates of the strength–duration time constant suggest that TMS activates relatively large diameter axons (Peterchev et al. 2013). Because axonal threshold is sensitive to the direction of applied current different directions of induced current at the same cortical point can stimulate different populations of neurons depending on their orientation with respect to the stimulus. Invasive recordings of descending corticospinal activity demonstrate that a monophasic, PA-directed current over the human central sulcus evokes the lowest threshold volley (I1-wave) at a latency that is 1.0–1.4 ms longer than a volley induced by LM oriented TMS or electrical stimulation (D-wave; Di Lazzaro et al. 2008). Reversing the current direction to AP results in more complex and variable responses in which the lowest threshold volley often has a longer and more variable latency (Di Lazzaro et al. 2008; Hamada et al. 2013). Our latency data match these observations, with PA-LM latencies being significantly shorter than AP-LM latencies together with a broader distribution of individual AP-LM latencies (Fig. 2).

Models of cortical stimulation suggest that network interactions of excitatory and inhibitory circuits interact with intrinsic neuronal properties (model of neural oscillation) to produce the I-wave periodicity of CSN discharge (Esser et al. 2005; Di Lazzaro et al. 2012). Importantly, even direct monosynaptic inputs to CSNs would set up the same reverberating activity because of the recurrent collateral projections from corticospinal axons back to the cortex. Thus, PA-TMS might activate early inputs to the CSNs, and initiate activity in the oscillating circuit, whereas AP-TMS might activate a variety of later arriving inputs that could initiate oscillations at a later time. This would explain why the latency of the I-waves as well as their temporal dispersion can differ slightly between AP- and PA-TMS (Di Lazzaro et al. 2012). Further support for the hypothesis that AP and PA-TMS-induced descending activity might be elicited by different neuronal mechanisms stems from a recent TMS study reporting a stronger influence of short-latency afferent inhibition (SAI) on I-waves generated by PA than those generated by AP current direction (Ni et al. 2011). Studying associative cortico-cortical plasticity, Koch et al. (2013) reported an impact of coil orientation (PA, AP) on bidirectional long-term potentiation or depression induced in M1 after paired stimulation with the posterior parietal cortex (PPC). Plasticity induced using the same interval between PPC and M1 (5 ms, PPC first) led to MEP suppression if the associative protocol employed a PA pulse whereas there was facilitation if an AP pulse had been used. This suggests that the elements activated by AP and PA pulses respond differently to pairing with PPC inputs. Furthermore, these effects were mediated by different patterns of cortical rhythms as revealed by TMS/electroencephalography experiments (Veniero et al. 2013).

One source of inputs to M1 that could be activated by TMS pulses are projections from premotor areas. Although stimulation of neurons in the ventral premotor cortex does not always cause CSNs in M1 to discharge (but see Maier et al. 2013), it does increase the amplitude of late as well as early I-waves that are evoked by direct M1 stimulation a few milliseconds later. The implication is that premotor inputs excite some of the same circuitry that participates in I-wave generation (Shimazu et al. 2004; Lemon 2008; Di Lazzaro and Ziemann 2013). A similar result can be observed in humans. TMS applied over left dPMC during tonic contraction of the target muscle, facilitated MEPs evoked by subsequent stimulation of left M1. Importantly, this occurred at interstimulus intervals of 1.2, 2.4–2.8 and 4.0 ms which again suggests that subliminal input from premotor cortex was utilizing circuits involved in production of I-waves (Groppa et al. 2012). Similar results have been obtained for double-pulse stimulation over vPMC and M1 (Baumer et al. 2009).

Origin of TMS and fMRI Correlations

In the present study, functional connectivity was assessed via fMRI using the same motor task as for TMS latency recordings (i.e., constant FDI-contraction). Higher connectivity reflects stronger correlation of the activation time courses between the respective regions (functional connectivity, Friston 1994). Although we cannot infer with certainty the number of synapses involved in mediating this correlation, it seems likely that the smaller the number of synapses involved, the greater the synchronization of activity between the presynaptic and postsynaptic neurons (i.e., CSN). Conversely, a large number of synapses between these cells will give rise to more variance and noise and lower the probability of producing correlated activity. Thus, given that CSNs are situated within the chosen seed-voxel, it is reasonable to assume that functional connectivity estimates might be influenced by the number of intercalated synapses between this and other sites.

It is well known that stimulation produces larger effects if it is applied during preactivation of the stimulated tissue (Barker et al. 1985; Di Lazzaro et al. 1998b). Thus, TMS will be biased to activate any pathways that are also activated during voluntary muscle contraction. In the present instance, TMS will be biased to activate any shared pathways that show strong functional connectivity. Furthermore, since functional connectivity is strongest in direct connections, the TMS responses evoked in this way will have short onset latencies. The present results show that AP-TMS evokes shorter latency MEPs (relative to LM-TMS) in people in whom there is strong functional connectivity in pathways from premotor areas to M1. This would be compatible with the idea that AP-TMS activates relatively direct pathways from premotor cortex to M1 in all individuals. However, stimulation only evokes descending M1 output in those people in whom the excitability of the pathway is increased by a voluntary contraction. The greater the connectivity in this pathway the greater the probability that it will contribute to early onset of MEPs evoked by AP-TMS.

It is important to note 2 things about this explanation. First, we assume that all sites that have high functional connectivity feature relatively direct synaptic inputs to the seed voxel in M1. However, it is only those sites in premotor cortex and SMA that are correlated with AP-LM latency difference. This strongly suggests that AP stimulation activates inputs coming from these areas but not inputs from more posterior areas that also show high levels of functional connectivity. The second point is that the present explanation is still compatible with the general finding that the mean onset latency of AP stimulation is longer compared to PA stimulation. Since PA-TMS is thought to activate monosynaptic inputs to M1 it is usually assumed that AP stimulation either activates polysynaptic inputs and/or monosynaptic inputs at some distance from M1 (Fig. 1). However, the latency difference varies and our assumption is that those people who have the shortest latency AP responses are the ones who have the highest levels of activity in direct pathways with the shortest number of synapses.

Furthermore, this scenario might also explain why M1 connectivity did not relate to PA-LM latencies. Near threshold PA stimulation is thought to cause monosynaptic activation of CSNs within M1 in all individuals and its efficiency would depend only on the level of excitability in the CSNs themselves rather than any interposed interneurones (Fig. 1; Day et al. 1989; Di Lazzaro et al. 2008; Lemon 2008). In addition it could be that PA-TMS activates elements that are intrinsic to M1 and in this case we would not expect any relationship between functional connectivity to distant regions and MEP-onset latencies.

Implications

Ever since the introduction of TMS (Barker et al. 1985), it remains unknown what and where TMS stimulates. Furthering our insights into mechanisms underlying TMS-induced neural activity seems important, i.e., regarding evolving therapeutic applications of neural plasticity induction via repetitive TMS (rTMS). Various studies suggest the individual response to rTMS depends on several factors, such as age, gender, time of day, physical activity, and genetics (see review by Ridding and Ziemann 2010). Recently, it has been shown that individual AP-LM latencies correlated with another physiological parameter: the susceptibility to induction of neural plasticity via theta burst stimulation (TBS; Hamada et al. 2013). The shorter the onset latency of AP-TMS the less likely TBS was to produce its “canonical” effects (i.e., inhibition from continuous TBS and facilitation from intermittent TBS). As discussed by the authors, the biphasic (i.e., PA–AP) TMS pulse applied during TBS preferentially activates neurons during the reverse stimulation phase (i.e., AP; Maccabee et al. 1998; Di Lazzaro et al. 2001). Thus, MEPs evoked by AP stimulation might represent activation of the same elements as stimulated with TBS (Hamada et al. 2013). Therefore, our present data suggest that direct inputs from premotor areas do not contribute to the “canonical” effects of TBS paradigms. However, this hypothesis has to remain speculative and further research is needed to advance our insights into plasticity induction via rTMS and biomarkers to assess individual susceptibility. Identifying activated neuronal elements and furthering our insights into the mechanisms underlying TMS-induced activity may help to shed light on the variability of response to TMS.

We can only speculate whether or not similar relationships would have been found for resting-state fMRI data. While some studies suggest that TMS connectivity is related to resting-state fMRI (Koch et al. 2012), other studies showed stronger relationships between fMRI and TMS-induced effects in the activated motor system (Boudrias et al. 2012; Cardenas-Morales et al. 2014). Furthermore, fMRI resting-state connectivity estimates are rather weakly correlated with fMRI connectivity estimated for the activated motor system (Rehme et al. 2013). Therefore, it remains to be tested in the future whether differences in cortical latencies are also reflected by resting-state fMRI estimates.

Finally, we found no relation between MEP-onset latencies or motor system connectivity and RTs in a simple RT task. Thus, individual differences in the recruitment of cortico-cortical pathways following AP-TMS do not seem to impact on motor performance in a simple RT task. A possible explanation for this finding might lie in the distinctive activation of the cortical motor network by TMS compared with voluntary index finger movement as performed during the simple RT task. That functional connectivity assessed during FDI contraction did not correlate with reaction times is not surprising given the fundamental difference of the motor tasks (i.e., constant contraction vs. phasic reaction). Future studies are needed to determine whether the preferential activation of distinct cortico-cortical pathways by AP-TMS may hold an implication for other, e.g., more complex types of motor behavior in healthy subjects.

Limitations

We used MEP-onset latencies differences (AP-LM, PA-LM) as electrophysiological surrogates of the latency of CSN activation by TMS pulses of different orientation (see Materials and Methods for more detailed discussion). Given that LM stimulation at high intensities (i.e., 150% AMTlm) is thought to directly activate CSN axons (Werhahn et al. 1994), the latency differences represent the cortical delay in CSN activation, with more direct pathways presumably resulting in shorter latency differences. One might argue that MEP-onset latencies also depend on the stimulation intensity used to evoke the MEP, with higher intensities possibly resulting in shorter onset latencies. However, no significant correlations were evident between individual coil orientation-specific latencies and AMTs, which determined individual stimulation intensities. Furthermore, the correlation between M1-connectivity and AP-LM remained significant with both thresholds (AMTap, AMTlm) as well as the threshold difference (AMTap-lm) being included as covariates. Therefore, we can exclude the stimulation intensity to substantially bias latency differences as surrogates of cortico-cortical pathway recruitment.

If AP-LM or PA-LM latency differences represented the additional time to elicit descending activity following indirect, presumably trans-synaptic activation of CSNs (AP, PA) relative to direct depolarization of CSN axons (LM), all these measures should be inter-related across subjects, as all stimulations ultimately result in depolarization in CSNs, causing MEPs. Indeed, all latency measures (AP, PA, LM) were significantly correlated across subjects, corroborating our interpretation of AP-LM significance.

Finally, we used different contraction forces (10% for TMS, 5% for fMRI) to activate the motor system and allowed subjects to relax their hand muscles in between TMS recordings but not during the fMRI session because of methodological reasons. In case that one measurement had been more exhausting than the respective other measurement, this would have increased the variance of the data, and—as a consequence—would have made it more difficult to find a significant correlation. However, the fact that we found significant correlations despite putative variations in motor system activity underlines the robustness of our findings. In turn, this also implies that we have maybe missed additional correlations.

Conclusion

We present some of the first multimodal evidence for the involvement of premotor regions in the generation of I-waves in humans—supporting the hypothesis of orientation-dependent activation of distinct neuronal pathways. Furthermore, the response to AP stimulation was found to relate to functional M1-connectivity suggesting that individual preactivation and synaptic properties within the cortical motor network impact on stimulation responses following single-pulse TMS. Our finding suggests that functional preference of motor cortical networks substantially affects how TMS stimulates the neurons in the brain. Finally, our findings might help to further our insights into mechanisms underlying induction of cortical plasticity via rTMS

Funding

LJV and CG are supported by grants from the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG GR 3285/2–1 and GR 3285/5?1). MH was supported by the Japan Society for the Promotion of Science Postdoctoral Fellowships for Research Abroad. JCR has been supported by: the European Union (REPLACES, an FP7 collaborative project - 222918).

Notes

Conflict of Interest: M.H. serves as a medical advisor for Pfizer Japan Inc.

References

Arai
N
Okabe
S
Furubayashi
T
Terao
Y
Yuasa
K
Ugawa
Y
.
2005
.
Comparison between short train, monophasic and biphasic repetitive transcranial magnetic stimulation (rTMS) of the human motor cortex
.
Clin Neurophysiol
 .
116
:
605
613
.
Ashburner
J
Friston
KJ
.
2005
.
Unified segmentation
.
Neuroimage
 .
26
:
839
851
.
Barker
AT
Jalinous
R
Freeston
IL
.
1985
.
Non-invasive magnetic stimulation of human motor cortex
.
Lancet
 .
1
:
1106
1107
.
Baumer
T
Schippling
S
Kroeger
J
Zittel
S
Koch
G
Thomalla
G
Rothwell
JC
Siebner
HR
Orth
M
Munchau
A
.
2009
.
Inhibitory and facilitatory connectivity from ventral premotor to primary motor cortex in healthy humans at rest—a bifocal TMS study
.
Clin Neurophysiol
 .
120
:
1724
1731
.
Berardelli
A
Inghilleri
M
Cruccu
G
Manfredi
M
.
1990
.
Descending volley after electrical and magnetic transcranial stimulation in man
.
Neurosci Lett
 .
112
:
54
58
.
Biswal
B
Yetkin
FZ
Haughton
VM
Hyde
JS
.
1995
.
Functional connectivity in the motor cortex of resting human brain using echo-planar MRI
.
Magn Reson Med
 .
34
:
537
541
.
Boudrias
MH
Goncalves
CS
Penny
WD
Park
CH
Rossiter
HE
Talelli
P
Ward
NS
.
2012
.
Age-related changes in causal interactions between cortical motor regions during hand grip
.
Neuroimage
 .
59
:
3398
3405
.
Burke
D
Hicks
R
Gandevia
SC
Stephen
J
Woodforth
I
Crawford
M
.
1993
.
Direct comparison of corticospinal volleys in human subjects to transcranial magnetic and electrical stimulation
.
J Physiol
 .
470
:
383
393
.
Bzdok
D
Langner
R
Caspers
S
Kurth
F
Habel
U
Zilles
K
Laird
A
Eickhoff
SB
.
2011
.
ALE meta-analysis on facial judgments of trustworthiness and attractiveness
.
Brain Struct Funct
 .
215
:
209
223
.
Cardenas-Morales
L
Volz
LJ
Michely
J
Rehme
AK
Pool
EM
Nettekoven
C
Eickhoff
SB
Fink
GR
Grefkes
C
.
2014
.
Network Connectivity and individual responses to brain stimulation in the human motor system
.
Cereb Cortex
 .
24
:
1697
1707
.
Day
BL
Dressler
D
Maertens de Noordhout
A
Marsden
CD
Nakashima
K
Rothwell
JC
Thompson
PD
.
1989
.
Electric and magnetic stimulation of human motor cortex: surface EMG and single motor unit responses
.
J Physiol
 .
412
:
449
473
.
Diekhoff
S
Uludag
K
Sparing
R
Tittgemeyer
M
Cavusoglu
M
von Cramon
DY
Grefkes
C
.
2011
.
Functional localization in the human brain: gradient-echo, spin-echo, and arterial spin-labeling fMRI compared with neuronavigated TMS
.
Hum Brain Mapp
 .
32
:
341
357
.
Di Lazzaro
V
Oliviero
A
Profice
P
Saturno
E
Pilato
F
Insola
A
Mazzone
P
Tonali
P
Rothwell
JC
.
1998a
.
Comparison of descending volleys evoked by transcranial magnetic and electric stimulation in conscious humans
.
EEG Clin Neurophysiol
 .
109
:
397
401
.
Di Lazzaro
V
Oliviero
A
Saturno
E
Pilato
F
Insola
A
Mazzone
P
Profice
P
Tonali
P
Rothwell
JC
.
2001
.
The effect on corticospinal volleys of reversing the direction of current induced in the motor cortex by transcranial magnetic stimulation
.
Exp Brain Res
 .
138
:
268
273
.
Di Lazzaro
V
Pilato
F
Dileone
M
Profice
P
Oliviero
A
Mazzone
P
Insola
A
Ranieri
F
Meglio
M
Tonali
PA
et al
2008
.
The physiological basis of the effects of intermittent theta burst stimulation of the human motor cortex
.
J Physiol
 .
586
:
3871
3879
.
Di Lazzaro
V
Profice
P
Ranieri
F
Capone
F
Dileone
M
Oliviero
A
Pilato
F
.
2012
.
I-wave origin and modulation
.
Brain Stimul
 .
5
:
512
525
.
Di Lazzaro
V
Restuccia
D
Oliviero
A
Profice
P
Ferrara
L
Insola
A
Mazzone
P
Tonali
P
Rothwell
JC
.
1998b
.
Effects of voluntary contraction on descending volleys evoked by transcranial stimulation in conscious humans
.
J Physiol
 .
508
(2)
:
625
633
.
Di Lazzaro
V
Ziemann
U
.
2013
.
The contribution of transcranial magnetic stimulation in the functional evaluation of microcircuits in human motor cortex
.
Front Neural Circuits
 .
7
:
18
.
Esser
SK
Hill
SL
Tononi
G
.
2005
.
Modeling the effects of transcranial magnetic stimulation on cortical circuits
.
J Neurophysiol
 .
94
:
622
639
.
Fox
MD
Raichle
ME
.
2007
.
Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging
.
Nat Rev Neurosci
 .
8
:
700
711
.
Friston
KJ
.
1994
.
Functional and effective connectivity in neuroimaging: a synthesis
.
Hum Brain Mapp
 .
2
:
56
78
.
Groppa
S
Werner-Petroll
N
Munchau
A
Deuschl
G
Ruschworth
MF
Siebner
HR
.
2012
.
A novel dual-site transcranial magnetic stimulation paradigm to probe fast facilitatory inputs from ipsilateral dorsal premotor cortex to primary motor cortex
.
Neuroimage
 .
62
:
500
509
.
Hamada
M
Murase
N
Hasan
A
Balaratnam
M
Rothwell
JC
.
2013
.
The role of interneuron networks in driving human motor cortical plasticity
.
Cereb Cortex
 .
23
:
1593
1605
.
Koch
G
Bozzali
M
Bonni
S
Giacobbe
V
Caltagirone
C
Cercignani
M
.
2012
.
FMRI resting slow fluctuations correlate with the activity of fast cortico-cortical physiological connections
.
Plos One
 .
7
:
e52660
.
Koch
G
Ponzo
V
Di Lorenzo
F
Caltagirone
C
Veniero
D
.
2013
.
Hebbian and anti-Hebbian spike-timing-dependent plasticity of human cortico-cortical connections
.
J Neurosci
 .
33
:
9725
9733
.
Lemon
RN
.
2008
.
Descending pathways in motor control
.
Annu Rev Neurosci
 .
31
:
195
218
.
Maccabee
PJ
Nagarajan
SS
Amassian
VE
Durand
DM
Szabo
AZ
Ahad
AB
Cracco
RQ
Lai
KS
Eberle
LP
.
1998
.
Influence of pulse sequence, polarity and amplitude on magnetic stimulation of human and porcine peripheral nerve
.
J Physiol
 .
513
(2)
:
571
585
.
Maier
MA
Kirkwood
PA
Brochier
T
Lemon
RN
.
2013
.
Responses of single corticospinal neurons to intracortical stimulation of primary motor and premotor cortex in the anesthetized macaque monkey
.
J Neurophysiol
 .
109
:
2982
2998
.
Moisa
M
Siebner
HR
Pohmann
R
Thielscher
A
.
2012
.
Uncovering a context-specific connectional fingerprint of human dorsal premotor cortex
.
J Neurosci
 .
32
(21)
:
7244
7252
.
Ni
Z
Charab
S
Gunraj
C
Nelson
AJ
Udupa
K
Yeh
IJ
Chen
R
.
2011
.
Transcranial magnetic stimulation in different current directions activates separate cortical circuits
.
J Neurophysiol
 .
105
:
749
756
.
Passingham
RE
Ramnani
N
Rowe
JB
.
2004
.
The motor system
. In:
Frackowiak
RSJ
Friston
KJ
Frith
CD
Dolan
RJ
Price
CJ
Zeki
S
Ashburner
J
Penny
W
, editors.
Human brain function
 .
2nd ed.
Amsterdam
:
Elsevier
p
5
32
.
Peterchev
AV
Goetz
SM
Westin
GG
Luber
B
Lisanby
SH
.
2013
.
Pulse width dependence of motor threshold and input–output curve characterized with controllable pulse parameter transcranial magnetic stimulation
.
Clin Neurophysiol
 .
124
:
1364
1372
.
Rehme
AK
Eickhoff
SB
Grefkes
C
.
2013
.
State-dependent differences between functional and effective connectivity of the human cortical motor system
.
Neuroimage
 .
67
:
237
246
.
Ridding
MC
Ziemann
U
.
2010
.
Determinants of the induction of cortical plasticity by non-invasive brain stimulation in healthy subjects
.
J Physiol
 .
588
:
2291
2304
.
Rossi
S
Hallett
M
Rossini
PM
Pascual-Leone
A
,
Safety of TMSCG
.
2009
.
Safety, ethical considerations, and application guidelines for the use of transcranial magnetic stimulation in clinical practice and research
.
Clin Neurophysiol
 .
120
:
2008
2039
.
Rothwell
JC
.
1997
.
Techniques and mechanisms of action of transcranial stimulation of the human motor cortex
.
J Neurosci Methods
 .
74
:
113
122
.
Sakai
K
Ugawa
Y
Terao
Y
Hanajima
R
Furubayashi
T
Kanazawa
I
.
1997
.
Preferential activation of different I waves by transcranial magnetic stimulation with a figure-of-eight-shaped coil
.
Exp Brain Res
 .
113
:
24
32
.
Shimazu
H
Maier
MA
Cerri
G
Kirkwood
PA
Lemon
RN
.
2004
.
Macaque ventral premotor cortex exerts powerful facilitation of motor cortex outputs to upper limb motoneurons
.
J Neurosci
 .
24
:
1200
1211
.
Veniero
D
Ponzo
V
Koch
G
.
2013
.
Paired associative stimulation enforces the communication between interconnected areas
.
J Neurosci
 .
33
:
13773
13783
.
Werhahn
KJ
Fong
JK
Meyer
BU
Priori
A
Rothwell
JC
Day
BL
Thompson
PD
.
1994
.
The effect of magnetic coil orientation on the latency of surface EMG and single motor unit responses in the first dorsal interosseous muscle
.
EEG Clin Neurophysiol
 .
93
:
138
146
.
zu Eulenburg
P
Caspers
S
Roski
C
Eickhoff
SB
.
2012
.
Meta-analytical definition and functional connectivity of the human vestibular cortex
.
Neuroimage
 .
60
:
162
169
.