Transcallosal generation of phase-aligned beta bursts underlies TMS-induced interhemispheric inhibition Open Access

Beta oscillations, comprising non-rhythmic bursts of 13–30 Hz activity (Rayson et al., 2023; Shin et al., 2017; Szul et al., 2023), are commonly observed in Electroencephalography (EEG) and Magnetoencephalography (MEG) recordings of the primary sensorimotor cortex (SM1; Démas et al., 2020; Neuper et al., 2006; Pfurtscheller, 1981; Salenius & Hari, 2003). Numerous studies have implicated the beta oscillations in the balancing of excitatory and inhibitory processes within the SM1 (Khanna & Carmena, 2017; Sherman et al., 2016). Particularly, a suppression of beta bursts reflects the neuronal activation within a given somatotopic region, and an enhancement of beta bursts reflects the subsequent incoming neuronal inhibition (Heinrichs-Graham et al., 2017; Wessel, 2020). However, the ability of beta oscillations to synchronize across distant cortical areas (Buzsáki & Draguhn, 2004) has led researchers to investigate beta dynamics on a larger scale. This has provided consistent evidence that the beta suppression observed during bodily movement is typically bilateral (McFarland et al., 2000; Pfurtscheller & Lopes da Silva, 1999; van Wijk et al., 2012) and such bilaterality defines a sensorimotor beta-band network (Mary et al., 2017; Sugata et al., 2020). However, whether such large-scale bilateral beta oscillatory dynamics reflect the transcallosal regulation of excitation and inhibition between the two SM1s remains mostly unclear.

One mechanism theorized to realize transcallosal SM1 regulations is interhemispheric inhibition (IHI; Bologna et al., 2012; Duque et al., 2005, 2007; Ferbert et al., 1992; Fleming & Newham, 2017; Giovannelli et al., 2009). IHI is a process through which activation of one brain region selectively suppresses activation of its homologous region in the contralateral hemisphere (Perez & Cohen, 2009). IHI has been widely investigated with Transcranial Magnetic Stimulation (TMS) paradigms such as the ipsilateral Silent Period (iSP). The iSP manifests as a suppression of the Electromyography (EMG) activity of a contracted muscle after a TMS pulse is applied above the ipsilateral primary motor cortex (M1) representation of that muscle (Fling et al., 2013; Fleming & Newham, 2017; Giovannelli et al., 2009; Jung & Ziemann, 2006; Kuo et al., 2017). The TMS pulse inhibits the contralateral SM1 which disrupts its efferent output and results in the silenced contralateral (ipsilateral to the TMS pulse) EMG (Wassermann et al., 1991). Considering the strong implication of beta oscillations in cortical inhibition raises the question whether modulations of left-/right-hemisphere beta oscillations underpin the changes in M1 excitability captured by the TMS-evoked iSP.

Previous research with simultaneous TMS-EEG has shown that the potency of a TMS pulse is contingent on the amplitude, rate, and phase of naturally occurring beta bursts (Hussain et al., 2019). However, beta oscillations are highly variable across individuals (Illman et al., 2022), prompting the question whether this inter-individual variability results in inter-individual differences in IHI captured by the iSP. Furthermore, it is unclear exactly which feature of the beta bursts carries the inhibitory effect. Interestingly, single-pulse TMS above a certain brain region can alter the amplitude, as well as the phase of neuronal oscillations both within that given region (Premoli et al., 2017) and in another anatomically connected brain region (Erickson et al., 2024; Kawasaki et al., 2014). Therefore, it is possible that an iSP-evoking TMS pulse alters contralateral beta oscillations in a manner akin to the phase resetting theorized to underlie neural information transfer (Canavier, 2015; Voloh & Womelsdorf, 2016).

To elucidate the extent to which beta oscillations subtend IHI, we investigate the relationship between bilateral changes in beta amplitude during volitional movement and iSP magnitude. Subsequently, building on findings showing that SM1 beta oscillatory activity propagates to, and can be reliably assessed from peripheral muscle activity (Bourguignon et al., 2017; Bräcklein et al., 2022; Echeverria-Altuna et al., 2022; Mongold et al., 2022), we used the EMG of the contracted muscle during the iSP protocol as a proxy for assessing the effect of the TMS pulse on the contralateral cortical beta oscillations. Lastly, to explore the behavioral relevance of the beta oscillations and the iSP, we investigated their association with bimanual motor dexterity.

Forty (16 female; Mean ± SD age = 24.5 ± 4.4, range 20–49 years) adult volunteers with normal or corrected-to-normal vision and hearing participated in the study. All participants were right handed according to the Edinburgh Handedness Inventory (Oldfield, 1971), did not suffer from neurological or psychiatric disorders, did not take any medication known to alter cerebrocortical excitability, and had no contraindications to TMS. The experimental protocol was approved by the ethics committee of the CUB Hôpital Erasme (CCB, B4062023000234) and the experiment was conducted in accordance with the Declaration of Helsinki. All participants signed an informed consent prior to participation.

The experimental protocol was divided into three parts: an EEG assessment of cortical beta oscillations during volitional hand movement, a TMS assessment of IHI via the iSP protocol, and a behavioral assessment of bimanual dexterity. The three parts of the experiment were conducted in a randomized order within a single session. All participants successfully completed the experiment and no adverse effects of TMS were reported.

Participants’ task was to maintain a stable isometric pinch-grip contraction with the right thumb and index finger against a force transducer (RS Pro, P/N 1004-0.6 kg, Vishay Precision Group, Malvern, PA, USA) at 10 ± 3% of their maximum voluntary contraction (MVC) force. MVC was individually determined for each participant prior to the start of the experiment. Task compliance was ensured with real-time visual feedback prompting participants to increase their force when it subceeded 7% MVC and to decrease their force when it exceeded 13% MVC. While engaged in the isometric contraction with the right hand, once every 3–4 s (random jitter), participants were presented with an auditory tone prompting them to perform a fast closing/opening movement with the fingers of their non-contracted left hand (Fig. 1A). This task was performed for two blocks of 5 min with a short break in between.

Throughout the two blocks, EEG was recorded with 64 Ag/AgCl-sintered electrodes arranged according to an augmented 10–20 system (EEGO Mylab, ANT Neuro, Hengelo, The Netherlands). Electrode impedance was kept blow 20 kΩ. The recording was online referenced to electrode CPz, and sampled at 1 kHz. The signals of a three-axis accelerometer (ADXL335, AnalogDevices, Wilmington, MA, USA) attached to the index finger of the closing/opening left hand were recorded in synchrony with the EEG signals.

Surface EMG (COMETA Pico, COMETA, Milan, Italy) from the contracted right First Dorsal Interosseous (rFDI) muscle was recorded with a sampling frequency of 1 kHz. EMG was recorded in a monopolar configuration, with an “active” electrode placed above the rFDI muscle belly, and a “reference” electrode placed above the first metacarpal bone. The EEG and EMG acquisitions were synchronized via the delivery of digital synchronization triggers.

Participants were seated on a TMS chair and equipped with a neck brace for securing the head into position and minimizing unwanted movements. Surface EMG from the left and right FDI muscles was recorded as in the EEG task.

The hotspot corresponding to the hand area on the right M1 was identified via biphasic single-pulse TMS (MagPro X100, MagVenture, Farum, Denmark) with a hand-held figure-of-eight coil (7.5 cm external loop diameter), pointed 45° away from the midline. Each participant’s Resting Motor Threshold (RMT) was identified. RMT was defined as the minimal TMS intensity that could elicit a motor-evoked potential of at least 50 μV on 5 out of 10 stimulations (Rossini et al., 1994).

During the iSP task, participants held an isometric contraction with their right hand while completely relaxing their left hand. The parameters of the contraction and the visual feedback were identical to the ones during the EEG task, therefore, allowing for direct comparison of right FDI EMGs between the two tasks. While participants were holding the contraction, once every 5–6 s (random jitter) a TMS pulse was delivered to the previously identified hotspot with an intensity of 130% RMT (Fig 1B). The procedure was performed for 2 blocks of 4.5 min, resulting in a total of 100 iSPs per participant.

Participants completed the Purdue Pegboard Test (PPT; Tiffin & Asher, 1948). The PPT assesses the fine motor performance of the left hand (PPT_LH), right hand (PPT_RH), both hands in synchrony (PPT_BH), and both hands in alternation (PPT_Assembly; Fig. 1C). Before each subtest, the task was explained to the participants and they were allowed one practice trial. All scores were computed following the PPT guidelines (Tiffin & Asher, 1948). Two additional bimanual coordination scores were computed. The first coordination score (PPT_C1) was the PPT_Assembly divided by the geometric mean of the PPT_LH and PPT_RH. The second PPT coordination index (PPT_C2) was the PPT_Assembly divided by the PPT_BH.

EEG data were preprocessed offline with a custom MATLAB (Mathworks, Natick, MA) script and the FieldTrip toolbox (Oostenveld et al., 2011). Bad channels were identified following the guidelines of Bigdely-Shamlo et al. (2015) and topographically interpolated. EEG signals were re-referenced to a common average, filtered through 0.3–45 Hz, and Independent Components Analysis (Vigário et al., 2000) was used for further artifact suppression. Twenty independent components were evaluated from the data with a FastICA algorithm with dimension reduction, 25; nonlinearity, tanh (Hyvärinen & Oja, 2000; Vigário et al., 2000), and independent components corresponding to eye-blink, eye movement, and heartbeat artifacts were visually identified and removed. On average, four (SD = 1.3) independent components per participant were removed.

EEG signals were filtered through 5-Hz-wide frequency bands centered on 5–40 Hz by steps of 1 Hz. The bandpass filter used to this aim was designed in the frequency domain with zero-phase and 1-Hz-wide squared-sine transitions from 0 to 1 and 1 to 0 (e.g., the filter centered on 20 Hz rose from 0 at 17 Hz to 1 at 18 Hz and ebbed from 1 at 22 Hz to 0 at 23 Hz). This filtering approach is standardly applied to electrophysiological data for resolute time–frequency decompositions of the signals (e.g., Bourguignon et al., 2017; Mongold et al., 2022). Then, EEG signals were smoothly set to 0 (squared-cosine transition of 1 s) at timings 1 s around artifacts (time points where full-band EEG amplitude in at least one electrode exceeded 5 SDs above the mean), and to avoid edge effects, further analyses were based on time points at least 2 s away from artifacts and appearance of force visual feedback. Subband signals’ envelope was then extracted with the Hilbert transform and downsampled to 20 Hz. To avoid contamination by very slow envelope variations, these subband envelopes were divided by their low-pass filtered version at 0.1 Hz (squared-sine transition from 0.05 to 0.15 Hz). Such normalization ensures that envelopes fluctuate about the value 1, and are blind to changes slower than 0.1 Hz that cannot be ascribed to the task. As a further normalization, the envelopes were centered on 0 by subtracting the value 1, and converted to percentage of change by multiplying by 100%.

Subband envelopes were segmented into epochs from –1000 to 2500 ms relative to left-hand movement onset as identified in accelerometer recordings. Epochs which contained time points less than 2 s away from artifacts, or in which participants failed to perform a movement, or failed to maintain the force of the right hand at 10 ± 3% MVC were excluded from the analysis. The remaining epochs (113 ± 25, Mean ± SD across participants) were subsequently averaged across trials, giving rise to a time–frequency map of amplitude modulation for each EEG channel and each participant. Further analyses were focused on preselected electrodes spanning the left (C1, C3, C5, CP1, CP3, CP5, FC1, FC3, FC5) and right (C2, C4, C6, CP2, CP4, CP6, FC2, FC4, FC6) SM1s. We further identified within each SM1 electrode preselection the electrode with the maximal amplitude suppression in a preselected time–frequency window ranging from –500 to 1000 ms relative to left-hand movement onset and from 13 to 30 Hz (β_depth). Within the corresponding electrode, we further extracted the size of the cluster of significant beta suppression in the previously defined time–frequency window (β_size; see subsection Statistical Analyses). Lastly, to highlight left-/right-hemisphere imbalance in beta modulations, contrasts between the left and right indices of beta suppression were computed as their difference (left minus right) divided by their sum, leading to Δβ_size and Δβ_depth. These contrasts are smaller than 0 when beta amplitude is more suppressed in the right hemisphere (contralateral to the moving hand) compared with the left hemisphere (ipsilateral to the moving hand) and vice versa.

The EMG data from the contracted rFDI during the iSP protocol were preprocessed offline with a custom MATLAB script. Data were filtered between 20 and 295 Hz, followed by notch filters at 50 Hz and its harmonics, and rectified. The EMG was segmented into epochs from –100 to 200 ms relative to TMS pulse onset. The epochs were averaged within participants and normalized by their baseline mean (taken from 50 to 5 ms prior to the TMS pulse), and used for iSP quantification. The iSP onset and offset were identified with Mean Consecutive Difference thresholding (Garvey et al., 2001; Hupfeld et al., 2020) and used for computing iSP latency (the time between the TMS pulse and the iSP onset), iSP duration (the time between the iSP onset and offset), and normalized iSP area (the integral of the normalized rectified EMG signal between the iSP onset and offset; Hupfeld et al., 2020).

To further characterize the beta modulations during the voluntary movements and the TMS-iSP protocol, we used the unrectified rFDI EMG of the contracted right hand during each task as a proxy for the changes occurring in the contralateral SM1 (Bourguignon et al., 2017; Bräcklein et al., 2022; Echeverria-Altuna et al., 2022; Mongold et al., 2022). EMG signals were filtered in subbands for envelope extraction, normalized by their <0.1 Hz trend, segmented, and averaged as previously done with the EEG signals. In addition, to separate the contribution of evoked and induced responses, the same analysis was also conducted on EMG signals from which the average response was subtracted.

We further characterized the bursts underlying the beta modulations identified in the unrectified rFDI EMG during the volitional left-hand movements and the TMS stimulations (Rayson et al., 2023; Szul et al., 2023). To this end, individual beta bursts in the EMG data from each task were identified following the methodology of Tinkhauser et al. (2017) and O’Keeffe et al. (2020). Beta envelopes were obtained using the Hilbert transformation applied to the unrectified EMG signals filtered through 13–30 Hz. This filtering approach is commonly applied to electrophysiological data in preparation for beta burst extraction (e.g., O’Keeffe et al., 2020). The EMG envelopes were further normalized by their <0.1 Hz trend. We then identified, for each participant and condition separately, time intervals where the EMG envelope exceeded its 75^th percentile. Such intervals were considered as beta bursts when their duration was greater than 50 ms. The time of occurrence of each beta burst was defined as the middle of the corresponding time interval. The probability of beta burst occurrence and the average burst duration were computed within 200-ms windows spaced by 100 ms and centered on -900 to 2400 ms relative to movement or TMS onset.

Lastly, having identified TMS-evoked beta bursts in the proxy signal for the left SM1, we further investigated whether TMS-evoked iSP relates to a phase resetting of these beta bursts. Thus, the unrectified EMG was filtered as done for the beta burst analyses. Then, the EMG data were segmented into epochs from –200 to 500 ms relative to TMS pulse onset. Epochs were normalized by their root-mean-square amplitude (taken from 50 ms to 5 ms prior to the TMS pulse) and averaged across trials. Hilbert transformation was applied to the averaged EMG trace of each participant. Then, we determined whether Hilbert coefficients at time points from 0 to 200 ms by step of 10 ms tended to show a consistent phase across participants. Due to the uncertainty regarding EMG polarity (which is dependent on the choice of the active and reference electrodes that was not standardized across participants), phase consistency was assessed using an approach based on singular value decomposition (SVD). More precisely, for each time point, SVD was applied to all participants’ Hilbert coefficients arranged as a matrix with the real part of all participants’ coefficients stacked as the first column and the imaginary part as the second column. In this setup, the SVD identified potential Hilbert coefficient alignment in the complex plane, corresponding to a phase alignment across participants of the EMG beta oscillations after the TMS pulse. The first singular value of the decomposition was used to quantify the strength of the phase alignment across participants.

Non-parametric cluster-based tests were applied to each participant’s EEG data to assess the significance of β_size in the time–frequency window ranging from –500 to 1000 ms relative to left-hand movement onset and from 13 to 30 Hz separately for the left (ipsilateral) and right (contralateral) hemisphere electrodes. In brief, 10,000 surrogate amplitude modulation maps were built based on EEG subband envelopes segmented with respect to random onsets separated by 3–4 s. These surrogate maps served to identify β_size, for the genuine and each surrogate data, which was the size of the largest cluster of time–frequency bins (bin size: 50 ms × 1 Hz) displaying a larger amplitude suppression or enhancement than their percentile 95 across all maps. The p-value for β_size was obtained as the proportion of surrogate β_size values that were higher than the observed genuine value.

Analogous non-parametric tests were applied to the time–frequency maps of EMG amplitude.

The statistical significance of changes in beta burst properties (probability of occurrence and duration) in each time interval from 0 to 500 ms relative to movement or TMS onset was assessed with a t-test in comparison with baseline values (mean from –900 to –500 ms). Correction for multiple comparisons was performed using Bonferroni.

Surrogate data-based statistics were applied to evaluate the significance of the TMS-evoked phase alignment in the beta band of the right FDI EMG. To do so, the strength of the phase alignment across participants for each time point after the TMS pulse was compared with a null distribution obtained using the Hilbert coefficients with a phase randomly selected in the range [-π, π] (10,000 repetitions) while preserving the amplitude. The p-value was obtained as the proportion of values in the null distribution that were higher than the genuine singular value.

Redundancy analysis (RDA) was used to investigate the associations between all constructs of interest. RDA is a multivariate technique for modeling the effect of a linear combination of a set of related predictors on the linear combinations of a set of related outcomes (van den Wollenberg, 1977). In two RDA models, we assessed (i) the predictive power of all indices of left-/right-hemisphere beta suppression imbalance on all indices of iSP and (ii) the predictive power of all indices of left-/right-hemisphere beta imbalance and iSP together on all indices of bimanual motor dexterity. Statistical significance of each model was assessed with RDA Monte Carlo permutation tests with 10,000 permutations.

The RDA was followed up with pairwise Spearman correlations between all variables of interest.

The EEG analyses identified a large-scale suppression of the beta amplitude following left-hand movement (Fig. 2). More quantitatively, in the time interval from –500 to 1000 ms, β_size was significant in 35 (87.5%) participants (all p < 0.05) in the contralateral (right) hemisphere, and in 34 (85%) participants (all p < 0.05) in the ipsilateral (left) hemisphere. Across participants, the contralateral β_size was 100.5 ± 85.9 (Mean ± SD) and the ipsilateral β_size was 84.4 ± 73.3 time–frequency bins, with the contralateral β_size being significantly greater than the ipsilateral β_size (t₃₉ = 2.4, p = 0.024, d = 0.4). Likewise, across participants, the contralateral β_depth was 26.1 ± 6.2% suppression and the ipsilateral β_depth was 24.1 ± 4.8% suppression. The contralateral β_depth was also significantly greater than the ipsilateral β_depth (t₃₉ = 4.4, p < 0.0001, d = 0.7).

The RMT of our sample was 50.0 ± 8.6% of the maximal stimulation intensity of the TMS device.

An iSP was identified in all participants (Fig. 3), in the form of transiently decreased muscle activity ~50 ms after TMS stimulation, characterized by a latency of 39.3 ± 6.2 ms, a duration of 25.1 ± 7.6 ms, and a normalized area of 7.1 ± 3.5 mV ms. Of note, iSPs were closely followed by two successive peaks of increased muscle activity up until ~150 ms.

Figure 4A presents the group-averaged time–frequency map of the amplitude of the rFDI EMG signal in the left-hand movement task. It clearly highlights a large-scale suppression of the beta amplitude corresponding to the onset of the left-hand movement, reflecting that observed in the left SM1 (Fig. 2). In the time interval from –500 to 1000 ms, EMG β_size was 53.2 ± 60.9 time–frequency bins and was statistically significant in 29 (72.5%) participants (all p < 0.05).

Figure 4B presents the group-averaged time–frequency map of the amplitude of the rFDI EMG signal in the iSP task. Contrasting with Figure 4A, it shows a large-scale enhancement of spectral amplitude at frequencies up to 30 Hz after TMS pulse, along with a discernible maximum at beta frequencies localized between 50 and 200 ms. In the time interval from –500 to 1000 ms, EMG β_size was 42.1 ± 31.0 time–frequency bins and was statistically significant in 32 (80%) participants (all p < 0.05).

Figure 4C presents the same map obtained after subtracting each subject’s averaged unrectified EMG response, thereby focusing only on the induced response. Removal of the averaged EMG resulted in a broad-band attenuation of amplitude in the 5–30 Hz range, showing that a TMS-evoked response contributed largely to the initial amplitude modulation. The TMS-induced enhancement of amplitude was restricted to the low-beta frequency range. In the time interval from –500 to 1000 ms, this induced EMG β_size was 27.48 ± 25.5 time–frequency bins, which was statistically significant in 25 (63%) participants (all p < 0.05).

Figure 5 presents the evolution of beta burst characteristics (probability of occurrence and duration) as derived from the right FDI EMG relative to movement onset or TMS stimulation. Compared with baseline, the probability of observing a beta burst was significantly reduced in the time windows centered from 0 to 100 ms relative to hand movement (peak at the 0 ms time window, t₃₉ = 3.13, p_uncorrected = 0.0033, p_corrected = 0.0197) and significantly increased in the time windows centered from 0 to 200 ms relative to an ipsilateral TMS pulse (peak at the 0 ms time window, t₃₉ = -9.69, p_uncorrected < 0.00001, p_corrected < 0.00001; Fig. 5A). The average beta burst duration was not significantly modulated during the hand movement task (p_uncorrected > 0.05), and was significantly increased in the time windows centered from 100 to 200 ms relative to the ipsilateral TMS pulse (peak at the 100 ms time window, t₃₉ = -4.67, p_uncorrected = 0.00004, p_corrected = 0.0002; Fig. 5B). Therefore, the large-scale amplitude suppression during movement execution (Fig. 4A) reflects the suppression of beta bursts, and the large-scale amplitude enhancement following an iSP-inducing TMS stimulation (Fig. 4B, C) reflects TMS-evoked beta bursts.

For investigating the identified TMS-evoked beta bursts further, Figure 6 presents the time course of the averaged normalized non-rectified EMG traces filtered through 13–30 Hz in the TMS-iSP protocol. It highlights the presence of a beta oscillation starting from 20 to 30 ms after the TMS pulse. Importantly, the first peak of this oscillation appeared consistently at 40 ms across participants. And indeed, the phase consistency analysis revealed a significant phase alignment of Hilbert coefficients in the time interval 10 to 70 ms following the onset of the TMS pulse (all p < 0.05), with the beta phase alignment reaching maximal significance at 40 ms after the TMS pulse (p = 0.0003; Fig. 6). This demonstrates that 10 to 70 ms after the TMS pulse, ipsilateral EMG beta oscillations align in phase and then diverge over time, possibly due to interindividual variability in the shape of the beta bursts.

To ensure that the observed TMS-evoked beta burst and its phase alignment across participants were not simply a consequence of narrow band-pass filtering, we repeated our analysis on the EMG signals filtered between 31 and 48 Hz. This equally wide frequency band on the neighboring gamma range did not show a significant phase alignment of Hilbert coefficients in the time interval 10 to 70 ms following the onset of the TMS pulse (0.069 ≤ p ≤ 0.99). This suggests that the observed results in the beta band are not simply a consequence of the applied methodology.

Participants scored 15 ± 2 (range 11–19) on PPT_LH, 16 ± 2 (range 10–22) on PPT_RH, 12 ± 2 (range 7–16) on PPT_BH, and 37 ± 7 (range 22–51) on PPT_Assembly. The scores on all PPT subtests are consistent with the findings of other investigators and with PPT population norms from healthy adults (Gu et al., 2023; Yeudall et al., 1986), therefore, indicating that our participants completed the PPT adequately.

With respect to the additional bimanual coordination scores, participants scored 2.3 ± 0.4 (range 1.4–3.1) on PPT_C1 (PPT_Assembly/(√PPT_LH × √PPT_RH) and 3.1 ± 0.6 (range 2.0–5.9) on PPT_C2 (PPT_Assembly / PPT_BH). Given that twice as much time is allowed for the assembly task, the fact that PPT_C1 values were above 2 on average indicates that performance with the two hands working in coordination was better than independent performance with each hand individually. Besides, given that PPT_BH indicates a number of peg pairs while PPT_Assembly accounts for each element, a PPT_C2 below 4 indicates that performance with the two hands working in coordination was worse than performance with the two hands working in synchrony.

The RDA aimed at assessing the associations between the indices of left-/right-hemisphere beta imbalance (Δβ_size and Δβ_depth), iSP (latency, duration, and normalized area), and bimanual motor performance (PPT_BH, PPT_Assembly, PPT_C1, and PPT_C2) revealed that, collectively, the beta modulation indices did not explain a significant proportion of variance of the iSP indices (4.4%, F_2,37 = 0.85, p = 0.48). It also revealed that, collectively, the beta modulation and iSP indices did not explain a significant proportion of variance of bimanual dexterity (12.7%, F_5,34 = 0.99, p = 0.44).

The results of a follow-up univariate correlation analysis between all variables are presented in Figure 7. As expected, the contrasts of ipsilateral/contralateral beta imbalance were significantly correlated with each other (r = 0.45, p = 0.004). The iSP indices were also significantly correlated with each other (0.46 ≤ | r | ≤ 0.73, all p < 0.05) in the respective expected direction. However, no significant correlation between the Δβ_size and Δβ_depth and measures of the iSP was found (0.26 ≤ p ≤ 0.57) and, with the exception of a moderate significant correlation between the Δβ_size and the PPT_C1 (r = 0.37, p = 0.018), no index of left/right beta imbalance or of iSP significantly correlated with any measure of bimanual motor performance.

Our study aimed at elucidating the link between the movement-related bilateral modulations of SM1 beta oscillations and TMS-induced IHI processes, and investigating the behavioral relevance of the two neurophysiological phenomena. We did not find a significant association across participants in the left-/right-hemisphere imbalance in SM1 beta suppression during volitional hand movement, the iSP, and bimanual dexterity. However, we found that TMS-iSP is contingent on the rapid generation of a phase-aligned beta burst in the EMG proxy signal of the brain hemisphere contralateral to the stimulation, a process exactly opposite in nature to the bilateral beta burst suppression observed during volitional movement.

Using the FDI EMG of the contracted right hand as a proxy for the changes occurring in the contralateral SM1 (Bourguignon et al., 2017; Bräcklein et al., 2022; Echeverria-Altuna et al., 2022; Mongold et al., 2022), we demonstrated a striking divergence in beta SM1 oscillatory dynamics during volitional hand movement and during TMS-evoked iSP. Indeed, ipsilateral SM1 beta bursting activity was suppressed following a hand movement, consistent with an extensive literature (Démas et al., 2020; Kilavik et al., 2013; McFarland et al., 2000; Pfurtscheller & Lopes da Silva, 1999; van Wijk et al., 2012; Zaepffel et al., 2013), while delivering a TMS pulse to SM1 resulted in a significant enhancement of contralateral beta activity attributable to the generation of beta bursts. In addition, more pronounced ipsilateral beta amplitude suppression (relative to contralateral) was not significantly associated with classical iSP parameters. Therefore, the ubiquitous beta rhythm suppression—or decrease in beta burst occurrence—in the hemisphere ipsilateral to hand movement appears to receive minimal contribution from the IHI regulatory processes captured by the iSP. Even if naturally occurring IHI processes during volitional movement would encompass a contralateral beta burst generation of the type highlighted in our iSP results, the presence of such a hypothetical burst would be obscured by the co-occurring large-scale beta suppression. In any case, the absence of significant association between measures of movement-related beta modulation and TMS-evoked iSP suggests that these two phenomena reflect two distinct types of interhemispheric interactions occurring at different time scales.

The distinct nature of the beta rhythm changes during movement compared with TMS-iSP can be explained by the vastly distinct state of the motor system during each task (Fuggetta et al., 2005). The suppression in the EEG beta oscillations during volitional movements results from goal-directed action planning, followed by motor command formulation, initiation, and constant refinement, as orchestrated by a distributed prefrontal (Andersen & Cui, 2009), premotor (Halsband et al., 1993), SM1, and parietal (Riemann & Lephart, 2002) system. However, the IHI processes captured by iSP are manifestations of a very rapid and strong excitation of an SM1 hotspot in the absence of any volition to move. Such excitation does not have the crucial role of precisely executing a desired movement and, therefore, engages the distributed sensorimotor system in a vastly distinct way. Therefore, in the case of TMS-iSP, the TMS-evoked beta burst in the contralateral hemisphere causes a fleeting inhibitory process which interacts with the natural brain oscillations, yet lacks the functional relevance of the amplitude modulations of the beta rhythm during movement execution. Overall, these considerations suggest that TMS protocols aimed at assessing IHI indeed capture an inhibitory mechanisms across the two SM1s, however, this mechanism is only a restricted part of the complex neurophysiology of the interhemispheric signaling which realizes—and assures the precision of—volitional everyday movement.

We found that the beta enhancement observed in the proxy signal for the SM1 contralateral to that stimulated with TMS coincides with a consistent phase alignment of beta bursts both within and across participants. This alignment was most evident ~40 ms after the TMS pulse in the proxy rFDI EMG signal, but would likely occur ~20 ms after TMS in contralateral SM1 given the well-documented ~20-ms propagation delay from SM1 to hand EMG (Ibáñez et al., 2021). The beta phase alignment indicates that a phase-resetting mechanism underlies the iSP. Phase resetting effects of TMS have previously been reported for naturally occurring theta, alpha, and beta oscillatory activity within the stimulated SM1 (Kawasaki et al., 2014; Premoli et al., 2017), but not in the contralateral SM1 and not in relation to IHI. Expanding upon the perspective whereby beta oscillations consist of short-lived bursts (Bonaiuto et al., 2021; Enz et al., 2021; Shin et al., 2017; Szul et al., 2023), which have an inhibitory effect on the motor system, our data suggest that it is the phase of the initial peak of the burst that carries the inhibitory effect. This is evidenced by the near-exact co-occurrence of the initial peak of the beta bursts (Fig. 6) and the onset of the iSP (Fig. 3) at ~40 ms after the TMS pulse. Although generated by TMS, the identified beta bursts are comparable in duration (~160 ms; Fig. 6) with naturally occurring beta bursts documented with MEG (~150 ms reported by Law et al., 2022; ~167 ms reported by Shin et al., 2017; and ~180 ms reported by Szul et al., 2023) and with Electrocorticography (between ~150 and ~200 ms reported by West et al., 2023); however, precise comparison is made difficult by the heterogeneity in methods used to detect bursts and criteria used to define burst onset and offset. Still an interesting difference in the timing of the behavioral inhibitory effect between our TMS-generated bursts and naturally occurring bursts emerges, with our bursts having an inhibitory effect only between ~40 and 60 ms after their onset and naturally occurring bursts having a much longer inhibitory effect on behavior lasting for 200–400 ms after their onset (Law et al., 2022; Shin et al., 2017). These differences in behavioral inhibition can potentially be accounted for by different burst properties as well as different task demands. First, there is evidence that naturally occurring beta bursts, although quite diverse in shape, can be classified into motifs and that different shape motifs can potentially result in different inhibitory patterns (Rayson et al., 2023; Szul et al., 2023). The TMS-generated bursts could form a shape motif that is distinct from the motifs of naturally occurring beta bursts and thus likewise result in a different inhibition at the behavioral level. Tentative support for this comes from the noticeable difference in peak shape between our TMS-generated bursts (Fig.` 6) and the Wavelet-like naturally occurring beta bursts (Bonaiuto et al., 2021; Rayson et al., 2023); however, this is to be verified by future TMS-neuroimaging studies with unified methodology for extracting beta burst properties. Furthermore, although the exact neurophysiology of TMS-IHI remains unknown (Ziemann, 2013), a difference in beta shape motifs could result from the TMS procedure recruiting GABA-ergic inhibitory mechanisms distinct from the GABA-ergic thalamo-cortical and within-cortical signalling believed to generate naturally occurring beta bursts (Bonaiuto et al., 2021; Law et al., 2022) and thus have a different inhibitory outcome. Second, the behavioral inhibitory effect of beta bursts can differ between purely motor tasks—like the constant muscle contraction applied in our experiment—and tasks that involve both cognitive and motor components—like the reaction time task applied in other investigations. Accordingly, in the latter it could be the case that beta bursts have a comparable transient ~40 to 60 ms inhibitory effect on cognitive processing, but that this keeps impacting motor behavior for a longer time.

Therefore, the TMS-iSP protocol we present, able to generate contralateral beta bursts, could serve future investigations of the nature of the beta oscillations by comparing the properties of the TMS-generated burst with those of naturally occurring bursts recorded with EEG/MEG as well as to the predictions of computational models of neural oscillations (Law et al., 2022; West et al., 2023). Additionally, our results provide a promising avenue for developing novel neuromodulation protocols. Those are particularly important considering the often documented abnormalities of beta oscillations in disorders such as cerebral palsy (Démas et al., 2020), multiple sclerosis (Arpin et al., 2017), stroke (Shiner et al., 2015), and Alzheimer’s diseases (de Haan et al., 2008), in all of which a pattern of pathologically suppressed beta burst amplitude is associated with worse clinical outcomes. Therefore, the TMS-iSP protocol could be used in a biofeedback setup for non-invasive modulatory enhancement of beta oscillatory activity, which could lead to therapeutic effects.

Since in our study we assessed cortical beta oscillations, iSP, and bimanual dexterity non-simultaneously, we could only capitalize on inter-individual variability to identify brain–behavioral relationships The absence of significant association of beta suppression and iSP magnitudes with bimanual dexterity appears puzzling as the PPT is a valid test specifically designed to assess bimanual motor coordination (Lindstrom-Hazel & VanderVlies Veenstra, 2015) and successfully applied in numerous contemporary investigations (e.g., Gu et al., 2023; Hidese et al., 2018; Hinkle & Pontone., 2021; Mongold et al., 2024). Moreover, our participants achieved scores consistent with the normative data and with the results of other investigators (Gu et al., 2023), while displaying a substantial amount of performance variability that could have been accounted for by the differences in bilateral beta oscillatory activity or the iSP. However, previous research has found behavioral relevance of bilateral beta modulation and iSP predominantly with in-task assessments of motor performance (Kuo et al., 2019; Murase et al., 2004; Morishita et al., 2022; Sugata et al., 2020), while investigations quantifying motor performance with separately performed tests such as the PPT paint a less clear picture (Gu et al., 2023). Future studies should clarify the task specificity of the behavioral relevance of beta oscillations and IHI. If beta oscillations and IHI truly are the crucial motor system regulatory mechanisms they are theorized to be (Perez & Cohen, 2009; Engel & Fries, 2010), their role for motor behavior should likely extend beyond in-task effects.

Our investigation focused on a sample of right-handed participants and investigated IHI from the non-dominant towards the dominant hemisphere. However, there is evidence that left-handed participants have more efficient interhemispheric communication (Cherbuin & Brinkman, 2006, but see Raaf & Westerhausen, 2023, for a cautionary note on the matter), and that IHI tends to be stronger from the dominant towards the non-dominant hemisphere (Bäumer et al., 2007; Duque et al., 2007). Therefore, the results might have been more salient (i.e., the transcallosally generated beta bursts would be larger) if the TMS pulse had been applied over the dominant hemisphere. This should be verified in future studies applying our protocol in a mixed sample of left- and right-handed participants, while also manipulating the laterality of the TMS pulse.

Our study aimed at clarifying the beta oscillatory dynamics underlying TMS IHI, using a paradigm in which beta oscillations and IHI were assessed non-simultaneously. We found that the large-scale cortical beta amplitude suppression observed during bodily movement was not significantly associated with IHI assessed with the iSP. Moreover, we show that the iSP is contingent on the rapid generation of a high-amplitude phase-aligned beta burst in the hemisphere contralateral to the stimulation. This process is opposite in nature to the ipsilateral beta burst suppression observed during movement. Therefore, the phenomena observed in TMS-IHI studies may not generalize to natural motor control; however, the TMS iSP protocol offers a means to evoke and analyze beta bursts, as well as to support neuroimaging and computational modeling studies.

All data and code are publicly available on the Open Science Framework (https://osf.io/d5m74/).

Christian Georgiev: Conceptualization, Investigation, Formal analysis, Visualization, Writing—original draft. Scott J. Mongold: Conceptualization, Investigation, Writing—review & editing. Pierre Cabaraux: Investigation, Writing—review & editing. Gilles Naeije: Writing—review & editing. Julie Duque: Writing—review & editing. Mathieu Bourguignon: Conceptualization, Formal analysis, Visualization, Data curation, Supervision, Writing—review & editing.

Christian Georgiev was supported by an Aspirant Research Fellowship awarded by the Fonds de la Recherche Scientifique (F.R.S.-FNRS; Brussels, Belgium; grant 1.A.211.24F). Scott Mongold was supported by an Aspirant Research Fellowship awarded by the F.R.S.-FNRS (Brussels, Belgium; grant FC 46249). Pierre Cabaraux was supported by a Clinical Researcher Fellowship awarded by the F.R.S.-FNRS (Brussels, Belgium; grant 40024164). Gilles Naeije is postdoctorate Clinical Master Specialist at the FRS-FNRS (Brussels, Belgium). The project was supported by grants of the Fonds de la Recherche Scientifique (F.R.S.-FNRS, Brussels, Belgium; grant MIS F.4504.21), and of the Brussels-Wallonia Federation (Collective Research Initiatives grant) awarded to Mathieu Bourguignon. This publication was supported by the Walloon Region as part of the FRFS-WELBIO strategic axis.

The experimental protocol was approved by the ethics committee of the CUB Hôpital Erasme and all participants signed a detailed informed consent prior to participation.

The authors have no conflicts of interests to disclose.