Abstract
Voluntary movement control depends on plasticity in several interconnected brain regions, including the cerebellum (CB), primary motor cortex (M1), and posterior parietal cortex (PPC). It is thought that one role of the CB is to regulate communication between PPC and M1, but causal evidence for this regulatory role in humans remains lacking. Here, we evaluated how transiently altering activity in CB via intermittent theta burst stimulation (iTBS) affects PPC–M1 connectivity and plasticity by assessing the effectiveness of subsequent Hebbian-like cortical paired associative stimulation (cPAS) to PPC and M1. Using a within-subject design, we administered four different single-session stimulation conditions to the CB and parietal–motor pathway of the motor network and measured the aftereffects on plasticity (both sexes). We administered iTBS to the right CB or right visual cortex, followed by cPAS of a parietal–motor circuit in the left hemisphere. In a subset of participants, we performed two additional control conditions to assess the effect of CB iTBS alone and Hebbian-like cPAS of the PPC–M1 circuit alone. We evaluated motor-evoked potentials (MEPs) using single-pulse transcranial magnetic stimulation as a measure of motor cortical excitability before and after each plasticity induction protocol. Cerebellar iTBS reduced cPAS-induced plasticity in the parietal–motor circuit, as evidenced by a decrease in MEPs. These responses were selective, as no decreases in excitability were observed during the control experiments. These findings suggest that CB activity can modify distal neural activity in a network-connected parietal–motor circuit through heterosynaptic metaplasticity.
Significance Statement
Modulating plasticity in the motor system offers a promising approach to improving motor function. This study examines how the cerebellum (CB) influences plasticity in a brain network that controls movement. By combining noninvasive brain stimulation aimed at the CB with a multifocal technique designed to strengthen brain connections, we found that cerebellar input reduces the usual increase in plasticity between the parietal and motor regions. These findings highlight the CB's pivotal role in regulating brain activity across distant regions, providing new insights into its influence on motor control. This research highlights the potential of using noninvasive brain stimulation to target the CB as a therapeutic tool for correcting dysfunctional brain interactions in motor impairments.
Introduction
The control of goal-directed movements involves the coordinated activity between the cerebellum (CB) and various cortical regions, including the primary motor cortex (M1), posterior parietal cortex (PPC), premotor cortex, and prefrontal cortex (Shadmehr and Krakauer, 2008; Shadmehr et al., 2010; Manto et al., 2012; Caligiore et al., 2016b; Manto et al., 2022). The CB plays a critical role in motor control, including motor planning, fine motor coordination, and visuomotor adaptation (Wolpert et al., 1998; Bastian, 2006; Lang et al., 2017; Tzvi et al., 2021; Manto et al., 2022). However, evidence also suggests that a key function of the CB is to regulate communication between cortical areas crucial for accurate movement (D. Popa et al., 2013; Lindeman et al., 2021; McAfee et al., 2022).
Repetitive transcranial magnetic stimulation (rTMS) can be used to modulate cortical excitability and connectivity, inducing lasting local and remote effects across brain regions (Siebner et al., 2022). This makes rTMS an ideal tool for investigating cortical plasticity at the systems level (Fox et al., 2012; Goldenkoff et al., 2020). Using plasticity-inducing rTMS to influence cerebellar activity offers insights into its role in shaping plasticity within cortical networks controlling skilled motor actions (Shadmehr and Krakauer, 2008; Caligiore et al., 2016b; Buch et al., 2017). For example, several human studies suggest that plasticity in corticocortical communication depends on cerebellar activity, but they lack the spatial resolution (Casula et al., 2016; Koch et al., 2018) or causal methods (Halko et al., 2014; Rastogi et al., 2017) needed to conclusively demonstrate its role. While we have previously shown that the parietal–motor cortical pathway is subject to metaplasticity, with its expression dynamically modulated by prior excitability levels (Goldenkoff et al., 2024), it remains unclear whether such plasticity is influenced by a distant brain area within the motor network.
In the current study, we tested this hypothesized mechanism with a direct measure of causality that determined whether manipulating cerebellar activity with stimulation could exert a neuromodulatory effect on distant activity in a connected parietal–motor pathway involved in skilled hand action (Koch et al., 2007, 2008c; Ziluk et al., 2010; Vesia and Davare, 2011; Karabanov et al., 2013; Vesia et al., 2013, 2017; Lafleur et al., 2016; Schintu et al., 2016; Koch, 2020; Malderen et al., 2022). To achieve this, we employed a modulate-and-measure approach to assess a targeted downstream parietal–motor plasticity response to cerebellar stimulation. We used two rTMS methods to induce long-term potentiation (LTP)-like effects. We first used intermittent theta burst stimulation (iTBS; Huang et al., 2005) over the CB. We then applied cortical paired associative stimulation (cPAS) by using two coils simultaneously over two different sites to modulate inter-regional coupling in a parietal–motor circuit associated with goal-directed motor actions (Hebb, 1949; Markram et al., 2011; Koch et al., 2013; Chao et al., 2015; Goldenkoff et al., 2020, 2024). Using a within-subject design, we administered four different single-session stimulation conditions to the CB and parietal–motor pathway of the motor network and measured the aftereffects on plasticity. We measured motor-evoked potentials (MEPs) using single-pulse TMS as a readout of motor cortical excitability before and after each plasticity induction protocol (Chen, 2000; Chen and Udupa, 2009). Given induction of cerebellar plasticity by means of iTBS can modulate the neural activity of the interconnected parietal–frontal network in the contralateral hemisphere (Casula et al., 2016; Koch et al., 2018), we hypothesized that applying iTBS to CB before cPAS would prevent the Hebbian plasticity effects of cPAS-induced potentiation by repeated activation of the short-latency connection between PPC and M1.
Materials and Methods
Participants
Fourteen adults (10 females, ages 19–23 years) participated in the study after providing written, informed consent. Participants were screened for contraindications to MRI and TMS using standard MRI and TMS safety questionnaires (Keel et al., 2001; Rossi et al., 2011). All procedures were approved by the University of Michigan Institutional Review Board (HUM00157197/ HUM00129313) in accordance with the Declaration of Helsinki.
Structural MRI data acquisition
Whole-brain T1–weighted scans were acquired for all participants to define individualized TMS targets. MR data were acquired with a 3 T GE scanner (MR 750) with a 32-channel head coil. Functional data for resting-state (rs) scans were obtained using a one-shot multiband T2*-weighted echoplanar imaging (EPI) sequence sensitive to blood oxygenation level-dependent contrast (TR, 1,200 msec; TE, 30 msec; flip angle, 70°; 21 cm field of view; in-plane resolution, 2.4 × 2.4 mm; MB acceleration, 3). Each functional volume contained 51 contiguous 2.5-mm-thick axial slices separated by a 0 mm interslice gap acquired in an interleaved manner.
rs-fMRI data and preprocessing
rs-fMRI data were preprocessed through a pipeline to correct artifacts, minimize physiological noise, and standardize and align the data for analysis. This pipeline involved slice time correcting the data to reflect the same temporal instance and realigning all functional EPI volumes to the one with the minimum outlier fraction to correct for motion. We then normalized the data to the MNI152 template using a nonlinear warp. The RETROICOR method (Jo et al., 2010) was used to regress noise from physiological signals such as heart rate and respiration. We then shrank large spikes in the time series using the AFNI's 3dDespike tool. We then performed nuisance regression of head motion (three translation and three rotation parameters along with their temporal derivatives), cerebrospinal fluid, and white matter signals (Dijk et al., 2010; Jo et al., 2010). Regressors associated with large motion-related spikes (framewise displacement >0.2 mm) were also included in the model (Satterthwaite et al., 2013). Functional data were then smoothed with a 4 mm full-width at half-maximum kernel. No global signal regression or independent component analysis denoising was applied. The output of these preprocessing steps was a residual time series used in seed-to-voxel correlation analyses (see below).
TMS targets
TMS targets were defined for each participant using anatomical and rs scans. For the PPC stimulation target, we first identified the hand knob of the left precentral gyrus on the axial and sagittal slices of the anatomical scans (Yousry et al., 1997; Fig. 1, left panel). We then created a spherical region of interest (ROI) with a radius of 5 mm centered on this point. From this ROI, we extracted the mean preprocessed rs time series. This time series was used to calculate whole-brain seed-to-voxel Pearson's correlations using AFNI's 3dTcorr1D command. We next set a relatively lenient threshold of p < 0.001 on the resulting M1 correlation map and masked out all clusters smaller than 40 contiguous voxels. The PPC stimulation site was defined as the surviving cluster that was most strongly connected with M1 in the region of the left parietal cortex (Fig. 1, middle panel). A seed was then defined for this PPC area, and an additional rs-fMRI analysis was performed. A right hemisphere CB target was selected based on maximal correlation with the PPC seed using a frameless stereotactic neuronavigation system (Brainsight 2; Rogue Research). A visual cortex (VC) target with the lowest absolute correlation with PPC was selected as an active control site (Fig. 1, right panel). Stimulation at this site controls for the nonspecific effects of stimulation (cutaneous sensation, peripheral nerve sensation, induced current in the brain, etc.) The high-resolution T1–weighted scan for each participant was imported into Brainsight and coregistered to digitize anatomical landmarks for TMS coil placement during the experiment.
TMS target selection procedure. The M1 targets were selected based on the amplitude of MEP. A rs-fMRI analysis was then performed with the hand knob area as a seed. The maximum correlation with the parietal lobe was selected as our PPC target. A seed was then defined for this PPC area, and an additional rs-fMRI analysis was performed. A right hemisphere CB target was selected based on maximal correlation with the PPC seed. A right VC target with the lowest absolute correlation with left PPC was selected. Correlation maps and targets are depicted on cortical (fsLR; Glasser et al., 2013) and cerebellar (SUIT; Diedrichsen, 2006; Diedrichsen and Zotow, 2015) template surfaces for illustration purposes only. Actual targets were selected in Brainsight using native T1 space correlation maps.
TMS
MEPs elicited by a single TMS pulse with a monophasic waveform were recorded from the relaxed first dorsal interosseous and abductor pollicis brevis muscle in the right hand with Ag–AgCl surface electrodes in a tendon–belly arrangement. The electromyographic signal was amplified (1,000×), bandpass filtered (20 Hz–2.5 kHz; Model 2024F; Intronix Technologies), digitized at 5 kHz by an analog-to-digital interface (Micro 1401; Cambridge Electronics Design), and recorded by a computer using the Signal software (version 7; Cambridge Electronics Design) for off-line analysis.
iTBS
iTBS was delivered using a figure-8 coil with static cooling (MagPro MCF-B70) attached to a MagPro X-100 stimulator with MagOption (MagVenture). iTBS consisted of 10 bursts of high-frequency stimulation (a 2 s train of three biphasic waveform pulses at 50 Hz repeated every 200 ms) repeated every 10 s for a total of 190 s (600 pulses; Huang et al., 2005). The iTBS coil was held tangentially to the scalp for the CB and VC stimulation with the handle directed upward at 90°. The coil was held close to the scalp for sham stimulation but angled away, so no current was induced in the brain (Koch et al., 2020). Each participant's active motor threshold (AMT) was determined at the beginning of the experiment, defined as the minimum intensity required to produce MEPs of ≧200 μV in at least 5 of 10 trials, while the participant maintained a 20% maximum contraction in the targeted right-hand muscle (Huang et al., 2005). This yielded a mean percentage of maximum stimulator output (MSO) intensity of 42 ± 6 across participants. To minimize participant discomfort and decrease the interindividual difference in stimulation-induced effects in the underlying brain target, we opted to use a fixed MSO intensity of 35%, which is similar to standard iTBS protocols with 80% AMT (Huang et al., 2005) and within consensus recommendations for safety (Oberman et al., 2011; Rossi et al., 2021).
cPAS
Repetitive low-frequency pairs of cortical stimuli were delivered with a small figure-8 coil (D50 Alpha B.I., 5 cm diameter) to PPC and another figure-8 coil (D702, 7 cm diameter) to M1, each connected to a Magstim 2002 stimulator (Magstim). cPAS consisted of 100 pairs of monophasic waveform pulses at 0.2 Hz for ∼8.3 min (Koch et al., 2013; Goldenkoff et al., 2020, 2024). Coil 1 was positioned over the left PPC at a 10° angle to the midline, and Coil 2 was placed over the optimal scalp position for activation of the hand area of left M1 at a 45° angle to the midline, inducing a current in the posterior–anterior direction in the underlying cortical tissue. The interpulse interval (IPI) of the paired pulses differed for experimental conditions: PPC stimulation preceded M1 stimulation by 5 ms (Hebbian-like plasticity induction protocol) or by 500 ms (control for timing contiguity for paired associative plasticity in parietal–motor pathway; Koch et al., 2013; Johnen et al., 2015; Goldenkoff et al., 2020, 2024; Lazari et al., 2022). Measures of motor excitability for each participant were determined for the individual coils at the beginning of the experiment. Resting motor threshold (RMT) was defined as the minimum intensity required to produce MEPs of ≧50 μV in the relaxed targeted right-hand muscle in 5 of 10 consecutive trials (Rossini et al., 1994). The intensity of the first TMS pulse (Coil 1 to PPC) was set to 90% of RMT, while the second TMS pulse intensity (Coil 2 to M1) was set to produce an MEP of ∼1 mV in the relaxed targeted right-hand muscle. The same stimulation location and intensities for M1 and PPC were used throughout the experiment for each participant for each experimental day.
Experimental design
All participants underwent an fMRI scanning session on a separate day before the stimulation experiments. In the main experiment, all participants received iTBS to right CB (EXPCB condition) or right VC (CTRLVC condition), followed by cPAS in the left hemisphere. iTBS and cPAS protocols were separated by a 10 min interval. Here, we investigated whether cerebellar iTBS modulated the subsequent Hebbian-like plasticity effects produced by cPAS of interconnected parietal and motor areas. We selected VC, a region outside of the rs-fMRI connectivity network in our sample, to evaluate whether nonspecific effects of brain stimulation could have produced the effects on the subsequent induction of paired associative plasticity in the left parietal–motor pathway. Each experimental condition was performed on a separate day at least 5 d apart. The order was counterbalanced. During testing, participants were seated comfortably in an armchair with both hands relaxed. At the beginning of each testing session, we determined the optimal scalp position for eliciting MEPs in the targeted right-hand muscle by delivering single-pulse TMS with the D702 coil with minimal stimulation intensity. The lowest stimulation intensity needed to elicit a ∼1 mV MEP was determined for each participant before testing. The same TMS intensity was used to examine the changes in LTP-like plasticity in left M1 (Chen and Udupa, 2009) after the different plasticity induction protocols. Each testing session included baseline (pre-iTBS), post-iTBS, and post-cPAS assessments (Fig. 2). MEP amplitudes were measured before and 10 min after iTBS to examine the immediate effect of iTBS on motor excitability. In addition, MEP amplitudes after (10, 20, 30, 40, 50, and 60 min) the cPAS protocol were measured to examine the effect of iTBS over the 60 min following Hebbian cPAS to PPC and M1. Twenty-four MEPs were acquired at each point in time. Single-pulse TMS was applied every 5 s. MEP amplitudes (mV) were measured peak to peak in a time window between 15 and 100 ms after single-pulse TMS.
Experimental design. All participants underwent a rs-fMRI scan before the TMS experiments. The main experiment included 14 subjects who underwent iTBS in two different conditions: EXPCB (dark blue boxes) and CTRLVC (red boxes). The iTBS was delivered to a functionally connected CB region in the EXPCB condition or a nonfunctionally connected VC region in the CTRLVC condition. This was followed by cPAS to the PPC and the M1 with a 5 ms interstimulus interval (ISI) at 0.2 Hz for ∼8.3 min. Eight participants took part in two additional control experiments. In the CTRL500ms (light blue boxes) experiment, iTBS was administered to CB and cPAS with a 500 ms ISI. In the CTRLSHAM (gray boxes) experiment, iTBS was administered with the coil held close to the scalp but angled away from it, and cPAS was given at a 5 ms ISI. MEPs elicited by single-pulse TMS were measured at three different time points: baseline, 10 min after iTBS, and at six different time points for 1 h after cPAS (at 10, 20, 30, 40, 50, and 60 min). The sessions were conducted at least 5 d apart from each other.
We performed two additional control experiments on eight of the original participants using the same procedures as the main experiment. We had planned to run all 14 of the original participants through these control experiments. However, restrictions on human participant research during the COVID-19 pandemic led to the premature termination of the study. One control experiment (CTRL500ms condition) evaluated whether cerebellar iTBS could have produced the effects on excitability in the motor cortex alone instead of an induced change in paired associative plasticity in the human parietal–motor pathway. Previous evidence shows that the induced Hebbian associative plasticity produced by cPAS depends on the timing of stimuli between PPC and M1 stimulation (e.g., IPI of <20 ms; Koch et al., 2013). We, therefore, applied the identical cPAS protocol (i.e., delivered the same number of pulses at the same frequency and intensity) to the same targeted parietal and motor areas from the main experiment but with an IPI of 500 ms (Johnen et al., 2015; Lazari et al., 2022; Goldenkoff et al., 2024). One participant's data were excluded due to excessive muscle activity. The other control experiment (CTRLSHAM condition) verified whether the Hebbian cPAS protocol with an IPI of 5 ms between PPC and M1 delivered alone could induce the expected pattern of associative plasticity in a sample of eight participants from the main experiment.
Statistical analysis
Statistical analysis was performed using R (R Core Team, 2019; Version 3.6.1). To model the effect of cerebellar iTBS on motor cortical excitability before and after the application of cPAS, we fit a Bayesian hierarchical generalized linear model (GLM). Bayesian statistics provide numerous advantages, including making intuitive probabilistic statements about the range of credible values (e.g., Parameter x has a probability of 0.95 of being in the range of 1–5) and incorporating prior knowledge. As MEP amplitudes are strictly positive with positive skew, we specified a Gamma likelihood function and a log-link function. The model included fixed effects of stimulation condition (EXPCB, CTRLVC, CTRL500 ms, CTRLSHAM) and phase (pre-iTBS, post-iTBS, post-cPAS) and their interaction. The model additionally allowed intercepts to vary by subject. A posterior distribution over possible parameter values was sampled using Markov chain Monte Carlo (MCMC) sampling implemented in rstan (Stan Development Team, 2020; Version 2.21.2) via the brms package (Bürkner, 2017, 2018); Version 2.14.4). The model was specified as follows:
We ran four separate chains with 7,000 iterations each. The first 2,000 iterations were discarded as warm-ups. R-hat values were all very close to 1 (R-hat ≤ 1.001), and the effective sample size exceeded 5,000 for all parameters indicating that MCMC chains had converged, and minimal autocorrelation was in the sampling. Posterior predictive checks confirmed that distributional assumptions were met and that the specified model could generate data that qualitatively resembled the actual data. For each parameter in the model, we report the median, 95% highest density interval (HDI), and the probability of direction (pd). The HDI represents the interval for which all values within that interval have a higher probability density than points outside. Due to the log-link function, we exponentiate the median and 95% HDI values for reporting so that values represent multiplicative factors on the original response scale [e.g., a one-unit change in x is associated with an increase in MEP amplitude by a factor of exp(0.1) = 1.11]. pd is an index of effect existence (ranging from 50 to 100%), representing the probability that an effect goes in a particular direction (e.g., pd = 0.99 indicates that effect x has a 99% probability of being negative). Note that pd represents the probability that the effect is negative or positive before exponentiation and the likelihood that the effect is less than or greater than one after exponentiation. We consider a pd greater than 95% to be evidence for an effect and a pd < 95% to indicate limited evidence for an effect.
To examine the effect of phase for each stimulation condition individually, we took draws from the expectation of the posterior predictive distribution (
To characterize the effect of cerebellar iTBS across the 60 min following the cPAS protocol, we additionally fit a gamma GLM that included fixed linear and quadratic effects of time. Intercepts were again allowed to vary by subject:
To characterize the magnitude and consistency of the effects of cerebellar iTBS and cPAS alone, we analyzed a subset of participants who completed both the EXPCB and CTRLSHAM conditions (N = 8). We extracted the EXPCB post-iTBS time point as the MEPs collected at this time point reflect the effect of cerebellar iTBS alone. MEPs collected 10–60 min post-cPAS in the CTRLSHAM conditions were also extracted as these time points reflect the effect of cPAS without any iTBS priming. To assess the consistency of the iTBS and cPAS effect within an individual, we computed the correlation and associated Bayes factor (Ly et al., 2016) between the iTBS effect (percentage change in MEP amplitude at the post-iTBS time point of the EXPCB condition) and the cPAS effect (percentage change in MEP amplitude at 10–60 min post-cPAS of the CTRLSHAM condition). To assess differences in the magnitude of iTBS and cPAS effect, we performed paired Bayesian t tests (Rouder et al., 2009).
Results
Participants reported no undesirable side effects after stimulation. TMS targets for each participant are shown in stereotactic space overlaid on a template brain using the Brainsight software in Figure 3. RMT was 42.0 ± 5.8% of MSO for Coil 1, while the intensity to elicit an MEP amplitude of 1 mV in the targeted right-hand muscle was 49.2 ± 6.4% of MSO for Coil 2 across participants.
Stimulation targets for each participant are shown in stereotactic space overlaid on a template brain using the Brainsight software (N = 14). A, Dots indicate individual locations for the M1 (orange) and PPC (yellow) for cPAS. B, Dots indicate individual locations for the CB (blue) and VC (red) for iTBS.
Pre-iTBS, post-iTBS, and post-cPAS
To determine the impact of cerebellar iTBS on distal cortical physiology, we first evaluated MEP amplitudes before and after the application of iTBS. We fit a Bayesian hierarchical gamma GLM to MEP amplitudes with fixed effects of phase (pre-iTBS, post-iTBS, and post-cPAS), stimulation condition (EXPCB, CTRLVC, CTRL500 ms, CTRLSHAM), and their interaction. Post-iTBS MEP amplitudes increased 11.5% relative to pre-iTBS in the EXPCB condition (Fig. 4A; median, 1.115; 95% HDI [1.015, 1.234], pd = 98.97%; BF = 5.65). This result indicates that the application of iTBS to cerebellar nodes functionally connected with PPC increased MEP amplitudes prior to the application of cPAS, consistent with prior reports (Koch et al., 2008a).
In contrast, the CTRLVC condition that stimulated an active control site and the CTRLSHAM condition in which stimulation was directed away from the scalp exhibited little evidence for a change in MEP amplitude post-iTBS relative to pre-iTBS (Fig. 4A,B; CTRLVC, median, 1.053; 95% HDI [0.955, 1.164]; pd = 85.40%; BF = 0.17; CTRLSHAM, median, 1.039; 95% HDI [0.921 1.169]; pd = 73.93%; BF = 0.15). The GLM analysis found some evidence of a decrease in MEP amplitude between pre- and post-iTBS timepoints in the CTRL500 ms condition (Fig. 4B; median, 0.865; 95% HDI [0.722 1.019]; pd = 95.91%). However, the more targeted Bayes factor comparison revealed evidence in favor of the null hypothesis of no difference between pre- and post-iTBS (BF = 0.45).
We next examined the effect of cerebellar iTBS before and after the application of cPAS to PPC and M1. For this analysis, we collated the MEP values from all six time points following cPAS. In the EXPCB condition, post-cPAS MEP amplitudes reduced by 8.4% relative to MEPs post-iTBS (Figs. 4A, 5A; median, 0.915; 95% HDI [0.843, 0.986]; pd = 99.12%; BF = 39.64). In contrast, there was a 33.4% increase in MEP amplitudes following cPAS in the CTRLVC condition relative to the post-iTBS MEP amplitudes (Figs. 4A, 5A; median, 1.334; 95% HDI [1.211, 1.470]; pd = 100.00%; BF = 322,776,386). This positive change in MEP amplitude was also found in the CTRL500ms condition; MEP amplitudes increased by 22.4% post-cPAS relative to post-iTBS (Figs. 4B, 5B; median, 1.224; 95% HDI [1.079, 1.396]; pd = 99.86%; BF = 13.71). As MEP amplitude unexpectedly decreased from pre-iTBS to post-iTBS in the CTRL500ms condition, this increase could reflect a delayed effect of cerebellar iTBS on M1 excitability consistent with the EXPCB condition and prior work (Koch et al., 2008b). The suppression effect of cPAS observed after cerebellar iTBS is dependent on both the site of iTBS and the specific timing that is thought to induce changes in PPC–M1 connectivity. Similarly, we observed a 51.4% increase in MEP values in the CTRLSHAM condition post-cPAS relative to prior-cPAS (median, 1.514; 95% HDI [1.337, 1.726]; pd = 100.00%; BF = 19,269,214). This signifies that a Hebbian cPAS protocol between PPC and M1 delivered alone, with no prior neuromodulating stimulation to the CB, is associated with increased MEP amplitude. Together, these results demonstrate that iTBS to the CB, followed by Hebbian cPAS to PPC and M1, is associated with an inhibitory effect on MEP amplitude. This effect is specific to the EXPCB condition; MEP amplitudes in all three control conditions increased following cPAS relative to post-iTBS.
CB iTBS reverses the effect of subsequent parietal–motor plasticity induction protocol. A, Group (N = 14) average change in MEP amplitude for EXPCB and CTRLVC at each time point (expressed as percentage change from the average MEP amplitude recorded at the pre-iTBS time point). Error bars (pre-iTBS and post-iTBS) and shaded area (10–60 min Post-cPAS) indicate within-subject standard error (Morey, 2008). B, Sham CB iTBS and sham cPAS cannot explain the effect of CB priming on parietal–motor plasticity induction. Group (CTRL500ms, N = 7; CTRLSHAM, N = 8) average change in MEP amplitude at each time point (expressed as percentage change from the average MEP amplitude recorded at the pre-iTBS time point). Error bars (pre-iTBS and post-iTBS) and shaded area (10–60 min post-cPAS) indicate within-subject standard error (Morey, 2008).
Individual participant time courses. A, Change in MEP amplitude (expressed as percentage change from the average MEP amplitude recorded at the pre-iTBS time point) across time for each subject (colored lines) and condition (EXPCB and CTRLVC). Black dots and lines indicate the group average across subjects (same as in Fig. 4A; Morey, 2008). B, Change in MEP amplitude across time for each subject (colored lines) and condition (CTRL500ms and CTRLSHAM). Black dots and lines indicate the group average across subjects (same as in Fig. 4B).
Extended effects of modulatory stimulation protocols
We next examined the time course of MEP amplitude changes associated with cerebellar iTBS over the 60 min following cPAS. MEPs were collected at 10 min intervals for an hour following cPAS in all four conditions. An analysis of MEP amplitudes at the six time points following cPAS revealed suppression in the EXPCB condition compared with the facilitation of MEPs across the three control conditions. We found a robust negative linear effect and positive quadratic effect across the 60 min time window in the EXPCB condition (Figs. 4A, 5A; linear, median, 0.864; 95% HDI [0.787, 0.951]; pd = 99.91%; quadratic, median, 1.015; 95% HDI [1.001, 1.028]; pd = 98.46%), representing an initial sharp decrease in MEP amplitudes that plateaus over the 60 min time window.
In contrast, the three control conditions showed a positive linear and negative quadratic effect over time following cPAS. Modulating the VC instead of the CB before Hebbian cPAS (CTRLVC) is associated with somewhat larger MEP amplitudes over time (Figs. 4A, 5A; linear, median, 1.063; 95% HDI [0.979, 1.155]; pd = 92.20%; quadratic, median, 0.999; 95% HDI [0.982, 1.005]; pd = 87.16%). Similarly, refraining from delivering iTBS stimulation at all, as in the CTRLSHAM condition, also led to an increase in MEP values post-cPAS (Figs. 4B, 5B; linear, median, 1.476; 95% HDI [1.328, 1.624]; pd = 100.00%; quadratic, median, 0.959; 95% HDI [0.946, 0.972]; pd = 100.00%). Additionally, the robust negative linear effect observed in the EXPCB condition was not observed when cerebellar iTBS was followed by non-Hebbian cPAS as in the CTRL500ms condition (linear, median, 1.018; 95% HDI [0.903, 1.146]; pd = 62.13%; quadratic, CTRL500ms, median, 0.996; 95% HDI [0.979, 1.012]; pd = 68.03%), consistent with a sustained nonvarying increase in MEP amplitude across the 60 min interval. These results demonstrate a facilitation effect of cPAS when either paired with sham stimulation or stimulation outside of the network of interest that subsided over the 60 min time window and neither facilitation nor suppression with cerebellar iTBS alone.
Magnitude and consistency of iTBS and cPAS across individuals
In the subset of participants that completed both the EXPCB and CTRLSHAM conditions (N = 8), we examined the magnitude and consistency of effects for CB iTBS and cPAS alone. The EXPCB post-iTBS time point reflects the effect of cerebellar iTBS on motor excitability alone, and all post-cPAS time points (10–60 min) in the CTRLSHAM (sham CB iTBS followed by PPC–M1 cPAS) reflect the effect of cPAS alone (Fig. 6). Despite the small sample size, we found a strong correlation in effect magnitude between CB iTBS at the post-iTBS time point and all CTRLSHAM post-cPAS time points [EXPCB (post-iTBS)–CTRLSHAM (10, 20, 30, 40, 50, 60 min), r(6) = 0.769, 0.674, 0.331, 0.698, 0.819, 0.828; BF = 2.74, 1.79, 0.81, 1.97, 3.67, 3.90], indicating the effect of iTBS and cPAS was consistent within an individual. In other words, participants who showed an effect of cerebellar iTBS also tended to exhibit a cPAS effect. There was limited evidence for a difference in the magnitude of the effect of iTBS and cPAS (BF range, 0.45–1.55) other than 40 min post-cPAS. There was some evidence the effect of cPAS at 40 min was greater than the effect of iTBS (BF = 3.07).
Magnitude and consistency of cerebellar iTBS and PPC–M1 cPAS. Change in MEP amplitude (expressed as percentage change from the average MEP amplitude recorded at the pre-iTBS time point) for each subject who completed both the EXPCB and the CTRLSHAM condition (N = 8). The EXPCB post-iTBS time point reflects the effect of cerebellar iTBS alone and the CTRLSHAM post-cPAS time points reflect the effect PPC–M1 cPAS alone.
Discussion
The present study investigated how cerebellar activity modulates plasticity in an interconnected parietal–motor network associated with action control. Our findings demonstrate that altering cerebellar activity attenuates the plasticity-inducing effects of subsequent cPAS between the PPC and M1. Relative to stimulating a region out-of-network, sham CB stimulation, and non-Hebbian cPAS, modifying cerebellar activity blocked the induction of cortical associative plasticity in a distal parietal motor pathway previously implicated in motor actions. These results provide causal evidence for cerebellar regulation of cortical plasticity through heterosynaptic metaplasticity mechanisms.
The precise mechanism by which the CB modulates cortical plasticity is currently unclear. All connections between the cerebral cortex and CB are polysynaptic. Afferent fibers originating from the cerebral cortex reach the cerebellar cortex via the pons or the inferior olivary nucleus. Efferent connections are relayed to the cerebral cortex through the cerebellar nuclei and then the thalamus (Evarts and Thach, 1969; Kemp and Powell, 1971; Strick, 1985; Schmahmann and Pandyat, 1997). Inhibitory Purkinje cell axons that synapse on deep cerebellar nuclei are the sole output of the cerebellar cortex. The dentate nucleus has been shown to exert a tonic facilitatory influence on M1 (Liepert et al., 2004), and activation of Purkinje cells is associated with inhibition of the motor cortex (Ugawa et al., 1994, 1995, 1997; Pinto and Chen, 2001; Daskalakis et al., 2004; Reis et al., 2007). Based on these findings and the original theta burst stimulation study demonstrating that continuous theta burst stimulation (cTBS) is inhibitory and iTBS is excitatory within the motor cortex (Huang et al., 2005), one might expect that iTBS would activate Purkinje cells, resulting in inhibition of the dentate nucleus and disfacilitation of M1. Surprisingly, Koch and colleagues (2008) instead found that the inhibitory cTBS protocol reduces MEP amplitudes following single-pulse TMS applied to M1, and the excitatory iTBS protocol increases MEP amplitudes. This result led to speculation that theta burst stimulation protocols influence subpopulations of interneurons within superficial layers of the cerebellar cortex without directly affecting cerebellar cortex output (i.e., Purkinje cells). Further work combining TMS and EEG demonstrated that cerebellar iTBS decreased TMS-evoked potentials (TEPs) from 140 to 190 ms following TMS (Casula et al., 2016). Motor cortex TEPs during this time range have been associated with GABAB-mediated inhibitory postsynaptic potentials (Casula et al., 2014; Premoli et al., 2014a,b), indicating that cerebellar iTBS results in disinhibition of the motor cortex. This is consistent with our finding that MEP amplitude increased following the application of cerebellar iTBS in the EXPCB condition. We observed a decrease in MEP amplitude in the sham cPAS condition following the application of CB iTBS, which is opposite to our finding in the EXPCB condition and the results of Koch et al. (2008b). However, we found an increase in MEP amplitude that was sustained for the subsequent 60 min following the application of sham cPAS. As the cPAS with an IPI of 500 ms would not be expected to induce plasticity, we interpret this increase as an effect of cerebellar iTBS that is simply delayed relative to the EXPCB condition.
Following the application of iTBS to CB, we observed a robust reduction in motor excitability induced by Hebbian cPAS across the subsequent 60 min. It has been previously shown that cerebellar iTBS results in a topographically precise loss of the induction of plastic changes in M1 following the application of paired associative stimulation over the median nerve at the wrist and M1 (T. Popa et al., 2013). Both of these findings are consistent with the Bienenstock–Cooper–Munro (BCM) theory of a sliding threshold for the induction of LTP or LTD (Bienenstock et al., 1982). According to this rule, any change in the M1 homeostatic state conditions the subsequent induction of plasticity. Here, priming with a stimulation protocol that enhances M1 excitability (i.e., CB iTBS) prevented further facilitation of M1 excitability by subsequent stimulation (i.e., cPAS). In the framework of the BCM theory, cerebellar iTBS may have shifted the threshold for plasticity induction in the PPC–M1 circuit (Thomson and Sack, 2020; Sack et al., 2024), making it more difficult for cPAS to induce further potentiation. This metaplastic effect could serve as a homeostatic mechanism to prevent excessive potentiation and maintain network stability (Bear, 1995). Future work could further elucidate the metaplastic effects of cerebellar stimulation by comparing our protocol (CB iTBS→ PPC–M1 5 ms cPAS) to the application of cerebellar cTBS followed by PPC–M1 −5 ms cPAS. Both CB cTBS and PPC–M1 −5 ms cPAS applied alone have been shown to produce anti-Hebbian plasticity (i.e., reduction in motor excitability; Koch et al., 2013). Based on our current findings, we would predict a facilitative effect on MEP amplitude following PPC–M1 −5 ms cPAS if conditioned by CB cTBS. Again, using the framework of the BCM theory, CB cTBS would be expected to decrease M1 excitability and, therefore, reduce the threshold for plasticity induction in the PPC–M1 circuit by subsequent stimulation.
It was further shown that while cerebellar iTBS attenuated PAS-induced M1 plasticity, it had no effect on plasticity induced by the subsequent application of iTBS to the motor cortex (T. Popa et al., 2013). This suggests that cerebellar priming influences how information is conveyed to M1 rather than directly modulating plasticity within M1.
Our study focused on neurophysiological measures of plasticity (i.e., MEPs). Future studies incorporating behavioral tasks assessing motor learning or adaptation could provide a more comprehensive understanding of the functional relevance of cerebellar-modulated cortical plasticity. Specifically, investigating how cerebellar iTBS affects the acquisition and consolidation of new motor skills or the adaptation to perturbations during reaching movements could elucidate the behavioral consequences of cerebellar-induced metaplasticity (Koch et al., 2020). Multimodal research integrating these plasticity measures with rTMS and neuroimaging during skilled actions could define network-level effects for skilled motor function, providing a broader picture of the modulatory changes that underlie behavior (Fox et al., 2012; Johnen et al., 2015; Sale et al., 2015; Polanía et al., 2018; To et al., 2018; Spampinato et al., 2024). Using dual-site TMS during action planning also may provide insights into how CB affects the processing of action-specific parietal–motor functions (Davare et al., 2010, 2011; Vesia and Davare, 2011; Koch, 2020).
There is a growing appreciation for using cerebellar neuromodulation as a therapeutic strategy for guiding plasticity to improve movement behavior (França et al., 2018; Manto et al., 2022; Spampinato et al., 2024). The ability to noninvasively modulate cortical plasticity through cerebellar stimulation opens new possibilities for investigating and potentially treating disorders characterized by maladaptive plasticity or impaired motor learning, such as stroke, ataxia, and Parkinson's disease (Grefkes and Fink, 2014; Liew et al., 2014; Raffin and Siebner, 2014; Raffin and Hummel, 2017; França et al., 2018; Potvin-Desrochers and Paquette, 2021; Manto et al., 2022). For example, recent work demonstrated that cerebellar iTBS combined with physical therapy improves gait and balance functions in individuals with hemiparesis (Koch et al., 2018). The results of the present study further our understanding of how rTMS can be employed to improve motor function and lend credence to the idea that stimulation can rectify aberrant functional connections observed in patients (Vercammen et al., 2010; Fox et al., 2012; Hartwigsen and Volz, 2021). Future research building on these results could lead to more network-targeted and effective neuromodulatory interventions (Caligiore et al., 2016a,b; Beynel et al., 2020; Horn and Fox, 2020; Panda et al., 2024), ultimately improving outcomes for individuals with neurological and sensorimotor disorders.
Footnotes
We are grateful to Amir Mashni, Hannah Lavis, Danielle Destiny, Semat Adekoya, Julie Waineo, Jacob Dattilio, Claire Shea, and Megan Gordon for their assistance in data collection and analysis. The work was supported by R21NS118055 and S10OD026738 from the National Institutes of Health, the University of Michigan School of Kinesiology Pilot Research Grant, and Mcubed seed-funding from the University of Michigan Office of Research (UMOR). E.R.G. was supported by the National Science Foundation Graduate Research Fellowship Program (NSF GRFP). A postdoctoral fellowship from the National Institutes of Health (F32MH124268) supported J.A.B.
↵*E.R.G. and J.A.B. contributed equally to this work and share first authorship.
↵‡M.V. is the senior author.
The authors declare no competing financial interests.
- Correspondence should be addressed to Michael Vesia at mvesia{at}umich.edu.
