Time dependent effects of cerebellar tDCS on cerebello-cortical connectivity networks in young adults

Abstract

The cerebellum is involved in non-motor processing, supported by topographically distinct cerebellar activations and closed-loop circuits between the cerebellum and the cortex. Disruptions to cerebellar function may negatively impact prefrontal function and processing. Cerebellar resources may be important for offloading cortical processing, providing crucial scaffolding for normative performance and function. Here, we used transcranial direct current stimulation (tDCS) to temporarily alter cerebellar function and subsequently investigated resting state network connectivity. Critically, what happens to these circuits if the cerebellum is not functioning optimally, or after stimulation, remains relatively unknown. We employed a between-subjects design with 74 participants total (38 female; M= 22.0 years, SD= 3.45), applying anodal (n=25), cathodal (n=25), or sham (n=24) stimulation to the cerebellum to examine the effect of stimulation on cerebello-cortical resting state connectivity in young adults. We predicted increased functional connectivity following cathodal stimulation and decreased functional connectivity following anodal stimulation. We found, anodal stimulation resulted in increased connectivity in both ipsilateral and contralateral regions of the cortex, perhaps indicative of a compensatory response to degraded cerebellar output. Additionally, a window analysis also demonstrated a time dependent nature to the impacts of cerebellar tDCS on connectivity, particularly with cognitive regions of the cerebral cortex. This work suggests that when cerebellar outputs are degraded, in this case by tDCS, the cerebellum offloads its processing responsibility which encourages more cortical regions to engage to compensate for the degraded cerebellar output. This results in in differences in cortical activation patterns and performance deficits. These results might inform and update existing compensatory models, which focus primarily on the cortex, to include the cerebellum as a vital structure involved in the scaffolding of cortical processing.

Introduction

Viral tract tracing in non-human primates has shown topographically distinct loops connecting the cerebellum and cortex. Anterior regions of the cerebellum such as Lobules I-IV and V have projections to the primary motor cortex, whereas posterior lobules, such as Crus II, project to the prefrontal cortex (Kelly & Strick, 2003). Converging evidence in human neuroimaging parallels the viral tract tracing work in primates (Clower et al., 2005; Coffman et al., 2011; Hoshi et al., 2005; Kelly & Strick, 2003). Both human functional magnetic resonance imaging (fcMRI; Krienen & Buckner, 2009) and diffusion-weighted imaging (Salmi et al., 2010) found segregated fronto-cerebellar networks, linking the cerebellum to the motor cortex, the dorsal lateral prefrontal cortex, the medial prefrontal cortex, and the anterior prefrontal cortex in humans. O’Reilly and colleagues (2010) have furthered this understanding by demonstrating functional zones in the cerebellum such that anterior regions of the cerebellum have specific associations with motor, somatosensory, visual, and auditory cortices; whereas, posterior lobes have projections to, and functional networks with, the prefrontal and posterior parietal cortices (Bernard et al., 2012; Diedrichsen et al., 2019; Guell et al., 2018; King et al., 2019; O’Reilly et al., 2010). Behaviorally, work has suggested connectivity between the PFC and cerebellum might predict performance on learning executive function tasks (Reineberg et al., 2015), highlighting the non-motor contributions of the structure to broader networks. Further, degraded gray and white matter integrity observed in aging (Bernard et al., 2013) and disease (Allen et al., 2007; Bai et al., 2009; Bernard & Mittal, 2014; Mittal et al., 2014) have behavioral implications as the internal models used by the cerebellum for automaticity may not be utilized as effectively due to disruptions in cerebellar function and network connectivity (Bernard et al., 2013; Bernard & Seidler, 2014; Maldonado, et al., 2022; Miller et al., 2013). Thus, understanding these networks and alterations therein could be the basis for an improved understanding of cerebellar contributions to behavior.

Transcranial direct current stimulation (tDCS) is a noninvasive neuromodulation technique that typically increases (anodal) or decreases (cathodal) neural activity in the cerebral cortex using a small amount of electrical current, impacting cortical excitability (Brunoni et al., 2012; Nitsche & Paulus, 2000; Priori, Berardelli, Rona, Accornero, & Manfredi, 1998), and in turn behavior (Coffman et al., 2014). However, anodal stimulation to the cerebellum excites inhibitory Purkinje cells, which ultimately decreases signal to the cortex, whereas cathodal stimulation to the cerebellum inhibits the same cells, resulting in increased signal to the cortex (Ghez, 1991; Grimaldi et al., 2016). The impact of anodal cerebellar tDCS is of particular interest, as it may provide insights into cerebello-cortical interactions when cerebellar output is degraded (Ferrucci et al., 2016), improving our understanding of cerebellar involvement in non-motor cognitive processing. Limited work in healthy populations has found anodal stimulation increases functional connectivity from the cerebellum to cortical areas involved in the motor control of speech (Turkeltaub et al., 2016), language and speech motor regions in the left hemisphere (D’Mello et al., 2017; Turkeltaub et al., 2016), and spelling (Sebastian et al., 2017). Additionally, anodal stimulation reduced cerebellar ataxia symptoms and improved cerebellar output in individuals with cerebellar ataxia (Benussi et al., 2017). Further, anodal stimulation to the cortex also increases cortico-cerebellar connectivity when applied to the dorsal lateral prefrontal cortex (Abellaneda-Pérez et al., 2020), motor cortex (Cummiford et al., 2016) and the right posterior parietal cortex (Callan et al., 2016).

Thus, cerebello-cortical connectivity has been well-documented, and work focused on language networks has indicated that these networks are subject to modulation via cerebellar tDCS (D’Mello et al., 2017; Turkeltaub et al., 2016). Critically, recent functional imaging work, coupled with tDCS, demonstrated that disrupting the cerebellum resulted in greater cortical processing, suggesting a cerebellar scaffolding mechanism (Maldonado, et al., 2022). That is, the cerebellum provides resources when the cortex is taxed. Here, we were interested in understanding how connectivity between the cerebellum and the cortex might support this mechanism. That is, if the cerebellum is disrupted, will the cortex recognize the lack of processing input from the cerebellum, and in turn increase connectivity with cortical regions to help maintain task performance (Bernard, 2022)? Thus, we used tDCS and fcMRI to better understand cerebello-cortical functional connectivity. Though past work has suggested anodal stimulation increases connectivity between the cerebellum and language networks in the cortex specifically (D’Mello et al., 2017; Turkeltaub et al., 2016), we predict cathodal stimulation will inhibit the inhibitory Purkinje circuit, resulting in a decrease in inhibitory function on the deep cerebellar nuclei (DCN), output regions of the cerebellum, resulting in an increase in cerebellar output to the cortex. Anodal stimulation will excite the inhibitory Purkinje circuit, resulting in an increase in inhibitory function on the DCN and a decrease in cerebellar output to the cortex. This is consistent with past work suggesting cerebellar tDCS manipulates the inhibitory nature of Purkinje cells (Galea et al., 2009; Ghez, 1991; Grimaldi et al., 2016), and in turn impacts patterns of task-based cortical activation (Maldonado et al., 2022). To this end, we conducted a priori connectivity analyses, focusing primarily on Lobule V, Crus I, Crus II, and Lobule VIIIa to assess the polarity specific changes in cerebello-cortical connectivity; but also completed exploratory analyses using all lobules of the right cerebellum to further unpack our findings.

Methods

Participants

Seventy-five healthy, young adults participated in this study and were provided monetary compensation for their time. Exclusion criteria included left handedness, history of neurological or mood disorders, skin conditions, pregnancy, and history of concussion. Data from one participant was not used because the participant did not wish to finish the experiment after providing consent. Thus, seventy-four right-handed participants (38 female) ages 18 to 30 (M= 22.0 years, SD= 3.45) were included in the analyses, a sample in line with typical imaging studies (Szucs & Ioannidis, 2020). Participants were randomly assigned to either the anodal (n=25), cathodal (n=25), or sham (n=24) stimulation condition. All procedures completed by participants were approved by the Texas A&M University Institutional Review Board and conducted according to the principles expressed in the Declaration of Helsinki.

Procedure

Resting state data was collected as part of a larger imaging study (Maldonado et al., 2022) that took approximately two hours to complete. Functional imaging and behavioral data were also collected but described elsewhere (Maldonado et al., 2022). In the context of this larger study, the resting state scan was one of the first scans completed, and all resting state data were collected within 45 minutes of stimulation.

Following the completion of the consent form, participants completed a basic demographic survey, followed by tDCS (see below for details). Participants were blind to the stimulation condition and were tDCS naïve. Following stimulation, participants were escorted to the scanner for the brain imaging protocol. It is thought tDCS lasts up to 90 minutes (Batsikadze et al., 2013, 2019; Galea et al., 2009; Nitsche & Paulus, 2001) and on average, there was a 15.53 minute (SD=5.28; range 8–40 minutes) delay between the end of stimulation and the beginning of the resting state scan. During this delay, the tDCS equipment was removed, participants were escorted to the scanner, completed a final safety check for scanning and then situated in the scanner. Once ready to scan, participants completed a localizer scan that lasted about 90 seconds, before they completed two 4-minute resting state scans. Seven participants had a delay longer than 20 minutes (cathodal=3, anodal=3, sham=1) between the end of stimulation and the beginning of the resting state scan. These delays were primarily due to issues with situating people in the scanner, or a delay because of equipment malfunctions. Critically, removing their data did not meaningfully affect the average time between the end of stimulation and the beginning of the resting state scan. Additionally, removing outliers (data points 3 SD above mean; n=2) from the connectivity analysis did not change the results. This would suggest the effect of tDCS was still present during scanning. After the resting state scans, participants completed several functional and structural scans as part of a separate study (Maldonado et al., 2022). Once the scanning procedures were complete, participants completed a tDCS sensation survey and were debriefed. No adverse events were reported in the tDCS sensation survey.

tDCS Parameters

Participants were fitted with a classic two electrode montage to administer either cathodal, anodal, or sham stimulation using a Soterix 1×1 tES system. One 5×5cm electrode was placed two cm below and four cm lateral of the inion over the right cerebellum. In addition to convention (Ferrucci et al., 2015), there cerebellum has contralateral projections to the cortex such that the right cerebellum projects to the left frontal lobe (Palesi et al., 2017; Ramnani, 2006). Because connectivity associated with many non-motor cognitive processes projects to the left frontal lobe (Buckner, 2013; Diedrichsen et al., 2019), we chose to stimulate the right cerebellum to better understand how cerebellar tDCS might affect connectivity associated with non-motor cognitive processing. The second 5×5cm electrode was placed on the right deltoid. Both electrodes were affixed using elastic bands. Each electrode was placed in a pre-soaked sponge, which had 6 mL of saline solution added on each side.

Once electrodes were placed, stimulation was set to 1.0 mA for thirty seconds to ensure the electrodes made a good connection with the scalp. If contact quality was below 50%, adjustments, such as moving hair to increase the electrode’s contact with the scalp, were made and contact quality was rechecked. Following a successful re-check, participants completed a 20-minute stimulation session at 2 mA (Ferrucci et al., 2015; Grimaldi et al., 2014, 2016). During the stimulation conditions, maximum stimulation intensity was reached in 30 seconds and maintained for 20 minutes, and then would return to 0 mA. During sham conditions, to simulate the active stimulation conditions, the stimulator turned on and maximum stimulation intensity (2 mA) was reached within 30 s, but immediately returned to 0 mA across 30 s. During this ramp down, the timer starts so that the participant experiences a 20-minute session but does not receive active stimulation. At the conclusion of the 20-minute session, the same ramp-up/ramp down procedure is used and the stimulation session is complete.

Using the Soterix HD-Explore software (Huang et al., 2017), we simulated the electric field (EF) created by our montage (Figure 1) and we found a max field intensity of 0.42 V/m when placing the stimulating electrode over B12 (approximately 2 cm lateral and 4 cm proximal of the inion on a 332 channel montage). Though there is some bleeding of the EF into the occipital lobe, most intensity changes were located within the right cerebellum. The stimulation seemed to primarily target Lobules IV-X (.2-.42 V/m), with some effect on Lobules I-V (.1-.25 V/m).

Figure 1.

(A) Soterix simulation model with the active electrode over B12 and the return over the right shoulder. We found a max field intensity of 0.42 V/m. The colorbar indicates the strength of the electric field using V/m. (B) SimNibs simulation with the active electrode over the right cerebellum (X: 42.31, Y: −81.91, Z: −66.24). We found an average peak intensity around .25 V/m. The colorbar indicates the strength of the electric field using V/m. Table 1 summarizes true values.

SimNibs4 (Thielscher et al., 2015) was also used to model the electric field following cerebellar tDCS using participant structural images. We first created head meshes based on T1 structural images using CHARM (Puonti et al., 2020). With the resulting head mesh, we simulated the electric field for each participant. The simulation included an active 2.5×5cm electrode in a 5×5 cm saline-soaked sponge placed over the right cerebellum (X: 42.31, Y: −81.91, Z: −66.24) with the same parameters used for the return placed on the neck (X: 54.06, Y: −36.17, Z: −117.58), to approximate a deltoid return placement. Default conductivities were used, though the values for saline and sponges were changed to 1.4 S/m (Datta et al., 2009). When simulations were completed, the CLOS Pipeline (Rezaee & Dutta, 2019) was used to extract the electric field values from each lobule of the cerebellum, which were then plotted on a cerebellar flatmap (Diedrichsen & Zotow, 2015). Figure 1 displays the average electric field for each lobule in each stimulation condition. Table 1 reports the average electric field for each lobule and the average max peak intensity for each lobule in each stimulation condition in V/m. Briefly, the electric field seems to peak in Lobules VII and VIII (~.25 V/m). Though the max intensities differ between the Soterix and SimNibs models, the general pattern of impacts are parallel across the modeling approaches.

Table 1.

Mean and max electric fields by stimulation condition and lobule in V/m.

	Cathodal				Anodal
Region	Min	Max	Mean	Mean Max	Min	Max	Mean	Mean Max
Right I-IV	0.037	0.152	0.058	0.125	0.031	0.090	0.052	0.115
Right V	0.044	0.173	0.070	0.130	0.038	0.113	0.064	0.119
Right VI	0.053	0.217	0.087	0.165	0.046	0.146	0.081	0.148
Right Crus I	0.064	0.258	0.104	0.199	0.049	0.193	0.099	0.191
Right Crus II	0.081	0.266	0.125	0.231	0.057	0.228	0.120	0.208
Right VIIb	0.097	0.290	0.152	0.256	0.072	0.259	0.147	0.234
Right VIIIa	0.095	0.298	0.153	0.256	0.078	0.260	0.148	0.236
Right VIIIb	0.087	0.256	0.132	0.228	0.075	0.217	0.125	0.213
Right IX	0.066	0.231	0.096	0.162	0.051	0.171	0.090	0.154
Right X	0.051	0.225	0.084	0.156	0.051	0.129	0.078	0.143

fMRI data acquisition

Resting state fMRI data was collected at the Texas A&M Translational Imaging Center with a 3-T Siemens Magnetom Verio scanner using a 32-channel head coil. Two blood oxygen level dependent (BOLD) whole brain scans with a multiband factor of 4 were collected in the absence of any task (number of volumes = 114, repetition time [TR] = 2000 ms, echo time [TE] = 27 ms; flip angle [FA] = 52°, 3.0 × 3.0 × 3.0 mm voxels; 56 slices, interleaved, slice thickness=3.00mm, field of view (FOV) = 300 × 300 mm; time = 4:00 min). The scans were collected in opposite encoding directions (anterior → posterior and posterior → anterior). During the scan participants viewed a centrally located fixation cross and were asked to stay awake while “thinking about nothing in particular”. An additional high resolution T1 weighted whole brain anatomical scan was taken (sagittal; GRAPPA with acceleration factor of 2; TR = 2400 ms; TE = 2.07 ms; 0.8 × 0.8 × 0.8 mm3 voxels; 56 slices, interleaved, slice thickness= 0.8mm; FOV = 256 × 256 mm; FA = 8°; time = 7:02 min) for data normalization.

Rs-fMRI data pre-processing

Both anatomical and functional images were collected using DICOM format but were converted to NIFTI files and organized into a Brain Imaging Data Structure (BIDS) using bidskit (v 2019.8.16; Mike Tyszak, 2016).

Rs-fMRI data were pre-processed using CONN (Whitfield-Gabrieli & Nieto-Castanon, 2012) standalone toolbox (v20b). Functional and anatomical data were preprocessed using a flexible preprocessing pipeline (Nieto-Castanon, 2020b) including realignment with correction of susceptibility distortion interactions, slice timing correction, outlier detection, direct segmentation and MNI-space normalization, and smoothing. Functional data were realigned using SPM realign & unwarp procedure (Andersson et al., 2001), where all scans were coregistered to a reference image (first scan of the first session) using a least squares approach and a 6 parameter (rigid body) transformation (Friston et al., 1995), and resampled using b-spline interpolation. Temporal misalignment between different slices of the functional data (acquired in interleaved Siemens order) was corrected following SPM slice-timing correction (STC) procedure (Henson et al., 1999; Sladky et al., 2011), using sinc temporal interpolation to resample each slice in a BOLD timeseries to a common mid-acquisition time. Potential outlier scans were identified using ART (Whitfield-Gabrieli et al., 2011) as acquisitions with framewise displacement above 0.9 mm or global BOLD signal changes above 5 standard deviations (Nieto-Castanon, 2022; Power et al., 2014), and a reference BOLD image was computed for each subject by averaging all scans excluding outliers. Functional and anatomical data were normalized into standard MNI space, segmented into grey matter, white matter, and CSF tissue classes, and resampled to 2 mm isotropic voxels following a direct normalization procedure (Calhoun et al., 2017; Nieto-Castanon, 2022) using SPM unified segmentation and normalization algorithm (Ashburner, 2007; Ashburner & Friston, 2005) with the default IXI-549 tissue probability map template. Last, functional data were smoothed using spatial convolution with a Gaussian kernel of 5 mm full width half maximum (FWHM).

Functional Connectivity Analysis

Functional data were denoised using a standard denoising pipeline (Nieto-Castanon, 2020a) including the regression of potential confounding effects characterized by white matter timeseries (5 CompCor noise components), CSF timeseries (5 CompCor noise components), motion parameters and their first order derivatives (12 factors) (Friston et al., 1996), outlier scans (12 factors) (Power et al., 2014), and linear trends (2 factors) within each functional run, followed by bandpass frequency filtering of the BOLD timeseries (Hallquist et al., 2013) between 0.008 Hz and 0.09 Hz. This was applied to the resting state data in order to focus on slow-frequency fluctuations while minimizing the influence of physiological, head-motion and other noise sources (Hallquist et al., 2013). Potential confounding effects to the estimated BOLD signal were determined and removed separately for each voxel and for each subject and functional run/session, using Ordinary Least Squares (OLS) regression to project each BOLD signal timeseries to the sub-space orthogonal to all potential confounding effects using an anatomical component-based noise correction procedure (aCompCor) (Behzadi et al., 2007).

Seed to voxel analysis.

Seed to voxel analyses, were conducted using whole lobular (Bernard et al., 2012; Grami et al., 2021) ROIs in the right cerebellar hemisphere. The whole lobular approach was determined a priori. Though we examined all lobules (Right Lobules I-X), we focused our reporting on Lobules V-VIIIa. The more posterior regions seem to be most impacted by stimulation, while Lobules V and VI provide an important point of contrast as they seemingly received less of the stimulation. Fisher-transformed bivariate correlation coefficients between each lobular ROI BOLD timeseries and each individual voxel BOLD timeseries were assessed for the anodal, cathodal, and sham stimulation groups individually. Individual scans were weighted by a boxcar signal characterizing each individual task or experimental condition convolved with an SPM canonical hemodynamic response function and rectified. Group level comparisons were as follows: anodal>sham, cathodal>sham, and anodal>cathodal. Results were thresholded using a combination of a cluster-forming p < 0.001 voxel-level threshold, and a familywise corrected p-FDR < 0.05 cluster-size threshold (Chumbley et al., 2010). Because of the delay between the end of stimulation and the beginning of the resting state scan (M=15.53 minutes, SD=5.28 minutes), we controlled for this variable in our analyses by entering the time delay in our model as a covariate.

Window analysis.

We used the standalone CONN toolbox (v20b) (Whitfield-Gabrieli & Nieto-Castanon, 2012) dynamic variability in seed-based connectivity pipeline to evaluate sources of temporal variability in functional connectivity patterns. We created two-time windows that were 120 seconds long, with onsets at 0 (early window) and 120 (late window) seconds post scan onset. Each individual window is treated as a separate condition, and weighted GLM is used to compute corresponding time-specific measures. Variability of these measures across time is then computed, creating dynamic variability in seed-based connectivity maps which highlight correlations in temporal tied variability between a seed region and a voxel in the brain for each window. That is, this analysis will examine how connectivity between a cerebellar lobule and a voxel change over time. We assessed anodal, cathodal, and sham stimulation individually and the contrasts between them for each cerebellar seed region within each time window and compared changes in connectivity. Based on the simulations reported above, we focused on Crus I and II, though other lobules (Lobule I-VI and IX-X) were included, in an exploratory fashion. Results were thresholded using a combination of a cluster-forming p < 0.001 voxel-level threshold, and a familywise corrected p-FDR < 0.05 cluster-size threshold (Chumbley et al., 2010).

Results

Connectivity patterns for each stimulation group are not reported here. However, connectivity patterns were broadly consistent with canonical cerebello-cortical networks (Bernard et al., 2012; Diedrichsen et al., 2019; King et al., 2019; O’Reilly et al., 2010) for the sham and active stimulation groups. In brief, for Lobules I-V we primarily see positive connectivity to motor regions, such as the precentral gyrus, in addition to other memory, language and visual processing regions of the cortex. Posterior regions of the cerebellum correlated with regions within the frontal, parietal, and temporal lobes of the cortex, as expected. While all stimulation contrasts were examined, we only reported significant group contrasts (i.e. Anodal > Cathodal) that emerge within each lobule below. Additionally, using our a priori hypotheses and the simulations reported above, we focused on Crus I and II, though other lobules (Lobule I-VI and IX-X) were also reported on, in an exploratory fashion.

Whole Brain Resting fMRI Analysis

Motor Lobules.

When examining motor-oriented lobules (Lobules I-V), we found lower connectivity between Lobules I-V and regions in the pre- and post-central gyrus and other motor regions in the cathodal stimulation group compared to the sham group (Figure 2A; Table 2). There were no significant effects for the Anodal >Sham or Anodal>Cathodal contrast.

Figure 2.

A. The cathodal stimulation group had lower connectivity (relative to sham) to motor regions, with some modulation in cognitively oriented regions in the frontal lobe. B. The anodal stimulation group showed higher connectivity (relative to cathodal) to regions in the occipital lobe. The color bars denote the percentile range of z-scores. The maps are thresholded such that only significant (p-FDR < 0.05) results are presented.

Table 2.

MNI coordinates of the local maxima of brain regions showing functional connectivity with the lobules of the right cerebellar hemisphere by stimulation contrast.

Stimulation Contrast	Seed	Region included in cluster	Cluster Size	MNI Coordinates of Cluster			t-Value	p-FDR
				x	y	z
Cathodal>Sham	Lobule I-IV	R Precentral gyrus	310	30	−4	42	−5.29	< .001
		R Supplemental Motor Area
		R Middle Frontal Cortex
		Posterior Cingulate Gyrus
		R Insular Cortex	86	32	10	−6	−5.75	0.009
		R Putamen
	Lobule V	R Precentral Gyrus	144	32	−6	42	−4.37	< .001
		R Middle Frontal Gyrus
		L Postcentral Gyrus	128	−32	−24	38	−5	< .001
		L Superior Frontal gyrus
		L Precentral Gyrus
	Lobule X	R Intracalcarine Cortex	107	2	−82	14	−4.84	0.003
		L and R Supracalcarine Cortex
		R Cuneal
Anodal>Cathodal	Crus II	L and R Intracalcarine Cortex	120	4	−68	14	4.52	0.002
		L and R Supracalcarine Cortex
		R Cuneal
		Precuneus
	Lobule X	L Superior Parietal Lobule	94	−38	−56	64	−4.59	0.011
		L Lateral Occipital Cortex
		R Supplemental Motor Area

Non-Motor Lobules.

For the cognitively oriented lobules (Lobule VI-Crus II), we found higher connectivity between Crus II and visual regions such as the intra and supracalcarine cortex in the anodal stimulation group compared to cathodal (Figure 2B; Table 2). Additionally, connectivity was lower between Lobule X and visual regions such as the right intra and supracalcarine cortex, and the cuneus in the cathodal group compared to the sham group (Table 2). We also found lower connectivity between Lobule X and visual regions such as the lateral occipital lobe, as well as cognitive regions in the superior parietal lobe in the anodal stimulation group compared to the cathodal group (Table 2). There were no significant effects for the Anodal >Sham contrast.

Together, cathodal stimulation might result in lower connectivity between motor lobules in the right cerebellum and bilateral cognitive regions of the cortex. Anodal stimulation over cognitively oriented lobules in the right cerebellum result in higher contralateral connectivity in areas related to vision processing.

Window Analysis

To further investigate the impact of cerebellar tDCS on connectivity, an exploratory window analysis was used to investigate how connectivity patterns vary across the duration of the scan. Below we describe general patterns within an early and late window. Both windows were 120 seconds long with the early window starting at scan onset and the late window starting at 120 seconds after scan onset.

Motor Lobules.

When examining connectivity of motor lobules in the cerebellum, there was a general trend in which stimulation in the cathodal group lowered connectivity from Lobules I-IV to ipsilateral motor areas, such as the precentral gyrus, and contralateral cognitive areas including the paracingulate and cingulate gyri. This was particularly evident in the early window compared to the sham group (Table 3). In the later window the cathodal group had generally reduced connectivity from Lobules V to motor regions, such as the supplemental motor area compared to the sham group. This may suggest that cerebellar stimulation to motor regions differentially impacted connectivity over time post stimulation, which might result in differing behavioral impacts over the course of task completion, when tDCS is used in task paradigms.

Table 3.

MNI coordinates of the local maxima of brain regions showing functional connectivity with the lobules of the right cerebellar hemisphere by stimulation contrast within each window.

Window	Stimulation	Seed	Region included in cluster	Cluster Size	MNI Coordinates of Cluster			t-Value	p-FDR
					x	y	z
Early	Anodal > Sham	Crus II	L and R Intracalcarine Cortex	199	−4	−68	12	4.9	<.001
			L and R Supracalcarine Cortex
			L and R Lingual Gyrus
			Precuneus
	Cathodal > Sham	Lobules I-IV	R Precentral gyrus	196	20	−16	42	−4.71	< .001
			R Supplemental Motor Area
			R Middle Frontal Gyrus
			Posterior Cingulate gyrus
			Anterior Cingulate gyrus	60	−10	16	38	−4.8	0.001
			L Paracingulate gyrus
	Anodal > Cathodal	Lobule VI	R Lingual Gyrus	151	28	−48	−6	−5.39	< .001
			R Parahippocampal gyrus
			R Temporal Fusiform Cortex
			R Temporal Occipital Fusiform Cortex
			Posterior Cingulate Gyrus
		Crus II	L and R Intracalcarine Cortex	74	6	−68	16	4.23	0.024
			L and R Supracalcarine Cortex
			L and R Lingual Gyrus
			R Cuneal
			Precuneus
		Lobule X	L Superior Parietal Lobe	85	−36	−56	64	−5.06	< .001
			L Lateral Occipital cortex
			R Lateral Occipital cortex	67	18	−72	62	−5.6	< .001
Late	Anodal > Sham	Crus I	R Middle Frontal Gyrus	173	44	30	24	5.07	< .001
			R Frontal Pole
			R Inferior Frontal Gyrus
			L Temporal Occipital Fusiform Cortex	80	−34	−60	−18	4.8	0.009
			L Occipital Fusiform Gyrus
			L Cerebellar Lobule VI
		Crus II	L and R Paracingulate Gyrus	118	−4	44	20	−4.85	0.003
			L and R Superior Frontal gyrus
			R Frontal Pole
		Lobule VIIIb	R Intracalcarine Cortex	116	6	−66	10	−4.81	0.002
			R Lingual Gyrus
			Precuneus
			L Cuneal	94	−20	−74	20	−4.41	0.005
			Precuneus
			L Occipital Cortex
	Cathodal > Sham	Lobule V	Anterior Cingulate Gyrus	73	26	−12	40	−4.72	0.027
			Supplemental Motor Area
		Lobule IX	L Frontal Pole	68	−42	36	−2	4.59	0.036
			L Frontal Orbital Cortex
			L Inferior Frontal Gyrus
			R Caudate	55	10	−12	22	4.59	0.002
	Anodal > Cathodal	Lobule VI	R Middle Frontal Gyrus	110	42	22	38	5.05	0.003
Late>Early	Anodal > Sham	Crus I	Lateral Occipital Cortex	67	−40	−78	−4	4.71	0.028
		Crus II	L Lateral Occipital Cortex	118	−40	−74	−14	4.92	<.001
			L Occipital Fusiform Gyrus
			L Crus I
			L Cerebellar Lobule VI
			L Occipital Pole	90	−20	−96	0	4.47	0.003
			L Lateral Occipital Cortex
			L Occipital Fusiform Gyrus
			R Occipital Pole	70	28	−90	0	5.31	0.008
			R Lateral Occipital Cortex
			R Superior Frontal Gyrus	49	28	−2	70	4.96	0.032
			R Precentral Gyrus
		Lobule VIIb	R Thalamus	76	6	−24	10	4.44	0.015
	Anodal > Cathodal	Crus II	L Superior Frontal Gyrus	64	−20	20	54	−4.39	0.032
			L Frontal Pole	47	−34	48	−2	−4.87	0.049
			R Frontal Operculum Cortex	45	36	14	8	4.44	0.049
			R Insular Cortex
		Lobule VIIb	L Frontal Pole	114	−34	40	−18	−4.8	0.001
			L Frontal orbital Cortex
		Lobule X	L Middle Temporal Gyrus	81	−44	−50	8	4.37	0.004
			L Lateral Occipital Cortex
			L Angular Gyrus

Non-Motor Lobules

When examining cognitive regions of the right cerebellum, there was a general trend in which anodal stimulation over the right cerebellum resulted in greater temporal variability in connectivity across windows, particularly when analyzing the right Crus II seed region. Much of the variability in connectivity during the early window was found between Crus II in the cerebellum and visual regions of the occipital lobe such as the bilateral intracalcarine cortex, supracalcarine cortex, and the precuneus in the anodal stimulation group when compared to the sham and cathodal stimulation group (Figure 3A; Table 3). During the late later window, however, connectivity from right Crus I in the anodal stimulation group was greater in language processing centers in the ipsilateral middle and inferior frontal gyrus compared to the sham group (Figure 3B; Table 3). Interestingly, there was less connectivity in the anodal stimulation group between Crus II and cognitive centers in the frontal lobe, compared to sham (Figure 3C; Table 3). This continues to demonstrate temporal variability in cerebello-cortical connectivity following tDCS and perhaps suggest a time dependent nature to cerebellar tDCS effects.

Figure 3.

The temporal dynamics of connectivity of both Crus I and II was impacted by anodal stimulation (relative to sham). A. The anodal stimulation group had greater connectivity (relative to sham) between Crus I and visual regions in the occipital lobe during the early time window. B. The anodal stimulation group had greater connectivity (relative to sham) between Crus I and cognitively oriented regions in the frontal lobe during the late time window. C. The anodal stimulation group had lower connectivity (relative to sham) between Crus II and cognitively oriented regions in the frontal lobe during the late time window. D. The anodal stimulation group had greater connectivity (relative to sham) between Crus II and visual regions in the occipital lobe during the late time window (relative to early windows). E. The anodal stimulation group had lower connectivity (relative to cathodal) between Crus II and cognitively oriented regions in the frontal lobe during the late time window (relative to early windows). The color bars denote the percentile range of z-scores. The maps are thresholded such that only significant (p-FDR < 0.05) results are presented.

In more posterior cerebellar lobules, such as right Lobules VIIIb-X, the anodal stimulation group had lower connectivity to visual attention centers in the contralateral superior parietal lobule and the lateral occipital cortex compared to the cathodal stimulation group (Table 3). This was particularly evident in the early window in Lobule X in the anodal stimulation group. In the later window, the anodal stimulation group saw lower connectivity to visual regions of the occipital lobe such as the bilateral intracalcarine cortex compared to sham. Interestingly, in the late window, connectivity between Lobule IX and X and frontal regions was larger in the cathodal stimulation group to compared to sham (Table 3).

Lastly, we also assessed a contrast between windows (Late>Early) and generally found that when looking at Crus I and II, the anodal stimulation group had higher connectivity to bilateral visual regions in the occipital lobe in the late time window compared to the early window, relative to sham (Table 3; Figure 3D). However, the anodal stimulation group saw lower connectivity during the late window compared to the early window between Crus II and contralateral frontal lobe regions, such as the Superior Frontal Gyrus, relative to the cathodal stimulation group (Table 3; Figure 3E). There were no significant effects for the Cathodal>Sham contrast.

Together, this continues to suggest a time dependent variability for cortical impacts of cerebellar tDCS, and perhaps begins to indicate a polarity specific effect on connectivity, which could in turn impact behavior.

Discussion

Cerebellar tDCS can be used to understand how cerebellar networks with the cortex differ after perturbation, which can in turn provide insights into aging as well as neurological and psychiatric illnesses where the cerebellum and cerebello-cortical networks are impacted (Ferrucci et al., 2016; van Dun & Manto, 2018). To this point, however, most work has focused on language networks, with little work looking at other nonmotor processing domains using a whole brain approach. The current work used tDCS over the right cerebellum to better understand cerebello-cortical connectivity. We found stimulation primarily affected cerebellar connectivity to motor and vision processing centers, with some modulation of cortical regions primarily implicated in cognition. Generally anodal stimulation resulted in increased connectivity in both ipsilateral and contralateral regions of the cortex, perhaps indicative of a compensatory response to degraded cerebellar output. A window analysis also demonstrated high temporally variable connectivity patterns that could suggest a time dependent nature to the impacts of cerebellar tDCS on connectivity. This was particularly notable for connectivity with cognitive and associative areas of the cerebral cortex. Together, the data suggest the effect of stimulation might be time dependent for both motor and cognitive processing, such that behavioral change following stimulation might differ depending on when behavior is observed compared to stimulation offset. A better understanding of how temporal dynamics of connectivity changes across time might help interpret behavioral effects of cerebellar stimulation. Our findings are discussed in more detail below.

Connectivity Patterns and Stimulation

Motor Regions.

Broadly, when examining connectivity in the sham stimulation group we found the cerebello-cortical connectivity patterns one might to expect between the cerebellum and motor regions in the cortex (Buckner et al., 2011; King et al., 2019). When comparing stimulation groups, connectivity between the cerebellum and motor regions (e.g., precentral gyrus and supplemental motor area) was lower in the cathodal group compared to sham. A window analysis demonstrated that the cathodal stimulation group had lower connectivity to motor regions in the cortex compared to the sham group, during early windows. In late windows, the cathodal stimulation group had decreased connectivity to inhibitory centers (anterior cingulate gyrus). Critically, the cathodal stimulation group generally saw lower connectivity relative to sham. We predicted only the anodal stimulation group would experience lower connectivity (Galea et al., 2009; Grimaldi et al., 2016), though variability in how cerebellar tDCS affects behavior is not uncommon, particularly when examining motor function (Buch et al., 2017). As such, this mixed literature also would support higher connectivity. Continued work is needed to determine why this variability exists, though our work here may provide some initial insights.

First, the cerebellum is a complex structure and it is difficult to know how it might be impacted by stimulation, as the effect depends on current flow direction, relative to axonal position (Chan & Nicholson, 1986). This, coupled with the curvature of the cerebellum, may contribute to the variable effects seen here and elsewhere (Buch et al., 2017). Our simulations support this notion that the cerebellum is difficult to stimulate as we only found small effects of stimulation materialize across lobules using participant data. This lends itself to the ongoing discussion about individual differences in the responsiveness of tDCS (Labruna et al., 2019). Similarly, recent work suggested that stimulating different cerebellar regions can have different behavioral and neural effects (Rice et al., 2021).

Second, it is still not clear which cells types are actually influenced by cerebellar tDCS (Grimaldi et al., 2016). Our hypotheses are based on the modulation of Purkinjie fibers, but their activity is not readily captured by fMRI, instead climbing fibers might drive the fMRI signal (Grimaldi et al., 2016). Though past work suggests Purkinjie cells respond to cerebellar tDCS (Grimaldi et al., 2016), it is unclear whether other cells are also modulated. Our electric field simulation suggests minimal changes in the electric field in motor lobules of the cerebellum, likely due the position of these lobules relative to the electrode. There is a significant amount of tissue matter preventing the current from the electrode from reaching these cerebellar lobules. Further, this change is likely not enough to produce direct change in connectivity. However, our results do demonstrate some change in connectivity of Lobules I-V following stimulation, perhaps an indirect response resulting from connectivity changes following stimulation in other regions, such as Crus I and II, which did experience meaningful change in field intensity and was likely influenced by stimulation. That is, connectivity changes between Crus I or Crus II and the cortex might modulate the connectivity between other cerebellar lobules, and in turn, the cortex following stimulation due to the closed loop circuits between the cerebellum and cortex (Kelly & Strick, 2003; Middleton & Strick, 2001). Past work has demonstrated a compensatory response in the cortex following cerebellar stimulation (Maldonado et al., 2022). The changes in connectivity between cerebellar regions not directly affected by stimulation and the cortex might be the result of a compensatory mechanism.

Together, these data provide further evidence for how cerebellar tDCS might affect cerebello-cortical connectivity. This is particularly pertinent when considering how the impact of stimulation may vary based on the cerebellar lobules that are targeted and the effect stimulation might have on connectivity with other cortical regions.

Non-motor regions.

Stimulation to nonmotor regions resulted in significant connectivity differences when compared to the sham stimulation group, but this connectivity was primarily to visual processing regions in the occipital lobe and language centers in the frontal lobes. Critically, our window analysis demonstrated that modulation of connectivity is not consistent, but variable, and perhaps even time dependent, such that variability of connectivity to language centers was not evident until a later time window. This is likely the result of stimulation reaching nonmotor lobules in the cerebellum at levels large enough to result in change in connectivity. This is likely due to these lobules (i.e. Crus I and II) being directly under the electrode. These results were partially in line with past work from Galea et al. (2009) that suggest that anodal stimulation would increase, and cathodal stimulation would decrease, cortical connectivity. This is also consistent with past cerebellar tDCS and imaging work that found that anodal tDCS to the right cerebellum increased Crus I and II activation and connectivity in cortical language centers (D’Mello et al., 2017) and areas involved in the motor control of speech (Turkeltaub et al., 2016). However, this is the first study to demonstrate impacts to frontal, temporal, and occipital regions associated with cognitive processing outside of language regions because of cerebellar stimulation.

The general pattern of results, particularly in the window analysis, seems to suggest that anodal stimulation modulates connectivity to the cortex, resulting in bilateral changes in connectivity. Task based fMRI work conducted by Maldonado, et al. (2022) suggested that anodal tDCS over the right cerebellum resulted in increased cortical activation. We argued that anodal stimulation disrupted cerebellar processing resulting in a greater need for cortical scaffolding. This idea is supported in past work which suggested anodal stimulation would negatively impact cerebellar output (Galea et al., 2009), and in turn activation of the cerebral cortex (Grimaldi et al., 2016). Here, this down regulation in the cerebellum following anodal stimulation might result in the need for greater connectivity to the ipsilateral hemisphere to compensate for the degraded communication between the cerebellum and contralateral cortical regions. As this did not consistently occur in the cathodal stimulation group, we suggest that the decreased output of the cerebellum is driving what we argue may be compensatory connectivity. Past work suggests the cerebellum might support cortical processing, such that degraded cerebellar output results in poor task performance (Bernard, 2022; Bernard et al., 2020; Bernard & Seidler, 2014; Schmahmann et al., 2019). Put simply, automatic processing relies on internal models and cerebellar processing, freeing up cortical resources, particularly if tasks become increasingly complicated. Perhaps this impact is not widespread, but hemisphere dependent. That is, decreased contralateral cortical connectivity resulted in the ipsilateral cortex upregulating cortical processing, which might have implications during functional task performance (Maldonado, et al., 2022).

Rather unexpectantly, cathodal stimulation did not seem to meaningfully impact non-motor connectivity to the cortex when compared to the anodal and sham stimulation groups. This is not in line with our hypothesis, as past work focused on language processing has demonstrated increased connectivity and behavioral performance following cathodal stimulation (Pope & Miall, 2012; Turkeltaub et al., 2016). It is possible that connectivity in young adults is already efficient, and cathodal stimulation does not reliably provide any change in connectivity, whereas there is room for degradation following anodal stimulation, as noted above. Critically, this work here, to our knowledge, is the first attempt to assess the effect of cathodal stimulation over the right cerebellum on cerebello-cortical connectivity broadly. That is, we were looking outside of the motor domain at more general connectivity patterns with the cortex. Though past work might predict stable, polarity specific changes in activation and connectivity following cerebellar tDCS (Galea et al., 2009; Grimaldi et al., 2016), the broader literature has been mixed as anodal and cathodal stimulation to the cerebellum has both improved (Maldonado & Bernard, 2021; Mannarelli et al., 2019; Pope & Miall, 2012; Wynn et al., 2019) and hindered cognitive task performance (Boehringer et al., 2013; Ferrucci et al., 2008; Maldonado & Bernard, 2021; Spielmann et al., 2017). However, some imaging work does provide encouraging data to support polarity specific effects, such that anodal stimulation might result in downregulation in the cerebellum (Macher et al., 2014; Maldonado et al., 2022) and cathodal stimulation resulted in the predicted disinhibition of the DCN (Küper et al., 2019).

One notable consideration is that the temporal variability occurred in such a small window of time and might call into question the stability of the tDCS effect. The current understanding is that cerebellar tDCS provides a generally consistent and sustained change, that lasts for 30–90 minutes following stimulation, depending on the stimulation parameters (Batsikadze et al., 2013, 2019; D’Mello et al., 2017; Galea et al., 2009; Küper et al., 2019; Macher et al., 2014; Maldonado, Jackson, et al., 2022; Nitsche & Paulus, 2001; Rice et al., 2021; Turkeltaub et al., 2016). The trend we see here might be the dynamic change in connectivity needed for the stable effect of cerebellar tDCS demonstrated in past work. It is also possible that we are observing changes in correlation dynamics that are distinct from broader changes in function and a reflection of the changes in cerebellar function. Perhaps, as the result of the short resting state timeframe, we are quantifying the active change in connectivity and the peak of the effect of stimulation occurs farther from stimulation offset than originally thought. It is also possible that we are seeing changes in long term potentiation (LTP; Miterko et al., 2019) within the cerebellum, which is affecting broader connectivity networks. Recent work using animal models has demonstrated anodal tDCS effects are dependent on LTP-like activity in Purkinje cells (Das et al., 2017; Gutierrez-Castellanos et al., 2017; Johansson et al., 2015). This gradual strengthening in the cerebellum might affect how cerebello-cortical connectivity changes over time and explains some of the variability found in the current data. This might have specific methodological implications, such that timing between cerebellar tDCS and task completion, or the oscillation in the effect on connectivity, might be a factor to consider when constructing stimulation procedures.

Limitations

Several limitations might affect interpretation of the current work. First, stimulation parameters, such as electrode placement and size, might not be conducive to modifying activations in Lobules I-V in the cerebellum (Rampersad et al., 2014). Indeed, investigation of electric field distributions in the cerebellum suggest anterior regions of the cerebellum, such as Lobules I-V, and the Vermis are not well stimulated (Rampersad et al., 2014). This is the result of the curvature of the cerebellum and its relative anatomical location below the brain. Second, the use of a between subject design might introduce bias as stimulation groups are comprised of different people which could increase the error introduced by individual variability, such as bone density, fluid dynamics, and head fat, etc. While a within subject design might be ideal, a between subjects’ design mitigates issues with attrition. Additionally, the current work used tDCS naïve, randomized and condition blind participants and our sample was in line or exceeded previous samples collected in similar imaging (Szucs & Ioannidis, 2020) and tDCS (Mitra et al., 2019) studies. Further, we do acknowledge the scan duration was short, lasting only four minutes. As our scan was part of a larger study, we chose four minutes to maximize stimulation effectiveness across the study procedure and reduce participant burden. Past work suggests the most common scan duration is five to seven minutes (Van Dijk et al., 2010), as the strength of the functional connectivity is considered stable. Though more recent work has demonstrated that scans between three and eleven minutes is enough to relivable evaluate functional connectivity (Braun et al., 2012). Our scan is within that range. Lastly, though we found a time dependent effects of cerebellar tDCS, we do acknowledge that a small-time window was used. What is considered an early time window for some participants might qualify as a late time window for others. While this does not negate the current work, future work may look to increase the scan duration and have more control in the delay between stimulation offset and scan onset to find a clear understanding of cerebellar tDCS and connectivity.

Conclusions

The current work aimed to better understand the impact of cerebellar tDCS on cerebello-cortical connectivity. Stimulation group differences in connectivity provided novel and important insights into the time dependent effects of tDCS on cerebellar networks. This work provides new insights into cerebello-cortical connectivity in the face of perturbation which has implications for aging and both neurological and psychiatric disease. Specifically, initial support for the cerebellum as a component for existing scaffolding models might be evident in the current data, further supporting its inclusion in models describing frontal lobe activation in health and disease. However, at this point additional work that further considers tDCS stimulation parameters and functional boundaries is still warranted.