Subjects. Nine young healthy individuals participated in this study (mean age 31.22; SD 4.55; 5 males). All subjects were consistently right handed, as determined by the Edinburgh Handedness Inventory [40] with a mean laterality quotient of 97. Subjects were recruited from an open access advertisement posted on a website for subject recruitment (www.forsøgsperson.dk). All subjects gave their written informed consent. The experimental protocol (H-18031987) has been approved by the Regional Committee on Health Research Ethics of the Capital Region of Denmark, and the Declaration of Helsinki. All methods were performed in accordance with approved institutional guidelines and regulations.
Study design. Participants underwent a single pcASL-MRI session that lasted approximately 50 minutes. The experimental procedures are illustrated in Fig. 5. Bipolar aTDCS was applied with the anode targeting left primary motor cortex (M1) and the cathode on the right side of the forehead (supraorbital (SO) region). During stimulation, participants were resting in the MRI scanner with their eyes focusing on a fixation cross displayed in the middle of a screen.
The experiment included four aTDCS blocks. During a TDCS block, one of four current intensities were applied. Target current intensity was set at 0.5 mA, 1.0 mA, 1.5 mA or 2.0 mA in a given block. The order of blocks was pseudorandomized and participants were blinded. Each aTDCS block lasted 10 minutes and consisted of 4 min no-stimulation baseline followed by alternating epochs of aTDCS (38 s) and periods without TDCS (58 s). During an aTDCS epoch, stimulation current was linearly ramped up to the target intensity within 4s, continuously applied at target intensity for 30 s, and then ramped down again within 4s.
After each aTDCS block, the participants answered a series of questions through the MR-speaker system about how they had experienced the preceding aTDCS block, using a six-level Visual Analogue Scale (VAS)(see Supplementary Table 1).
We included finger-tapping (FT) blocks without stimulation at the beginning (FT-pre) and the end (FT-post) of the pcASL-MRI experiment to compare rCBF changes evoked by aTDCS with rCBF change during voluntary motor activity. During FT blocks participants tapped their right (dominant) index and thumb together paced by a blinking cross (2 Hz). Each FT-block was four min, consisting of interleaved 32 s epochs of FT and 32 s of rest.
Transcranial DC stimulation in the MRI scanner. We applied aTDCS at 0.5, 1.0, 1.5 and 2.0 mA. The center of the anodal electrode corresponding to the C3 location of the 10/20 EEG system [41], with 2-mm thick 7x5 cm rubber electrodes (NeuroConn, Illmenau, Germany). Connector plugs were arranged pointing down towards the ear for the anodal electrode, and horizontally outwards for the cathodal electrode. Electrodes were applied with a thin layer of ten-20 conductive gel (Weaver and Company, Aurora, Colorado, US) and fixated with a net cap. Current was applied by a battery-operated DC-stimulator with MR compatible stimulation cables and filter-boxes (NeuroConn Illmenau, Germany). For safety reasons, the total impedance was kept below 15 k, which included the two 5 k resistors in the cables.
MRI Image acquisition. Images were acquired on a Phillips 3 Tesla MR Achieva scanner (Philips, Best, Netherlands) using a 32-channel head coil. A localizer scan assessed the head-position prior to a structural T1-weighted whole-brain scan using a 3d-TFE multi-shot sequence (TR/TE = 6.0/2.7 ms; flip angle = 8°; FOV= 245 FH 245 AP 208 RL mm3; isotropic resolution = 0.85 mm3) and a T2-weighted whole-brain scan using a 3D-TFE multi-shot sequence (TR/TE = 2500/265 ms; flip angle = 90°; FOV= 245 FH 245 AP 190 RL mm3; isotropic resolution = 0.85 mm3).
Pseudo-continuous (pc)ASL was acquired as a single run per block, resulting in six ASL-MRI runs per participants, corresponding to the four aTDCS and two FT blocks per participant (Fig. 5). The labeling plane was positioned where the Vertebral and Internal Carotic Artery are parallel (approximately at C2 level) and angled perpendicular to the vessel orientations. pcASL images were acquired with background suppression by two pulses, pulsed continuous labeling, label duration of 1650 ms, and post-label-duration of 1200 ms, using an echoplanar imaging (EPI) readout. Image resolution was 3 x 3 x 4 mm. A single ASL volume consisted of 17 slices with a gap of 0.5 mm, covering the pericentral cortices and adjacent frontoparietal regions of both cerebral hemispheres. Duration for each ASL dynamic was 2 x 4.0 sec, with 78 dynamics per aTDCS block, and 32 dynamics per FT block.
Data analysis:
All ASL-MRI data was analyzed using FSL software, Wellcome Centre for Integrative Neuroimaging, University of Oxford. (https://www.win.ox.ac.uk).
Pre-processing of ASL-MRI data is described in Appendix A. ASL-MRI data were analyzed using FSL FEAT. For each participant, we did six separate first level analyses, one for each of the four aTDCS runs, and two for FT-runs. Voxel-wise changes in rCBF were analyzed by fitting a General Linear Model (GLM) to the time series of each ASL-MRI run/block modelling either one of the four aTDCS intensities or the movement sequences. The GLM featured three regressors, as previously described in Moisa et al. [42] (further details describing our GLM in Appendix A). The positive z-activation map from the perfusion regressor was used for group level analyses, described below.
Definition of volumes of interests
The left M1-HAND was our primary volume of interest (VOI) . We also wanted to determine specific functional changes in sub-regions within the primary VOI, based on depth and proximity to the anodal electrode and we planned to include the region with the highest induced E-field as an additional VOI:
Functional defined VOI of the left M1-HAND (M1FT). We used the average FT-pre activation as a functional localizer for this VOI (illustrated in Fig. 2, Fig. 4A, and Supplementary Fig. 2). The VOI was derived from a whole-brain voxel-wise group analysis using a one-sided t-test with a corrected statistical threshold of p<0.05 set for family wise error (FWE) cluster level correction and a cluster-defining threshold that was set to p <0.001 at the voxel level (corresponding to Z=3.1).
Anatomical defined VOI of the left primary sensorimotor cortex (SM1anat). To avoid bias from the motor-task guided VOI (M1FT), we additionally defined an anatomical VOI around the left primary sensory-motor cortex based on the probabilistic atlas from FSL’s Juelich map, including areas BA4a, BA4b BA3a, BA3b and BA6 for the left hemisphere. The probabilistic area masks were thresholded at 50% relative to their peak values and binarized. In addition, the extend of the mask was restricted to fit area around M1-HAND by including only voxels with MNI coordinates between -52 mm and -25 mm in left-right direction and ≥36 mm in inferior-posterior direction (Fig. 3).
Sub-regional spherical VOI’s in precentral cortex. Secondary VOI analyses included three precentral sub-regions and the frontal cortical site that was exposed to the maximal electrical field during aTDCS. We defined three VOIs within the left pre-central gyrus (10 mm spheres) using the FT-pre as a functional localizer. The VOIs were placed at the regional activity peaks within M1FT, (Fig. 4A). Two of the VOIs were placed at the superficial and deep point of highest regional activation in the primary motor cortex, “Mdeep” and “M1superficial” (center coordinates x,y,z: [-38, -24, 46] and [-38, -24, 60]). The third VOI was placed in a functional activated area corresponding to area 6 in the Juelich atlas in FSL, the dorsal premotor cortex “PMd” (center coordinate [-38, -6, 60]).
We performed additional analysis that considered the current distribution in each subject. We used individual structural scans to simulate the induced electric field for each subject, using SimNIBS v.3.2.2 [21]. For details on E-field simulations, please see Appendix A (and Fig. 4). To define a VOI for the maximal electrical field each individual E-field simulation at 1 mA were overlaid in MNI-space, considering only positions at which the gray matter of a least 5 subjects overlapped. A 10mm VOI was placed at the gray matter location at maximum of the averaged E-field (sphere center coordinate: -40, 28, 42), corresponding to the left dorsolateral prefrontal cortex (DLPFC) (bottom right panel in Supplementary Fig. 3).
Statistical inference on group level
Regional perfusion changes in the cortical VOIs. We ran six GLM analyses at group level, one for each aTDCS condition, corresponding to the four aTDCS intensities, and the two finger tapping sessions. The parameter estimates from the perfusion regressor from the first level analyses was fitted into a second level GLM that modelled the group average, corresponding to a one-sided t-test for each condition. In each model, the z-scores were averaged within our pre-defined volumes of interest (VOI definitions described in section 2.5.3). For each VOI, we submitted the spatially averaged z-scores into a two-way repeated measure ANOVA for within subject factor “conditions” with 6 levels (4 aTDCS intensities, 2 movement sequences). We performed post-hoc tests with pairwise comparisons, p-values were corrected for multiple comparisons using the Bonferroni method. For all ANOVA’s, we used Mauchly’s test to test for sphericity, and corrected with the Greenhouse Geisser method when sphericity was violated. ANOVA and post hoc tests were done using version 25 of the SPSS statistics software package (IBM, Armonk, New York, USA).
Correlation between regional E-field and regional perfusion change. We explored the relation between the highest induced E-field and perfusion response on an individual level. Here we simulated the E-field (at 0.5, 1.0, 1.5, 2.0 mA) in all four VOIs (M1deep, M1superficial, PMd, DLPFCE-field) in all subjects. For each intensity and each VOI, we ran linear regression analyses on the mean perfusion activation and E-field, to see if there were any correlation between the individual perfusion changes and estimated individual E-field strength (see Supplementary Fig 4). We used Bonferroni correction between the 16 comparisons, with an alpha level of 0.05.
Voxel-based exploratory analyses. We performed a complementary voxel-based analysis to test for linear increase or decrease in perfusion throughout all aTDCS intensities, including all voxels. We applied a cluster-size threshold of p = 0.05 and voxel-wise threshold of z = 3.1 or p = 0.001. We additionally explored finding potential activity with a more liberal voxel-wise threshold of z = 2.3 or p = 0.01.
Analysis of psychometric data. For details on analysis of VAS-scale rated sensory experience, please see Appendix A.