Reorganizing the intrinsic functional architecture of the human primary motor cortex during rest with non-invasive cortical stimulation

Abstract

The primary motor cortex (M1) is the main effector structure implicated in the generation of voluntary movements and is directly involved in motor learning. The intrinsic horizontal neuronal connections of M1 exhibit short-term and long-term plasticity, which is a strong substrate for learning-related map reorganization. Transcranial direct current stimulation (tDCS) applied for few minutes over M1 has been shown to induce relatively long-lasting plastic alterations and to modulate motor performance. Here we test the hypothesis that the relatively long-lasting synaptic modification induced by tDCS over M1 results in the alteration of associations among populations of M1 neurons which may be reflected in changes of its functional architecture. fMRI resting-state datasets were acquired immediately before and after 10 minutes of tDCS during rest, with the anode/cathode placed over the left M1. For each functional dataset, grey-matter voxels belonging to Brodmann area 4 (BA4) were labelled and afterwards BA4 voxel-based synchronization matrices were calculated and thresholded to construct undirected graphs. Nodal network parameters which characterize the architecture of functional networks (connectivity degree, clustering coefficient and characteristic path-length) were computed, transformed to volume maps and compared before and after stimulation. At the dorsolateral-BA4 region cathodal tDCS boosted local connectedness, while anodal-tDCS enhanced long distance functional communication within M1. Additionally, the more efficient the functional architecture of M1 was at baseline, the more efficient the tDCS-induced functional modulations were. In summary, we show here that it is possible to non-invasively reorganize the intrinsic functional architecture of M1, and to image such alterations.

Figure 1. BA4 labelling and registration to the functional space.

An example of the quality of the cortical segmentation and labelling of BA4 is depicted, which was performed with Freesurfer in one of the participants. Panels A and B show the BA4 label over the pial and inflated brain surface respectively. Panel C shows the left BA4 label registered in the original T1 image. The red line shows the portion of the cortical segmentation performed by Freesurfer that corresponds to the left BA4 label (in this image left is right). Notice the tDCS electrode over the scalp of the subject is positioned over the central sulcus. Panel D shows the registration of the cortical segmentation and BA4 label in the native functional space of the same subject (first row: first 14 images of the native functional space. Second row: cortical segmentation registered from Freesurfer to the native functional space was overlapped. Third row: BA4 label (red) registered from Freesurfer space to the native functional space). For visualization purposes, we show the approximate location of the tDCS electrode over left BA4 in figure S1.

Figure 2. Global network metrics.

Shown are the results of the global network parameters that were calculated in the present study (mean connectivity degree (K) and the small-world parameters gamma, lambda and sigma [32]) calculated at each threshold T (0.25–0.35 in increasing steps of 0.002) before and after each Sham (A), Cathodal (B) and Anodal (C) tDCS. The second row - second column of each panel show that the approximate number of nodes (M1 voxels) of each undirected graph was ∼470. As expected, the mean connectivity degree monotonically decreases as T increases. M1 has salient small-world properties i.e. lambda≈1, gamma≫1, thus sigma≫1 , . No significant differences were observed for any of the network metrics before and after each of the tDCS sessions (P>0.05 paired two-tailed t-tests). Error bars represent the s.e.m.

Figure 3. Graph parameter statistics at the BA4 cortical surface.

(A) Shown is the flattening of the left BA4 (green labelled region) obtained from the left hemisphere surface average subject, which was used to project the statistical maps. Panels B to D show the ANOVA for the interaction effects (time*stimulation) Montecarlo cluster corrected at p<0.05 for the nodal connectivity degree maps (B), clustering coefficient maps (C) and the characteristic path length maps (D). Panels E to H show post hoc paired t-tests for the following contrasts: (E) After_Cathodal – After_Sham in the clustering coefficient maps; (F) After_Anodal – Before_Anodal in the clustering coefficient maps; (G) After_Anodal – After Sham in the characteristic path length maps; (H) After_Anodal – Before_Anodal in the characteristic path length maps. Notice that the L maps were L_rand/Li normalized, which means that the values of L in the significant cluster are lower after stimulation (see methods section).

Figure 4. Dependency of the tDCS-induced effects on baseline functional architecture.

Panel A shows that the effect of nodal clustering coefficient (C) increase found in the cluster of Figure 3F strongly correlated with baseline C (P = 0.0051; R² = 0.46). Panel B shows that the positive decrease found in the characteristic path length (L) maps that was found after anodal tDCS (Figure 3H) also has a positive correlation with the baseline L (P = 0.002; R² = 0.51).

Figure 5. Highly efficient nodes within M1.

The average L maps for all subjects and all before-tDCS fMRI scans were averaged. Nodes that showed the highest L_rand/Li values (i.e. nodes that communicate more efficiently within M1) were mapped over the flattened BA4. As an exploratory threshold we used the 15% of the voxels that showed the highest L_rand/Li values.