tACS of the Cerebellum and the Motor Cortex Entrains the Spiking Activity of the Cells in Motor Thalamus in a Frequency Dependent Manner

Abstract

Transcranial AC stimulation (tACS) of the cerebellum can entrain spiking activity in the Purkinje cells (PCs) of the cerebellar cortex and, through their projections, the cells in the cerebellar nuclei (CN). In this paper, we investigated if the cells in the motor thalamus (Mthal) can also be modulated (i.e. spikes entrained) via the CN-Mthal projections in rodents. A total of 82 thalamic cells were found, presumably in the Mthal by their stereotaxic coordinates, that were modulated by tACS of the cerebellum. Out of the 346 cells isolated, the thalamic cells with shorter action potentials and regular firing patterns had a higher probability of modulation by cerebellar stimulation than the cells with wider action potentials. The modulation level had a tuning curve with a maximum around 100-200 Hz. Spike histograms over the stimulation cycle transitioned between unimodal and bimodal distributions depending on the frequency. Most cells had a unimodal distribution at low frequencies, a bimodal distribution for frequencies between 80-125 Hz, and then a unimodal one for frequencies above 150 Hz. In addition, tACS of the motor cortex (MC) was also tested in a subset of thalamic cells. Unlike cerebellar stimulation, modulation levels peaked at two distinct frequencies, presumably due to entrainment through multiple MC-Mthal pathways with different preferred frequencies. The results demonstrate the feasibility of modulating a deep brain structure such as the thalamus through multi-synaptic pathways by stimulation of the cerebellar cortex (and the motor cortex) using a non-invasive neuromodulation method.

Fig. 1.

Experimental setup for recording of thalamic cells during transcranial cerebellar and motor cortex stimulation. (A) Locations of the craniotomy atop the skull for targeting the Mthal with a tetrode and the stimulation electrodes on the posterior skull and over the motor cortex are shown with small red circles. (B) VL and VM sections of Mthal are illustrated in a coronal section. (C) The tetrode tip was dyed with DiD’ leaves a mark in the Mthal in a representative case near the VL/VM border. (D) Connection of the stimulation and recording electrodes to the animal and the equipment.

Fig. 2.

Stereotaxic coordinates (depth, mediolateral-ML and anteroposterior-AP) of 346 cells recorded in the Mthal. The cells that are modulated by cerebellar stimulation (connected) are shown in red.

Fig. 3.

Clustering of spiking activity for a typical recording from the thalamus with a tetrode. (A) Scatter plots of the first principal component from each of the tetrode channels before clustering. PC1 of channels 1–3 are plotted along the x, y, and z axes, and PC1 of channel 4 is encoded by the color scale. Dark blue cluster in panel A is background noise. (B) Scatter plots after selecting 4 clusters where each cluster in different color belongs to a different cell. (C) Aggregate of raw data within the stimulation cycle (top) and after removing the extracted stimulation artifact and filtering (bottom). (D) Firing patterns of the thalamic cells in all four channels of the tetrode. Cells belonging to different clusters are marked with asterisks in different colors. (E) Aggregate of action potentials belonging to different clusters, plotted by spike triggering. (F) Probability distribution (histogram) of spike timings locked to the AC stimulation cycle for each cluster.

Fig. 4.

Response of a sample thalamic cell to AC sinusoidal stimulation of the cerebellar cortex with frequencies ranging from 1 Hz to 400 Hz at 200 μA. (A) Probability distribution of spikes over the AC stimulation cycle at each frequency. Red traces are the stimulus current waveforms (arbitrary units). (B) Corresponding PLV1 and PLV2 at each frequency.(C) Probability distribution of spiking activity in the polar plane.

Fig. 5.

PLV-frequency plots for thalamic cells during stimulation of the cerebellum (43 cells in 19 rats). (A)PLV1. (B) PLV2. (C) Percentage of unimodal responses(blue) and bimodal responses (red) at each frequency, with the number of cells indicated. For the unimodal case, Wilcoxon signed-rank comparisons between the peak frequency (125 Hz) and the other tested frequencies showed significant differences after Holm–Bonferroni correction. PLV at 125 Hz exceeded values at 1–60 Hz and 400 Hz (all adjusted p <0.04), with consistently large effect sizes (r >0.59), highlighting a significant peak at 125 Hz. In the bimodal case, significant differences were again observed, with PLV at 125 Hz greater than at 1–60 Hz and 150–400 Hz(all adjusted p < 0.008). The correlation coefficients were large (r > 0.74), further confirming a robust peak response at 125 Hz. (The effect size in Wilcoxon signed-rank is based on correlation coefficients, here and inFig. 11 [25]).

Fig. 6.

Mean phase vectors of the response to AC stimulation (80, 100, 125, 150 and 200 Hz) in thalamic cells (43 cells in 19 rats). (A-B) Unimodal. (C-D) Bimodal. B and D show the average phase shifts for each vector group, coded in different colors, as a function of frequency. A linear line-fit to the phase shifts has a slope of 0.034 rad/Hz for the bimodal peaks (D) and 0.036 rad/Hz for the means of the vector groups of the unimodal responses (B). The different colors indicate two groups of vectors classified into separate clusters by k-means clustering. Unimodal cases were represented with a single vector, while the bimodal cases were represented with two vectors, one for each peak in the histogram. To cluster the vectors into two groups, k-means clustering was repeated 1000 times, and the most frequently occurring centers were taken as the final cluster centers. In the bimodal case, we ensured that the two vectors of the same response were not placed in the same cluster. The only assumption made was that the phase rotates counterclockwise with increasing frequency, consistent with the expected propagation delay, and the cluster centers were assigned based on this assumption. For the unimodal cases, the number of vectors is the number of unimodal responses at the given frequency, as marked in Fig. 5C, whereas for bimodal cases, the count is doubled since each histogram was represented with two vectors.

Fig. 7.

Clustering of thalamic cells (346 cells in 24 rats) based on spike-timing and -shape based parameters. (A) Action potential waveform of 346 cells (modulated cells are plotted in red). Inset: Definitions for Peak-to-Trough Time (PTT), Repolarization Time (RT) time, and Half-Width. (B) k-means clustering by the first and second principal components can divide the cells into two clusters (circles and hexagrams). (C) Action potential shapes for the first cluster. (D) Action potential shapes for the second cluster. (E) distribution of cells according to LMI and PTT. (F) distribution of cells according to CV2 and half-width. The cells that can be modulated by cerebellar stimulation marked in red in all plots.

Fig. 8.

Distribution of shape and timing parameters within clusters created by PCA (top row, cluster 1 and 2) and groups by connectedness to the cerebellum (bottom row). (A-G) Distribution of cells within Cluster 1 and Cluster 2 according to shape (A-C) and timing parameters (D-G). In (A-G) the connected cells are marked with red dots as opposed to not-connected cells in grey. (H-N) Distribution of cells within the groups of connected and not-connected cells according to shape (H-J) and timing parameters (K-N). In (H-N) the cells in Cluster 1 are marked with blue dots as opposed to the cells in Cluster 2 in green. The effect size (η²), the square of the correlation coefficient from Wilcoxon rank sum test, is marked in each subplot [25].

Fig. 9.

Probability distribution of thalamic cells by depth for different groups segregated according to their action potential shape (narrow vs. wide) and ability to be modulated by cerebellar stimulation (connected vs. not connected). Plots are smoothed in MATLAB using ksdensity with 100 bins.

Fig. 10.

Response of a thalamic cell to sinusoidal stimulation of motor cortex with frequencies ranging from 1 Hz to 1000 Hz at 400 μA. (A) Probability distribution of spikes by the AC stimulation cycle. Red traces are the stimulus current waveform in arbitrary units. (B) Corresponding PLV1 and PLV2 at each frequency. (C) Probability distribution of spiking activity in the polar plane.

Fig. 11.

PLV-frequency plots for thalamic cells while stimulating the motor cortex (15 cells in 6 rats). (A) PLV1. (B) PLV2. (C) Percentage of unimodal (blue) and bimodal responses (red) at each frequency, with the number of cells indicated within the bars. In the unimodal case, Wilcoxon signed-rank tests revealed that the peak frequency (200 Hz) produced significantly higher PLV values than 1–80 Hz after Holm–Bonferroni correction (all adjusted p < 0.005). The correlation coefficients were r > 0.81. In the bimodal case, peak frequency (125 Hz) emerged as significantly higher than 1–60 Hz (r > 0.68, all adjusted p < 0.05).