Direct effects of transcranial electric stimulation on brain circuits in rats and humans

Transcranial electric stimulation is a non-invasive tool that can influence brain activity; however, the parameters necessary to affect local circuits in vivo remain to be explored. Here, we report that in rodents and human cadaver brains, ~75% of scalp-applied currents are attenuated by soft tissue and skull. Using intracellular and extracellular recordings in rats, we find that at least 1 mV/mm voltage gradient is necessary to affect neuronal spiking and subthreshold currents. We designed an ‘intersectional short pulse’ stimulation method to inject sufficiently high current intensities into the brain, while keeping the charge density and sensation on the scalp surface relatively low. We verify the regional specificity of this novel method in rodents; in humans, we demonstrate how it affects the amplitude of simultaneously recorded EEG alpha waves. Our combined results establish that neuronal circuits are instantaneously affected by intensity currents that are higher than those used in conventional protocols.


Figure 5d
Sine 2 NA (3)  Notes below explain the rationale for the choice of stimulation parameters in different experiments.
(1) In contrast to the human measurements, neuron stability and brain state changes limited the duration of the experiment available for rodent data collection. In the intracellular experiments, we chose DC stimulation (right side was the cathode) because artifact issues are easier to deal with DC stimulation (only onset and offset artifacts had to be removed). In experiments where sine waves were used, the positive half of the sine wave corresponded to left-anodal stimulation, while the negative half to the right-anode. (2) Each trial consisted of 3 x 2.5 µs pulses repeated at 133 kHz (100% duty cycle) for 500 ms and followed by 1 s pause (see also Supplementary Figure 2a). (3) To achieve high signal-to-noise ratio in the cadaver experiments, we first identified the strongest stimulation intensity, which did not saturate the amplifier. This variability across brains should not affect our results since we demonstrate a nearly perfect linear correlation between with the applied intensity and induced fields (Figure 4). In a subset of experiments, we used the same intensity (1 mA) for scalp, skull and brain surface stimulation ( Figure 5). (4) In the human ISP experiments, 6 x 10 µs pulses were repeated at 16.66 kHz (100% duty cycle).
The amplitude of the pulses was modulated by a 1-Hz sine wave, linearly ramping up from zero to maximum in 6 seconds, then ramping down to zero in 6 seconds (see also Supplementary  Figure 2b). After each 5-min stimulation epoch, the subjects were asked: did you see 'sparks'? Did you feel dizzy? Did you feel taste in your mouth? What was it like? Subjects were not asked to give a magnitude but they spontaneously reported mild, moderate and strong dizziness.
In three subjects (III/2, III/4 and III/6 -eyes open sessions) the protocol was repeated while they were asked to keep the eyes open for both the stimulation and control periods. These sessions were used to further estimate the severity of dizziness during ISP and Shuffled ISP protocols only, and were not included in the analysis of EEG. Note that all three subjects reported more severe dizziness compared to the eyes closed sessions, and in one occasion an altered peripheral vision. These side effects were completely absent when the Shuffled ISP protocol was applied.
* This subject was excluded from the analysis due to excessive electrical artefacts.
**At the termination of both ISP and Shuffled stimulations the subjects were asked: Which epoch was more unpleasant, the first or the second? *** This subject reported altered peripheral vision at 2 mA ISP but not at 7 mA.
Subjective discomfort varied from mild sensation to burning feeling of the scalp but its magnitude was not quantified.
Supplementary Figure 1. Circuit schematics of ISP Stimulator and artefact removal. (a) Left: Schematics of fast-pulse ground-independent signal-splitter circuit for one electrode pair. Driver TTL signal is generated by an external pulse generator, which is advancing a decade counter. Counter is driving six identical bipolar switching modules, each built of four phototransistors. Right: An alternative solution used a microcontroller (Microchip PIC18F4525) and digital isolators (ADuM1400) that allow more flexibility of stimulation patterns. Ground-independent switching is performed by high-speed analog switches (ADG412) instead of phototransistors. (b) EEG traces during ISP stimulation. Example trace showing EEG recording before (top left trace) and after artefact removal (bottom trace). Right panels: corresponding power spectra of EEG traces shown on the left. Stimulus intensity = 7 mA. (c) Alpha-band filtered EEG signals recorded by the left and right occipital leads (left panels). Note that the phase and amplitude of alpha waves in the two hemispheres vary relatively independently from each other under both control and ISP stimulation (2 mA and 9 mA) conditions, implying that the traces are free of common electrical artefacts. Timelag of cross-correlogram peaks is also similar under control and ISP stimulation conditions (red vertical bars denote correlogram peak and trough of the 0 mA condition for better visibility). Instantaneous frequencies of the EEG traces from the two hemispheres vary from event to event (Pearson's linear correlation; R= -0.0024, 0.01 and -0.004; P = 0.89, 0.57, 0.82; n=3053, 3092 and 3098 from a single subject, at 0, 2 and 9 mA intensities, respectively). Note that stimulation-induced artefacts are expected to have constant phase and amplitude ratios at all recording positions. Full data distributions are shown on the scatter plots.
Supplementary Figure 2. Illustration of ISP protocol.
(a) Upper part shows the schematics of the recording and stimulating setup in rodent experiments. Neuronal activity was recorded from both hemispheres simultaneously (white circles corresponds the location of craniotomies). The ISP was alternatingly focused to left or right hemisphere in an interwoven fashion, so that neurons in the left (or right) hemisphere were more strongly modulated by ISP focused to the left (or right) hemisphere. Spiking activity of some neurons contralateral to the focused hemisphere was suppressed, possibly because of the opposite geometric orientation of neurons compared to the ipsilateral hemisphere (note the different orientation of the schematic neuron in the white circle). Lower part shows the schematics of the stimulation sequence for two consecutive trials. Each trial consisted of 3 x 2.5 µs pulses repeated at 133 kHz (100% duty cycle) for 500 ms and followed by 1 s pause (Supplementary Table 3). Note that the duration of the 2.5 µs pulses are shown disproportionally longer for better visibility. (b) Upper part shows the position of the recording (P3 and P4) and stimulating electrodes in human measurements. Lower part shows a single trial, which consisted of 6 x 10 µs pulses repeated at 16.66 kHz (100% duty cycle). The amplitude of the pulses was modulated by a 1-Hz sine wave, linearly ramping up from zero to maximum in 6 seconds, then ramping down to zero in 6 seconds. Please note that the length of the 10 µs pulses and the ramping time are shown disproportionally for better visibility.
Supplementary Single session example shows distribution maps with voltage mode (upper row, 1, 3 and 5 V from left to right, respectively), and current mode (lower row, 50, 100 and 150 μA intensities from left to right, respectively). Intensities for current mode stimulation were chosen to match the calculated current intensity in voltage mode stimulation sessions (see Materials and Methods). Note identical distribution maps. (b) Equivalent circuit schematic for the application of multiple independent stimulating pairs in an intersectional arrangement, resembling gamma-ray radiosurgery. Note that due to the common conductive medium, the currents from the two stimulators couple serially, mimicking the effect of one large surface electrode pair and/or increased stimulus intensity, but they don't reach spatial selectivity (c) As predicted from the model on panel b, shifting the relative phase of the sinusoidal stimuli from two independent stimulator pairs reduce the induced field in both the horizontal and coronal planes. Alpha power band is highlighted by squares. Each panel shows the difference between stimulation (5 min) and the preceding nonstimulated 1-min long control periods (as shown in Figure 8a). Note that anodal stimulation of the respective hemisphere at 6 mA ISP strongly increases alpha power, while shuffled ISP stimulation is much less effective (for quantification and statistics, see also Figure 8b). Power in the beta band is also increased by stimulation. Based on n = 6 subjects. Abdominal wall stimulation (6 mA ISP) had no effect (n = 2 subjects).