Stimulator and tissue phantom platform

To evaluate the effects of tTIS, we developed a temporal interference stimulator capable of generating currents with specific frequencies and intensities (Fig. 1A, see supplementary materials for details). In addition, we used a tissue phantom within a cylindrical configuration to mimic the conductivity properties of head tissue (Figs. 1B and 2A).

Fig. 1

Overview. A Design of the tTIS stimulator. B Schematic of phantom. C Visualization of a finite element model of a mouse brain. D Schematic of the experimental setup. tTIS is applied to head-fixed mice. Meanwhile, the value of Ca2+ signals in the SC is recorded by fiber photometry, and pupil activity is recorded using a macro lens. E Two pairs of tTIS electrodes stimulate a conscious head-fixed mouse with currents I₁ and I2. F Representative image of GCaMP6s virus expression in the SC of a C57BL/6J mouse (green, GCaMP6s; scale bar, 200 μm). G Mean values of eye movement amplitudes (upper green curve) and Ca2+ signals in the SC (middle red curve) in three mice during tTIS (carrier frequency: 2000 Hz, difference frequency: 1 Hz, current: 1 mA; lower blue curve).

Fig. 2

The Experimental Protocol. A Experimental Equipment of the Phantom: The tissue model is constructed using a petri dish with a diameter of 90 mm. The tissue model is filled with NaCl solution, and the concentration is adjusted so that the impedance between each pair of copper electrodes is 3 kΩ. B Flowchart of the Computational Model: Preparation: An FEM model with 5.8 million voxels is used. The electrode positions are located on the model according to the experimental design, and the Leadfield matrix (LFM) of the brain (811,011 voxels) is calculated. Stimulation: The electrical field strength at the corresponding ROI under different stimulation parameters is calculated. The results are linearly interpolated for better visualization. C Flowchart of Animal Experiments: Preparation: Virus injection and fiber optic implantation in half of the mice. One week later, electrode fixation surgery is performed on all mice. Acclimation: After two days, all mice are acclimated for three days (~30 min/day) to get used to the head immobilization system. Data Collection: Ca2+ signals and eye movements are recorded 5 s before and 10 s during stimulation.

The tTIS stimulator consists of two independent and controllable constant-current sources, each of which can output a current of 0–5 mA at a frequency of 0–10 kHz with an adjustment accuracy of 0.1 Hz. This source comprises a direct digital synthesis signal generator, an array of constant-current feedback circuits, an inverse reference circuit, and various control circuits (Fig. S1A, B). Such a design ensures a stable current output across a range of loads. The stimulator’s performance underwent extensive testing under different loads, validating its ability to consistently output current and effectively adjust frequencies, even with varying load resistances (Fig. S1C–E).

We then measured the amplitude of interferential electric field envelope modulation using a tissue phantom. This phantom, constructed from a 90 mm diameter dish, features two pairs of copper electrodes symmetrically mounted on the edge of the dish, connected to the tTIS stimulator. Filled with a sodium chloride solution, the salt concentration was adjusted to achieve an interelectrode impedance of 3 kΩ at 2 kHz and 1 mA of alternating current. We also examined the impact of electrode size on the measurements.

The electrical field within the phantom was ascertained using two orthogonal dipole electrodes, spaced 5 mm apart and made from medical stainless-steel needle electrodes. These electrodes were precisely positioned within the phantom using a stereotaxic instrument, advancing in 5-mm steps. The resulting signals were captured on an oscilloscope to record the envelope waveform. To enhance accuracy, measurements at each location were averaged thrice, reducing noise influence. For electrical field mapping, the MatLab interp2 function was applied to linearly interpolate the measurement points, thereby generating comprehensive electrical field maps.

Simulation-based on the finite element method

We applied the finite element method (FEM) to simulate the electrical fields generated by transcranial electrical stimulation (Fig. 1C). The FEM model was built based on Digimouse [27], a whole-body computed tomography dataset of a mouse with a resolution of 0.1 mm × 0.1 mm × 0.1 mm by Alekseichik et al. [28]. To establish the experimental setup, we utilized a pair of copper pins 1 mm in diameter as skull electrodes. Each skull electrode was paired with a 10-mm diameter electrode attached to the ipsilateral cheek. Notably, the needle electrodes did not penetrate the skull, while the cheek electrodes were 5 mm thick saline electrodes on the skin. All electrodes were positioned in the same manner as in the animal experiments. The patch electrode that fitted the cheek was added to the model using SimNIBS 3.2 [29]. Itk-snap [30] and iso2mesh [31] were used to implement the operations of targeting and modifying the FEM model. The model was then converted to segmentation and incorporated the copper pins using Itk-snap [30]. Lastly, the tetrahedral-based FEM model with 5,298,777 tetrahedral elements was loaded using Gmsh [32] and exported as a Nastran Bulk Data File for further processing (Fig. 2B). COMSOL Multiphysics 4.3 (COMSOL, Inc., Burlington, MA) was used to set conductivities and calculate physics equations. Conductivities were defined as follows (in S/m): white matter and grey matter, 0.126; CSF, 1.654; bone, 0.01; scalp, 0.465; eyeballs, 0.5; copper, 6.7; silicone rubber, 29.4; saline, 1.0. The type of study was steady-state current studies and the initial state electric field was set to zero. Based on the model, we generated a lead field matrix (LFM), which served as an intermediary for calculating the field distribution. The LFM enabled the computation of the field generated by any given current injections. Considering that our analysis was limited to the brain area and involved four channels, the resulting matrix had a shape of [4, 811011, 3], where 3 represents the three orthogonal directions.

We performed three simulation experiments, involving the application of 2 mA of both tTIS and tACS. For tACS, the electrical field was directly obtained by injecting current into either two needle electrodes or two pairs of skull-cheek electrodes. On the other hand, to obtain the electrical field distribution for tTIS, we calculated the results for the one-pair skull-cheek electrodes separately. When two high-frequency currents intersect within the brain, they generate interference, resulting in a low-frequency envelope due to the beat frequency phenomenon. Individually, each frequency is too high to drive neural firing effectively [15, 33, 34]. However, this low-frequency beat can modulate neuronal activity. Subsequently, we determined the envelope of the electrical field distribution using the equation (1) as described in [33]:

$$\left|}_\left(\overrightarrow\right)\right|=2\cdot \text\left(\left|}_\left(\overrightarrow\right)\right|,\left|}_\left(\overrightarrow\right)\right|\right)$$

where \(E\left(\overrightarrow\right)\) is the electrical field at the location r, and E1 and E2 are the electrical fields from one pair of electrodes.

Animal experimentsExperimental preparation

All procedures were approved by the Animal Care and Use Committees of the Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences. An overview of the animal experiments is summarized in Figure 2C. Six adult (4 to 5 months old) male C57BL/6J mice were used. All mice were maintained on a 12/12-h light/dark cycle at 25 °C. Food and water were provided ad libitum. The AAV-syn-GCaMP6s virus was used in the fiber photometry experiments. C57BL/6J mice were anesthetized with pentobarbital (1% m/v, 10 mL/kg) and fixed on a stereotaxic apparatus (RWD, Shenzhen, China). During the surgery, the mice were anesthetized with isoflurane (1.5%) and placed on a heating pad to maintain body temperature at 35 °C. A 10 μL microsyringe with a 33-Ga needle was connected to a microliter syringe pump (UMP3/Micro4; WPI, USA) and used for virus injection into the SC (coordinates (in mm): AP, −4; ML, 0.8; DV, −1.6). A 200-μm optical fiber (NA: 0.37; Newdoon, China) was chronically implanted in the SC (coordinates: AP, −4; ML, 0.8; DV, −1.4) of C57BL/6J mice 2–3 weeks following virus expression for fiber photometry experiments. Then, a nylon head plate was implanted into the skull to fix the head. After surgery, all mice were allowed to recover for at least 1 week.

After virus injection and optical fiber implantation, we performed a third operation on C57BL/6J mice (n = 3). Moreover, C57BL/6J mice (n = 3) without virus injection and optical fiber implantation were subjected to the same procedure. The mice were anesthetized with 4% isoflurane maintained at 1.5%. Erythromycin eye ointment was applied to the eyes. The scalp and face were shaved and disinfected with 70% ethanol. We used two stainless-steel cranial pins (0.8 mm × 4 mm) as electrodes, which were fixed on the skull with dental acrylic. The skull electrodes were located at stereotactic coordinates (relative to bregma) of 4 mm anteroposterior, 2.5 and −1.7 mm mediolateral. Two days later, head-fixed awake mice were habituated to restraint for three consecutive days for 30 min each. During this time, the animals were fixed on top of a polystyrene foam ball. They could remain stationary or run and try to turn the ball forward or backward. According to experimental observations, the mice were largely at rest throughout both the experimental period and the application of electrical stimulation. To prevent mice from getting fatigued, each experiment was completed within 2 h.

Histology and microscopy assessments

After the final experiment on each mouse, 1% pentobarbital sodium (10 mL/kg) was injected intraperitoneally for deep anesthesia. Each mouse was transcardially perfused with 1 mol/L phosphate-buffered saline (PBS), followed by ice-cold 4% paraformaldehyde (PFA) in 1 mol/L PBS. The brain was removed and submerged in 4% PFA at 4 °C overnight, post-fixed, and then transferred to 30% sucrose to equilibrate. Coronal brain sections (30 μm) were cut on a cryostat (CM1950; Leica, Germany). The sections were collected and stored in 24 well plates in turn. PBS solution (0.01 mol/L) was added to the plates for short-term storage. The sections of the SC brain area were stained using immunofluorescence as follows: the sections were rinsed with 1 mol/L PBS three times (5 min/time), then incubated with 1:10000 DAPI dye solution for 10 min, and then rinsed with 1 mol/L PBS three times (5 min/time). Next, the sections were attached to the slides, and the anti-fluorescence quenching agent was used to seal the slides. Finally, the Olympus VS120 virtual microscope slide scanning system was used for image observation. The location of the ROI was traced by referring to the mouse whole-brain atlas [35].

Neural recording and eye movement recording during tTIS

In this study, Ca2⁺ signals were recorded using a fiber photometry system (Inper, Hangzhou, China). Two weeks after the AAV-syn-GCaMP6s virus injection, an optical fiber (NA: 0.37; Newdoon, China) was implanted into the SC. Spontaneous nystagmus refers to those occurring in the absence of any external stimulation. Induced nystagmus, on the other hand, was recorded during the application of electrical stimulation, including changes in gaze direction, either horizontally or vertically (gaze-evoked nystagmus), or from head movements parallel to the plane of the horizontal semicircular canal (head-shaking nystagmus). It can also arise from other types of stimuli, such as vibration (vibration-induced nystagmus), hyperventilation (hyperventilation-induced nystagmus), or changes in posture (positional nystagmus) [36]. To distinguish between the two, we monitored the eye movements during baseline (no stimulation) and compared these to eye movements during electrical stimulation. Spontaneous movements occurred randomly, while induced movements were temporally correlated with the stimulation onset and showed a more consistent directionality and frequency pattern in response to the stimulation. The eye movements of the head-fixed mice were recorded by a 200 Hz macro lens (Point Gray FL3-U3-13EAM) with 1280 × 1024 resolution. The shooting distance was 30 cm, the magnification was 1.6, and the depth of field was >2 mm. We fixed a 940-nm high-power infrared light-emitting diode below the front of the eye for illumination (Fig. 1D). A high transmittance infrared filter was applied to remove the reflection of the infrared lamp on the eyeball for extracting pupil activity.

The head-fixed mice were stimulated using two skull electrodes. During tTIS, two electrically isolated currents I1​ and I2​ were applied transcranially via electrodes connected to the stimulator by thin silver wires (Fig. 1E). Current I1​ was applied via the skull electrode that was located at the coordinates AP −4 mm, ML 2.5 mm relative to bregma. Current I2​ was applied via the skull electrode located 4.2 mm laterally to the I1​ electrode (distance between centers of electrodes). Each skull electrode was paired with an 8 mm diameter cloth electrode attached to the ipsilateral cheek. The stimulation time was 20 s, repeated five times, with at least 30 s rest time between each trial. During transcranial stimulation, the two cranial electrodes were paired. In the time window of 60 s before and after stimulation, we recorded Ca2⁺ signals in the SC with an optical fiber recording system (Inper, Hangzhou, China), and simultaneously photographed the pupil position of mice with a macro lens. We recorded the pupil and Ca2⁺ signals of three mice with optical fiber implantation and the pupil data of three mice without optical fiber. A representative image of GCaMP6s virus expression in the SC is shown in Fig. 1F. During the non-stimulation period, Ca2⁺ signals and eye movements were minimal, so we used the first 5 s as control data. Upon initial stimulation, both signals and movements increased significantly but stabilized as the stimulation continued. Therefore, for data analysis, we used the first 10 s of stimulation as experimental data. The injected tTIS current, the recorded Ca2⁺ signals, and the induced pupil movements are shown in Fig. 1G.

Data analysis and statistics

First, InperDataProcess V0.2.3 (Inper, Hangzhou) was used to correct the baseline of the original data and reduce the photobleaching caused by long-term recording. We then subtracted the scaled 405-nm trace from the 470 nm trace to generate the corrected 470 nm signal [37]. Custom MatLab (The MathWorks Inc.) scripts were developed for further analysis using R2017a. Signals were analyzed as ΔF/F = (F−Fbaseline)/Fbaseline, where Fbaseline was defined as the baseline fluorescence within 5 s before stimulation. Then, a Gaussian function was used to smooth the data.

We used the DeepLabCut Toolbox to identify the center of the pupil in the videos [38]. The pupil displacement in the 2D image was converted to a rotation angle based on the estimated eyeball radius (1.67 mm) for adult C57BL/6 mice [39]. We also used a Gaussian function to smooth eye movement data. To analyze the synchronous Ca2⁺ signals, all eye movement data were down-sampled to 100 Hz. To quantify eye movement amplitudes, the initial eye position was determined as the average eye position in the 5 s window before tTIS, and the position was set to a 0° angle.

Statistical analyses and graph plotting were performed with Prism 8.0 (GraphPad Software) and MatLab 2017a (MathWorks). Pairwise conditional GC was computed using the multivariate GC MatLab Toolbox [40]. All values are presented as the mean ± SEM. Non-parametric Wilcoxon signed-rank tests were applied for two-group comparisons. *P <0.05, **P <0.01, and ***P <0.001.