The chloroplatinic acid (H₂PtCl₆) and lead acetate [Pb (CH₃COO)₂] were purchased from Sinopharm Chemical Reagent (China). 3,4-ethoxylenedioxythiophene (EDOT) was purchased from Aladdin (China). Poly(sodium-4-styrenesulfonate) (PSS) was obtained from Herochem (China). 2-methyl-2-thiazoline (2MT) was purchased from Shanghai Acmec Biochemical Co., Ltd (China). 1,1′-Dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) was purchased from Beyotime Institute of Biotechnology (China). The stereotaxic frame (51600) was purchased from Stoelting (USA). Micropositioner (model 2662) was purchased from David KOPF instrument (USA).

Six female mice (C57BL/6J) weighing 20–25 g were used in the experiment, which were purchased from Charles River Laboratory Animal Technology Co, Ltd. All mice were housed at 20°C with a 12-h light: 12-h dark reversed light cycle. All animal experiments were conducted with the permission of the Beijing Association on Laboratory Animal Care and approved by the Institutional Animal Care and Use Committee at Aerospace Information Research Institute, Chinese Academy of Science (AIRCAS).

The MEAs, based on micro-electro-mechanical system (MEMS) technology, were developed for electrophysiological recording in MA of freely behaving mice. The MA area is similar to an irregular triangle and is located at a depth of about 5 mm in the brain (Supplementary Figure S1A). The widest part of MA is only about 500 μm, and the longitudinal length is about 1,000 μm. In order to match the shape and size of the MA, there were eight electrode recording sites (15 μm in diameter) on the tip of every shank forming a 2 × 8 array (Supplementary Figure S1B). The MEAs were designed to consist of two 7-mm-long shanks. Each shank has a width of 180 μm, a thickness of 30 μm, and a shank-to-shank distance of 80 μm.

The MEA consists of three layers, namely, base layer Si (30 μm), metal layer Ti/Pt (30 nm/250 nm), and insulating layer SiO₂/Si₃N₄ (300 nm/500 nm) (Supplementary Figure S1B). The electrode manufacturing process flow is shown in Figure 1A. The fabrication process can be briefly divided into three parts: First, the Ti/Pt conductive layer pattern (including recording sites, bonding pads, and conducting wire) was formed by photolithography, sputtering, and lift-off. Then, SiO₂/Si₃N₄ insulating layer was deposited, and the recording sites and bonding pads were exposed by the second photolithography and CHF₃ reactive ion etching (RIE). Finally, each electrode is released by KOH wet etching.

To decrease the electrode impedance and improve the signal-to-noise ratio, PtNPs and PEDOT:PSS were successively electroplated onto the bare microelectrodes. The schematic diagram of PtNPs/PEDOT:PSS electrode modification is shown in Figure 1B. Chloroplatinic acid (48 mM in PBS) and lead acetate (4.2 mM in PBS) are mixed at a volume ratio of 1:1 and kept for 12 h to obtain a platinum (Pt) black electrolyte solution. EDOT solution (20 mM) was obtained by adding 0.0284 g of EDOT (20 mM) and 0.206 g of PSS (0.1 M) to 10 ml of deionized water and treating with ultrasound for 1 h. Electrode surface modification was completed by an electrochemical workstation (Gamry Reference 600, Gamry Instruments, USA). First, Pt black nanoparticles were electroplated by chronoamperometry (-0.8 V, 50 s) using a Pt wire as the counter electrode and then PEDOT by cyclic voltammetry (CV) (0–1 V, 8 cycle) using an Ag/AgCl as reference electrode and a Pt wire as the counter electrode.

Microscope and scanning electron microscope (SEM) were used to observe the surface morphology before and after electrode modification. Both CV and electrochemical impedance spectroscopy (EIS) were carried out in phosphate-buffered saline solution (PBS, 0.1M, PH = 7.4) with an Ag/AgCl as the reference electrode and a Pt wire as the counter electrode. Charge storage capacity (CSC) was measured by CV from −0.6 V to 0.8 V at a scanning rate of 0.1 V/s. Electrode stability was tested by CV sweeping from −0.6 V to 0.8 V (6,000 cycle) and ultrasonic oscillation (50 W, 100 min) in an ultrasonic cleaner.

The operating table and all surgical instruments are cleaned and disinfected in advance. The mice were anesthetized by the isoflurane (3% for induction and 1% for maintenance) and then fixed in the stereotaxic frame. Then, the mice scalp was cut off to expose the skull, and the head was adjusted to a level: one craniotomy for implanting MEA to MA (AP: −1.5 mm, ML: 2.0 mm) and the other three for skull nail as reference electrodes. The MEAs were advanced to a depth of 5.2 mm from the cortical surface at a speed of 5 μm/s using a micropositioner. Finally, the MEAs were fixed by dental cement, and the mice were put on the heating pad until woke up.

After the experiment, the mouse was injected with 10% chloral hydrate into deep anesthesia and perfused through the heart with 0.9% NaCl and 4% paraformaldehyde in sequence. The brain was removed and put into 20% sucrose solution for 24 h followed by 30% sucrose solution for 48 h for postfixation. Then, 40 μm thick sections were prepared with a freezing microtome and were observed under a microscope. According to the mouse brain atlas (Paxinos and Franklin, 2001), the electrode implantation position is mainly concentrated in MePD (posterodorsal part of MA). The schematic diagram of electrode positions of six experimental mice is shown in Supplementary Figure S2.

As shown in Figure 2A, the mice with MEAs were placed in a 20 × 20 × 16 cm (L × W × H) shielded box with a 2 × 2 cm paper on the bottom. The electrophysiological signals in MA were recorded in real time, and the behavior changes of mice were recorded simultaneously by a camera (Hikvision, China). The recording video frame rate is 25 Hz. Before the experiment, the mice were allowed to habit in the box for 15 min until they were free to go anywhere in the box. First, electrophysiological signals and behavior video were recorded for 5 min as a control group. Then, 5 μl of 2MT (25%) was added onto the bottom paper through the pipette, and the data were continuously recorded for 10 min.

In the experiment, we observed that the mice would exhibit thigmotaxis (a special behavior that prefers to be close to the wall) and even jump after dropping 2MT. This behavior would last about 2 min, and it is defined as the avoidance stage. Then, the mice exhibited strong freezing, and it is defined as the freezing stage. In order to facilitate statistics and analysis, we divided the entire behavior stage into the control stage, avoidance stage, and freezing stage, with durations of 5 min, 2 min, and 5 min, respectively (Figure 2B). In order to analyze the synchronous changes in mouse behavior and electrophysiological signals, the freezing and spike firing rate are computed in 30-s bins.

All data are expressed as mean ± SE. Significance analysis uses the one-way ANOVA or one-way repeated measures ANOVA test by Origin. The analysis of behavioral characteristics is completed with the help of behavioral analysis software EthoVision. Among them, freezing is defined as a state in which there is no movement except breathing-related movement and lasts for more than 1.5 s (Müller and Fendt, 2006). The time of freezing (in seconds) and the velocity (in centimeters/second) were collected with EthoVision. The freezing rate in this study is the percentage of freezing time to the total time.

Electrophysiological signals are obtained by plugging MEAs into the homemade electrophysiological recording system (Xu et al., 2018; He et al., 2021). The signal sampling rate is 30 K Hz. A high pass filter was applied at 250 Hz to obtain spikes, and a low pass filter was applied at 250 Hz to view the local field potential (LFP). The software written by our laboratory was used for spike sorting/clustering. Units were isolated based on waveform parameters including peak amplitude and principal components. Clusters containing similar valid waveforms were manually defined, and the single units did not contain a refractory period <1 ms. MANOVA test (p < 0.05) was used to validate the single-unit nature of spikes (Friend et al., 2015). Neuroexplorer (Plexon) was used for further analysis.

The detection performance of bare electrode is poor due to its large impedance. PtNPs modification is beneficial to reduce the impedance of the electrode and obtain a better signal-to-noise ratio. PEDOT modification can improve the biocompatibility and long-term signal detection ability. In this study, we first electroplated PtNPs on the electrode and then electroplated PEDOT:PSS on the PtNPs-modified electrode. The PtNPs/PEDOT:PSS-modified electrode is shown in Figure 3A. In order to compare the electrode performance of different modification methods, PEDOT:PSS, PtNPs, and PtNPs/PEDOT:PSS were plated on the bare electrode according to the method described earlier, and the morphology of modified electrode sites was characterized using a SEM, as shown in Figures 3B–D. The morphology of PEDOT:PSS modification is relatively flat. However, the specific surface area of the electrode did not increase significantly. The PtNPs-modified electrode formed an obvious particle bulge on the surface, increasing the specific surface area of the electrode. The better biocompatibility of PEDOT has been confirmed by many studies (Kozai et al., 2015; Pranti et al., 2018; Dijk et al., 2020). In order to ensure lower impedance and better biocompatibility, we electroplated PEDOT:PSS on PtNPs-modified electrode. The morphology of PtNPs/PEDOT:PSS-modified electrode is shown in Figure 3D. PEDOT fills the gap formed by PtNPS, like covering the surface of PtNPs with a thin film to make the electrode surface smoother.

As shown in Figure 4A, the average impedance of PtNPs/PEDOT:PSS-modified electrode is 27.0 ± 2.3 KΩ, which decreased by more than 90% compared with the bare electrode and was also only a quarter of the PEDOT-modified electrode. In terms of phase delay, the average phase delay of PtNPs/PEDOT:PSS-modified electrode at 1 kHz is about −12.30 ± 0.52° (Figure 4B), which is lower than that of PtNPs-modified electrode (-17.77 ± 2.00°). The average phase delay of PtNPs and PtNPs/PEDOT:PSS-modified electrodes at useful bandwidth for recording spikes (1–3 kHz) is shown in Supplementary Figure S3. CSC is another common indicator for electrode material characterization (Yamaguchi et al., 2016; Davidsen et al., 2019). We compared the CSC of the PtNPs and PtNPs/PEDOT:PSS-modified electrode. The results show that the CSC of PtNPs/PEDOT:PSS-modified electrode is almost twice that of PtNPs, which shows the good performance of PtNPs/PEDOT:PSS modification (Figure 4C). The electrode durability is particularly important for the electrophysiological measurement of freely behaving mice. In order to test the electrochemical and mechanical stability of the electrode, we performed the CV sweeping and ultrasonic oscillation on the electrodes (Figure 4D). After 6,000 cycles of CV sweeping, the impedance of PtNPs-modified electrodes increased significantly, but the impedance of PtNPs/PEDOT:PSS-modified electrodes hardly changed. In addition, the average impedance of PtNPs/PEDOT:PSS-modified electrodes was already less than that of PtNPs-modified electrodes. At the same time, the average phase delay of PtNPs-modified electrode increased to −50°, while PtNPs/PEDOT:PSS-modified electrode could still be maintained within −30° (Supplementary Figure S4). After 100 min ultrasonic oscillation, the impedance and phase delay of PtNPs-modified electrode increased further, while the phase delay of PtNPs/PEDOT:PSS electrode did not change significantly in this process. This proved the good stability of PtNPs/PEDOT:PSS modification.

In addition, we tested the background noise of PtNPs electrode and PtNPs/PEDOT:PSS electrode in PBS. The results show that PtNPs/PEDOT:PSS electrode has less background noise (Supplementary Figure S5A,B). In the electrophysiological test experiment, the PtNPs/PEDOT:PSS-modified electrode shows the characteristics of low noise and high signal-to-noise ratio. As shown in Supplementary Figure S5C, the noise of the signal measured by the modified electrode is about ± 15 μV, and the signal-to-noise ratio can reach about 10.

We recorded the electrophysiology of MA and behavior of mice simultaneously. In terms of behavior, the mice freely explore anywhere in the box in the control stage. After adding 2MT, the mice will walk along the edge of the box most of the time and exhibit thigmotaxis, an index of increased anxiety (Kennedy et al., 2020). In the avoidance stage, the mice will want to approach and explore the odor source, but they are more likely to walk rapidly along the wall of the behavior box or freeze, and even individual mice exhibit jump. The mice's movement velocity increased and decreased sharply, indicating that the mice were panicked (Figures 5A,B). After entering the freezing stage, the mice curled up in a corner and showed strong freezing defense behavior. The velocity of mice also reached the lowest (refer to Supplementary Materials for video examples). In order to confirm the reliability of the results, we replaced 2MT with deionized water in the control experiment. The results showed that the dripping water had no effect on the behavior of mice, which indicated that the defense behavior of mice was induced by 2MT (Supplementary Figure S6).

In terms of electrophysiology, the spike firing rate increased significantly in the avoidance and freezing stages, which reflected that the neurons in MA were activated after fear (Figure 5C). The LFP also became more intense and different. It can be seen from the enlarged figure at the bottom that the LFP amplitude becomes larger in the avoidance stage. In the freezing stage, in addition to the larger amplitude compared with the control stage, there are more dense spines in the LFP waveform, which may be caused by the increase in action potential.

We also counted the freezing rate, spike firing rate, and LFP power to quantitatively analyze the three stages of fear. It can be seen from Figure 6A that the average freezing rate of six mice is <10% in the control stage. Then, the freezing rate gradually increased in the avoidance and freezing stages, indicating that 2MT induced a strong defense response in mice. At the same time, we counted the action potential firing rate of 69 neurons in six mice. We found that the firing rate and the freezing rate had the same trend in the three stages, and the firing rate in the freezing stage was significantly different from that in the control stage (Figure 6B). The LFP power is mainly concentrated in the low-frequency band, and we counted the LFP power of four different frequency bands, namely, δ (1–4 Hz), θ (4–8 Hz), α (8–13 Hz), and β (13–30 Hz) (Figure 6C). We found that the LFP power also has a similar trend, that is, the δ and θ band powers in the avoidance stage and freezing stage are significantly greater than that in the control stage. The difference is that the LFP power in the avoidance stage is higher than that in the freezing stage. We suspect that the avoidance stage may involve a variety of behaviors and psychology of mice: avoidance, flight, exploration, and freezing. This leads to more neurons participating in it, which increased LFP power. In addition, there was no significant change in neural activity after dropping the same volume of water, as shown in Supplementary Figure S7.

We differentiated neurons by waveforms into two classes, namely, type 1 and type 2, according to the principles commonly applied to cortical extracellular recordings. Specifically, we extracted two features of spike waveform, namely, symmetry and pulse width, and combined them with the clustering algorithm to divide the detected neurons into type 1 neurons and type 2 neurons (Barthó et al., 2004; Riera et al., 2014; English et al., 2017) (the parameters for this calculation are shown in Supplementary Figure S8). The classification results are shown in Figure 7A. Among the 69 neurons, there were 46 type 1 neurons and 23 type 2 neurons. Supplementary Figure S9 shows the average waveforms of these 69 neurons. At the same time, we counted the proportion of neurons significantly activated in the freezing stage. Neurons whose firing rate was significantly higher (one-way ANOVA, p < 0.05) in the freezing stage than in the control stage were considered activated neurons. The results showed that nearly 87% of type 2 neurons and 37% of type 1 neurons were significantly activated (Figures 7B,C). Then, we calculated the firing rate of activated neurons in three stages. Compared with the control stage, the firing rate of type 1 neurons and type 2 neurons in the freezing stage increased by 283 and 240%, respectively (Figures 7D,E). In addition, we calculated the synchronous changes in the firing rate of different types of neurons and the freezing rate (bin = 30 s) (Supplementary Figure S10). It can be seen that the firing rate of activated neurons has the same trend as the freezing rate of mice (Supplementary Figures S10A,C). We then calculated the correlation between the firing rate of activated neurons and the freezing rate in mice. In addition, we found a good linear correlation between the firing rate of partially activated neurons and the freezing rate. Supplementary Figure S11 shows a linear fitting coefficient between the firing rates of all activated type 1 neurons (n = 17) and type 2 neurons (n = 20) with the freezing rate of mice. The average value of type 2 neurons fitting coefficient is greater than that of type 1 neurons. But there was no significant difference between them. The correlation coefficient distribution of activated type 1 neurons is relatively scattered, and the correlation coefficient distribution of type 2 neurons is relatively concentrated. The correlation coefficient between the firing rate of the two types of neurons and the freezing rate can reach up to 0.86 and 0.69 (Figures 7F,G).