NR2B is an essential subunit of NMDARs, plays a crucial role in neuronal excitability, and is associated with the learning and memory function of the nervous system. Increased expression or activation of NR2B could revert the learning and memory abilities of mice with impaired cognitive function [22,23]. To confirm the details of THz photon absorption, we calculated from molecular dynamics simulations that the absorption spectrum of the glutamate-binding pocket of NR2B shows a clear fingerprint peak near 34.5 ± 1 THz, while this main absorption peak was located for glutamate molecules (Fig. 2A). Notably, the THz photons corresponding to frequencies of ~34.5 THz are well outside the strong absorption range of the surrounding water molecules (the fingerprint peaks are approximately between 0 and 30 THz) [24]. The THz photons (close to 34.5 THz) are mainly selectively absorbed by the binding pocket of NR2B due to the limited absorption by water molecules, while glutamate molecules are largely unabsorbed.

Patients with PTSD often exhibit abnormal learning and memory abilities [25]. The imaging findings revealed a noteworthy decrease in hippocampal volume among individuals diagnosed with PTSD, resulting in an impairment of its cognitive functioning [26–28]. Here, we proved a new method to improve learning and memory abilities in PTSD rats. We used THz photons to irradiate the CA3 subregion of the hippocampus in animal brains by delivering THz photons through optical fibers and cannulas into the target brain area (Fig. 1). We determined the coordinates of this area from the brain atlas as (1.5, −2.16, −3.5) and fixed the optical fiber on the brain stereotaxic device to facilitate the control of the optical fiber to accurately reach the target position. We set 3 power densities at low, medium, and high levels (22, 44, and 88 W/cm²).

The single prolonged stress (SPS) method is a commonly used animal model of PTSD that can not only accurately simulate the clinical symptoms of PTSD patients but also show abnormalities in the hypothalamic–pituitary–adrenal (HPA) axis. It is postulated that SPS is an appropriate animal model of PTSD [29,30]. We developed an animal model of PTSD using SPS method and conducted experiments following the procedure shown in Fig. 2B. To investigate the effects of THz photon irradiation on cognitive impairment induced by SPS, we performed a novel object recognition test, a shuttler box test, a Y-maze test, and a Barnes maze test reflecting hippocampus-dependent learning and memory abilities. The results showed that the proportion of time animals in the model group spent exploring new objects was significantly reduced, while THz photon (34.5 THz) irradiation increased the proportion of time spent exploring new objects in PTSD rats (Fig. 2C, increased by about 10.0%, P < 0.001). However, not all irradiated rats performed this way. This effect could only be achieved under certain conditions (44 W/cm²). For the shuttler box test, after training, we recorded initiative and passive avoidance times and calculated the accuracy as a reflection of learning and memory function. The results showed that the accuracy of PTSD rats was significantly lower than that of healthy animals, and THz photon irradiation improved the accuracy by ~100% (Fig. 2D, P < 0.001). This means that THz photon (34.5 THz) irradiation improved working memory in PTSD rats. In these 2 tests, all rats in the sham group showed similar behaviors to the rats in the model group (P = 0.13.5; P = 0.2503), indicating that the method of embedding cannulas to establish THz photon radiation did not affect the learning and memory ability of rats. Subsequently, we used the Barnes maze test and Y-maze test to further explore the effect of THz photon irradiation on the learning and memory ability of animals. The results showed that in the Barnes maze, the numbers of field entries in the model group were significantly higher than those in the healthy-control group and lower than those in the medium-power-density group (Fig. 2E; was about 5 times lower, P = 0.001; P = 0.0005). Similar results were observed in the Y-maze test, where the proportion of time for PTSD animals to explore new arms was significantly reduced, while the proportion of time for animals to explore new arms significantly recovered after medium-power-density THz irradiation (Fig. 2F; increased by about 40%, P = 0.0329; P = 0.0357). These 4 behavioral experiments all demonstrated that 34.5-THz photon irradiation of the hippocampal CA3 subregion could effectively recover the learning and memory abilities of PTSD animals.

Based on the results of behavioral, we further investigated the brain regions after THz photon irradiation, including the detection of glutamate concentration in the hippocampal CA3 subregion tissue and the assessment of NR2B phosphorylation levels. First, we detected the concentration of glutamate (glu) in the hippocampal CA3 subregion tissue and medial prefrontal cortex (mPFC) tissue. The results showed that there was a significant decrease in glu concentration in the hippocampal CA3 subregion tissue but an increase in glu concentration in the mPFC tissue (Fig. 3A and B; P = 0.004; P = 0.0385). Glutamate bound to NR2B promoted the phosphorylation level of NR2B. Therefore, detecting the phosphorylation level of NR2B could serve as an indicator of the binding ability between glutamate and NR2B. The western blotting results showed that there was increased phosphorylated NR2B (p-NR2B, increased by about 50%) and total NR2B (increased by about 40%) (Fig. 3, C to F; P = 0.0439; P = 0.0115) in the hippocampal CA3 subregion. This result suggested that the decreased glutamate concentration in the hippocampal CA3 subregion was caused by the enhanced binding of glutamate to NR2B by THz photon irradiation. Moreover, glutamate binding to NR2B promoted the phosphorylation of NR2B.

Furthermore, patients with PTSD often exhibit alterations in HPA axis activity, and changes in cortisol are associated with symptoms of PTSD [31]. We detected changes in hormone levels associated with the HPA axis in the peripheral blood. The results showed that the adrenocorticotropic hormone (ACTH) levels of PTSD rats were significantly increased by approximately 200% and were restored to normal levels after THz irradiation (Fig. 3G; P = 0.0002; P = 0.0007). The results showed a significant decrease in corticosterone (CORT) levels in peripheral blood in the model group, and THz irradiation also restored CORT levels (Fig. 3H; P < 0.0001; P = 0.0016).

To investigate the efficiency of THz photons (34.5 THz) on the utilization of glutamate molecules by the NR2B receptor, we compared the effect of THz photons stimulation on the binding free energy (G) and the binding interaction modes between NR2B and glutamate using molecular dynamics methods. The glutamate receptor (binding core region) forms nonbinding interactions with glutamate (surface shape) mainly through several sets of key amino acids and has a typical pocket structure, as shown in Fig. 4A. At room temperature (300 K), after kinetic simulations for ~100 ns to reach thermodynamic equilibrium, 4 stable hydrogen bonds form between amino acids Tyr186, Gln78, Glu111, and Tyr32 within the binding pocket, and 6 sets of amino acids form hydrophobic interactions. Interestingly, as shown in Fig. 4B, we found that 2 new hydrogen-bonding interactions between amino acids Gln78 and Thr187 were added after applying THz photons (frequency ν = 34.5 THz, intensity A = 0.5 V/nm), which resulted in a decrease in the free energy of glutamate receptor-glutamate binding by ~31.98 kJ/mol, indicating that the THz photons enhanced the binding capacity between NR2B and glutamate (Fig. 4C). In addition, we found that the enhancement of the binding capacity by THz photons increases with increasing intensity A from 0 to 2.0 V/nm. Furthermore, we calculated the energy contribution to the difference in binding free energy (ΔG) before and after THz photon exposure and found that it was mainly the electrostatic component that produced a change of ~32.4 kJ/mol, which was related to that of the hydrogen-bonding effect in Fig. 4D. We finally also assigned contributions to the binding free energy differences (ΔG) of the constituent amino acids of the NR2B receptor and found that it was mainly the amino acids Gln78 and Thr187 that produced relatively large changes in ΔG (Gln78) = 18.27 and ΔG (Thr187) = 11.27 kJ/mol, respectively (Fig. 4E). Therefore, our simulation suggests that THz photons (34.5 THz) have resonance absorption effects on the binding pocket of the NR2B receptor and then enhance the utilization of glutamate.

Under different power densities, heat may be generated by irradiation. We used a temperature-sensitive probe to detect the temperature in the irradiated region. We inserted the optical fiber into the brain at a depth of 3.5 mm and inserted the temperature-sensitive probe into the brain at a 37° angle with the optical fiber at a depth of 5 mm. The end of the probe was located 0.5 mm from the end of the optical fiber (Fig. 5A and B). The measurement results showed that when the laser started to irradiate, the tissue temperature began to rise, and within 10 min, the temperature rise in the low and medium-power-density groups was not significant (ΔT1 = 0.7 °C, ΔT2 = 0.6 °C), while the temperature increase in the high-power group was more significant, reaching 2 °C. Referring to the research results of predecessors, under irradiation with visible light or mid-infrared light, the temperature of the brain will increase to a certain extent (~2 °C), which is similar to our results (Fig. 5C), and the increase in this degree will not have an effect, inhibitory or otherwise, on the brain [16,32–34]. From this, we obtained a method that can perform THz photon irradiation on specific brain regions of animals, and this method has no significant thermal effect.

We also investigated whether THz photon irradiation causes damage to neurons. We designed a light path for neuronal irradiation (Fig. 5D). The beam emitted by the laser was reflected by a total mirror, changing its direction vertically upwards. Then, a lens was used for focusing, and a horizontal scanning displacement platform was placed at the focal point. The IBIDI dish with cultured neurons was placed here for scanning and irradiation. Four sets of irradiation parameters were set: 53.6 THz-3.6 W/cm², 53.6 THz-6.6 W/cm², 44.5 THz-3.6 W/cm², and 44.5 THz-6.6 W/cm². After 24 h of irradiation, an MTS (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt) reagent kit was used to detect cell activity. The results showed that there was no change in cell viability in any of the 4 groups (Fig. 5E). We then demonstrated the variation in NR2B expression after irradiation by THz photons. The results showed that 34.5 THz-6.6 W/cm² THz photon irradiation increased the expression of NR2B; however, changing the photons length of power did not achieve this effect (Fig. 5F; P = 0.0447). The results indicated that THz photon (34.5 THz) irradiation can increase the expression of NR2B in neurons cultured in vitro. This result indicated that THz photon irradiation did not cause damage to neurons. Thus, these conditions of THz photon irradiation were neuro-safe.

In the in vivo experiments, the laser used was a 34.5-THz frequency QCL (frequency 34.5 THz, pulse duration 2 μs, repetition rate 200 kHz, duty cycle 40%). Once the THz photons emitted by the QCL passed through the THz lens and was focused, it was then coupled into the fiber by using a 3-dimensional mobile platform. The optical fiber was standard polycrystalline infrared AgCl:AgBr polycrystalline fiber (PIR240/300, art photonics) with a core diameter of 240 ± 15 μm, a clad diameter of 300 + 0/−15 μm, and a numerical aperture of 0.3 ± 0.03. To precisely target a specific brain area, a stereotaxic apparatus was utilized to implant a metal tube at a predetermined position (±1.5 mm, −2.16 mm, −3.5 mm, determined according to the rat brain atlas) in the animal's head. The metal cannula was purchased from RWD Life Science Co., Ltd (RWD, 62049), and the inner and outer diameters is 0.45 and 0.64 mm, respectively. After the cannula was fixed, the optical fiber was then inserted into the metal tube, guided by the stereotaxic apparatus, to reach the designated brain area (3.5 mm deep into the brain). This setup enabled precise irradiation to be performed on the specific brain region with accuracy.

Tissue from the CA3 subregion was isolated for assaying. Tissues and cells were lysed by radioimmunoprecipitation assay lysis buffer, and total protein was extracted for western blotting experiments. The primary antibodies were NR2B (Cell Signaling Technology, 14544s) and glyceraldehyde phosphate dehydrogenase (Cell Signaling Technology, 5174s), and the secondary antibody was horseradish peroxidase-linked anti-rabbit immunoglobulin G (Cell Signaling Technology, 7074s). After 24 h of irradiation, the cell activity was detected using the MTS reagent kit (Promega, G3581).

Sprague-Dawley rats were purchased from Sibeifu (Beijing) Biotechnology Co., Ltd. All rats were randomly divided into cages and fed freely every day, with 12 h of light exposure per day. After 1 week of adaptive feeding, the experiment was conducted. In the experiment, animals were divided into a healthy-control group, a model group, a sham surgery group, a low-power-density group (22 W/cm²), a medium-power-density group (44 W/cm²), and a high-power-density group (88 W/cm²). Except for the healthy-control group, all other groups underwent modeling operations. Modeling was performed by adopting the SPS method. The animal was first restrained for 2 h, followed by forced swimming for 10 min. After being removed and wiped dry, each animal was anesthetized with isoflurane. After the animal completely lost consciousness, it continued to be anesthetized for 15 min. Then, the animal was placed back in the breeding cage and allowed to wake up. After the modeling was completed, the animals were kept undisturbed for 6 days. After the establishment of the animal model, the animals were subjected to tube embedding surgery, and the coordinates of the hippocampal CA3 area (1.5, −2.16, −3.5) were determined based on the brain atlas. Double cannulas were customized according to these coordinates, and the cannula was inserted at the predetermined location using a brain stereotaxic locator. The cannula was fixed with dental cement. Three days after surgery, THz radiation treatment was started, and the animals were anesthetized for radiation. Each side of the hippocampus was irradiated for 10 min once a day for 3 days. Behavioral experiments were conducted on animals after irradiation. All animal experiments had undergone animal ethics review and approved by Institutional Animal Care and Use Committee (review number: IACUC-DWZX-2021-737).

The learning and memory abilities of the animals were tested by behavioral experiments, including novel object detection, shuttle box experiments, Barnes maze, and Y-maze experiments. The novel object detection experiment tests the situational memory ability of animals. In the novel object detection experiment, first, the animals were placed into the experimental box to adapt for 5 min. Then, 2 objects were placed into the experimental box, and the animals were familiarized with them for 5 min. After 2 h, one of the objects was replaced with a new object, and the animal was placed into the box for testing. A camera was used to record the animal’s behavior, and RWD analysis software was used for analysis.

The shuttle box experiment tests the working memory ability of animals. In this experiment, the animals were first given acousto-optic combined stimulation and then given acousto-optic stimulation for 10 s, followed by electric stimulation for 10 s. The 2 stimuli were combined for training. The behavior of animals escaping to the safety area after receiving acousto-optic stimulation was recorded as active escape, and the behavior of animals escaping to the safety area after receiving electric stimulation was recorded as passive escape. The animals were trained 20 times a day, with a 30-s interval between each training, and the number of active evasions was observed. When the number of active evasions of healthy animals reached at least 18, testing of the animals began. Finally, the active avoidance rate of animals was recorded as the evaluation standard of the working memory ability of the animals.

The Barnes maze is an improved version of the water maze experiment, which tests the spatial memory ability of animals. The experimental maze consisted of a circular platform with a circle of holes with a diameter of 10 cm around the platform. One of the holes is fixed as the target hole. The animals were first placed in the maze to adapt for 5 min, and after an interval of 2 h, the animals were placed in the maze to train them to find the target hole; the time to find the target hole and the number of times exploring nontarget holes were recorded. The animals were trained twice a day, and when the number of times the healthy-group animals explored nontarget holes was less than 2, training was stopped, and testing was started.

Eighteen-day-old pregnant mice were purchased from Sibeifu Biological Co., Ltd., and their fetuses were used for primary neuronal culture. IBIDI Petri dishes were used to culture the isolated neurons. First, Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 medium (Gibco, 11330057) containing 10% fetal bovine serum (Gibco, 10099141C) was cultured in a CO₂ incubator for 4 h and then changed to neurobasal medium containing B27 (Gibco, 17504001), glutamine (Gibco, 25030081), and penicillin-streptomycin solution (Gibco, 15140122). The medium was changed every 2 days. After 7 d of cell culture, irradiation began. The lighting conditions were 53.6 THz-3.6 W/cm², 53.6 THz-6.6 W/cm², 34.5 THz-3.6 W/cm², and 34.5 THz-6.6 W/cm². The movement speed of the displacement table was 0.2 mm/s, and the scanning area was a square with a side length of 4.4 cm. It was determined that all areas with cells were irradiated. After irradiation, the cells were placed in an incubator and cultured for 24 h.

In PTSD animals with HPA axis disorder, ACTH and CORT were selected for detection (Cloud-clone, CEA836Ra; CEA540Ge). Under anesthesia, peripheral blood was extracted from the animal using the abdominal aorta blood collection method, and plasma was separated for use in the experiment. An enzyme-linked immunosorbent assay (ELISA) detection kit was used to detect ACTH and CORT levels. In addition, ELISA kits were also used to detect Glu levels in hippocampal CA3 tissue (Cloud-clone, CES122Ge). The animal’s hippocampal CA3 tissue was isolated for homogenization, and ELISA experiments were performed on the tissue homogenates.

Our simulations were designed to investigate the molecular dynamics process of glutamate utilization by the NR2B receptor and to compare the binding efficiency supplied by THz photons. Our model is based on the NR2B receptor core model [24] (Protein Data Bank ID: 1us4) and is initially placed in the center of a cubic simulation box with a side length of 8.0 nm. In addition, the simulation box was filled with 14,515 water molecules (SPCE model), 41 potassium ions, and 45 chloride ions (i.e., salt concentration: 0.15 M) to maintain electrical neutrality of the whole system. We used the Charmm-36 force field and periodic boundary conditions [47], the Ewald connected element algorithm for electrostatic treatment [48], and the Velocity-Verlet algorithm [49] for solving the equations of motion. The truncation of the Lennard–Jones interaction and the real space part of the Ewald sum were 1.0 nm, and the convergence factor of the Ewald sum was 1.20 nm. The simulated system undergoes temperature and pressure equilibration at room temperature (300 K), resulting in the complete dissolution of water and ions around the protein to achieve a dynamic steady state. To include the effect of THz photons on the NR2B receptor, THz photons (close to 34.5 THz) were added to the whole simulated system. Since the intensity ratio of the electromagnetic component of the photons is equal to the speed of light, THz photons can be described by the equation E(t) = A*u*cos(ωt+φ) [50], where A denotes the maximum amplitude of the electric field, which determines the intensity of the electric component of photons, ω is the angular frequency, and u and φ denote the polarization direction and its phase, which are set to (0,0,1) and 0, respectively. The frequency of photons, ν, is related to angular frequency ω by the formula v = ω/2π. In this way, THz photons are involved in molecular dynamics simulations through electric field forces associated with E and the charges on all the atoms.

We simulated the binding pocket of the NR2B receptor and the absorption spectrum of glutamate on the basis of the molecular dynamics approach of the classical GROMACS code [51]. The absorption spectra (Fig. 2A) were calculated from the Fourier transform of the total charge current velocity autocorrelation function of the simulated system [52]. We set the time interval for the sampling of the spectra to ~1 fs and the total sampling time to 50 ps. The absorption spectra were calculated on the basis of the Fourier transform of the autocorrelation function of the total charge current: J(t) = Σ_iq_iv_i(t) [52]^, where q_i is the charge of the i-th atom and v_i(t) is the velocity of the i-th atom at time t. Note that the 4 residue sets that were uniquely identified as key amino acid residues of the protein pocket, Tyr32, Gln78, Glu111, and Tyr186, were used in the calculation of the binding pocket absorption spectra.

In addition, the following 4 terms can be used to calculate the free energy of binding (G_binding): between glutamate and the NR2B receptor in solution: G_binding = E_vdw + E_elec + G_polar + G_nonpolar, i.e., van der Waals free energy (E_vdw), electrostatic free energy (E_elec), polar or electrostatic solvation free energy (G_polar), and nonelectrostatic solvation free energy (G_nonpolar) [53–56]. The g_mmpbsa tool [55] was used here to assess G_binding based on simulated trajectories. Using a sampling interval of 0.5 ns, after simulating the system time for 50 ns to reach equilibrium, we continued the run for 10 ns to calculate the binding free energy, and the average binding free energy over time is calculated based on the bootstrap method. g_mmpbsa also calculates the contribution of residues to the total binding energy and identifies ligand–receptor binding at key residues in ligand–receptor binding.

The experimental data in this article are all represented as the mean ± SD, and t test or one-way analysis of variance (clearly provided in the figures) using GraphPad Prism 6.0 software were used for difference testing.