Introduction

Multiple studies have demonstrated that GABAergic interneurons, especially parvalbumin-expressing fast-spiking interneurons (PVINs), play a critical role in attenuating pathological hypersynchronicity of excitatory networks and decreasing seizure susceptibility1,2,3,4,5,6. Decreased excitability of PVINs produces increased activity of pyramidal neurons, facilitating spontaneous epileptic seizures6. It is well-established that PVINs play an integral role in generating or regulating gamma oscillations (30–80 Hz)7,8,9, which reflects rhythmic interactions between inhibitory interneurons and excitatory neurons10,11. Altered gamma activity is associated with cognitive dysfunction, psychiatric disorders, and epilepsy12,13,14,15,16. In hAPP-J20 mice, decreased Nav1.1 expression led to inhibitory hypofunction of PVINs, resulting in reduced gamma sctivity, increased epileptiform spikes, and network hypersynchronization5,17. Reductions in gamma activity tended to correlate with increased epileptiform discharges in patients with epilepsy18. Furthermore, silencing PVINs in an absence epilepsy model resulted in spikes with characteristics similar to absence seizures1. Thus, seizure induction may be attenuated by enhancing PVIN activity and gamma oscillations.

Optogenetic activation of PVINs at 40 Hz in vivo has been proven to induce robust gamma oscillations19,20. Intriguingly, gamma entrainment using sensory stimuli (GENUS), a non-invasive, exogenous stimulation by rhythmic 40 Hz light flicker, strengthens gamma activity in the primary visual cortex (V1) and suppresses spike generation13,15,20,21. Recent studies have demonstrated that GENUS increases neuronal survival, enhances synaptic plasticity, facilitates long-term depression, and reduces hippocampal neuroinflammation in the mouse models of Alzheimer’s disease (AD), cerebral ischemia, or neurodegeneration13,14,15,20,22. Studies conducted on animals with AD and patients with mild-to-moderate AD have revealed that chronic GENUS ameliorated learning and memory impairments13,14,15,23, hippocampal atrophy23,24, as well as deficits in daily activities23,25 and sleep quality25. Gamma activity, spikes, and cellular alterations are influenced by GENUS, and are all associated with epilepsy5,17,18,26,27,28,29. However, GENUS has yet to be investigated to determine if it has a role in antiepileptogenesis.

Therefore, we assessed the protective effects of chronic GENUS on epileptogenesis using pilocarpine-induced status epilepticus (SE) mice, a well-established model of temporal lobe epilepsy30,31. We discovered that SE mice treated with GENUS for 6 weeks exhibited reduced neuronal damage, improved behavioral performance, and decreased seizure susceptibility. In addition, by assessing neural circuit function utilizing c-Fos brain mapping, neuronal ablation, and chemogenetics, we identified that GENUS-entrained 40 Hz activity relied on a specific pathway. This pathway consisted of ON-OFF direction-selective retinal ganglion cells (ooDS-GCs) that innervate glutamatergic neurons in the shell of the dorsal lateral geniculate nucleus (dLGN shell), which in turn activate PVINs in the superficial layer 2/3 of V1 (superficial V1).

Results

GENUS enhanced gamma activity and decreased the spike rate

Altered gamma activity has been observed in neurological and psychiatric disorders12,13,14,20,32. Accordingly, we recorded local field potential (LFP) neural activity over the parietal cortex of epileptic mice 6 weeks after SE induction because the mice had entered a period of spontaneous recurrent seizures with frequent epileptiform spikes30 (Fig. 1a). First, we identified electroencephalography (EEG) segments with and without frequent high-amplitude discharges, i.e., spikes and baseline (Supplementary Fig. 1a). The power spectral density (PSD) analysis revealed enhanced gamma activity in the baseline compared to EEGs coupled with spikes (Fig. 1b, c). Next, the relative power for the frequency bands in the baseline and spikes was analyzed on 20–25 EEG segments, each with a 10 s segment duration. Compared to the baseline, the relative power for EEGs with frequent spikes was clearly increased in the delta (1–4 Hz) and alpha (8–13 Hz) bands, drastically decreased in the gamma (30–80 Hz) band, and not significantly altered in the theta (4–8 Hz) and beta (13–30 Hz) bands (Fig. 1d and Supplementary Fig. 1b). The relative power for individual frequencies in the gamma band was analyzed further. For each frequency, the decrease in relative power for the EEG with frequent spikes was twofold more than the baseline (Fig. 1e and Supplementary Fig. 1c). These results suggested reduced gamma activity was associated with increased epileptiform spikes in epileptic mice.

Fig. 1: GENUS enhanced gamma activity and decreased spike rate.
figure 1

a Experimental schedule for power spectrum analysis and spike detection in epileptic mice. b A time-frequency spectrogram of EEG with or without frequent spikes. c PSD of EEG recordings (0–100 Hz; 15 segments). d Relative power in different frequency bands (n = 12 mice/group). e Relative power at specific frequencies within the gamma band (n = 12 mice/group). f A time-frequency spectrogram of EEG during 40 Hz light flicker (50 Hz notch). g PSD of EEG recordings during 40 Hz light flicker (0–100 Hz, 15 segments). h Relative power in response to light flicker at different frequencies (n = 10 mice/group). i Spike rate in response to light flicker at different frequencies (n = 10 mice/group. j Spike rate detected before, during, and after 40 Hz light flicker (n = 9 mice/group). k A 2-h recording of EEG obtained 6 weeks after SE (the red line denotes the amplitude threshold for spike detection). l Location and number of spikes in the EEG shown in (k). m A time-frequency spectrogram of the EEG presented in Fig. 1k. Data were means ± SEM (d, e, h–j); *P < 0.05, **P < 0.01, ***P < 0.001. Two-way ANOVA with Bonferroni post hoc test for (d, e). Repeated measures two-way ANOVA with Bonferroni post hoc test for (h, i). Repeated measures one-way ANOVA with Bonferroni post hoc test for (j). All EEGs were recorded over the parietal cortex (AP: −2.0; ML: ± 2.0; DV: −0.5 mm). Source data are provided as a Source Data file.

Full size image

Recent studies have shown that GENUS can drive 40 Hz oscillations in mice with cognitive dysfunction13,14,15,21. Thus, we used a custom-made light-emitting diode (LED) device to deliver GENUS at 40 Hz. Consistent with previous studies, the 40 Hz power in LFP was significantly enhanced at the 40 Hz light flicker (Fig. 1f, g). We also observed that light flickering at a specified frequency could entrain corresponding frequency oscillations in LFP (Fig. 1h and Supplementary Fig. 1d). A reduced spike rate has been observed in high-intensity gamma activity and 40 Hz flicker5,20. Therefore, to determine whether this phenomenon was universal in light flicker-entrained gamma activity, we exposed SE mice to light flicker at individual gamma frequencies. The spike rate was significantly decreased in light flicker at 40 Hz but not at other frequencies (Fig. 1i, k–m and Supplementary Fig. 1e). Epileptiform discharges exhibited a strong correlation with non-rapid eye-movement (NREM) sleep33,34. We detected spikes in SE mice during NREM sleep before, during, and after 40 Hz light flicker (Fig. 1j and Supplementary Fig. 1f). Compared with the spike rate during NREM sleep before 40 Hz flicker, the spike rate was significantly reduced during NREM sleep after 40 Hz flicker. Moreover, the suppressive effect of 40 Hz flicker on epileptic spikes persisted for at least one hour. Notably, spikes are used as an electrical signature for massive and synchronous discharges generated by excitatory neurons that may progress to epileptic seizures35,36,37,38,39,40. Thus, GENUS may be a tool to suppress excitatory networks and seizures in epileptic mice.

Chronic GENUS attenuated hippocampal damage and seizure susceptibility

In mice with cognitive dysfunction, GENUS improved gamma activity and hippocampal neuroprotection14,15,20, suggesting that GENUS could affect brain regions beyond V1, including the hippocampus13,14,15. To examine the potential role of GENUS in antiepileptogenesis, SE mice received GENUS 1 h/day for 6 weeks, starting on the 4th day after mice recovered from SE (Fig. 2a). The SE mice treated with GENUS displayed measurable amelioration of neuronal loss, astrocytosis, and microgliosis in the hippocampus compared to SE mice without GENUS treatment (Fig. 2b–e and Supplementary Fig. 2a–g). In addition, hippocampal microglia in GENUS-treated SE mice exhibited significantly smaller cell bodies and shorter processes compared to the activated microglia in SE mice (Fig. 2f, g and Supplementary Fig. 2h, i).

Fig. 2: Chronic GENUS attenuated hippocampal damage and seizure susceptibility.
figure 2

a Experimental schedule for VEEG recordings and immunofluorescence staining after chronic GENUS treatment. b Immunofluorescence images of neurons, astrocytes, microglia in the hippocampal dCA3 region. Scale bar = 200 μm, zoom-in scale bar = 20 μm. c Number of neurons (NeuN+) in the dCA3 region (n = 5 mice/group). d Number of astrocytes (GFAP+) in dCA3 region (n = 5 mice/group). e Number of microglia (Iba-1+) in dCA3 region (n = 5 mice/group). f Mean process length of microglia in dCA3 region (n = 5 mice/group). g Mean cell body diameter of microglia in dCA3 region (n = 5 mice/group). h Number of seizures per day (n = 8 mice/group). i Number of spikes per hour (n = 8 mice/group). j Total duration of seizures (n = 8 mice/group). Data were means ± SEM (c–j); *P < 0.05, **P < 0.01, ***P < 0.001. Two-way ANOVA with Bonferroni post hoc test for (c–g). Two-sided Student’s t test for (h, j). One-way ANOVA with Bonferroni post hoc test for (i). Data presented in (c–g) were derived from the corresponding data in Supplementary Fig. 2e–i. Source data are provided as a Source Data file.

Full size image

Seizure activity during epileptogenesis was analyzed using video electroencephalography (VEEG) recordings (Supplementary Fig. 3a–c and corresponding Supplementary Movies 1–3). As described in previous studies27,28,30, the SE mice presented more frequent seizures than the control group (Fig. 2h). Epileptiform spikes and seizure frequency and duration were dramatically reduced in GENUS-treated SE mice compared to non-treated SE mice (Fig. 2h–j). We subsequently analyzed the frequencies of convulsive and non-convulsive seizures separately. Compared with SE mice without treatment, those receiving chronic GENUS exhibited a marked reduction in convulsive seizure frequency (Supplementary Fig. 2j). However, no significant difference was observed between the two groups in the frequencies of non-convulsive seizures (Supplementary Fig. 2k). Epileptic mice exhibited a lower baseline number of non-convulsive seizures compared with convulsive seizures31,41. Nevertheless, the neuroprotective effects induced by chronic GENUS treatment, which transformed convulsive seizures into non-convulsive seizures, cannot be excluded. These results indicated that chronic GENUS protected against neuronal loss and gliosis and ameliorated brain hyperexcitability in epileptic mice. In addition, we noted that spikes were detected in the control mice, which were consistent with observations reported in previous studies4,42. The primary cause of spikes may be attributed to postsurgical damages43, despite our efforts to minimize damage to the brain cortex by avoiding excessively deep insertion of epidural screws.

Chronic GENUS improved epilepsy-related behavioral performance

Because psychiatric conditions are associated with epilepsy44, we evaluated the effect of chronic GENUS on epilepsy-related behaviors during epileptogenesis (Fig. 3a). After GENUS treatment for 6 weeks, mice were evaluated using the sucrose preference test (SPT), open field test (OFT), and elevated plus maze (EPM) for anhedonia and anxiety-related behaviors associated with epilepsy (Fig. 3b). SE mice treated with GENUS showed a significant increase in consumption of the sucrose solution compared to SE mice without GENUS (Fig. 3c, d). In the OFT, chronic GENUS significantly increased the total distance, number of times the mice reared, and the number of lines crossed in epileptic mice (Fig. 3e–g), but not the time spent in the center area (Fig. 3h). In the EPM, epileptic mice treated with GENUS exhibited more activity in the open arms, including distance, entries, and duration (Fig. 3i–l). We also assessed hippocampus-dependent spatial learning and memory using the Morris water maze (MWM). In the place navigation test, the escape latency for each group decreased gradually over the four training days, and no significant differences were observed among the three groups (Fig. 3m). In the spatial probe test on the 5th day, the SE mice not exposed to GENUS took the longest time to reach the hidden platform (Fig. 3n). The GENUS-treated SE mice showed significantly improved spatial memory, with more visits to the target quadrant and platform location than SE mice not exposed to GENUS (Fig. 3o, p). Thus, chronic GENUS treatment for SE mice reduced epilepsy-related anhedonia, anxiety behaviors, and cognitive dysfunction. Taken together, chronic GENUS treatment appeared to have a potential role in antiepileptogenesis and epilepsy-related comorbidity modification.

Fig. 3: Chronic GENUS improved epilepsy-related behavioral performance.
figure 3

a Experimental schedule for behavioral performance after chronic GENUS treatment. b Movement traces in OFT, EPM, and MWM. c Total water consumption in SPT (n = 10 mice/group). d Ratio of sucrose solution intake to total water consumption in SPT (n = 10 mice/group). e Total distance traveled in OFT (n = 12 mice/group). f Number of rearing in OFT (n = 12 mice/group). g Number of cross lines in OFT (n = 12 mice/group). h Ratio of time spent in center area in OFT (n = 12 mice/group). i Distance traveled in open arms in EPM (n = 11 mice/group). j Ratio of time spent in open arms in EPM (n = 11 mice/group). k Number of entries into open arms in EPM (n = 11 mice/group). l Ratio of entries into open arms in EPM (n = 11 mice/group). m Escape latency in place navigation test in MWM (n = 11 mice/group). n Escape latency in spatial probe test in MWM (n = 11 mice/group). o Number of visits to target quadrant in spatial probe test in MWM (n = 11 mice/group). p Number of visits to platform in spatial probe test in MWM (n = 11 mice/group). Data were means ± SEM (c–i, k–m, o) and medians (IQR) (j, n, p); *P < 0.05, **P < 0.01, ***P < 0.001. One-way ANOVA with Tukey HSD post hoc test for (c–i, k, l, o). Kruskal–Wallis test with Bonferroni post hoc test for (j, n, p). Repeated measures two-way ANOVA with Bonferroni post hoc test for (m). Source data are provided as a Source Data file.

Full size image

GENUS-entrained gamma activity was associated with V1 PVINs

Studies have shown that PVINs in the cortex can be activated and synchronized by optogenetic stimulation at the gamma frequency8,45. The selective activation of PVINs at 40 Hz significantly increased gamma power19,45. These data indicate that PVINs were preferentially correlated with gamma oscillations in the cortical circuit. Therefore, we determined whether PVINs were activated by assessing c-Fos expression in several brain areas associated with gamma activity and the visual circuit, including the visual cortex, hippocampus, and thalamus10. We observed that numerous PVINs were distributed in the hippocampus and visual cortex, especially in V1 (Supplementary Fig. 4a, b).

Mice were placed in a dark chamber for 6 h to eliminate the influence of visual stimulation on PVIN activation. Then the mice received a 200 Hz or 40 Hz light flicker for 1 h and were euthanized immediately to map c-Fos expression. Animals that received 200 Hz light flicker, simulating a light environment without visible flickering, were considered as the control group. C-Fos was highly expressed in the hippocampus and visual cortex after light flickering at 200 Hz or 40 Hz (Supplementary Fig. 4c). However, there was no prominent difference in c-Fos expression between the 200 Hz and 40 Hz flicker (Supplementary Fig. 4d, e). After PVIN+ + c-Fos+ double-labeled cells were identified, a significant increase in activated PVINs was only observed in the V1 at the 40 Hz flicker compared with the 200 Hz flicker (Fig. 4a, b).

Fig. 4: GENUS-entrained gamma activity was associated with V1 PVINs.
figure 4

a Immunofluorescence images of PVINs and c-Fos (PV+ + c-Fos+ double-labeled) in the V1 region. Scale bar = 150 μm, zoom-in scale bar = 100 μm. b Ratio of PVINs+ + c-Fos+ double-labeled cells to total PV+ cells (n = 10 mice/group). c Immunofluorescence images of PVINs (PV+) in the V1 region. Scale bar = 400 μm, zoom-in scale bar = 100 μm. d Sketch illustrating PVINs ablation in the V1 region. e Number of PVINs in the bilateral V1 region (n = 7 mice/group). f Relative power at 40 Hz in response to 40 Hz light flicker in PV-Cre mice with or without PVINs ablation (n = 7 mice/group). g Ratio of relative power between 40 Hz light flicker and occluded flicker in PV-Cre mice with or without PVINs ablation (Normalization relative to the mean relative power at occluded flicker; n = 7 mice/group). h A time-frequency spectrogram of EEG recorded from PV-Cre mice with or without PVINs ablation. i PSD of EEG recordings obtained from PV-Cre mice with or without PVINs ablation (20–60 Hz, 6 segments, 50 Hz notch). Data were means ± SEM (b, e–g); *P < 0.05, **P < 0.01, ***P < 0.001. Two-way ANOVA with Bonferroni post hoc test for (b, g). Two-sided Student’s t test for (e). Repeated measures two-way ANOVA with Bonferroni post hoc test for (f). All EEGs were recorded over the parietal cortex (AP: −2.0; ML: ± 2.0; DV: −0.5 mm). Source data are provided as a Source Data file.

Full size image

Finally, we evaluated whether PVINs in V1 were necessary to GENUS entrain gamma activity. A Cre-dependent virus encoding taCasp3 (AAV2/9-taCasp3-TEVp) was delivered bilaterally into the V1 of PV-Cre mice to ablate PVINs selectively (Fig. 4c–e). We determined that the PVIN ablation remarkably minimized the GENUS-entrained 40 Hz power 3 weeks after injection (Fig. 4f–i). These results indicated that GENUS-induced 40 Hz oscillations could be attributed to PVIN activation in V1.

PVIN ablation in V1 abolished chronic GENUS-induced hyperexcitability reduction and hippocampal neuroprotection

To determine the effects of chronic GENUS-entrained gamma activity on reduced hyperexcitability and hippocampal neuroprotection, SE was induced in mice with bilateral PVIN ablation in the V1 (i.e., 3 weeks after injection of virus encoding taCasp3). Following 6 weeks of treatment with GENUS, mice were examined with SPT, OFT, EPM, and MWM (Fig. 5a). The results indicated that chronic GENUS treatment did not improve behavioral performance in SE mice with PVIN ablation (Fig. 5b–e and Supplementary Fig. 5a–k). Neuronal damage in the hippocampus was also assessed and the neuroprotective effect induced by chronic GENUS in SE mice was counteracted by PVIN ablation (Fig. 5f–k and Supplementary Fig. 5l–p). Furthermore, the network excitability of mice with PVIN ablation was tested. Mice were injected with pilocarpine (250 mg/kg, i.p.) 30 min after scopolamine (1 mg/kg, i.p.) administration and their seizure activity was monitored for 60 min. Enhanced seizure severity and decreased survival probability in mice with PVIN ablation were observed (Fig. 5l and Supplementary Fig. 5q), indicating a hyperexcitable neuronal network resulted from bilateral PVIN ablation in the V1. We tested the influence of chronic GENUS treatment on the excitability of the epileptic brain in mice with PVIN ablation. After 6 weeks of GENUS treatment, no significant amelioration in epileptiform spike, seizure frequency, and duration was observed in SE mice with PVIN ablation compared with SE mice not treated with GENUS (Fig. 5m, n and Supplementary Fig. 5r). Compared to the control group, naive mice with PVINs ablation in the V1 did not exhibit significant neuronal damge, poor behavioral performance, or hyperexcitability, suggesting that ablation of PVINs in the V1 was unlikely to facilitate the development of epilepsy over a 6-week period. Together, these results indicated that bilateral ablation of PVINs in the V1 abolished chronic GENUS-induced mitigation of epilepsy-related behavioral performance, hippocampal damage, and seizure susceptibility in epileptic mice.

Fig. 5: PVIN ablation in V1 abolished chronic GENUS-induced hyperexcitability reduction and hippocampal neuroprotection.
figure 5

a Experimental schedule for behavioral performance, VEEG recordings, and immunofluorescence staining after chronic GENUS treatment. b Movement traces in OFT, EPM, and MWM. c Total distance traveled in OFT (n = 8 mice/group). d Ratio of entries into open arms in EPM (n = 8 mice/group). e Number of visits to the platform in the spatial probe test in MWM (n = 8 mice/group). f Immunofluorescence images of neurons (NeuN+), astrocytes (GFAP+), and microglia (Iba-1+) in hippocampal dCA3 region. Scale bar in NeuN+ and GFAP+ images = 200 μm. Scale bar in Iba-1+ images = 150 μm, zoom-in scale bar = 20 μm. g Number of neurons (NeuN+) in dCA3 region (n = 5 mice/group). h Number of astrocytes (GFAP+) in dCA3 region (n = 5 mice/group). i Number of microglia (Iba-1+) in dCA3 region (n = 5 mice/group). j Mean process length of microglia in dCA3 region (n = 5 mice/group). k Mean cell body diameter of microglia in dCA3 region (n = 5 mice/group). l Survival probability over a 60 min duration following pilocarpine administration (n = 12 mice/group). m Number of seizures per day (n = 8 mice/group). n Total duration of seizures (n = 8 mice/group). Data were means ± SEM (c–e, g–k, m, n) and rate (l); *P < 0.05, **P < 0.01, ***P < 0.001. One-way ANOVA with Bonferroni post hoc test for (c–e, m, n). Two-way ANOVA with Bonferroni post hoc test for (g–k). Kaplan–Meier analysis with Log Rank (Mantel–Cox) test for (l). Data presented in (g–k) were derived from the corresponding data in Supplementary Fig. 5l–p. Source data are provided as a Source Data file.

Full size image

GENUS-entrained gamma activity required the activation of the visual circuit related to the dLGN shell

Our results suggested that GENUS-entrained gamma activity played a protective role in epileptogenesis by activating PVINs in the V1. Therefore, we explored the visual circuit associated with GENUS in activating PVINs. We identified discrete brain regions involved in GENUS activity using c-Fos mapping. Mice were treated with 200 Hz or 40 Hz flicker for one h after six h dark adaptation. The expression of c-Fos was examined in brain regions associated with the visual circuit or gamma activity10. Compared to the 200 Hz flicker, the 40 Hz flicker induced a substantial increase in the expression of c-Fos in the dLGN shell (Fig. 6a and Supplementary Fig. 6a–c). The proportion of c-Fos in the dLGN shell to total c-Fos within the dLGN was distinctly higher in mice treated with 40 Hz flicker compared to 200 Hz flicker exposure (Fig. 6b).

Fig. 6: PVIN ablation in V1 abolished chronic GENUS-induced hippocampal neuroprotection and hyperexcitability suppression.
figure 6

a Immunofluorescence images of c-Fos+ cells in the dLGN. b Ratio of c-Fos+ cells in the dLGN (n = 8 mice/group). c Immunofluorescence images of neurons (NeuN+) in the dLGN shell. d, j, o, v Relative power at 40 Hz in response to 40 Hz flicker (n = 6 mice/group for (d, j); 5 mice/group for (o, v)). e, l, p, w A time-frequency spectrogram of EEG. f, m, q, x PSD of EEG recordings (20–60 Hz, 6 segments, 50 Hz notch). g Immunofluorescence images of PV+ + c-Fos+ double-labeled cells in the V1. h Ratio of PV+ + c-Fos+ double-labeled cells in the V1 (n = 10 mice/group). i Immunofluorescence images of PV+ + mCherry+ double-labeled cells in the superficial V1. k Relative power at 30 Hz in response to 30 Hz flicker (n = 6 mice/group). n Immunofluorescence images of CaMKII+ + mCherry+ double-labeled cells in the dLGN shell. r Sketch of monosynaptic circuit tracing procedure. s Immunofluorescence images of PV+ + mCherry+ double-labeled cells in the superficial V1. t Ratio of PV+ + mCherry+ double-labeled cells in the superficial V1. u Immunofluorescence images of CART+ + mCherry+ double-labeled cells in the retina. Data were means ± SEM (b, d, h, j, k, o, v) and rate (t); *P < 0.05, **P< 0.01, ***P < 0.001. Two-way ANOVA with Bonferroni post hoc test for (b, h). Repeated measures two-way ANOVA with Bonferroni post hoc test for (d). Repeated measures one-way ANOVA with Bonferroni post hoc test for (j, k, o, v). All EEGs were recorded over the parietal cortex (AP: − 2.0; ML: ± 2.0; DV: − 0.5 mm). Scale bar = 200 μm for (a, c, g); 400 μm for (i, n, s); 1000 μm for (u). Zoom-in scale bar = 40  μm for (a, c); 50 μm for (g, i, n, s, u). Source data are provided as a Source Data file.

Full size image

It is well-accepted that the dLGN is a hub connecting the retina to the visual cortex in visual perception46,47. To verify whether the dLGN shell was a relay region for GENUS-entrained gamma activity, we selectively ablated neurons in the dLGN shell using an intracranial injection of AAV2/9-taCasp3-TEVp with AAV2/9-EGFP-P2A-CRE (Fig. 6c and Supplementary Fig. 6d, e). Three weeks after the injection, the GENUS-entrained 40 Hz power was substantially decreased in mice with neuronal ablation in the dLGN shell compared with control mice (Fig. 6d–f and Supplementary Fig. 6f). These results indicated that 40 Hz oscillations entrained by GENUS required the activation of neurons in the dLGN shell.

Neurons in the dLGN shell preferentially target the superficial V146. Therefore, based on the Allen Brain Atlas (http://atlas.brain-map.org/), we analyzed PVIN activation by identifying PVIN+ + c-Fos+ double-labeled cells in the individual layers of V1. There were no differences in PVIN distribution and total neuronal activation in the corresponding layer of V1 between mice that received 200 Hz or 40 Hz flicker (Fig. 6g and Supplementary Fig. 6g, h). However, activated PVINs in layer 2/3 of V1 (superficial V1) were increased significantly in mice treated with the 40 Hz flicker compared to the 200 Hz flicker exposure (Fig. 6g, h). We chemogenetically manipulated PVINs in the superficial V1 to assess their function in GENUS-entrained gamma activity. AAV2/9-DIO-hM4D(Gi)-mCherry was injected into layer 2/3 of V1 to infect the PVINs in PV-Cre mice (Supplementary Fig. 6i). After 3 weeks, 83.70% of the mCherry-labeled neurons were PVINs (Fig. 6i and Supplementary Fig. 6j). The increased 40 Hz power induced by GENUS was notably reduced approximately 40 min after the clozapine-N-oxide (CNO) injection (Fig. 6j and Supplementary Fig. 6k). However, the increased 30 Hz power entrained by 30 Hz flicker did not markedly decrease after CNO administration (Fig. 6k and Supplementary Fig. 6l). We immediately switched light flickering to 40 Hz flicker following the 30 Hz flicker and observed that the reduced 40 Hz power after CNO administration during GENUS was persistent, indicating that decreased 40 Hz power resulted from PVIN inhibition but was not attributed to other interventions such as closing the eyes (Fig. 6l, m). Therefore, PVINs in the superficial V1 were involved in GENUS-entrained gamma activity.

Next, we delivered AAV2/Retro-CRE-EGFP bilaterally into layer 2/3 of V1 and infected the dLGN shell neurons with AAV2/9-DIO-hM4D(Gi)-mCherry (Supplementary Fig. 6m). After 3 weeks of recovery, 76.70% of mCherry-labeled V1-projecting neurons in the dLGN shell were co-labeled with CaMKII (Fig. 6n and Supplementary Fig. 6o). There also were numerous cells co-labeled with mCherry and VGLUT2 (Supplementary Fig. 6n). These results confirmed that excitatory neurons in the dLGN shell carried major retinal inputs to the superficial V147. Thus, chemogenetic inhibition of V1-projecting neurons in the dLGN shell minimized the effect of GENUS on 40 Hz entrainment (Fig. 6o–q and Supplementary Fig. 6p). In addition, we delivered AAV2/1-CRE-EGFP into the dLGN shell while delivering AAV2/9-DIO-mCherry into the superficial V1 (Fig. 6r). 81.98% of PVINs in layer 2/3 of the V1 located near the injection site were labeled with mCherry (Fig. 6s, t). Thus, the pathway of the dLGN shell to the superficial V1 was responsible for GENUS-entrained gamma activity.

The dLGN shell contains the axonal terminations of ooDS-GCs, which form the retino-geniculo-superficial V1 circuit46,48. We examined whether ooDS-GCs activation was required for GENUS-entrained gamma activity. AAV2/Retro-CRE-EGFP was injected into the dLGN shell, and AAV2/9-DIO-hM4D(Gi)-mCherry was injected binocularly to infect the retinas (Supplementary Fig. 6q). Three weeks later, the majority of mCherry-labeled retinal ganglion cells (RGCs) were immunopositive for the neuropeptide cocaine- and amphetamine-regulated transcript (CART) (Fig. 6u and Supplementary Fig. 6r). Repressing the dLGN shell-projecting ooDS-GCs by CNO significantly decreased the enhanced 40 Hz power entrained by GENUS (Fig. 6v–x and Supplementary Fig. 6s).

To exclude any potential neuroactive effects derived from CNO and its metabolites, AAV2/9-DIO-mCherry was injected into the dLGN shell and the vitreous chamber. We observed that GENUS-induced gamma oscillations were not significantly decreased after CNO administration in the mice devoid of DREADDs expression in the dLGN shell or retina (Supplementary Fig. 6t–ac). Therefore, the activation of the retina-dLGN shell-layer 2/3 of the V1 pathway was required for gamma entrainment induced by GENUS.

Inactivation of the retina-dLGN shell-superficial V1 pathway diminished the seizure susceptibility reduction induced by chronic GENUS

The above observations indicated that activation of the retina-dLGN shell-layer 2/3 of the V1 pathway was necessary for GENUS-entrained gamma activity. To evaluate the role of this neuronal pathway in the seizure susceptibility reduction induced by GENUS, we conducted a 72 h VEEG recording to assess SRSs following chronic GENUS treatment in mice with the pathway inactivated (Fig. 7a).

Fig. 7: Inactivation of the retina-dLGN shell-superficial V1 pathway diminished the seizure susceptibility reduction induced by chronic GENUS.
figure 7

a Experimental schedule for SE mice with the pathway inactivated after chronic GENUS treatment. b Sketch illustrating neuronal ablation. c Number of PVINs in the superficial V1. d Mean fluorescence intensity of vGLUT2 in the dLGN shell. e Number of ooDS-GSs in the retina. f Immunofluorescence images of PVINs (PV+) in the superficial V1. g Immunofluorescence images of vGLUT2 (vGLUT2+) in the dLGN shell. h Immunofluorescence images of ooDS-GSs (CART+) in the retina. i A time-frequency spectrogram of EEG. j Relative power at 40 Hz in response to 40 Hz flicker. k PSD of EEG recordings (20–60 Hz, 6 segments, 50 Hz notch). l Spike rate in mice with PVINs ablation in the superficial V1. m Seizures per day in mice with PVINs ablation. n Total seizure duration in mice with PVINs ablation. o Spike rate in mice with excitatory neurons ablation. p Seizures per day in mice with excitatory neuron ablation. q Total seizure duration in mice with excitatory neurons ablation. r Spike rate in mice with ooDS-GSs ablation. s Seizures per day in mice with ooDS-GSs ablation. t Total seizure duration in mice with ooDS-GSs ablation. Data were means ± SEM (c–e, j, l, n, o, r, t) and medians (IQR) (m, p, q, s); *P < 0.05, **P < 0.01, ***P < 0.001. n = 8 mice/group for (c–e, j, l–t). Two-sided Student’s t test for (c–e). Repeated measures two-way ANOVA with Bonferroni post hoc test for (j). One-way ANOVA with Bonferroni post hoc test for l, n, o, r, and t. Kruskal–Wallis test with Bonferroni post hoc test for (m, p, q, s). All EEGs were recorded over the parietal cortex (AP: − 2.0; ML: ± 2.0; DV: − 0.5 mm). Scale bar = 500 μm for (f); 200 μm for (g); 1000 μm for (h). Zoom-in scale bar = 50 μm for (f); 20 μm for (g); 100 μm for (h). Source data are provided as a Source Data file.

Full size image

To specifically inactivate neurons within the retina-dLGN shell-superficial V1 pathway, we delivered AAV2/9-taCasp3-TEVp into the bilateral superficial V1 of PV-Cre mice, the dLGN shell of vGLUT2-Cre mice, and the retina of Cartpt-Cre mice to ablate PVINs, excitatory neurons, and ooDS-GCs, respectively (Fig. 7b). After a 3-week recovery period, compared to mice without neuronal ablation, the number of PVINs in the superficial V1 and ooDS-GCs in the retina decreased by 67.91% and 70.59%, respectively, and the mean fluorescence intensity of vGLUT2 in the dLGN shell reduced by 65.07% (Fig. 7c–h). Inactivation of these specific neurons significantly diminished the enhanced 40 Hz power entrained by GENUS (Fig. 7i–k). Following 6 weeks of GENUS treatment, more epileptiform discharges and SRSs were observed in SE mice with the pathway inactivated compared to those with the pathway intact (Fig. 7l–t), suggesting that inactivation of this pathway abolished the reduced seizure susceptibility induced by chronic GENUS treatment. Taken together, these findings indicate that activation of the retina-dLGN shell-superficial V1 pathway is required for GENUS-induced reduction in seizure susceptibility.

Discussion

GENUS is a cost-effective, non-invasive intervention for cognitive dysfunction13,14,15,20,23,24,25. In this study, chronic GENUS proved to be an effective treatment for neuronal damage, seizure susceptibility, and epilepsy-related behaviors in pilocarpine-induced epileptic mice. We demonstrated that ooDS-GCs in the retina, excitatory neurons in the dLGN shell, and PVINs in layer 2/3 of V1 are critical components of the neural circuitry involved in the gamma entrainment and seizure susceptibility reduction induced by GENUS.

Epileptogenesis is a process by which the normal brain transitions into an epileptic brain that exhibits an imbalance between excitatory and inhibitory activities4,16,26,44,49. PVINs are major inhibitory neurons and are necessary for local oscillations3. Decreased PVIN excitability observed in the mutant mice with insufficient neddylation reduced GABA transmission and increased network excitability, which then contributed to the occurrence of spontaneous epileptic seizures6. Scn1a gene re-expression restored the excitability of Scn1a-deficient PVINs and attenuated epileptic phenotypes in Dravet syndrome mice50. This study revealed that chronic GENUS treatment suppressed seizure susceptibility and mitigated epileptogenesis in SE mice. Furthermore, PVIN ablation in V1 abolished the antiepileptogenesis effect induced by chronic GENUS. These results established a critical role for PVINs in regulating the excitability of neural networks.

Epileptiform spikes are one of the biomarkers of an epileptic brain and can be used to locate the epileptogenic focus36,37,39,51. CXCR5 deficiency in cortical neurons results in neuronal hyperexcitability, causing an increased spike frequency and higher susceptibility to epilepsy40. In a human study, there was a notable similarity in the probability distribution of spikes and seizures, indicating that when seizures are likely to occur, the probability of epileptic spiking increases35. In hAPP-J20 mice, the spike rate was inversely related to gamma oscillation intensity, suggesting a network hypersynchronicity could emerge when gamma activity is reduced5. It is accepted that the rhythmic firing of PVINs is a primary source of gamma activity3,8,9,10,32. In a genetic model of absence epilepsy, PVIN inhibition in the cortico-thalamocortical network generated spikes that were similar to absence seizures1. Stimulating PVINs at the gamma frequency induced an elevated gamma power in the LFP of the somatosensory cortex; conversely, suppressing PVINs reduced gamma oscillations5,8,19. These findings suggest that enhancing gamma through PVIN activation might inhibit synchronous discharges in an epileptic brain, attenuating seizure susceptibility.

Consistent with previous studies13,14,15,22, GENUS (40 Hz light flicker with 12.5 ms on and off) could increase power in LFP at 40 Hz recorded over the hippocampus of epileptic mice. Intracranial EEGs primarily reflect LFP generated by synaptic potentials related to the synchronous activity of local neuronal assemblies18,52. Thus, in this study, a significantly reduced spike rate was observed at 40 Hz light flicker but not at other frequencies, indicating that GENUS reduced excessive hypersynchronicity in the neuronal network. It appears that the hippocampal network was affected by the increased gamma activity induced by GENUS because an elevated 40 Hz power during GENUS was observed in brain areas beyond V113,14,15,22,23,53. However, another recent study reported that GENUS could entrain gamma in V1 but not in deep structures, including the hippocampus21. Nevertheless, GENUS has been shown to be beneficial to the hippocampus by enhancing neuroprotection, reducing neuroinflammation, and improving synaptic plasticity14,15. Hippocampal neuronal loss, gliosis, and pro­inflammatory responses are commonly observed in epileptogenesis27,28,30,54 and in epilepsy-related comorbidities, including depression, anxiety, and cognitive dysfunction55,56,57. In this study, after GENUS treatment for 6 weeks, SE mice demonstrated considerable diminution in hippocampal damage, seizure susceptibility, and epilepsy-related behavioral performance, indicating that chronic GENUS alleviated epileptogenesis and epilepsy-related comorbidities. Thus, decreased hippocampal hyperexcitability and improvements in neuroprotection induced by chronic GENUS in epileptic mice could contribute to its effects on antiepileptogenesis and comorbidity modification. However, it remains unclear how GENUS regulates hippocampal activity.

We observed that the GENUS-entrained gamma activity was attributed to PVINs in V1. This is consistent with the report that PVINs in V1 are preferentially correlated with cortical gamma oscillations8. Our experiment revealed that the chronic GENUS-induced antiepileptogenesis was abolished with bilateral V1 PVIN ablation. Nevertheless, in the investigation of the visual circuit related to 40 Hz flicker, we found that inhibition of PVINs in layer 2/3 of V1 significantly decreased 40 Hz entrainment by GENUS, revealing that PVINs in superficial V1 are the specific cells contributing to gamma enhancement induced by GENUS.

The dLGN constitutes the principal relay station by which the retina transmits visual information to the visual cortex46,47,48. The dLGN shell in rodents comprises a thin lamina residing just beneath the optic tract48. We observed that GENUS significantly increased neuronal activation in the dLGN shell. Following neuronal ablation in the dGLN shell, GENUS did not impact 40 Hz entrainment. These results suggested that the dLGN shell in the visual circuit was involved in GENUS-entrained gamma activity. The dLGN is primarily composed of excitatory thalamocortical relay neurons that provide glutamatergic excitation to V158,59. Furthermore, neurons in the dLGN shell preferentially target superficial V146. We used virus tools and chemogenetics to demonstrate that the majority of PVINs in layer 2/3 of V1 were innervated by neurons in the dLGN shell. Inhibition of V1-projecting neurons in the dLGN shell clearly minimized GENUS-entrained 40 Hz power, indicating that the dLGN shell-superficial V1 pathway is related to GENUS-entrained gamma activity.

RGCs are responsible for receiving visual stimuli and include more than 30 distinct cell types47. Among RGCs, ooDS-GCs project to neurons in the dLGN shell, which, in turn, project to the superficial V146,47,48. ooDS-GCs respond robustly to motion stimuli in a specific direction and are enriched in CART48,60. RCGs respond to the onset and termination of a stationary flashing spot61. Therefore, these properties of ooDS-GCs make them the optimal candidate for involvement in GENUS-entrained gamma activity. Using chemogenetics, we observed that the critical cell type in the retina mediating the relay of GENUS to through the thalamocortical network is ooDS-GCs. The 40 Hz power increment induced by GENUS was significantly decreased when ooDS-GCs were inhibited. Finally, we observed that more SRSs occurred in the epileptic mice with the inactivation of PVINs in the superficial V1, excitatory neurons in the dLGN shell, or ooDS-GCs in the retina, even when they were treated with chronic GENUS. Therefore, the activation of retinal ooDS-GCs-excitatory neurons in the dLGN shell-PVINs in the superficial V1 pathway is required for GENUS-induced gamma entrainment and seizure susceptibility reduction.

In summary, these results provided compelling evidence that GENUS can alleviate epileptogenesis through a dedicated thalamocortical visual circuit. However, some limitations should be considered in this study. First, only a pilocarpine-induced epilepsy model was utilized. Although this epileptic model is the most commonly used to reproduce temporal lobe epilepsy and hippocampal sclerosis30, the etiology and pathophysiology during epileptogenesis are convoluted26,44. Thus, it is critical to determine whether chronic GENUS-induced antiepileptogenesis in our model is translatable to other models or humans. Second, we did not use multisite silicon probes to record the LFP of the hippocampus and V1 synchronously due to the limitations of our equipment. It remains unclear how GENUS affects the synchronization of excitatory neurons and pathological changes in the hippocampus of epileptic mice. Further investigation of the underlying mechanisms is warranted.

Methods

Animals

Adult male mice aged 8–20 weeks were used in this study. Animals were housed in a temperature-controlled (22–25 °C) specific-pathogen-free (SPF) room with a 12 h light–dark cycle (lights on at 7 AM) and constant air humidity (approximately 50%). Mice had free access to food and water. Protocols for this research were approved by the Institutional Animal Care and Use Committees of Anhui Medical University. Animal suffering and the number used were minimized throughout the study. The SPF C57BL/6J mice were purchased from Ziyuan Experimental Animal Science and Technology Company (SCXK[Zhe] 2019–0004, Hangzhou, China). PV-Cre mice (Stock No. 017320) and vGLUT2-Cre mice (Stock No. 028863) were obtained from the Jackson Laboratory. The Cartpt-Cre mice with a C57BL/6J background were generated using the CRISPR/Cas9 system by Gene&Peacebiotech Co., Ltd. (Jiangsu, China). The strategy for the knock-in region was Exon 3. The targeted genes of F0 mice were amplified by PCR and sequenced, and the chimeric mice were crossed with wild-type C57BL/6J mice to obtain F1 Cartpt-Cre mice.

Induction of SE

Mice were pretreated with scopolamine hydrobromide (1 mg/kg, intraperitoneal (i.p.), Glpbio, USA) to prevent peripheral cholinergic effects. After 30 min, pilocarpine hydrochloride (250 mg/kg, i.p., TargetMol, USA) was administered to induce SE; repetitive injections (50 mg/kg, i.p.) were administered as needed every 30 min. Seizure activity was recorded in five stages as previously described with minimal modifications49,62: Stage 1, face and vibrissae twitching, hyperactivity; stage 2, head nodding, myoclonic jerks, or unilateral limb clonus; stage 3, bilateral limb clonus or whole body paroxysmal convulsions; stage 4, tail hyperextension with persistent generalized convulsions; and stage 5, rearing and falling with generalized rigidity (Supplementary Movie 4). SE was defined as the occurrence of stage 3 or higher seizures that progressed to repeated behavioral seizures. Death was scored as 6 in the evaluation of seizure susceptibility after PVIN ablation in V1. Two hours after SE induction, diazepam (2 mg/kg, i.p., Kelun Pharmaceutical Company, China) was administered to terminate the seizures.

Once the required number of SE mice for the experiment was achieved, each mouse was marked with a number. The labels were entered into Excel (Microsoft Office 2016, USA), and SE mice were randomly assigned to either the SE group or GENUS group using Excel software to generate a table of random numbers.

Electrode implantation

Mice were anesthetized with isoflurane (3% induction, 1.5% maintenance, RWD Life Science, China) and fixed in a stereotaxic apparatus (RWD Life Science, China). A rostrocaudal incision was made in the skin over the skull after disinfection with 1% iodophor (Lircon Medical Technology Company, China). The underlying connective tissue was removed by wiping the skull surface with 3% hydrogen peroxide (Hengjian Pharmaceutical Company, China) to expose the skull. Two coordinate points, bregma and lambda, were used to ensure the skull was oriented horizontally. Holes were drilled through the skull, utilizing the Mouse Brain in Stereotaxic Coordinates63. The coordinates (mm), defined as anterior-posterior (AP) from bregma, medial-lateral (ML) from the midline, and dorsal-ventral (DV) from the skull surface. Bilateral EEG recording electrodes (Nanjing Greathink Medical Technology, China) were epidurally implanted over the parietal cortex (AP: −2.00, ML: ±2.00, DV: −0.50)43. Another two electrodes were positioned in the skull over the cerebellum as the ground and reference. Two copper wires were inserted into the cervical muscles to record electromyography (EMG). The electrodes were anchored with screws and fixed in place with dental cement. Following surgery, the mice were injected with ibuprofen (5 mg/kg, subcutaneously, Easton Pharmaceutical Company, China) and amoxicillin (50 mg/kg, i.p., Brilliant Pharmaceutical Company, China) and recovered from anesthesia on a heated pad before being returned to their cages. The mice were allowed to recover for a minimum of 7 days before conducting EEG recordings. Detailed information about electrodes and histological images was shown in the Supplementary Fig. 7a–d.

Visual stimulation protocol

The visual stimulation equipment consisted of four modules (Supplementary Fig. 8a), including a low-voltage power supply (24 V), pulse modulator (1–999 Hz), LED strip (24 V, 20 W, and 4000 K), and a cage with one transparent side (30 × 17 × 20 cm; Luminance: 991.96 ± 226.29 Lux) (Supplementary Fig. 8b). LED flicker is a periodic square-wave light stimulus with a 50% duty cycle. For electrophysiological recording, the mice were placed in the EEG laboratory for 120 min then transferred to the cage for light stimulation. A multi-channel physiological signal recording system (Nanjing Greathink Medical Technology, China) was used to record the LFP. Each animal was subjected to 200 Hz flicker (namely, a mimetic light circumstance) for 40 min to obtain the EEG baseline. The pulse modulator was switched to a 40 Hz flicker (light on and off for 12.5 ms) for another 40 min. In the chemogenetics cohorts, the flicker frequency and duration were adjusted according to experimental requirements.

For immunohistochemistry, mice were placed in a dark chamber for 6 h to deactivate neurons related to light stimulation entirely, then the mice received 200 Hz or 40 Hz flicker for 60 min. During the chronic GENUS period, SE mice received light treatment 72 h after recovering from SE. They received 40 Hz flicker once a day for 6 weeks utilizing 60 min of stimulation each time. All experiments were conducted between 8 AM to 12 PM.

We monitored the movement trajectories of animals in the cage during a 60 min GENUS treatment for 3 consecutive days (Supplementary Fig. 8c). In all, 93.33% (14/15) of the mice received efficient light stimulation for no less than 50 min during the entire GENUS treatment (Supplementary Fig. 8d). GENUS can drive 40 Hz oscillations in mice even when they were not directly looking at the light (Supplementary Fig. 8e–h). This may benefit from our equipment design, where the cage has only one transparent side.

LFP power spectrum

EDFbrowser software (version 2.04, https://www.teuniz.net/edfbrowser) was used to analyze the relative power and PSD of the LFP. The EEG frequency was categorized based on the recommendation of the International Federation of Clinical Neurophysiology11, which included delta (1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), and gamma (30–80 Hz). EEG data were down-sampled to 500 Hz with a bandpass filter at 1–100 Hz or 10–100 Hz in the relative power analysis of the frequency bands or individual gamma frequency, respectively. In the relative power analysis (Window: Hamming; Blocksize: 5000 samples; Overlap: 50%; window width: 10 s; step size: 10 s), 20–25 segments for each EEG trace with 10 s of each segment were obtained for analysis. In the GENUS experiment, a relative power analysis was performed for the EEG traces without frequent spikes or apparent artifacts. In the PSD analysis (Segment length: 10 s; Block length: 3 s; Overlap: 50%; Window: Hamming; Logarithmic), the EEG data were obtained utilizing a notch filter at 50 Hz and analyzed for PSD.

Spike detection

AcqKnowledge 4.2.0 software (BIOPAC System, California, USA) was used for epileptiform spike detection. The EEG signal was down-sampled with 500 Hz and then bandpass filtered at 5–80 Hz via an infinite impulse response with a 50 Hz notch filter. EEG traces with obvious artifacts were discarded. A spike was defined as a sharp deflection with an amplitude greater than the median plus 4.5 times the standard deviation of the filtered signal with a duration of 20–100 ms, avoiding contamination of the standard deviation4,20,64. After the spikes were screened based on the amplitude threshold, the morphology of each spike was further confirmed through manual identification65,66. Spikes significantly deviating from the EEG baseline and those lacking rapid negative or positive polarity were discarded (Supplementary Fig. 1a and corresponding Supplementary Table 1a, b, Supplementary Fig. 9a, b). During SRSs recording, EEG traces obtained from the first 24 h were used for spike detection. The effective analysis time showed no significant differences among groups (Supplementary Fig. 9c–g).

Spike detection during NREM sleep

To comprehensively evaluate the effect of 40 Hz light stimulation on epileptiform discharges, we detected spikes that occurred during NREM sleep. Mice were allowed to habituate to the EEG recording cage for 24 h. First, the video EEG signals were acquired for 3 h during 200 Hz flicker as the baseline. Then, the pulse modulator was switched to 40 Hz flicker for 60 min. Finally, we switched the pulse modulator back to 200 Hz for 60 min after 40 Hz flicker was completed. EEG and EMG signals were bandpass filtered from 1 to 100 Hz and 10–100 Hz, respectively, with a 50 Hz notch filter. The PSD of EEG at 1–30 Hz was calculated as described in “LFP power spectrum”. Vigilance states were identified manually by visual inspection of EEG, EMG, and the PSD of EEG. Wakefulness was defined as high EMG with high frequency and low amplitude EEG. NREM sleep was defined as low-frequency and high-amplitude EEG with low EMG. During rapid eye-movement sleep, EEG showed high frequency and low amplitude signals, while EMG remained low67,68 (Supplementary Fig. 7e). In this study, we analyzed the signals with a duration of one hour before and after 40 Hz flicker. Subsequently, spikes were detected using the method described in “Spike detection”. The effective analysis times were 26.86 ± 7.17 min, 12.30 ± 4.62 min, and 25.55 ± 5.58 min for before light stimulation, during 40 Hz light flicker, and after light stimulation, respectively.

Sucrose preference test

Mice were raised individually and given a free choice between two bottles, one containing a 1% sucrose solution and the other containing purified water. An adaptation period of 48 h was utilized. The bottle positions were changed every 12 h. Food and water were withheld for 12 h during the testing period, and then mice were allowed to access the 1% sucrose solution and purified water freely for 12 h. The bottle positions were changed at the sixth hour. The percentage of sucrose solution consumption compared to total fluid consumption was calculated as the sucrose preference index.

Open field test

Mice were introduced into the center of an open, square, white box (50 × 50 × 50 cm) and allowed to roam freely for five min. Movements were recorded using a video camera placed over the box. The bottom was divided into 16 squares. The four central squares constituted the “center area,” and the remaining squares were designated the “peripheral area.” VisuTrack (Shanghai XinRuan Information Technology, China) or ANY-maze software (version 4.99 m, Stoelting, USA) was used to analyze the movement paths. After each trial, the box was cleaned with 75% alcohol to eliminate the influence of odor on the next mouse.

Elevated plus maze

The elevated plus maze consisted of two open arms (30 cm) and two closed arms (30 cm). The maze was positioned 75 cm above the floor. Each arm was 6.5 cm wide, and the closed arms were enclosed in 15 cm high non-transparent walls. Each mouse was initially placed in the intersection of the open and closed arms, facing one of the open arms and allowed to freely explore for 5 min. Movement patterns were recorded using a video camera placed over the apparatus and analyzed with VisuTrack software (Shanghai XinRuan Information Technology, China). The maze was cleaned with 75% alcohol between each test.

Morris water maze

The Morris water maze was composed of a circular black pool (diameter 120 cm, height 40 cm) with a circular platform (diameter 7.5 cm) submerged one cm below the water. The pool was divided into four quadrants. The platform location was set as the target quadrant. Four markers with different shapes and colors were posted on the walls of the four quadrants. Nontoxic white pigment was added to water to make it opaque. In the place navigation test, the mice entered water from the four quadrants with their head initially facing the wall. They were allowed to swim freely for 60 s to explore the maze, find the platform, and stay on it for 15 s. Mice were trained four times with an interval of at least 20 min each day for 4 days. The time spent traveling from the starting point to the platform was recorded as the escape latency. Mice were guided to the platform and allowed to stay on it for 15 s if they could not find the platform in 60 s. In that case, the escape latency was recorded as 60 s. The platform was removed in the spatial probe test on the 5th day. The mice were released at a point opposite the target quadrant and allowed to swim for 90 s. The swim path for each mouse was recorded and analyzed using VisuTrack (Shanghai XinRuan Information Technology, China) or ANY-maze software (version 4.99 m, Stoelting, USA). After each trial, the mice were wiped dry and kept warm before returning to their cage.

VEEG recording

Electrode implantation and postoperative care were conducted as described in “Electrode implantation.” After recovery for 7 days, a multi-channel physiological signal recording system (Nanjing Greathink Medical Technology, China) was used to record the VEEG. Mice were allowed to habituate to the test cage for 24 h, then EEG traces were acquired over 3 days69. Animals were freely moving and had access to food and water during VEEG recording. EEG data were down-sampled using SIRENIA® SEIZURE (Pinnacle Technology, USA) at 500 Hz and bandpass filtered from 5 to 80 Hz with a 50 Hz notch filter. Epileptic seizures were identified by a cluster of repetitive epileptiform spikes that persisted for more than 10 s with amplitude three times more than the baseline EEG70. Spontaneous behavioral seizures captured on video were scored based on an adapted Racine scale, as described in the section of “Induction of SE”49,62. SRSs were considered as non-convulsive if the mouse presented electrographic seizures but had no stage 3–5 behavioral seizures (Supplementary Fig. 3a, b and corresponding Supplementary Movies 1 and 2)31,41. Convulsive seizures were defined as a pattern of electrographic seizures accompanied by spontaneous behavioral seizures of at least stage 3 (Supplementary Fig. 3c and corresponding Supplementary Movie 3)31,41. Two investigators, who were blinded to the experimental protocols, analyzed the EEG recordings and videos. Any discrepancies during analysis were resolved by the senior investigator (Y.W.).

Virus injection

The surgical procedure before sphenotresia was performed as described in “Electrode implantation.” The coordinates (mm) were as follows: V1 (AP: −3.40; ML: ±2.50; DV: −1.20); the dLGN shell (AP: −2.30; ML: ±2.15; DV: −2.55); layer 2/3 of V1 (AP: −3.40; ML: ±2.50; DV: −0.75). Small holes were drilled through the skull, and 50 to 250 nl of virus was injected bilaterally via a microsyringe into the target area at a rate of 20 nl/min. Following fluorescent tracing, viral tools were employed: AAV2/9-hSyn-EGFP-P2A-CRE-WPRE-hGH pA (5.22E + 12 vg/mL) and AAV2/9-EF1α-flex-taCasp3-TEVp-WPRE-hGH pA (5.40E + 12 vg/mL) were used for neuronal ablation. AAV2/1-hSyn-CRE-EGFP-WPRE-hGH pA (1.04E + 13 vg/mL) was injected in the dLGN shell to infect postsynaptic neurons in V1. AAV2/9-EF1α-DIO-hM4D(Gi)-mCherry-WPREs (5.25E + 12 vg/mL) and AAV2/9-EF1α-DIO-mCherry-WPRE-hGH pA (5.18E + 12 vg/mL) were used to Cre-dependently express the inhibitory designer receptors exclusively activated by designer drugs (DREADD), mCherry was used as a control. AAV2/Retro-hSyn-CRE-EGFP-WPRE-hGH pA (5.22E + 12 vg/mL) was injected in the V1 and the dLGN shell to retrogradely express Cre-recombinase in neurons projecting to the injection site.

For intraocular injections, mice were anesthetized with isoflurane (3% induction, 1.5% maintenance, RWD Life Science, China) and fixed in a stereotaxic apparatus (RWD Life Science, China). 0.5% tropicamide (Bausch Lomb FREDA Pharmaceutical Company, China) and 1% phenylephrine hydrochloride (Harvest Pharmaceutical Company, China) were applied to dilate the pupils for 10 min. Erythromycin ointment (Renhe Pharmaceutical Company, China) was used to prevent the eyes from drying and infection. The eyeballs were penetrated at the corneoscleral limbus with the tip of a 26-gauge needle. Then, 1.0 µl of virus was injected into the vitreous chamber bilaterally using a microsyringe (Gaoge, 1.0 µl, China). AAV2/9-EF1α-DIO-hM4D(Gi)-mCherry-WPREs (5.25E + 12 vg/mL) was used to Cre-dependently express inhibitory DREADDs, and AAV2/9-EF1α-DIO-mCherry-WPRE-hGH pA (5.18E + 12 vg/mL) was used as a control.

To verify the antiepileptogenesis induced by chronic GENUS through the retina-dLGN shell-superficial V1 pathway, AAV2/9-EF1α-flex-taCasp3-TEVp-WPRE-hGH pA (5.40E + 12 vg/mL) was used for specific neuronal ablation, including PVINs in the superficial V1 of PV-Cre mice, excitatory neurons in the dLGN shell of vGLUT2-Cre mice, and ooDS-GSs in the retina of Cartpt-Cre mice.

After the injection was completed, the microsyringe was maintained in its position for 10 min and then slowly withdrawn. The EEG electrode implantation and postoperative care were the same as mentioned in “Electrode implantation.” We manipulated the neural circuit using hM4D (Gi) to produce inhibition. CNO (2 mg/kg, APExBIO, USA) was injected intraperitoneally. All AAV tools were obtained from BrainVTA (Wuhan, China).

Immunofluorescence staining

Mice were anesthetized using pentobarbital sodium (40 mg/kg, i.p., YAOPHARMA, China) and perfused transcardially with 30 ml of saline, followed by 20 ml of 4% paraformaldehyde (Servicebio, China). The brains were removed and fixed in fresh 4% paraformaldehyde overnight, then the brains were dehydrated in 30% sucrose solution for 48 h. Subsequently, the brains were embedded in OCT compound (Biosharp, China) and serially cut into rostrocaudal sections (40 μm, Leica, CM3050S, Germany). To minimize selective bias, brain sections were harvested from the locations approximately −1.58 mm and −2.70 mm from bregma for conducting dorsal and ventral hippocampal pathological examinations, respectively. In each animal, three sections spaced at intervals of 120 μm (i.e., one section was selected from every three consecutive sections) were used. The sections were immersed in an anti-freeze solution (phosphate buffer solution (PBS):ethanediol:glycerol = 5:3:2) and stored at −20 °C. The eyeballs were fixed in 4% paraformaldehyde for 4 h, then the retinas were isolated and underwent immunofluorescence staining immediately.

The brain sections and retinas were permeabilized using 0.3% Triton X-100 (Servicebio, China) for 60 min at room temperature (RT). After three PBS washes, the sections were blocked in QuickBlock™ blocking buffer (Beyotime, China) at RT for 2 h and then incubated with antibodies overnight at 4 °C: rabbit anti-NeuN (1:400, Cat. #24307S, CST, USA), mouse anti-GFAP (1:400, Cat. #3670S, CST, USA), rabbit anti-Iba-1 (1:400, Cat. 019-19741, Wako, Japan), rabbit anti-parvalbumin (1:600, Cat. #80561S, CST, USA), mouse anti-parvalbumin (1:600, Cat. NBP2-50038, NOVUS, USA), rabbit anti-c-fos (1:600, Cat. #2250S, CST, USA), rabbit anti-CART (1:100, Cat. #14547S, CST, USA), rabbit anti-CaMKII (1:400, Cat. ab52476, Abcam, USA), rabbit anti-VGLUT2 (1:200, Cat.GTX133142, GeneTex, USA). The sections were washed five times in PBS and incubated with secondary antibody at RT for 2 h in the dark: goat anti-rabbit Alexa Fluor® 488 (1:400, Cat. GB25303, Servicebio, China), donkey anti-rabbit Cy3 (1:400, Cat. GB21403, Servicebio, China), goat anti-mouse Alexa Fluor® 488 (1:400, Cat. GB25301, Servicebio, China), donkey anti-mouse Cy3 (1:400, Cat. GB21401, Servicebio, China). After five PBS washes, the sections were stained with DAPI (Servicebio, Cat. GDP1024, China) for 5 min at RT, then the DAPI was washed off and the sections were coverslipped using an antifade mounting medium (Servicebio, Cat. G1401, China).

Images were acquired at the same exposure using a fluorescence microscope (Leica, DM6 B, Germany). ImageJ (version 1.54 f, USA) was used to count fluorescent-positive cells. In the analysis of microglial cell body diameters and process lengths, immunostaining images (×20) were derived from stacks of 15–25 images at 2.0 μm z-intervals from 40-μm sections using an inverted confocal microscope (Zeiss, LSM 980, Germany). Microglia within hippocampal CA1, CA3, and DG regions were quantified. The cell bodies were defined as substantial Iba-1 staining bodies with a clear center location and blue nuclear DAPI staining. Areas with slight Iba-1 staining without clear cellular characteristics were not included in the analysis. The process lengths were measured using the NeuronJ plugin in ImageJ. Three frames (150 × 150 μm), normalized by ImageJ in each hippocampal region, were used for analysis. The mean value derived from three sections was used as the final result for one animal. Microglial skeletons were constructed using the skeletonize tool in ImageJ to clarify the microglial processes71. For brain sections co-labeled with c-Fos and PVINs, including the visual cortex (V1, V2L, and V2M) and hippocampus (dCA1, dCA3, vCA1, and vCA3), confocal image stacks from 40 μm sections at 2.0 μm z-intervals were collected using an inverted confocal microscope LSM 980. These regions on each section were delineated using ImageJ based on the Mouse Brain in Stereotaxic Coordinates63. C-Fos positive nuclei were identified when they exhibited a clear increase in immunoreactivity with the appropriate size and shape compared to the background. The cell bodies of PVINs were defined as substantial PV-positive bodies with a clear central location and blue nuclear DAPI staining. A PV+ + c-Fos+ double-labeled cell was recognized when a PV-positive cell contained c-Fos-positive puncta. The full field of the target regions on each section was chosen for quantification of c-Fos-expressing PVINs (normalized to the total number of PVINs). Values from three sections were averaged to determine the mean ratio of c-Fos-positive PVINs in each mouse. All images were acquired and processed with identical parameter settings. The experimenter performing the image analysis was always blinded to the treatment.

Nissil staining

To display histological images of the brain-implanted electrodes, Nissl staining was conducted. The hippocampal sections were prepared as mentioned in the section “Immunofluorescence staining". The sections were immersed in the Nissil solution (Cat. G1036, Servicebio, China) for 5 min and then washed with distilled water. After being differentiated with 0.1% glacial acetic acid (Cat. A801294, MACKLIN, China), the sections were dried in an oven at 37 °C, followed by sealing with neutral gum (Cat. WG10004160, Servicebio, China). They were subsequently observed with a fluorescence microscope (Leica, DM6 B, Germany).

Statistical analysis

The data distribution was evaluated by calculating the kurtosis and skewness. Data were presented as means ± SEM or medians (interquartile ranges [IQR]) for variables with normal distribution or skewed distribution, respectively. Student’s t test, Kruskal–Wallis test or Mann–Whitney U test, and one-way or two-way analysis of variance (ANOVA) followed by Bonferroni or Tukey HSD post hoc test were used to analyze differences among groups. Paired-samples t test or repeated measures two-way ANOVA were used to analyze the relative power before and after treatment with GENUS or CNO. The differences in escape latency in the Morris water maze among the groups were analyzed using repeated measures two-way ANOVA. Kaplan–Meier analysis was used to analyze survival probability after SE induction. The seizure score within 30 min following pilocarpine administration was analyzed using Generalized Linear Models. Statistical analysis was carried out using SPSS software (version 22.0, IBM Inc., USA). All graphs were prepared using GraphPad Prism (version 8.4.2, USA). The images were processed using Photoshop CS6 (Adobe Systems, USA). All schematic diagrams within the experimental schedule were generated using Adobe Illustrator 2025 (Adobe Systems, USA). Based on different experimental protocols, we employed five mice per group for hippocampal pathological tests, 6–12 mice per group for epidural EEG recording, and 8–12 mice per group for behavioral experiments. The specific sample number for each subfigure was detailed in its legend. A P value < 0.05 was considered statistically significant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.