Preferential pruning of inhibitory synapses by microglia contributes to alteration of the balance between excitatory and inhibitory synapses in the hippocampus in temporal lobe epilepsy

Abstract Background A consensus has formed that neural circuits in the brain underlie the pathogenesis of temporal lobe epilepsy (TLE). In particular, the synaptic excitation/inhibition balance (E/I balance) has been implicated in shifting towards elevated excitation during the development of TLE. Methods Sprague Dawley (SD) rats were intraperitoneally subjected to kainic acid (KA) to generate a model of TLE. Next, electroencephalography (EEG) recording was applied to verify the stability and detectability of spontaneous recurrent seizures (SRS) in rats. Moreover, hippocampal slices from rats and patients with mesial temporal lobe epilepsy (mTLE) were assessed using immunofluorescence to determine the alterations of excitatory and inhibitory synapses and microglial phagocytosis. Results We found that KA induced stable SRSs 14 days after status epilepticus (SE) onset. Furthermore, we discovered a continuous increase in excitatory synapses during epileptogenesis, where the total area of vesicular glutamate transporter 1 (vGluT1) rose considerably in the stratum radiatum (SR) of cornu ammonis 1 (CA1), the stratum lucidum (SL) of CA3, and the polymorphic layer (PML) of the dentate gyrus (DG). In contrast, inhibitory synapses decreased significantly, with the total area of glutamate decarboxylase 65 (GAD65) in the SL and PML diminishing enormously. Moreover, microglia conducted active synaptic phagocytosis after the formation of SRSs, especially in the SL and PML. Finally, microglia preferentially pruned inhibitory synapses during recurrent seizures in both rat and human hippocampal slices, which contributed to the synaptic alteration in hippocampal subregions. Conclusions Our findings elaborately characterize the alteration of neural circuits and demonstrate the selectivity of synaptic phagocytosis mediated by microglia in TLE, which could strengthen the comprehension of the pathogenesis of TLE and inspire potential therapeutic targets for epilepsy treatment.


| INTRODUC TI ON
Epilepsy is a disease characterized by recurrent seizures resulting from the hyperexcitability and hypersynchrony of neurons. 1,2 It affects over 65 million people worldwide and is triggered by a variety of factors, including traumatic brain injury, brain tumors, stroke, etc. 3 Temporal lobe epilepsy (TLE), originating from tissue damage in the hippocampus or amygdala, is the most common form of epilepsy. 4 Upon the onset of seizures, the brain typically undergoes reorganization of neural circuits and activities. Furthermore, people will eventually develop spatial learning and memory deficits with the progression of epilepsy. 5 Despite the existence of effective treatments that primarily target ion channels to inhibit the abnormal firing of neurons, approximately one-third of patients with TLE do not respond to medical treatment, making it the most prevalent form of drug refractory epilepsy. 6 Thus, there is a critical need to elucidate the pathogenesis of epilepsy, which may facilitate the development of more effective and non-invasive therapies.
Until now, it has been generally assumed that the emergence of epilepsy is strongly interconnected with the reconfiguration of neural circuits in the hippocampus. 7 The trisynaptic circuit is one of the most classic neural circuits in the hippocampus in which the dendrites of granule cells in the dentate gyrus (DG) of the hippocampus receive an excitatory transmission from the perforant path axons, and then project through the mossy fibers to the dendrites of the pyramidal cells in the cornu ammonis 3 (CA3), and finally project to pyramidal cells of CA1 via Schaffer's veins. [8][9][10] In this circuit, a seizure originates in a primary lesion and projects to another region where the response is intensified, even amplifying paroxysmal firing and spreading the seizure to other brain regions. The discharge of epilepsy is thus prolonged in space and time, resulting in a seizure more intense than if the primary lesion had caused seizures. 11,12 Accumulated evidence indicates that the balance between excitatory and inhibitory synapses (E/I balance) may play a crucial role in aberrant electrical signals in neural circuits. 2,13 The mainstream opinion on the underlying mechanism of epilepsy is that the E/I balance is disturbed, where excitatory synaptic transmission is enhanced while inhibitory synaptic transmission is weakened during epileptogenesis. 14,15 Glutamate is crucial for the development of mammalian learning, memory, and cognitive function. However, excessive glutamate synaptic transmission can also result in neurotoxic effects and lead to a variety of neurological disorders such as epilepsy. 16,17 The distribution and overall changes of excitatory/inhibitory synapses during epileptogenesis have been briefly studied in the hippocampus in experimental animal models and clinical brain specimens. In a rat model of TLE, it was determined that the excitatory synaptic protein vesicular glutamate transporter 1 (vGluT1) is predominantly expressed in the stratum radiatum (SR) of CA1, the stratum lucidum (SL) of CA3, the outer and inner molecular layer (OML, IML), and the polymorphic layer (PML) of the DG and subiculum (SUB) in the rat hippocampus. 18 Moreover, vGluT1 immunoreactivity in the human epileptic hippocampus was prominent in the stratum pyramidale (SP) of the CA, SR, SL, and the IML of the dentate gyrus (DG) and SUB.
Intriguingly, the intensity of vGluT1 in human epileptic hippocampal SR of CA1 and the SLM and OML of DG was significantly lower than in rats. 19 In addition, vGluT1 protein levels were higher in all hippocampal subregions of patients with non-hippocampal sclerosis (non-HS) than in patients with HS and the control group. 19,20 Despite their small number, interneurons are essential for the homeostasis of neural circuits in the brain, where they attach to specific excitatory inputs and provide feedforward and feedback inhibition to pyramidal neurons, maintaining homeostasis of neural activity in the brain. 21,22 Gephyrin, an inhibitory postsynaptic protein, is widely distributed in the CA1, CA3, and DG of TLE rats. 23 In addition, the onset of seizures reduced gephyrin expression in the hippocampus and adjacent neocortex. 24 Furthermore, gephyrin significantly downregulated in the human temporal neocortex in patients with mesial temporal lobe epilepsy (mTLE). 25 Although pervasive studies have documented these changes, no studies have clarified how excitatory and inhibitory synapses change in the region-specific network of the trisynaptic circuit during epileptogenesis. Moreover, the expression of glutamate decarboxylase 65 (GAD65) was found to increase in Ihara's epileptic rats and older spontaneously epileptic rats. 26 GAD65 in remaining hippocampal GABA neurons also increased in pilocarpine-treated rats. 27 However, a decrease of GAD65 was observed in the sensorimotor cortex in TLE. 28 Given these, controversy still exists regarding the alteration of the number of synapses in different epilepsy nidus and models.
Glia cells are vital components of nerve cells, with microglia and astrocytes constituting the largest population. [29][30][31] Abnormal activation and release of inflammatory mediators by microglia and astrocytes exacerbate epileptogenesis. [32][33][34] However, recent evidence has pointed out that microglia and astrocytes also contain anticonvulsant properties. [35][36][37] Microglia are resident immune cells in the brain 38 that engulf cell debris, facilitate neural development, and prune synapses, thereby preserving brain homeostasis. 31 In the physiological microenvironment of the brain, microglia assume a ramified form to maintain the structure and function of synapses via their processes. 39 Diseases such as epilepsy lead to the formation of ameboid microglia, which prune abnormally formed synapses. [39][40][41][42] In some cases, however, microglia can also prune both excitatory and inhibitory synapses, but the E/I ratio continues to favor excitation. 43,44 Interestingly, microglia tend to prune synapses with less active inputs. 45 Unfortunately, the selectivity of microglial phagocytosis in epilepsy is unknown. Furthermore, it remains to be determined how the E/I balance shifts in epilepsy after microglial pruning.
In this study, we generated a kainic acid (KA)-induced epilepsy model in rats, which were examined at various time points to explore the changes in the number of excitatory and inhibitory synapses, along with the preference for microglial phagocytosis in the hippocampus. Electroencephalography (EEG) recording demonstrated stable spontaneous recurrent seizures (SRS) in rats 14 days after KA administration. We observed a continuous increase in excitatory synapses during epileptogenesis, while inhibitory synapses decreased significantly. In addition, vGluT1 occupied a high proportion in the SR, SL and PML during epileptogenesis. Moreover, we found that microglia preferentially pruned inhibitory synapses after the formation of recurrent seizures. Consistent results were also validated regarding the selectivity of microglial phagocytosis in the resected hippocampus in mTLE patients. Taken together, these findings can aid in elucidating the potential mechanism by which microglia contribute to synaptic alterations in TLE-derived neural circuits, inspiring novel therapeutic interventions targeting microglia for TLE along with its complications.

| Animals
Male Sprague Dawley (SD) rats (180-200 g) were used and purchased from Shanghai SLAC Laboratory Animal Co., Ltd (certificate: SCXK 2022-0004). Animals were kept in the Laboratory Animal Center of Zhejiang University and were group housed in three per cage, in stainless cages under a 12-h/12-h light/dark cycle at room temperature of 22 ± 1°C, humidity of 50 ± 20% with food and water ad libitum. The animal protocols were approved by the Animal Care and Use Committees at the Zhejiang University School of Medicine and were conducted in accordance with the policies of institutional guidelines on the care and use of laboratory animals.

| Human samples
The human cases investigated in this study were obtained from the First Affiliated Hospital, Zhejiang University School of Medicine.
The information of patients is shown in Table 1. One brain slice of every patient was used for Nissl staining to confirm that the site of surgery was taken in the temporal lobe and hippocampus.
Otherwise, the patients were excluded. Patients (n = 6) underwent surgical resection treatment of a varying part of the temporal lobe for medically intractable TLE. The whole protocols were approved by the Ethics Committee of the First Affiliated Hospital, Zhejiang University School of Medicine (No. 531) and followed the guidelines of the Declaration of Helsinki. Informed consent was obtained from the families concerned before specimen collection.

| Kainic acid-induced epilepsy model
Rats were randomly assigned to one of five groups: control group, which received an intraperitoneal (i.p.) injection of saline, and 7, 14, 28, and 63 days following the induction of epileptic seizure, which received an i.p. injection of 18 mg/kg kainic acid (ab120100, Abcam). Each group consisted of 6 SD rats. The Racine scores were applied to rate seizure severity: category 1, immobility and facial twitch; category 2, head nodding; category 3, forelimb clonus; category 4, rearing; and category 5, rearing and falling. 46 According to previous studies, the onset of epilepsy was defined when at least two generalized motor seizures (stage 4-5) were observed 48 h post status epilepticus. 47

| Tissue preparation
The human cases were fixed in 10% formalin fixing solution (P2161, Solarbio, CHN) and dehydrated with ethanol. After paraffin embedding, 5 μm sections were cut using a rotary slicer (HM325, Thermo) and were prepared for Nissl staining and immunofluorescence.
Anesthetized rats were transcardially perfused with phosphatebuffered saline (PBS), after which the rat brains were fixed with 4% paraformaldehyde (PFA) overnight at 4°C, then moved to a 15% sucrose solution overnight, and finally to a 30% sucrose solution.
A sliding microtome (HM520, Microm) was used to cut 30 μm brain slices, which were then preserved in a cryoprotectant at −20°C.
Three animals per group were used to harvest 3-4 matching slices between 4.56-and 3.60-mm posterior to the bregma in one hemisphere of the rat brain for each experiment.

| Synaptic puncta staining and analysis
All confocal images of vGluT1 or GAD65 were obtained using Z-stack (1 μm step) with a 63-objective oil with 2× zoom. Synaptic puncta analysis was performed with ImageJ. For all rats in the experiment, images were captured with a 1 airy unit (AU) pinhole while keeping the gain and offset settings constant. Three sites per slice were captured for analyzing. For the analysis, the Gaussian blur of each image was set to 1 and all contrasted photos were set to the same threshold, then the total area of vGluT1 and GAD65 were calculated in the same regions of each rat, Age-matched control rats were employed to exclude the age-related dynamics of all parameters, the results of rats at different ages were normalized to the age-matched control ( Figure S1, S2). For display, post-analysis photos were brightened and contrasted over the whole image.
The lysosomal marker, immunolabeling of CD68 was applied to identify the phagocytic microglia. Co-labeling of lysosomal marker CD68 and vGluT1 or GAD65 was used to evaluate the synaptic pruning of microglia. Sections were imaged with the 63× objective oil with 2× zoom using the confocal microscope at 1 μm Z-stack step.
Images on each stack were overlaid for microglia counting. Three sites per slice were selected for counting. ImageJ was used to remove the noise from the image. The Gaussian blur per image was set to 1, and plot profile in ImageJ was used to reconstruct the microglia and synaptic volumes. The number of IBA1 positive cells co-stained with DAPI was counted manually.

| Electroencephalography recording
The KA model of epilepsy has been expounded previously. Rats typically develop SRSs 2 weeks after the onset of status epilepticus (SE). 50 14 days after KA injection, rats were surgically implanted with epidural electrodes in the cortex for EEG recording. The rats were first anesthetized with 3% pentobarbital sodium. Then recording electrodes were implanted on the left side of the rat (AP: − 4.08 mm; ML: 3.6 mm; DV: − 0.8 mm). And two screws were placed on the right and inferior sides of the cerebellum's skull to serve as reference and ground electrodes for EEG recording. The screws were welded through copper wire and fixed to the electronic pins with dental cement. Rats were allowed 7 days of recovery following the surgery.
Rats were placed in a 35 cm 2 square cage to move freely for 1 day prior to EEG recording at a sampling rate of 1 kHz (Spike 2, CED) synchronized with a digital video monitoring system. EEG was acquired using standard alternating current amplifiers without bandpass filters. Rats were monitored 12 h per day for seven consecutive days.
EEG data analysis was carried out by two trained observers who were not aware of the experimental groups. The video was evaluated when necessary to rule out sources of artifacts. All EEG data from each monitoring session was reviewed for hippocampal paroxysmal discharges (HPDs). The baseline of epilepsy in each rat is stable and reproducible, and seizures do not occur in clusters, which is a key prerequisite for reliable pharmacological studies. HPDs were defined as a spike in frequency ≥ 2 Hz, amplitudes at least 3-fold higher than baseline and duration longer than 10 s. Spikes with a baseline of fewer than 5 s between them were deemed to be the same seizure. 51 The number of HPDs, seizure duration and spiking amplitude per 24 h were counted.

| Kainic acid induces spontaneous recurrent seizures 14 days after status epilepticus onset
Previous studies demonstrated that the number of synapses could be altered in TLE. 52 Here, we conducted comprehensive research on the number of synapses in the hippocampus of rats, which was

| Kainic acid induces the alteration of excitatory synapses in rats' hippocampus
To explore the changes in excitatory synapses, we evaluated the changes in vGluT1, an excitatory synapse marker, in each functional region of the hippocampal CA1, CA3, and DG after KA induction (Figure 2A-C).
The quantitative analysis of immunofluorescence in rats' hippocampus indicated that in the SP of CA1 and CA3, KA induced a significant elevation of the vGluT1 + area during periods of epileptogenesis (7-28 days) and returned to control levels in a chronic phase of epi-  Figure 2D,E). In the granular cell layer (GCL) of the DG, however, there were no discernible changes in the vGluT1 + area across these time points following KA induction ( Figure 2F). Regarding the SR of the hippocampal CA1, the SL of CA3 and the PML of DG, the vGluT1 + area elevated incrementally from day 7 to 28, but returned to the con-

| Kainic acid induces the alteration of inhibitory synapses in rats' hippocampus
To quantify the changes in inhibitory synapses, we used immunostaining for GAD65, an inhibitory synapse marker in each functional region of the hippocampal CA1, CA3, and DG after KA induction ( Figure 3A-C). The quantitative analysis of immunofluorescence in the rat hippocampus indicated that KA induced a distinct increase in the GAD65 + area in the SP of CA1 on days 14 and 28.
The GAD65 + area in the SP of the CA3 increased significantly during epileptogenesis (days 7-28), which may signal the aggregation of inhibitory synapses, and returned to baseline levels during the chronic phase of epilepsy (day 63) (CA1 SP: **p < 0.01, 1.58 ± 0.08 vs.
1.00 ± 0.06, 28 days vs. control group, Figure 3D,E). In the GCL of the DG, however, no changes in the GAD65 + area were noted across these time points following KA induction ( Figure 3F)    In addition, the activation and phagocytosis of microglia in the SR, SL and PML were investigated. In the SR of CA1, we perceived abundant activated microglia on from days 14 to 63, whereas in the SL of CA3 and PML of the DG, activated microglia were discerned from days 7 to 63 (SR: ***p < 0.001, 6.28 ± 1.01 vs.
1.17 ± 0.23, 28 days vs. control group, ***p < 0.001, 6.72 ± 1.12 vs.  Figure 4O). In addition, we discovered a linear correlation between the number of activated microglia and CD68 + microglia in the CA1, CA3 and DG regions of the hippocampus ( Figure 4P-U). These results confirmed that activated microglia conducted phagocytosis during the period of SRS in the KAinduced TLE models.

| Kainic acid-induced activated microglia preferentially prune inhibitory synapses in the rats' hippocampus during spontaneous recurrent seizures
Microglia are thought to inappropriately activate and prune excess synapses during the development of other neurological diseases. 53,54 To further characterize microglia-mediated synapse pruning in TLE, IBA1, CD68, and vGluT1 or IBA1, CD68, and GAD65 triple staining was performed to determine the specificity of microglial phagocytosis ( Figure 5A,B). In TLE rats 28 and 63 days after KA induction, quantification of CD68 and vGluT1 immunofluorescent signals within IBA1 + microglia, as well as the quantification of CD68 + microglia with engulfed vGluT1 revealed that microglia did not prune excitatory synapses ( Figure 5C).
In the GCL of DG, though, no significant change was observed after the KA induction ( Figure 5J). These data supports a selective pruning of synapses by microglia during epileptogenesis.

| Microglia selectively prune inhibitory synapses in the hippocampus in TLE human patients
To further validate the selectivity of microglial phagocytosis in the human mTLE hippocampus, we next collected 18 brain tis-  Figure 6G).
Quantification of the number of CD68 + microglia with engulfed GAD65 or vGluT1 signals also revealed the preferential pruning of GAD65 by microglia, rather than vGluT1(***p < 0.001, 4.17 ± 1.35 vs. 0.28 ± 0.11, GAD65 vs. vGluT1, Figure 6E). Together, immunostaining in human tissue provides compelling evidence that microglia can modify the present neural circuits through specific phagocytosis of inhibitory synapses in TLE.

| DISCUSS ION
In this study, we conducted detailed research to determine how the number of excitatory and inhibitory synapses changed at multiple typical time points following KA induction. We discovered that the Glutamate is a prominent neurotransmitter in the brain that mediates excitatory synaptic transmission and is essential for synaptic plasticity. 55 We used vGluT1 to label excitatory synapses to investigate the variation in the number of excitatory synapses. We discovered that the number of excitatory synapses in the PML of the DG increased dramatically from days 7 to 28 after KA induction, which may indicate an increase in excitatory projections from the PML to the SL during epileptogenesis. In the DG of seizuresensitive gerbils, vGluT1 immunoreactivity was also enhanced. This is consistent with mossy fiber sprouting (MFS), a typical pathological change in epilepsy. Following epileptic damage by KA injection, mossy fibers lose projections from pyramidal cells in the CA3 and hilus and thus sprout abnormally into the IML and SL to reconstruct the synapse network. 56,57 This increase in excitatory input peaked on day 28, which is consistent with the sprouting number of mossy fibers, as previous studies in a model of KA-induced epilepsy in rats found that the density of abnormal synaptic connections formed by MFS was significantly lower at 7 days after status epilepticus than at 28 days. 58 The number of excitatory synapses decreased to the control level on day 63, which may be attributed to the massive neuronal death associated with HS 63 days after KA injection in rats.
However, there was no significant change in the number of excitatory synapses in the GCL at any time point. Similar results were also found in a pilocarpine-induced rat model, where there was a tiny difference in granule cell activity between pilocarpine-induced and control rats. 57 This may be interpreted as the massive death of inhibitory synapses in the SP of the CA3, the inhibitory projection via Schaffer's veins may correspondingly increase. 63  Our studies also demonstrate the synaptic pruning of microglia in rats 63 days following KA treatment. 34 Thus, the compensatory increase of inhibitory synapses in the SP and SR of CA1 could be predominant, which may result in the temporary increase of inhibitory synapses before day 63. Intriguingly, according to electrophysiological investigations, animal models of epilepsy exhibited reduced inhibition of CA1 pyramidal cells. 65 Given the inhibitory presynaptic marker that we employed, it is also necessary to incorporate postsynaptic markers to examine the number of functional synapses.
Microglia regulate the equilibrium of synapses during neurodevelopment by creating new synapses and removing non-functional ones. 66 According to our findings, the number of microglia was highly activated during spontaneous seizures. This is consistent with earlier research indicating that during the chronic phase of epilepsy, microglia were strongly active. 67 However, activated microglia might not possess phagocytic function; therefore, we co-stained microglial markers with lysosomal markers to quantify the number of phagocytic microglia. Interestingly, 14 days post-KA induction, the number of activated microglia exhibited a great quantity, but the number of CD68 + microglia was well below that on days 28 and 63, which reveals that microglia began to conduct prominent phagocytosis on day 28. 34,64 In addition, 63 days post-KA induction, the number of activated microglia reduced significantly compared to 28 days, which may be due to a large amount of neuronal death and synaptic damage that occurred during HS, which led to a considerable decrease in the number of microglia. A prior study also supports this phenomenon, which found amoeboid microglia expressed at a low level in human epileptic temporal lobes. 39 In addition, the weakened activation of microglia was accompanied with neuron loss in a mouse model of TLE. 68 However, CD68 + microglia still surpassed the control and were close to that on day 28, suggesting that microglia continued to conduct phagocytosis even during HS, which consisted of a major cause of cognitive impairment in epilepsy. Notably, when comparing the quantity of microglia in the CA1 with that in the CA3 and DG 28 days after KA induction, we discovered that the numbers of activated microglia and CD68 + microglia were substantially higher in the CA3 and DG than in the CA1. This is in line with the substantial neuronal death found in CA3 and DG in rats treated with KA, indicating the significance of CA3 and DG in the development of epilepsy. 69 In addition, we found that microglia preferentially pruned inhibitory synapses in epileptic brains. It is noteworthy that although the number of activated microglia on day 14 was considerably more than that on day 63, CD68 + microglia or microglia with engulfed GAD65 signal on days 28 and 63 were superior to that on day 14. Considering the statistical analysis above, it can be ratiocinated that microglia did not exert noticeable phagocytosis until 28 days after the KA induction. We also utilized human samples to demonstrate the validity of this conclusion. However, the mechanism of the preferential pruning by microglia is still obscure. Studies indicate that the synaptic pruning of microglia could be regulated by the classical complement molecules C3 and C1q. 64 Microglia preferentially prune inhibitory synapses in progranulin-deficient mice by activating C1q. 70 C1q was abundantly deposited and co-localized with inhibitory synapses engulfed by microglia in the multiple sclerosis (MS) hippocampus. 71 In addition to disease models, GABA-receptive microglia also remodel C1q tagged inhibitory synapses in mouse postnatal development. 72 Our prior research also found that intranasal infusion of C3a reversed microglia-mediated synaptic phagocytosis in epilepsy after captopril treatment. 34 These results suggest that microglia during SRS may highly express GABA receptors and C1q may act as an "usher" to help microglia recognize inhibitory synapses, all of which deserve further investigation.
We applied presynaptic markers (vGluT1 and GAD65) to investigate changes in the number of excitatory and inhibitory synapses.
Notwithstanding, combining presynaptic markers with postsynaptic markers can better reflect the function of synapses in neural circuits, 73 which will be further explored. In addition, electrophysiological studies may be added to confirm changes in the number of synapses in the future. In fact, postsynaptic density-95 protein (PSD95) exhibited decreased expression in epileptic rats. 74 Additionally, we noticed that some Iba + puncta colocalized only with synaptic puncta but not with CD68. Actually, in addition to phagocytosis, microglia generate microglial process pouches that enveloped the dendrites of neurons in mice treated with KA for 6 h without performing phagocytosis. 75 In complex febrile seizures, microglia could also displace GABAergic synapses by P2Y 12 receptor. 66 Thus, the function of microglia on synapses in epilepsy may be bidirectional. The specific role may be carried out depending on the stage of disease progression. Moreover, our study only tested male SD rats. However, the reactivity of microglia can differ in sex dimorphism, given microglia adopted ameboid morphology and exhibited increased phagocytic capacity only in the developing brains of males, while females possess decreased phagocytic capacity. 76 Furthermore, cerebrovascular reactivity reproducibility was slightly higher in males. 77 And there was a higher expression level of proteins involved in neurovascular metabolism and oxidative respiration in females, which is an essential element for the phagocytic function of microglia. 78,79 Therefore, studying both sexes should be a future subject.
In sum, our study demonstrated that microglia selectively pruned inhibitory synapses to modify neuronal circuits in the hippocampus, resulting in changes in the number of excitatory and inhibitory synapses, which ultimately led to E/I imbalance and epilepsy formation.
Our findings highlight a biological mechanism by which microglia contribute to synapse degradation. This underscores the importance of microglia in the pathophysiology of epilepsy and implies that microglia can be targeted for the prevention and treatment of epilepsy in clinical settings.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors have no financial conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

CO N S E NT FO R PU B LI C ATI O N
Not applicable.