Microglia induce auditory dysfunction after status epilepticus in mice

Auditory dysfunction and increased neuronal activity in the auditory pathways have been reported in patients with temporal lobe epilepsy, but the cellular mechanisms involved are unknown. Here, we report that microglia play a role in the disinhibition of auditory pathways after status epilepticus in mice. We found that neuronal activity in the auditory pathways, including the primary auditory cortex and the medial geniculate body (MGB), was increased and auditory discrimination was impaired after status epilepticus. We further demonstrated that microglia reduced inhibitory synapses on MGB relay neurons over an 8‐week period after status epilepticus, resulting in auditory pathway hyperactivity. In addition, we found that local removal of microglia from the MGB attenuated the increase in c‐Fos+ relay neurons and improved auditory discrimination. These findings reveal that thalamic microglia are involved in auditory dysfunction in epilepsy.


| INTRODUCTION
Epilepsy is a chronic brain disorder that involves synchronized neuronal bursting (Dominguez et al., 2005).Temporal lobe epilepsy (TLE), a type of epilepsy that accounts for approximately 60% of all epilepsies seen in adults, has been reported to be associated with auditory symptoms in some patients (Furia et al., 2022;Téllez-Zenteno & Hernández-Ronquillo, 2012).Auditory impairments include tinnitus, auditory hallucinations, and impaired sound discrimination (Han et al., 2011;Hauf et al., 2013;Meneguello et al., 2006).These symptoms make it difficult to process auditory information and reduce the patient's ability to communicate and their quality of life.The cellular mechanisms underlying auditory dysfunction in TLE remain unknown but have been reported to be associated with damage to the auditory regions.It has also been suggested that neuronal activity increases in the auditory cortex of TLE patients with auditory dysfunction (Hauf et al., 2013).In particular, Knipper et al. (2022) have demonstrated that increased neuronal activity in the central auditory pathway, resulting from decreased inhibitory activity, may underlie various auditory dysfunctions.
Because the epileptogenic process involves a collapse of the balance between excitatory and inhibitory synapses in some brain regions (Andoh et al., 2019;Fritschy, 2008), excessive neuronal Tasuku Araki and Toshimitsu Hiragi contributed equally to this study.activity may also occur in the auditory pathway due to the dysfunction of inhibitory synapses.In the primate and rodent auditory pathways, the medial geniculate body (MGB) of the thalamus, particularly the ventral MGB (vMGB), is the primary sensory input to the primary auditory cortex (A1) (Anderson & Linden, 2011;Felix et al., 2018) (Figure 1a).The vMGB receives both feedforward and feedback F I G U R E 1 Interictal spikes in the auditory pathway after status epilepticus.(a) A schematic diagram of the mouse auditory pathway.The inhibitory circuits of the auditory pathway mainly consist of feedforward inhibitory projections from the inferior colliculus (IC) to the medial geniculate body (MGB) (Peruzzi et al., 1997) and feedback inhibitory projections from the primary auditory cortex (A1) to the MGB through the thalamic reticular nucleus (TRN) (Yu et al., 2009).CN, cochlear nucleus; SOC, superior olivary nucleus.(b) A schematic diagram (left) showing the electrode positions (red, MGB; blue, A1) and Nissl staining (right) to identify the location of electrode implantation.Arrowheads indicate electrode traces in the MGB (top) and A1 in coronal sections (bottom).Scale bar, 500 μm.(c) The experimental paradigm for LFP recordings.One week after the date of electrode implantation (Day -9, Surgery), PBS was injected (Day -2) as a control for kainic acid (KA) injection (Day 0), and LFP was recorded on Day -1 (Prerecording; Pre-Rec.).On Day 0, KA was injected to induce SE, and LFPs were recorded on Days 1, 3, 6, 10, and 14.After recording on Day 14, mice were perfused, and histological staining was performed to confirm the electrode traces (b).(d) Left, a photo of a freemoving mouse with implanted electrodes.Right, a representative trace of LFPs was recorded from the MGB and A1 at Day 3. The green line indicates a threshold value of 500 μV from the baseline, and signals beyond this threshold are indicated with black points as the interictal spike.(e) Representative traces of bandpass filtered LFPs at 1-100 Hz simultaneously recorded from the MGB, A1, and the prefrontal cortex (PFC).(f) Frequency of interictal spike per second (event frequency) on each recording date under free-moving conditions.The number of mice used was as follows: MGB, n = 7-11; A1, n = 7-11; PFC, n = 3-7.Significant differences were determined by Student's t-test versus Pre after one-way ANOVA, followed by Bonferroni correction (*p < 0.05).The data are shown as the mean ± SEM.(g) Correlation of the timing of the onset of interictal spike in the MGB and A1.n = 7-11 mice.Significant differences were determined by paired t-test versus 0, followed by Bonferroni correction (*p < 0.05).The data are shown as the mean ± SEM.(h) The probability of observing an interictal spike at A1 when a spike is detected at MGB. (i) The probability of observing an interictal spike at MGB when a spike is detected at A1. inhibitory projections from the inferior colliculus (IC) (Peruzzi et al., 1997) and the thalamic reticular nucleus (TRN) (Yu et al., 2009), respectively.These inhibitory circuits play a crucial role in the sorting and processing of auditory information sent to the A1 (Rauschecker et al., 2010).Thus, changes in neuronal activity may disrupt auditory information processing in the brain, resulting in impaired auditory function.
In recent years, synaptic regulation by the brain-resident macrophage, microglia, has received considerable attention.Microglia play an important role in regulating the removal and formation of synapses during development and learning (Andoh & Koyama, 2021;Cornell et al., 2022).
Disruption of synaptic regulation by microglia has also been implicated in diseases such as autism and Alzheimer's disease.However, despite the importance of the balance between excitatory and inhibitory synaptic function in epilepsy, few studies have examined microglial regulation and the reality remains elusive (Andoh et al., 2019).Interestingly, it has recently been reported that microglia insert their processes between the neuronal somata and GABAergic synapses to displace synapses in mouse brains after febrile seizures (Wan et al., 2020).This is called synaptic stripping, and has been reported as a mechanism of synapse removal by microglia in the brain during inflammation (Chen et al., 2014;Trapp et al., 2007).Because GABAergic signaling can be depolarizing in the developing brain, GABAergic synaptic stripping may suppress excitability and be neuroprotective (Kasahara et al., 2019;Wan et al., 2020).However, in the adult brain, reducing hyperpolarizing GABAergic synapses may increase neuronal excitability through similar mechanisms.
Therefore, in this study, we tested the hypothesis that the disruption of inhibitory synaptic inputs in the vMGB causes increased neuronal activity in the auditory pathway, leading to impaired auditory discrimination after seizures.Kainic acid (KA), a glutamate receptor agonist, is used to generate animal models of epilepsy (Rusina et al., 2021).Systemic administration of KA to mice induces status epilepticus (SE), and although recurrent seizures are rare, pathological changes such as neuronal cell death and changes in neuronal activity are observed after several days.We recorded interictal spikes in the MGB and the A1 after SE.We found that auditory function, which was assessed by the auditory discrimination avoidance task, was impaired after SE.Using whole-brain and confocal fluorescence imaging, we found that microglia reduced inhibitory synapses from relay neurons in the vMGB, which persisted for 8 weeks after SE.Furthermore, we performed local injection of clodronate into the MGB to deplete microglia, and the reduction of microglia in the vMGB resulted in the increased c-Fos expression in vMGB relay neurons.Finally, depletion of microglia from the MGB also improved MGB-mediated auditory discrimination deficits.Thus, this study reveals a previously unknown role for microglia in auditory dysfunction after seizure.

| Animals
Experiments were performed with the approval of the animal experiment ethics committee at the University of Tokyo (approval number: P29-11, P29-14, P29-15, P4-4) and according to the University of Tokyo's guidelines for the care and use of laboratory animals.C57BL/6J male mice (SLC, Shizuoka, Japan) and CX3CR1 GFP/GFP male mice (Jackson Labs, Farmington, CT, USA, RRID:IMSR_JAX:005582) were housed in cages under standard laboratory conditions (a 12-h light/dark cycle, free access to food and water).Male CX3CR1 GFP/GFP mice were crossed with female C57BL/6J mice to obtain CX3CR1 GFP/+ mice.Mice were randomly assigned to experimental groups.Mice were 8-10 weeks old at the time of SE induction.Only male mice were used in this study to exclude the effect of the reproductive cycle on seizures (Rusina et al., 2021).All efforts were made to minimize the animals' suffering and the number of animals used.

| Seizure induction
In the KA-induced SE experiments, the ionotropic glutamate receptor agonist KA (30 mg/kg, i.p.) (Tocris Bioscience, Bristol, UK) was injected into 8-to 10-week-old C57BL/6J mice (22-28 g) and CX3CR1 GFP mice (22-28 g).KA-induced seizure severity was scored as follows according to Racine's scale with some modifications (Racine, 1972): stage 1, freezing behavior; stage 2, rigid posture with a straight and rigid tail; Stage 3, repetitive head bobbing and rearing into a sitting position with a shaking forelimb; stage 4, rearing and falling, jumping, and running with a period of total stillness; stage 5, continuous stage 4-level behaviors; and stage 6, loss of posture and generalized convulsive activity, usually preceding death.After KA injection, the mice progressed through the different stages, usually entering stage 1 in the first 15 min after the injection and reaching Stage 3 in 30-45 min.In this study, only animals with seizures that reached stage 5 and lasted at least 60 min were used.When mice did not reach stage 5 within 60 min of the first injection, they received an additional KA injection (5 mg/kg, i.p.) every 30 min.If no seizures were observed 30 min after the first injection, they were given another dose of KA (10 mg/kg, i.p.), followed by the standard 5 mg/kg dosing every 30 min until they reached stage 5.The duration of the seizure varied from 2 to 4 h.Control mice received phosphatebuffered saline (PBS) injection and exhibited no abnormal behaviors.
Up to four additional doses of KA were administered, and mice that did not reach stage 5 after four doses were excluded from the analysis.Diazepam (DZP) (10 mg/kg i.p.) was administered 30 min before KA treatment, and no seizures beyond Stage 3 were observed in the DZP-treated group even after four additional doses of KA.The number and percentage of mice that survived after KA treatment and average dose of KA in each experimental group were calculated as follows: SE, 34/65 mice (52%), 45.0 mg/kg (survived), 40.5 mg/kg (dead); Clo-SE, 15/32 mice (47%), 42.9 mg/kg (survived), 41.Three mice in the SE group and one mouse in the Clo-SE group were excluded from subsequent experiments because they did not exhibit stage 4 or higher seizures after the maximum dose (60 mg/kg).

| Auditory discrimination avoidance task
Mice were handled for 4 days (3 min for each mouse per day) before the start of the task.The auditory discrimination avoidance task was conducted using the Habitest system (Coulbourn Instruments, Holliston, MA) as previously described (Kolodziej et al., 2016;Kurt & Ehret, 2010).Briefly, a training session consisted of 3 min of habituation, 30 six-second Go trials, and 30 six-second NoGo trials.Each trial started with light stimuli.During the Go trial, tone stimuli with increasing frequencies (4-8 kHz, 65-75 dB) were delivered, and the mice had to cross the partition.A correct move was classified as Hit and a failure as Miss.In the NoGo trial, a falling tone stimulus (8-4 kHz, 65-75 dB) was delivered, and the mice had to stay in the same room without crossing the partition.The result of not moving was classified as correct rejection, and the result of moving by mistake was classified as false.If mice did not respond correctly to the Go and NoGo trials, a 150 μA electric shock was delivered as punishment.Go trials and NoGo trials were randomly performed interposed with 20-s intervals.Training was conducted for 10 or 12 days, and mice that were significantly different in the proportion of hits and false (χ 2 test, p < 0.05) on at least 3 days during each training period were used as individuals that were able to learn.Mice with no significant difference in the χ 2 test during the training period were excluded from the experiment as they were not able to learn the task.On the next day, mice received intraperitoneal KA injection, and 3 days later, test sessions were performed.To confirm that the mice did not lose their ability to hear sound itself, instead of comparing ascending tone (4-8 kHz) to descending tone (8-4 kHz), test compared pure tone (8 kHz) to silence (but with a light on to signal the start of the task).The number and percentage of mice that were able to learn the auditory discrimination task is as follows: pure tone vs. silent, 10/10 (100%); 8-4 kHz versus 4-8 kHz discrimination, 13/20 (65%); 8-4 kHz versus 4-8 kHz discrimination with clodronate, 17/31 (54%).

| Stereotaxic intra-MGB injection of clodronate
Mice anesthetized with isoflurane (Pfizer Inc., New York, NY) were placed in a stereotaxic frame following subcutaneous administration of ropivacaine (100 μL; Sandoz K.K., Tokyo, Japan).Clodronate (50 mg/mL, 300 nL; D4434, Sigma-Aldrich, St. Louis, MO) in PBS was pressure-injected into the MGB (3.1 mm posterior and 2.1 mm lateral to the bregma, 3.0 mm deep from the brain surface) at a rate of 50 nL/min using a syringe pump (KD Scientific Inc., Holliston, MA) connected to glass pipettes (30-0034; Harvard Apparatus, Holliston, MA).After 2 days of recovery from the surgery, mice received an intraperitoneal KA injection, and after an additional 2 days of recovery, mice completed the auditory discrimination avoidance task (test session).

| Surgery for electrode implantation
The mice were anesthetized with isoflurane gas (1%-2%), and circular craniotomies were made using a high-speed drill at the indicated coordinates.An electrode assembly that consisted of six tetrodes was stereotaxically implanted above the MGB (3.1 mm posterior and 2.1 mm lateral to the bregma, 3.0 mm deep from the brain surface) and the A1 (2.9 mm posterior and 4.2 mm lateral to the bregma, 0.8 mm deep from the brain surface).The tetrodes were generated from 17-μmwide polyimide-coated platinum-iridium (90/10%) wires (California Fine Wire, Grover Beach, CA), and the electrode tips were plated with platinum to lower the electrode impedances to 200-300 kΩ.The LFP recording using tetrodes is our conventional recording method for freely moving mice and we here used this method to record LFP signals without isolation of single unit spikes.In some mice, an additional stainless steel screw electrode was implanted as a reference electrode on the skull above the prefrontal cortex (PFC) (2.0 mm anterior and 0.5 mm lateral to the bregma).The recording device was secured to the skull using dental cement (Re-fine Bright, Yamahachi Dental Mfg. Co., Aichi, Japan).After all surgical procedures were completed, anesthesia was discontinued, and the animals were allowed to awaken spontaneously.Following surgery, each animal was housed in a transparent Plexiglas cage with free access to water and food for at least 7 days.

| Electrophysiological recording
Mice were connected to the recording equipment via Cereplex M, a digitally programmable amplifier, which was placed close to the animal's head and was connected to the Cerebus recording system (Blackrock Neurotech, Salt Lake City, UT), a data acquisition system, via a lightweight multiwire tether and a commutator.Recordings were made up to 14 days after SE induction and neuronal activity was recorded over 30 min every recording day, with a selected subset of 10 min being utilized for subsequent analysis.To minimize the potential confounding effect of movement-related artifacts being falsely detected as interictal spikes, data from mice exhibiting significant movement were excluded from the analysis.During these recordings, 1 if there was an interictal spike, and 0 if there was no interictal spike.Pearson's correlation coefficients were calculated for 6000 bins.Recordings were made up to 14 days after SE induction, after which the mice were sacrificed to confirm electrode location.

| Confirmation of electrode locations by histological analysis
The mice were given an overdose of isoflurane and then were perfused intracardially with 4% paraformaldehyde (PFA) in PBS (pH 7.4) and decapitated.The brains were left for more than 6 h after decapitation so that electrode tracks were indented into the tissue.After dissection of the brain, each brain was fixed overnight in 4% PFA and then equilibrated with 30% sucrose in PBS overnight.Frozen coronal sections (50 μm) were cut using a microtome and processed for cresyl violet staining.The positions of all electrodes were confirmed by identifying the corresponding electrode tracks in the histological tissue.

| LFP analysis
LFP signals during quiescent states were analyzed.LFP traces were bandpass filtered at 1-100 Hz, and interictal spikes were detected at the local maximum or minimum at which the amplitude of a transient deflection exceeded a threshold of 500 μV (>10 standard deviations from the baseline before seizure induction).Interictal spikes detected within a 500-ms bin were regarded as a single interictal spike.This analysis was conducted using MATLAB.

| Immunohistochemistry
Animals were intracardially perfused with cold PBS followed by 4% PFA in 0.1 M PBS, and the brains were quickly removed, placed in 4% PFA, and postfixed overnight.Continuous sagittal or coronal sections of the brain (40 μm thick) were prepared using a Microtome Cryostat HM520 (Microm International GmbH, Walldorf, Germany).
Sections were permeabilized and blocked with 10% goat serum and 0.3% Triton X-100 in PBS.The sections were then incubated overnight at 4 C under agitation with the following primary antibodies in PBS containing 10% goat serum and 0.3% Triton X-100: rabbit anti-NeuN (1:1000 dilution; ab177487, Abcam, Cambridge, USA), mouse anti-NeuN (1:1000 dilution; MAB377, EMD Millipore, Billerica, MA), mouse anti-GAD67 (1:500 dilution; MAB5406B, EMD Millipore), rabbit anti-Iba1 (1:500 dilution; 019-19741, Wako, Osaka, Japan), guinea  Images were collected at 1024 Â 1024 pixels and stacked and analyzed using ImageJ software (NIH).Z-series images were collected with step sizes of 2.00 μm (for the 10Â objective lens), 1.00 μm (for the 40Â objective lens), or 0.33 μm (for the 100Â objective lens).Cell density and interaction analysis was established by stacking 11 z-series images collected with the 40Â objective lens, and regions of interest (ROIs) for the analysis were similarly established for the images in each dataset.Cellular density and interaction analysis were conducted by stacking 11 z-series images acquired through a 40Â objective lens.ROIs were established accordingly for each dataset to facilitate analysis.Cell density assessment involved manual examination, wherein the position of nuclei was verified using Hoechst staining and the Cell Counter plug-in.For our wrapping analysis, we used unstacked Z-sequence images.We defined the "wrapping" by microglia as instances where both the microglial soma and its processes were in direct adjacency (within a distance of 1 pixel or co-localized) with the neuronal soma (see Figure S3 for a more comprehensive visual representation of this definition).For synapse analysis, 4 Z-series images collected with a 100Â objective lens were used.Quantification of synaptic punctuation around neurons was performed using 7 Z-series images of neurons taken randomly from within the vMGB region using 100Â objective lens with 3Â zoom and by referring to methods from previous studies (Liu et al., 2014).Using ImageJ, the colocalization of immunostained markers was confirmed by confocal three-dimensional reconstruction, and thresholding for quantifying fluorescent images was performed by a researcher who was blinded to the experimental conditions.All images presented are the maximum intensity projections of z stacks of individual optical sections.Representative images of half-brains were obtained by a BZ-X700 phase contrast microscope (Keyence, Osaka, Japan) with 4Â objectives.

| Microglial morphology analysis
Microglial morphology analysis was performed utilizing ImageJ, following methodologies described in previous studies (Green et al., 2022;Young & Morrison, 2018).In brief, the images were initially adjusted for brightness and contrast and then converted to 8-bit.Preprocessing steps included the application of an Unsharp mask and Despeckle filters, followed by adjustment of thresholds to generate binary images.Further processing involved applying the Despeckle and Close filters to the binary images.To encompass microglial cell bodies and processes, ROIs were defined and consistently applied across all cells.The paintbrush tool was utilized to connect branches fragmented during image processing, ensuring the removal of non-microglial fragments, while maintaining fidelity to the original images.Finally, the binarized images were skeletonized and used for subsequent analysis.Manual tracing with reference to Hoechst staining was used to determine cell body sizes.

| Whole-brain clearing and light-sheet microscopy
Transparent imaging of the CX3CR1 GFP/+ mouse brain was performed using the CUBIC-HV method (Susaki et al., 2015).First, for delipidation, brains fixed with 4% PFA were incubated with CUBIC-L (T3740, Tokyo Chemical Industry, Tokyo, Japan) diluted to 50% in water at 37 C overnight.Then, the samples were incubated in 100% CUBIC-L at 37 C for 5 days, and the solution was changed once during the incubation.Delipidated brains were washed three times in PBS containing 0.05% NaN 3 and stored in PBS/0.05%NAN 3 at 4 C until use.
Nuclear staining was performed with propidium iodide (PI) (5 μg/mL; Thermo Fisher, P3566) in CUBIC-HV1 3D nuclear staining buffer (Tokyo Chemical Industry, S5890) at 37 C for 4 days.For immunostaining, brains were incubated in CUBIC-HV1 3D immunostaining buffer (Tokyo Chemical Industry, I1142) at 32 C for 1.5 h and then incubated for 3 weeks at 37 C in a buffer containing rabbit anti-GFP Alexa647 (1:20 dilution; Thermo Fisher, A31852).After washing 3 times with immunostaining wash buffer at 4 C, the samples were postfixed in 1% formalin solution for 24 h.The samples were washed with PBS and finally clarified by 1 day of preliminary replacement with CUBIC-R+(N) (T3983, Tokyo Chemical Industry) diluted 1:1 with water, followed by refractive index adjustment with the original solution for 2-3 days at room temperature.Immunostained brains were imaged using a light-sheet fluorescence microscope (MVX10 (Olympus)-based custom-built Light sheet microscope, sCMOS camera (Neo-5.5-CL3,Andor), and Lasers (OBIS, Coherent) with a step size of Z = 10 μm.

| Data representation and statistical analysis
The data were pooled from at least three independent experiments.
The data were collected and statistically analyzed by a researcher who

| SE induced interictal spikes in the auditory pathway
First, we investigated the changes in neuronal activity in the auditory pathway after KA-induced SE (Lévesque & Avoli, 2013).LFPs were recorded from the A1 and MGB (Figure 1b) and the PFC as a control before and after KA-induced SE (Figure 1c).An interictal spike with amplitudes ranging from 500 to 2500 μV (Figure 1d) was detected in all electrodes placed in both the MGB and A1, but not in the PFC, after SE (Figure .1e).Neuronal activity was recorded in the MGB and A1 on Day 1, and a gradual decrease in frequency was observed in the MGB starting on Day 3, but activity in the A1 was maintained until Day 14 (Figure 1e,f).In the MGB and A1, which have reciprocal projections, the timing of the interictal spike was highly correlated and maintained up to Day 14 (Figure 1g-i).These results suggest that neuronal activity in the auditory pathway is increased and maintained after SE in mice.

| Reduction of inhibitory synapses around neuronal soma after SE is mitigated by microglial removal
To examine microglial changes after SE, we performed whole-brain tissue clearing and light-sheet microscopy of CX3CR1 GFP/+ mice, in which microglia are labeled with GFP.We found that SE induced an increase in GFP fluorescence intensity, which is indicative of clustering microglia, in the MGB as well as in the hippocampus, which has a high susceptibility to KA-induced SE (Movies S1 and S2).
Consistent with light-sheet microscopy results, fluorescence confocal microscopy of immunostained brain sections revealed a clustering of Iba1 + microglia with relatively large soma and loss of NeuN + neurons in the dorsal MGB (dMGB) after SE (Figure 2a-f).We also found that the neuronal loss preceded microglial clustering (Figure S1A-F).Therefore, to examine the involvement of microglia in the histological changes in the MGB after SE, we first established a strategy to deplete MGB microglia by local and bilateral injection of clodronate (50 mg/mL, 300 nL), which is known to effectively deplete mouse macrophages and microglia (Torres et al., 2016), into the MGB (Figure S2A-C).There was no discernible effect of clodronate treatment on either the dosage of KA or SE survival, as detailed in Section 2. The results showed that while clodronate significantly reduced microglial density, it did not inhibit SE-induced neuronal loss (Figures 2d-f and S2D-G).In contrast to the dMGB, SE did not induce neuronal loss in the vMGB (Figure 2g,h).In the vMGB, although no microglial clustering was observed, the density of microglia was increased and was suppressed by clodronate administration (Figure 2i).
Because of the loss of neurons in dMGB, the neural activity observed in electrophysiology experiments may reflect vMGB.
Because neural activity alters the interaction between microglia and neurons, we examined these contacts.We found that the percentage of microglia that wrapping the neuronal soma was increased after SE (Figures 3a-c and S3).In addition, the percentage of neuronal soma wrapped by microglia also tended to be increased, although not significantly (Figure 3d).These microglial responses were probably caused F I G U R E 2 Microglial clustering in vMGB after status epilepticus.(a) Experimental paradigms for the control, SE, and clodronate (Clo)-SE groups.Clodronate (50 mg/mL, 300 nL) was administered locally to bilateral MGBs using a microsyringe (see also Figure S2).by KA-induced SE rather than directly by KA because these changes were suppressed when SE was blocked by DZP (Figure S4A-D).
Microglia wrapping the neuronal soma has been previously reported to be associated with microglial stripping of inhibitory synapses (Chen et al., 2014;Trapp et al., 2007).Since inhibitory synaptic boutons are abundant in the vMGB (parvalbumin+, VGAT+, and GAD67+, Figure S5A-H), we next examined inhibitory synaptic inputs to the neuronal soma and found that SE decreased the percentage circumference of neuronal soma occupied by GAD67 (Figure 3e).Surprisingly, removal of microglia by clodronate restored the percentage of the neuronal soma periphery occupied by GAD67, suggesting that microglia are involved in the reduction of these inhibitory synapses.In addition, the total area of GAD67 signals in the vMGB was not affected in the SE group, indicating that the inhibitory projections to the vMGB itself were not affected by the experimental conditions (Figure 3f).Reduced inhibitory synaptic inputs to vMGB neurons may affect the regulation of neuronal activity.To evaluate our hypothesis, we assessed the expression of the immediate-early gene c-Fos, an indicator of increased neuronal activity.As hypothesized, we observed an increase in c-Fos expression in vMGB neurons after SE (Figure 3g,h), which is consistent with our electrophysiological observations (Figure 1).In contrast to inhibitory synaptic boutons, the overall distribution of excitatory synapses was not affected (Figure S6A-F), and vGlut1 + excitatory presynaptic boutons were not decreased (Figure S6G-I).Together, these results suggest that microglia may reduce inhibitory synaptic inputs to the relay neurons in the vMGB, thereby increasing the activity of these neurons.

| Auditory dysfunction in mice after SE was suppressed by microglial depletion
We next examined the possible role of MGB microglia in behavioral auditory dysfunction after SE.Auditory discrimination deficits, that is, the difficulty in discriminating different sounds (Johnsrude et al., 2000), occur in mTLE patients (Han et al., 2011;Meneguello et al., 2006).We performed a sound discrimination avoidance task, which has been used in mice as a test of discrimination between different sounds (Kolodziej et al., 2016;Kurt & Ehret, 2010).In this task, mice had to discriminate between different patterns of the Go cue sound (ascending; 4-8 kHz) and the NoGo cue sound (descending; 8-4 kHz) in order to avoid electric foot shocks (Figure 4a).The results were classified into four categories: hit, miss, correct rejection, and false, depending on the type of trial and the response of the mice (Figure 4a).Mice were subjected to 30 Go and 30 NoGo trials per day, and the hit rate (hit/[hit + miss]) and the false rate (false/[false + correct rejection]) were analyzed and compared (Figure 4b).After 10 days of training, the mice gradually discriminated the two different sounds, resulting in the hit rate being higher than the false rate (Figure 4c).In the test session following the training, the hit rate was significantly higher in control mice (Figure 4c), whereas there were no significant differences between the hit and false rates in the test session when SE was induced prior to the test session (Figure 4d).These changes seen in the mice after SE could be due to the mice no longer being able to hear these sound frequencies, or due to a change in motivation for the trial.Therefore, a similar test was then conducted using a Go cue (8 kHz) and a NoGo cue (no sound, only a light indicating the start of the trial).We confirmed that mice after SE avoided electric shocks, providing corroborating evidence that their auditory acuity remained intact (Figure S7).Together, these data Three days later, we intraperitoneally injected PBS as a control or KA to induce SE in mice.We found that the clodronate treatment to mice decreased the hit rate in the test session, but auditory discrimination was still possible (Figure 4f).We also found that clodronate attenuated the SE-induced impairment of auditory discrimination (Days 24-26, Figure 4g).Thus, these results suggest that MGB microglia contribute to auditory discrimination deficits after SE.

| Long-term persistence of synapse reduction and impaired auditory behavior
It has been suggested that synapse stripping on neuronal soma by microglia, unlike synaptic phagocytosis, is reversible (Chen et al., 2014;Trapp et al., 2007) and does not involve the loss of presynaptic structure.Therefore, we examined whether synaptic reduction at the soma and auditory dysfunction after SE persist over time (Figure 5a-d).Eight weeks after SE induction, similar to 6 days after SE (Figure 2), we observed an increased density of microglia (Figure 5e) with no accompanying change in neuronal density (Figure 5f), increased wrapping of neuronal soma (Figure 5g) and reduction of inhibitory synapses (Figure 5h) in the vMGB, and clustering of microglia (Figure 5i) and a loss of neurons (Figure 5j) in the dMGB.Changes in microglial morphology were also observed up to 8 weeks post-SE, with microglia having fewer projections and larger cell bodies (Figure S8).
Finally, we tested auditory discrimination 8 weeks after SE in mice that had undergone the behavioral test 6 days after SE (Figure 6a).We started the test at Day 67 (8 weeks after SE induction) and found that control mice were able to discriminate sound differences with a significant difference in hit rate and false rate as early as 3 days after starting the test (Day 70, Figure 6b), while mice that experienced SE did not during the 7-day test session (Figure 6c).In summary, synaptic reduction and impaired auditory discrimination were maintained as long as 8 weeks after SE, which highlights the long-lasting effects by microglia in the auditory pathway.Patients with TLE often suffer from auditory abnormalities, such as auditory hallucinations and impaired auditory discrimination, which are associated with seizures (Han et al., 2011;Hauf et al., 2013;Meneguello et al., 2006) and human brain imaging data suggest that these patients exhibit functional changes in auditory regions of the brain (Hauf et al., 2013).Because auditory abnormalities significantly reduce the quality of life of patients with TLE, the resolution of these deficits is an important issue, but the cellular mechanisms linking seizures and auditory abnormalities remain unclear.Here, we used a KA-induced SE mouse model in which auditory pathway hyperactivity was confirmed to highlight the role of thalamic microglia in auditory dysfunction.

| Histological changes in the MGB after SE
Our results suggest that the auditory pathway in mice is vulnerable to SE in terms of neuronal loss and microglial changes, but the reasons for this remain unexplained.While there was considerable variability in hippocampal microglial accumulation among the mice, microglial accumulation in the MGB was observed in all the mice.In this study, we found that DZP suppressed both KA-induced seizures and microglial activity as well as neuronal changes in the MGB (Figure S4).However, in contrast to the MGB, microglial accumulation was observed in the hippocampus, suggesting the possibility of a different mechanism.Previous studies have reported that the neurological damage observed in areas other than the hippocampus occurs in animals with severe seizures during SE (Sperk et al., 1985).Therefore, it is likely that the neuronal loss in the MGB was induced by the propagation of neuronal hyperactivity rather than by the toxicity of KA itself.Among the KA receptors, GluK2, a subunit of the KA receptor, has been shown to play a key role in KA-induced neuronal death (Mulle et al., 1998).However, it has been reported that GluK2 is barely expressed in the MGB (Wisden & Seeburg, 1993).Furthermore, a previous study showed that SE induced neurodegeneration in the MGB after pilocarpine treatment in mice (Wang et al., 2008).Taken together, these findings suggest that the high vulnerability of the MGB is caused by overactivity during seizures, rather than a direct effect of KA. dMGB resulted in neuronal loss after SE, whereas vMGB did not show such an effect (Figure 2e,h).These differences may be explained by differences in cellular characteristics or projection in each subregion.It has been reported that the calcium-binding proteins calbindin and Calretinin are expressed in dMGB neurons (Lu et al., 2009).The expression of such calcium-binding proteins has been proposed to be associated with vulnerability in neurodegenerative diseases (Fairless et al., 2019), which may explain the damage in dMGB.On the other hand, vMGB had more projections from PV and GAD67-positive cells than other subregions (Figure S5A).Such inhibitory input may contribute to survival by suppressing neuronal hyperexcitability during SE.Differences in degeneration between dMGB and vMGB in disease models have received little attention.Therefore, further detailed investigation may provide new insights into auditory symptoms in various neurodegenerative diseases.

| Microglial function in the MGB after SE
Microglia play a critical role in maintaining brain homeostasis by extending their projections into the surrounding areas and monitoring the environment.During epilepsy, they modify their interactions with neurons in response to increased neuronal activity and contribute to inflammatory responses and the clearance of dead cells following neuronal death (Hiragi et al., 2018).The microglial response to neuronal death in the acute phase of epilepsy model has primarily been investigated in the hippocampus, where regions such as the CA3 and CA1 are susceptible to hyperexcitation induced by pilocarpine and KA (Araki et al., 2020;Borges et al., 2003;Wyatt-Johnson et al., 2017).
Neuronal death typically occurs within hours after SE, followed by microglial accumulation starting approximately 3 days after SE, which has been observed to persist for 2 to 4 weeks.In our study, we observed neuronal death 1 day after SE and subsequent microglial accumulation 3 days after SE in the dMGB (Figure S1), demonstrating a similar temporal pattern as reported previously.Interestingly, changes in microglial morphology were also observed in the vMGB, where neuronal death did not occur.This may indicate a spread of the inflammatory response from the dMGB or increased neuronal activity.
Compared to the acute phase, microglial changes in the chronic phase, which occur several months after SE, remain unclear.Our results showed that microglial changes persisted even 8 weeks after SE (Figures 5 and S8).Notably, Benson et al. reported that the expression of inflammation-related mRNAs in microglia was suppressed at 3 weeks after SE, but gradually recovered between 5 and 12 months later (Benson et al., 2015).Studies using explanted human tissue from patients with epilepsy have also reported a tendency for microglia to adopt an amoeboid morphology in areas of significant neuronal loss (Morin-Brureau et al., 2018).Furthermore, seizures have been shown to promote the release of inflammatory factors by microglia.These long-term effects may reflect tissue sclerosis following neuronal death and increased neuronal activity.
Our finding that microglia reduced inhibitory synapses on the soma in the vMGB suggests that microglia may also regulate brain function in various brain regions other than the hippocampus during neuronal hyperactivity (Abiega et al., 2016;Araki et al., 2020;Avignone et al., 2008;Eyo et al., 2014;Luo et al., 2016;Wyatt-Johnson et al., 2017).In this study, we observed an increase in microglial wrapping of the neuronal soma and a decrease in inhibitory synapses, suggesting a link to synaptic stripping.Synaptic stripping is the physical disruption of synaptic transmission by the insertion of microglial processes between the presynaptic and postsynaptic regions (Trapp et al., 2007).Previous studies have also suggested that microglia strip or phagocytose inhibitory synapses during conditions of health and disease, including during epilepsy (Carrillo et al., 2020;Chen et al., 2014;Hashimoto et al., 2023;Wan et al., 2020;Wittner & Magl oczky, 2017).
Although the molecular mechanisms underlying synaptic stripping remain largely unclear, a recent study has implicated the involvement of neuronal ATP-microglial P2Y12R signaling.Following febrile seizures in developing mice, microglia were reported to displace GABAergic presynapses projecting to cortical neurons in a P2Y12Rdependent manner (Wan et al., 2020).Contacts between microglia and neurons after SE have also been shown to be mediated by CX3CL1-CX3CR1, and IL1-β signaling between neurons and microglia, as well as reduced extracellular Ca 2+ levels (Eyo et al., 2014(Eyo et al., , 2015(Eyo et al., , 2016)).Therefore, it would be interesting to determine whether these molecular signaling pathways are involved in the reduction of inhibitory synapses (Figure 3e) and the auditory dysfunction (Figure 4c,d) observed in this study.
Although it has been suggested that microglial contact with neuronal soma after inflammation is reversible (Chen et al., 2014), in our results, microglial morphological changes and synaptic reduction persisted for 8 weeks after SE induction (Figure 5h).This may be due to long-term inflammation and increased neural activity in the brain, unlike inflammation caused by LPS and other factors.However, spontaneous convulsive seizures are rare in the mouse model of systemic administration of KA, and in this study, we did not observe recurrent seizures in mice used for long-term experiments.
Various mechanisms have been reported for the regulation of synapses by microglia (Andoh & Koyama, 2021).In our study, the total area of GAD67 in MGB was unchanged (Figure 3f), and we found no evidence of regulation by phagocytosis, such as the presence of inhibitory synapses inside microglia.However, because microglia were depleted from the MGB, we cannot rule out the possibility that microglia are involved in auditory dysfunction by mechanisms other than the wrapping of neuronal somata and the reduction of inhibitory synapses.

| Relationship between auditory dysfunction and anatomical changes in the MGB
Although several studies have documented depression-like behavior and hyperactivity in rodent models of KA administration (Tchekalarova et al., 2011;Zeidler et al., 2018), our study did not include assessments of motor behavior or motivation.Notably, in the discrimination avoidance task, a less pronounced reduction in the hit rate was observed in SE and Clo-SE mice on the first day of the testing session i (Figure 4c-f).In addition, the same mice exhibited a higher false rate, denoting inadvertent movements during the NoGo trials.If these differences are due to increased motor activity, one would expect a concomitant increase in the hit rate along with the false rate.However, the hit rate remained lower than that observed at the end of the training session.In addition, SE mice performed similarly to controls in a simple Go/NoGO task (Figure S7), suggesting that these differences are not due to increased motor behavior.In addition, the clodronate + PBS-treated group tended to have a decreased hit rate during the test period compared to the PBS-only group (Figure 4c,f).Although the training and test periods differ between the two groups, it is worth noting that mice previously trained in this task were able to discriminate immediately, even after an 8-week break.
This suggests that the effect of the different pre-test intervals may be minimal.However, given that the removal of microglia from the developing medial nucleus of the trapezoid body has been shown to affect innervation and auditory thresholds (Milinkeviciute et al., 2019), we cannot rule out the possibility that the decrease in the hit rate is due to either the removal of microglia or the effects of clodronate itself.This hypothesis certainly warrants further investigation in future studies.
We have observed distinct anatomical changes in the MGB in mice after SE, including an extensive loss of neurons in the dMGB and decreased inhibitory synaptic input in the vMGB.However, it remains unknown how these anatomical changes lead to auditory dysfunction.
Removal of MGB microglia by clodronate treatment prevented the loss of auditory discrimination ability after SE (Figure 4g).Regarding the regional specificity of microglial removal, it is possible that the needle insertion during administration could affect microglia in the adjacent areas.While the hippocampus may have been affected in this study, we believe that the effect on other auditory regions, such as A1, is minimal.In the vMGB, removal of microglia also prevented a decrease in inhibitory synaptic inputs and an increase in c-Fos staining (Figure 3e,h).Our results suggest that the reduction of inhibitory synapses in vMGB relay neurons may contribute to the loss of auditory discrimination ability.However, a limitation of our study is that we did not directly confirm the effect of clodronate on neuronal activity.
Instead, we have only established a correlation between neural activity, as assessed by c-Fos, and behavioral tests.The vMGB is the primary site of auditory projections to the A1 (Anderson & Linden, 2011;Felix et al., 2018) and receives both feedforward and feedback inhibitory inputs from the IC and TRN, respectively (Peruzzi et al., 1997;Yu et al., 2009) (Figure 1a).In particular, inhibitory inputs from the TRN play a critical role in the frequency selective neuronal tuning of auditory information in the vMGB (Cotillon-Williams et al., 2008).The reduction of inhibitory synapses could have disrupted this tuning and attenuated specific responses to sound frequency information, resulting in impaired auditory discrimination.
In contrast, removal of microglia was not sufficient to prevent neuronal loss in dMGB (Figure 2e).This result suggests that the contribution of dMGB neurons to sound discrimination is limited.This is a surprising result given the importance of dMGB in auditory frequency discrimination tasks (Chen et al., 2019).MGB subregions differ in their projection patterns and roles in auditory information processing (Bartlett, 2013;Lee, 2015); the dMGB does not have projections to A1 but instead has connections to the non-primary auditory cortex and the amygdala.Although the details of the function of the dMGB are still unclear, it has been proposed to be involved in the emotional processing of auditory information and the integration of information with information provided by other senses, such as vision (Bartlett, 2013;Lee, 2015).
The interictal spike frequency in the MGB showed a decrease 6 days after SE (Figure 1f); however, the diminished performance in the auditory discrimination task persisted for 8 weeks thereafter (Figure 6).Notably, the presence of a significant difference in c-Fos expression on Day 6 (Figure 3h) suggested a potential increase in MGB activity, despite the reduction in interictal spike.However, one of the inherent limitations of our study is that the interpretations are derived from the correlation of different experiments, and a more cohesive approach would be beneficial in future studies.It is important to recognize that the LFP serves as a signal reflecting the electrical activity of multiple neurons, with a particular focus in this study on the large-amplitude waveforms observed during the interictal spike.
Thus, a more comprehensive understanding of neuronal circuit changes can be achieved by examining individual neuron firing rates and synaptic transmission through simultaneous in vivo multicellular recordings and in vitro slice patch clamp recordings.
Considering the sustained high activity correlation between the MGB and A1 for up to 14 days after SE (Figure 1g), it is plausible that the MGB and A1 mutually influence auditory discrimination behavior.
From the present results, it remains a challenge to determine whether the observed increase in the A1 cortex is a consequence or a cause of the observed changes in the vMGB.The intricate reciprocal projections between A1 and MGB, along with their complex regulatory mechanisms, preclude us from drawing definitive conclusions based solely on the results of this study.In particular, dMGB serves as the primary pathway for conveying projections to A1. Research on tinnitus has suggested that reduced TRN-derived inhibition may lead to increased neuronal activity in the MGB, followed by increased activity in the A1, potentially contributing to the onset of tinnitus (Leaver et al., 2011;Rauschecker et al., 2010).Furthermore, auditory cortical L6 CT neurons have direct excitatory projections to the MGB, and can modulate sound responsiveness (Guo et al., 2017).In addition, the administration of bicuculline, a selective GABA A receptor antagonist, to A1 has been reported to enhance neuronal activity in A1 and induce c-Fos expression in the MGB (Guo et al., 2007).Therefore, further investigation is warranted to explore the involvement of A1 in auditory dysfunction following SE in future studies.
Thus, although the molecular mechanisms underlying seizure-induced synaptic stripping and the role of microglia in auditory dysfunction beyond their synaptic stripping function need to be further investigated, this study reveals a previously unknown role for microglia in auditory dysfunction in the brain after seizures.
significant movement was determined based on electromyography (EMG) signals.The signals were band-pass filtered between 20 and 200 Hz, and the root mean square (RMS) values of the filtered signals were computed using a 1-s bin size.Due to the variability in baseline movement levels among the mice, establishing a universal EMG threshold for movement definition wasn't feasible.Instead, we manually tailored the EMG thresholds for each mouse, relying on a visual evaluation of their EMG RMS traces.For a majority of the subjects, movement periods were defined when the EMG signals surpassed roughly 5 standard deviations (SDs) above the average of baseline EMG RMS traces.During this process, mouse position traces were employed as a reference but were not subjected to further analysis.Electrical signals were sampled at 2 kHz and low-pass filtered at 500 Hz.The correlations between A1 and MGB were calculated for each interictal spike that exceeded the threshold value.A matrix was created for each electrode with a time bin of 100 ms: pig anti-Iba1 (1:300 dilution; 234,004 and 234,308, Synaptic Systems, Göttingen, Germany), guinea pig anti-c-Fos (1:1000 dilution; 226,308, Synaptic Systems), rabbit anti-S100β (1:1000 dilution; ab52642, Abcam), mouse anti-calretinin (1:1000 dilution; MAB1568, EDM Millipore), guinea pig anti-parvalbumin (1:500 dilution; 195,004, Synaptic Systems), guinea pig anti-VGAT (1:500 dilution; 131,004, Synaptic Systems), guinea pig anti-vGluT1 (1:500 dilution; 135,304, Synaptic Systems), and rabbit anti-Homer1 (1:500 dilution; 160,002, Synaptic Systems).After extensive washing in PBS, the sections were incubated with appropriate secondary antibodies conjugated with Alexa Fluor dyes (488, 594, 647; 1:500; Thermo Fisher, Waltham, MA) and Hoechst 33342 dye (1:1000; Life Technologies, Carlsbad, CA) for 4 h at 4 C.After three rinses, the samples were enclosed with Mountant Permafluor (Thermo Fisher Scientific).
was blinded to the experimental conditions.The analyses used included Tukey's test after one-way or two-way ANOVA, Bonferroni's multiple comparison test after two-way RM-ANOVA, and an unpaired two-tailed t test with 95% confidence.All p values and sample sizes are indicated in the figure legends.No statistical methods were used to predetermine sample sizes, but our sample sizes were chosen based on those generally employed in the field, and the normal distribution of data was statistically confirmed before performing parametric analysis.All statistical analyses were performed using Prism 9.4.0 software (GraphPad, San Diego, CA).

F
I G U R E 4 Depletion of MGB microglia attenuates auditory dysfunction after status epilepticus.(a) Schematic diagrams of the auditory discrimination avoidance task.During the daily task, a Go cue (65-75 dB, ascending frequency of 4-8 kHz) and a NoGo cue (65-75 dB, descending frequency of 8 to 4 kHz) were randomly presented 30 times each.If mice made an incorrect action during the 6-s period of cue presentation, an electric foot shock of 150 μA was delivered to the scaffold grating.(b) Experimental paradigm of the auditory discrimination avoidance task for the control and SE groups.(c, d) The hit rate in Go trials and the false rate in NoGo trials in the control (c) and SE (d) groups.The hit rate was significantly higher than the false rate during the test session in the control, while there was no significant difference (NS) in the SE group.n = 5 mice.(e) Experimental time course of the auditory discrimination avoidance task including the depletion of microglia with clodronate.The experimental groups included the clodronate alone group (Clo) and clodronate + SE group (Clo-SE).(f, g) The hit rate in Go trials and the false rate in NoGo trials in the Clo (f) and Clo-SE (g) groups.The hit rate was significantly higher than the false rate during the test session in both the Clo and Clo-SE groups from Days 24 to 26. n = 4 mice.Significant differences were determined by Bonferroni's multiple comparison test after two-way RM-ANOVA (*p < 0.05; **p < 0.01, ***p < 0.001; and ****p < 0.0001).(c, d) F (51, 272) = 2.109, p < 0.0001; (f, g) F (57, 228) = 1.888 p = 0.0006.The data are shown as the mean ± SEM. indicate that the impairment of auditory discrimination ability occur in the KA-induced SE mouse model.Next, to assess the involvement of MGB microglia in the impairment of auditory discrimination, we injected clodronate into the bilateral MGBs after the training period (Figure 4e).

F
I G U R E 6 Maintained auditory dysfunction 8 weeks after SE.(a) Schematic diagrams of the auditory discrimination avoidance task.The task was performed 8 weeks later in mice that had learned to discriminate sounds in Figure 4b.(b, c) The hit rate in Go trials and the false rate in NoGo trials in the auditory discrimination task (the experimental paradigm is the same as that in Figure 4) in the control (b, n = 5 mice) and SE (c, n = 5 mice) groups 8 weeks after SE.The hit rate was significantly higher than the false rate in the control group but not in the SE group.Significant differences were determined by Bonferroni's multiple comparison test after two-way RM-ANOVA (**p < 0.01; ***p < 0.001; and ****p < 0.0001).F (21, 112) = 0.9752, p = 0.4988.The data are shown as the mean ± SEM.