Microglia depletion improves hippocampal circuit function after mild traumatic brain injury in male mice

Abstract

Traumatic brain injury (TBI) affects over 69 million people every year, and mild traumatic brain injury (mTBI) accounts for 70–90 % of cases. TBI has two components: i) primary injury – direct damage to the tissue from the mechanical impact and ii) secondary injury – additional or extended damage to the tissue from the ensuing biochemical and physiological processes such as neuroinflammation. Neuroinflammation triggered in part by activated microglia, determines whether the post-injury outcome is recovery or long-term neurodegeneration. Microglia, key components of the neuroinflammatory process, release cytokines such as TNF-α, which affect neuronal activity. Our study investigated the effects of acute microglia depletion on hippocampal neurophysiology in male mice (7–10 days after mTBI), a time window that allows us to target sub-acute microglial responses post-injury. An additional objective of the study was to determine if the pro-inflammatory cytokine TNF-α contributed to the injury-induced network excitability shifts in the hippocampal circuitry. We demonstrate that depleting microglia with PLX-3397 treatment for 7–10 days after mTBI restores network excitability in hippocampal area CA1 and the dentate gyrus (DG). Furthermore, treatments with thalidomide and etanercept show that TNF-α plays a role in altering the network excitability after mTBI. These findings provide new insights into the physiological changes after injury and highlight potential targets for future interventions to specifically address the detrimental effects of chronic inflammation.

1. Introduction

Traumatic brain injury (TBI) results from a sudden force or blow to the head. Over the last three decades, there has been a dramatic increase in the incidence of TBI globally. It is estimated that 69 million people worldwide sustain at least one TBI annually; the majority (70–90 %) of whom suffer a mild TBI (mTBI) (Dewan et al., 2018; Centers for Disease Control and Prevention, 2014, 2015). Current treatments for mTBI are mostly palliative and do not correct the abnormal changes underlying the observed cognitive and physiological deficits. Consequently, there is an urgent need for treatments that restore the underlying physiological alterations and reduce the growing socio-economic burden.

The lack of effective mechanism-based treatments for mTBI is partly due to the complexity of the injury process. Broadly, injury has two components – primary and secondary injury (Smith, 2013; Giza and Hovda, 2014). Primary injury is the immediate damage to cells and the tissue structure from the impact. Secondary injury is the additional or extended damage to cells and the tissue from the physiological changes that occur as a result of the impact. Secondary injury can occur from metabolic and oxidative stress (Giza and Hovda, 2014) and/or persistent inflammation (Risbrough et al., 2022; Bergold, 2016).

1.1. Neuroinflammatory response post TBI

Neuroinflammation is the immune response of the central nervous system (CNS). One hallmark of TBI is striking reactive changes in microglia, resident immune cells of the brain (Loane and Kumar, 2016; Donat et al., 2017; Shao et al., 2022; Wofford et al., 2019). Upon binding to damaged-associated molecular patterns (DAMPS), which are released from cells suffering a primary injury, microglia are activated and release pro-inflammatory chemokines and cytokines that orchestrate the neuroinflammatory response (Loane and Kumar, 2016; Donat et al., 2017). Pro-inflammatory cytokines begin to rise within minutes to hours after the injury (Bourgognon and Cavanagh, 2020). This is typically followed by an increase in anti-inflammatory molecules which contribute to restorative mechanisms in later stages (Loane and Kumar, 2016; Jassam et al., 2017). While acute neuroinflammation is necessary for recovery, the shift from release of pro-to-anti-inflammatory molecules does not always occur after mTBI, leading to persistent inflammation and further damage (Schimmel et al., 2017).

Existing work on mTBI underscores the importance of understanding physiological changes in the neuronal network that create the cognitive deficits in mTBI patients. We have previously reported injury-induced regional shifts in the network excitability in hippocampal circuitry that contribute to learning and memory deficits after mTBI. To elaborate, we found increased excitability in the dentate gyrus (DG) together with decreased excitability in area CA1 (Folweiler et al., 2018, 2020; Johnson et al., 2014; Paterno et al., 2016; Smith et al., 2015; Wolf et al., 2017). Inflammatory markers can modulate circuit function and mTBI-associated alterations via synaptic and ion channel interactions (Beattie et al., 2002; Stellwagen et al., 2005; Stellwagen and Malenka, 2006; Tyagi et al., 2024). For example, TNF-α has been shown to directly affect the insertion of AMPA and GABA receptors in the post-synaptic membrane, and is crucial for excitatory synaptic signaling (Santello and Volterra, 2012), plasticity (Singh et al., 2022), as well as the hippocampal E/I balance (Wang et al., 2018). Taken together, these findings suggest that dysregulated cytokine signaling after injury may worsen post-injury impairments.

1.2. Microglia depletion post-injury

CSF-1R inhibitors (e.g., PLX-3397, PLX-5622) have been used as tools to temporarily deplete microglia and assess their effects on TBI outcomes. A few studies have demonstrated that microglia depletion post-injury contributes to decreased pro-inflammatory cytokines and chemokine gene expression, reduced brain tissue loss, decreased astrocyte reactivity, as well as increased neuronal recovery and connectivity (Boland and Kokiko-Cochran, 2024; Witcher et al., 2021). Specifically, depletion for 5 days after controlled-cortical impact (CCI) injury using PLX-3397 resulted in decreased brain tissue loss and increased neuronal recovery (Wang et al., 2022). Treatment with PLX-3397 before injury shows decreased neuronal apoptosis, increased neurite growth and preserved dendritic spines (Wang et al., 2019, 2020). Treatment with PLX-5622 for one week, 28 days post-CCI improves cognitive recovery and neuronal connectivity up to 3 months post-injury (Henry et al., 2020; Witcher et al., 2021). Interestingly, a study reported that treatment with PLX-5622 for 21 days before and continued up to 19 weeks after CCI do not alleviate symptoms (Willis, 2020), suggesting that PLX treatment may be beneficial only transiently after injury. Taken together, depletion via CSF-1R inhibition provides insights into microglia function after mTBI. However, careful consideration of the limitations of PLX and injury timeline is crucial in designing studies aimed at understanding and modulating microglial functions post-injury. In particular, all CSF-1R inhibitors exhibit off-target effects on peripheral immune cells, complicating the interpretation of microglia-specific roles. While PLX-5622 offers greater specificity for CSF-1R, PLX-3397 holds translational relevance since it is FDA-approved for human use—making it a promising, albeit less selective, tool for bridging preclinical and clinical research.

1.3. Significance and innovation

Targeting microglia is a promising therapeutic approach for neurodegenerative diseases (Asai et al., 2015; Dagher et al., 2015; Yang et al., 2018); however, the translational application of microglia depletion after TBI requires a better understanding of its impact on neuronal network function. Additionally, given that TBI and inflammation are both complex, and microglia are not the only contributors to secondary processes, depletion alone is unlikely to repair every observed dysfunction. Here, we designed experiments to explore a mechanism through which microglia depletion leads to improved cognitive function and neuronal recovery. To our knowledge, this is the first study to investigate microglia depletion, TNF-α release, and network function after mTBI electrophysiologically. Our findings provide new insights into microglial modulation of neural circuitry after mTBI.

2. Materials and methods

2.1. Animals

Experiments were performed on 7–10-week-old, male C57BL/6J mice (The Jackson Laboratory, Bar harbor, ME, USA). All animals were housed in groups of 2 – 5 in the animal facility with standard day-night cycle (lights off: 18:00 h – 6:00 h; lights on: 6:00 h – 18:00 h) and free access to food and water. 208 mice were investigated and randomly assigned to one of the following experimental groups: TBI, Sham, TBI depletion, and Sham depletion. Experimenters were blinded to group assignments. All procedures were conducted in accordance with guidelines and regulations of the Children’s Hospital of Philadelphia (CHOP) Institutional Animal Care and Use Committee (IACUC). To obtain data at similar time points, behavioral, electrophysiological, immunohistochemistry, and Luminex assay experiments were performed on separate cohorts of animals.

2.2. Lateral Fluid Percussion Injury (lFPI)

lFPI is a well-validated model of mild to moderate TBI; it closely replicates the pathology and symptoms observed in humans after concussion (Carbonell and Grady, 1999; Thompson et al., 2005; Xiong et al., 2013; Lyeth, 2016). Before the injury, animals were randomly assigned to two groups: sham (surgery only) and lFPI (surgery + injury). To allow mice to recover from the craniectomy, lFPI was performed over two days: on day 1 a craniectomy was performed and on day 2 the injury (Smith et al., 2015; Witgen at al., 2005; Farrugia et al., 2023).

On day 1, mice were anesthetized (100 mg/kg of ketamine and 10 mg/kg of xylazine) and their right parietal bone between lambda and bregma was carefully removed. This was done without breaching the dura. A 3 mm diameter saline-filled Luer-loc needle was then glued to the skull over the craniectomy window. On day 2, animals were anesthetized with isoflurane and connected to the lFPI device (Department of Biomedical Engineering, Virginia commonwealth University, Richmond, VA). A 10–15 ms fluid pulse (1.4–1.8 atm) was delivered to cause mTBI. Following the injury the needle hub was removed, the skin over the scalp sutured, and the animals allowed to recover on a heating pad in a supine position. Righting times (time it took the mice to turn themselves into the prone position) were noted to determine injury severity (Grin’kina et al., 2016; Morehead et al., 1994). Sham animals underwent the same procedures except for the impact with the fluid pulse. Animals with a righting time greater than 6 min (Hylin et al., 2013; Shultz et al., 2011; Eakin and Miller, 2011) and/or breach of the dura (Witgen et al., 2005) were excluded from the study. Animals were allowed to recover from the procedures for 7 days after sham/injury.

2.3. Microglia depletion

After injury, mice assigned to the depleted group– were placed on Plexxikon (PLX) 3397 obtained from Selleckchem. PLX-3397 was incorporated into AIN-76A standard chow (Research Diets, Inc.) at 600 mg/kg. Mice in this group remained on this diet until euthanasia (see Fig. 1A for timeline).

Fig. 1.

Microglial density increases after injury while both injury and depletion lead to MAC2 expression by hippocampal macrophages. (A) Schematic of experimental timeline: 7–10-weeks-old C57BL/6J mice were subjected to sham surgery or TBI followed by microglia treatment for 7–10 days. The mice were then used for behavioral, electrophysiological, or immunohistochemistry staining experiments. (B) Representative images of IBA1⁺ and MAC2⁺ cells in the ipsilateral (to injury) area CA1 of the hippocampus in sham, injured, sham-depleted, and injured-depleted mice. Scale bar: 100 μm. (C-F) Image enlargements from areas surrounding white arrows in Fig. B. Scale bar: 40 μm (G-H) Quantification of cells showing a significant interaction effect between groups in number of IBA1⁺ cells per field; two-way ANOVA: F(1, 9) = 9.432, p = 0.0133 and MAC2⁺ cells; two-way ANOVA: F(1, 9) = 30.59, p = 0.0004. (H) Quantification of percentage of MAC2⁺/IBA1⁺ cells revealed only a main effect of depletion; two-way ANOVA: F(1, 9) = 1051, p < 0.0001. (J) Mice transplanted with GFP⁺ monocytes were used as positive controls for MAC2 staining. Scale bar: 100 μm. (K) Image enlargement of area surround white arrow in Fig. J. Scale bar: 40 μm. Data shown as mean ± SEM; Sham: N = 3; Injured: N = 4; Injured Depletion: N = 3; Sham Depletion: N = 3. Asterisks for multiple comparisons indicate *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

2.4. Immunohistochemistry (IHC)

2.4.1. Staining protocols

Mice were perfused through the left ventricle of the heart using normal 9 % saline followed by 4 % paraformaldehyde (PFA). Extracted brains were incubated at room temperature in PFA for an additional 90 mins and kept in 1 % phosphate buffer saline (PBS) overnight at 4 °C. The next day, brains were sliced into 50 μm sections on a vibratome (VT1000S, Leica Biosystems, Buffalo Grove, IL, USA). Slices were stored at 4 °C in 24-well plates. For IHC, slices were transferred to a 12-well plate and incubated in 50 μl of 5 % normal goat serum (NGS), 100 μl of 1 % BSA and 100 μl of 10x PBS-T (0.3 % triton) at room temperature (RT) for 60 min. Slices were washed twice with 1 % PBS and incubated with rabbit anti-IBA1 (1:500; Thermo Fisher Scientific Abcam; MA5 36257), and anti-Mouse/Human Mac-2 (1:500; Cedarlane; CL8942AP), in 1 % PBS at RT for one hour, then at 4 °C overnight. The next day, slices were washed four times with 1 % PBS and incubated for 75 min at RT with DAPI (1:500), Alexa Fluor 594-conjugated Goat Anti-Rat IgG (1:200; Jackson ImmunoResearch), and Alexa Fluor 588-conjugated Goat Anti-Rabbit IgG (1:200; Jackson ImmunoResearch) in 1 % PBS. After incubation, slices were washed three times with 1 % PBS and mounted on slides (Frosted, 75 mm, Fisher Scientific) with Fluoromount aqueous mounting medium (Sigma-Aldrich) and stored at 4 °C.

2.4.2. GFP⁺ transplanted monocyte positive control

A set of Cx3cr1-CreER+/−; Csf1R fl/fl host animals with GFP⁺ transplanted monocytes (Bastos et al., 2025; Aisenberg et al., 2025) were used as a control to confirm MAC2 expression by infiltrating monocytes (Hohsfield et al., 2022). Monocyte transplanted tissues were generated using methods described in Bastos et al. (2025) that achieve high brain parenchymal chimerism with donor monocyte derived macrophages. Briefly, Cx3cr1CreER+/−;Csf1rfl/fl hosts were treated with tamoxifen on p1 and p2 to deplete endogenous microglia, followed by intracranial transplantation of 3×10⁵ GFP⁺ monocytes on p3. Transplanted host animals were then harvested at 2 months of age. For harvesting of brain tissue, mice were anesthetized and perfused with 20 ml cold 1x PBS, then 20 ml cold 4 % PFA. Brain was then dissected and drop-fixed overnight in 4 % PFA before cryopreservation in 30 % sucrose. After cryopreservation, brains were frozen in OCT and stored at −80 °C until sectioning.

2.4.3. Confocal imaging

Confocal microscopy was performed using a Leica Sp8 confocal microscope (Leica, Dee Park, IL, USA). Images were acquired with a 20x/0.75 oil immersion lens at a resolution of 1024 × 1024 pixels. Z-stacks were collected at 30 μm intervals across the full depth of the tissue section. Laser power, gain, and offset settings were kept constant across all groups to allow for direct comparison during quantification. Images from Fig. 1B were enlarged to better visualize and highlight morphological differences. For display purposes, we enhanced the contrast of the images in Fig. 1B–F and J–K.

2.4.4. Quantification of IHC

IBA1⁺ and MAC2⁺ cells were counted using the Cell Counter plugin in Image J Fiji. Cell counts from 3 sections containing dorsal to ventral hippocampus (Bregma −1.82 mm to −2.54 mm; Paxinos and Franklin, 2001) were averaged for each mouse. To calculate the percentage of IBA1⁺ cells that co-expressed MAC2, we first merged the IBA1 and MAC2 channels to identify double-positive cells in each slice. We then counted the total number of MAC2⁺/IBA1⁺ cells and divided this by the total number of IBA1⁺ cells per slice. The resulting percentages were averaged across slices to obtain a final value for each mouse. All cell densities are reported in cells per mm² per animal.

2.5. Contextual Fear Conditioning (FC) and Extinction (FE)

Contextual fear paradigm, a well-established experimental model of fear-based learning (Shechner et al., 2014; Maren et al., 2013; Witgen et al, 2005; Cole et al., 2010), was used to assess hippocampus dependent cognitive functions (Holland and Bouton, 1999). The apparatus consisted of a rectangular chamber (21.6 cm × 17.8 cm × 12.7 cm) with a stainless-steel grid floor. Electric foot shock (0.7 mA) (Med Associates, St. Albans, VT, USA) was delivered through the stainless-steel bars and a LED light outside the chamber provided white illumination. The intensity of the unconditioned stimulus was measured and confirmed with a current test package before placing the mouse inside the chamber (Med Associates, St. Albans, VT, USA). On days 8 and 9 mice were handhandled for 3 mins in a holding room (Hurst and West, 2010). They were then placed in individual boxes on days 10–12 to allow them to acclimate to the behavior room for 40 mins and handled for 3 mins each. On day 13, mice were placed in the conditioning apparatus and were subjected to a 6 mins conditioning protocol. First, they were allowed to habituate for 3 mins inside the chamber and then shocked for 2 s thrice (aversive stimulus) at 1 min intervals. On day 14, mice were placed back in the chamber to assess whether they learned to associate their environment (i.e., the chamber) with the aversive stimulus (i.e., the foot shock). A camera installed in the apparatus was used to observe behavior for 5 mins. The percent time spent freezing was calculated by dividing the total time spent freezing by the total observed time for each animal, as a proxy of fear response (Valentinuzzi et al., 1998; Fanselow, 1990). To assess inhibitory learning or FE (Chang et al., 2009), mice were placed in the chamber again on Day 15, and their fear response measured (Fig. 2A). Many of the same mice used in the FC experiments were also utilized for FE. Our FE cohorts were smaller because we added the extinction paradigm later during data collection. By this point, the first cohort of mice used in the FC experiments had been euthanized and could no longer be utilized for our FE experiments.

Fig. 2.

Injured-depleted mice show improved learning after injury compared to injured mice with no treatment. (A) Schematic of FC and FE paradigm used to test learning and memory after mTBI and PLX treatment. (B) Injured-depleted animals spent more time freezing like sham and sham-depleted animals. Data shown as mean ± SEM; Sham: N = 8; Injured: N = 8; Injured Depletion: N = 10; Sham Depletion: N = 7. There is an interaction effect between injury and treatment; two-way ANOVA: F (1, 27) = 4.532, p = 0.0425. (C) Injured-depleted animals demonstrate decreased fear response on Day 15 similar to sham and sham-depleted animals. Data shown as mean ± SEM; Sham Day 14: N = 6; Sham Day 15: N = 6; Injured Day 14: N = 5; Injured Day 15: N = 5; Injured Depletion Day 14: N = 5; Injured Depletion Day 15: N = 5; Sham Depletion Day 14: N = 4; Sham Depletion Day 15: N = 4. There was a significant difference between Day 14 (M = 32.5, SD = 16.03) and Day 15 (M = 13.67, SD = 14.21) in sham animals; t(5) = 4.056, p = 0.0098, between Day 14 (M = 49.33, SD = 22.05) and Day 15 (M = 15.6, SD = 7.50) in injured-depleted animals; t(4) = 4.225, p = 0.0134, and between Day 14 (M = 42.750, SD = 23.307) and Day 15 (M = 26, SD = 22.045) in sham-depleted mice; t(3) = 3.210p = 0.0490. However, there was no significant difference between Days 14 and 15 in injured mice; t(4) = 2.296, p = 0.0833. Asterisks for multiple comparisons indicate *p < 0.05, **p < 0.01.

2.6. Electrophysiology

2.6.1. Slice preparation

Coronal sections were cut as described previously (Smith et al., 2015; Witgen et al., 2005; Palmer et al., 2016). Briefly, mice were deeply anaesthetized with isoflurane and decapitated. Brains were quickly (less than 1 min) extracted and placed in ice-cold-oxygenated (95 % O₂/5% CO₂) sucrose-based artificial cerebral spinal fluid (aCSF) composed of (in mM): sucrose, 202; KCl, 3; NaH₂PO₄, 2.5, NaHCO₃, 26; glucose, 10; MgCl₂, 1; CaCl₂, 2. Coronal sections (350 μm) were cut in sucrose aCSF with a vibratome (VT1200S, Leica Biosystems, Buffalo Grove, IL, USA) and transferred to an incubator filled with oxygenated (95 %O₂/5% CO₂) aCSF solution composed of (in mM): NaCl, 130; KCl, 3; NaH₂PO₄, 2.5, NaHCO₃, 26; CaCl₂,10; MgCl₂, 1. Slices were placed in a water bath (32–33 °C) for 60 mins and then at room temperature.

2.6.2. Field Excitatory Postsynaptic Potential (fEPSP) recordings

For electrophysiology, slices were placed in an interface brain slice chamber system BSC1 (Scientific System Design Inc.) and irrigated with aCSF at a rate of 2.5 ml min⁻¹. An upright microscope (MZ7.5; Leica Microsystems, Deerfield, IL, USA) was used to visualize slices through a 4X objective. Photomicrographic images of slices were referenced to ensure electrode placement was consistent between slices. Recording electrodes were fabricated using borosilicate glass (World Precision Instruments, Sarasota, FL, USA, #1B150F-4), pulled to a tip resistance of 2–4 MΩ and filled with aCSF. A Multiclamp 700B amplifier and pClamp11 data acquisition software (Molecular Devices, Sunnyvale, CA, USA) were used to record fEPSPs. A non-concentric bipolar stimulating electrode (World Precision Instruments, Sarasota, FL, USA, #ME12206) was used for stimulation. To record Schaffer Collateral evoked fEPSP, the recording electrode was placed in the Stratum Radiatum of area CA1. To record the medial perforant path (MPP) evoked fEPSP, the stimulating and recording electrodes were placed in the inner molecular layer (imDG). To record the lateral perforant path (LPP) evoked fEPSP the electrodes were placed in outer molecular layer (omDG). Input-output curves (50–500 μA stim, 50 μA increments, 8 s inter stimulus interval) were determined for all three evoked fEPSP (see Supplementary Fig. A.1 for illustration of electrode placements).

Several drugs were used to isolate and confirm waveform components (see Supplementary Fig. A.2). MPP and LPP evoked fEPSPs were confirmed with DCG-IV (1 μM, Tocris Bioscience), a mGluR6 agonist used to block MPP) and LAP4 (20 μM, Tocris Bioscience), a selective group III mGluR agonist to block LPP. In addition, D-(–)-2-Amino-5-phosphonopentanoic acid (APV; Abcam, 50 μM) and 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX; Abcam, 20 μM) were used to block ionotropic glutamatergic transmission and confirm fEPSP. This was followed by Tetrodotoxin (TTX; Abcam, 0.4 μM), a voltage-gated sodium channel blocker to confirm fiber volley. Slopes of fEPSP and fiber volley were measured in pClamp11 (Molecular Devices).

Thalidomide (5 μM, Tocris Bioscience), an immunosuppressant (Moreira et al., 1993; Corral and Kapplan, 1999; Keenan et al., 1991) was bath-applied in some slices from sham and injured animals to determine the effects of TNF-α on neural circuits. Etanercept, a drug that acts as a TNF receptor by binding to TNF-α and β (Pan and Gerriets, 2019), was injected intraperitoneally (1 mg/kg, MedChem Express) in mice 2 h before electrophysiology experiments 7 days post-injury (dpi).

2.7. Luminex inflammatory panel

Cytokine and chemokine quantification in brain tissue of injured and control mice was performed using a pre-mixed commercial kit from Millipore (MCYTOMAG-70 K-14PX) containing the following analytes: TNF-α, Il-6, IL-10, IL-1β, IFNγ, GM-CSF, IL-1α, IL-17A/CTLA8, IL-2, IL-4, IL-12p40, IL-12p70, M–CSF, and IL-13. The entire ipsilateral (to injury) side of brains were collected at 6 h, 12 h, 48 h, and 7 days after in sham, injured, injured-depleted, and sham-depleted mice. Samples were homogenized in 1 ml of T-PER-cOmplete protein buffer (Thermo Scientific) using a Model PRO-200 double insulated micro-homogenizer at 60 HZ (PRO Scientific Inc. Oxford, CT, USA). Homogenates were centrifuged at 9000g for 10 min, and the supernatant was collected, aliquoted, and stored at −80°C. Samples were thawed for cytokine analysis. Assays were performed by the Human Immunology Core at the Perelman School of Medicine at the University of Pennsylvania, following manufacturer protocol. Data were acquired on a FlexMAP 3D quantification instrument, and analysis was done using the Luminex^® xPONENT^® 4.2 (Diasorin) and Bio-Plex Manager^™ Software 6.1 (BioRad). Data quality was determined by ensuring that the standard curve for each analyte had a five-parameter logistic coefficient (5PL R²) value > 0.95 with or without minor fitting using the xPONENT software. The percent recovery of standard curves was between 70–130 % for all analytes. The intraassay coefficient of variation was <10 % for each inflammatory biomarker measured. Each cohort had 4–5 biological replicates. Out of the 14 cytokines tested, only 8 yielded concentrations within the standard range of the assay for the respective cytokines (see Supplementary Table A.1 and Supplementary Fig. A.3). If the analytes included measurements that were out of the range of the standard curve, they were excluded from further analyses.

2.8. Sex differences

Physiological (Somach et al., 2023), structural, and behavioral differences (Rubin and Lipton, 2019) exist between male and female mice after mTBI. Therefore, electrophysical and behavioral changes between injured and sham mice were assessed in both sexes. Due to the lack of behavioral and electrophysiological deficits at 7 dpi in female mice (Supplementary Fig. A.4), which may reflect a delayed injury response or sex-specific resilience mechanisms, female mice were excluded from further experiments.

2.9. Statistical analysis

All analyses were performed using Prism 10.3.0 (GraphPad Software, La Jolla, CA, USA), RStudio 4.3.3, and MATLABR2024a (MathWorks) and assessed at a p < 0.05 level. Prior to statistical testing, all datasets were assessed for normality using the Shapiro–Wilk test and for homogeneity of variances using the Bartlett test. Assumptions for parametric analyses were met for all datasets.

2.9.1. Immunohistochemistry

Differences between groups were assessed using a two-way ANOVA to evaluate the main and interaction effects between injury conditions (sham vs. injured) and treatment (PLX vs. no PLX). For significant interactions (p < 0.05), post hoc tests were performed using Fisher’s LSD.

2.9.2. Contextual Fear Conditioning (FC) and Extinction (FE)

Freezing behavior was recorded at 5 s intervals with a total of 60 possible freezing instances. A two-way ANOVA was used to assess main effects of condition (sham vs. injured) and treatment (PLX vs. no PLX) as well as interaction effects. If the interaction was significant (p < 0.05), Fisher’s LSD post hoc tests were performed to evaluate specific interactions. Additionally, paired-t-tests were used to evaluate differences between freezing responses on Day 14 and Day 15 in the same mice.

2.9.3. Field Excitatory Postsynaptic Potential (fEPSP) recordings

Slopes of the evoked fEPSPs were determined from the initial linear portion of the response in Clampfit 11.7 (see Fig. 3A). Slope data from 2 to 3 slices in each animal were averaged for currents in the 50–500 μA range. These were used to create I/O curves for each animal. To statistically assess the effects of treatment (PLX vs. no PLX) on condition (sham vs. injured), a mixed-effect two-way ANOVA model was used. Mixed-effect models account for repeated recordings on the same animal by including slice as a random effect; condition and treatment as the independent variables and fEPSP slopes as the dependent variable. Main and interaction effects between injury condition and treatment used an F-test with a Satterthwaite correction for degrees of freedom. Statistically significant effects were analyzed by Wald tests to evaluate specific comparisons of interest (e.g., sham vs. injured, injured vs. injured-depleted, sham vs. sham-depleted). The same analysis was used to compare the effects of thalidomide and etanercept on injured slices.

Fig. 3.

Microglia depletion restores excitability shifts in area CA1 and DG at 7 dpi. (A) Schematic representation of the typical fEPSP waveform recorded from area CA1. The key components of the trace are labeled as: (1) Stimulus artifact which is the response evoked by current from extracellular electrical stimulus, (2) Fiber volley, an indirect measure of pre-synaptic axonal activity, (3) Slope which represents the rapid depolarization of a population of post-synaptic neurons and is used as a measure of excitability, and (4) fEPSP also known as the field excitatory post-synaptic potential. Stimulus artifacts were partially removed or fully removed from representative traces. (B) I/O curve of fEPSP slopes in area CA1 shows increased excitability to sham levels in injured-depleted mice. Data shown as mean ± SEM; Sham: N = 13; Injured: N = 12; Injured Depletion: N = 13; Sham Depletion: N = 15. Mixed effect two-way ANOVA: significant interaction effects: F(1, 196) = 11.73; p = 0.0007. (C) I/O curves of fEPSP slopes in omDG show decreased excitability (to sham levels) in injured-depleted mice. Data shown as mean ± SEM; Sham: N = 9; Injured: N = 8; Injured Depletion: N = 10; Sham Depletion: N = 7. Mixed effect two-way ANOVA: significant interaction effects: F (1,133) = 7.289, p = 0.0078. (D) I/O curve of fEPSP slopes in imDG shows increased excitability (to sham levels) in injured-depleted mice. Data shown as mean ± SEM; Sham: N = 13; Injured: N = 11; Injured Depletion: N = 11; Sham Depletion: N = 12. Mixed effect two-way ANOVA: significant interaction effects: F (1, 163) = 8.983, p = 0.0032. Asterisks for multiple comparisons indicate, **p < 0.01.

2.9.4. Luminex panel

For Luminex panels in Fig. 4, concentrations of each inflammatory biomarkers were measured at the different timepoints. A two-way ANOVA was used to look at main and interaction effects between condition and timepoints. Significant interactions were analyzed by Tukey’s multiple comparisons test to compare between groups (injured, sham, injured-depleted) at different timepoints. For Fig. 5, a two-way ANOVA was also used to evaluate main and interaction effects of injury condition and treatment at 7 dpi. If interactions were significant, Tukey’s multiple comparisons test was used for specific comparisons.

Fig. 4.

IL-6 and TNF- α levels increase at 6 h in injured and injured-depleted mice. (A-H) Cytokine concentrations at 6 h, 12 h, and 48 h post-injury in sham, injured, and injured-depleted mice. (A) IL-6 concentrations were significantly different across time points; two-way ANOVA: F(2, 33) = 6.699, p = 0.0036; but no interaction effects were noted between groups; F(4, 33) = 1.670, p = 0.1805. (B) M–CSF concentrations showed no significant interaction effects; two-way ANOVA: F(4, 33) = 1.338, p = 0.2757, but main effects across timepoints were significant; F(2, 33) = 4.649, p = 0.0166. (C) TNF-α concentrations, were significantly different across time and treatment condition; two-way ANOVA: F(4, 36) = 5.513, p = 0.0014. (D) IL-17 showed no significant interaction effects between groups; two-way ANOVA: F(4, 36) = 2.280, p = 0.317, with only a difference between injured and injured-depleted animals at 6 h (p = 0.0797). (E-H) No significant changes were observed across groups in IFNγ; two-way ANOVA: F(4, 30) = 0.4931, p = 0.7408, IL-1β; F(4, 33) = 1.342, p = 0.2751, IL-12(p70); F(4, 33) = 0.5092, p = 0.7293, and IL-2; F(4, 33) = 0.7922, p = 0.5387. Data shown as: mean ± SEM: N = 4–5 in each group, Asterisks for multiple comparisons indicate *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 5.

Depleting microglia reduces TNF-α levels at 7 dpi. (A-H) Concentration of inflammatory markers at 7 dpi in sham, injured, sham-depleted, and injured-depleted cohorts. (A) IL-6 concentrations are significantly increased in injured-depleted mice compared to injured and sham animals with no treatment; significant main effects of treatment; two-way ANOVA: F(1, 13) = 19.61; 0.0007. (B) M–CSF concentrations showed significant main effects of treatment; two-way ANOVA: F(1,15) = 283.1, p < 0.0001, but no interaction effects; F(1, 15) = 3.461, p = 0.0825. (C) TNF- α concentrations showed interaction effects of treatment and injury; two-way ANOVA: F(1, 16) = 40.07, p < 0.0001. (D) No significant interaction effects were found in IL-17; two-way ANOVA: F(1, 16) = 0.4831, p = 0.4970 but there were significant main effects of treatment; F(1, 16) = 26.36, p = 0.0001. (E) IL-1 β showed no interaction effects; two-way ANOVA: F(1,15) = 1.179, p = 0.2085, but significant main effects of treatment; F(1, 15) = 13.83, p = 0.0021. (F) IL-2 concentrations also showed no interaction effects; two-way ANOVA: F (1,15) = 0.8561, p = 0.3695, but significant effects of treatment; F(1, 15) = 30.10, p < 0.0001. (G & H) No significant effects were found between groups in IFNγ; two-way ANOVA: F(1, 14) = 0.0026, p = 0.9593, and IL-12(p70) concentrations; F(1,16) = 3.001, p = 0.1024. Data shown as: mean ± SEM: N = 4–5 in each group. Asterisks for multiple comparisons indicate *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

2.9.5. Fiber volley vs. fEPSP ellipses

Error ellipses indicating the limits of the SEM of slopes and fiber volley were used to illustrate the variability in the fEPSP and fiber volley slope. The fiber volley and fEPSP slope ratio was used to compare groups at different stimulation strength (see Supplementary Fig. A.5). Two-way ANOVA was then used to assess main and interaction effects of condition and stimulation strength. For significant comparisons, Tukey’s multiple comparisons test was used. For Fig. 7A–C, the independent variables were the stimulation strength (50–500 μA) and condition, described as sham vs. injured, and for Fig. 7D–F, condition was described as before and after thalidomide bath application.

Fig. 7.

Excitability shifts in hippocampal circuitry point to pre-synaptic dysfunction after injury which is restored by application of thalidomide. (A-F) Fiber volley and fEPSP slope, at simulation ranging from 50 to 500 μA with error ellipses indicating 95% confidence. Figs. 7A–7C demonstrate presynaptic dysfunctions, as the relationship between the fiber volley and the fEPSP does not significantly change after injury in all three subregions of the hippocampus. In figures 7D–7F these disruptions are restored to sham levels post-thalidomide application in both area CA1 and omDG. However, in imDG (Fig. 7F), the correlation between fiber volley and fEPSP in injured vs. sham animals is different as shown by the lack of change in fiber volley slope and decreased overlap of the ellipses. This suggests that network excitability may be restored through a combination of presynaptic and postsynaptic means (see Supplementary fig. A.5 for details on analysis).

3. Results

3.1. Microglial density increases after injury while both injury and depletion lead to MAC2 expression by hippocampal macrophages

After TBI, increases in number of microglia are reported (Donat et al., 2017; Nagamoto-Combs et al., 2007); therefore, we sought to confirm changes in microglial density in our mTBI model. We investigated these changes 7 dpi because the week after injury is an important therapeutic window (Grady and Lam, 1995; Mohamadpour et al., 2019). In particular, we were interested in the cumulative inflammatory response during this window (Simon et al., 2017; Postolache et al., 2020). Furthermore, we wanted to use these values to compare and confirm if we achieved significant microglia depletion. The experimental timeline is summarized in Fig. 1A. Representative images of IBA1⁺ and MAC2⁺ cells in the ipsilateral (to injury) CA1 region of the hippocampus from sham, injured, injured-depleted, and sham-depleted mice are illustrated in Fig. 1B. IBA1⁺ cells in the depleted groups exhibited a strikingly different morphology compared to the non-treated groups (Fig. 1C–F), consistent with prior studies (Elmore et al., 2015; Zhan et al., 2020). Our positive control using mice with GFP⁺ transplanted monocytes confirmed MAC2 expression by monocyte-derived macrophages (Fig. 1J–K). Quantification of IBA1⁺ cells revealed a significant increase in microglia density after injury (Fig. 1G). Specifically, there was an approximately 87 % reduction in microglia counts in injured animals and an 84 % decrease in uninjured animals following PLX-3397 treatment. Additional images were collected in the cortex and thalamus to provide a sampling across multiple regions (see Supplementary Fig. A.6 and Table A.2). Injured mice showed a significant increase in MAC2⁺ cells compared to sham mice, raising the possibility that macrophages may have infiltrated the brain following injury (Fig. 1H); contributing to the observed increase in IBA1⁺ cells. MAC2⁺ cells were also present in both injured and sham mice that were treated with PLX and made up the majority of IBA1⁺ cells following treatment (Fig. 1I), consistent with a prior study (Zhan et al., 2020). Together, this suggests that mTBI leads to microgliosis with increased MAC2 expression, consistent with a combination of monocyte infiltration and endogenous microglial reactivity to injury.

3.2. Microglia depleted injured mice show improved learning post-injury compared with injured mice

Contextual FC and FE engage the hippocampus (Maren and Fanselow, 1997; Kim and Cho, 2020). Although other brain regions are involved, the hippocampus is crucial for encoding and recalling the context associated with the memory. We have previously reported deficits in contextual FC, and excitatory/inhibitory (E/I) shifts in area CA1 and DG (Witgen et al., 2005) after mTBI. Here, we asked how depletion of microglia affects the deficits in contextual FC and FE. In our contextual FC paradigm, mice were handled on days 8––12 after mTBI, conditioned on day 13 and tested on days 14 – 15 (Fig. 2A). Injured animals demonstrated decreased freezing compared to sham animals on Day 14. In contrast, injured-depleted animals exhibited increased freezing (similar to sham and sham-depleted) compared to injured animals with no treatment. This indicates their ability to associate the context (the box) with the shock (Fig. 2B). On Day 15, we placed the same mice in the box again with no shock, to assess FE. No significant differences were observed between Day 14 and Day 15 in the injured group (Fig. 2C), suggesting that injured animals may be slower at learning following mTBI. However, injured-depleted animals showed a significant reduction in freezing behavior, similar to sham and sham-depleted animals on Day 15 (Fig. 2C). The preserved FE behavior indicates that injured-depleted animals quickly learn that the context is not always aversive.

3.3. Microglia depletion restores normal excitability in area CA1 and DG at 7 dpi

To identify alterations in the hippocampal network, we performed a series of extracellular recordings, stimulating afferent pathways that connect to pyramidal neurons in area CA1, and DG granule cells in the inner molecular layer (imDG) as well as outer molecular layer (omDG). We measured the fEPSP slope at each stimulation strength, by fitting a straight line between two cursors: one positioned after the fiber volley and the other, before the population spike. (Fig. 3A). We confirmed decreased network excitability in area CA1 after injury (Fig. 3B). We also found a significant increase in the fEPSP slopes in omDG and a decrease in imDG after injury (Figs. 3C and 3D). Interestingly, excitability returned to sham levels in these regions in injured-depleted mice (Fig. 3B–D). Given that microglia play a role in synaptic functions, plasticity, as well as excitatory and inhibitory cellular responses (Santello and Volterra, 2012; Singh et al., 2022; Wang et al., 2018), we asked if depletion contributes to pathway dysfunctions. To answer this, we performed experiments in sham-depleted animals. We found that sham-depleted mice exhibit a significant decrease in area CA1 excitability compared to sham controls (Fig. 3B). A slight but not significant increase in excitability in omDG and a slight decrease in excitability in imDG was observed (Figs. 3C and 3D). These findings suggest that microglia are important for maintaining network excitability under normal conditions and depleting them may affect some network properties in the hippocampus. To confirm our pathways, fiber volleys, and fEPSPs, a series of drugs (described in Supplementary A.2) were bath-applied at the end of some of our recordings.

3.4. IL-6 and TNF-α levels increase at 6 h in injured and injured-depleted mice

Microglial participation in the inflammatory response consists of several components, a major one being the release of inflammatory cytokines, many of which have been linked to time-specific changes in brain function. Studies show that both injury (Vedantam et al., 2020) and PLX treatment (Jin et al., 2017) alter cytokine levels as early as 24 h. However, there is no literature available on expression levels within 12 h after mTBI. Consequently, we looked at early time points to obtain a more comprehensive expression profile of the inflammatory molecules. Our results show that IL-6 concentrations were significantly elevated in injured and injured-depleted mice compared to sham at 6 h (Fig. 4A). Additionally, macrophage colony stimulating factor (M–CSF), a pro-and-anti-inflammatory cytokine involved in the regulation and proliferation of monocytes and macrophages under homeostatic conditions (Stanley and Chitu, 2014; Ushach and Zlotnik, 2016), significantly increased in injured and injured-depleted mice at 6 h, compared to sham animals (Fig. 4B). TNF-α and IL-17 concentrations were also significantly elevated in injured-depleted animals at 6 h post-injury (Fig. 4C and D). However, no changes in IFN-γ, IL-1β, IL-2, or IL-12(p70) were observed between groups (Fig. 4E–H); suggesting that these cytokines may not be involved in the early response to mTBI. These findings indicate that depletion of microglia does not affect the early expression of inflammatory cytokines.

3.5. Depleting microglia reduces TNF-α levels at 7 dpi

Given that our behavior and recording experiments were performed at 7 dpi, we asked what inflammatory molecules were at play in sham, injured, sham-depleted and injured-depleted mice. We found that IL-6 concentrations were significantly increased in injured-depleted animals compared with sham and injured with no treatment (Fig. 5A). Concentrations of M–CSF were significantly increased in injured-and-sham-depleted mice compared to injured and sham animals which may be an attempt to compensate for microglia depletion (Fig. 5B). TNF-α concentrations showed the largest difference between groups, including a 2-fold decrease in injured-depleted animals compared to injured mice with no treatment. (Fig. 5C). Interestingly, the cytokines IL-17, IL-1β, and IL-2 (Fig. 5D–F) showed a significant decrease in both sham-depleted and injured-depleted cohorts compared to injured animals with no treatment, likely due to the immunomodulatory effects of CSF1R inhibition itself. Neither IFNγ nor IL-12 (p70) were different between the 4 groups (Fig. 5G and H). Notably, these findings highlight that microglia depletion alters the concentration of most cytokines at 7 dpi, corresponding to the onset maximal microglia depletion.

3.6. Thalidomide restores excitability in hippocampal circuits at 7 dpi

Pro-inflammatory cytokines regulate neurotransmission in the hippocampus in both normal tissue (Beattie et al., 2002; Stellwagen et al., 2005; Stellwagen and Malenka, 2006), and in neurological conditions including Alzheimer’s disease (Singh et al., 2019) and mTBI (Batsaikhan et al., 2019; Palmer et al., 2022; Lin et al., 2020). Bath application of thalidomide, which was used as a TNF- α synthesis inhibitor in a mouse model of Alzheimer’s disease, restored long-term potentiation (LTP) – an important phenomenon in learning and memory (Singh et al., 2019). Additionally, thalidomide analogs have been shown to modulate the inflammatory response after mTBI (Batsaikhan et al., 2019; Palmer et al., 2022; Lin et al., 2020). Given the large and significant reduction in microglia due to PLX treatment, and the detrimental effects of neuroinflammation (Loane and Kumar, 2016; Donat et al., 2017; Shao et al., 2022; Wofford et al., 2019; Bourgognon and Cavanagh, 2020; Jassam et al., 2017), we hypothesized that a corresponding decrease in pro-inflammatory modulators after microglia depletion contributes to the improvements we observed in hippocampal-dependent behavior and network excitability. To test this hypothesis, we investigated whether suppressing the immune response would yield results comparable to microglia depletion post-injury. In particular, we asked whether pro-inflammatory cytokine suppression improves network excitability. To determine this, we prepared slices from injured mice at 7 dpi, and bath applied thalidomide. Thalidomide restored normal network excitability in all three areas tested – it increased excitability in area CA1 (Fig. 6A) and imDG (Fig. 6C), and decreased excitability in omDG (Fig. 6B). To confirm that the drug was not simply inhibiting any acute inflammation caused by the slicing procedure, we washed in thalidomide in sham slices. Thalidomide significantly increased excitability in area CA1 in sham slices but did not significantly affect excitability in the DG (Fig. 6A–C).

Fig. 6.

Thalidomide restores excitability in hippocampal circuits at 7 dpi. (A) I/O curves of fEPSP slopes in area CA1 show no interaction effects; however, the significant increase in excitability in both injured and sham slices post 15-minute bath application of thalidomide suggest main effects of treatment; two-way ANOVA: F(1, 13) = 9.107, p < 0.0099. (B) Outer molecular layer Dentate Gyrus (omDG) show decreased excitability in injured slices after thalidomide application with no changes in sham slices; two-way ANOVA F(1, 117) = 18.23, p < 0.0001. (C) Inner molecular layer Dentate Gyrus (imDG) also revealed no changes in sham slices but an increase in excitability in injured animals post-thalidomide application; two-way ANOVA: F(1, 117) = 5.777, p = 0.0319. Data shown as mean ± SEM; Injured: N = 8; Sham: N = 7. Asterisks for multiple comparisons indicate ****p < 0.0001.

3.7. Excitability shifts in hippocampal circuitry point to pre-synaptic dysfunction after injury which is restored by application of thalidomide

We probed the mechanism through which thalidomide restores network excitability by analyzing the relationship between the fiber volley (i.e., the presynaptic action potential recorded extracellularly) and the fEPSP in slices from injured and sham mice (Fig. 7A–C). We observed a proportional change in the fiber volley and fEPSP slope in all three areas, i.e., the fEPSP (postsynaptic) and the fiber volley (presynaptic) changed by the same proportion for both sham and injured animals. This suggests that the changes in fEPSP slope were due primarily to changes in activation of the presynaptic axons and/or pre-synaptic terminals. As a result, in CA1 and imDG the fEPSP and fiber volley proportionally decreased (Figs. 7A and 7C), while in omDG they proportionally increased (Fig. 7B) after injury. Interestingly, thalidomide reversed the disruption in all three regions (Fig. 7BD–F).

3.8. Etanercept (ETA), a TNF-α and β blocker, injected 2 h before slice electrophysiology restores excitability shifts at 7 dpi

Given the known off-target effects of thalidomide (Pellacani and Eleftheriou, 2020) and that TNF-α had the largest change between groups at 7 dpi (Fig. 5C), we asked if specifically targeting TNF-α produces the same results observed in Fig. 6. To this end, we used etanercept (ETA), a soluble TNF-α and β receptor which binds TNF-α and β and prevents them from binding to their endogenous receptors. ETA was injected intraperitoneally in injured and sham mice 2 h before hippocampal slice preparation (Campbell et al., 2007) and restored excitability to sham levels in CA1 and DG (Fig. 8A–C). These findings further support our hypothesis that increased levels of TNF- α in injured slices contribute to disruptions in hippocampal circuit function.

Fig. 8.

Injection of etanercept (ETA), two hours before slice preparation restores excitability shifts at 7 dpi. (A) I/O curve of fEPSP slopes in area CA1 show significantly increased excitability (to sham levels) in injured mice injected with ETA two hours before slice preparation. Mixed effect two-way ANOVA: F(1, 62) = 13.34, p = 0.0003. (B) Outer molecular layer Dentate Gyrus (omDG) showed a significant decrease in excitability in injured slices of mice injected with ETA; F(1, 62) = 6.190, p = 0.0025 (C) Inner molecular layer Dentate Gyrus (imDG) show a significant increase in excitability in injured slices of injected mice; F(1, 62) = 3.291, p = 0.0122. Data shown as mean ± SEM; Injured + ETA: N = 6; Sham + ETA: N = 4; Injured + Saline: N = 5; Sham + Saline: N = 4. Asterisks for multiple comparisons indicate *p < 0.05, **p < 0.01, ***p < 0.001. (D) Representative fEPSP traces from injured and sham animals injected with ETA or saline.

4. Discussion

Our study investigated the effects of microglia depletion on hippocampal network excitability and behavior at 7 dpi, with a focus on neuroinflammatory modulation, and more specifically TNF-α. We performed these experiments by pharmacologically inhibiting the CSF-1R using PLX-3397, administered through diet. We demonstrate that microglia depletion post mTBI leads to restoration of hippocampal network excitability, likely through neuroinflammation modulation. We also confirm that microglia density significantly increases after mTBI and demonstrate that PLX-3397 can effectively reduce microglial populations in the hippocampus by approximately 84 % in sham animals and 87 % in injured mice. Interestingly, we found that while microglia depletion significantly alters the concentration of cytokines at 7 dpi, it has no effect within hours, likely due to the fact that PLX-3397 takes days to significantly deplete microglia. Our findings also have implications for understanding the role of TNF-α in circuit recovery, given that injecting ETA into mice at 7 dpi restores network excitability similarly to thalidomide and microglia depletion.

At 7 dpi, the observed increase in IBA1⁺ cells in the hippocampus is consistent with previous studies reporting both increased microglia counts and activation in response to injury (Donat et al., 2017; Witcher et al., 2021). This increase is also consistent with evidence that microglia proliferate post-injury (Urrea et al., 2007; Willis et al., 2024), and while initially protective, may subsequently contribute to cognitive dysfunctions during the later stages of injury (Saba et al., 2021). Immunostaining demonstrated a significant increase in MAC2 expression by parenchymal macrophages following injury. MAC2, a mammalian lectin, is associated with monocyte infiltration (Hohsfield et al., 2022) and with microglial injury responses (O’Koren et al., 2019), where it is thought to have a protective role (Yu et al., 2024). Interestingly, it is also induced in microglia by PLX treatment (Zhan et al., 2020). Based on this information, we can speculate that microglial state change and/or monocyte infiltration may also contribute to the positive impacts of microglia depletion. These possibilities can be addressed using genetic knockout and fate mapping, respectively (Liu et al., 2019).

Behavioral outcomes were significantly improved in injured animals that underwent microglia depletion, as evidenced by their performance in contextual FC. Specifically, injured-depleted mice displayed freezing behavior comparable to sham animals during both acquisition and extinction phases, suggesting that microglia depletion mitigates cognitive deficits, which aligns with previous reports (Boland and Kokiko-Cochran, 2024). Among the improvements in injured-depleted mice, the restored network excitability in hippocampal circuitry was noteworthy. We specifically targeted the Schaffer collateral–CA1 pathway because it is a well-characterized excitatory input from CA3 to CA1 pyramidal cells and is relevant for studying synaptic plasticity and circuit integrity after injury (Cohen et al., 2007; Kwon et al., 2018). Similarly, the outer and inner molecular layers of the DG were examined due to their roles in receiving input from the entorhinal cortex (Borzello et al., 2023). These regions are also highly sensitive to inflammatory (Parnet et al., 2002), and synaptic disruptions post-TBI (Atkins et al., 2011) and represent major hubs for hippocampal information processing. Our electrophysiological findings suggest that depleting microglia post-injury can help restore network dysfunctions associated with learning and memory, and in particular, they support the conclusion that immune suppression, and more specifically blocking TNF-α improves synaptic function post-injury. Microglia interact with neurons via signaling molecules and play key roles in synaptic pruning during development and disease (Hong et al., 2016a; Hong et al., 2016b; Cornell et al., 2022). When activated, however, they can impair synaptic plasticity and cognition, particularly in area CA1 (Cornell et al., 2022; Chen et al., 2023). Given these roles in neural circuit function, their involvement in synaptic dysfunction following mTBI is not surprising.

We observed some disruptions in depleted mice with no injury, specifically in area CA1, where we report decreased excitability in sham-depleted animals. Microglia are dynamic in the healthy brain; they are crucial for neuronal surveillance during post-natal development and are involved in shaping synapses to refine neuronal connectivity. Considering that area CA1 is extremely vulnerable to CNS alterations (e.g., injury, ischemia, hypoxia) (Schmidt-Kastner and Freund, 1991; Kirino, 2000), it is likely that this subregion responds quickly to the absence of microglia. Although these differences were not significant in DG, there is a trend in the expected direction (i.e., slight increase in omDG and slight decrease in imDG), suggesting that with longer depletion time points, we could potentially find additional disruptions in these subregions of the hippocampus in sham-depleted mice. Additionally, the absence of deficits in fear conditionning in sham-depleted mice, despite the deficit in hippocampal network excitability indicates that our behavioral paradigm does not solely rely on hippocampus (Gerwitz et al., 2000). Future studies should incorporate additional assessments, such as spatial object recognition, to better evaluate hippocampal circuitry, and other brain regions, such as the amygdala (Palmer et al., 2016).

Given that little is known about the early trajectory of pro-inflammatory cytokines after mTBI in mice, we chose our depletion timeline to investigate how microglia depletion, occurring in the context of an ongoing inflammatory response, influences hippocampal function and immune signaling. Prior studies have examined depletion before or long after injury, with variable outcomes (Boland and Kokiko-Cochran, 2024). By targeting the early post-injury window, our model closely mimics a realistic therapeutic scenario, where intervention would be initiated after the injury. Our Luminex analysis revealed changes in cytokine levels post-depletion that are relevant for identifying appropriate therapeutic windows after concussion. For example, TNF-α showed elevated concentrations at 6 h in injured-depleted mice, followed by a gradual decrease to sham levels by 7 dpi. Literature suggests that TNF-α typically peaks at 24 h (Hayakata et al., 2004), but can remain elevated up to a year post-injury in humans (Chaban et al., 2020). In a rat model, concentrations are elevated up to 72 h in severe TBI but not mTBI (Knoblach et al., 1999); therefore, it is possible that in mTBI, TNF-α release takes longer. We also observed that M–CSF, which has both pro-and anti-inflammatory properties, was significantly elevated in injured-depleted mice at 7 dpi compared to both sham and injured groups. This increase in M–CSF could be a compensatory mechanism in depleted mice in response to the decrease in microglia to replenish these cells (Green et al., 2020). Additionally, M–CSF can induce M2 polarization of macrophages (Martinez and Gordon, 2014), thereby promoting tissue repair post-injury (Li et al., 2021).

Bath application of thalidomide to hippocampal slices from injured mice restored normal excitability in both DG and area CA1 subregions of the hippocampus, similar to the effects of microglia depletion. Changes of similar magnitude were not observed in sham animals, except in area CA1, suggesting that while the thalidomide effect in CA1 might not be entirely injury-specific, it would nonetheless still be restorative for injured mice. Other cytokines are also altered by thalidomide, beyond TNF-α (Ye et al., 2006; Jung et al., 2019), and could therefore contribute to alterations in non-injured slices. To further understand how thalidomide alters network excitability, we generated graphs to examine the relationship between the fEPSP and the presynaptic fiber volley slopes. This correlation provides insight regarding whether any observed change in the fEPSP is likely to have a presynaptic (axons and/or pre-synaptic terminals) versus a postsynaptic component. It also answers whether therapeutic interventions can specifically target and restore said disturbances at their source. Our results revealed that deficits across all hippocampal regions in injured mice are partially driven by pre-synaptic alterations and immunosuppression mitigates the dysfunction in these presynaptic fibers, including axons which originate in the entorhinal cortex (EC) (Amaral and Witcher, 1989; Jones et al., 1993). Disruptions in the EC can alter oscillatory activity in the hippocampus (Ortiz and Gutiérrez, 2015), highlighting the critical role of EC-hippocampal connectivity (Colgin, 2015; Fernández-Ruiz et al., 2021; Mizuseki et al., 2009; Schomburg et al., 2014). Future work using patch clamping methods to identify the precise cellular mechanism for the changes observed would further investigate how microglia affect network excitability properties.

Injection of ETA has been shown to produce improvements in TBI outcomes in rodents (Chio et al., 2010, 2013; Tuttolomondo et al., 2014) and humans (Tobinick et al., 2014); however, its role in circuit function is not fully understood. The use of ETA, administered acutely 2 h before slice preparation at 7 dpi, restored excitability (similar to thalidomide and PLX treatments) in CA1, imDG, and omDG in injured mice, further supporting the role of TNF-α in modulating circuit dysfunction following injury. Given that TNF-α levels may rise more gradually in mTBI compared to more severe injuries (i.e., begin to increase 7 days post mTBI), these results indicate that targeted cytokine interventions during this specific window may represent a promising therapeutic strategy for further investigation. It is important to note that previous studies have administered ETA via intracerebroventricular (Corrigan et al., 2015) or perispinal injection (Tobinick, 2010). In our study, IP administration allows for systemic modulation of TNF-α, which is particularly relevant at 7 dpi, when secondary cascades (e.g., blood brain barrier disruption and peripheral immune infiltration) are known to contribute to increased pro-inflammatory cytokines and circuit impairments (Lozano et al., 2015; Schwulst et al., 2013). Nonetheless, future studies should further examine the effect of ETA injections directly into the brain post mTBI, which could be beneficial to investigate region-specific deficits. Additionally, ETA has been shown to reduce other pro-inflammatory cytokines (Chio et al., 2010), and while the primary mechanism of this drug is to mimic endogenous TNF-α-and-β receptors, its broader anti-inflammatory effects may also contribute to improved circuitry in injured mice.

There are limitations to be considered when interpreting our results. Although microglia depletion was achieved using PLX-3397 diet for 7 days, the effects of this treatment on microglia function may not fully capture the complexity of microglial responses in vivo, since long term effects of PLX post-TBI were not assessed in the current work. Future studies with longer depletion timelines could provide further details on chronic effects of depletion. PLX-3397 also has off-target effects on peripheral immune cells and other receptors which may alter the overall inflammatory response (Claeys et al., 2023). We could assess these confounding effects in the future by introducing an inducible genetic knockout of CSF-1R (Bastos et al., 2025) in microglia to deplete them, which would help identify the potential effects attributable to other mechanisms. Additionally, microglia are essential in the healthy brain which means that microglia depletion can only be used as a short-term therapeutic intervention. In the future, it will be important to assess the effects of depletion followed by repopulation on network excitability. While our study was designed to assess injury-related changes during the clinically relevant therapeutic window following mTBI (Grady, 1994; Mohamadpour et al. 2019), it is important to note that this window does not fully capture the long-term progression of TBI-related impairments, which may continue to evolve well beyond the acute phase. It should also be noted that our Luminex panel originally included 13 cytokines, but only eight produced values within the assay’s standard range. This may suggest the other cytokines are not involved in mTBI or may simply reflect a technical limitation, as brain tissue is more difficult to process for Luminex analysis compared to serum or cell culture (Staples et al., 2013). Future studies could use CSF for greater sensitivity.

5. Conclusion

While the specific mechanisms linking microglia depletion to changes in network excitability are not fully understood, our results highlight the potential impact of TNF-α release on synaptic function. Specifically, we introduce the idea that by decreasing pro-inflammatory regulators during specific time points post mTBI, we may be able to promote normal network excitability. Our results also emphasize the need for careful consideration of the broader effects of microglia depletion, such as the infiltration of peripheral immune cells, particularly in healthy brains as demonstrated by the circuit function deficits observed in CA1 as well as MAC2 expression in sham-depleted animals.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.bbi.2025.106178.

Data availability

Data will be made available on request.