Inhibition of p38 MAPK after repetitive mTBI ameliorates immune signaling and functional deficits

Abstract

Repetitive mild traumatic brain injuries (rmTBI) sustained within a window of vulnerability can lead to long-term neurological impairments. Following mechanical impact, increasing evidence suggests that the brain undergoes an inflammatory cascade, leading to chronic neuroinflammation and persistent neurological deficits. While the p38 MAPK signaling pathway is implicated in severe TBI, its role in rmTBI is unclear. This study investigated whether small molecule inhibition of p38 MAPK with SB239063 reduces inflammatory response and functional deficits after repetitive mTBI in a mouse weight-drop model. In females, p38 MAPK inhibition reduced synaptic loss, cytokine upregulation, microglial reactivity, functional deficits, and transcript alterations while upregulating protective pathways. In males, p38 MAPK inhibition attenuated microglial changes and transcript alterations but had limited functional effects. Together, these findings suggest the role of p38 MAPK in driving injury consequences in a sex-dependent manner and highlight therapeutic potential for p38 MAPK inhibition after repetitive mTBI.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40478-026-02226-w.

Introduction

Traumatic brain injuries (TBI) affect over 70 million people each year, resulting in an approximately $400 billion healthcare burden [1]. Among TBI cases, the majority (approximately 80%) are classified as mild (mTBI). An estimated 10–40% of mTBI patients experience significant functional deficits, including headache, blurry vision, and depression, that can persist for months [2–4]. Moreover, repetitive mTBI (rmTBI) sustained within a window of vulnerability after the initial insult can increase susceptibility to injury​, leading to exacerbated long-term deficits [2, 5, 6] and the development of persistent pathological changes, including AD-like pathology [7]. Unfortunately, existing treatments for mTBI/rmTBI primarily focus on addressing symptoms rather than underlying injury mechanisms. New therapeutic strategies that target molecular mechanisms of injury, such as the NFκB and MAPK signaling pathways, are needed.

mTBIs induce an injury cascade comprised of the primary injury, which occurs at the time of impact, and the secondary injury, which unfolds over days-to-weeks and involves many processes including excitotoxicity, mitochondrial dysfunction, altered neurotransmission, and an inflammatory response [8–11]. Among these, the inflammatory response, which consists of microglial and astrocytic reactivity, cytokine production, and proinflammatory signaling cascades, has emerged as a critical process contributing to long-term neurological deficits following mTBI/rmTBI [12–15]. While the benefits of microglial depletion after TBI have received considerable attention [13, 16], phospho-protein signaling pathways represent central and targetable mechanisms of the inflammatory response. These phospho-protein signaling pathways regulate downstream transcriptional factors that drive production of pro-inflammatory molecules and injury pathology [17–20]. Indeed, several studies have identified the essential role of phospho-signaling pathways in driving pathology after severe forms of TBI [12, 21, 22], and similar patterns have been demonstrated in mTBI/rmTBI [6, 23, 24].

Our group and others have identified elevated p38 mitogen-activated protein kinase (MAPK) signaling as being involved in both mild and severe TBI models [11, 12, 25, 26]. This pathway, which has been shown to be activated in microglia after severe TBI, has known roles in altering multiple cellular functions [20], including the modulation of microglial proinflammatory phenotypes [12, 25, 27, 28], cytokine expression [12, 14, 29], synaptic dysfunction [27, 30], neuronal apoptosis [28], and cognitive dysfunction in animal models [27, 28, 31]. Microglial deletion of the p38α MAPK isoform has been shown to exhibit protective effects after controlled cortical injury [12, 27]. Moreover, studies targeting receptors upstream of p38 MAPK, such as TREM1, have been shown to elicit reduced inflammatory responses, edema, and functional deficits [12, 27, 28, 32] following brain injury. While p38 MAPK has been well-studied in the context of severe TBI, we have little understanding of its role in mTBI. Given that mechanisms of severe and mild TBI may be distinct [33] and that our previous work has demonstrated that phosphorylated p38 MAPK co-labels with neurons, not microglia, after rmTBI [11], it is important to elucidate the role of p38 MAPK in the context of rmTBI.

Herein, we investigated the role of p38 MAPK following rmTBI using a closed head weight-drop model. p38 phosphorylation was inhibited with the clinically-relevant small molecule inhibitor, SB239063 (SB), administered 30 min after each injury. We quantified the effects of p38 MAPK inhibition on protein and transcriptional alterations associated with the immune response, synaptic function, microglial phenotype, and functional outcomes. We hypothesized that acute inhibition of p38 MAPK signaling would attenuate injury-induced functional and neurological consequences. We found significant sex disparity in the response to injury with current injury paradigm and in the response to p38 MAPK inhibition, consistent with a previous report using the same injury model [10]. Generally, females had a more striking response to injury in most of the markers quantified. Thus, we focus the results section on the female response, and we summarize similarities and differences between sexes.

Results

Because we observed sex differences in almost all outcome metrics, we report the results for females and males separately. We focus on females in the main text and present all results for males in the supplement.

Inhibition of p38 MAPK attenuates injury-induced synaptic loss post rmTBI

We began our study by examining the effect of p38 MAPK inhibition on postsynaptic density protein 95 (PSD95), a critical regulator of synaptic plasticity and function and an established marker of synaptic integrity [34], following rmTBI. While we observed no changes in PSD95 at 4-h post-injury, pronounced decreases were observed at 4-weeks in females (Fig. 1; Fig. S1-3). Visual inspection of immunohistochemistry (IHC) images demonstrated that the cortical PSD95 labelling was qualitatively reduced at 4-weeks post-injury compared to sham-injured animals and that this effect was ameliorated in five once-daily closed-head injuries (5xCHI) + SB treated mice (Fig. 1A; Fig. S1). Quantitative Western blot assessment of PSD95 revealed a significant reduction in 5xCHI + vehicle mice compared to sham-injured controls (5xCHI + vehicle vs. Sham, p < 0.001) that was absent in 5xCHI + SB mice (5xCHI + SB vs. 5xCHI + vehicle, p = 0.6; Fig. 1B-C, Fig. S2). Together, these data indicate that SB protects against post-synaptic density loss after 5xCHI.

Fig. 1.

Acute inhibition of p38 phosphorylation suppressed 5xCHI-induced synaptic loss 4-weeks post final CHI in females. A. Representative IHC images in the cortex showing PSD95 stain (red) and DAPI (blue) (scale bar: 200 μm, representative sections from n = 5 mice/group; Fig. S1). B. Representative Western Blot from 4 blots for PSD95 and ɑ-Tubulin (Fig. S2). C. Box plot of quantification of Western blot for PSD95 expression normalized to ɑ-Tubulin (n = 8–12/group, mean ± SEM, Wilcoxon rank sum test with Bonferroni adjustment). Each dot denotes an individual animal: blue circles = sham, red triangles = 5xCHI + vehicle, pink squares = 5xCHI + SB. Navy bars denote mean ± SEM. All p-values reflect Wilcoxon rank sum tests with Bonferroni adjustment for multiple comparisons. IHC = immunohistochemistry. PSD95 = post synaptic marker 95

Acute inhibition of p38 MAPK mitigates injury-induced microglial changes post-rmTBI

Because p38 MAPK is known to affect microglial function [12, 19, 25, 27, 28] and existing literature has revealed both acute and prolonged changes in microglia post-injury [15, 35], we next investigated the effect of p38 MAPK inhibition after injury on microglial markers at 4-h and 4-weeks post-injury in the visual cortex.

At 4-h post-injury, ELISA showed that Tmem119 was significantly increased in 5xCHI + vehicle compared to sham-injured female controls (5xCHI + vehicle vs. sham, p = 0.008; Fig. 2A), and this increase was attenuated by SB treatment (5xCHI + SB vs. 5xCHI + vehicle, p = 0.028; Fig. 2A). IHC for Tmem119 showed an increase in Tmem119-positive cells in 5xCHI + vehicle compared to sham-injured controls, as confirmed by cell counts, which was absent in 5xCHI + SB (Fig. 2B-C, Fig. S4). Moreover, in sham-injured controls, Tmem119-positive cells exhibited a ramified morphology indicative of a homeostatic microglial state [36]; whereas in 5xCHI + vehicle treated animals, Tmem119-positive cells demonstrated a dense, amoeboid shape, consistent with reactivity [36]. This morphological change was less pronounced in 5xCHI + SB than in 5xCHI + vehicle mice (Fig. 2B, Fig. S4A). Similar Tmem119 IHC results were observed in both the visual and frontal cortices (Fig. S4B), and similar trend was observed in a subset of samples labeled with Iba1 (Fig. S5). In contrast to these changes in Tmem119, there was no significant change in CD68 at 4-h post-injury (Fig. 2D). Taken together, these results suggest an increase in acute microglial recruitment and reactivity with injury that is attenuated by p38 MAPK inhibition without changes in phagocytic activity.

Fig. 2.

Acute Inhibition of p38 phosphorylation suppressed 5xCHI induced changes in microglial activation at 4-h and 4-weeks post final CHI in females. The top panels (A-D) represent data from tissues collected 4-h post-injury and the bottom panels (E–H) represent data from tissues collected 4-weeks post-injury. A. Box plot of cortical Tmem119 expression at 4-h post-injury (n = 5/group, Wilcoxon rank sum test, the y-axis represents ELISA results in a.u.). B. Representative images of immunohistochemistry for marker of Tmem119 stain (red) and DAPI (blue) (scale bar: 20 μm, representative sections from n = 3–5 mice/group; Fig. S4). C. Box plot of quantification of Tmem119 positive cells counts from immunohistochemistry in panel B and Figure S4. D. Box plot of cortical CD68 expression at 4-h post-injury (n = 5/group, Wilcoxon rank sum test, the y-axis represents ELISA results in a.u.). E. Box plot of cortical Tmem119 expression at 4-weeks post-injury (n = 16–18/group, Wilcoxon rank sum test, the y-axis represents ELISA results in a.u.). F. Representative images of immunohistochemistry for marker of Tmem119 stain (red) and DAPI (blue) (scale bar: 20 μm, representative sections from n = 5 mice/group; Fig. S6). G. Box plot of quantification of Tmem119 positive cells counts from immunohistochemistry in panel B and Figure S6. H. Box plot of cortical CD68 expression at 4-weeks post-injury (n = 16–20/group, Wilcoxon rank sum test, the y-axis represents ELISA results in a.u.). In all plots, each dot denotes an individual animal: blue circles = sham, red triangles = 5xCHI + vehicle, pink squares = 5xCHI + SB, light blue circle from sham group denoted sham + SB; dark blue circle denoted sham + vehicle. Navy blue lines denote mean ± SEM, p-values reflect Wilcoxon rank sum tests with Bonferroni adjustment for multiple comparisons. a.u. = arbitrary units. Tmem119 = transmembrane protein 119. CD68 = cluster of differentiation 68

At 4-weeks post-injury, Tmem119 was significantly decreased in both 5xCHI + vehicle and 5xCHI + SB groups compared to sham-injured female controls (5xCHI + vehicle vs. Sham, 5xCHI + SB vs. Sham; p < 0.01; Fig. 2E). Although IHC showed limited qualitative differences in Tmem119-positive cells between experimental groups, as confirmed by cell counts, similar morphological changes in 5xCHI + vehicle compared to 5xCHI + SB and sham-injured controls found at 4-h post-injury were also observed at 4-weeks post-injury (Fig. 2F-G; Fig. S6). A similar morphological pattern, although less pronounced, was observed in a subset of samples labeled for Iba1 (Fig. S5). Moreover, CD68 was significantly increased in 5xCHI + vehicle compared to sham-injured controls (5xCHI + vehicle vs. sham, p = 0.042; Fig. 2H), and this increase was attenuated by SB treatment (5xCHI + SB vs. 5xCHI + vehicle, p = 0.044, Fig. 2H). Taken together, these results suggest a long-term change in microglial state from homeostatic to phagocytic that is attenuated by p38 MAPK inhibition.

Together, these data indicate that SB administered 30 min after each injury attenuated microglial activation and morphological changes caused by 5xCHI.

Inhibition of p38 MAPK attenuates acute inflammatory response post rmTBI

We next used a Luminex multiplexed immunoassay to examine 32 cytokines in both hippocampal and cortical tissues in females collected 4-h post-injury (Fig. 3A). Statistical significance of all analytes is provided in Table S1 for females and Table S3 for males. Because the effects of injury on cytokines were more pronounced in the hippocampus, we focused our analysis on this region (Fig. 3). Discriminant partial least squares regression (D-PLSR) identified a composite profile of cytokines, latent variable 1 (LV1), that distinguished groups. Latent variable 2 (LV2) showed little separation between groups (Fig. 3B). Scores on LV1 showed that 1) the majority of the 5xCHI + vehicle group clustered separately from the 5xCHI + SB group, and 2) the 5xCHI + SB group and sham-injured controls grouped closely together, indicating that the effect of injury was attenuated by SB treatment (Fig. 3B). LV1 consisted of a weighted profile of cytokines correlated with 5xCHI (positive) or sham (negative) (Fig. 3C).

Fig. 3.

Acute inhibition of p38 phosphorylation suppressed 5xCHI induced changes in cytokine expression 4-h post final CHI in females. A. Multiplexed Luminex analysis of 32 cytokines (columns) expressed in the hippocampus 4-h after 5xCHI and sham-injury. Each row in the z-scored heat map denotes an individual animal from either the sham (bottom rows) or injured (top rows) group. B. Scores plot of discriminant partial least square regression. LV1 explained 42% of the variance, and the LV2 explained an additional 11%. Each dot denotes one sample. C. Discriminant partial least squares regression identified a variable, LV1, that separated samples by experimental group. LV1 consists of a weighted profile of cytokines that were up regulated in either 5xCHI (positive weights) or sham-injured controls (negative weights) (mean ± SD across LV1 generated for all models in a leave one out cross validation). D. Box plots for individual cytokines (mean ± SEM, Wilcoxon rank sum test). In all plots, each dot denotes an individual animal: blue circles = sham, red triangles = 5xCHI + vehicle, pink squares = 5xCHI + SB. Light blue circle from sham group denoted sham + SB; dark blue circle denoted sham + vehicle. Navy blue lines denote mean ± SEM. All p-value reflect Wilcoxon rank sum tests with Bonferroni adjustment for multiple comparisons. a.u. = arbitrary units

Among cytokines with the highest weights on LV1, several exhibited significant elevations with injury that were attenuated by SB treatment. Specifically, vascular endothelial growth factor (VEGF), Eotaxin, Granulocyte–macrophage colony-stimulating factor (GM-CSF), and anti-inflammatory interleukin-13 (IL-13) were all increased in 5xCHI + vehicle compared to sham-injured controls (5xCHI + vehicle vs. Sham, p < 0.05) and decreased in 5xCHI + SB compared to 5xCHI + vehicle mice (5xCHI + vehicle vs. 5xCHI + SB, p < 0.05; Fig. 3D), indicating that injury-induced cytokines were reduced by SB treatment. Similar trends were observed for Interferon-gamma (IFN-γ), Interleukin-1 beta (IL-1β), and Interleukin-6 (IL-6) but did not reach significance (p > 0.05; Table S1). A parallel analysis of cortical tissue revealed less prominent distinction between groups on LV1 compared to hippocampal tissue (Fig. S7; Table S1). Nevertheless, among cytokines with the highest weights on LV1 in cortical tissue, four exhibited significant elevation with injury that was attenuated with SB treatment. Specifically, macrophage inflammatory protein 1β (MIP-1β), a pro-inflammatory molecule that has been implicated in the development of neuroinflammation after brain injury, along with G-CSF, Eotaxin, and IL-5 were all increased in 5xCHI + vehicle compared to sham-injured controls (5xCHI + vehicle vs. Sham; p < 0.05, Fig. S7D), but this increase was absent in 5xCHI + SB (5xCHI + SB vs. Sham p > 0.05, Fig. S7D). Eotaxin was impacted by injury and SB treatment in both hippocampal and cortical tissues.

Together, these data indicate that acute inhibition of p38 MAPK attenuates acute cytokine expression after 5xCHI in females, with more pronounced effects in the hippocampus.

Cytokines impacted by injury and SB treatment co-label with neurons

Having found that VEGF, Eotaxin, IL-13, and GM-CSF were impacted by both injury and SB treatment (Fig. 3), we next asked if these cytokines would be associated with neurons given our previous findings [10, 11]. IHC at 4-h post-injury revealed that VEGF and Eotaxin co-labeled with the neuronal marker NeuN in the cortex and hippocampus in the 5xCHI + vehicle group and that this co-labelling was absent in the 5xCHI + SB group (Fig. 4; Fig. S8). We further noted that the co-labelling between these markers and NeuN was also absent in sham-injured controls (Fig. 4; Fig. S8). Moreover, visual inspection revealed enlarged, diffuse NeuN labeling in 5xCHI + vehicle but defined, smaller NeuN labelling in sham-injured controls and 5xCHI + SB animals, suggesting cellular stress that is attenuated by SB. Collectively, our data suggest that cytokines impacted by injury are closely associated with neurons, and that neuronal p38 MAPK signaling may contribute to post-rmTBI immune signaling, as supported by reduced neuronal co-labeling following SB treatment (Fig. 4; Fig. S8).

Fig. 4.

Inhibition of p38 MAPK attenuated co-labels between cytokines and NeuN 4-h post final CHI in females. Cytokines upregulated with 5xCHI and normalized with SB (Eotaxin, VEGF) (red) co-stained with NeuN (green) and DAPI (blue) show neuronal co-labeling at 4-h post-injury in females (scale bar: 20 μm, representative sections from n = 3–5 mice/group; Fig. S7)

Inhibition of p38 MAPK ameliorated functional changes at 4-weeks post final injury.

Given that functional deficits represent a major consequence of rmTBI, we investigated whether inhibition of p38 MAPK could ameliorate clinically relevant phenotypes for mTBI, including depression [2, 3], motor function [2], and anxiety [4]. TST for depressive-like behavior showed that 5xCHI + vehicle in females had significantly decreased immobility time compared to sham-injured animals (5xCHI + vehicle vs. Sham, p = 0.015) at 4-week post-injury, suggesting an increase in hyperactive behavior (Fig. 5A). This difference was not significant in 5xCHI + SB compared to the sham-injured controls (5xCHI + SB vs. Sham, p = 0.28). Similarly, locomotor deficits accessed by the rotarod showed that 5xCHI + vehicle females had significantly decreased fall time (latency to fall) compared to sham-injured controls (5xCHI + vehicle vs. Sham, p = 0.0067), suggesting a decrease in motor coordination. This difference was not significant in 5xCHI + SB compared to sham-injured controls (Fig. 5B, 5xCHI + SB vs. Sham, p = 0.20). Open field assessment of anxiety-like behavior at 4-weeks post-injury showed that 5xCHI + vehicle females showed a trend toward increased time spent in the open field center compared to sham-injured controls (5xCHI + vehicle vs. Sham, p = 0.12; Fig. 5C), which was absent in the 5xCHI + SB group (5xCHI + SB vs. Sham, p = 0.76). Because variability was observed across functional outcomes, we performed a post hoc power analysis for each measure (Table S2). While most effects of injury (sham vs. 5xCHI) were large, the effect of SB treatment was small-to-moderate, hence results should be interpreted with caution.

Fig. 5.

Inhibition of p38 phosphorylation ameliorated 5xCHI induced functional changes at 4-week post final injury. A. Normalized TST immobility time for all groups. All TST immobility time was normalized by the mean immobility time from the sham group within each cohort (n = 8–12/group). B. Rotarod fall time for all groups. All Rotarod fall time was normalized by the mean fall time from the sham group within each cohort (n = 9–11/group). C. Normalized time spent in open field center for all groups (n = 8–12/group). In all plots: each dot denotes an individual animal: blue circles = sham, red triangles = 5xCHI + vehicle, pink squares = 5xCHI + SB. Navy blue lines denote mean ± SEM, All p-value reflect Wilcoxon rank sum tests with Bonferroni adjustment for multiple comparisons

SB activated protective genes and normalized injury-induced transcriptional changes at 4-weeks post 5xCHI

We concluded by asking if we could identify enduring molecular effects of SB treatment consistent with functional protection. To evaluate these changes, we used bulk RNA-seq to profile changes in gene expression at 4-weeks post-injury. Of the 9,636 genes that remained after filtering (Methods), we identified 170 differentially expressed genes (DEGs) (86 upregulated; 84 downregulated) in the 5xCHI + vehicle group compared to sham-injured controls (Fig. 6A; Data S1). We also identified 291 DEGs (155 upregulated DEGs; 136 downregulated DEGs) in the 5xCHI + SB group compared to sham-injured controls (Fig. 6B; Data S2).

Fig. 6.

SB activated protective genes and normalized injury-induced genes 4-weeks after 5xCHI. A. Volcano plot for differentially expressed genes (DEGs) between 5xCHI + vehicle and sham-injured controls and B. between 5xCHI + SB and sham-injured controls. DEGs have p-values ≤ 0.05 (above dashed horizontal line) and corresponding log2 fold change |log2FC|≥ 0.32. In the volcano plot, each dot represents a single differentially expressed gene (DEG); the x-axis indicates the log₂ fold change of gene expression, and the y-axis represents the –log₁₀(p-value). Venn diagrams highlight SB normalized (left most) and SB regulated (right most) DEGs. The red circle represents DEGs in the 5xCHI + vehicle vs. sham-injured control comparison, the pink circle represents DEGs in the 5xCHI + SB vs. sham-injured control comparison. The overlap of these circles (middle) denotes DEGs unimpacted by SB treatment, the left most unique section (darkest red) denotes SB normalized DEGs, the right most unique section (lightest link) denotes and SB-specific DEGs for C. upregulated and D. downregulated. Top DEGs by highest |log2FC| are listed to the right. E. Enriched Gene Ontology (GO) biological processes in 5xCHI + vehicle vs. 5xCHI + SB comparison. Each dot represents a significantly enriched GO term. The color gradient indicates the adjusted p-value, with darker colors representing greater statistical significance. Dot size corresponds to the number of genes associated with each GO term (gene set size), and the x-axis represents fold enrichment. F. Volcano plot for differentially expressed genes (DEGs) between 5xCHI + vehicle and 5xCHI + SB. DEG = differentially expressed gene. GO = gene ontology

Among the upregulated DEGs in 5xCHI + vehicle animals, 78 were normalized by SB (i.e., 5xCHI + SB was not significant compared to Sham) and 6 were not affected by SB (i.e., 5xCHI + SB was significant compared to Sham). Additionally, 130 DEGs were upregulated specifically by SB (Fig. 6C). Among the top 6 SB-normalized DEGs with the highest log2FC, many were associated with immune-related functions: transcription factors and cytokine regulation (Camk4) [37], innate immune sensing (Polr3f) [38], lymphocyte differentiation (Tcf3) [39], and immune cell development (Etv6) [40]. The normalization of these DEGs by SB suggests that p38 MAPK activates downstream immune signaling post-injury. In contrast, among the top 6 DEGs (greatest log2FC) upregulated in 5xCHI + vehicle that were not normalized by SB, many possessed cell homeostatic functions: signal transduction and infection (Ppm1m) [41], cell structure and regulation of neuronal morphology (Marcksl1) [42], cell migration (Kank2) [43], and protein turnover (Rnf150) [44]. These DEGs represent injury-upregulated homeostatic functions that were not directly impacted by SB treatment. Moreover, the top SB-specific DEGs with the highest log2FC primarily possessed functions related to injury response and cell recovery: immune related functions (Bag6) [45], mitochondrial integrity (Immt) [46], cell metabolism (Insig1 [47], Ggt7 [48]), and oxidative damage control (Prxl2c) [49]. These transcriptomic changes suggested that SB normalized the injury-induced inflammatory response and upregulated pathways associated with cell recovery and maintenance.

Among the downregulated DEGs in 5xCHI + vehicle animals, 79 were normalized by SB (i.e., 5xCHI + SB was not significant compared to Sham) and 7 were not affected by SB (i.e., 5xCHI + SB was also significant compared to Sham). Additionally, 148 DEGs were downregulated specifically by SB (Fig. 6D). The top 6 SB-normalized DEGs with the lowest log2FC were primarily involved in neuronal and cell maintenance: transcription regulation (Cdk19) [50], Nuclear Factor kappa-light-chain-enhancer of activated B cells (NFκB) regulated cell survival (Usp31) [51], neural connectivity (Wapl) [52], ion transport (Fxyd5) [53], and lipid metabolism and neuronal peroxisomes function (Acbd5) [54]. Normalization of these DEGs by SB suggests its role in restoring post-injury cellular functions related to cellular homeostasis and neuronal function, including stress signaling, neuronal connectivity, and metabolic balance. In contrast, the top downregulated DEGs in 5xCHI + vehicle that were not normalized by SB possessed a variety of cell maintenance functions: metabolic regulation (Tbc1d1) [55], hormonal signaling (Inha) [56], intracellular transport (Cep126) [57], and lysosomal function (Lamp2) [58]. Collectively, these DEGs represent injury-downregulated pathways that were not directly impacted by SB treatment. Moreover, most of the top SB-specific downregulated DEGs are involved in injury response: stress response (Stub1 [59], Ermp1 [60]), epigenetic regulation of T cell differentiation (Cxxc1) [61], T cell anergy (Dtx1) [62], and epithelial immune function (Pax6) [63]. Collectively, these gene expression changes are consistent with SB downregulating immune and stress responses, while restoring injury-downregulated overall cellular homeostasis and neuroprotection.

Further, we identified 104 differentially expressed genes (DEGs) (57 upregulated; 47 downregulated) in the 5xCHI + SB group compared to 5xCHI + vehicle group (Fig. 6E; Data S3). To gain further insight into the biological pathways modified by SB and because individual gene changes were relatively modest, we focused interpretation on pathway-level patterns using gene set enrichment analysis (GSEA) [64] to identify biological processes enriched in 5xCHI + SB vs. 5xCHI + vehicle (Fig. 6F; Data S4). This analysis identified significantly enriched gene sets related to the ensheathment of neurons, protein synthesis (cytoplasmic translation, translation at synapses), ATP production (aerobic respiration, oxidative phosphorylation), and ribosomal function (ribosome biogenesis, rRNA processing) enriched in 5xCHI + vehicle compared to 5xCHI + SB. Biological processes related to neurodevelopment (e.g., brain development, neuron projection guidance), synaptic function (regulation of synapse structure or activity, modulation of chemical synaptic transmission), cognitive processes (learning or memory, cognition, behavior), and signaling pathways (regulation of small GTPase-mediated signal transduction, protein autophosphorylation, G protein-coupled receptor signaling pathway) were enriched in 5xCHI + SB compared to 5xCHI + vehicle. These processes are associated with brain recovery, ranging from molecular, cellular, and functional levels. Collectively, these transcriptional changes are consistent with SB attenuating injury-induced prolonged adaptive responses, while promoting neuroprotective pathways at 4-weeks post-injury.

Inhibition of p38 MAPK partially alleviated injury-induced response post rmTBI in male mice

In general, male mice exhibited less pronounced response to injury compared to females, and the male response to SB treatment was less prominent than that of females (Fig. 7).

Fig. 7.

Summary of response to injury and SB treatment in females and male. Summary of responses for female and male mice across major outcome metrics. Injury refers to the comparison between 5xCHI + vehicle and sham-injured controls; SB refers to the comparison between 5xCHI + SB and 5xCHI + vehicle. An upward arrow indicates an increase, a downward arrow indicates a decrease, and Ø indicates little or no change. Grey arrows denote trends that did not reach statistical significance (p > 0.05), while black arrows denote significant changes (p < 0.05)

In terms of PSD95, injury had little effect at 4-h in either sex (Fig. S9). There was, however, a difference between sexes at 4-weeks, where females showed a significant effect of injury that was not present in males (Fig. S10, Fig. 2).

In terms of microglial response, at 4 h post-injury males exhibited more pronounced microglial changes in the hippocampus, while females showed greater effects in the cortex (Fig. S11-S13, Fig. 3A-D). In males, hippocampal CD68 was significantly upregulated in the 5xCHI + vehicle group compared to sham-injured controls and significantly decreased in the 5xCHI + SB group compared to 5xCHI + vehicle, suggesting microglial phagocytic activity is acutely increased with 5xCHI and attenuated by SB (Fig. S12). Qualitatively, at 4-h post-injury, males and females showed similar shifts in microglial morphology from a ramified shape in sham-injured controls to an amoeboid morphology in the 5xCHI + vehicle group, which was ameliorated by SB in both sexes (Fig. S14). In contrast to females, limited microglial changes were observed in males at 4-weeks post injury (Fig. S11-S12; Fig. S15).

In terms of cytokines at 4-h post-injury, the male hippocampal and cortical cytokine profiles demonstrated weak effects of injury (Fig. S16-S17; Table S3). Only IL-17, IL-5, GM-CSF, IL-1ɑ were unregulated in males (5xCHI + vehicle vs. Sham, p < 0.05), and among these only injury-induced IL-17 changes were attenuated with SB treatment (5xCHI + vehicle vs. 5xCHI + SB, p = 0.032; Fig. S16A). Histological examination paralleled the Luminex findings, showing co-labelling between cytokines (VEGF and Eotaxin) with NeuN in all conditions regardless of injury or SB treatment (Fig. S18-S19). Together, these results indicate that males exhibited 1) less pronounced cytokine response to injury and 2) a sex-specific cytokine response to SB treatment.

In terms of functional changes, 5xCHI + vehicle males had a trend toward increased immobility time after injury compared to sham-injured animals (5xCHI + vehicle vs. Sham, p = 0.18), which was absent in the 5xCHI + SB group compared to sham-injured controls. In contrast, the 5xCHI + SB group had significantly higher immobility time than the 5xCHI + vehicle group (5xCHI + vehicle vs. 5xCHI + SB, p = 0.041; Fig. S20A). In terms of rotarod, both 5xCHI + vehicle and 5xCHI + SB treated males had decreased fall time compared to sham-injured controls (5xCHI + vehicle vs. Sham, p = 0.043; 5xCHI + SB vs. Sham, p = 0.071; Fig. S20B). These data indicated limited effect of SB treatment on rotarod, contrasting with female results. In terms of open field, at 4-weeks post-injury, 5xCHI + vehicle males had a trend towards decreased time spent in open field center compared to sham-injured controls (5xCHI + vehicle vs. Sham, p = 0.18; Fig. S20C). Due to the variability observed across functional outcomes, we performed a post hoc power analysis for each measure (Table S4) and revealed moderate effects of injury (sham vs. 5xCHI) and of SB treatment, hence results should be interpreted with caution.

Finally, in terms of transcriptional profile, males exhibited responses to injury and SB treatment that were largely similar to those observed in females. We identified 196 DEGs in 5xCHI + vehicle compared to sham-injured controls (Fig. S21; Data S5). We also identified 129 DEGs in 5xCHI + SB compared to sham-injured controls (Fig. S21; Data S6). SB normalized most of the DEGs changed in 5xCHI + vehicle animals, demonstrating its effect in alleviating injury-induced transcriptional dysregulation. Moreover, 123 DEGs were regulated specifically by SB treatment, including genes associated with promotion of neuronal development and the attenuation of genes associated with stress response. GSEA analysis identified negatively enriched GO terms associated with adaptive responses and positively enriched terms related to cellular homeostasis in 5xCHI + SB compared to 5xCHI + vehicle (Data S7-S8), consistent with findings in females.

Taken together, these data suggest that acute inhibition of p38 MAPK in males partially ameliorates injury-induced changes, including TST immobility scores, acute microglial changes, IL-17 cytokine expression, and transcriptional profiles, while having limited effects on PSD95 loss, longer-term microglial activation, and rotarod fall time.

Discussion

In this study, we evaluated the effects of a translationally relevant p38 MAPK inhibitor administered after repetitive mTBI on a battery of functional and molecular outcomes. We hypothesized that pharmacologic inhibition of p38 MAPK would ameliorate injury-induced functional deficits and normalize changes in microglial and immune markers, synaptic markers, and transcriptional profiles. To our knowledge, this is the first work to evaluate the effects of p38 MAPK inhibition in a mild repetitive model of TBI.

One striking finding from our data is the presence of marked sex-based differences in the response to injury in most of the outcome metrics measured (Fig. 7). For example, our TST data showed that male injured mice had increased immobility after injury compared to sham controls, while females had decreased immobility (Fig. 5; Fig. 7; Fig. S20). This observation is consistent with previous reports of increased depressive-like behavior in males [65] and increased activity in females [66] following weight drop and controlled cortical impact (CCI) injury models, respectively. In terms of microglial response, females showed no change in the phagocytic marker CD68 at 4 h after injury with significant upregulation by 4 weeks (Fig. 2), whereas males exhibited upregulated CD68 at 4-h after injury that resolved by 4-weeks (Fig. S12). Similarly, females showed significant upregulation of the homeostatic marker Tmem119 at 4-h post-injury and downregulation at 4-weeks (Fig. 2), whereas males had no significant changes in Tmem119 (Fig. S12-S14), suggesting a longer-term microglial response in females, consistent with literature generally demonstrating stronger immune response in females [67]. Sex-based differences were also observed in the cytokine response to injury, with females exhibiting more robust changes in cytokine expression at 4-h post final closed-head injury (CHI) (Fig. 3; Fig. S16), consistent with previous reports [10, 67]. These findings demonstrate distinct immune response patterns between the sexes characterized by longer-term microglial response and acute cytokine response in females, versus acute microglial response and minimal cytokine response in males. These differences may arise, in part, from 1) a sex-distinct baseline immune profile, and/or 2) sex-dependent alterations in temporal immune kinetics following repetitive injury, as we previously demonstrated in 3xTg Alzheimer’s mice exposed to rmTBI [10]. We also observed sex-specific differences in the response to SB treatment following injury. For example, SB reduced injury-induced cytokine upregulation in females, but had limited effects in males (Fig. 3; Fig. S16–17). Similarly, SB ameliorated injury-induced rotarod changes in females but had little effect in males (Fig. 5; Fig. S20). One possible source of the sexual dimorphism in the injury and treatment response may be due to the specifics of our experimental design, e.g., the time points that were sampled, innate sex dependence of injury, etc. The importance of each of these parameters should be evaluated in future work.

Our finding that SB protected PSD95 expression 4-weeks post-injury was paralleled with the finding that TST and rotarod function was also protected in females (Fig. 1, Fig. 5). Given that reduction of PSD95 is associated with cognitive deficits across various neurological conditions [34, 68], these findings are consistent with prior work linking MAPK signaling to synaptic health [69, 70]. The neuroprotective effects of SB are bolstered by our RNA-seq analysis at 4-weeks, which showed upregulation of GO terms with SB treatment post-injury associated with synaptic organization, cognitive processes, and neurodevelopment (Fig. 5). Thus, in females our functional, protein, and RNA-seq data collectively support the potential for p38 MAPK inhibition to protect synaptic integrity and reduce cognitive vulnerability after rmTBI.

Another finding of our work is that SB treatment attenuated multiple cytokines that were over expressed in females at 4-h post-injury (Fig. 3). Among these, histological examination suggested both Eotaxin and VEGF co-labeled with the neuronal marker NeuN in 5xCHI + vehicle mice, while SB diminished expression of each of these cytokines in 5xCHI + SB compared to 5xCHI + vehicle animals. These data are consistent with our previous findings of co-labeling between cytokines increased by injury and neuronal marker NeuN [10, 11], and they suggest a role of neurons in facilitating brain immune signaling following rmTBI. Indeed, neurons are known to express various cytokine receptors and activate phospho-protein signaling pathways such as MAPK in response to inflammation [30, 71]. Increasing evidence also suggests that neurons actively contribute to the secretion of cytokines and other immune molecules under both homeostatic and pathological conditions [70, 72–74]. Furthermore, multiple studies have reported evidence of co-localization of immune signaling proteins and/or their transcripts with neurons for the cytokines we observed to be affected by injury and altered with SB treatment, i.e., Eotaxin [75], GM-CSF [76], IL-13 [77], VEGF [78]. Although neuronal cytokine expression is a relatively new area of study, native state proteomic labeling from our group and collaborators suggests that a subset of cytokines may be closely associated with neuronal compartments [79] and necessitates future focused studies to define neuronal signaling proteomics. Given the temporal alignment between cytokine expression and both early microglial recruitment indicated by increased Tmem119 expression, and longer-term microglial reactivity suggested by increased CD68 expression in females (Figs. 2 and 3), it is plausible that the effect of SB treatment on microglial activity may be mediated, at least in part, through modulation of neuronal p38 MAPK signaling. However, these results do not preclude the possibly that microglial p38 MAPK is also altered following rmTBI, which has been previously shown [12].

Our RNA-seq analysis further supports a multifaceted role for p38 MAPK inhibition following rmTBI. Our data suggested that acute SB treatment not only normalized injury-induced gene expression changes but also activated a complementary set of genes associated with cellular maintenance, metabolic support, and neuroprotection. These findings suggest that p38 MAPK inhibition exerts effects beyond the suppression of injury-driven pathways and instead impacting border signaling landscape during post-injury recovery. One plausible mechanism is that transient inhibition of p38 MAPK activity led to altered downstream MAPK signaling, potentially permitting compensatory engagement of parallel pathways, such as JNK and ERK, which have been implicated in neuronal survival [80–82], synaptic plasticity [80–82], repair processes [80], and memory formation [80, 81]. Moreover, prior studies have demonstrated crosstalk between p38 and other MAPK family members, including coordinated regulation via shared phosphatases and reciprocal modulation of activation, further supporting the idea that modulation of one MAPK branch can shape activity in others [83]. Consistent with this interpretation, gene set enrichment analysis revealed coordinated upregulation of pathways related to neuronal structure and protein synthesis in SB-treated injury-animals. Together, these data suggest that transient p38 MAPK inhibition may alter the post-injury transcriptional profile from an adaptive stress response to restoration of brain homeostasis.

Taken together, our findings build on the growing body of literature supporting the role of MAPK signaling in TBI [12, 14, 25–28, 32, 84]. Specifically, p38 MAPK signaling has been associated with neurological outcomes across various moderate to severe injury models, including midline fluid percussion [25], weight drop [27, 32], CCI [12, 26], and lateral fluid percussion [84] injury models. While most existing studies utilize transgenic knockout or broader pathway modulation, our work contributes to this literature by demonstrating the protective effects of p38 MAPK inhibition using a small-molecule in a repetitive mild TBI model. Given the binding specificity, relatively short half-life, scalability and accessibility offered by clinically relevant small molecule inhibitors, pharmacological inhibition of p38 MAPK could offer a pathway-specific, translatable, and effective therapeutic approach for repetitive mTBI. In this study, we administered SB239063 30 min post-injury to enhance translational relevance and to test whether a brief early intervention could yield lasting protection against downstream injury consequences, motivated in part by the transient activation of p38 MAPK signaling following brain injury. Future studies will need to evaluate longer or delayed post-injury dosing paradigms to further enhance clinical relevance, particularly in critical care settings where treatment initiation may occur beyond the acute phase. Related MAPK inhibitors, e.g., Neflamapimod and Losmapimod, are being evaluated in clinical trials for Alzheimer’s disease and muscular dystrophy [69, 85]. To our knowledge, this work is the first to investigate the effects of small molecule inhibitor SB239063 as a potential therapeutic treatment administered post-injury.

The current study has several limitations that motivate future studies. First, we limited our functional assessments to those without a pronounced visual component given the vision loss observed in the injury model. Given the changes we observed in the post synaptic density marker PSD95, future work should investigate the effect of p38 MAPK inhibition specifically on cognitive deficits, possibly by utilizing another rmTBI model that does not elicit vision loss. Indeed, prior reports in more severe models of TBI link MAPK inhibition and cognitive benefits, supporting the premise that p38 MAPK inhibition after rmTBI may improve cognitive outcomes [14, 28, 32, 84]. Second, the changes in PSD95 (Fig. 1) merit future evaluation of electrophysiologic outcomes to fully understand the potential cognitive benefits of SB treatment in the context of brain injury. Third, while our bulk RNA-seq analysis highlights complementary protective effects of p38 MAPK inhibition (Fig. 6), i.e., a downregulation of injury-induced inflammatory response and an upregulation of homeostatic and cell-maintenance genes, we did not perform parallel validation of specific DEGs e.g., with qPCR or protein level immunoassays. Future work should prioritize targeted validation of these transcriptomic findings to strengthen mechanistic interpretation. Moreover, future single-cell RNAseq analysis can be used to better illuminate how neuronal and glial populations respond to rmTBI and SB treatment. Recent work has already begun investigating these cell type specific changes. For example, Zhuang et al. [86] used single-nucleus RNA-seq in a drop-weight mTBI model and found acute oligodendrocyte–neuron disconnection and later noradrenergic hyperconnectivity in the brainstem, motivating similar analyses in cortical tissue in our model. Wu et al. [87] applied single-cell RNA-seq in adolescent weight-drop rmTBI mice and identified disruption of genes involved in neuronal proteostasis and inflammasome activation; these findings motivate the need for understanding cell-type specific transcriptional changes with age. Fourth, a limitation of this study is that estrous-cycle stage was not monitored in female mice. While recent evidence suggests that estrous-cycle related variability has a relatively modest effect on injury outcome [88, 89], some of our tests, e.g., TST, can still be influenced by cycle stage. Future studies can incorporate cycle monitoring to better delineate hormone-dependent effects on behavioral and molecular responses after rmTBI. Finally, this study aimed to identify a clinically-relevant treatment for rmTBI with less emphasis on identifying a cell-type specific mechanism responsible for the neurological consequences post-injury. Thus, our future work will utilize a transgenic approach to interrogate the contribution of p38 MAPK from neurons and glia.

In total, our work characterized the effects of p38 MAPK inhibition after rmTBI considering the key experimental variables of sex and time point post-injury. Our data suggest multidimensional protective effects of SB treatment, particularly in females, that include the alleviation of injury-induced PSD95 loss, attenuation of immune and glial markers, reduction of functional changes, and normalization of injury-induced transcriptional changes alongside modulation of protective pathways that support restoration of brain health, sustained from 4-h to up to 4-weeks post-injury. Notably, the less pronounced effects observed in males likely reflect the reduced severity of injury in male mice, rather than diminished efficacy of SB treatment. Given that most prior p38 MAPK inhibition studies have been conducted only in males or without reporting sex-specific effects [12, 14, 26, 28, 32], our findings highlight a critical gap and motivate future work investigating sex-dependent MAPK signaling in the context of rmTBI. As a small molecule inhibitor, SB presents therapeutic potential for not only rmTBI but also related neurodegenerative conditions. Further, the co-labeling between neurons and cytokines modulated by injury and SB treatment supports the role of neuronal p38 MAPK as an important factor in the pathogenesis of rmTBI. Together, these findings support further investigation of SB in preclinical studies and suggest the pathogenic relevance of neuronal p38 MAPK post-injury.

Materials and methods

Study protocol

All protocols were approved by the Emory University Institutional Animal Care and Use Review Office. Male and female C57BL/6J mice (Jackson Laboratory, strain 0033930) were housed in the university animal facility with a 12-h light/dark cycle. Food and water were provided ad libitum. Mice aged to 2–4 months were randomly assigned to one of two groups: five once-daily closed head injuries (5xCHI) spaced 24-h apart, or five once-daily sham injuries. Within each group, mice were then randomly assigned to treatment or vehicle groups. The treatment group was given the p38 MAPK small molecule inhibitor SB239063 (SB) intraperitoneally (20 mg/kg in physiological saline); the vehicle group was given physiological saline. Given that a prior study using the same injury model reported p-p38 MAPK activation at 30-min post 1xCHI and 3xCHI [90], treatment/vehicle was injected 30 min after each CHI or sham injury to align with the activation window and to enhance translational and clinical applicability. As such, we had a total of 4 experimental groups: 5xCHI + vehicle, 5xCHI + SB, sham + vehicle, sham + SB. Animals were sacrificed by cervical dislocation under 5% isoflurane (1L/min, 100% oxygen) at either a short or long time point. The short time point was 4-h after the final CHI, and the long time point was ~ 1 month after the final CHI (between 18 and 34 days). In the mice who were sacrificed at the long time point, a battery of functional assessments was conducted prior to euthanasia, starting ~ 2 weeks after the final CHI. After euthanasia, brain tissues were collected. Left hemispheres were microdissected into 3 sections: somatomotor cortex, visual cortex, and hippocampus. These sections were flash frozen in liquid nitrogen and stored at -80°C for molecular analysis. Right hemispheres were fixed with 4% paraformaldehyde followed by paraffin processing and embedding for immunohistochemistry analysis.

Closed Head Injury model

Animals were subject to a weight drop closed head injury (CHI) model [91]. For this model, animals were anesthetized using 3–5% isoflurane (1 L/min, 100% oxygen). The anesthetized animal was positioned on a task wipe (Kimwipes®, Kimberly-Clark, Irving, TX), grasped by the base of the tail, and the head was positioned under a guide tube. A 54 g weight was dropped down a 0.96 m guide tube (49035K85, McMaster-Carr, Elmhurst IL) such that impact occurred on the dorsal aspect of the head between the approximate location of the coronal and lambdoid structures (note, the skull was intact for all injuries). On impact, the mouse penetrated the task wipe and underwent rapid, unrestricted rotation of the head in the anterior–posterior plane. Following injury, animals were monitored continuously until they regained consciousness and righting reflex. Neither skull fracture nor hemorrhage were observed in any of the injured animals, consistent with previous research [92]. Sham-injured mice were age- and sex-matched and received the same exposures to anesthesia but were not subject to closed-head injury.

Western blot, ELISA, and luminex multiplexed immunoassays

To investigate the effect of p38 MAPK inhibition on the brain inflammatory response after 5xCHI, we quantified 3 classes of proteins: a molecular marker of pathology (post synaptic density marker 95; PSD95), glial phenotypic markers (microglial activation marker CD68 and homeostatic marker Tmem119), and 32 cytokines. For these analyses, we used a total of 30 samples (n = 5 for 5xCHI + vehicle and 5xCHI + SB per sex, n = 2 for sham + SB per sex, n = 3 for sham + vehicle) collected at 4-h after the final injury and a total of 121 samples (females: n = 16 for 5xCHI + SB, n = 18 for 5xCHI + vehicle, n = 12 for sham + vehicle, and n = 4 for sham + SB; males: n = 21 for 5xCHI + SB, n = 25 for 5xCHI + vehicle, n = 19 for sham + vehicle, and n = 6 for sham + SB) collected at 1-month post final injury. Cortical and hippocampal tissues from the left hemisphere were lysed in 8 M Urea buffer containing Complete Mini tablets (Roche #11836153001) and protein concentrations were determined using a Pierce BCA Protein Assay (Thermo Fisher #23,225).

To quantify a marker of pathological changes, PSD95 expression was quantified via Western blot. After protein concentration was determined, equal amounts of protein were loaded on SDS-PAGE gels, transferred onto a PVDF membrane, and blocked with 5% BSA buffer diluted in 1x TBST for 1-h at room temperature followed by overnight incubation of anti-PSD95 antibody (1:1000; GeneTex) and anti-ɑ-Tubulin antibody (1:10,000; Sigma-Aldrich) at 4°C. On the next day, membranes were incubated with Alexa Fluor-conjugated secondary antibodies (1:2000; ThermoFisher Scientific) for 1-h at room temperature and imaged using a LiCor Odyssey DLx Imager. PSD95 signal intensity was quantified using Image Studio™ Software. For densitometry, we circled a box around the PSD95 band in each lane and normalized to the ɑ-tubulin signal. The final PSD95/ɑ-Tubulin ratio was used for quantification. All antibody information is provided in Table S5.

To quantify microglial markers, macrophage and microglia activation marker cluster of differentiation 68 (CD68) and microglial homeostatic marker transmembrane protein 119 (Tmem119) [36], we utilized an enzyme-linked immunosorbent assay (ELISA). Prior to analysis, lysates were thawed on ice and centrifuged at 4 °C for 10 min at 15,500 g. Protein concentrations were normalized with respective assay diluents from each kit to 0.05 μg/μL in 100 ul for the CD68 and Tmem119 ELISA kits (LifeSpan Biosciences LS-F11095 and LS-F52734). The loaded protein concentrations were selected to fall within the linear range of absorbance vs. protein concentration for detectable analytes. Because whole-tissue lysates can introduce assay interference, analyte quantification was performed using a linear range–based approach rather than reliance on standard curves alone [93]. For each assay, background measurements with assay buffer in the absence of biological samples were subtracted.

To quantify cytokines, we utilized the Milliplex® MAP Mouse Cytokine/Chemokine 32-Plex kit (Eotaxin, G-CSF, GM-CSF, IFN-γ, IL-1α, IL-1β, IL-2, IL-3, IL-4, IL-5, IL-6, IL-7, IL-9, IL-10, IL-12p40, IL-12p70, IL-13, IL-15, IL-17, IP-10, KC, LIF, LIX, MCP-1, M-CSF, MIG, MIP-1α, MIP-1β, MIP-2, RANTES, TNF-α, and VEGF) (Millipore Sigma MCYTMAG-70 K-PX32). Specifically, G-CSF in male hippocampal tissue, MCP-1 and LIX in male cortical tissue did not fall within a linear range of bead fluorescent intensity vs. protein concentration and were therefore excluded in our analysis. Prior to analysis, lysates were thawed on ice and centrifuged at 4 °C for 10 min at 15,500 g. A linear range of bead fluorescent intensity vs. protein concentration was conducted prior to running the Luminex panel for detectable analytes. Protein concentrations were normalized with Milliplex® MAP Assay Buffer (EMD Millipore, Billerica, MA) to 6 μg protein per 12.5 μL.

Immunohistochemistry

To complement the protein analysis, immunohistochemistry (IHC) was conducted to visualize and quantify Tmem119, ionized calcium-binding adaptor molecule 1 (Iba1), and PSD95. For IHC, right brain hemispheres were fixed in 4% paraformaldehyde and then processed and embedded in paraffin. Tissue slices were cut into 10 μm sagittal sections using a rotary microtome and mounted onto glass microscopy slides. Tissue sections were deparaffinized in xylenes, cleaned with 100% ethanol then 95% ethanol, and rehydrated with deionized water. Antigen retrieval was performed with a microwave by boiling slides in 10 mM sodium citrate buffer at pH 6.0. Slides were then rinsed with 1x TBST and then left to dry. A hydrophobic ring was drawn around each individual slice using a PAP pen (Enzo Life Sciences), and then the slices were blocked with 5% BSA buffer diluted in 1x TBST for 4-h. Samples were then incubated with rabbit-anti Tmem119 (1:500; Abcam), rabbit-anti Iba1 (1:500, Abcam), or rabbit-anti PSD95 (1:1000; GeneTex) diluted in blocking buffer overnight at 4 °C. Slides were then rinsed in 1x TBST and incubated with Alexa Fluor 555 (1:200; ThermoFisher) diluted in blocking buffer. Slides were counterstained with 1 μg/mL DAPI and then rinsed in water. Slides were then mounted with 10% glycerol in Phosphate-Buffered Saline (PBS). Samples were imaged using epifluorescent microscopy on a Zeiss Axio Observer Z.1 inverted microscope with a 40 × lens and halogen bulb illumination using Zeiss filter set 49 to image DAPI and Zeiss filter set 20 to image Alexa Flour 555. Images were processed using ImageJ.

IHC was also conducted using a serial staining approach to co-label cytokines VEGF and Eotaxin with the neuronal marker NeuN. Note, the quality of NeuN stain in the nuclei may be compromised due to artifacts of the drop-fixation protocol, as we have found in previous publication [10, 94]. For the IHC, a protocol was utilized to prevent overlapping of different antibodies on a slice at the same time. This protocol was identical to the ones discussed above to stain for Tmem119 and PSD95 up until the addition of the primary antibody. After the slices were blocked with 5% BSA buffer diluted in 1x TBST for 4-h, the slides were incubated with chicken-anti NeuN (1:800; GeneTex) diluted in blocking buffer overnight at 4 °C. Slides were then rinsed in 1x TBST and incubated with Alexa Fluor 488 (1:200; ThermoFisher) diluted in blocking buffer. Slides were then rinsed with 1 × TBST and blocked again with 5% BSA buffer diluted in 1x TBST for 4-h. After blocking for the second time, slides were then incubated with either goat-anti Eotaxin (1:100; R&D Systems) or mouse-anti VEGF (1:100; ThermoFisher) diluted in blocking buffer overnight at 4 °C. Slides were then rinsed in 1xTBST and incubated with Alexa Fluor 647 (1:200; ThermoFisher) diluted in blocking buffer. Slides were counterstained with 1 μg/mL DAPI and then rinsed in water. Slides were then mounted with 10% glycerol in PBS. Samples were imaged using epifluorescent microscopy on a Zeiss Axio Observer Z.1 inverted microscope with a 40 × lens and halogen bulb illumination using Zeiss filter set 49 to image DAPI, Zeiss filter set 38 to image Alexa Fluor 488, and Zeiss filter set 50 to image Alexa Flour 647. Images were processed using ImageJ.

Functional assays

A battery of functional assays consisting of four testing paradigms was conducted over 2 weeks, starting 2–4 weeks after the last closed-head injury (CHI)/sham-injury (n = 12–15/group; Table S4). Because the injury model induces vision impairments (Fig. S22) [95, 96], we limited functional outcomes to those that are less vision dependent. The testing paradigms consisted of assessments for visual acuity and contrast, antidepressive-like behavior, locomotion, and anxiety-like behavior. For each paradigm, mice were placed in the testing room at least 1 h prior to testing to acclimate to the environment.

Visual acuity and contrast: To assess visual function, we measured the optomotor response (OMR) [97] using an OptoMotry vision tracking system (Cerebral-Mechanics, Lethbridge, AB, Canada). Awake mice were placed on a small, elevated platform (5.3 cm diameter) in the center of the 39 × 39 × 32.5 cm (L x W x H) testing chamber. The 4 walls of the chamber were computer monitors that displayed a black and white vertical sine wave grating that was rotated across the screens at 12 degree/second. A video camera placed above the platform was used to record head movements during the experiment. To determine visual acuity, vertical bands were displayed at 100% contrast starting at 0.067 cycles/degree, and the spatial frequency was adjusted following a staircase paradigm until the mouse no longer displayed reflexive head movement. Visual acuity was defined as the highest spatial frequency at which the reflex was observed. To determine contrast sensitivity, the vertical bands were displayed at a spatial frequency of 0.067 cycles/degree and the contrast was adjusted from 100% following a staircase paradigm until the animal no longer displayed reflexive head movement. Contract sensitivity was defined as the lowest contrast in which head movements were observed and was reported as the reciprocal of the Michelson contrast, which accounts for the screen’s luminance [97]. For each mouse, gratings were rotated both clockwise and counterclockwise to separately stimulate the responses of the left and right eyes, respectively [97]. Results are reported as the average across the right and left eyes.

Antidepressive-like behavior: To assess antidepressive-like behavior, mice underwent a tail suspension test (TST). For TST, mice were suspended for 6 min at a height of 0.5 m using a piece of medical grade tape placed approximately 1 cm from the tip of the tail [98]. Immobility time was defined as the total time the mouse spent in the following states: passive hanging, lack of body motion, lack of tail climbing motion, pendulum motion due to previous mobility, or motion of only the hind or forward limbs. Immobility time was quantified by two blinded researchers; a third researcher reviewed periods of disagreement between researchers via an in-house designed graphical user interface (MATLAB, Mathworks) to arrive at a final immobility score. Higher immobility time is considered depression-like behavior. The testing apparatus was thoroughly cleaned with 70% ethanol between mice.

Locomotion: To assess locomotor ability, mice were tested using a Rotamex-5 apparatus (0254-2002L, Columbus Instruments, Columbus, OH) with a 3 cm diameter rod and 44.5 cm fall height from the rod center. Mice were habituated to the rotarod instrument by allowing them to explore the rod at rest for 2-min. During this habituation period, mice were placed back on the rod if they fell off. Mice then completed three 3-min trials at constant rotation of 4.0 cm/s (25.46 rpm). The latency to fall within the 3-min trial time was recorded. At least 5 min rest was provided between trials. The final latency time was calculated as the mean latency of all 3 trials. The rotarod instrument was thoroughly cleaned with 70% ethanol between each mouse.

Anxiety-like behavior: To assess anxiety-like behavior, mice were tested in an open field arena. Mice were allowed to explore the 40 × 40 cm arena for 10 min while their movements were recorded from above. Software (EthoVision, Noldus Information Technology) was used to quantify the total time spent in and total entries into the 20 cm × 20 cm center zone. The arena was thoroughly cleaned with 70% ethanol between each mouse.

Adjustment for batch effects

To account for cohort variances observed in measurements of CD68, TMEM119, and all functional assessments, collected data were normalized to the mean score of sham-injured animals within each batch.

Bulk tissue RNA sequencing and analysis

We conducted RNA sequencing (RNA-seq) on a randomly selected representative subset of 37 female and 24 male samples collected 4-weeks post-injury. RNA from somatosensory cortex was extracted using the miRNeasy Micro kit (Qiagen #217,084). Extracted RNA was sent to Admera Health, LLC (South Plainfield, NJ) for sequencing, alignment, and calculation of the count matrix. The count matrix yielded 55,488 non-zero analytes. We filtered out transcripts with fewer than 200 counts in at least 5% of the samples, leaving 10,077 transcripts. Unrecognized genes in org.Mm.eg.db and predicted genes (e.g., Gm10033) were filtered out, leaving 9,636 transcripts. The remaining transcripts were normalized by the ratio of median method in R using the DESeq2 package, available on Bioconductor [99]. Outlier detection was conducted in R by calculating Mahalanobis distance of each point from the data’s centroid within each grouping of sex and conditions using the ClassDiscovery package [100]. RNA data from 3 female samples and one male sample were discarded because they exceeded the cutoff threshold of α = 0.001. To account for batch variances and RNA integrity number (RIN) variance observed in bulk-tissue RNA-seq analysis, the limma::RemoveBatchEffect function in R was utilized to conduct computations.

The Wilcoxon rank-sum test was performed using the coin package in R to identify genes that were significantly upregulated or downregulated between groups. Wilcoxon rank-sum test was performed here instead of DESeq2 because DESeq2 yields less conservative adjusted p-values in small cohorts. The threshold for differentially expressed genes (DEGs) was set to p ≤ 0.05, log2FC ≥ 0.32 (upregulated) and log2FC ≤ -0.32 (downregulated) to allow for detection of subtle changes and patterns. Following the Wilcoxon rank-sum test, volcano plots were generated using EnhancedVolcano package in R. Venn diagrams were generated using ggplot2, ggvenn, and ggpubr packages in R. All plots were modified and/or created using Illustrator.

Gene Set Enrichment Analysis (GSEA) was performed using the clusterProfiler package in R to identify Gene Ontology (GO) biological processes that were significantly enriched in either group [64]. The appropriate background was generated by using the genes that remained post-filtering. Permutation test, with the Benjamini–Hochberg false discovery rate (FDR) correction identified statistically significant enriched terms (FDR adjusted p ≤ 0.05). The ggplot2 package in R was used to generate dot plots to visualize enriched GO biological processes.

The gene expression FASTQ files and count matrix associated with this study have been deposited in the Gene Expression Omnibus (GEO).

Statistical analysis

Data were analyzed and figures were generated via RStudio (Boston, MA) using the R programming language. Data processing was conducted using the tidyverse collection of packages. Heatmaps were generated with the R package heatmap3; bar graphs and dot plots were generated using the packages ggplot2, ggpubr, ggbeeswarm, and ggprism. Clustering was conducted using the hclust function of the stats package in R using Euclidean distance in the unweighted pair group method with arithmetic mean. For functional and protein analysis, Bonferroni adjustment was applied to correct for multiple comparisons. If no difference was detected by Wilcoxon-rank sum test between sham + SB and sham + vehicle groups, we merged the groups to increase the N in our sham group. Post hoc power analysis was conducted in R for all functional outcome measures given the substantial variability observed.

Multivariate cytokine data were analyzed via discriminant partial least squares regression (D-PLSR) analysis to identify axes, called latent variables (LVs), which consist of profiles of cytokines that separate samples based on discrete variables (e.g., sham vs. 5xCHI). D-PLSR analyses were performed with the R package ropls. Data were z-scored prior to inputting into the algorithm. Error bars for LV analyte weightings were calculated by iteratively excluding samples 100 times and regenerating the D-PLSR model in each run. The mean and standard deviation (SD) of the analyte weightings across these runs were used to generate error bars to provide an indication of the variability within each cytokine among the D-PLSR model.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

CL, SET, and MNG conducted data analysis, prepared figures, and wrote the manuscript. SET, AMR, JPC, PVM, PFS, PIS, AH, AG, and CL conducted animal experiments and curated animal data. CL, MNG, AES, EP conducted molecular assays and experimentation. CL, LBW, EMB contributed to transcriptional profiling design, computational analysis, and biological interpretation. LBW and EMB conceived the study and supervised the research. MNG, SET, LBW and EMB and revised the manuscript. All authors reviewed the manuscript.

Funding

This work was supported by the National Institutes of Health under Award No. 1 R01 NS115994 (LBW/EMB) and by support from the George W. Woodruff School of Mechanical Engineering Faculty Fellowship at Georgia Tech (LBW).

Data availability

The gene expression FASTQ files and count matrix that support the findings of this study have been deposited in the Gene Expression Omnibus (GEO) repository under series record GSE298240.

Declarations

Competing interests

The authors declare no competing interests.