Impaired Cortical Neuronal Homeostasis and Cognition after Diffuse Traumatic Brain Injury Are Dependent on Microglia and Type I Interferon Responses

Abstract

Neuropsychiatric complications including depression and cognitive decline develop in the years after traumatic brain injury (TBI), negatively affecting quality of life. Microglial and type 1 interferon (IFN-I) responses are associated with the transition from acute to chronic neuroinflammation after diffuse TBI in mice. Thus, the purpose of this study was to determine if impaired neuronal homeostasis and increased IFN-I responses intersected after TBI to cause cognitive impairment. Here, the RNA profile of neurons and microglia after TBI (single nucleus RNA-sequencing) with or without microglia depletion (CSF1R antagonist) was assessed 7 dpi. There was a TBI-dependent suppression of cortical neuronal homeostasis with reductions in CREB signaling, synaptogenesis, and synaptic migration and increases in RhoGDI and PTEN signaling (Ingenuity Pathway Analysis). Microglial depletion reversed 50% of TBI-induced gene changes in cortical neurons depending on subtype. Moreover, the microglial RNA signature 7 dpi was associated with increased stimulator of interferon genes (STING) activation and IFN-I responses. Therefore, we sought to reduce IFN-I signaling after TBI using STING knockout mice and a STING antagonist, chloroquine (CQ). TBI-associated cognitive deficits in novel object location and recognition (NOL/NOR) tasks at 7 and 30 dpi were STING dependent. In addition, TBI-induced STING expression, microglial morphological restructuring, inflammatory (Tnf, Cd68, Ccl2) and IFN-related (Irf3, Irf7, Ifi27) gene expression in the cortex were attenuated in STING^KO mice. CQ also reversed TBI-induced cognitive deficits and reduced TBI-induced inflammatory (Tnf, Cd68, Ccl2) and IFN (Irf7, Sting) cortical gene expression. Collectively, reducing IFN-I signaling after TBI with STING-dependent interventions attenuated the prolonged microglial activation and cognitive impairment.

Graphical Abstract

• TBI suppressed neuronal homeostasis while increasing IFN-I gene expression in microglia.

• Microglial depletion reversed 50% of TBI-induced neuronal gene changes.

• STING-inhibition reversed inflammation, microglial priming, and cognitive deficits.

Introduction:

Traumatic brain injury (TBI) increases the prevalence of neuropsychiatric illness, with 15–30% of TBI patients experiencing cognitive decline, often comorbid with depression [1–7]. Furthermore, TBI correlates with an increased risk of neurodegenerative diseases including chronic traumatic encephalopathy (CTE) and Alzheimer’s disease [8–13]. In clinical studies of TBI, elevated metabolic activity, white matter abnormalities, and reactive microglia were detectable months to years after injury [14–16]. Functional magnetic resonance imaging (fMRI) analyses showed chronically activated microglia/macrophages in former NFL players with a history of concussions [17] with inflammation detectable prior to cognitive impairment [18]. Positron emission tomography (PET) studies revealed microglial activation in the thalamus was associated with lower cognitive processing speed [9, 15]. Thus, it is plausible that chronic neuroinflammation and microglia dysfunction underlie these complications.

Mounting evidence indicates that chronic inflammatory processes persist after TBI (diffuse and focal) in rodents [19–21]. These inflammatory processes ultimately affect neuronal homeostasis and physiology [21]. One pathway that could be key in the transition from acute to chronic inflammation and neuronal imbalance is the type I interferon (IFN-I) response. IFN-I activation in the brain is associated with immune responses to viral pathogens [22] and they increase the anti-viral properties of immune cells [23]. Interferons bind to IFNAR1, and receptor activation drives Stat1, Stat2, and Irf9 signaling to control gene transcription associated with anti-viral responses. Nonetheless, IFN-I pathways are evident after TBI in the absence of a virus. For example, Ifnβ mRNA levels were higher in individuals who died shortly after TBI [24]. This increased IFN-I signaling may augment inflammatory responses and decrease functional outcomes following TBI.

In rodent studies of TBI, the sub-acute phase (7 dpi) after diffuse TBI was dominated by a robust IFN-I response [21, 25, 26]. Specifically, single-cell RNA sequencing (scRNA-seq) showed enriched IFN-I responses (Irf3, Irf5, Irf8, Tbk1, IFNAR2, Stat1) within trauma associated microglia 7 dpi [21]. Another study showed sorted microglia and astrocyte profiles enriched in IFN response at 7 dpi in lateral fluid percussion injury (FPI) [26]. In addition, aged mice after TBI had amplified IFN-I responses and enhanced gliosis and neuroinflammation in the cortex [27, 28]. Sustained depletion of microglia using a CSF1R antagonist PLX5622 prior to a TBI attenuated neuroinflammation and ablated the IFN-I response. These reductions corresponded with improved cortical dendritic complexity, neuronal physiology, and prevented cognitive impairment [21]. Forced turnover of microglia at 7 dpi, when IFN-I signaling was elevated in the cortex, also reduced cortical inflammation, cognitive impairment, and microglial priming 30 dpi. [29]. Microglial priming represents an increase in the inflammatory profile of microglia with a corresponding enhanced response to a secondary immune challenge [30]. Anti-viral gene expression induced by IFN-I has a significant overlap with proteins that are chronically expressed on microglia after TBI (30 dpi) including MHC II-related (H2-eb1), dendritic cell-related (Itgax), and phagocytic-related genes (Cd68) [31]. Thus, the intersection between microglia and IFN-I signaling following TBI may be critical in the transition from acute inflammation to chronic microglial priming.

A key upstream regulator of IFN-I is the stimulator of IFN genes (STING). STING is a stress-responsive endoplasmic reticulum protein responsible for inducing IFN-I genes. In the context of viral infection or injury, tissue damage increases cytosolic double-stranded DNA (dsDNA), and mitochondrial DNA (mtDNA) that are sensed by the cGAS-STING pathway [32, 33]. Activation of the STING pathway precedes IFN-I release [34] and, in the brain, is predominately localized to microglia [32]. STING activation promotes IFN-I responses that enhance the transcription factors IRF3 and NF-κB [35, 36] to promote a diverse array of IFN-I and NF-κB-mediated responses. Increased STING was detected in human brain samples 3 hours after fatal brain injury [37]. Notably, this pathway is inhibited by chloroquine (CQ), an FDA approved anti-viral medication capable of crossing the blood brain barrier (BBB) [38]. Taken together, the IFN-I pathway may represent a viable target for reducing neuroinflammation following TBI.

Based on these data, we surmise that the IFN-I pathway is key in the transition from the acute response to TBI to chronic inflammation, neuronal dysfunction, and microglia priming. Thus, the purpose of this study was to determine if impaired neuronal homeostasis and increased IFN-I responses intersected after TBI to cause cognitive impairment. As such, we targeted the IFN-I pathway genetically (STING^KO mice) or pharmacologically (CQ administration).

Materials and Methods:

Mice:

Adult (2–4 month-old) male C57BL/6 mice were purchased from Charles River Laboratories (Wilmington, MA). B6(CG)-Sting1^tm1.2Camb/J; Stock No: 025805 were purchased from the Jackson Laboratory and then bred in-house. Mice were group housed under a 12/12 light-dark cycle with ad libitum access to food and water. Mice were randomly assigned to groups with mixed treatment and injury groups in each cage. All procedures were performed in accordance with the National Institute of Health Guidelines for the Care and Use of Laboratory Animals and the Public Health Service’s Policy on Human Care and Use of Laboratory Animals and the Guide for the Care and Use of Laboratory Animals were approved by The Ohio State University Institutional Laboratory Animal Care and Use Committee.

Midline Fluid Percussion Injury (mFPI):

Mice were subjected to a midline diffuse TBI using a fluid percussion injury (FPI) apparatus as described [21, 25, 29, 39]. Briefly, mice were anesthetized in an isoflurane chamber at 2–3% with a flow rate of 0.8 liters/min. After the surgical site was shaved, mice were secured to the stereotax (Stoelting Co., Cat# 51731) and maintained under anesthesia with a mask attachment (Stoelting Co., Cat# 51609M). The surgical site was prepared with aseptic technique, using alternating applications of iodine and 70% ethanol. Mice received a 3mm craniectomy between the landmark sutures bregma and λ, and a rigid Luer-loc needle hub was secured over the craniectomy site. Following this procedure, mice were moved to a heated (37 ºC) recovery cage and monitored until fully conscious (upright, responsive, and walking). After recovery, mice were briefly re-anesthetized in an isoflurane chamber at 5% (flow rate 0.8 liters/min) for 5 min. The Luer-loc hub was filled with saline, and the hub was attached to the injury device. Once a positive toe-pinch response was elicited (~30 s), a 10 ms pulse of saline (1.2 atm; 670–720 mV) was imposed on the dura. Immediately after injury, the hub was removed, dural integrity was confirmed, and mice were moved to recovery in a heated cage. In these studies, control mice were naïve and uninjured.

Post-Op Care:

Mice with TBI were monitored for 1 h post-injury, then allowed to recover overnight in a heated recovery cage with accessible food and hydrogel. The next day, mice were returned to their home cages. Based on our experimental design, no analgesics were provided in these studies. Mice were weighed and checked for signs of lethargy (lack of movement) and infection (redness and pus around the incision site) daily throughout the experiments (7 or 30 days). Removal criteria included a loss of 20% of baseline bodyweight, sustained lethargy, paralysis, or surgical site infection. In the current study, no TBI-injured mice met this exclusion criteria. Wound clips (7 mm) were used to close the incision site and were removed between 10–14 dpi.

Plexxikon (PLX) 5622 Administration:

PLX5622 was provided by Plexxikon Inc. (Berkley, CA) and formulated in AIN-76A rodent chow by Research Diets Inc. at a concentration of 1,200 mg/kg. Standard AIN-76A diet was provided as a vehicle control. Mice were provided ad libitum access to PLX5622 or vehicle diet for 7 days to deplete microglia prior to TBI. Mice were maintained on the experimental diets for the duration of the study. This dose and time was validated in our lab to deplete ~96% of microglia [21].

Chloroquine (CQ) Administration:

Chloroquine is an inhibitor of the cGAS-STING pathway [40]. Chloroquine (80mg/kg in sterile saline; Cayman Chemical, Cat#14194) was administered intraperitoneally (i.p.) 1h following TBI and every 24 hours post-injury up to 6 dpi.

Nuclei Isolation:

For single nucleus RNA-sequencing, each group (n=3) was sacrificed simultaneously, then pooled. Cortices were extracted then placed into 2mL Dounce homogenizers with 1 mL of homogenization buffer consisting of 1.5 M NIM1 Buffer (250 mM sucrose, 25 mM KCl, 5 mM MgCl2, 10 mM Tris Buffer pH 8 in MilliQ water), 1uM DTT, 0.4 U/uL Enzymatics RNAase-Inhibitor (#Y9240L), 0.2 U/uL Superase-Inhibitor (Thermo Fisher Scientific, #AM2694), and .1% Triton X-100 (Sigma Aldrich, #9036–19-5). Cortices were homogenized, filtered using a 40 μM strainer and homogenates were clarified. Samples were resuspended in a PBS buffer with RNase Inhibitors (0.05U/ul of Enzymatics RNAase-Inhibitor and Superase-Inhibitor) and re-pelleted. To remove myelin debris, samples were incubated with Myelin Removal Beads II (Miltenyi Biotec, #130–096-731) for 15 minutes at 4°C. Samples were washed (50% PBS and 50% PBS + 1% BSA) and re-pelleted. Supernatant was removed and samples were resuspended in 1 mL of wash buffer. One LS column (Miltenyi Biotec, #130–042-401) was used to filter each of the samples, which were then pelleted and resuspended in 1 mL of wash buffer. Nuclei were counted with AO/PI (Logos Biosystems, #F23001) on a Luna-FL Cell Counter and fixed with a Nuclei Fixation Kit (Parse Biosciences, #SB1003) per the manufacturer’s instructions followed by rapid freezing at −80°C.

Single-Nuclei Barcoding and Sub-library Generation:

The Parse Biosciences Whole Transcription Kit was used to barcode and generate eight separate sub-libraries with 12,500 nuclei per sub-library according to the manufacturer’s instructions. This kit is optimized for a maximum of 100,000 nuclei. Initial quality control experiments were completed to determine the number of nuclei necessary for barcoding. Based on these results, three mice were pooled together (per experimental group) to obtain the required final number of nuclei for the snRNA sequencing. DNA concentration was measured by Qubit 4 Fluorometer and a Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, #Q32851). A Bioanalyzer 2100 with a High Sensitivity DNA Assay chip was used to control quality of sub-libraries before samples were sequenced. A portion of the RNA libraries was sequenced using to saturation to determine number of nuclei and read depth required for optimal data output. Based on our saturation curve, RNA was sequenced at a depth of 44,000 reads per nuclei using a NovaSeq S4 at the Advanced Genomics Core at The University of Michigan. The experiment design with pooling and replicate numbers for snRNAseq is consistent with previous published studies (Gage et al., 2016, Schmitt et al., 2022, Qu et al., 2022, Raamsdonk et al., 2023, Zhang et al., 2023).

Single-Nuclei Sequencing Data Processing:

Each fastq.gz file was downloaded and aligned to the Genome Reference Consortium Mouse Reference 39 (mm39) using the Parse Biosciences pipeline. Matrices were downloaded and manually filtered in RStudio using Seurat (v4.1.1) Low-quality nuclei and doublets were filtered using Seurat in R. Cell-type identification was done using previously established markers: endothelial cells (Flt1), astrocytes (Slc1a3), oligodendrocytes (Mag), microglia (Csf1r), and neurons (Syt1). Differential gene expression was performed using the FindMarkers feature of Seurat with non-parametric Wilcoxon rank sum test. Pathway and master regulators analyses were performed with Ingenuity Pathway Analysis (IPA; Qiagen).

Immunohistochemistry and Analysis:

Mice were perfused with phosphate buffered saline (PBS) followed by 4% PFA. Brains were removed, post-fixed and dehydrated in 30% sucrose. Coronal sections (30 μm) were collected, washed, blocked, (0.1% Triton X, 5% BSA, and 5% NDS) and incubated with primary antibodies for anti-IBA1, anti-GFAP, or anti-STING (rabbit anti-IBA1, 1:1000, Wako Cat#019–19471, RRID:AB_2665520; goat-anti-GFAP, 1:500, Abcam Cat#ab53554, RRID:AB_880202, rabbit anti-STING, 1:200, Proteintech, #19851–1-AP, RRID:AB_10665370). Next, sections were washed and incubated with an appropriate fluorochrome-conjugated secondary antibody (donkey anti-rabbit, or anti-goat; AlexaFluor 488/594/647; Invitrogen) then mounted and cover-slipped with Fluoromount (Beckman Coulter, Inc., Fullerton, CA). Fluorescent labeling was imaged using an EVOS FL Auto 2 imaging system (Thermo Fisher, Waltham, MA). To determine percent area labeling of IBA1⁺, GFAP⁺, or STING⁺ single channel images were converted to 8-bit TIFF format and constant thresholds were used to quantify positively labeled pixels (ImageJ Software). Rod morphology of Iba1⁺ cells were quantified based on length-to-width ratios as previously described [25, 41]. Values from 4–6 images per mouse were averaged and used to calculate group averages and variance from each group. Images were analyzed by an investigator blinded to treatment groups.

Percoll Enrichment of Brain Myeloid Cells:

CD11b⁺ cells were enriched from whole brain homogenates as described [42–44] with minor modifications. In brief, brains were manually homogenized using Potter homogenizers, and resulting homogenates were centrifuged at 600g for 6 min. Supernatants were removed and cell pellets were resuspended in 70% isotonic Percoll (GE-Healthcare, Catalog #45–001-747). A discontinuous isotonic Percoll density gradient was then layered as follows: 50%, 35%, and 0% (PBS). Samples were centrifuged for 20 min at 2000 g, and cells were collected from the interphase between the 70% and 50% Percoll layers. These cells were referred to as enriched brain CD11b⁺ cells based on previous studies demonstrating that viable cells isolated by Percoll density gradient yields 90% CD11b⁺ cells [43].

RNA Extraction and qPCR:

Microdissected cortices were flash-frozen and RNA was later extracted using TriReagent (Sigma-Aldrich, Cat# T9424), normalized by concentration, and reverse-transcribed to cDNA (High Cap RT Kit, Cat# 4368813; Thermo Fisher). Percoll-enriched myeloid cells were lysed, stored at −80°C, and total RNA was extracted using the Picopure RNA Isolation Kit (ThermoFisher, KIT0204). Target genes (Tmem173 (Sting), Mm01158117_m1; Ifi27a, Mm00835449_g1; Irf3, Mm00516784_m1; Irf7, Mm00516791_g1, Ccl2, Mm00441242_m1, H2-eb1, Mm00439221_m1, Cd68, Mm03047343_m1, Tnf Mm00443258_m1, Clec7a Mm01183349_m1, and Itgax Mm00489701_m1) and reference gene (Gapdh, Mm99999915_g1) expression was measured using a QuantStudio 6 (Thermo Fisher) and data were analyzed using the comparative threshold method (ΔΔCt) with data expressed as fold-change from control.

Novel Object Recognition (NOR) and Location (NOL):

Novel object recognition (NOR) and novel object recognition (NOL) tasks were conducted as previously described [21, 29]. Briefly, these tests involved four 10 min phases each separated by 24h: habituation (no objects), acclimation (2 objects), recognition (2 objects, with one new object), and location (2 objects, one new location). Discrimination index in the recognition and location trials was determined [(time_novel-time_familiar)/time_total] x100 [29, 45]. Videos were analyzed by an investigator blinded to treatment groups.

Statistical Analysis:

GraphPad Prism (Version 9; San Diego, CA) was used for analysis of variance (ANOVA) of histological and behavioral data. One-way or two-way ANOVA was used as appropriate to determine main effects and interactions between factors. Tukey’s test for multiple comparisons was used for post-hoc analysis when main effects and/or interactions were determined. p<0.05 was considered statistically significant. Statistical analysis for snRNA-sequencing using Seurat are described above. Outlier data values that were more than two standard deviations from the mean were excluded.

Results:

RNA profiles of cells in the cortex 7 days after TBI.

Our previous reports indicate that microglia and IFN-I responses are associated with the transition from acute to chronic inflammation in the cortex after moderate diffuse TBI in mice [21, 25]. Thus, this study aimed to determine the degree to which TBI-induced neuronal profiles and IFN-I responses intersected to cause cognitive impairment. In the first experiment, the RNA profiles of neurons and microglia after TBI (snRNA-seq, 7 dpi) with or without microglia depletion (CSF1R antagonist, PLX5622) were assessed. snRNA-seq captures neurons and other cell types that are difficult to isolate at a higher ratio than traditional scRNA-seq [46]. snRNA-seq was used to resolve neuronal RNA profiles more selectively [47].

Here, microglia were depleted prior to midline fluid percussion (TBI) using a CSF1R antagonist, PLX5622 (Fig.1A). Mice were maintained on vehicle or PLX diet for the duration of the experiment. Cortical nuclei were isolated, fixed, and barcoded at 7 dpi. snRNA-seq was performed on nuclei from the cortex. [47]. Fig.1B shows that there were 26 clusters of cortical cells. The influence of the four treatment groups (Veh-Control, PLX-Control, Veh-TBI and PLX-TBI) on cortical clusters is shown (Fig.1C). Cell-specific gene expression for each cluster is listed in Fig.1D and is shown in Fig.1E. The percentage of nuclei based on cell type is represented in Fig.1F . Overall, this snRNA-seq approach provided a high resolution of nucleus RNA profiles for cortical neurons compared to the other cell types.

Fig. 1. RNA profiles of cells in the cortex 7 days after TBI.

A) Adult C57BL/6 mice were provided diets formulated with either vehicle (Veh) or PLX5622 (PLX) for 7 d. Mice were uninjured (Con) or were subjected to mFPI (TBI). Mice were maintained on vehicle or PLX diet for the duration of the experiment (14 d). At 7 dpi, the cortex was microdissected, pooled (3 mice per group) and nuclei were collected. Nucleus RNA profiles were determined 7 dpi by snRNA-seq. B) UMAP plots show 25 specific subsets of cortical cells based on expression of cell-specific genes. C) UMAP plot with the distribution of cells based on the four treatments groups at 7 dpi. D) Cortical nuclei clusters were identified based on cell specific gene expression: upper layer neurons (Cux1 &Cux2), layer 4 neurons (Rorb), deep layer neurons (Foxp2), excitatory neurons (Slc17a7), inhibitory neurons (Gad1&2), oligodendrocytes (Mag), endothelia (Flt1), microglia (Csf1r), and astrocytes (Slc1a3). E) Dot plot shows relative expression of genes used to differentiate the cortical nuclei clusters. F) Representative percentages of each cell type are shown. Clustering and differential expression were determined using Seurat in R. Samples for each group were pooled (n=3).

Cortical neuronal clusters and RNA profiles 7 dpi were influenced by microglia.

To determine the interactions between TBI and microglial depletion on cortical neuron gene expression, neuronal clusters were subset in R based on Syt1 expression. The neuronal nuclei (24,863 nuclei) were then sub-clustered using Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) to differentiate cortical neurons. Fig.2A shows that there were 15 distinct clusters of neurons (cortical layers 1–6, excitatory neurons, and inhibitory neurons). Cell-specific gene expression for each cluster is shown in Fig.2B. The influence of the four treatment groups (Veh-Control, PLX-Control, Veh-TBI and PLX-TBI) on the neuronal clusters is represented in Fig.2C. Fig.2C&E show that RNA profiles of upper layer neurons (Cux1), layer 4 neurons (Rorb), deep layer neurons (Foxp2), inhibitory neurons (Gad1&2), and excitatory neurons (Slc17a7) were detected.

Fig. 2. Cortical neuronal clusters and RNA profiles 7 dpi were influenced by microglia.

Continuing with the snRNA-seq experiment outlined in Fig.1A, neuronal clustering and profiles were determined. A) UMAP plots show 15 subsets of cortical neurons sub-clustered using Syt1 expression. B) Neuronal clusters represented based on expression of cell specific genes. C) UMAP plot with the distribution of neurons based on the four treatments groups at 7 dpi. D) Syt1+ cortical nuclei clusters of neurons were sub-clustered based on cell specific gene expression: upper layer neurons (Cux1&Cux2), layer 4 neurons (Rorb), deep layer neurons (Foxp2), excitatory neurons (Slc17a7), inhibitory neurons I (Adarb2, Gad1&2), and inhibitory neurons II (Gad1&2). E) Pie charts of percentages of clusters of the four treatment groups at 7 dpi. F) Top genes from each representative cluster are shown (p-adj <.05). G) IPA of canonical pathways of neuronal clusters (NC) 0–6, and 12 (p-adj<.05, z>2). Clustering and differential expression were determined using Seurat in R. Pooled samples for 3 replicates.

The influence of TBI and microglia depletion on neuronal clustering are represented in the pie charts (Fig.2E). The most prominent influence of TBI on cortical neurons at 7 dpi was the reductions in the percentage of neuronal clusters (NC): upper layer neurons NC1 (25% to 6%), layer 4 neurons NC2 (16% to 6%), and deep layer neurons: NC3 (17% to 6%). Parallel to these reductions 7 dpi, there were TBI-dependent increases in the percentage of TBI-upper layer neurons NC0 (not present to 30%), TBI-deep layer neurons NC5 (not present to 15%) and inhibitory neurons NC4 (5% to 10%). The TBI-dependent influence on neuronal clusters was attenuated by microglia depletion (PLX-TBI). For example, the PLX-TBI group had less NC0 (16%) and NC5 (12%) and had more NC4 (14%), NC1 (8%), and NC3 (8%) compared to the Veh-TBI group. Moreover, there was a unique cluster, PLX-TBI deep layer neurons NC12 (5%), that was only present in the PLX-TBI group. Examples of differentially expressed genes (DEGs: top 15 genes) in NC0–6, & NC12 are listed (Fig.2F). Overall, TBI and microglial influences on the pattern of neuronal clustering were evident 7 dpi.

DEGs for each of these clusters (NC 0–6 and NC12) were next used in IPA analysis of canonical pathways (Fig.2G). For example, synaptogenesis signaling and CREB signaling pathways were enhanced in NC1 (Control-upper layer neurons) and NC3 (Control-deep layer neurons), but these pathways were decreased in NC0 (TBI-upper layer neurons) and NC5 (TBI-deep layer neurons). TBI-induced suppression was also evident for oxytocin signaling and S100 family signaling within these clusters. Collectively, there were robust influences of TBI, and interactions between TBI and microglia depletion on specific neuronal populations in the cortex 7 dpi.

Neuronal RNA profiles and pathways in the cortex 7 dpi were influenced by microglia.

Continuing with the snRNA-seq analyses from Figs.1&2, the pie chart (Fig.3A) represents the percentage of the five different neuronal populations detected in all four experimental groups. Notably, clusters were pooled based on their subtype. Deep layer (37%) and upper layer cortical neurons (27%) represented approximately 60% of the total cortical neurons. At 7 dpi, there was a TBI-dependent suppression of gene expression within these neurons (across the five sub-populations). The DEGs for each neuronal cluster were used in IPA analysis of canonical pathways (Fig.3B). Only two canonical pathways, RhoGDI (Rho GDP-dissociation inhibitor) and PTEN (phosphatase and tensin homolog) signaling, were increased in neurons 7 dpi (p_ad<0.05; composite z-score >2, Fig.3B). A myriad of pathways, however, were decreased in neurons 7 dpi. Canonical pathways reduced by TBI included CREB signaling, synaptogenesis signaling, calcium signaling, and synaptic migration (p_ad<0.05; composite z-score >2, Fig.3B). IPA analysis also showed most differentially expressed (activation z-score) master regulators influenced by TBI (Fig.3C). For instance, anti-inflammatory related regulators, IL4R and PIAS1, were reduced in all five neuronal groups. In addition, HDAC5 was increased and CREBBP was decreased in upper layer neurons at 7 dpi. Amyloid precursor protein (APP) was enhanced by TBI (7 dpi) in deep layer cortical and inhibitory neurons. Last, CX3CR1 and FMR1 were enhanced 7 dpi in excitatory neurons.

Fig. 3. Neuronal RNA profiles and pathways in the cortex 7 dpi were influenced by microglia.

Continuing with the snRNA-seq experiment outlined in Fig.1A, neuronal clustering and profiles were determined A) Percentages of each neuron sub-type are shown. B) IPA (Canonical Pathways) of differentially expressed genes within all neuronal groups between TBI-Veh and Con-Veh (p-adj <.05). C) IPA (Master Regulators) of differentially expressed genes within all neuronal groups between TBI-Veh and Con-Veh (p-adj< .05). D) Pie-charts reflects the average proportion of genes whose expression was reversed (white) or unaffected (gray) by microglia depletion. Percentages of overall genes differentially expressed genes within all neuronal groups between TBI-Veh and Con-Veh (p-adj<.05). E). Subset of top genes increased or decreased within each neuronal group between TBI-Veh and Con-Veh (p-adj<.05). F) IPA (Canonical Pathways) of differentially expressed genes reversed by microglial depletion in representative neuronal groups between TBI-Veh and TBI-PLX. Each arrow represents a change in IPA z-score of ±2.

The graphical representation of different cortical neurons (Fig.3D) shows the percentage of DEGs influenced by TBI in each neuronal cell type that were either reversed or unaffected by microglia depletion. Microglial depletion with PLX5622 reversed 34–60% of the TBI-induced DEGs across the five neuronal subpopulations. For example, layer 4 cortical neurons had the highest number of DEGs influenced by TBI (4,298) and 60% of these were reversed by microglial depletion (Fig.3D). Fig.3E shows the top 5 genes increased or decreased by TBI in each neuronal cluster. IPA analysis of the DEGs revealed microglial depletion (PLX5622) reversed the suppression of synaptogenesis, CREB signaling, synaptic migration, and IL-4 signaling (p_ad<0.05; composite z-score >2) (Fig.3F). TBI-associated increases in PTEN and PPAR (peroxisome proliferator-activated receptor) signaling 7 dpi were also reversed by microglial depletion (Fig.3F). PTEN associated layer specific effects of microglial depletion are shown (Fig.3G). Microglial depletion selectively reversed p53 and p14 signaling (excitatory neurons), mitochondrial dysfunction signaling (layer 4 & excitatory neurons), and sonic hedgehog signaling (layer 4 neurons). Overall, there was a microglia-dependent component (34–60% dependent on cortical layer) to the TBI-induced reduction in neuronal homeostasis in the cortex.

Microglia RNA profiles and pathways in the cortex 7 dpi.

Continuing with the snRNA-seq analyses in Fig.1, nucleus RNA expression in microglia from the cortex was assessed. It is important to highlight that the PLX treatment groups did not have enough microglial nuclei to analyze (n<30), so comparisons shown were between Veh-TBI and Veh-Control groups. There were 365 nuclei from microglia and three primary microglia clusters (MC): MC0, MC1, and MC2 (Fig.4A). The majority of the microglia in the Veh-TBI were in MC1 (Fig.4B). The volcano plot shown represents the effect of TBI in microglia (Fig.4C). There were 1282 microglial DEGs after TBI in the cortex (7 dpi) with 348 genes increased and 898 genes decreased. IPA analysis of canonical pathways of the DEGs induced by TBI are shown (Fig.4E). The pathways increased in microglia by TBI include the inflammasome pathway, Trem1 signaling, PI3K/AKT signaling, Pyroptosis signaling, Th1 pathway, Macrophage classical activation pathway, and ISGylation signaling pathway. The pathways decreased in microglia by TBI include SNARE signaling and cAMP-mediating signaling. Furthermore, many of the DEGs associated with canonical pathways increased by TBI in microglia 7 dpi were related to STING mediated IFN-I production (Tbk1, cGas, Sting1) or IFNAR mediated response to IFN-I (Stat1, Ifnar2, Mx1, Mx2, Oasl2, and Usp18) (Fig.5F). The same genes were compared across all neuronal groups in Fig.4G. No IFN-I genes were differentially expressed in neurons after TBI. Collectively, these preliminary snRNAseq data of microglia are interpreted to indicate that was an enhanced IFN-I and STING response in microglia 7 dpi.

Fig. 4. Microglia RNA profiles and pathways in the cortex 7 dpi.

Continuing with the snRNA-seq experiment outlined in Fig.1A, microglia clustering and profiles were determined A) Microglial populations were subset using CSF1R expression and sub-clustered using UMAP. B) UMAP plot with the distribution of microglia based on the Veh-Control and Veh-TBI groups at 7 dpi. C) Volcano plot genes increased or decreased within microglia between TBI-Veh and Con-Veh (p-adj <.05). D) Number of differentially expressed genes (DEGs) increased or decreased by TBI (p-adj <.05). E) IPA (Canonical Pathways) of differentially expressed genes within all microglial groups between TBI-Veh and Con-Veh (p-adj <.05). Each arrow represents a change in IPA z-score of ±2. F) Representative dot plot of “STING Mediated IFN-I Production” (Tbk1, cGas, Sting1, and “IFNAR medicated IFN-I Response” (Stat1, Ifnar2, Mx1, Mx2, Oasl2, Usp18) positive microglia showing relative expression. G) Representative dot plot of “STING Mediated IFN-I Production” (Tbk1, cGas, Sting1, and “IFNAR medicated IFN-I Response” (Stat1, Ifnar2, Mx1, Mx2, Oasl2, Usp18) in neurons showing relative expression.

Fig. 5. TBI-induced cognitive deficits and neuroinflammation 7 dpi were attenuated by global knockout of STING.

A) Adult male C57BL/6 STING^WT and STING^KO mice were subjected to midline fluid percussion injury (TBI) or left as controls. Cognition, STING expression and neuroinflammation were assessed 7 dpi. Cognition was determined using the novel object recognition (NOR) and location (NOL) tests (n=11–12) B) Total exploration (seconds) of the objects in NOR. C) Percent time exploring the novel object and D) Discrimination index of time exploring the novel object. E) Total exploration (seconds) of the objects in NOL. F) Percent time exploring the object in the novel location. G) Discrimination index of time exploring the object in the novel location. In a separate experiment, STING^WT and STING^KO were subjected to control or midline fluid percussion injury (TBI) and STING protein levels were determined in the cortex. H) Representative images of STING (red) and DAPI (blue) labeling from the cortex 7 dpi. I) Percent area of STING labeling in the cortex 7 dpi (n=4–6). In a separate experiment, STING^WT and STING^KO were subjected to control or midline fluid percussion injury (TBI) and mRNA levels of J) Irf3, K) Ifi27 and L) Cd68 were determined in percoll enriched microglia collected 7 dpi (n=4–6). In addition, M) mRNA levels of Irf7, Ccl2, Tmem173 (Sting), Tnf, and MhcII were determined in the cortex. Bars represent the mean ± SEM, and individual data points are provided. Means with (*) are significantly different from control groups (p<0.05). Means with (†) are different (p< 0.05) from the WT-TBI group.

TBI-induced cognitive deficits and neuroinflammation 7 dpi were ameliorated by global knockout of STING.

Our previously published scRNA-seq data [21] and the snRNA-seq data on microglia shown here (Fig.4) indicate that there is a robust augmentation of IFN-I responses during the sub-acute phase of neuroinflammation after TBI (7 dpi). Thus, we surmise that the IFN-I pathway is key in the transition to chronic inflammation, neuronal imbalance, and microglia priming. STING is a critical regulator of the production of IFN-I genes. [21, 25, 27, 32, 33] Therefore, wild type (STING^WT) and global knockout (STING^KO) mice were subjected to control or TBI (mFPI), and cortical and hippocampal mediated cognition was assessed 7 dpi using NOR/NOL (Fig.5A). As expected, there were no differences in total time spent exploring between treatment groups (Fig.5B&E). Fig.5B-D shows that there were TBI-dependent reductions in NOR 6 dpi (TBI, F_1,43 = 84.0, p<0.0001). Time spent with the novel object was influenced by TBI and STING (interaction, F_1,43 = 28.0, p<0.0001, Fig.5C&D). Post-hoc analyses indicated that STING^WT-TBI mice investigated the novel object for less time compared to all other groups, including the STING^KO-TBI mice (p<0.0001, Fig.5C&D). These effects and interactions were also evident in NOL 7 dpi (Fig.5E-G). Time spent exploring the object in the novel location was influenced by TBI (F_1,43 = 60.5, p<0.0001) and STING (interaction, F_1,43 = 20.3, p<0.0001). For instance, STING^WT-TBI mice investigated the object in the novel location for less time compared to all other groups, including the STING^KO-TBI mice (p<0.0001, Fig.5F&G). Taken together, the cognitive impairments in TBI STING^WT mice were absent in the TBI-STING^KO mice 7 dpi.

Next, STING protein was assessed in the cortex 7 dpi. Fig.5H&I show a TBI-dependent increase in STING protein in the cortex at 7 dpi (F_1,18 = 13, p<0.05). STING protein expression was influenced by TBI and STING (interaction, F_1,18 = 13, p<0.0001, Fig.5H&I). In addition, STING protein was not detected in the cortex in either STING^KO group. This provides validation of the STING knockout mice. Furthermore, the increase in STING protein in the cortex 7 dpi was ablated in the STING^KO-TBI mice (p<0.0001). Specifically, STING protein was increased in STING^WT-TBI compared to all others (p<0.0001, for each) Overall, STING protein was increased by TBI in the cortex and this effect was ablated by the knockout.

In a related study, mice were injured as described, and samples were collected 7 dpi. Microglia were percoll enriched from the brain and the RNA levels of genes associated with IFN-I and inflammation were determined. The mRNA levels of IFN-I (Irf3, Ifi27) and inflammatory (Cd68) genes were increased in enriched microglia 7 dpi compared to controls (Fig.5J-L) (TBI F_1,16 = 8.1, p<0.01, for each). This TBI-associated increase in IFN-I-related genes in microglia was influenced by STING (interaction, F_1,16 = 13.6, p<0.002, for each). For instance, STING^WT-TBI had the highest RNA levels of these genes 7 dpi compared to all other groups including STING^KO-TBI (p<0.05, for all). Thus, TBI induction of STING-IFN-I responses in microglia was attenuated in global STING^KO mice.

In a similar experimental design, several inflammatory (Ccl2, Tnf, H2-eb1) and IFN-I-related genes (Tmem173, Irf7) were determined in the cortex 7 dpi (Fig.5M). Consistent with the STING protein data, Tmem173 (Sting) RNA was ablated in the STING^KO mice. In addition, TBI increased several inflammatory (Ccl2, Tnf, and H2-eb1 (MhcII)) and IFN-I (Irf7) in the cortex (F_1,20 = 27.2, p<0.0001, for each). Again, the increase in these genes influenced by TBI and STING (interaction, F_1,20 = 25.4, p<0.001, for each). Post hoc analysis confirmed that TBI-induced genes in the cortex of STING^WT mice (Irf7, Ccl2, Tnf, & H2-eb1) were attenuated in the STING^KO mice (Fig.5M, p<0.05, for each). Overall, IFN-I-related inflammation in the cortex 7 dpi was attenuated in TBI-STING^KO mice.

TBI-induced microglia restructuring and rod-shaped microglia formation 7 dpi were attenuated by global knockout of STING.

We have previously reported astrocyte and microglia restructuring 7 dpi in the cortex [25, 27]. Thus, astrocyte (GFAP) and microglia (IBA-1) morphology were assessed in the cortex 7 dpi of control (STING^WT) and knockout (STING^KO) mice (Fig.6A). As expected, GFAP⁺ labeling of astrocytes was increased 7 dpi in the cortex (TBI, F_{1, 19} = 22.9, p<.0001). This increase, however, was independent of STING (Fig.6B&C).

Fig. 6. TBI-induced microglia restructuring and rod-shaped microglia formation 7 dpi were attenuated by global knockout of STING.

A) Adult male C57BL/6 STING^WT and STING^KO mice were subjected to control (uninjured) or midline fluid percussion injury (TBI). GFAP and IBA-1 labeling was assessed in the cortex 7 dpi (n=5–6). B) Representative images of GFAP (red) and DAPI (blue) labeling (10×) in the cortex 7 dpi. Inset panel shows zoomed GFAP labeling. Right panel shows representative pseudo-skeletonized GFAP labeling (white) in STING^WT- TBI and STING^KO-TBI groups at 7 dpi. C) Percent-area of GFAP labeling in the cortex 7 dpi. D) Representative images of IBA-1 (green) labeling (10×) in the cortex 7 dpi. Inset panel shows zoomed IBA-1 labeling. Right panel shows representative shows pseudo-skeletonized Iba1 labeling (white) in STING^WT- TBI and STING^KO-TBI groups at 7 dpi. E) Percent-area of IBA-1 labeling in the cortex 7 dpi. F) Representative images IBA-1 labeling of rod-shaped microglia (20×) in the cortex 7 dpi. Inset panel shows zoomed IBA-1 labeling of rod-shaped microglia. Right panel shows representative pseudo-skeletonized IBA-1 labeling of rod microglia (white) in STING^WT- TBI and STING^KO-TBI groups at 7 dpi. G) Number of Iba1+rod microglia per 20×field in the cortex 7 dpi. Bars represent the mean ± SEM, and individual data points are provided. Means with (*) are significantly different from control groups (p<0.05). Means with (#) tend to be significantly different (p<0.1).

TBI also increased the morphological restructuring of microglia (% area of IBA-1⁺ labeling) in the cortex 7 dpi (F_1,18 =100.1, p<0.0001, Fig.6D&E). This TBI-associated increase in microglia restructuring was influenced by STING (interaction, F_1,18 = 50.7, p<0.0001, Fig.6D&E). Post hoc analysis confirmed that STING^WT-TBI had the highest percent area of IBA-1⁺ labeling in the cortex compared to all other groups including the STING^KO-TBI mice (p<0.0001, Fig.6D&E).

Restructuring of cortical microglia was also represented by increased rod-shaped microglia 7 dpi (Fig.6F&G). Here, TBI increased the number of rod microglia in the cortex 7 dpi (F_{1, 18} = 7.5, p<0.05). This TBI-associated increase in the percentage of rod-shaped microglia was dependent on STING (interaction, F_{1, 18} = 5.2, p<0.05). Thus, STING^KO-TBI mice had fewer rod-shaped microglia in the cortex at 7 dpi compared to the STING^WT–TBI mice (p<0.05, Fig.6G). Taken together, microglia morphological restructuring and rod formation 7 dpi was dependent on the STING-IFN-I pathway.

TBI-induced cognitive deficits and neuroinflammation 30 dpi were attenuated by global knockout of STING.

Another critical issue after TBI is persistent neuroinflammation evident 30 dpi. We have reported neuroinflammatory processes persisted in the cortex 30 dpi [39] and these were microglia dependent [21, 29]. Therefore, STING^WT and STING^KO mice were subjected to TBI (mFPI) or remained controls, and cognition was assessed at 30 dpi using NOR/NOL (Fig.7A). As expected, there were no differences in total time spent exploring between treatment groups (Fig.7B). Fig.7B-C show that there were TBI-dependent reductions in NOR 29 dpi (F_{1, 45} = 16.9, p<0.001). Time spent with the novel object was influenced by TBI and STING (interaction, F_{1, 44} = 22.7, p<.0001 Fig.7B&C). Post-hoc analyses indicated that STING^WT-TBI mice investigated the novel object for less time compared to all other groups, including the STING^KO-TBI mice (p<0.0001, Fig.7B&C). These effects and interactions were also evident in NOL 30 dpi (Fig.7D-F). Again, time spent exploring the object in the novel location was influenced by TBI (F_{1, 47} = 12.0, p<0.002) and STING intervention (interaction, F_{1, 42} = 7.4, p<0.01). For instance, STING^WT-TBI mice investigated the object in the novel location for less time compared to all other groups, including the STING^KO-TBI mice (p<0.0001, Fig.7E&F). Taken together, the cognitive impairments 30 dpi in STING^WT mice were absent in the TBI-STING^KO mice.

Fig. 7. TBI-induced cognitive deficits and neuroinflammation 30 dpi were attenuated by global knockout of STING.

Adult male C57BL/6 wild type (STING^WT) and global STING knockout mice (STING^KO) were subjected to control (uninjured) or midline fluid percussion injury (TBI). Cognition and neuroinflammation were assessed 30 dpi. Cognition was determined using the novel object recognition (NOR) and location (NOL) tests (n=12–13). A) Total exploration (seconds) of the objects in NOR. B) Percent time exploring the novel object and C) Discrimination index of time exploring the novel object. D) Total exploration (seconds) of the objects in NOL. E) Percent time exploring the object in the novel location. F) Discrimination index of time exploring the object in the novel location. G) In a separate experiment, STING^WT and STING^KO were subjected to control or midline fluid percussion injury (TBI) and GFAP H) and Iba1 I) labeling were determined in the cortex. J) In addition, mRNA levels of Irf7, Ifi27, Clec7a, Tnf, H2-eb1, and Itgax were determined in the cortex. Bars represent the mean ± SEM, and individual data points are provided. Means with (*) are significantly different from control groups (p<0.05). Means with (†) are different (p< 0.05) from the WT-TBI group.

Next, astrocyte and microglia morphology were assessed in the somatosensory cortex 30 dpi in STING^WT and STING^KO mice (Fig.7G-I). For astrocytes, GFAP⁺ labeling in the cortex remained elevated 30 dpi (TBI, F_{1, 15} = 43.4, p<0.001, Fig.7G&H). This TBI-associated increase in astrocyte restructuring was influenced by TBI and STING intervention (interaction, F_{1, 15} = 11.4, p<0.005, Fig.7H). Post hoc analysis confirmed that STING^WT-TBI had the highest percent area of GFAP⁺ labeling in the cortex compared to all other groups including the STING^KO-TBI mice 30 dpi (p<0.005, Fig.7H). TBI-associated morphological restructuring of microglia (% area of IBA-1⁺ labeling) in the cortex was still evident 30 dpi (TBI, F_{1, 14} = 9.2, p<0.05, Fig.7G&I). This TBI-associated increase in microglia restructuring tended to be influenced by TBI and STING intervention (interaction, F_{1, 14} = 4.1, p=0.06, Fig.7I). STING^WT-TBI tended to have the highest percent area of IBA-1 labeling in the cortex compared to all other groups including the STING^KO-TBI mice 30 dpi (p=0.1, Fig.7G&I).

Next, several IFN-I (Irf7, Ifi27) and priming-related (Clec7a, Itgax, and MhcII) genes were determined in the cortex 30 dpi. As previously reported [21, 29], IFN-I genes were attenuated by 30 dpi (Fig.7J). Genes associated with priming (Clec7a, Itgax, and MhcII) were all increased in the cortex 30 dpi (TBI, F_1,17 = 12.75, p<0.05, for each). Expression of H2-eb1 (MhcII) and Itgax (Cd11c) 30 dpi were influenced by STING (F_1,17 = 7.1, p<0.05, for each). Post hoc analysis confirmed that the TBI-induced MhcII in the cortex of STING^WT mice was attenuated in the STING^KO mice (p<0.05). Itgax levels were the same between control STING^WT and TBI STING^KO mice (, Fig.7J). Clec7a tended to be influenced by STING (F_1,18 = 3.5, p=0.07) and levels were the same between control STING^WT and TBI STING^KO mice (p<0.05, Fig.7J). Overall, genes associated with microglial priming persisted in the cortex 30 dpi and these genes were reduced in global STING^KO mice.

TBI induced neuroinflammation and cognitive deficits (7 dpi) were attenuated by the STING antagonist, chloroquine.

Our findings indicate that TBI-induced inflammatory gene expression, microglial activation, and cognitive deficits were attenuated in global STING^KO mice. Thus, these STING-driven IFN-I responses 7 dpi are critical propagators of neuroinflammation. As such, we next sought to determine if pharmacological inhibition of STING also reduced the TBI-associated neuroinflammation and cognitive impairments. Here, chloroquine (CQ) was used. We used CQ as it is an FDA-approved anti-viral medication with blood-brain barrier (BBB) penetrability[40]. While other STING inhibitors exist, they require central (ICV) administration because they are not BBB permeable.

In these experiments, C57BL/6 mice were subjected to TBI (mFPI) or left as control. Vehicle or CQ intervention was administered i.p. starting 1 h after TBI. Mice received injections of vehicle or CQ every 24 h for 6 consecutive days. We selected this time-point as it represents a key shift from sub-acute to chronic inflammation after TBI and is associated with increased IFN-I signaling. Cognition was assessed 7 dpi using NOR/NOL (Fig.8A-F). As expected, there were no differences in total time spent exploring between treatment groups (Fig.8A&D). Fig.8B-C shows that there were TBI-dependent reductions in NOR 6 dpi (F_1,30 = 51, p<0.005). Time spent with the novel object was influenced by TBI and CQ intervention (interaction, F_1,30 = 40, p<0.0001, Fig.8B&C). Post-hoc analyses indicated that Vehicle-TBI mice investigated the novel object for less time compared to all other groups, including the CQ-TBI mice (p<0.001, Fig.8B&C). These effects and interactions were also evident in NOL 7 dpi (Fig.8E-F). Again, time spent exploring the object in the novel location was influenced by TBI (F_1,31 = 23, p<0.0001) and CQ intervention (interaction, F_1,31 = 10, p<0.005). For instance, Vehicle-TBI mice investigated the object in the novel location for less time compared to all other groups, including the CQ-TBI mice (p<0.0005, Fig.8E&F). Taken together, the cognitive impairments 7 dpi in Vehicle-TBI mice were absent in the CQ-TBI mice.

Figure. 8. TBI induced neuroinflammation and cognitive deficits (7 dpi) were attenuated by the STING antagonist, chloroquine.

Adult male C57BL/6 mice were subjected to control (uninjured) or midline fluid percussion injury (TBI). 1h after TBI, mice received i.p. injection of saline (vehicle) or chloroquine (80mg/kg). Cognition and neuroinflammation were assessed 7 dpi. Cognition was determined using the novel object recognition (NOR) and location (NOL) tests (n=8–9). A) Total exploration (seconds) of the objects in NOR. B) Percent time exploring the novel object and C) Discrimination index of time exploring the novel object. D) Total exploration (seconds) of the objects in NOL. E) Percent time exploring the object in the novel location. F) Discrimination index of time exploring the object in the novel location. Following cognitive testing, mice were sacrificed, and cortical sections were collected for IHC and RNA analyses. G) GFAP and H) IBA-1 labeling were determined in the cortex. I) mRNA levels of Irf7, Ccl2, Tnf, H2-eb1, and Cd68, were determined in the cortex. Bars represent the mean ± SEM, and individual data points are provided. Means with (*) are significantly different from control groups (p<0.05). Means with (†) are different (p< 0.05) from the WT-TBI group.

In a similar experimental design, astrocyte (GFAP) and microglia (IBA-1) morphology were assessed in the cortex 7 dpi of vehicle and CQ treated mice (Fig.8G-I). GFAP⁺ labeling of astrocytes was increased 7 dpi in the cortex (TBI, F_{1, 25} = 13.8, p<.05). This increase, however, was independent of CQ (Fig.G&H). For microglia, TBI increased the % area of IBA-1⁺ labeling in the cortex 7 dpi (F_1,25 = 5.0, p<0.05, Fig.8G&H). Veh-TBI tended to have higher % area of IBA-1⁺ labeling in the cortex than Veh-Controls (p=0.06), but CQ-TBI was not different from CQ-CON mice (Fig.8I).

After the completion of the cognitive testing, several inflammatory (Ccl2, Tnf, Cd68, H2-eb1) and IFN-I-related (Tmem173, Irf7) genes were determined in the cortex 7 dpi (Fig.8J). There was a main effect of TBI on inflammatory and IFN-I-related genes (F_1,29 = 22.1, p<0.0001, for each, (Fig.8J). As well, there was a main effect of CQ treatment on inflammatory and IFN-I-related genes (F_1,29 = 4.1, p<0.05, for each). Again, the increase in these genes (Ccl2, Cd68, Tmem173, Irf7) after TBI was influenced by CQ (interaction, F_1,29 = 5.2, p<0.05, or tended to be influenced by CQ (Tnf, H2-eb1) (interaction F_1,29 = 7.7 p<0.1). Post hoc analysis confirmed that the TBI-induced genes in the cortex of Veh-TBI were attenuated (Ccl2, Cd68, Tmem173, Irf7, p<.05 or tended to be attenuated (H2-eb1, Tnf, p<0.1) in the CQ-TBI mice (Fig.8J). Overall, IFN-I-related inflammation in the cortex 7 dpi was reduced in mice receiving CQ intervention after TBI.

Discussion

The purpose of this study was to determine the degree to which neuronal gene expression and IFN-I responses intersected after TBI to cause cognitive impairment. Here, novel data show TBI-dependent suppression of cortical neuronal homeostasis 7 dpi. Up to 50% of DEGs after TBI in cortical neurons were dependent on microglia. Moreover, there were reductions in neuronal pathways associated with CREB signaling, synaptogenesis, and synaptic migration after TBI and these deficits were reversed by microglia depletion. The RNA signature in microglia 7 dpi was consistent with a robust IFN-I response, a key inflammatory pathway activated by STING. TBI-induced cognitive deficits were reversed by both genetic and pharmacological intervention of the STING pathway. In addition, TBI-induced STING induction, microglia morphological restructuring, inflammatory, and IFN-I-related gene expression within the cortex were attenuated in STING^KO mice compared to STING^WT mice. CQ, a STING inhibitor, reversed TBI-induced inflammatory and IFN-I-related gene expression in the cortex. Last, the prolonged inflammatory profile and gliosis in the cortex at 30 dpi were STING dependent.

One novel finding of this study was the profound effect of TBI on specific neuronal populations in the cortex. For example, our snRNA-seq data provides evidence of neuronal dysfunction 7 dpi, across five groups of cortical neurons including deep cortical neurons, layer 4 cortical neurons, upper layer cortical neurons, excitatory neurons, and inhibitory neurons. This enhanced resolution of neuronal RNA profiles was the primary reason that the snRNA-seq approach was used in this study [47]. Consistent with our previous scRNA-seq study [21], there was a clear influence of TBI on the RNA profile (nucleus) of cortical neurons. Analyses of IPA master regulators influenced by TBI revealed a number of anti-inflammatory related regulators, IL4R and PIAS1, were reduced in all five neuronal groups. Other master regulators were layer specific, including HDAC5 and CREBBP which were increased in upper layer neurons and decreased in deep layer neurons. Furthermore, APP was increased in deep layer neurons while CX3CR1 and FMR1 were increased in excitatory neurons. CX3CR1 is a microglial receptor, but the ligand, CX3CL1, is highly expressed in neurons and this pathway is influenced by TBI and influences neuronal homeostasis [48, 49]. We interpret these changes as a link to CX3CL1 in neurons. For example, IPA target molecules connected to CX3CR1 regulation to genes associated with post-synaptic densities (Dlg4, Fyn, and Nsf). Overall, there was a suppression of critical pathways associated with growth and plasticity across the five cortical neuronal subtypes including, synaptogenesis signaling, CREB signaling, S100 signaling, and IL-4 signaling. In parallel, neuronal pathways associated with cellular growth and survival were suppressed.

Microglia ablation prior to TBI resolved between 30–60% of the DEGs across the cortical neuron subtypes. These reversals were directly related to the pathways associated with growth and plasticity. This microglia-dependent reversal was lower than what was detected from scRNA-seq (e.g., 90% reversal) [21]. Nonetheless, there were fewer neurons and less specific neuronal subtype resolution in our previous scRNA-seq dataset compared to the current snRNA-seq data set [21]. The IPA reversals show that major pathways influenced 7 dpi were reversed by microglia depletion. Overall, TBI induces neuronal dysfunction 7 dpi in the cortex, and approximately 50% of these changes in RNA were microglia dependent.

Related to the points above, microglial influence on cortical neurons after TBI was associated with subtype and the spatial dynamics of the cortical structure. For instance, the upper layer cortical neurons (layers 1–3) are closest to the pia matter and the initial TBI. In these Cux1/2+ neurons, 47% of the DEGs (1445 genes) induced by TBI (7 dpi) were reversed by microglia depletion. Moreover, the TBI-associated upper neuronal profile (NC0) was reduced from 30% to 16% with microglia depletion. Below the upper layer neurons are the layer 4 neurons (internal granular layer), which receive inputs from the thalamus [50]. Within these layer 4 neurons (Rorb⁺), 60% of the DEGs (2540 genes) induced by TBI were reversed by microglia depletion. Next are the deep layer neurons (layers 5&6) which are closest to the white matter [51]. In these Foxp2⁺ neurons, 34% of the DEGs (908 genes) induced by TBI (7 dpi) were reversed by microglia depletion. The effect of microglia on neuronal profiles was higher in cortical areas close to the primary injury site. As one moves deeper into the cortex from the pia matter, the imbalance in the neurons after TBI was less dependent on microglia.

Another relevant finding from the snRNA-seq data set was that PTEN signaling was increased in all neuronal populations of the cortex following TBI. PTEN is a master inhibitor of cell growth, survival, and differentiation [52]. Overexpression of PTEN leads to reduced neuronal growth, axon growth, and dendritic branching [53]. Moreover, inhibition of PTEN following middle cerebral artery occlusion (MCAO) improved neuronal survival, dendritic arborization, and motor deficits [54]. Other studies reported that inhibiting PTEN following either a single severe TBI [48] or repetitive mild TBIs improved functional outcomes [55]. In the current study, microglial depletion reversed the increased PTEN signaling 7 dpi. Moreover, there were layer-specific effects of microglial depletion related to PTEN-regulated pathways. For example, p14, p53, and sonic hedgehog signaling were only reversed in excitatory neurons. Other pathways, such as mitochondrial dysfunction, were only reversed in layer 4 neurons. Overall, microglia-dependent suppression of neuronal homeostasis in cortical neurons 7 dpi was associated with increased PTEN-mediated pathways.

Another relevant finding was microglia, not neurons, had RNA evidence of IFN-I production and response. For example, the microglia RNA profile 7 dpi had evidence of STING-dependent production of IFN-1 (cGas, Tbk1, Sting1). In addition, genes associated with the IFN-I receptor (Ifnar2, Stat1) and interferon stimulated genes (ISGs) (Mx1, Mx2, Oasl2) were also increased in the microglia 7 dpi. These ISGs are highly conserved markers for cells stimulated by interferons [56]. While other intracellular receptors that induce IFN-I expression exist (MAVS, RIG-1, etc.) these pathways were unaffected in our current or previous studies [21, 25, 27]. One plausible explanation is that RIG-1 and MAVS detect single/double-stranded RNA, present with viral infection while STING detects double-stranded DNA, present with cellular stress or injury [32]. These genes were selected based on previous reports of TBI induced IFN-I microglia detected using scRNA-seq after TBI [21]. The increase in IFN-I signaling (detected by snRNA-seq) was not present within the neuronal populations after TBI. Moreover, the STING-dependent IFN-I response in TBI microglia 7 dpi was also reflected by increases in the canonical pathways associated with IFN-I (Inflammasome pathway, PI3K/AKT signaling, Th1 pathway, Macrophage classical activation pathway, and ISGylation signaling pathway). Notably, the low number of microglia that were resolved by snRNAseq is limiting and these data should be considered preliminary. Nonetheless, these preliminary data provided evidence of STING dependent IFN-I production and increased IFN responsiveness in microglia 7 dpi .

The preliminary snRNA-seq data for microglia are consistent with both scRNA-seq data and NanoString RNA data from the cortex 7 d after diffuse TBI (mFPI) [21, 25] showing robust IFN-I signaling in microglia in the sub-acute phase following diffuse TBI (mFPI). In addition, studies using lateral FPI reported IFN-I response at 7 dpi in both enriched astrocytes and microglia [26]. Studies of focal TBI (CCI) had increased neuronal expression of STING following injury [37, 59]. In fact, the IFN response in focal or severe TBI was evident within 24 hours in mice [24] and detected in humans following death from severe TBI [24, 37]. Focal TBI induces tissue cavitation and neuronal death, which is distinct to severe brain injury. Tissue cavitation and neuronal death are not present in the diffuse model of TBI induced by mFPI [37]. Consistent with this notion, a previous study using diffuse TBI (mFPI) also did not detect increased IFN-I expression within neurons [21]. Thus, neuronal IFN-I production following traumatic CNS injury may depend on significant tissue damage while the glial production of IFN-I may be present after diffuse injury in the sub-acute phase of injury response.

A key finding of this study was that IFN-I, microglia activation, and pro-inflammatory signaling following TBI were dependent on STING. For instance, STING^KO mice had reduced IFN-I and pro-inflammatory gene expression 7 and 30 dpi. At the sub-acute time point (7 dpi) following injury, STING protein expression was elevated in the cortex and this induction was ablated in the STING^KO mice. These changes were paralleled by reduced cortical inflammation (Ccl2, Tnf, MhcII) and IFN-I gene expression (Tmem173, Irf7) in TBI-STING^KO mice 7 dpi. These reductions were paralleled by decreased IFN-I-related (Irf3, Ifi27) and pro-inflammatory (Cd68) genes within enriched TBI-STING^KO microglia. Multiple measures of microglial activation (IBA1 % area, rod formation, enriched RNA) were attenuated in STING^KO groups at both 7 dpi and 30 dpi. These findings are consistent with previous reports that indicate STING^KO mice are protected following focal TBI (CCI) [32, 37]. Based on the data here we contend that IFN-I inhibition prevented the induction of pathological gene expression. It is also possible that STING^KO mice had increased protective gene expression. For instance, a previous study using CCI detected increased anti-inflammatory mRNA after administration of a PERK-STING inhibitor (GSK2656157) [59]. The novelty of the current study is that the STING intervention was effective in reducing neuroinflammation at subacute and chronic time points following diffuse TBI. In addition, these previous studies with CCI assessed gene expression acutely (24 h) and did not assess cognition. We also show that the reduction of IFN-I and pro-inflammatory gene expression were mirrored using CQ administration. For example, chloroquine attenuated the pro-inflammatory (Tnf, Cd68, Ccl2) and IFN-I (Tmem173, Irf7) profiles of cortical RNA. Similarly, CQ prevented microglial morphological restructuring 7 dpi. Overall, STING induction mediated IFN-I and pro-inflammatory gene expression following diffuse TBI.

Another point for discussion is that astrogliosis was STING independent 7 dpi. For example, STING knockout and CQ treatment at 7 dpi did not reduce astrocyte restructuring following TBI. These data are consistent with a previous study that found pretreatment using a STING agonist (DMXAA) prior to injury did not alter astrocyte morphological restructuring 7 dpi [27]. However, when GFAP⁺ percent area was examined 30 dpi, TBI-STING^KO mice had returned to baseline levels while TBI-STING^WT mice were still increased. This matches well with previous work that reported astrogliosis remains 30 dpi following diffuse TBI (mFPI), [39, 60] and is dependent on STING [37]. This reduction in GFAP⁺ percent area between day 7 and 30 dpi may represent a shift away from the traditional timeline of astrogliosis due to reduced pro-inflammatory gene expression through STING^KO mice. Overall, chronic astrogliosis is dependent on increased STING expression.

Another relevant finding of this study was that cognitive impairments 7 and 30 dpi were dependent on STING. For example, there were cognitive impairments evident in the NOR and NOL memory tasks 7 and 30 dpi in STING^WT-TBI mice, but not in STING^KO-TBI mice. Pharmacological inhibition of STING with CQ also blocked TBI-induced cognitive deficits at 7 dpi. These memory tasks requires intact cortical and hippocampal connections [61]. These data are consistent with previous work showing that microglia depletion (CSF1R antagonism) prior to TBI ablated IFN-I responses in microglia and reversed TBI-associated memory deficits at 30 dpi [21]. These data are also consistent with other studies (CCI) that have inhibited components of the type I interferon pathway (IFN-β or STING) and found reduced inflammatory gene expression, motor deficits, and cognitive deficits [62, 63]. By preventing STING-mediated induction of IFN-I signaling following TBI, cognitive burden was reduced following TBI.

One notable point for discussion is that the cortical mRNA 30 dpi was consistent with genes associated with microglial priming [39, 60]. Primed microglia are hyper-reactive to secondary immune challenge (liposaccharide) 30 dpi and this resulted in exaggerated pro-inflammatory cytokine expression and augmented behavioral and cognitive deficits [39, 60]. Here, we showed that cortical RNA levels of H2-eb1, Itgax, and Clec7a were increased 7 dpi and remained elevated following 30 dpi. This primed cortical RNA profile, which are hallmarks for microglial priming following TBI [39, 60], was absent in the STING^KO-TBI mice. These changes were paralleled by a reduction of morphological restructuring of microglia 30 dpi in STING^KO mice. The TBI-induced IFN-I and pro-inflammatory signaling in microglia during the subacute phase of injury (7 dpi) which give way to increased expression of priming-associated genes and morphological restructuring at 30 dpi in the cortex, was STING dependent. Thus, STING was necessary for the inflammatory-mediated processes that drive microglial activation at 7 dpi toward a primed profile evident at 30 dpi in the cortex.

It is important to highlight that the RNA profiles of neurons after TBI with STING interventions were not assessed in the current study. This is a limitation of the current paper. Nonetheless, the expectation, based on the reduction of cognitive deficits and inflammatory gene expression, is that neuronal dysfunction in the cortex would be improved 7 dpi with these STING interventions (KO and CQ) similar to the effect of microglia depletion in Figs.2-3. A proportion of TBI effects would be reversed, but not all. There are multiple components of the neuronal response to injury that are microglial and inflammatory independent [21, 29] Physical damage to axons are independent of microglial depletion [21]. Similarly, our previous microglial intervention studies with TBI show that attenuation of RNA profiles in neurons, microglia, and cortical samples corresponded with improvements in neuronal plasticity and physiology [21, 25, 29]. Collectively, reducing type I interferon signaling using microglial depletion or STING intervention improved inflammatory and cognitive outcomes after TBI.

In summary, TBI induced a STING-dependent IFN-I response sub-acutely that led to neuronal dysfunction and cognitive impairment. Microglial depletion reversed approximately 50% of the TBI-induced cortical neuronal DEGs. The transition from acute to chronic inflammation after diffuse TBI in mice was dependent on the STING pathway. This is a pathway that can be targeted using pharmacological interventions. This is of particular importance as it relates to inflammatory events and subsequent microglial priming. Reducing IFN-I signaling after TBI with STING-dependent interventions attenuated microglial activation, cognitive impairment, and evidence of a primed profile of microglia 30 dpi. Therefore, inhibiting STING and subsequently IFN-I signaling may represent a viable intervention for reducing TBI-associated neuroinflammation and cognitive deficits.

Data Availability:

The data that support the findings of this study are available from the corresponding author upon reasonable request.