Abstract
While originally identified as an antiviral pathway, recent work has implicated that cyclic GMP-AMP-synthase-Stimulator of Interferon Genes (cGAS-STING) signaling is playing a critical role in the neuroinflammatory response to traumatic brain injury (TBI). STING activation results in a robust inflammatory response characterized by the production of inflammatory cytokines called interferons, as well as hundreds of interferon stimulated genes (ISGs). Global knock-out (KO) mice inhibiting this pathway display neuroprotection with evidence that this pathway is active days after injury; yet, the early neuroinflammatory events stimulated by STING signaling remain understudied. Furthermore, the source of STING signaling during brain injury is unknown. Using a murine controlled cortical impact (CCI) model of TBI, we investigated the peripheral immune and microglial response to injury utilizing male chimeric and conditional STING KO animals, respectively. We demonstrate that peripheral and microglial STING signaling contribute to negative outcomes in cortical lesion volume, cell death, and functional outcomes postinjury. A reduction in overall peripheral immune cell and neutrophil infiltration at the injury site is STING dependent in these models at 24 h. Transcriptomic analysis at 2 h, when STING is active, reveals that microglia drive an early, distinct transcriptional program to elicit proinflammatory genes including interleukin 1-β (IL-1β), which is lost in conditional knock-out mice. The upregulation of alternative innate immune pathways also occurs after injury in these animals, which supports a complex relationship between brain-resident and peripheral immune cells to coordinate the proinflammatory response and immune cell influx to damaged tissue after injury.
Significance Statement
The innate immune STING pathway triggers harmful neuroinflammation after traumatic brain injury with support from human and preclinical models indicating that this pathway is active hours to days after injury. Our findings in a preclinical cortical contusion model suggest that STING signaling specifically from peripheral immune cells or microglia drive this process to affect injury outcome. Activation of STING in microglia as early as 2 h postinjury drives a distinct transcriptional program that influences neutrophil and peripheral immune cell influx, which dictates injury outcomes. These findings shed light on the acute temporal changes in a cell-type specific manner where innate immunity drives subsequent events that affect secondary injury processes after insult.
Introduction
Traumatic brain injury (TBI) is a leading cause of death and disability across the globe, and in the United States alone an estimated 2.8 million individuals sustain a TBI annually; albeit this number is well underreported (Roozenbeek et al., 2013; Peterson et al., 2018). Following the initial mechanical damage upon injury, various cellular processes termed “secondary injury” occur in the hours to weeks after impact, exacerbating tissue damage. A key aspect of secondary injury is neuroinflammation (Jassam et al., 2017; Simon et al., 2017); however, the neuroinflammatory response is highly complex with an ever-changing, heterogeneous environment of resident and infiltrating immune cells at the site of trauma and in the periphery, which is not fully understood. Some of the first inflammatory events immediately following TBI are the release of damage-associated molecular patterns (DAMPs) from injured cells (McKee and Lukens, 2016) and microglial activation as early as 30 min after insult (Davalos et al., 2005; Nimmerjahn et al., 2005; Roth et al., 2014).
As a DAMP-sensing pathway, recent work has shown that the cyclic GMP-AMP-synthase-Stimulator of Interferon Genes (cGAS-STING)-Type I interferon (IFN) signaling pathway plays an important role in the deleterious neuroinflammatory response to TBI. Preclinical brain injury models utilizing global genetic knock-outs (KO) of cGAS (Fritsch et al., 2022), STING (Abdullah et al., 2018; Fritsch et al., 2022), IFNβ (Barrett et al., 2020), or the Type I IFN receptor, IFNAR (Karve et al., 2016; Todd et al., 2023) largely resulted in neuroprotection. RNAseq analysis has shown that STING-IFN transcriptional signatures are present 7 d postinjury in a fluid percussion injury paradigm, and the effects of injury are augmented when mice are pretreated with a STING agonist (Wangler et al., 2022). Aged animals display heightened cGAS-STING-IFN signaling after injury, indicating that this pathway is clinically relevant across the age spectrum (Barrett et al., 2021; Wangler et al., 2022). All this work points to the potential positive benefits of STING inhibition, but there are key gaps in knowledge that have yet to be addressed. First, it is currently unknown which cell types STING-dependent signaling originates from after injury.
Data from central nervous system (CNS) infection (Reinert et al., 2016; Katzilieris-Petras et al., 2022), TBI (Fritsch et al., 2022), and stroke (Peng et al., 2020), as well as RNA sequencing databases (Zhang et al., 2014, 2016), suggest that STING is highly expressed in the microglia when compared with other CNS cell types. Furthermore, many types of peripheral immune cells, except for neutrophils (Xia et al., 2015), express STING, and DAMPs have been found to be circulating in serum and cerebrospinal fluid after injury (Marchi et al., 2013; Kigerl et al., 2014; Kayhanian et al., 2022). To address this gap in knowledge, we sought to determine which cell types are responsible for STING-mediated neuroinflammation following neurotrauma to better understand how this pathway is sustained days after insult.
Our previously published data showed STING expression elevated in the brain as early as 2 h postinjury (Fritsch et al., 2022), and patients display increased expression >6 h postinjury (Abdullah et al., 2018). However, it is unclear how important this increase in STING expression at acute time points is for shaping the neuroinflammatory milieu. Furthermore, Type I IFNs at these later time points can be sustained by autocrine or paracrine signaling mechanisms (González-Navajas et al., 2012) by cell types that do not express STING but do express IFNARs. So, studies have demonstrated that peripheral immune cell IFN signaling is important for neuroprotection 7 d postinjury in a controlled cortical impact (CCI) model using chimeric IFNAR KO animals (Karve et al., 2016), or that there is chronic tissue protection and improved function outcomes from IFNβ KO (Barrett et al., 2020) mice, may not necessarily be STING dependent. Finally, cGAS-STING is not the only DAMP-sensing pathway that mediates IFN production (Manes and Lita-Nazar, 2021). It is unclear which part of the early neuroinflammatory landscape after injury is STING dependent. This information will be important for establishing therapeutic windows if STING were to be considered for patient care. The availability of cGAS or STING antagonists for the treatment of cancer or autoimmune conditions is already moving forward, and is an extremely attractive option as many companies are currently in different stages to reach clinical trials for FDA approval (Sheridan, 2019; Mullard, 2023).
To address these questions, we utilized a CCI injury paradigm in conditional microglial-specific STING KO animals to assess histological and chronic functional outcomes. We further evaluated the acute transcriptional programming at 2 h driven by STING activated microglia, finding that STING influences immune cell recruitment correlating with injury outcome. We also found that STING expression increased 2 h postinjury in circulating white blood cells (WBCS), and bone marrow chimeric STING KO mice displayed significant protection and less immune cell infiltration after injury as well. Bone marrow studies also support the notion that STING signaling in WBCs in turn affects microglia. Our findings contribute to a broader understanding concerning the timeline of acute activation of neuroinflammation and the compartmentalized signaling to drive latter events influencing leukocyte infiltration into the parenchyma, which affects secondary brain injury.
Material and Methods
Animals
All mice were housed in a pathogen-free facility on a 12 h light/dark cycle at Virginia Tech and provided standard rodent diet and water ad libitum. C57/Bl/6J-TMEM173gt/J (STING KO), B6;SJL-Sting1tm1.1Camb/J (STINGfl/fl), B6.129P2(Cg)-Cx3cr1tm2.1(cre/ERT2)Litt/WganJ (CX3CR1CreER/CreER), and C57BL/6-Tg(CAG-EGFP)131Osb/LeySopJ (GFP+ WT) mice were purchased from Jackson Laboratories. We bred STING KO (gt/gt) mice to GFP+ transgenic mice to generate heterozygous animals for both the STING gt mutation and the GFP transgene. Those heterozygous animals were then crossed to establish a colony of STING KO (gt/gt) GFP+ animals. To generate STINGfl/fl, CX3CR1CreER+ (cKO) and CX3CR1CreER+ animals, STINGfl/fl mice were bred to CX3CR1CreER/CreER animals to obtain heterozygous STINGfl+, CX3CR1CreER+ mice. These mice were then bred to obtain cKO, STINGfl/fl, and CX3CR1CreER+ genotypes. Considering that these percentages of wanted genotypes were low from this crossing, we also optimized crosses using combinations of homozygous and heterozygous mice for the floxed and Cre genotypes. Mice were all on C57Bl/6J background. All experiments were conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals and under approval of the Virginia Tech Institutional Animal Care and Use Committee.
Generation of microglial-specific STING KO animals
To activate Cre recombinase, 6- to 8-week-old male animals received 100 mg/kg tamoxifen (Sigma-Aldrich) in corn oil via intraperitoneal injection for 5 consecutive days. Animals underwent CCI surgery 1 month after the last dose of tamoxifen.
Generation of bone marrow chimeric animals
Mice received two x-ray irradiated doses of 550 rad spaced 6 h apart to ablate bone marrow cells (BMCs) at 8 weeks of age. Animals were placed on 1 mg/ml gentamicin sulfate (MWI) autoclaved water for 3 d prior to and 2 weeks following irradiation. BMCs from donor animals were collected by flushing bone marrow and lysing red blood cells before resuspending cells in sterile PBS. Within 24 h of the second dose of irradiation, recipient animals were reconstituted with 4 million BMCs via tail vein injection. CCI surgery was performed 1 month after reconstitution.
CCI injury
Male mice age 8–12 weeks were anesthetized with an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg), then positioned in a stereotactic frame. Body temperature was maintained at 37°C, monitored with a rectal probe and a controlled heating pad. A 4 mm craniotomy was made with a portable drill over the right parietal-temporal cortex (−2.5 mm A/P and 2.0 mm lateral from bregma). Moderate CCI was induced with an eCCI-6.3 device (Custom Design and Fabrication) using a 3 mm impact tip at an angle of 70°, 5.0 m/s velocity, 2.0 mm impact depth, and 100 ms dwell period. The incision was closed with Vetbond tissue adhesive (3 M), and the animals were placed into a heated cage and monitored every 20 min until they fully recovered from anesthesia. Animals also received subcutaneous injection of Buprenorphine SR (1 mg/kg) after surgery.
Rotarod
Gross motor coordination and function was evaluated by Rotarod (Columbus Instruments) testing. Ten-week-old animals were trained for 4 consecutive days; then, a baseline measurement was taken on the fifth day using an initial velocity set at 5 rpm with an acceleration of 0.1 rpm/s. Each animal underwent four trials per day with a 2 min rest between each trial. The average time of the four trials was used for analysis. Each animal's performance was compared with its baseline measurement, and the average performance for all animals was reported.
Novel object recognition (NOR) test
Mice were habituated in the experimental room for 1 h before testing. All equipment was cleaned with 70% ethanol after each mouse completed the assay. Cognitive short-term nonspatial hippocampal-mediated memory (Clark et al., 2000; Lueptow, 2017) was tested using novel object recognition (NOR) as described previously (Okyere et al., 2018). All objects used were custom designed and 3D printed either in the shape of a cone or paraboloid 5 inches tall and base diameter of 1.5 inches. Briefly, mice were habituated to an empty NOR box 1 d prior to the experimental baseline. Mice were introduced to two identical objects, 14 cm high, spaced 20 cm apart, and 10 cm from the wall (Day −1, 19, 35 d postinjury) and then short-term memory was tested 1 h later by replacing one familiar object with a new or “novel” object in a 40 cm × 40 cm arena surrounded by 20 cm high walls covered by clear Plexiglas. Time of exploration of the familiar and novel object was recorded over 5 min by a video camera positioned over the chamber and later manually scored by an observer blinded to the treatment. Exploration was defined as touching the objects with the nose and/or forepaws. Preference for the novel object was calculated as follows: X = (b/(a + b)) * 100, where X is the preference index, b is the time spent exploring the novel object, and a is the time spent exploring the identical object.
Adhesive tape removal test
Mice were habituated in the experimental room for 1 h before testing. All equipment was cleaned with 70% ethanol after each mouse completed the assay. Somatosensory and motor functions were tested using the adhesive tape removal test as previously described (Bouet et al., 2009). This test allows for the examination of long-term functional deficits and recovery. Briefly, mice were habituated to an empty, transparent testing box for 3 consecutive days prior to the experimental baseline testing. After a 60 s habituation period in the testing box, mice were scruffed, and their forepaws were pressed with equal pressure onto two previously cut adhesive tape strips (∼0.3 cm × 0.4 cm) before being placed back into the testing box. Immediately, a four-channel timer was started with four different end points: contact time of left paw to mouth, contact time of right paw to mouth, time to remove left paw tape, and time to remove right paw tape. Thereafter, the mouse was removed from the testing box and returned to its home cage. Functional deficits on the contralateral side to brain injury were calculated as follows: X = (b−a)/(a + b) * 100, where X is the asymmetry score, b is the time to remove the right paw tape, and a is the time to remove the left paw tape.
Histology and immunofluorescence staining
Mice were deeply anesthetized with an intraperitoneal injection of ketamine (500 mg/kg) and xylazine (10 mg/kg) 24 h or 14 d post-CCI at the indicated time points and tissue was fixed with transcardial perfusion. Animals were perfused with an ice-cold PBS followed by an ice-cold 4% paraformaldehyde pH 7.4 in PBS. Brains were removed and cryoprotected in 22.5% sucrose in 1× PBS at 4°C. Tissue was frozen on dry ice and embedded in O.C.T. (Tissue-Plus O.C.T. Compound) and then coronally sectioned (30 µm thickness) through the lesion site (−1.1 to −2.6 mm posterior to the bregma) using a cryostat (CryoStar NX50, Thermo Scientific). Serial sections 300 µm apart were stained with cresyl violet (Electron Microscopy Sciences).
To identify cells undergoing apoptosis, slides were fixed in 10% formalin (Fisher Chemicals) for 5 min, washed with 1× PBS, permeabilized in 2:1 ethanol:acetic acid at −20°C for 10 min and 0.4% Triton X-100 (Fisher Scientific) for 10 min, and then washed with 1× PBS and TUNEL stained according to the manufacturer's suggestions (DeadEnd Fluorometric TUNEL System). To label neurons, slides were blocked in 0.2% Triton X-100, 2% cold water fish gel (Sigma) in PBS for 30 min, then stained with Nissl (1:100 in block, NeuroTrace 530/615 Red Fluorescence Nissl, Invitrogen), and mounted with DAPI Fluoromount-G (SouthernBiotech). Representative images were taken at 20× on a Nikon C2 confocal microscope. Maximum intensity projections were created in Nikon NIS-Elements.
For immunohistochemistry, slides were either incubated with 0.4% Triton X-100 in PBS for 30 min and then blocked in 10% NGS, 1% BSA in PBS for 1 h, or were blocked in 0.2% Triton X-100, 2% cold water fish gel in PBS for 30 min. Slides were then incubated at room temperature overnight in primary antibody diluted 1:100–1:200 in block. Slides were washed with 1× PBS, then incubated for 1.5 h at room temperature in the appropriate fluorescently-conjugated secondary antibody (Thermo Scientific) diluted 1:250 in PBS. After additional washes in 1× PBS, sections were mounted with DAPI Fluoromount-G (SouthernBiotech).
To identify IgG deposition, slides were blocked in 0.2% Triton X-100, 2% cold water fish gel in PBS for 30 min, then incubated at room temperature in fluorescently-conjugated anti-mouse secondary antibody (Thermo Scientific) diluted 1:250 in PBS. After additional washes in 1× PBS, sections were mounted with DAPI Fluoromount-G.
Antibodies
The following antibodies were used for this study: Iba1 (Wako, catalog #NC9288364), Arg1 (Abcam, catalog #ab60176, CCR2 (Abcam, catalog #ab216863), Ly6G (Abcam, catalog #ab25377) (see note below about specificity), and CD31 (R&D systems, catalog #AF362).
Estimating lesion size, IgG deposition, and cell numbers
The StereoInvestigator software (MicroBrightField) and an Olympus BX51TRF motorized microscope (Olympus America) was used by a blinded investigator to estimate lesion size and number of TUNEL+/Nissl+, GFP+/Iba1+/Arg1+, and CCR2+/Ly6G+. Five coronal serial sections spaced 300 µm apart around the lesion center were used for each animal. For lesion volume (mm3), cresyl violet stained slides (described above) were visualized at 4× under brightfield illumination using StereoInvestigator's Cavalieri estimator. A grid with 100 µm spacing was overlaid on the ipsilateral lesion site and markers were placed over contused tissue, identified by reduced staining intensity, abnormal morphology, and pyknotic neurons as previously described (Fritsch et al., 2022). The Cavalieri estimator software used the number of markers, section thickness, section interval, and number of sections to estimate volume of the lesion. The volume of IgG deposition was estimated using the same method, but with the appropriate fluorescent filter.
To approximate cell numbers, StereoInvestigator Optical Fractionator was used with a grid size of 500 µm × 500 µm. A contour was drawn around the lesion site at 4×, then approximately 100 randomized sites were evaluated at higher magnification for each animal. The section thickness was estimated every five sites to improve cell count estimate accuracy. The number of cells per contour, average estimated section thickness, section interval, and the number of sections were used to estimate the number of cells within the lesion. For estimating the number of apoptotic (TUNEL+) cells and apoptotic neurons (TUNEL+Nissl+), a counting from of 100 µm × 100 µm at 60× magnification was used. For estimating the number of infiltrating peripheral immune cells, a 150 µm × 150 µm contour and 40× magnification was used. Ly6G (Abcam, catalog #ab25377) has been found to cross react with Ly6C, so CCR2+/Ly6G+ double positive cells were considered monocytes and counted as CCR2+. All analysis was completed by a blinded investigator.
Microglial reactivity assessment
Two 40× magnificent z-stack Nikon C2 confocal images were utilized for each animal to assess microglial reactivity. Images were taken along the medial and lateral edge of the lesion site on the tissue section with the largest lesion area. Maximum intensity projection images were loaded into FIJI and Iba1+ cells were manually counted using the FIJI Cell Counter. Cells were classified as either mostly ameboid/activated (defined as Iba1+ cells with 3 or fewer processes) or mostly ramified/resting (Iba1+ cells with more than three processes). Approximately 25 cells were counted for each animal across two random images. Data is presented as the percentage of counted cells in each category. Images were taken and analyzed by a blinded investigator.
Flow cytometry
WBCs were collected as described and resuspended in PBS with 1% BSA. For evaluating peripheral immune cell infiltration in bone marrow chimera mice, animals were deeply anesthetized with a ketamine (500 mg/kg)/xylazine (10 mg/kg) cocktail and hand perfused with 50 ml of ice-cold PBS to remove blood. Cortical tissue was rapidly dissected in ice-cold PBS and dissociated in papain following the Worthington Papain Dissociation System protocol (Worthington Biochemical Corporation). Cells were resuspended in PBS with 1% BSA. GFP+ single cells were identified with a BD FACSAria Fusion Cell Sorter (BD) using the 530 nm laser. Analysis was performed using FlowJo software (FlowJo LLC). Data are given as the percentage of single cells that were GFP+.
WBC isolation and complete blood counts
Mice were deeply anesthetized with a ketamine (500 mg/kg)/xylazine (10 mg/kg) cocktail and approximately 1 ml of whole blood was collected and immediately mixed with 100 µl of 500 mM EDTA. For complete blood counts (CBCs), samples were processed by Virginia Tech Animal Laboratory Services. For WBC isolation, red blood cells were lysed with ACK Lysis Buffer (Quality Biological) twice, then washed with 1× PBS and resuspended for the appropriate downstream application.
Enriched microglia isolation
The microglia were isolated and enriched as previously described (Fritsch et al., 2022). Briefly, mice were deeply anesthetized with a ketamine (500 mg/kg)/xylazine (10 mg/kg) cocktail and hand perfused with cold PBS to remove blood. Following the removal of the brain, the cortical tissue was rapidly dissected in ice-cold PBS and dissociated in papain for 45 min at 37°C using the Worthington Papain Dissociation System (Worthington Biochemical Corporation), then resuspended in RPMI media with 1% pen-strep (Gibco), 10% FBS, and 2 mM L-glutamine (Gibco). Cells were plated on petri dishes and incubated for 1 h at 37°C. Microglia have intrinsic in vitro properties to allow them to stick to the dish, while nonadherent cells were washed off with 1× HBSS (Woolf et al., 2021).
RNA sequencing
Animals were deeply anesthetized with a ketamine (500 mg/kg)/xylazine (10 mg/kg) cocktail and hand perfused with cold PBS to remove blood. Cortical tissue was rapidly dissected in ice-cold PBS and stored in RNAlater (Sigma-Aldrich) until RNA extraction. RNA was isolated with Direct-zol RNA Miniprep Plus (Zymo Research Corp.) per the manufacturer's protocol. Isolated RNA was sent to MedGenome (MedGenome Inc.) for RNA sequencing using the Illumina TruSeq stranded mRNA kit for library preparation and the NovaSeq system for sequencing. RNA-seq data has been deposited on GEO depository: GSE223784.
Bases with quality scores less than 30 and adapters were trimmed from raw sequencing reads by Trim Galore (v 0.6.4). After trimming, only reads with length greater than 30 bp were mapped to mm10 by STAR (v 2.7.1a) with average mapping efficiency 88.5%. Raw counts and normalized counts for each gene were outputted by RSEM (v1.2.28). The raw counts were used to identify differentially expressed genes by DESeq2 (v 1.36.0). Only genes with an average transcripts per million (TPM) greater than 5 in at least one group, adjusted p-value less than 0.05 and at least 1.5-fold change were considered as differentially expressed genes. All the differentially expressed genes were used for GO enrichment with R package clusterProfiler (v4.4.4) and org.Mm.eg.db (v3.15.0). The top 10 most significant biological processes (BP) terms were used to generate a GO circle plot with R package GOplot (v1.0.2).
Bone marrow-derived macrophages
Following euthanasia via isoflurane, the femur and tibia bones were isolated. The bones were cleaned, and the ends were cut off and then flushed with complete RPMI media containing 10% FBS, 1% penicillin/streptomycin, and 1% L-glutamine. Cells were spun down and resuspended in ACK Lysis Buffer (Thermo Scientific) to lyse red blood cells and washed with sterile PBS. Three million bone marrow cells were plated per well in 6-well tissue culture treated plates with RPMI containing 10 ng/ml M-CSF (PeproTech). Three days after plating, a fresh media was added. Morphology was verified after 6 d in culture; then adherent cells were treated with 50 μg/ml DMXAA (Sigma-Aldrich) dissolved in DMSO or vehicle (DMSO) for 3 h. Cells were washed with cold HBSS, then directly resuspended in TRIzol Reagent (Invitrogen) for RNA isolation.
Real-time qPCR
cDNA was synthesized using an iScript cDNA Synthesis Kit (Bio-Rad), then qPCR was run on the CFX96 System (Bio-Rad) using SYBR Green PCR Master Mix (Bio-Rad), 10 ng of cDNA and 0.4 mM of each primer set. Each sample was run in technical triplicates and expression levels were normalized to GAPDH. Fold change was calculated by comparative CT method (Schmittgen and Livak, 2008) using CFX Maestro software (Bio-Rad). Primer efficiency was determined using a 4-point log concentration curve.
Primers are as followed: CCL7: GTC CCT GGG AAG CTG TTA TCT TCA, TGC TAT AGC CTC CTC GAC CC; GAPDH: AAT GTG TCC GTC GTG GAT CTG A, AGA TGC CTG CTT CAC CAC CTT CTT; GJA1: CGG AAG CAC CAT CTC CAA CT, CCA CGA TAG CTA AGG GCT GG; IFNB1: AAC TCC ACC AGC AGA CAG TG, GGT ACC TTT GCA CCC TCC AG; CCL2 (MCP-1): TCA CCT GCT GCT ACT CAT TCA CCA, TAC AGC TTC TTT GGG ACA CCT GCT; MBP: TCA CGA CCC CGG AAC ATA GT, TGG GGT GTT CAA GAG TGG TG; TMEM173 (STING): GCC TTC AGA GCT TGA CTC CA, GTA CAG TCT TCG GCT CCC TG; TMEM119: AGA CCC TTC TGC TTC CCC TT, CCC AGT ATG TGG GGT CAC TG; CDH5 (VE-cadherin): AGG ACA GCA ACT TCA CCC TCA, AAC TGC CCA TAC TTG ACC GTG; RBFOX3 (NeuN): CAC TCT CTT GTC CGT TTG CTT C, CTG CTG GCT GAG CAT ATC TGT A; CCR2: GGG CTG TGA GGC TCA TCT TT, TGC ATG GCC TGG TCT AAG TG; CX3CR1: GTG AGT GAC TGG CAC TTC CTG, AAT AAC AGG CCT CAG CAG AAT C.
Statistical analysis
Data were analyzed with GraphPad Prism 9. A student's two-tailed t test was used for comparison of two groups, and a one-way or two-way ANOVA with Tukey's or Šidák's multiple-comparisons test was used to compare more than two groups as appropriate. Differences were considered statistically significant at p < 0.05. Data are reported as mean ± SEM with n values given in the figure legend.
Results
Loss of STING in the CNS or peripheral immune cells confers neuroprotection
Our work and others demonstrated that global loss of STING results in reduced cytokine production, reduced cell death, and smaller lesion size up to 2 weeks postinjury in a CCI mouse model (Abdullah et al., 2018; Fritsch et al., 2022). The bone marrow adoptive cell transfer data suggested that the IFNAR signaling in the periphery contributes to CCI outcome (Karve et al., 2016), so we quantified STING mRNA expression and found it was elevated in the blood 2 h postinjury (t(4) = 6.005, p = 0.0039, unpaired t test) (Fig. 1A). To clarify whether brain-resident or infiltrating peripheral immune cells participate in the detrimental STING-mediated inflammatory milieu after TBI, we generated bone marrow chimeric animals (Fig. 1B,C). Irradiated recipient STING KO and wild-type (WT) animals received bone marrow cells (BMCs) from GFP+ STING KO or GFP+ WT donors, generating four groups (WTWTBMC, KOKOBMC, WTKOBMC, and KOWTBMC).
Genetic loss of STING in adoptive cell transfer rodent models display neuroprotection after controlled cortical impact injury. A, qPCR for STING mRNA expression for WBCs isolated from WT animals 2 h after sham or CCI surgery. B, Cartoon diagram developing bone marrow chimeric animals. Panel made with BioRender. C, Cartoon diagram of experimental timeline. Following irradiation, recipient animals received 4 million bone marrow cells from GFP+ donor animals (WT or STING KO) via tail vein injection. Animals underwent CCI surgery 28 d later to allow time for reconstitution of the immune system. D, Flow cytometry of white blood cells confirming successful reconstitution of the immune system in chimeric animals. Plots are representative of one animal. Mean percentage of GFP+ cells 94.78%; ±0.88% SD; n = 4. E, Representative confocal images from naive, uninjured chimeric animals 1 month after receiving bone marrow. Tissue was stained for CD31 (red) to indicate vasculature and DAPI (blue) to show nuclei. E1,E2, Insets from (E) at the indicated locations. Arrows = examples of GFP+ cells localized in vessels. Scale bar = 1 mm, Inset scale bar = 10 μm. F, Flow cytometry of central nervous system cells in naive, uninjured chimeric GFP+ reconstituted animals. Plots are representative from one animal. Mean percentage of GFP+ cells detected in the brain parenchyma 0.22%; ±0.037% SD; n = 4. G, Quantification of lesion volume 24 h after injury. H, Representative images of sectioned cresyl violet stained brains 24 h after CCI from bone marrow chimeric animals. Scale bar = 1 mm. Each dot represents one animal. *p < 0.05; **p < 0.01; ****p < 0.0001. Error bars = SEM.
To verify successful reconstitution of the peripheral immune system, WBCs from chimeric animals were removed and sorted via flow cytometry 28 d after receiving GFP+ donor bone marrow. Approximately 94.78% (±0.88% SD, n = 4) of the isolated WBCs were GFP+, confirming reconstitution of the recipient animal's peripheral immune system (Fig. 1D). To determine whether GFP+ immune cells infiltrated the brain or replaced resident microglia during the procedure, naive chimeric animals were euthanized 28 d after reconstitution to check for GFP+ cells in the brain. Naive chimeric animals displayed GFP+ cells retained in the vasculature, but did not appear to penetrate or integrate into the parenchyma (Fig. 1E). When assessed by flow cytometry, only 0.22% (±0.037% SD, n = 4) of brain cells were GFP+ (Fig. 1F). These GFP+ were most likely retained after perfusion, which likely explains the small percentage of detected GFP+ cells by flow cytometry.
Next, we evaluated chimeric animals for neuroprotection after injury to determine if STING signaling in the periphery contributed to outcome. Chimeric KOKOBMC animals had significantly reduced lesion size compared with WTWTBMC, and mice lacking STING in either the brain (KOWTBMC) or hematopoietic cells (WTKOBMC) had smaller lesions than WTWTBMC animals (F(3,25) = 12.12, p = 0.00004339, ANOVA) (Fig. 1G,H). Neither WTKOBMC or KOWTBMC group reached the full level of protection seen in the KOKOBMC animals (Fig. 1G,H). These data suggest that both brain and peripheral STING signaling contributes to tissue loss after TBI and a loss of STING in either compartment results in neuroprotection.
Peripheral immune cell infiltration is driven in part by the CNS and monocytes through STING
Given that STING signaling is known to promote the production of chemokines such as monocytes chemoattractant protein 1 (MCP1) and CXCL10 that attract immune cells in other disease contexts (Jin et al., 2013; Vonderhaar et al., 2021), we next sought to determine whether loss of STING in chimeric animals would affect peripheral immune cell infiltration at the injury site. Ipsilateral cortical tissue was microdissected and prepared for flow cytometry to detect the percentage of GFP+ cells. WTWTBMC animals had significantly higher levels of peripheral immune cell infiltration at 24 h postinjury when compared with animals with either KO BMCs or who were a KO recipient (F(3,12) = 22.92, p = 0.0000295, ANOVA) (Fig. 2A,B). Interestingly, both groups of crossed genotypes had comparable levels of infiltration compared with KOKOBMC (Fig. 2A,B). To confirm these findings, we performed stereological counting of GFP+ cells from the ipsilateral hemisphere 24 h postinjury. In support of the flow cytometry data, we saw that chimeric mice lacking STING globally (KOKOBMC) or in either the brain (KOWTBMC) or periphery (WTKOBMC) had similarly reduced infiltration (F(3,19) = 22.54, p = 0.00000178, ANOVA) (Fig. 2C,D). Changes in blood–brain barrier (BBB) permeability may account for the similar infiltration patterns between groups. Staining for IgG deposition, we observed comparable reductions in BBB permeability indicated by reduced IgG staining for KOKOBMC, KOWTBMC, and WTKOBMC animals compared with WTWTBMC animals (F(3,20) = 7.741, p = 0.0013, ANOVA) (Fig. 2E), suggesting that animals lacking STING had reduced BBB breakdown after injury.
Loss of STING in donor or recipient mice reduces neutrophil infiltration 24 h after injury. A, Quantification of GFP+ cells identified by flow cytometry, given as a percentage of single cells 24 h after injury. B, Cortical brain tissue from bone marrow chimeric animals was isolated 24 h after injury and dissociated for flow cytometry. Representative plots are combined data from four representative animals for each group. Gating scheme included debris removal (B1) and selection of single cells (B2, B3) prior to gating for GFP+ cells (B4). GFP intensity gates were set based on white blood cells from GFP+ control animals. C, Stereological quantification of GFP+ cells in the lesion site 24 h after injury. D, Representative confocal images from GFP+ chimeric animals 24 h after injury in the perilesional cortex. DAPI counterstain is in blue. Scale bar = 100 μm. E, Quantification of IgG deposition 24 h after injury. F, Stereological quantification of GFP+/Iba1+ or GFP+/Iba1− in the lesion site 24 h after injury. G, Percentage of GFP+/Iba1+ and GFP+/Iba1− normalized to the total number of immune cells. H, Stereological quantification of GFP+/Iba1+/Arg1+ or GFP+/Iba1+/Arg1− in the lesion site 24 h after injury. Each dot represents one animal. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Error bars = SEM.
To determine if this pattern of infiltration may be explained by an alteration in the number of WBCs between genotypes before injury, CBCs were performed with STING KO and WT mice. However, WBC counts were not altered between genotypes, and while there was a significant difference in the number of platelets and monocytes (Table 1), both values were within normal range (O’Connell et al., 2015). This indicated that it was unlikely that STING greatly shifted WBC composition prior to injury and activation.
CBC data from naive WT and STING KO animals
To determine the types of cells infiltrating the injury site 24 h postinjury, we performed stereological counting of cells in serial sections through the lesion. We detected GFP+ cells that were either Iba1+ (monocyte/macrophage lineage) or Iba1−, with neutrophil infiltration predominating the GFP+ population as previously reported for this time point (Hallenbeck et al., 1986; Soares et al., 1995; Fig. 2F). While the number of GFP+Iba1+ cells did not greatly differ between chimeric groups, except between WTWTBMC and KOKOBMC mice, GFP+Iba1− cells were significantly reduced in KOKOBMC, KOWTBMC, and WTKOBMC mice compared with WTWTBMC mice (F(3,19) = 4.344, p = 0.0172, ANOVA), (F(3,19) = 19.42, p = 0.0000052, ANOVA) (Fig. 2F). Considering that the proportion of monocytes and neutrophils were unchanged (Fig. 2G), this observed change occurred due to the total reduction in infiltrating cells overall in these groups.
Co-labeling with arginase 1 (Arg1) was used to evaluate the inflammatory phenotype of infiltrating immune cells; monocytes of an anti-inflammatory state tend to overexpress Arg1, and thus we identified GFP+Iba1+Arg1− cells as more proinflammatory and GFP+Iba1+Arg1+ cells as more anti-inflammatory (Munder et al., 1999, Viola et al., 2019). Although monocytes only made up 20% of all infiltrating peripheral immune cells at this time point (Fig. 2G), the KOKOBMC and KOWTBMC groups had fewer number of GFP+Iba1+Arg1− cells than WTWTBMC mice indicating that these monocytes may be less proinflammatory (F(3,19) = 4.906, p = 0.0109, ANOVA) (Fig. 2H).
Neutrophils express STING at very low levels and lack upstream cGAS for canonical activation (Xia et al., 2015), these changes in peripheral immune cell infiltration are likely driven by the resident cells in the CNS and circulating monocytes (Fig. 2A,C,F). We also observed STING mRNA increased 2 h after injury in the peripheral blood (Fig. 1A), so we isolated bone marrow-derived macrophages (BMDMs) to profile their response in producing chemoattractant cytokines. Treating BMDMs with the STING agonist DMXAA resulted in a robust increase in IFNβ expression in WT cells, which was absent in STING KO BMDMs, as anticipated (F(3,16) = 579.4, p < 0.00000001, ANOVA) (Fig. 3A). WT BMDMs had significant increases in the mRNA expression of the chemokines MCP-1 and CCL7, but this response was not upregulated in KO BMDMs (F(3,14) = 287.0, p < 0.0000001, ANOVA), (F(3,14) = 127.0, p < 0.00000001, ANOVA) (Fig. 3B,C). The expression of receptors for these chemokines, CCR2 and CXC3CR1, were unchanged with DMXAA treatment and did not differ between genotypes (Fig. 3D,E). Taken together, these data suggest that STING signaling in BMDMs upregulates cytokines that have the ability to greatly influence the types and number of myeloid cells entering the brain.
BMDMs upregulate chemoattractant cytokines after STING activation. A–E, qPCR of mRNA expression for (A) IFNB1 (B) MCP1 (C) CCL7 (D) CCR2 and (E) CX3CR1 from WT and STING KO BMDMs treated with vehicle or 50 ng/ml DMXAA for 3 h. Each dot represents one experimental replicate. *p < 0.05; ****p < 0.0001; ns = not significant. Error bars = SEM.
Loss of microglial-specific STING is neuroprotective
Chimeric KOWTBMC mice also displayed significant protection after injury, so we sought to clarify which cell type in the brain primarily mounts the STING response. Our previous work and others suggested that microglia have high levels of basal STING expression (Zhang et al., 2014, 2016; Fritsch et al., 2022); therefore, we developed a tamoxifen-inducible, microglia-specific conditional STING KO mouse (STINGfl/fl CX3CR1CreER/+; cKO). CX3CR1 is a chemokine receptor that is present on many myeloid-derived cells, including monocytes, natural killer cells, dendritic cells, and microglia (Jung et al., 2000). In this model, tamoxifen administration conditionally knocks out STING in CX3CR1-expressing cells. Due to the turnover of peripheral immune cells, but not microglia (Réu et al., 2017), repopulation of peripheral CX3CR1-expressing cells 1 month after tamoxifen administration generates a STING conditional KO specific to microglia.
Enriched populations of microglia isolated from uninjured cKO animals 1 month after tamoxifen treatment showed approximately a 75-fold decrease in STING mRNA relative to control animals (STINGfl/fl), indicating that STING was ablated from the microglia (t(11) = 5.552, p = 0.0002, unpaired t test), (t(7) = 2.419, p = 0.046141, unpaired t test) (Fig. 4A,B). To ensure that WBCs recovered STING expression after tamoxifen injections, WBCs were isolated from naive, uninjured animals 1 month after tamoxifen administration. cKO animals displayed no change in STING mRNA in peripheral immune cells between genotypes indicating STING loss at this 1 month time-point post tamoxifen administration was specific to microglia (Fig. 4C). To evaluate whether loss of microglia-specific STING was sufficient for the neuroprotective effects seen in global or chimeric KOWTBMC mice, lesion volume was assessed 24 h after injury. We observed a significant reduction in lesion volume in cKO animals relative to STINGfl/fl and CX3CR1CreER/+ controls at this acute time point (F(2,16) = 11.62, p = 0.0008, ANOVA) (Fig. 4D,E). Furthermore, cKO animals showed a significant reduction in apoptotic cell death for Nissl+ cells (neurons) (F(2,16) = 7.851, p = 0.00421, ANOVA) (Fig. 4F,G). Additionally, the density of apoptotic neurons was significantly reduced in the cKO animals (Fig. 4F), suggesting neurons are protected by the loss of microglial STING. Considering that STINGfl/fl and CX3CR1CreER/+ controls displayed similar lesion volumes and TUNEL+Nissl+ cells, we used CX3CR1CreER/+ animals for subsequent comparisons as the estrogen responsive Cre was knocked into exon 1 (Parkhurst et al., 2013). cKO animals also showed significantly reduced lesion volume 14 d after injury when compared with controls, suggesting that loss of microglial STING is still beneficial (t(12) = 4.799, p = 0.0004, unpaired t test) (Fig. 4H,I).
Loss of microglial-specific STING expression reduces lesion volume and decreases neuronal cell death. A, Microglia were isolated from naive cKO (STINGfl/fl, CX3CR1CreER/+) and control STINGfl/fl animals 1 month after the final dose of tamoxifen was administered. qPCR for STING mRNA. B, qPCR for cell-type specific genes was used to confirm microglial enrichment in (C). n = 5. Gene expression quantified was used to detect microglia (Tmem119), astrocytes (Gja1), oligodendrocytes (Mbp), neurons (Rbfox3), and endothelial cells (Cdh5). C, WBCs were isolated from naive cKO and STINGfl/fl animals 1 month after the final dose of tamoxifen was administered. qPCR to quantify STING mRNA. D, Quantification of lesion size 24 h after CCI in STINGfl/fl, CX3CR1CreER/+, and cKO animals. E, Representative images of CX3CR1CreER/+ and cKO sectioned cresyl violet stained brains 24 h after injury. Scale bar = 1 mm. F, Quantification of apoptotic neurons identified as cells that were TUNEL+ Nissl+ 24 h postinjury. G, Representative confocal images of CX3CR1CreER/+ and cKO 24 h after injury. Slides were stained for TUNEL (white), Nissl (red), and DAPI (blue). Scale bar = 1 mm. H, Quantification of lesion size 14 d after CCI in CX3CR1CreER/+ and cKO animals. I, Representative images of CX3CR1CreER/+ and cKO sectioned cresyl violet stained brains 14 d after injury. Scale bar = 1 mm. Each dot represents one animal. *p < 0.05; **p < 0.01; ***p < 0.001; ns = not significant. Error bars = SEM.
Next, to determine whether this reduction in lesion volume and neuroprotection in cKO animals corresponded to changes in functional outcome, we assessed gross motor deficits with two behavioral paradigms. First, we performed Rotarod testing in cKO and CX3CR1CreER/+ animals for gross motor coordination deficits. Baseline (pre-injury) performance was comparable between groups (CX3CR1CreER/+ baseline time, 104.8 ± 6.6 s, n = 12; cKO baseline time, 109.0 ± 6.9 s, n = 10). Four days after injury, cKO animals had significantly reduced motor deficits than control animals (genotype × time, F(3,60) = 4.000, p = 0.0116; time, F(3,60) = 21.26, p = 0.000000002, two-way ANOVA) (Fig. 5A), suggesting loss of microglial STING offers functional improvement. No significant difference in performance was seen at more chronic time points, though this may be because both groups generally have substantial improvement on Rotarod testing within 1 week of injury (Onyszchuk et al., 2007). Considering gross motor deficits were improved in the sub-acute time point, we performed the adhesive tape removal test, which requires fine motor skills and sensorimotor feedback. Again, baseline performance was comparable between groups, but was not different (CX3CR1CreER/+ baseline asymmetry score: 7.8 ± 7.9, n = 7; cKO baseline asymmetry score: 5.9 ± 15.5, n = 9). cKO animals showed improved recovery to baseline, reflected by analyses revealing significance between time and genotype factor (time, F(4.251, 59.51) = 9.043, p = 0.000006; genotype, F(1,14) = 10.18, p = 0.0065, two-way ANOVA) (Fig. 5B). We also evaluated short-term memory recognition and retention, but no distinguishable differences were seen between groups in the NOR test across all time points (Fig. 5C). Together, these data demonstrate that loss of microglial STING confers neuroprotection and functional recovery to motor deficits after brain injury.
Functional improvements in motor coordination are exhibited by microglial-specific STING KO mice after injury. A, Rotarod performance of CX3CR1CreER/+ and cKO animals 4, 7, and 14 d after injury. n = 12 for CX3CR1CreER/+ and 10 for cKO. B, Adhesive tape removal test performance of CX3CR1CreER/+ and cKO animals at 1, 3, 7, 14, 21, 28, and 35 d postinjury. n = 7 for CX3CR1CreER/+ and 9 for cKO. C, Novel object recognition performance of CX3CR1CreER/+ and cKO animals. Preference index, the ratio of preference for the novel object to total time exploring all objects x 100, was evaluated between following CCI injury at 1, 19, and 35 days postinjury. n = 8 for CX3CR1CreER/+ and 8 for cKO. Each dot represents one animal. **p < 0.01. Error bars = SEM.
We previously reported that 2 h after injury, the Type I IFN transcriptional response was active in the injured cortical tissue with concomitant increases in STING mRNA expression (Fritsch et al., 2022). To assay for transcriptional changes to the injured cortical tissue 2 h postinjury, we performed a whole transcriptome RNA-seq expression analysis to understand how the loss of STING in microglia affected the inflammatory milieu. RNA-seq expression analysis revealed that many genes did not differ in the uninjured (contralateral) hemisphere between genotypes with only one gene, Galnt2, reaching significant downregulation in cKO mice (Fig. 6A, Extended Data 6-1). However, a pronounced upregulation of transcriptional changes occurred after injury regardless of genotype; using p < 0.05 and a threshold of 1.5-fold change in expression as a cutoff for analysis (Fig. 6A–C, Extended Data 6-2 and 6-3). Our analysis identified over 240 differentially expressed genes for cKO and 222 for CX3CR1CreER/+ when comparing the uninjured and injured cortical tissue between genotypes, and surprisingly, many of the top regulated genes (Cxcl2, Ccl3, Ccl4, Il1a, Tnf), producing cytokines and chemokines, and the downregulated gene (Tmem119), involved in microglial identity, were the same (Fig. 6B,C). There were very few downregulated genes that reached our cutoff for analysis. Further analysis of the expression profile of these upregulated genes changed 2 h postinjury with gene ontology (GO) revealed a significant increase in expression of genes related to innate immune processes and the chemotaxis and recruitment of peripheral immune cells to the injured tissue (Fig. 6D,E). When analyzing the upregulated genes from injured tissue compared with the contralateral side between genotypes, many genes overlapped (51.8%) with differentially expressed genes in the cKO (27.1%) and CX3CR1CreER/+ (21.1%) represented lower percentages (Fig. 6F). Due to the overlap of top genes between genotypes, we decided to check the levels of STING mRNA 2 h postinjury and found a significant increase in STING signaling with both CX3CR1CreER/+ and cKO genotypes (F(3,13) = 10.09, p = 0.0011, ANOVA) (Fig. 6G). Considering microglia only comprise 5–10% of the brain (Lawson et al., 1990), these data suggest STING signaling is active and upregulated by other resident CNS cells in the brain even upon the loss of STING in the microglia (Fig. 4A).
Transcriptomic profile of microglial-specific STING KO mice reveal significant upregulation of chemotactic and peripheral immune cell signaling signatures acutely after injury. A–C, Volcano plots of significantly upregulated (orange dots) and down regulated (blue dots) in STING-ablated microglia 2 h postinjury with comparisons between CX3CR1CreER/+ and cKO between (A) contralateral hemispheres (B) CX3CR1CreER/+ contralateral (uninjured) and ipsilateral (injured) hemispheres and (C) cKO contralateral (uninjured) and ipsilateral (injured) hemispheres. D,E, Genes from (D) CX3CR1CreER/+ and (E) cKO mice revealed GO terms associated with chemotaxis, myeloid and leukocyte migration, response to molecules of bacterial origin, and ERK1/2 signaling. F, Venn diagram showing overlap of significant upregulated genes between genotypes of CX3CR1CreER/+ and cKO animals. G, qPCR of STING expression from injured ipsilateral or contralateral cortices from CX3CR1CreER/+ and cKO 2 h after injury. Each dot represents one animal. *p < 0.05; **p < 0.01; ns = not significant. Error bars = SEM. Gene lists comparing contralateral hemispheres 2 h postinjury between cKO and CX3CR1CreER/+ animals are supplied in Extended Data 6-1, while separate gene expression lists comparing ipsilateral and contralateral hemispheres for CX3CR1CreER/+ mice and cKO mice are provided in Extended Data 6-2 and 6-3, respectively.
Figure 6-1
Differentially expressed gene lists comparing contralateral hemispheres 2 hours post injury between CX3CR1CreER/+ and cKO genotypes. n=3 per genotype. Download Figure 6-1, XLSX file.
Figure 6-2
Differentially expressed gene lists comparing ipsilateral and contralateral hemispheres 2 hours post injury for CX3CR1CreER/+ mice. n=3. Download Figure 6-2, XLSX file.
Figure 6-3
Differentially expressed gene lists comparing ipsilateral and contralateral hemispheres 2 hours post injury for cKO mice. n=3. Download Figure 6-3, XLSX file.
Our data suggested significant protective effects in cKO mice, but transcriptional signatures initially appear similar, so we performed additional analysis to understand if there were changes in magnitude between commonly shared genes or if specific biological pathways were more represented in one genotype. Similar changes in the magnitude of the transcriptional response for most of these overlapping upregulated genes 2 h after injury were seen, but genes that were the top regulated Cxcl2, Ccl3, Ccl4, Il1a, and Tnf appeared more so at higher levels in cKO mice (Fig. 7A). When examining differentially expressed genes in-depth by genotype, cKO animals highly upregulated hemoglobin subunit genes, Tlr2 which is a toll-like receptor involved in a STING-independent innate immune pathway, and the chemokines CCL7 and Cxcl10 (Fig. 7B). Considering our in vitro data suggested that the loss of STING inhibited a robust response upregulating CCL7 (Fig. 3C), these data suggest the loss of microglial STING could cause the upregulation of other innate immune pathways in other cell types. The loss of microglia STING also appeared to dampen other innate immune pathways considering the upregulation of Nlrp3, which is a component of the inflammasome, was seen in CX3CR1CreER/+ animals, but not found to be changed in cKO mice (Fig. 7C). GO analysis revealed the CX3CR1CreER/+ injured tissue mounted a response upregulating the IL-1β pathway, which was also absent in the cKO injured tissue (Fig. 7D).
Comparative and differential gene signatures reveal interleukin-1β signaling is blunted when microglial STING signaling is ablated. A–C, Heat maps of upregulated genes that are (A) similar or distinct to (B) CX3CR1CreER/+ or (C) cKO genotype when comparing the fold change in expression normalized to the contralateral hemisphere. Gray bar color indicates the gene was not significantly changed and counted as a differentially expressed gene. D, Gene ontology analysis dot plot of upregulated genes in (A–C). The color of each dot represents the false discovery rate of each term involved in the analysis. The size of each dot represents the gene counts of this term involved in the GO analysis.
Considering these changes in transcriptomics and our chimeric adoptive cell transfer experiments, we profiled the recruitment of immune cells 24 h postinjury in our cKO animals to determine if microglial STING signaling also influences this aspect of neuroinflammation. IgG deposition indicating BBB breach was attenuated in the cKO group as compared with controls (t(8) = 2.631, p = 0.0301, unpaired t test) (Fig. 8A,B). Using Ly6G (neutrophils) and CCR2 (monocytes) as markers for infiltrating immune cells (see Material and Methods), stereological quantification found a significant reduction in overall immune cell influx in cKO mice (t(8) = 3.751, p = 0.0056, unpaired t test) (Fig. 8C). The number of neutrophils (Ly6G+ cells) was significantly decreased in the injured cortices of cKO animals, though no significant changes were detected in the monocyte (CCR2+) population (t(8) = 2.366, p = 0.045562, unpaired t test) (Fig. 8D,F), and there was no skewing in percentages of Ly6G+ and CCR2+ cells at 24 h between groups (Fig. 8E). This indicated that STING signaling from microglia influenced neutrophil infiltration to the injury site. In contrast, WTKOBMC mice also displayed a reduction in neutrophil infiltration, where STING is intact in the CNS. This suggests that peripheral immune STING signaling may also in turn affect microglia.
Loss of microglial-specific STING expression blunts neutrophil infiltration. A, Quantification of IgG deposition 24 h after injury. B, Representative confocal images of CX3CR1CreER/+ and cKO 24 h after injury. Slides were stained for IgG (red) and DAPI (blue). Scale bar = 1 mm. C,D, Stereological quantification of (B) CCR2+ and Ly6G+ cells (total number of immune cells) or (C) CCR2+ Ly6G− and CCR2−Ly6G+ or in the lesion site 24 h after injury. E, Percentage of CCR2+Ly6G− and CCR2−Ly6G+ normalized to the total number of immune cells detected. F, Representative confocal images from cKO or CX3CR1CreER/+ animals 24 h postinjury stained with CCR2 (purple) and Ly6G (green). Scale bar = 100 μm. Insets are from the squared indicated locations. Red arrows = examples of Ly6G+ cells. Yellow arrows = examples of CCR2+ cells. Inset scale bar = 100 μm Each dot represents one animal. *p < 0.05; **p < 0.01. Error bars = SEM.
Next, we profiled the phenotypic microglia morphology at the injury site in both chimeric and conditional mouse models. Although microglia identity and categorization has shifted as our understanding deepens, microglial morphology still reveals insight into their function with more homoeostatic microglia display surveilling processes, while microglia in an “activated” state often show pronounced ameboid-like phenotypes (Paolicelli et al., 2022; Fig. 9A). Chimeric KOKOBMC, KOWTBMC, and WTKOBMC mice displayed a higher percentage of these less-branched microglia in the peri-lesioned region around the injury site at 24 h as compared with WTWTBMC (F(3,19) = 5.933, p = 0.004939, ANOVA) (Fig. 9B). This significant change in microglial morphology was also found at the same time point in cKO animals (t(8) = 4.690, p = 0.0016, unpaired t test) (Fig. 9C). These data suggest STING signaling from immune cells, as well as intrinsic STING signaling, influence microglia at acute time points after injury.
Ameboid microglial morphology is attenuated after the loss of STING in both chimeric and conditional KO models. A, Example confocal images showing microglia classified as activated (top) or ramified (bottom). Scale bars = 10 μm. B,C, Percentage of microglia in either an activated or ramified morphology in (B) bone marrow chimeric or (C) microglial-specific KO mice 24 h after injury. Each dot represents one animal. *p < 0.05; **p < 0.01; ***p < 0.001. Error bars = SEM.
Discussion
Here, we report that microglial or peripheral immune STING signaling cause detrimental outcomes with concomitant increases in peripheral immune cell and neutrophil infiltration and cause reactive microglia phenotypes immediately after brain injury. Our results in both adoptive cell transfer and microglia conditional KO models demonstrate that the loss of STING provides significant neuroprotection, dampening cortical tissue loss, which, in the microglial conditional KO, results in improved behavioral outcomes chronically. Both models provide evidence that STING-mediated neuroinflammation is triggered as early as 2 h after injury and shapes the neuroinflammatory milieu and that both STING activation from resident CNS and peripheral immune cells converge to contribute to injury outcome, as no adoptive cell transfer model was as protective as the KOKOBMC cohort.
Recent studies have demonstrated that the Type I IFN response that could be mediated by cGAS-STING signaling promotes detrimental neuroinflammation in the brain following injury (Karve et al., 2016; Abdullah et al., 2018; Barrett et al., 2020; Sen et al., 2020; Fritsch et al., 2022) and stroke (Gamdzyk et al., 2020; Li et al., 2020; Peng et al., 2020; Kong et al., 2022). However, these studies haven't implicated which cell types were involved, nor have they implicated that leukocyte and neutrophil infiltration in the injured brain were being influenced by STING signaling from microglia or from the WBCs themselves.
Surprisingly, we found neutrophil infiltration to be perturbed in both chimeric and microglia specific models where neutrophils express STING at very low levels and lack upstream cGAS for canonical activation (Xia et al., 2015). Human data demonstrate a robust display of neutrophil infiltration and activation after injury (Hausmann et al., 1999; Rhind et al., 2010; Liao et al., 2013), which is a result that is recapitulated in preclinical injury models (Clark et al., 1994; Roth et al., 2014). Neutrophils can form extracellular traps (NETs) made of chromatin-DNA fibers studded with globular domains made of granule proteins to trap pathogens and microbial foreign materials (Papayannopoulos, 2018). Recent work has demonstrated that NET formation positively correlates with injury severity in humans and degradation of NETs by administration of recombinant DNase-I ameliorates neutrophil infiltration, edema, and functional deficits and restores cerebral blood flow (Vaibhav et al., 2020). NETs activate macrophages to stimulate cGAS-STING-IFN production (Apel et al., 2021) and could be a source of DAMPs during injury. In this vein, a recent study demonstrated NET formation influenced TBI outcome, and that the protective effects of the pharmacological NET inhibitor Cl-amidine was abolished by the addition of STING agonists (Shi et al., 2023). Only a few studies have examined neutrophil trafficking in response to STING activation. One study reported that intratumoral injection of cGAMP, which activates STING, resulted in increased neutrophil trafficking possibly through CXCL1/CXCL2/CXCR2 signaling (Nagata et al., 2021). Another study also observed in mouse models of Alzheimer's disease that neutrophil infiltration was decreased in STING KO mice, which they attributed to mitochondrial DNA acting as a DAMP (Xia et al., 2024).
Our findings previously implicated mitochondrial DNA as a potential DAMP released after injury (Fritsch et al., 2022). However, we found CXCL1 and CXCL2, chemokines that are produced by and attract neutrophils (Girbl et al., 2018, Capucetti et al., 2020), are still highly upregulated 2 h postinjury in our cKO mice (Fig. 7A). More work is required to confirm this work on the protein level in a cell type-specific manner, or it could suggest that other chemokines could be responsible for STING-driven neutrophil attraction like IL-1β. Also, evidence indicates that monocytes, at least in part, drive neutrophil infiltration because monocyte depletion by liposome clodronate causes a significant reduction of infiltrating neutrophils after CCI (Makinde et al., 2017). In our chimeric experiments, KOWTBMC mice displayed significantly fewer proinflammatory Arg1− monocytes (Fig. 2H), but to better understand this relationship, another time point at 3 d postinjury in cKOs or chimeric mice would be required because very few monocytes infiltrate the brain at 24 h.
Our data also suggests that IL-1β signaling could be a target of interest, considering this pathway was transcriptionally activated by microglia after injury and absent in the cKO mice (Fig. 7D). IL-1β has previously been shown to be produced immediately after injury, predominantly secreted by microglia (Kamm et al., 2006; Hutchinson et al., 2007), and its attenuation has been shown to be protective to oligodendrocytes, dampens microglia/macrophage activation, and reduces neurodegeneration (Lu et al., 2005; Flygt et al., 2018; Ozen et al., 2020). Conflicting data suggested that IL-1β secretion may be NLRP3 inflammasome mediated and not dependent on the cGAS-STING pathway (Abdul-Sater et al., 2013), while other evidence in the context of viral immunity indicate that IL-1β signaling activates the cGAS-STING Type I IFN response (Aarreberg et al., 2019). Our data is supportive of both of these notions in the context of brain injury. While cKO mice did not upregulate IL-1β signaling at the transcription level (Fig. 7D), levels of NLRP3 mRNA were increased only in the CX3CR1CreER/+ injured brains (Fig. 7C). Considering that transcriptomics in this study were run on injured cortical tissue, dissection into which cells are producing which signals would provide further mechanistic insight to tease apart whether these changes in NLRP3 originated from microglia. A recent study that pharmacologically depleted microglia after injury did find significant downregulation of NLRP3 (Witcher et al., 2021), but our data also supports the notion that loss of microglial STING signaling could potentially be compensated for by other cell types after injury (Fig. 7G).
Our data suggests that a primary driver for changes in peripheral immune recruitment and infiltration in the brain may not be due to STING-induced changes to cytokine receptor expression on BMDMs. Our work in vitro suggests that while STING activation causes an upregulation in chemokines that bind CCR2 and CX3CR1, they themselves are unaltered (Fig. 3D,E). CCR2 is the receptor for CCL2/MCP-1, as well as one of the receptors for CCL7 (Palomino and Marti, 2015), chemokines that recruit monocytes into the brain after neurotrauma in both humans (Semple et al., 2010) and rodents (Popiolek-Barczyk et al., 2020). CCR2 KO animals showed reduced peripheral immune cell infiltration, as well as reduced Type I IFN signaling in microglia after experimental TBI (Somebang et al., 2021). In support of a detrimental role for infiltration of CCR2+ cells after injury, another group utilizing bone marrow chimeric mice showed that animals lacking hematopoietic CCR2 had reduced motor deficits and reduced infiltration of inflammatory monocytes concluding that the infiltration of CCR2+Ly6Chi inflammatory monocytes was largely responsible for the disability seen after intracerebral hemorrhage (Hammond et al., 2014). Recent work has also suggested not only infiltration levels but more so their phenotype dictate TBI outcome (Kowalski et al., 2022).
We utilized the CX3CR1CreER/+ mouse to genetically ablate STING in our cKO model which has only one copy of functioning CX3CR1 in the C57Bl/6J background; however, previous reports have demonstrated CX3CR1 signaling does contribute to the harmful neuroinflammatory effects seen in brain injury. CX3CR1 is expressed on microglia and myeloid cells, which attracts the CX3CRL1 ligand (Combadiere et al., 1998). Previous data in brain injury and stroke rodent models found that the complete loss of CX3CR1 is largely protective in the C57Bl/6J background (Tang et al., 2014; Febinger et al., 2015; Zanier et al., 2016). However, a loss of one copy of CX3CR1 in this strain did not have an effect on monocyte subsets or neutrophils after injury (Makinde et al., 2017). In the CD-1 background, loss of one copy of CX3CR1 did have a protective effect after injury (Soliman et al., 2021). Strain related differences after other injury paradigms, such as hypoxia and in the development of posttraumatic epilepsy after CCI injury, have been noted (Hunt et al., 2009; Bolkvadze and Pitkänen, 2012; Sheldon et al., 2018). Intracerebral injections of the bacterial adjuvant, LPS, showed increased neuroinflammation, heightened TNFα cytokine production, and increased CD11b+ immune cell infiltration in CD-1 as opposed to C57Bl/6 mice (Nikodemova and Watters, 2011) indicating a more robust immune response after injury between strains may account for this discrepancy. Further research in STING signaling after injury in other strains may be warranted.
Our data profiles the acute consequences of STING signaling immediately after injury, but STING Type I IFN signaling has been reported to persist in the brain days to weeks past the initial injury. We speculate that the initial events that trigger STING signaling could also be sustained by infiltrating immune cells, which we observed in this study, that also in turn stimulate and sustain IFN signaling through IFNARs with or without STING. In any case, our study demonstrates STING signaling is detrimental, triggered early through peripheral immune or microglial signaling, and shapes the neuroinflammatory landscape to influence outcome. Further elucidation of the influence that peripheral immune cells and microglia have on each other and the different mechanisms that sustain proinflammatory events will require future work in order to tease out these mechanistic outcomes.
Footnotes
This work was supported by the National Institutes of Health (R35GM142368) and Commonwealth Health Research Board (208-05-20) (A.M.P.).
*L.E.F. and C.K. contributed equally to this work.
The authors declare no competing financial interests.
- ↵Correspondence should be addressed to Alicia M. Pickrell at alicia.pickrell{at}vt.edu.
