Simulated wildfire smoke particulate matter elicits STING-dependent astrocyte-mediated inflammatory signaling and neurotoxicity

Abstract

Changing global wildfire landscapes necessitate exploration of the effects of exposure to wildfire smoke on health and disease. Exposure to this toxicant is not only associated with acute cardiopulmonary dysfunction, but is increasingly recognized as a serious risk factor for neurodegenerative diseases such as Alzheimer’s disease (AD) and Parkinson’s disease (PD). However, the cellular and molecular mechanisms which underlie this association are not well understood. One potential mechanism linking neurotoxic environmental exposures with neurodegeneration is glial-mediated neuroinflammation, which may be influenced by innate immune signaling through the cGAS-STING pathway in response to damaged or mislocalized DNA. To address this hypothesis, we here exposed primary astrocyte-enriched mixed glial cultures to laboratory generated wildfire smoke particulate matter extract and subsequently examined their reactivity and inflammatory signaling by applying cutting-edge techniques in high content microscopy, deep learning-based image analysis, and transcriptomics. We found that wildfire smoke exposure elicits DNA damage and results in STING signal transduction in astrocytes, including the production and release of inflammatory cytokines, as well as STING-dependent neurotoxicity. To better understand the mechanisms underlying this phenomenon, we integrated transcriptomic data from in vitro and in vivo wildfire smoke exposures studies, which revealed central hubs for functional enrichment surrounding interferon signaling. Together, these data strongly identify STING signaling as a central modulator of astrocyte-mediated inflammation resulting from wildfire smoke exposure.

1. Introduction:

Wildfire events are increasing in frequency, intensity, and duration globally concurrent with warmer and drier climates (1). This expanded wildfire landscape poses significant health risks as a result of greater contribution of smoke particulate matter toward total ambient air pollution (2). Wildfire smoke is a complex, heterogenous mixture that varies significantly in chemical composition based on parameters surrounding combustion including fuel source, moisture content, season, and geographic distance traversed between initial burn and downwind exposure (3). Although these variables make determining the toxic potential of smoke across fire events challenging, it remains well established that this toxicant contains substantial amounts of fine particulate matter (PM_2.5) and ultrafine particulate matter (PM_0.1), which constitute size fractions of air pollution that are capable of bypassing the lung-blood barrier and entering systemic circulation or directly entering the brain parenchyma via olfactory nerve bundles (4, 5). Accordingly, spillover of wildfire smoke into the periphery has been increasingly associated with longitudinal health risks outside of the acute cardiopulmonary deficits classically linked to its inhalation (6). This includes a growing body of epidemiologic and experimental literature pointing to neurologic consequences associated with air pollution exposure (7, 8). Wildfire smoke has specifically been shown to play a role in this context given strong associations with cognitive decline, accelerated epigenetic aging, and risk of neurodegenerative disease (9–12). Still, the cellular processes underlying this systemic dysfunction remain poorly understood.

Neurodegenerative diseases such as AD, PD, and related disorders share a number of common histopathologic features including aggregation of misfolded proteins and neuroinflammation, characterized by activation of glial cells and sustained production of inflammatory mediators that can exacerbate neuronal vulnerability to toxic injury and contribute to aberrant neuro-immune interactions (13). Astrocytes, the most abundant glial cell in the central nervous system (CNS), are critically involved in maintaining neuronal homeostasis, regulating synaptic transmission, and preserving blood–brain barrier (BBB) integrity (14). However, during age-related pathology or neurodegenerative disease, astrocytes are known to undergo phenotypic changes to various reactive states that can be neuroprotective or neurotoxic (15–17). Environmental stressors, including airborne particulate matter, have been shown to bias astrocytes toward pro-inflammatory reactive states, suggesting that astrocyte dysfunction may be a key interface between these exposures and neurodegenerative processes (18–20). Importantly, data have emerged demonstrating the ability of wildfire smoke exposure to alter brain endothelial cell tight junction integrity, induce neuroimmune modulation, and alter transcriptomic patterns relevant to neurodegeneration (21–24). Despite this, the molecular mechanisms by which wildfire smoke exposure may contribute to astrocyte-mediated neuroinflammation are not well understood.

The stimulator of interferon genes (STING) pathway has come under scrutiny as a critical regulator of innate immune signaling in the CNS (25). Activated by cytosolic DNA, STING signaling promotes type I interferon production that can contribute to pro-inflammatory transcriptional programs (26). While initially identified in the context of peripheral antiviral immunity, accumulating evidence implicates STING activation in glial cells as a contributor to neuroinflammation, age-related inflammatory priming, and neurodegenerative pathology (27, 28). Mitochondrial dysfunction, oxidative stress, and DNA damage, hallmarks of both aging and wildfire smoke exposure, have been shown to contribute to STING activation and cytotoxicity in immortalized cell lines (29, 30). These observations suggest that this pathway could be an important mechanistic link between exposure to environmental neurotoxins and sustained glial reactivity.

Therefore, we examined the effects of wildfire smoke extract on primary astrocyte-enriched murine mixed glial cells in vitro to identify inflammatory mechanisms linking exposure to this toxicant and astrocyte reactivity. By applying high content microscopy and deep learning-based image analysis, we aimed to define how environmental exposures can contribute to innate immune signaling in this cell type, promoting neuroinflammatory states relevant to age-related neurodegeneration. Combined with transcriptomic analysis, these findings identify the cGAS-STING signaling pathway in astrocytes as an important target of wildfire smoke exposure in astrocytes relevant to inflammatory injury to neurons.

2. Methods

2.1. Simulated wildfire smoke generation and extraction:

Simulated laboratory wildfire events were carried out using a tube and ring furnace instrument mounted on a linear actuator as previously described (31). All burns were conducted at 450°C to simulate smoldering combustion parameters with dried Douglas fir needles as the fuel source. This is consistent with other groups modeling wildfire smoke in a laboratory setting (32). Smoke particulate was collected using quartz filters placed at the furnace exhaust duct throughout the duration of each burn. Filters were stored at −80°C prior to use in downstream assays. Particles were methanol extracted in conical tubes via ice bath sonication for 15 minutes. Following sonication, solvent exchange proceeded under N₂ gas. Smoke extract was resuspended in sterile dimethyl sulfoxide (DMSO) for downstream biophysical characterization and in vitro experimentation (FIG. 1A).

Figure 1. Biophysical characterization of wildfire smoke particulate extract size.

A) Scheme illustrating the particle collection and extraction protocol followed to suspend particulate in DMSO. B-C) Representative transmission electron micrographs depicting low magnification and high magnification views of smoke particulate used to calculate sizes of FPM and UFPM. D) Histograms demonstrating the relative size of UFPM and FPM fractions of smoke extract particulate matter measured via transmission electron microscopy. E) Frequency plot highlighting the two peaks of UFPM and FPM sized particles quantified via dynamic light scattering.

2.2. Smoke extract particulate matter biophysical characterization:

To quantify the size of smoke extract particulate matter, we applied Transmission Electron Microscopy (TEM, JEOL JEM-2100F) and Dynamic Light Scattering (DLS, Malvern Zetasizer Nano ZS), as previously described (33). Briefly, to quantify the size of particles via TEM, the Record Measurements feature of Adobe Photoshop (v. 27.2.0) was utilized to measure particle size across greater than 2000 particles from 10 images/size fraction at appropriate magnifications (FPM and UFPM). For DLS analysis, smoke extract particulate was diluted to 0.1 mg/mL in DMSO and 494.1 kcps were captured over 60 s at 25°C. DLS results represent the average of 3 technical replicates. Samples were ice bath sonicated immediately prior to particle size measurement.

2.3. Primary cell culture isolation and maintenance:

Primary astrocyte-enriched mixed glial cultures and primary neuronal cultures were established as previously reported (34). Briefly, P0-P1 murine neonates from C57Bl/6 (wildtype/WT) or Tmem173^goldenticket (STING^goldenticket/STING^gt) strains were sterilized by 70% ethanol and decapitated under isoflurane anesthesia. The cerebral cortex and midbrain were then placed in ice cold MEM/EBSS containing 2X PSN and dura and pia mater were carefully removed under a dissecting microscope. Cells were isolated using dispase digestion (37°C) of cleaned parenchymal tissues in glass beakers via gentle stir bar agitation. Collected cells were pelleted via centrifugation (4°C) and quantified using a Countess automated cell counter. Viable glial cells were seeded at a density of 750,000 cells/T75 flask in MEM/EBSS media supplemented with 10% FBS and 1X PSN and grown for 2 weeks, or until reaching 75% confluency, prior to passaging. All mixed glial cultures were established to be astrocyte-enriched by following media change protocols as previously described (35). Viable neurons were seeded at a density of 8,000 cells/well in black polystyrene flat-bottom 96-well plates (Costar) in Neurobasal A medium supplemented with 0.5 mM L-glutamine, 2% B-27 supplement, AraC, 3% FBS, and 1X PSN and grown for 9 days in vitro (DIV) with media changes occurring 24 hrs after isolation and every subsequent 48 hrs. All cells were maintained in humidified incubators at 37°C and 5% CO₂.

2.4. Cell viability, fluorescence reporter, and DNA damage assays:

Primary astrocyte-enriched glia between passage 2-3 were plated in 96-well plates at a density of 8,000 cells/well in DMEM/EBSS media supplemented with 10% FBS and 1X PSN (AstroDMEM) for viability assay conducted using a Tecan SPARK plate reader. Cells were allowed to grow for 48 hrs before dosing with either DMSO (1:1000 in AstroDMEM) as a vehicle control or ½ log doses of wildfire smoke extract in DMSO (1:1000 in AstroDMEM) ranging from 1 to 1000 ng/mL for pre-determined time points before being assayed for viability. Additionally, experiments were conducted in WT cells co-administered smoke extract and selective pharmacologic STING inhibitor H-151, as well as STING^gt cells. After 12 hrs, dosing media was vacuum aspirated and cells were washed with PBS (pH 7.0-7.2). Cells were then incubated with propidium iodide (PI), calcein AM (Calc), and Hoechst 33342 in FluoroBrite DMEM for 10 minutes in a humidified incubator at 37°C and 5% CO₂ as reported previously (36). Normalized measurements were determined as percent viability of negative control (DMSO – VEH) and positive control (1X PERM BUFF) values. Internal assay controls of no cells with dye, cells with no dye, and 1X PERM BUFF with no dye were also utilized in each experiment for assay validation. The same viability assay was applied to assess primary neurotoxicity following 24 hr treatment with glial-conditioned media collected after 24 hr treatment with smoke extract particulate. All viability data presented in this manuscript represent one of three individual replicates. Separate 96-well plates were prepared as described above using astrocyte-enriched mixed glial cultures established from either NF-κB^EGFP or IFN-β^EYFP mice, as shown previously (37). To assess NF-κB activity or IFN-β production, cells were treated with wildfire smoke extract and assayed every 3 hrs over a 24 hr period in FluoroBrite DMEM enriched with 10% FBS and 1X PSN. Excitation and emission parameters used to assess GFP and YFP were: GFP ex. 485/em. 535 and YFP ex. 500/em. 530. Normalized measurements were established via a terminal Hoechst stain as described above and comparison to media only, untreated, vehicle treated, or 100 ng/mL poly dA:dT DNA treated cells. DNA damage was assessed via the EpiQuik In Situ DNA Damage kit (p-γH2A.X Ser139, Epigentek, Farmingdale, NY, USA) following the manufacturer’s protocol.

2.5. Multiplex immunocytochemistry:

Astrocyte-enriched primary glia were plated in AstroDMEM onto Poly-L-Lysine coated 12 mm diameter German #1.5 glass covertiles at a density of 50,000 cells/well in 24 well-plates. Cells were allowed to grow for 48 hrs before dosing with either DMSO (1:1000 in AstroDMEM) as a vehicle control or a sublethal dose of 100 ng/mL wildfire smoke extract in DMSO (1:1000 in AstroDMEM) for pre-determined durations. After the desired number of hrs, dosing media was vacuum aspirated, cells were washed 1X with TBS (pH 7.6) and all wells were immediately fixed with iced cold methanol via incubation at −20°C for 15 minutes. Following this fixation, methanol was vacuum aspirated and covertiles were washed 3X with 0.05 M TBS (TBS, pH 7.6) at room temperature (RT). Immunocytochemistry proceeded via the following protocol: permeabilization with 0.2% Tris in TBS for 30 min at RT, washing 3X in TBS for 10 min at RT, blocking with 1% Donkey serum in TBS for 30 min at RT, washing 3X in TBS for 10 min at RT, primary antibody cocktail incubation in TBS for 1 hr at 37°C, washing 3X in TBS for 10 min at RT, secondary antibody cocktail incubation in TBS for 30 min at RT, washing 3X in TBS for 10 min at RT, Hoechst 33342 incubation in TBS for 3 min at RT, and washing 3X in TBS for 10 min at RT. Following the final wash step, coverslips were mounted onto charged glass slides using ProLong Diamond Antifade hardset mounting media. The concentrations and lot numbers of all antibodies used in this study can be found in Tables 1 and 2. After curing, all slides were stored in the dark at 4°C until imaging.

Table 1.

Primary antibodies

Brand	Product #	Host	Antigen	Lot #	Dilution
Invitrogen	PA5-105674	Rabbit	pSTING (S366)	ZH4407747	1:250
Invitrogen	PA5-36775	Mouse	pIRF3 (S385)	ZH4407671	1:500
Abcam	AB235197	Goat	ALDH1L1	1006150-35	1:500
Abcam	AB4674	Chicken	GFAP	1088394-2	1:1000

Table 2.

Secondary antibodies

Brand	Product #	Host	Antigen	Fluorophore	Lot #	Dilution
Invitrogen	A78948	Donkey	Chicken	AF488	2850198	1:500
Invitrogen	A31572	Donkey	Goat	AF555	2831376	1:500
Invitrogen	A31573	Donkey	Rabbit	AF647	2997084	1:500

2.6. High content slide scanning fluorescence microscopy:

Stained slides were allowed to warm to RT in the dark before being loaded into an Olympus SLIDEVIEW^™ VS200 microscope (Evident, Waltham, MA, USA) equipped with a Hamamatsu ORCA-Fusion camera (C14440-20UP, Hamamatsu Photonics, Shizuoka, Japan). Images were collected using Olympus ASW software. Whole covertile scans for downstream signal quantification were obtained via an Olympus UPLXAPO20X (NA 0.8) objective at the same exposure parameters, lamp intensity, camera gain, and filter settings. High magnification representative images of single cells used in this publication were obtained using an Olympus UPLXAPO60X (NA 1.42) oil immersion objective (working distance 0.15 mm). Three-dimensional renders of z-stack representative images were constructed using IMARIS for Neuroscientists software (version 10.2.0, Bitplane, South Windsor, CT, USA).

2.7. Visiopharm deep learning-based automated signal quantification and analysis:

Montage scans of cell populations encompassing entire cover tiles were uploaded to Visiopharm Oncotopix (v. 2025.08.2) to generate custom deep learning-based neural networks. Nuclei segmentation networks were generated by training off of images across each time point and treatment condition to detect Hoechst-positive area. Nuclear borders were dilated by 5 μm to represent a cytosolic region of interest (ROI) for biomarker quantification. The intensity of ALDH1L1 and GFAP within each perinuclear ROI, as well as the number of p-STING+ foci (Ser365) or nuclear p-IRF3+ area (Ser385), were collected to measure astrocytic STING activation. Networks were trained for approximately 100,000 iterations and monitored every 25,000 iterations for efficacy. For all immunocytochemical image analysis, a background channel and secondary antibody only controls were used to validate signal positivity. All networks were trained and applied by experimenters blinded to experimental conditions.

2.8. ELISA cytokine/chemokine analysis:

Astrocyte-enriched primary glia were plated into 6-well plates at a density of 500,000 cells/well in AstroDMEM. Cells were allowed to grow for 48 hrs before dosing with either DMSO (1:1000 in AstroDMEM) as a vehicle control or 100 ng/mL of wildfire smoke extract in DMSO (1:1000 in AstroDMEM) for 24 hrs. The conditioned media of the cells was then collected, centrifuged, and stored at −80°C until downstream assessment. Murine pre-frontal cortical samples (~15 mg tissue) derived from a previously published study cohort were protein extracted by chilled homogenization via Qiagen TissueLyser LT in RIPA buffer supplemented with PMSF protease inhibitor (24). Any experiments involving animal models were conducted with Institutional Animal Care and Use Committee (IACUC) approval. Briefly, in that study mice were whole body exposed to smoldering douglas fir needle combustion smoke for 2 hr/day, 5 day/week, for 16 weeks to approximate a cumulative mid-career exposure dose of a wildland firefighter (24). Subsequently, all extracted protein samples were pelleted via centrifugation prior to storage at −80°C until analysis. The protein concentration of each sample was determined using a Rapid Gold BCA Protein Assay (Pierce Biotechnology, Waltham, MA; catalog # A55860). For in vitro cytokine/chemokine analysis, 96-well plates were prepared using Milliplex Mouse Cytokine/Chemokine Magnetic Bead Panel - Immunology Multiplex Assay Kit (Millipore, Burlington, MA; cat #MCYTOMAG-70K). All plates were run on a Luminex200 platform, and the data was analyzed using Belysa^® software as previously described (38). To compare in vivo and in vitro protein levels for STING-relevant cytokines or transcription factors, Abcam ELISA kits were used to assess NF-κB (Cat. No. ab176648), TNF-α (Cat. No. ab208348), and IFN-β (Cat. No. ab252363). All samples were run in triplicate.

2.9. RNA sequencing and bioinformatics:

Astrocyte-enriched primary glia were plated in AstroDMEM onto T25 flasks at a density of 250,000 cells/flask. Cells were allowed to grow until they reached 75% confluency, before dosing with either vehicle control or a sublethal dose of 100 ng/mL wildfire smoke extract for 12 hr. Approximately 7.5x10⁵ cells per sample (n = 3 samples/group; 6-8 mice/sample) were RNA extracted via Qiagen RNEasy prep kit. Samples were checked for quality and purity via nanodrop spectroscopy, agarose gel electrophoresis, and Agilent 2100. RNA sequencing was performed on a NovaSeqX 10B flow cell and demultiplexed with DRAGEN BLC-Convert (v. 4.1.23). The quality of the raw reads was assessed using FastQC version 0.12.1. Adapter sequences were removed using Cutadapt (v. 4.4). Subsequently, clean reads (~62,750,000 reads/sample) reads were aligned to the Mouse reference (GRCm39) using HISAT2 (v. 2.2.1). Differential expression analysis of wildfire smoke-exposed versus vehicle treated samples was performed via DeSeq2 (v1.50.2) through one scaling normalized factor. Thresholds of overrepresentation analysis via Benjamini Hochberg corrected false discovery rate (FDR) < 0.05 and log2foldchange > 0.5 were used to determine differentially expressed genes (DEGs). DEGs were analyzed for Kyoto Encyclopedia of Genes and Genomes (KEGG) and Reactome pathways enrichment using clusterProfiler (v4.18.2) with significance determined by FDR < 0.05 as compared to vehicle controls. Integration of newly generated RNAsequencing data with previously published work from our group in murine pre-frontal cortex was conducted by comparing DEGs that were enriched vs. depleted between groups (24). Subsequently, Gene Ontology (GO) Biologic Process (BP) pathways were queried for concordantly dysregulated genes via REVIGO and protein protein interaction (PPI) network maps were constructed via the STRING database plugin for Cytoscape (39). Visualization of each network map was conducted using the Freuchterman–Reingold force-directed layout in Gephi. Top hits with validated primers/probes were used to perform qPCR validation as previously described (40).

2.10. Statistics:

All data presented in this manuscript as mean ± SEM, unless otherwise noted in respective figure caption. Experimental values from each sex and experimental group were subjected to outlier analysis and exclusion via ROUT (α = 0.05). Statistical tests were determined by the number of variables being compared between groups via student’s t-test, Welch’s t-test, or two-way ANOVA with Bonferroni post-hoc correction. All statistical analysis was performed using GraphPad Prism (version 10.1.1; Graph Pad Software, San Diego, CA, USA) unless otherwise stated in the above methods.

3. Results

3.1. Simulated wildfire smoke particulate matter is within a neurotoxic size fraction:

Characterization of wildfire smoke particulate matter was performed using two biophysical techniques, transmission electron microscopy (TEM) and dynamic light scattering (DLS). TEM analysis indicated that the average diameter of particles in the FPM fraction was 541.4 nm ± 280.4 (FIG. 1B–D). In the UFPM size fraction, mean particle diameter was 60.89 nm ± 21.02 (FIG. 1B–D). These particle sizes were separately confirmed by DLS, which demonstrated two peaks in particle size, one at 466.8 nm ± 123.1 and the other at 81.70 nm ± 13.02 (FIG. 1E). Consistent with previous reports (33), wildfire smoke particulate extract was found to be heterogenous with regard to size distribution with a polydispersity index of 0.510.

3.2. Smoke particulate extract results in DNA damage and STING-dependent glial cytotoxicity:

Next, we sought to determine the toxicity of applying wildfire smoke particulate matter extract on astrocyte-enriched primary glia via liquid suspension dosing. We utilized a permeability-based cell viability assay to establish the LD₅₀ = 734.3 ng/mL at 12 hrs (FIG. 2A). Subsequent pharmacologic inhibition (H-151) or genetic ablation of STING (STING^goldenticket/STING^gt knockout mice) increased the LD₅₀ to 1035 ng/mL or greater (FIG. 2A). The observed cytotoxicity was accompanied by an increase (mean difference = 1.213; p = 0.0497). in p-γH2A.X (Ser 139) following 3 hrs of wildfire smoke extract treatment at a sublethal dose of 100 ng/mL (FIG. 2B). We then queried the effects of this genotoxicity on downstream STING activation by applying neural network-based image analysis to quantify perinuclear pSTING+ foci (Ser 366) in populations of astrocytes expressing GFAP or ALDH1L1 across greater than 15,000 individual cells (FIG. 2C–D). This revealed a time-dependent increase in pSTING+ foci in distinct astrocyte subpopulations, with a substantial induction of pSTING in both GFAP^hiALDH1L1^hi cells and GFAP^hiALDH1L1^lo cells at 6 hrs after treatment with smoke extract (mean differences = 1.496 and 1.592; p = 0.0477 and 0.0392, respectively) (FIG. 2F). No changes were observed in the number of perinuclear pSTING+ foci in other astrocyte subpopulations at the 6 hr time point or in any astrocyte subpopulation at the 3 hr time point (FIG. 2E).

Figure 2. Smoke-induced cytotoxicity and STING activation in astrocytes.

A) Survival curves depicting % glial cell viability measured at 12 hr exposure to increasing doses of smoke extract particulate compared to DMSO vehicle across wild-type, wild-type + STING inhibitor H-151, and STING^goldenticket conditions. B) Bar graph depicting phosphorylated yH2A.X, a marker of double stranded DNA damage and repair, via in situ fluorescence assay; * depicts p-value < 0.05 as measured by Welch’s t-test. C-D) Representative immunofluorescent images displaying cells from the 6 hr time point of smoke extract exposure and associated 3-D renders created in IMARIS; arrowheads display pSTING+ foci within the quantifiable region of the perinuclear mask. E-F) Bar graphs demonstrating the # of perinuclear pSTING+ foci per cell for each condition and cell population bin; the number of cells assessed within each group is labeled within the corresponding bar; * depicts p-value < 0.05 as measured by one-way ANOVA.

3.3. Wildfire smoke extract elicits STING-dependent inflammatory signal transduction cascades:

Following the observation that wildfire smoke extract activates STING signaling in astrocytes, we sought to determine if this signal transduction persisted downstream at the transcript or protein level. Accordingly, we surveilled the subcellular localization of phospho-IRF3 (pIRF3), a downstream transcriptional effector of the cGAS-STING pathway, which is classically shown to induce expression of type-I-interferons (41). This analysis proceeded as described above and revealed a time- and STING-dependent increase in nuclear pIRF3+ area in GFAP^hiALDH1L1^hi (mean difference = 1.483; p-value = 0.0265) and GFAP^hiALDH1L1^lo (mean difference = 1.557; p = 0.0289) cells following 9 hrs of exposure to smoke extract (FIG. 3A–F). Because pSTING can activate both IRF-3-dependent as well as NF-κB-dependent inflammatory pathways, we examined the level of NF-κB activity in primary glia derived from NF-κB^EGFP reporter mice, which revealed substantial peak NF-κB activity at the 9 hr time point (mean difference = 2.125; p = 0.0113), which was prevented by co-treatment with the pharmacologic STING inhibitor, H-151 (FIG. 3G). To establish the impact of these signaling pathways on immune modulatory outcomes, we also employed primary glia derived from IFN-β^EYFP reporter mice. This revealed a significant increase in IFN-β production over 24 hrs (FIG. 3H) with the greatest difference against STING-inhibited cultures at the final time point (mean difference = 17.28; p = 0.0282).

Figure 3. STING-dependent increase in inflammatory signal transduction cascades.

A-D) Representative immunofluorescent images displaying cells from the 9 hr time point of smoke extract exposure and associated 3-D renders created in IMARIS; arrowheads display pIRF3+ foci within the nuclear mask. E-F) Bar graphs demonstrating the normalized pIRF3+ nuclear area per cell for each condition and cell population bin; the number of cells assessed within each group is labeled above or within the corresponding bar; * depicts p-value < 0.05 as measured by two-way ANOVA. G) Quantification of NF-kB activity normalized to LPS stimulation via fluorescent reporter astrocytes; * depicts p-value < 0.05 as measured by a mixed model approach compared to H-151 co-treatment. H) Determination of IFN-b production normalized to dsDNA treatment via fluorescent reporter astrocytes; * depicts p-value < 0.05 as measured by a mixed model approach compared to H-151 co-treatment.

3.4. Astrocyte-conditioned media contains brain-relevant neurotoxic pro-inflammatory cytokines following wildfire smoke exposure:

Cytokines and chemokines secreted in conditioned media from astrocytes exposed to wildfire smoke extracts were examined using Luminex ELISA and these data were correlated with neurotoxic endpoints. Of 16 total analytes examined by this method, only 8 had values above the limit of detection (FIG. 4B). Of those, the three cytokines showing the greatest fold-change between wildtype and STING^gt cells were IL-6, TNF-α, and IFN-β (mean difference = 2.430, 2.908, and 1.785, respectively). Interestingly, mice exposed to the same type of simulated wildfire smoke particulate matter demonstrated similar changes in inflammatory cytokines and signaling factors in the pre-frontal cortex, with increased expression of NF-κB (p65) (mean difference = 5.506; p = 0.0121), TNF-α (mean difference = 50.02; p = 0.0450), and IFN-β (mean difference = 26.10; p = 0.0069), when comparing wildfire smoke exposed animals against filtered air controls (FIG. 4D). To determine the effects of these cytokines on neuronal viability, we conducted neurotoxicity screening (FIG. 4C) using glial conditioned media following 24 hrs of smoke treatment or control, which revealed a STING-dependent effect on neuronal viability between wildtype and STING^gt glia (mean difference = 37.04; p = 0.0067).

Figure 4. Smoke-induced changes in transcriptional profile map to neurotoxic cytokines.

A) Scheme depicting harvest of glial-conditioned media for cytokine and chemokine profiling as well as treatment of primary neuronal cultures. B) Heatmap demonstrating most enriched protein constituents within wild-type or STING^goldenticket glial conditioned media following 24 hr smoke extract treatment. C) Bar graph depicting modulation of primary neuronal viability following 24 hr treatment with glial-conditioned media derived from wild-type or STING^goldenticket astrocytes treated for 24 hr with smoke extract; **, **** denote p-value < 0.01, 0.0001, respectively, as measured using two-way ANOVA; main effects or interaction effect highlighted in upper righthand corner of the graph. D) Illustrative scheme depicting murine smoke exposure and associated harvest of pre-frontal cortex brain tissue for protein characterization; bar graphs depicting NF-kB, TNF-a, or IFN-b levels in brain parenchyma measured using ELISA; *, ** denotes p-value < 0.05, 0.01, respectively, as measured using Welch’s t-test.

E) Scheme illustrating RNA isolation from primary wild-type astrocytes following 12 hr treatment with smoke extract; venn diagram depicts unique increased or decreased DEGs measured via RNAsequencing; PCA plot demonstrating clear clustering by experimental treatment; volcano plot annotated to show top increased or decreased DEGs following treatment. F) Bubble plot highlighting the top 10 enriched KEGG pathways across all DEGs; size and color maps correlate to gene ratio and raw gene count, respectively. G) Validation of marker genes identified via RNAseq using validated probe-based qPCR.

3.5. Wildfire smoke induces similar immunomodulatory transcriptomic disruption between astrocyte-enriched cultures and murine brain:

To identify signaling mechanisms underlying astrocyte-mediated STING-dependent neurotoxicity in culture, we conducted RNAsequencing on cell populations exposed to wildfire smoke particulate extract or vehicle control for 12 hrs. This revealed 603 unique increased transcripts and 670 unique decreased transcripts (FIG. 4E). Unbiased clustering clearly demarcated samples according to exposure status using PCA analysis (FIG. 4E) and dendrogram heatmap construction (SUP. FIG. 1). Top hits that were increased or decreased in expression are annotated on a volcano plot (FIG. 4E) including: semaphorin 3A (Sema3a, ↑), pleckstrin homology like domain family B member 2 (Phldb2, ↑), interferon-induced protein 44 (Ifi44, ↑), dedicator of cytokinesis 10 (Dock10, ↓), C-C motif chemokine ligand 2 (Ccl2, ↓), and hematopoietic prostaglandin D synthase (Hpgds, ↓). Pathways overrepresentation analysis also revealed the top 10 KEGG pathways across all DEGs to include: chemokine signaling pathway, B cell receptor signaling pathway, complement and coagulation cascades, neutrophil extracellular trap formation, PI3K-Akt signaling pathway, cytokine-cytokine receptor interaction, cytoskeleton in muscle cells, NF-kappaB signaling pathway, toll-like receptor signaling pathway, and ECM-receptor interaction (FIG. 4F). To further interrogate the relevance of these DEGs to organism-level dysfunction, we integrated transcriptomic profiles from smoke extract-treated cell culture and murine pre-frontal cortex following whole-body inhalation exposure to the same type of simulated smoke (FIG. 5A) and analyzed functional enrichment patterns for concordantly disrupted DEGs (31 down, 48 up). These genes mapped to GO BP pathways including downregulation of immune system process, lymphocyte activation, leukocyte activation, and B cell activation as well as upregulation of cell adhesion and connective tissue development (FIG. 5B,D). When assessing the protein-protein interactions that exist between these concordantly disrupted genes, we revealed a central node in the down network (FIG. 5C) surrounding integrin subunit alpha M (Itgam) and a central hub in the up network (FIG. 5E) surrounding interferon-induced protein 44 (Ifi44), interferon-induced transmembrane protein 1 (Ifitm1), and interferon-stimulated gene 15 (Isg15).

Figure 5. Integrated analysis of astrocyte transcriptional changes compared to murine brain.

A) Scheme illustrating RNA isolation and sequencing from primary wild-type astrocytes following 12 hr treatment with smoke extract or murine brain pre-frontal cortex following occupationally relevant whole-body inhalation exposure. B, D) Venn diagrams depict conserved decreased (B) or increased (D) DEGs after integrating across models; bubble plots highlight the top 5 enriched Gene Ontology Biologic Process pathways across conserved DEGs; size and color maps correlate to gene ratio and raw gene count, respectively. C, E) Protein-protein interaction network maps generated via STRING algorithm and visualized using force-directed Fruchterman-Reingold layout; node size correlates with degree of interconnectedness; edge color intensity warmth correlates with stronger association score.

4. Discussion

Changing climates world-wide have increased the severity and extent of fire seasons across greater portions of the year (42), resulting in much larger geographic areas exposed to smoke (43). Accordingly, wildfire smoke poses increasing health risks when compared to historical contributors of ambient air pollution, especially given the emergence of data highlighting greater pulmonary toxicity and inflammatory sequelae for this toxicant compared to other speciated FPM (44). Increasing reports have demonstrated the ability of wildfire smoke particulate to escape the lung and traverse systemic circulation, including work which has shown the propensity of this toxicant to induce expression of matrix metalloproteases capable of increasing the permeability of the blood-brain-barrier (BBB) and contributing to altered expression of tight junction proteins in the brain vasculature (21, 45). Astrocyte endfeet surround the microvasculature of the CNS and are therefore amongst the first parenchymal cells of the brain to encounter blood-borne particulates and associated inflammatory mediators. Here, we established the size fraction of laboratory generated wildfire smoke extract PM to be consistent with previous reports (FIG. 1D–E), showing UFPM of a size capable of entering the brain parenchyma, including via direct nose-to-brain transport following inhalation (46, 47). Accordingly, we sought to understand the effects of acute, low-dose simulated wildfire smoke extract particulate matter exposure on astrocytic inflammatory activation in vitro.

Particulate matter in air pollution is increasingly associated with age-related neurodegeneration, including AD, PD, and related dementias (48–50). These disorders share common hallmark features of neuronal loss, pathologic protein aggregation, and glial-mediated neuroinflammation (51, 52). Data from our group and others has highlighted how inflammatory signaling between reactive astrocytes and microglia amplifies neuronal injury in the context of neurotoxic exposures (19, 34). Further, we have shown that loss of function of inflammatory signaling factors in astrocytes can protect against neurodegeneration under similar neurotoxic environmental insults (17, 20, 53). In the present study, we initially determined the cytotoxicity of liquid suspension exposure to smoke extract at low concentrations in astrocyte-enriched mixed glial culture media (FIG. 2A). Both pharmacologic and genetic inhibition of STING markedly decreased the cytotoxicity of wildfire smoke particulates in primary astrocytes, indicating that STING directly modulates cellular injury associated with activation of innate immune inflammatory signaling (FIG. 2A). This phenomenon of STING-mediated cell death has been observed in other disease states, including cancerous cell proliferation (54) and is being examined as a potential therapeutic in that context (55, 56). Concurrent with cytotoxicity elicited by smoke extract, we also report induction of double-stranded DNA damage and repair activation following treatment, evidenced by increased levels of p-γH2A.X (FIG. 2B). This is consistent with previous reports highlighting the propensity for damaged DNA to activate STING signaling to induce cell death, despite the exact mechanism of cytosolic DNA release remaining ill-explored (57). Separately, we observed a marked increase in levels of STING activation as measured by high-content microscopy and AI-based image analysis (FIG. 2C–F). The cGAS-STING pathway has been increasingly associated with aging and neurodegeneration (27, 58). Further, this has been shown to be specific to astrocytes in the MPTP model of Parkinson’s disease (59). Together these data warranted subsequent investigation of these findings in the context of astrocytic inflammatory signaling.

To interrogate the effects of wildfire smoke exposure on astrocytic reactivity, we queried signal transduction downstream of STING activation. Initial characterization of this included investigation of compartmental localization of pIRF3 via high-content microscopy and AI-based image analysis (FIG. 3A–D), which revealed a substantial increase in nuclear area covered by this transcription factor at 9 hrs post-exposure (FIG. 3F). Nuclear IRF3 translocation is critical to astrocyte inflammatory reactivity (60). We then wanted to determine if this transcription factor activity correlated with production of inflammatory cytokines and chemokines. In primary astrocytes from NF-κB^EGFP and IFN-β^EYFP transgenic reporter mice, there was a STING-dependent increase in NF-κB activity as well as IFN-β production following treatment with smoke extract (FIG. 3G–H). Activation of NF-κB signaling has previously been associated with exposure to wildfire smoke (61, 62). Further, astrocytes have been shown to be the predominant producers of type-I-interferons in the brain under a range of conditions (63–66). Together, these immune modulatory cascades play integral roles in neuroinflammatory signaling under aging and neurodegeneration (67–69). This led us to examine the production of additional STING-associated pro-inflammatory cytokines following smoke exposure, which revealed substantial induction of IL-6 and TNF-α (FIG. 4B). These cytokines can adversely modulate neural function (70) and contribute to neurotoxicity (71). Interestingly, these cytokines are also shown to be elevated in human blood, both following acute exposure to woodsmoke in a laboratory setting and in an occupational firefighting context (72, 73). Accordingly, we identified a STING-dependent neurotoxic effect following treatment of primary neurons with GCM derived from smoke-treated glia (FIG. 4C). This is consistent with studies demonstrating that astrocytes directly contribute to neuronal loss in aging and neurodegenerative disease states (74–76). Further, these same pro-inflammatory signals were found to be enriched in the brains of smoke-exposed mice compared to filtered air controls (FIG. 4D) despite vast differences in the dose and duration of exposure to this toxicant. It is worth noting that the temporal dynamics of glial responses warrant future exploration of these phenomena across a more longitudinal time course of smoke exposure. Additionally, despite the association of these inflammatory cytokines with neuronal apoptosis, further investigation of the impact of exposure to wildfire smoke on neurodegenerative phenotypes and histopathology is necessary to determine disease relevance.

We next performed RNAseq and functional enrichment analysis to interrogate the mechanisms by which wildfire smoke contributes to astrocyte reactivity. When assessing the impacts of wildfire smoke exposure on astrocyte-enriched glial cells, the top DEGs included Sema3a, which has been shown to possess a central role in astrocytic regulation of synaptic plasticity (77), as well as Dock10 and Ccl2, which have been linked to astrocytosis and glial-immune crosstalk in neuroinflammatory disease states (78, 79). Another major DEG was Ifi44, which has been previously implicated in glial response to DNA damage and under neurodegenerative states (80, 81). KEGG pathways analysis confirmed the enrichment of such responses, with top hits including cytokine and chemokine signaling (FIG 4F). We then investigated whether these changes would be upheld across an integrated analysis of our present in vitro study and a previously published in vivo transcriptomic dataset from our group (24). Still, protein–protein interaction network mapping revealed overlapping genes predominantly annotated to type I interferon response elements, forming a highly interconnected hub within the network of conserved upregulated genes (FIG. 5E). Given that interferon responses are strongly associated with cytosolic nucleic acid sensing, these findings further corroborate that wildfire smoke activates innate immune pathways in glia, including cGAS-STING. Persistent interferon production has been linked to neurotoxic glial reactivity, positioning this pathway as a mechanistic link between environmental exposure and chronic neuroinflammation (82, 83). It is worth pointing out that smoke dosimetry varies substantially between in vitro and in vivo studies because of differences in experimental models, exposure routes, and exposure duration (84). These comparisons are further confounded by the highly heterogenous and volatile nature of this toxicant which significantly inhibits the potential to track delivery across different tissue types (85). Still, the convergence of interferon-related genes across these model systems highlights a potentially conserved transcriptional response to wildfire smoke extract in the brain.

5. Conclusion

Consistently worsening wildfire seasons underscore the need to better understand smoke-associated health effects, particularly those linking peripheral immune activation with long-term neuroinflammatory effects in the brain. We sought to address this knowledge gap by identifying potential mechanisms linking particulate matter exposure with aberrant glial reactivity and inflammatory signaling. The data presented here strongly support the involvement of astrocyte-mediated STING signaling and interferon responses in neurotoxicity following smoke exposure, including across in vitro and in vivo model systems. Future work should examine the effects of astrocyte-specific STING modulation on neuropathology and phenotypic changes in glia relevant to neurodegeneration in vivo.

Highlights.

• Simulated wildfire smoke (WFS) consists primarily of ultrafine particulate matter

• Sub-lethal WFS exposure activates cGAS-STING in cultured astrocytes

• STING-dependent inflammatory cytokines from glia mediate neuronal injury

• Interferon gene expression across systems suggests a conserved response to WFS

Appendix A:

Supplementary materials will be made available online following acceptance.

Data Availability

RNAsequencing data will be made available via publicly accessible data repository upon acceptance.