STING-dependent sensing of self-DNA drives silica-induced lung inflammation

Silica particles induce lung inflammation and fibrosis. Here we show that stimulator of interferon genes (STING) is essential for silica-induced lung inflammation. In mice, silica induces lung cell death and self-dsDNA release in the bronchoalveolar space that activates STING pathway. Degradation of extracellular self-dsDNA by DNase I inhibits silica-induced STING activation and the downstream type I IFN response. Patients with silicosis have increased circulating dsDNA and CXCL10 in sputum, and patients with fibrotic interstitial lung disease display STING activation and CXCL10 in the lung. In vitro, while mitochondrial dsDNA is sensed by cGAS-STING in dendritic cells, in macrophages extracellular dsDNA activates STING independent of cGAS after silica exposure. These results reveal an essential function of STING-mediated self-dsDNA sensing after silica exposure, and identify DNase I as a potential therapy for silica-induced lung inflammation.

O riginally associated with mining and stone industry, new causes of silicosis include denim sand blasting [1][2][3][4][5] , and the handling of frac sand for shale gas industry 6 . Indeed, drilling and fracking processes produce fine particles, such as silica particles that are retained in the lungs of silica-exposed workers and urban residents, and may lead to severe lung damage, silicosis, or idiopathic pulmonary fibrosis 7,8 . Silicosis is a chronic progressive fibrotic lung inflammation associated with increased cancer, tuberculosis, and chronic obstructive pulmonary disease 9 .
Here, we show that STING is activated in the lung tissue from patients with fibrotic interstitial lung disease (ILD). Mouse airway exposure to silica microparticles induces cell death, self-dsDNA leakage, and inflammatory response through STING-dependent type 1 IFN signaling and downstream CXCL10 expression. Interestingly, patients with silicosis exhibit increased circulating self-dsDNA, together with increased concentrations of CXCL10 in sputum. DNA is central as degradation of extracellular DNA by DNase I in vivo prevents the STING pathway activation and silica-induced lung inflammation. DNA sensor cGAS contributes to STING activation after silica in vivo exposure. Thus, STING, by sensing dsDNA from dying cells plays a key role in silicainduced lung inflammation and DNase I treatment abrogates this response.

Results
Airway silica induces self-dsDNA release and IFN-I response. Silica microparticles intratracheal exposure induced self-dsDNA release in the bronchoalveolar space (Fig. 1a). This was accompanied by the overexpression of STING (Tmem173) and cGAS (Mb21d1) genes in the lungs 7 days (Fig. 1b, c) and 4 weeks after silica exposure ( Supplementary Fig. 1a-c). The overexpression of Irf3, Irf7, and Ifnβ genes was in line with an engagement of the STING pathway at 4 weeks ( Supplementary Fig. 1d, e). Silica induced type I IFNs and downstream CXCL10 expression either at day 7 or 4 weeks following exposure (Fig. 1d, e; Supplementary  Fig. 1e,f). At 4 weeks, silica induced inflammatory cytokines, including IL-1β, TNF, CXL10, and IFN-γ in the lung, together with lung inflammation (Supplementary Fig. 1f and g). The levels of extracellular dsDNA in the bronchoalveolar lavage fluid (BALF) correlated with Ifnα and Ifnβ overexpression in silicaexposed WT mice on day 7 (Fig. 1f), and already at 24 h (Supplementary Fig. 1h-j), suggesting that this early dsDNA release might induce type I IFN expression. We verified in TLR2/TLR4 double-deficient mice that the effect of silica was not due to endotoxin or other PAMPs ( Supplementary Fig. 1k-n). We then assessed the contribution of type I IFN and IL-1 pathways by exposing IFNAR-and IL-1R1-deficient mice to silica.
Interestingly, the level of dsDNA in the BALF was not affected by the absence of functional IFNAR pathway, indicating that silica exposure induced dsDNA release upstream of type I IFN response. However, the inflammatory response in response to silica was reduced in the absence of either type I IFN or IL-1 signaling pathways, including neutrophil recruitment and CXCL10 release, while IFN-αβ production was unaffected by the absence of IL-1 signaling ( Supplementary Fig. 1o-p), confirming the role of the type I IFN pathway for mounting the inflammatory response to silica.
We then asked whether self-dsDNA leakage was the result of cell death induced by silica exposure. Indeed, cell death was highly increased in the lung of WT mice exposed to silica microparticles (Fig. 1g), correlating with dsDNA release in the airways (Fig. 1h). We further characterized the source of self-dsDNA released in the BALF of WT mice exposed to silica. Nuclear DNA (nDNA) and mitochondrial DNA (mtDNA) were increased in the BALF after silica exposure (Fig. 1i). To better characterize the mode of cell death induced by silica microparticles in the lung, we first assessed caspase 3 cleavage as a marker of apoptosis. We show that silica exposure induced caspase 3 cleavage in WT mice, indicative of silica-induced apoptosis (Fig. 1j, k). The contribution of pyroptosis and necroptosis in silica-induced cell death was assessed by immunoblots of cleaved gasdermin D (GSDMD) and phosphorylated mixed lineage kinase domain-like pseudokinase (MLKL) (Fig. 1j). Indeed, silica exposure induced the cleavage of GSDMD and phosphorylation of MLKL in the lung of WT mice (Fig. 1j, l, m). In summary, airway silica exposure induces lung cell death through apoptosis, necroptosis, and pyroptosis, with nuclear and mitochondrial self-dsDNA release, triggering type I IFN pathway activation.
DNase I prevents airway silica-induced STING activation. To address the role of released endogenous dsDNA in silica-induced lung inflammation, we treated mice exposed to silica particles with DNase I (Fig. 1n). DNase I treatment reduced the amount of dsDNA recovered in the bronchoalveolar space 7 days post silica exposure (Fig. 1o). We show that silica induced the expression of STING protein and the activation of STING pathway, as documented by STING phosphorylation and dimerization, phosphorylation of TBK1 and IRF3 in the lung of exposed mice (Fig. 1p, q). STING dimer immunoblot was visualized under seminative conditions, while reducing conditions disrupted the dimeric form of STING (Fig. 1r). In vivo DNase I treatment prevented silica-induced STING expression, phosphorylation and dimer formation, TBK1 and IRF3 phosphorylation in lung homogenates (Fig. 1p, q), as well as STING overexpression in the lung parenchyma (Fig. 1s), demonstrating that extracellular dsDNA is central to STING activation pathway in the airways. Silica-induced cell death was reduced by DNase I treatment, in terms of caspase 3 cleavage and MLKL phosphorylation (Fig. 1p). Further, DNase I prevented silica-induced airway inflammation with a significant decrease of lung IFN-α and IFN-β, as well as of CXCL10 release (Fig. 1t). Importantly, the recruitment of neutrophils and macrophages, and protein extravasation in the BALF seen 7 days post silica exposure were largely reduced by DNase I treatment (Fig. 1u). Therefore, our data identify endogenous dsDNA as an integral actor of STING pathway activation and lung inflammation induced by silica exposure.
Circulating dsDNA and CXCL10 in silicosis patients. Although mining and stone industry were traditional sources of silica exposure, new sources such as sand blasting in the denim industry raise the expected prevalence of silicosis to over 96% 3,19 .  t Pulmonary IFN-αβ and CXCL10. u Neutrophils, macrophages, and protein extravasation in the BALF. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (Kruskal-Wallis test followed by Dunn post test). The correlation analysis was nonparametric (Spearman correlation). Data are mean ± SEM representative of two independent experiments (a-f: mice per group: n = 5 or 10 (WT NaCl), n = 8 or 10 (WT silica); g-h: mice per group: n = 5 (WT NaCl), n = 5 (WT silica); i-n: mice per group, n = 5 or 10 (WT NaCl), n = 5 or 10 (WT silica), and n = 5 (WT silica + DNase I)). Immunoblots representative of n = 6 samples from three independent experiments (j, p), or n = 8 samples from two independent experiments (r), quantified in bargraphs with n = 4. Each symbol represents an individual mouse. Source data are provided as a Source Data file Unique characteristics of these patients include their young age, a definite start of exposure, an intense exposure for ca. 4 + 2 years, and a rapid progression of the disease, often fatal 3,19 . Here, we investigated a cohort of 21 young men suffering from silicosis after exposure to silica in denim industry in the mid-2000s (Table 1). We hypothesized that silica might have caused airway cell damage and release of self-DNA able to trigger the type I IFN response, as in the silica murine model. Interestingly, patients with silicosis exhibited increased plasma levels of dsDNA (Fig. 2a). Further, in their sputum, the concentration of CXCL10 was significantly higher than in healthy controls (Fig. 2b). Thus, in silicosis patients, a previous intense exposure to silica leads to high self-dsDNA circulating levels and CXCL10 response.
Activated STING in fibrotic interstitial lung disease. Silica particles can induce fibrotic ILD 8 and we next asked whether STING pathway is activated in this pathology. To address this question, we analyzed pneumonectomy samples from ILD patients undergoing lung transplantation. We show for the first time that the STING pathway is activated in the lung of ILD patients, as seen by STING overexpression, phosphorylation and dimer formation, TBK1 and IRF3 phosphorylation, and downstream CXCL10 production ( Fig. 2c-e), correlating with epithelium damage (Fig. 2f). Interestingly, STING and CXCL10 expression were reduced ( Fig. 2c-e) in the lung of ILD patients undergoing cortisone therapy ( Table 2).
In vitro, silica exposure induced self-dsDNA release in human PBMCs, and this correlated with IFNβ overexpression (Fig. 2g-i) and CXCL10 production (Fig. 2j). STING was highly overexpressed in human PBMCs exposed to silica (Fig. 2k). This was similar to the response elicited by transfected cyclic dinucleotide c-di-AMP, a cDN that signals through STING to induce type I IFN response via a direct binding to STING protein. Silicainduced STING phosphorylation and dimerization, together with phosphorylation of TBK1 and IRF3 in PBMCs, were similar to cDN (Fig. 2l). To identify the source of self-dsDNA, we investigated cell death in silica-exposed PBMCs. Indeed, PBMCs stimulated with silica displayed a double Annexin V + /PI + -stained population, indicating that most cells were dead (Fig. 2m). Further, the level of Annexin V + /PI + PBMCs correlated with self-dsDNA release and IFNβ expression (Fig. 2n). To investigate whether self-dsDNA drove STING-mediated type I IFN response, silica-exposed human PBMCs were treated with DNase I, to reduce extracellular self-dsDNA accumulation in vitro ( Supplementary Fig. 2). DNase I-treatment of human PBMCs did not affect the Annexin V + /PI + population, but decreased extracellular self-dsDNA concentration (Supplementary Fig. 2a-c). Degradation of extracellular self-dsDNA by DNase I reduced STING activation, as seen through the reduction of STING expression, phosphorylation and dimerization, IFNβ expression, and CXCL10 production in human PBMCs exposed to silica (Supplementary Fig. 2d-g).   Table 1), as compared with healthy individuals (ILO 0). c-f Human lung tissue samples of ILD patients treated or not with cortisone (see Table 2). c Confocal images of DNA dye Draq5 (cyan) and STING (red). Bars, 50 µm. Bars, 20 µm for zoomed regions. d Immunoblots of phosphorylated-STING (p-STING), STING, TBK1, phospho-TBK1, IRF3, phospho-IRF3, and β-actin. STING dimers quantification. e CXCL10 levels. f Correlation between STING dimers and epithelial damage scored on human histological tissue sections for necrotic cells, desquamation or denudation, and flattening of the epithelial barrier (indicated by arrows). Control lung tissue (patient G) and fibrotic area (patient D). Bars, 100 µm. g-n Human PBMCs were stimulated with silica microparticles (250 µg/mL) for 18 h, transfected with c-di-AMP (6 µg/mL) or unstimulated. g Extracellular dsDNA in cell supernatant, h IFNβ transcript, i correlation between IFNβ transcripts and released dsDNA. (j) CXCL10. (k) Confocal images of DNA dye Draq5 (cyan) and STING (red). Bars, 5 µm. l Immunoblots of phospho-STING, STING, phospho-TBK1, TBK1, phospho-IRF3, IRF3, and β-actin. m Flow cytometry annexin V/PI analysis gated on singlet cells. n Correlations between dead cells and extracellular dsDNA or IFNβ transcripts. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (Student's t test). The correlation analysis was nonparametric (Spearman correlation). Data are mean ± SEM representative of two independent experiments. a, b Plasma samples from n = 10 healthy controls and n = 21 silicosis patients; b sputum samples from n = 9 healthy controls and n = 13 silicosis patients. c-f c: Confocal image representative of three patients; d: immunoblots shown from left to right for ILD patients H, G, C, A (see Table 2 Thus, STING pathway is activated in the lung of ILD patients, and in vitro silica exposure induced cell death, self-dsDNA release, STING activation, and type I IFN response in human PBMCs. In addition, DNase I treatment shows that self-dsDNA is central for STING-mediated type I IFN response to silica. Self-DNA sensing by cGAS/STING is key for silica response. We next investigated whether self-dsDNA sensing was specific of STING activation pathway after in vivo silica exposure. As TLR9 is involved in self-DNA recognition and type 1 IFN expression 20 , we tested the susceptibility of TLR9-deficient mice (TLR9 −/− ) to silica microparticles. TLR9 −/− mice responded to silica exposure similarly to WT mice in terms of cell recruitment, protein leak, IFN-β, and inflammatory cytokines release, although IFN-α protein was reduced in TLR9 −/− mice ( Fig. 3a, b).
Since STING is activated in the lung of silica-exposed mice ( Fig. 1p-s), we next investigated how central is STING activation in silica-induced lung inflammation. STING −/− mice exhibited a reduction of all parameters of silica-induced lung inflammation, including neutrophil and macrophage recruitment in the airways, total protein extravasation, CXCL10, IL-1β, and TNF ( Fig. 3a), recapitulating the phenotype of IFNAR −/− mice to silica exposure ( Supplementary Fig. 1o-p). Silica-induced IFN-α and IFN-β were reduced in STING −/− mice (Fig. 3b), as were the phosphorylation of TBK1 and IRF3 (Fig. 3c), while dsDNA levels were unchanged in the bronchoalveolar space (Fig. 3d).
As cGAS is one of the main DNA sensors associated with STING activation, we next assessed the contribution of cGAS in silica-induced lung inflammation. cGAS −/− mice exhibited reduced inflammatory cell recruitment and protein extravasation in response to silica (Fig. 3a). CXCL10, IL-1β, and TNF and type I IFNs (Fig. 3a, b) were also reduced in cGAS −/− mice 7 days after silica exposure. The upregulation of STING (Tmem173) transcripts seen in WT mice 7 days after silica exposure was further increased in cGAS −/− mice, while cGAS (Mb21d1) transcript upregulation was not affected by the absence of STING (Fig. 3b). STING pathway is essential for the early response to silica, as STING −/− mice were protected 24 h post silica exposure, while the contribution of cGAS was less pronounced in cGAS −/− mice at this early time point ( Supplementary Fig. 3a), suggesting the contribution of other DNA sensors. Interestingly, the level of dsDNA in the BALF 24 h post silica exposure was not affected by the absence of functional STING pathway, confirming that silica induced dsDNA release upstream of STING and type I IFN responses ( Supplementary Fig. 3a). STING pathway was still crucial at day 35 post silica exposure, as type I IFN response was abolished in the absence of either STING or cGAS (Supplementary Fig. 3b). Lung inflammation was also decreased in STINGand cGAS-deficient mice, with reduced cell infiltration and granuloma 35 days post exposure ( Supplementary Fig. 3c). Our findings indicate that STING drives lung inflammation at both early and late phases of silica exposure, while the role of cGAS DNA sensor was less prominent early after silica exposure.
Silica-induced inflammatory cell death depends on cGAS/ STING. DNA sensing might differ according to DNA source and localization, and we next sought to identify the cellular source of self-DNA in the airways after silica exposure. We characterized the inflammatory cell populations recruited to the lung and their cellular fate 7 days after silica exposure. Neutrophils, and to a lesser extent interstitial macrophages, and dendritic cells were increased in the lung after silica exposure, while alveolar macrophages were unaffected ( Supplementary Fig. 4). The source of self-dsDNA might include dying/dead cells. Using Annexin V/PI staining, we quantified early apoptotic (Ann V + /PI − ) versus dead cells (Ann V + /PI + ) (Fig. 4a-e). Silica exposure induced cell death of interstitial and alveolar macrophages and tenfold more of neutrophils, as well as early apoptosis in DCs from WT mice, while this response was reduced in the lung of STING −/− and cGAS −/− mice (Fig. 4a-e). Interestingly, silica exposure also induced cell death in epithelial and non-hematopoietic, resident lung cells independently of the cGAS/STING pathway (Supplementary Fig. 5). This is in line with the fact that self-dsDNA release was not affected by the absence of the cGAS/STING pathway (Fig. 3d). We next questioned whether STING could be involved in apoptosis, necroptosis, and/or pyroptosis-related cell   death. Indeed, silica induced apoptosis and necroptosis in a STING-dependent manner, since caspase 3 cleavage and MLKL phosphorylation were highly reduced in the lung of STING −/− mice (Fig. 4f). By contrast, silica-induced gasdermin D cleavage was still present in STING −/− mice, showing that STING is not essential for pyroptosis cell death after silica exposure (Fig. 4f). We further addressed the involvement of the type I IFN pathway in early silica-induced cell death. Lung cell death was partially reduced in IFNAR −/− and STING −/− mice 24 h post silica exposure ( Supplementary Fig. 6a-c), although there was already a STING/IFN-I-independent release of dsDNA in the airways (Supplementary Fig. 6d). At this early stage, caspase 3 cleavage and MLKL phosphorylation occurred in the absence of the IFN-I pathway ( Supplementary Fig. 6e). These results show that, beside causing cGAS/STING-independent epithelial cell death, silica induces necroptosis and apoptosis in a STING-dependent manner in the lung 7 days post silica exposure.
STING modulates NLRP3 expression after silica exposure. The NLRP3 inflammasome has been recognized as an innate immune signaling receptor important for mediating cell responses and inflammation to crystals, such as silica 14,16,[21][22][23] . To dissect the effect of STING adaptor molecule on NLRP3 expression and pyroptosis, NLRP3-deficient (NLRP3 −/− ) and gasdermin Ddeficient (GSDMD −/− ) mice were exposed to silica  Fig. 7 and 8). The absence of NLRP3 led to a reduction of cell death in interstitial and alveolar macrophages but also of apoptosis in DCs after in vivo silica exposure (Supplementary Fig. 7a-c), which however did not translate into a reduction of dsDNA release in the airways ( Supplementary  Fig. 7d). Silica-induced lung inflammation was reduced in NLRP3 −/− mice as compared with WT mice (Supplementary Fig. 7e-f). Interestingly, type I IFN response was also reduced in the absence of NLRP3 after in vivo silica exposure ( Supplementary Fig. 7g, h). In order to understand how type I IFN signaling is reduced in NLRP3 −/− mice, we evaluated STING/TBK1 signaling in the absence of NLRP3 in vivo. Indeed, Tmem173 transcripts, STING protein expression, phosphorylation, and dimerization but also phosphorylation of TBK1 were partially reduced in the absence of NLRP3 in vivo ( Supplementary Fig. 7i-k). Conversely, STING is important for NLRP3 expression, as seen by the decrease of NLRP3 protein and gene expression in the lung of STINGdeficient mice after silica exposure ( Supplementary Fig. 7j, l, m). Although inflammatory cell death was reduced in GSDMD −/− mice, dsDNA release and lung inflammation were not affected 7 days post silica exposure (Supplementary Fig. 8). Thus, STING and NLRP3 pathways are linked to induce lung inflammatory response to silica.
Silica activates ROS, cell death, and STING in macrophages.
Macrophages play a key role in silicosis and inflammatory diseases, especially via their ability to internalize microparticles, nucleic acids, and dying cell-associated intracellular content 24 . To determine the role of macrophages following silica stimulation, we first evaluated their expression of co-stimulatory molecules in response to silica in vitro. We showed that the overexpression of CD40, C86, and MHC-II after macrophage stimulation with silica or cDN was abrogated in the absence of STING ( Supplementary  Fig. 9a). Following silica stimulation, macrophages underwent cellular stress with reactive oxygen species (ROS) production detected by MitoSOX immunofluorescence which co-localized with mitochondria revealed by MitoTracker (Fig. 5a, b). Interestingly, the induction of ROS after silica stimulation was impaired in STING −/− macrophages (Fig. 5a, b). Using live in vitro video microscopy, we documented that silica microparticles present in the culture were internalized by macrophages which underwent drastic morphological changes to engulf silica particles up to 12-24 h before cell death and silica particles externalization (Supplementary online video). Indeed, there was a sharp increase in Annexin V/PI double-positive macrophages 18 h post silica exposure (Fig. 5c), together with self-dsDNA release in the cell supernatant (Fig. 5d). Macrophage cell death after silica exposure was dependent on ROS induction as it could be partially prevented in the presence of ROS scavenger (Supplementary Fig.  10). Silica exposure induced apoptosis and necroptosis in macrophages, as revealed by cleaved caspase 3 and phosphorylated-MLKL (Fig. 5e). Interestingly, in silica-exposed STING −/− macrophages, caspase 3 cleavage and phospho-MLKL were reduced, as compared with WT macrophages (Fig. 5e). Gasdermin D was not cleaved after in vitro silica exposure, which may reflect the lack of inflammasome activation, due to the absence of the first NFκB signal (Fig. 5e) 25 . Interestingly, macrophages stimulated with silica exhibited Draq5-positive DNA-containing cytoplasmic structures co-localized with silica microparticles (Fig. 5f). The STING signaling pathway was activated in macrophages exposed to silica, including STING phosphorylation and dimerization, and phosphorylation of TBK1 and IRF3 (Fig. 5g). Indeed, STING was essential for silica-induced macrophage death, as STING −/− macrophages were protected (Fig. 5c). Further, intracellular DNA associated with silica microparticles induced STING overexpression and aggregation, as seen after transfection of c-di-AMP (cDN; Fig. 5h). Thus, in vitro silica exposure induced ROS, macrophage cell death, dsDNA release, intracytosolic DNA associated with silica microparticles, and activation of the STING pathway.
Silica induces STING-dependent apoptosis in DCs. Dendritic cells were recently shown to sense tumor dsDNA differently to macrophages 26 . To dissect whether the STING-dependent activation mechanism was different in DCs and macrophages, we next investigated the response of DCs to silica microparticles and the role of STING in this response. In vivo, DCs recruited to the airways after silica exposure initiated a cell death program (Fig. 4a, d). In vitro, BMDCs stimulated with silica microparticles for 24 h were activated, and the overexpression of CD40, CD86, and MHC class II was reduced in STING-deficient DCs (Supplementary Fig. 6b). Mitochondrial ROS expression was increased in silica-exposed wild-type DCs, but not in STING −/− DCs (Fig. 6a). This oxidative stress was associated with an increase in apoptotic Annexin V + /PI − and dead cells (Annexin V + /PI + ) 18 h post silica exposure (Fig. 6b), together with extracellular dsDNA release (Fig. 6c). Silica exposure induced STINGdependent apoptosis in DCs, as revealed by cleaved caspase 3, reduced in STING −/− DCs, while MLKL was not phosphorylated and gasdermin D was not cleaved (Fig. 6d). To examine whether dsDNA derived from DCs could be sensed by STING, DCs were double-stained with DNA dye Draq5 and STING-specific antibody. Interestingly, DCs stimulated with silica exhibited multiple DNA-containing subcellular structures inside the cytosol that colocalized with silica microparticles. There was a strong overexpression of STING and speck formation in these cells, beyond the perinuclear region (Fig. 6e). Further, the STING signaling pathway was activated in DCs after silica exposure via STING phosphorylation and dimerization and also phosphorylation of the TBK1/IRF3 axis (Fig. 6f). Together, these results indicate that silica is an activator of the STING signaling pathway in DCs through induction of mitochondrial stress, apoptosis-mediated cell death, and self-dsDNA release.
Extracellular DNA mediates silica macrophage activation. Recent work has indicated that immune cells can sense self-dsDNA present in the extracellular milieu 27 . We therefore asked whether the dsDNA released upon silica exposure could activate bystander macrophages, using DNase I to degrade extracellular dsDNA (Fig. 7a, b). Extracellular self-dsDNA accumulation after macrophage exposure to silica microparticles was prevented by DNase I treatment at 18 h (Fig. 7b). DNase I-treated macrophages showed a reduced pro-inflammatory response such as CXCL10, a complete abolition of Ifnα and Ifnβ message, and IFN-αβ protein overexpression in response to silica (Fig. 7b). To assess whether self-dsDNA is the major DAMP induced by silica, we next transfected DNA from silica-exposed lung into macrophages in vitro (Supplementary Fig. 11a). DNA from silica-exposed lung, but not from vehicle-treated mice, induced TNF proinflammatory cytokine release, and type I IFN responses, recapitulating silica macrophage activation, similar to c-di-AMP transfection ( Supplementary Fig. 11b). This response was STINGdependent as it was abrogated in STING −/− macrophages ( Supplementary Fig. 11b).
We next addressed the relevance of extracellular DNA sensing to DCs, by treating bone marrow-derived dendritic cells with extracellular DNase I (Fig. 7a). Surprisingly, extracellular self-dsDNA degradation did not affect silica-induced IFN-α and IFNβ protein overexpression by DCs (Fig. 7c). We next verified whether transfected self-dsDNA from silica-exposed lung could c Dendritic cell IFN-α and IFN-β protein concentrations determined in culture supernatants by multiplex immunoassay. d-i Mitochondrial DNA replication was inhibited using a low concentration of EtdBr (150 ng/mL) on day 7 of bone marrow cell culture. On day 11, bone marrow-derived DCs (e-h) and macrophages (i) were unstimulated, stimulated with silica (250 µg/mL), or transfected with c-di-AMP (6 µg/mL; cDN) for 18 h. Fold change of (e) mitochondrial DNA (mMitoF1 and mMitoR1) and (f) nuclear DNA (mB2MF1 and mB2MR1) in untreated versus EtdBr-treated DC exposed to silica, as compared to untreated unstimulated cells. g Immunoblot showing cGAS, phospho-STING, STING protein expression, and dimerization in DCs after EtdBr treatment, with β-actin as a reference. h CXCL10 level in DC supernatant quantified by ELISA and IFN-α and IFN-β concentrations determined by multiplex immunoassay. Tmem173 transcripts measured by real-time PCR. i CXCL10 level in macrophage supernatant quantified by ELISA and IFN-α and IFN-β concentrations determined by multiplex immunoassay. Tmem173 transcripts measured by real-time PCR. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (Mann-Whitney U analysis). Data are presented as mean ± SEM and are representative of two independent experiments with n = 4 independent cultures. Immunoblots are representative of n = 2 samples per condition (g). Source data are provided as a Source Data file activate DCs (Supplementary Fig. 11a). Indeed, transfection of DNA from silica-exposed lung, but not from vehicle-treated mice, induced a DC type I IFN response similar to c-di-AMP together with some induction of TNF and CXCL10 ( Supplementary  Fig. 11c), and this response was STING dependent as it was abrogated in STING −/− DCs (Supplementary Fig. 11c). These data demonstrate that silica particles induce type I IFN response through extracellular self-dsDNA release in macrophages, while the response of dendritic cells to silica is not affected by extracellular dsDNA degradation.
Silica-induced mitochondrial DNA sensed by STING in DC. As DCs were not affected by extracellular DNA degradation (Fig. 7c), we next hypothesized that mtDNA might be the source of self-dsDNA activating STING in DCs. To test this hypothesis, we depleted DCs mitochondrial DNA using a well-established protocol with low-dose ethidium bromide (EtdBr; Fig. 7d). EtdBr intercalating into mtDNA base pairs, prevents its replication, yielding a tenfold reduction of mtDNA while sparing genomic DNA replication 28,29 . EtdBr-treated DCs displayed a drastic reduction of mtDNA without affecting nDNA (Fig. 7e, f). STING protein expression, phosphorylation, and dimerization in response to silica was strongly reduced in DCs after EtdBr treatment, while the response to control c-di-AMP (cDN) was spared (Fig. 7g). cGAS protein levels were similar after silica or cDN stimulation, without or with EtdBr treatment (Fig. 7g). In mtDNA-depleted WT DCs, type I IFNs expression was totally abolished and CXCL10 strongly decreased (Fig. 7h), while the response to c-di-AMP was unaffected, excluding a cytotoxic effect of the EtdBr treatment. The abolishment of type I IFNs in mtDNA-depleted DCs may be a consequence of the impaired cGAS/STING pathway, as we documented no expression of Tmem173 transcript after EtdBr treatment (Fig. 7h). Moreover, the abolishment of Tmem173 transcript in mtDNA-depleted WT DCs indicates that mtDNA is central for regulating the cGAS/ STING pathway in DCs. However, mtDNA depletion did not affect the response of macrophages to silica (Fig. 7i), in line with the fact that the macrophage response was abolished by extracellular DNA depletion (Fig. 7b). Thus, mtDNA is the main source of DAMP activating STING pathway in DCs exposed to silica.
Silica-induced DC response depends on both cGAS and STING. To identify the DNA-binding sensor upstream of STING signaling in DCs, we next investigated the involvement of cGAS that drives type I IFN response in a STING-dependent manner. The response of DCs to in vitro silica stimulation by the induction of TNF, CXCL10, IFN-αβ message, and proteins, together with overexpression of STING and cGAS-coding genes was abolished in STING −/− and cGAS −/− DCs (Fig. 8a). DCs stimulation by c-di-AMP was dependent on both STING and cGAS pathways (Fig. 8a). Thus, in response to silica microparticles, DCs sense the endogenous dsDNA through both cGAS and STING promoting type 1 IFNs response.
cGAS-independent macrophage response to silica. Macrophages rely on extracellular DNA, but not on mtDNA to respond to silica exposure, however, they fail to upregulate cGAS coding gene (Fig. 8b), and we next questioned the contribution of cGAS for STING-mediated macrophage response. The increased release of TNF, CXCL10, together with Ifnα, Ifnβ, and STING-coding gene Tmem173 overexpression induced by silica microparticles in wild-type macrophages was abolished in STING −/− macrophages, or partially reduced for TNF (Fig. 8b). Interestingly, this response was largely independent of cGAS, since cGAS −/− macrophages responded to silica microparticles with a release of TNF, CXCL10, and an overexpression of Ifnα, Ifnβ, and Tmem173 similar to wild-type macrophages (Fig. 8b). Transfection with c-di-AMP induced a type 1 IFN response in wild-type macrophages that was abolished in STING −/− macrophages (Fig. 8b), while cGAS −/− macrophages responded to c-di-AMP, as expected (Fig. 8b). Altogether, in vitro macrophage stimulation by silica activates type 1 IFNs signaling in a STING-dependent manner, promoting inflammations, but this response is largely independent of cGAS DNA sensor.
Silica macrophage response involves STING, DDX41, and IFI204. We thus hypothesized that other upstream DNA sensors might mediate the response to silica and self-dsDNA in macrophages. Among DNA sensors activating STING, IFI204, the mouse ortholog of human IFI16, and the helicase DDX41 genes were overexpressed in silica-exposed macrophages (Supplementary Fig. 12a). Using siRNA ( Supplementary Fig. 12b), we could show that silica-induced activation of the STING pathway, including STING dimerization and phosphorylation of TBK1 and IRF3, was partly reduced after knockdown of IFI204 and more importantly after DDX41 knockdown ( Supplementary Fig. 12c), as were TNF, CXCL10, and type I IFN expression (Supplementary Fig. 12d). As expected, the response to c-di-AMP was largely dependent on DDX41 30 . IRF3 phosphorylation, which can be induced by different signaling pathways including TLR3, TLR4, or MAVS 31 was less affected by DDX41 knockdown. Thus, in macrophages in vitro, DDX41 and IFI204 DNA sensors contribute to STING-dependent type 1 IFN response to silica.
Hematopoietic cell STING mediates silica airway response. STING protein is expressed in endothelial and epithelial cell types, as well as in hematopoietic cells including macrophages, T cells, and DCs 32 . To study the cellular source of STING leading to silica-induced lung inflammation, bone marrow reconstituted STING −/− or wild-type mice were exposed to silica microparticles ( Supplementary Fig. 13). The inability of STINGdeficient mice to activate type I pathway in response to silica was corrected by WT hematopoietic cell reconstitution, and it was partially transferred to WT mice after reconstitution with STING −/− bone marrow cells. Indeed, silica-induced cell death, neutrophil recruitment in the airways, type I IFN expression, and the release of CXCL10 and TNF were reduced after reconstitution of WT mice with STING −/− hematopoietic cells ( Supplementary  Fig. 13a-k). Conversely, the reconstitution of STING −/− mice with WT BM restored the response to silica in terms of cell death, neutrophil recruitment in the airways, type I IFN expression, and CXCL10 and TNF release ( Supplementary Fig. 13c-k). Self-dsDNA release in the BALF after silica exposure was unaffected in the chimeric mice ( Supplementary Fig. 13i), similar to STING −/− mice (Fig. 3d). Thus, STING pathway in hematopoietic cells is crucial for silica-induced type I IFN response and lung inflammation. Nevertheless, the role of STING in stromal cells after silica exposure should not be excluded, since STINGcompetent stromal cells provide some CXCL10 response in WT mice reconstituted with STING −/− bone marrow mice, as compared with total STING deficiency ( Supplementary Fig. 13g).

Discussion
In this study, we show that self-DNA released by lung damage plays a key role for inducing a pro-inflammatory type I IFN response after experimental silica exposure. Indeed, silica particles induced reactive oxygen species, cell death, and release of self-dsDNA correlating with the type I IFN response. We identified STING as a critical mediator for self-dsDNA sensing and type I IFN response upon silica exposure. Importantly, silicosis patients had increased circulating dsDNA and CXCL10 in sputum, and STING was activated in lung tissue from ILD patients. Moreover, silica induced human PBMC cell death, and activation of STING and type I IFN pathways, which were dependent on extracellular dsDNA release. Using mice defective for different signaling pathways, we show that the pro-inflammatory type I IFN response is elicited by STING pathway activation via dsDNA released into the airways after silica in vivo exposure. We further show that silica exposure leads to activation of programmed cell death, apoptosis, necroptosis, and pyroptosis inducing dsDNA leakage. We identify both nuclei and mitochondria as a source of released self-dsDNA activating STING in vivo after silica exposure, with an increased proportion of mtDNA in the airways. Degradation of extracellular self-dsDNA by DNase I considerably reduced lung inflammation and type I IFN response in a STING-dependent manner. In vitro DNase I treatment showed that extracellular DNA is the main source of DNA activating STING pathway and type I IFN response in macrophages. Indeed, macrophages stimulated with silica promoted a cellular stress followed by cell death and extranuclear dsDNA leakage in a STING-dependent   20 . Using mtDNA inhibition by EtdBr intercalation, we excluded mtDNA as the source of DNA responsible for STING activation in macrophages after silica exposure. In DCs, mitochondria was the main source of self-DNA as mtDNA inhibition by EtdBr abolished type I IFN response, while extracellular dsDNA degradation with DNase I had no effect on silica-induced DC response. We showed that STING is more strongly activated in DCs than in macrophages, in line with the fact that self-dsDNA is more persistent inside DCs than macrophages. Indeed, in macrophages, endonucleases rapidly degrade cytosolic DNA, while DCs control mtDNA clearance via SIRPα signaling 26,33 . STING binds cDN, but whether it directly interacts with dsDNA is still a matter of debate 34,35 . The main cytosolic DNA sensors upstream of STING identified is cGAS, and other molecules contributing to cytosolic DNA sensing and type I IFN pathway activation in a STING-dependent manner include DDX41, IFI16 or mouse IFI204, DAI, RNA pol III, LRRFIP1, ZCCHC3 or meiotic recombination 11 homolog A (MRE11), IFI16 and IFI204 cooperating with cGAS, or ZCCHC3 acting as co-factor of cGAS [36][37][38][39][40][41][42][43][44] . Our findings indicate a role for cGAS DNA sensor for inducing type I IFN production after silica in vivo exposure. However, cGAS was dispensable for STINGmediated type I IFN response in the first 24 h post silica exposure, suggesting an early contribution of other DNA sensors. We thus investigated the DNA sensors triggered by silica exposure in vitro.
In DCs, silica exposure induced STING-dependent ROS production, cell death, and self-mtDNA release within DCs leading to type I IFN response. mtDNA engages cGAS-STING axis 20,33 . Indeed, in DCs, we identified the DNA sensor upstream of STING as cGAS, since cGAS-deficient DCs displayed a strong inhibition of type I IFN response after silica exposure. How mtDNA is released from the mitochondria to be sensed by cytosolic DNA sensors is still unresolved 20 . Several potential mechanisms were proposed: mtDNA may access the cytosol through initiation of apoptosis in a BAX and BCL-2-dependent manner 29,45 , activation of mitochondrial permeability transition pore 46,47 , or deficient control by transcription factor A mitochondria (TFAM) 45 .
In vitro, silica-exposed macrophages display extracellular dsDNA release, not from mitochondrial origin but likely from the nucleus, leading to STING-dependent type I IFN response, independent of cGAS but mainly via DDX41 and IFI204 DNA sensors. Furthermore, the question arises as to how self-dsDNA released extracellularly could be sensed by intracellular STING adaptor molecule in macrophages. Self-dsDNA co-localized with silica microparticles in the intracellular compartments of macrophages. Extracellular DNA bound to silica is likely transported from the extracellular milieu into macrophage cytosol 48 . For instance, nucleic acid receptors could be involved in promoting extracellular DNA uptake after silica exposure. Indeed, RAGE protein has been shown to interact with DNA promoting DNA uptake by cells via the endosomal route, inducing Nf-κB activation through TLR9 49 . Further, in pathological conditions, extracellular self-dsDNA can also get access to the cytosol through Fc receptors, forming a chromatin-immune complex able to be internalized by the host cell 50 . Beside, CARGO proteins or peptides such as LL37 may contribute to dsDNA transport into the cell 51 . Micronuclei formed after silica exposure may also act as DAMPs activating type I IFN response in a STING-dependent manner. Indeed, Mackenzie et al. recently reported a role for DNA sensors in sensing micronuclei arising from genome instability or micronuclear envelope breakdown 52 . Micronuclei formation might occur in macrophages after silica exposure as free dsDNA was found next to the nucleus. The remaining open questions include whether dsDNA-silica complexes activate STING and whether dsDNA dissociates from silica inside the cell.
NLRP3 inflammasome activation occurs after silica exposure inducing apoptosis/pyroptosis and the release of the proinflammatory cytokine IL-1β 14,16,53 . Here, we show that STING activation by self-dsDNA sensing is an important pathway to induce lung inflammation after silica exposure, similarly to NLRP3 inflammasome, and that the pathways are linked. Indeed, Gaidt et al. recently showed that STING activation orchestrates a lysosomal cell death program engaging NLRP3 inflammasome after cytosolic DNA recognition 17 . In silica-induced lung inflammation, both STING and NLRP3 are activated eliciting a cell death program and pro-inflammatory cytokines release including type I IFN. Interestingly, NLRP3 expression seems to be dependent on STING activation since STING −/− mice displayed a lower expression of the inflammasome NLRP3 after silica exposure. Conversely, in the absence of NLRP3, silica-induced STING expression and type I IFN response was reduced. We could speculate that STING by inducing reactive oxygen species and cell death leads to the release of DAMPs, including K + ions and self-DNA activating the inflammasome NLRP3. Further, our results indicate that NLRP3 inflammasome could interfere with the cytosolic surveillance system for activating DNA sensor pathways.
STING adaptor molecule is highly investigated in response to bacterial, viral, or self-DNA but the cellular source of STING activating the cytosolic surveillance system is poorly documented. Indeed, STING is highly expressed in hematopoietic cells but also in resident cells such as endothelial and epithelial cells 32 . To discriminate the role of STING in hematopoietic cells versus resident cells, chimeric bone marrow reconstituted STING −/− mice were exposed to silica microparticles. We found that STING-competent hematopoietic cells are crucial for inducing inflammatory cell death, type I IFN response, and the release of pro-inflammatory cytokines, including CXCL10 after silica exposure. STING expression by stromal cells seems to play a minor role, which may be indirect, in response to type I IFNs through IFNAR activation, leading to CXCL10 production. Indeed, silica microparticles cause an initial stress to the airways tissue, with connective tissue breakdown, lung epithelial cell death, and dsDNA release in the airway space independently of the STING pathway, and the persistence of microcrystals in the lung can entertain this original cellular stress 12,54 . Our results indicate that this first response of lung epithelial cell death to silica microparticles is cGAS/STING-independent, while a cGAS/ STING-dependent inflammatory response is induced by the released DNA, inducing ROS and further cell death.
The presence of nucleic acids in the cytosol represents a danger signal activating the cytosolic surveillance system but cells develop strategies to avoid an excessive immune response leading to autoimmunity: endo-and exonucleases such as MUS81, ERCC1-XPF, DNaseII, and TREX1 [55][56][57][58] . Investigation of the nucleases present in hematopoietic and stromal cells could be relevant to appreciate a direct or indirect activation of STING by nucleic acids or cGAMP, possibly through bystander cells, as suggested by Ablasser et al. 59 Self-DNA release including mtDNA has been implicated in various pathologies, such as infections, cancer, autoimmune, or inflammatory diseases, with so far limited implication of the STING pathway 20 . Extracellular mtDNA was recently detected in the BALF of idiopathic pulmonary fibrotic patients 60 , and mtDNA concentration in plasma was identified as a marker of disease progression. However, no mechanistic link has been discussed between mtDNA release and disease. On the other hand, STING has been implicated in pediatric ILD 61 , and gain-offunction mutations in Tmem173 were identified in a new autoinflammatory syndrome 59 (STING-associated vasculopathy with onset in infancy, SAVI), associated with ILD in five out of six patients 62 . Here, we show that silicosis patients with a history of intense silica exposure had increased circulating dsDNA and CXCL10 levels in sputum. Further, we document STING activation in the lung of patients with ILD. Airway silica exposure induced mitochondrial and nuclear self-dsDNA release that triggered STING-dependent type I IFN responses in mice. We propose that the DNA present in the airways activates STING pathway and contributes to the progressive lung inflammation.
In conclusion, the present findings establish a role for self-dsDNA released upon silica-induced injury in activating proinflammatory type I IFN response in a STING-dependent manner. DNase I treatment or targeting the sensing of self-dsDNA by STING may provide a therapeutic strategy to reduce lung inflammation after silica exposure and prevent the associated long-term development of silicosis, cancers, or COPD.

Methods
Tissue donors. Plasma and sputum samples were obtained from patients with silicosis and healthy controls, after written informed consent and with the approval of the local ethics committee (registration number 2018/41; ethics committee, Atatürk University, Faculty of Medicine, Erzurum, Turkey). The relevant patient data are displayed in Table 1.
Lung tissue explants were obtained from patients with proven diagnosis of fibrotic ILD undergoing lung transplantation. All tissue samples were obtained after written informed consent of each patient and with the approval of the local ethical committee (EC-N°1147/2015; ethics committee, Medical University of Vienna, Austria). The relevant patient data are displayed in Table 2.
In vivo animal experiments. Crystalline silica particles (1 mg/mouse in 0.9% saline) or saline vehicle was administered intratracheally (i.t.) to mice under isoflurane anesthesia. BALF was collected and cells were counted 72 . The supernatant of the first lavage was collected after centrifugation and stored at −80°C for dsDNA quantification (Quant-iT™ PicoGreen™). Lung tissues were fixed in 4% buffered formaldehyde overnight, paraffin-embedded, and 3-µm sections were stained with hematoxylin and eosin (HE). Lung IL-1β, TNF, and CXCL10 were quantified by ELISA (R&D systems). Protein extravasation in the BALF was measured by DC™ Protein Assay Kit (Bio-Rad ® ) and IFN-α and IFN-β proteins (Life Technologies ® ) were quantified in the lungs by Luminex Immunoassay System (MAGPIX-BIORAD ® ).