Microglia-derived extracellular vesicles trigger age-related neurodegeneration upon DNA damage

Significance

Neuroinflammation is a major risk factor for challenging neurodegenerative disorders. Our study shows that persistent DNA damage leads to the accumulation of cytosolic DNA fragments in microglia, stimulating a viral-like immune response in Er1^Cx/− and naturally aged mouse brains. We found that microglia release cytosolic DNAs in extracellular vesicles, causing neuronal cell death. Building on these findings, we developed an anti-inflammatory approach to target activated brain microglia in vivo, eliminate cytosolic DNAs, and postpone the early onset of neurodegeneration.

Abstract

DNA damage and neurodegenerative disorders are intimately linked but the underlying mechanism remains elusive. Here, we show that persistent DNA lesions in tissue-resident macrophages carrying an XPF-ERCC1 DNA repair defect trigger neuroinflammation and neuronal cell death in mice. We find that microglia accumulate dsDNAs and chromatin fragments in the cytosol, which are sensed thereby stimulating a viral-like immune response in Er1^Cx/− and naturally aged murine brain. Cytosolic DNAs are packaged into extracellular vesicles (EVs) that are released from microglia and discharge their dsDNA cargo into IFN-responsive neurons triggering cell death. To remove cytosolic dsDNAs and prevent inflammation, we developed targeting EVs to deliver recombinant DNase I to Er1^Cx/− brain microglia in vivo. We show that EV-mediated elimination of cytosolic dsDNAs is sufficient to prevent neuroinflammation, reduce neuronal apoptosis, and delay the onset of neurodegenerative symptoms in Er1^Cx/− mice. Together, our findings unveil a causal mechanism leading to neuroinflammation and provide a rationalized therapeutic strategy against age-related neurodegeneration.

Accumulation of DNA damage is a hallmark of age-associated neurodegenerative diseases, such as Alzheimer’s disease (1), (2), Parkinson’s disease (3), Huntington’s disease (4), and amyotrophic lateral sclerosis (5). In agreement, congenital DNA repair defects typically associate with neurological symptoms in man and corresponding animal models (6), (7). Irreparable DNA lesions inadvertently interfere with ongoing transcription in postmitotic neurons or obstruct the process of mRNA synthesis and DNA replication in dividing glial cells leading to cellular malfunction, senescence, or cell death (8), (9). To counteract DNA damage, mammalian cells rely on partially overlapping genome maintenance pathways to repair DNA lesions and preserve genome integrity (10).

Neuroinflammation is the activation of an innate immune response in the brain or spinal cord that often precedes neuronal tissue damage and degeneration (11), (12). Microglia, the resident phagocytes of the central nervous system (CNS), are phenotypically and developmentally distinct from peripheral or other tissue-resident macrophage populations and are ubiquitously distributed in the brain to support tissue maintenance and to safeguard the neuronal tissue against foreign pathogens and cellular debris (13). While transiently stimulated microglial cells offer protective effects for the brain, chronic activation leads to persistent release of inflammatory cytokines, overstimulating neuron receptors and causing subsequent damage to neuronal cells, thereby contributing to the premature onset of neurodegenerative disorders (14), (15). DNA damage–driven inflammation causally contributes to cellular malfunction and age-associated tissue degenerative changes (9), (16–20)–(20). However, the relative contribution of persistent DNA lesions in distinct cell types, e.g., neurons or glial cells to age-related neurodegeneration remains elusive.

XPF-ERCC1 (Xeroderma Pigmentosum F-Excision Repair Cross Complementation group 1), a heterodimeric endonuclease complex, is crucial for several DNA repair pathways repair, and for processing various noncanonical DNA structures that may interfere with replication, transcription, or DNA repair events (21–26). Mutations in XPF-ERCC1 in humans lead to Xeroderma Pigmentosum, XFE, or cerebro-oculo-facio-skeletal syndromes, characterized by premature aging features, including progressive neurodegeneration (27), (28). Likewise, animal models with inborn XPF-ERCC1 defects exhibit age-related pathologies, including cerebellar ataxia and cerebral atrophy (28). Using Er1^Cx/− mice carrying an engineered XPF-ERCC1 defect only in tissue-resident macrophages, we provide evidence for a fundamental mechanism wherein irreparable DNA damage triggers sustained immune activation, leading to neuronal cell death. Importantly, our findings underscore a targeted intervention strategy against neuroinflammation and neurodegenerative disorders.

Results

Loss of ERCC1 in Tissue-Resident Macrophages Triggers Progressive Ataxia in Mice.

Neuroinflammation, the innate immune response of the CNS (brain and spinal cord) to an inflammatory challenge, has recently emerged as a core feature of neurodegenerative diseases (13). Tissue-resident macrophages, e.g., brain microglia reside in distinct tissue environments and are vital for tissue homeostasis and defense against pathogens or environmental challenges (29).

To explore the role of brain-resident macrophages in responding to irreparable DNA lesions in vivo, we bred homozygous floxed Ercc1 (Ercc1^F/F) animals with mice carrying the CX3C chemokine receptor (Cx3cr)1-Cre transgene in an Ercc1 heterozygous background (hereafter referred to as Er1^Cx/− animals). The Cx3cr1 promoter was recently shown to be transiently active in neurons during mammalian development (30), (31) pointing to the use of Cx3cr1^CreER mice carrying a tamoxifen-inducible Cre allele as an alternative strategy for targeting microglia. However, tamoxifen is a potent genotoxin that indiscriminately triggers DNA damage to cells making it unsuitable for our studies (32), (33). Thus, we reevaluated the specificity of the constitutive original Cx3cr1-Cre transgene in our work. Confocal microscopy studies in brain cryosections and isolated microglia from Cx3cr1-Cre mice crossed with the Rosa YFP transgenic animals showed Cx3cr1-driven YFP expression in Macrophage-1 antigen (MAC1)⁺ tissue-resident macrophages confirming the specificity of Cx3cr1 promoter in microglial cells (34) (Fig. 1 A and B and SI Appendix, Fig. S1A). MAC1 (CD11b/CD18 heterodimer) is considered an acceptable marker for microglia and peripheral macrophages and TMEM119 can also be used as a selective microglia marker. The TMEM119-positive cell population does not differ from the MAC1-positive cell population in immunofluorescence studies of wt brain cryosections (SI Appendix, Fig. S1B). The specificity of Cx3cr1-Cre was also shown by absence of ERCC1 expression in isolated MAC1⁺ microglial cells from Er1^Cx/− animals expressing the Cx3cr1-Cre transgene compared to Ercc1^F/+ mice expressing the Cx3cr1-Cre transgene (hereafter referred to as wild-type; wt) (SI Appendix, Fig. S1C). Additionally, western blotting of whole cell extracts from Percoll fractions for the enrichment of microglia showed a reduction of ERCC1 expression levels in Er1^Cx/− compared to wt cells (Fig. 1C). In agreement, we find that ERCC1 is expressed in different areas of the CNS, i.e., cerebellum, cerebral cortex, and spinal cord indicating the normative ERCC1 expression levels in Er1^Cx/− cells other than the targeted cell population (Fig. 1D and SI Appendix, Fig. S1D). Last, western blotting of whole cell extracts from the brain and spinal cord confirmed the comparable ERCC1 expression levels in Er1^Cx/− and wt animals (Fig. 1E and SI Appendix, Fig. S1E). Together, our findings support recent observations showing that the Cx3cr1 promoter drives the expression of Cre recombinase largely restrictively in microglia even following excitatory injury (35). Er1^Cx/− mice were born at the expected Mendelian frequency and presented no developmental defects or other pathological features. At 6 mo of age (24 wk), however, we observed progressive signs of ataxia in Er1^Cx/− mice. When the 6-mo-old wt mice were suspended by their tails, the animals extended and swung their hind limbs to maintain balance (Movie S2). In contrast, Er1^Cx/− mice kept their hind limbs in a clasped position (Fig. 1F and Movie S1) and walked with a wide gait compared to age-matched littermate control animals. Rotarod assessments demonstrated a notable hind limb coordination deficiency in the 6-mo-old Er1^Cx/− animals compared to wt littermate controls (Fig. 1G). Beginning at 8 mo (32 wk) of age, Er1^Cx/− animals develop kyphosis (Fig. 1H) and fine tremor to front legs. The early onset of neuropathological features in Er1^Cx/− animals prompted us to assess the morphological and phenotypic characteristics of CNS-resident macrophages. Microglial cells in different CNS areas of 6-mo-old Er1^Cx/− animals formed finger-like protrusions (Fig. 1I and SI Appendix, Fig. S1F). Next, we analyzed the morphology of microglia in the cerebellum using skeletal analysis on Iba1-stained sections. In Er1^Cx/− microglia, we observed an increase in the number of junctions, triple junctions, and total length compared to wt microglia. Sholl analysis was applied to quantify the total number of intersections and the number of intersections relative to the radial distance from the individual cell soma. The total amount of Er1^Cx/− process interceptions was higher than their wt counterparts. Interestingly, Er1^Cx/−microglia displayed a larger soma and an increase in primary and medial process intersections (SI Appendix, Fig. S2A). These changes in microglial morphology are indicative of microglial activation, a hallmark associated with altered cellular locomotion, increased antigen presentation, and enhanced proliferation status (36). Consistently, flow cytometry analysis revealed an increase in the expression of MHC-II (expressed only on antigen-presenting cells) and CD86 (constitutively expressed on activated macrophages) proteins in microglial cells from Er1^Cx/− brains compared to wt controls (Fig. 1J). When compared to lipopolysaccharide-treated proinflammatory microglia, Er1^Cx/− microglia cells exhibit a distinct MHC-II/CD86 expression profile (SI Appendix, Fig. S2B). In contrast to the antigen presentation profile, transmigration for LPS-treated wt microglia is lower compared to both untreated wt and Er1^Cx/− microglial cells, supporting a different activation status between the Er1^Cx/−and bacterial toxin–treated cells (SI Appendix, Fig. S2C). These data are in agreement with previous findings showing that LPS treatment restricts microglia activation (37). Interestingly, transmigration using the Transwell™ chambers is higher for Er1^Cx/− microglial cells compared to wt ones only in the presence of the chemoattractant, ATP (SI Appendix, Fig. S2C). The morphology of Er1^Cx/− microglia, along with the MHC-II-CD86 expression and ATP-induced motility, argues for an inflammatory profile of these cells. A molecular switch that can modulate inflammatory responses is the signal transducer and activator of transcription 1 (STAT1), which is activated by the Janus Kinase (JAK) pathway in canonical interferon signaling and controls inflammation-related gene expression (38). Western blotting of whole cell extracts from microglia enriched Percoll fraction revealed an increase in both total STAT1 and phospho-STAT1 expression levels in Er1^Cx/− compared to wt cells (Fig. 1K). Of note, prolonged incubation periods of rat primary microglia with LPS+IFN-γ were shown to induce both the expression levels of activated STAT1 (p-STAT1) and total-STAT1 (39). To explore the proliferation status of microglia, we report on the numbers of the CD11b⁺CD45^lo population in wt and Er1^Cx/− mice (36). CD45 is a type I transmembrane molecule that is expressed at higher levels on the surface of lymphocytes and peripheral monocytes and at lower levels on resident microglial cells. We find that the percentage of CD11b⁺CD45^lo cells is comparable between the wt and Er1^Cx/− 6-mo-old mice (Fig. 1L and SI Appendix, Fig. S3A). Further analysis in Er1^Cx/− and wt brains showed that there is no significant difference in Annexin V ⁺ PI ⁺ cell populations gated for CD11b, indicating that Er1^Cx/− brain microglia do not undergo apoptosis and that are tolerant to intrinsic DNA damage (SI Appendix, Fig. S3B). Next, we examined whether the observed microglia priming originates from peripheral immune cell infiltration in Er1^Cx/− brains. H&E histological evaluation in the 6-mo-old Er1^Cx/− brains revealed no inflammatory foci (SI Appendix, Fig. S3C). Moreover, the difference in the percentage of Ly6C⁺ bone marrow-derived macrophages showed no statistical significance between Er1^Cx/− and wt brains (Fig. 1M). Consistently, the protein levels of CD45 were comparable between the 6-mo-old Er1^Cx/− and wt brains (SI Appendix, Fig. S3D).

Fig. 1.

Loss of ERCC1 in tissue-resident macrophages triggers progressive ataxia in mice. Cx3cr1-Cre-driven Rosa-YFP expression in (A), brain cryosections and (B) in isolated microglia. (C) Western blotting of ERCC1 protein in whole-cell extracts from the Percoll fraction of brains for the enrichment of microglia. GAPDH was used as a loading control. The graph represents ERCC1 protein levels normalized to GAPDH in Er1^Cx/−samples compared to corresponding wt controls. (D) Immunofluorescence staining of (i). ERCC1 and CALBINDIN (CALB) in mouse cerebella (indicated by the arrows). The numbers indicate the average percentage of ERCC1⁺ CALB⁺ ± SEM in Er1^Cx/− and wt cerebella (n > 70 cells per genotype). (ii and iii) ERCC1 in cortex and spinal cord cryosections (indicated by the arrows). The numbers indicate the ERCC1 Mean Fluorescence Intensity (MFI) of DAPI⁺ nuclei (n > 5 optical fields per genotype). Arrows indicate ERCC1⁺ cells. Single-channel images are shown in SI Appendix, Fig. S1D. (E) Western blotting of ERCC1 protein in whole-cell extracts from different CNS areas. TUBULIN (TUB) was used as loading control. Graph is shown in SI Appendix, Fig. S1E. (F) A photograph of a 32-wk-old Er1^Cx/− mouse and its control littermate depicting the hind limb paralysis developed in Er1^Cx/− mice. (G) A graph depicting the latency to fall (seconds on the rotating rod) during rotarod assessment of the motor coordination of 3-, 6-, 8, and 12-mo-old Er1^Cx/− mice and littermate wt controls, n = 6 mice per group (H). A photograph showing the kyphosis developed in 40-wk-old Er1^Cx/− mice. (I) MAC1 immunofluorescent staining of microglia cells (indicated by arrows) in CER, CTX, HIP, and SC cryosections of Er1^Cx/− and wt mice. Wider optical fields of the same pictures are shown in SI Appendix, Fig. S1F. (J) Activation status of Percoll-isolated microglia from Er1^Cx/− mice and wt littermates. The histograms overlay MHC-II and CD86 expression of CD11b⁺CD45^lo microglia cells (gating strategy shown in SI Appendix, Fig. S3A) from wt and Er1^Cx/− brains. Isotype controls are indicated with a gray dotted line. The graph shows the respective MFIs. (K) Western blotting of total STAT1 and phospho-STAT1 protein in whole-cell extracts from microglia enriched Percoll fraction of brain tissue. GAPDH was used as loading control. The graph represents t-STAT1 and pSTAT1 protein levels normalized to GAPDH in Er1^Cx/− samples compared to corresponding wt controls. (L) Flow cytometry analysis of Percoll-purified cells from 6-mo-old brains stained for CD11b and CD45. Representative flow cytometry plots show the gate for microglia, defined as the CD11b⁺CD45^lo population. The graph depicts the percentage of microglia cells in wt and Er1^Cx/− littermates. (gating strategy shown in SI Appendix, Fig. S3A). Statistical analysis indicated no significant differences. (M) Flow cytometry analysis of single-cell suspensions from Er1^Cx/− and wt mouse brains. The graph depicts the percentage of Ly6C⁺ cells. Statistical analysis indicated no significant differences. Cerebellum (CER), cortex (CTX), hippocampus (HIP), spinal cord (SC). Error bars indicate SEM among n ≥ 3 replicates, unless otherwise stated. The asterisk indicates the significance set at P-value: *≤0.05, **≤0.01 (two-tailed Student’s t test). Scale bar: 10 μm, unless otherwise stated.

Accumulation of Cytoplasmic Chromatin Fragments Triggers a Type-I IFN Response in Er1^Cx/− Microglia.

Next, we sought to determine whether ablation of ERCC1 in microglia is sufficient to lead to constitutive DNA damage signaling. The phosphorylated Ataxia telangiectasia-mutated protein (pATM) is a central mediator of the DNA damage response. Moreover, phosphorylated histone H2A.X (γ-H2A.X)-containing foci accumulate at sites of DNA breaks (40). Confocal microscopy studies showed an increase in the number of pATM⁺/MAC1⁺ and γ-H2A.X⁺/MAC1⁺ cultured microglia deriving from Er1^Cx/− brains compared to wt controls (SI Appendix, Fig. S4 A and B, as indicated). Consistently, the number of γ-H2A.X foci was higher in Er1^Cx/− nuclei from cultured microglia or microglia cells in brain cryosections compared to respective wt controls (SI Appendix, Fig. S4C, Fig. S4D, respectively), indicating that irreparable DNA lesions accumulate in the genome of Er1^Cx/− microglia cells.

DNA damage in the nucleus results in the accumulation of cytoplasmic DNA, notably in the form of micronuclei and cytoplasmic chromatin fragments (CCFs) or in DNA speckles (41). Our analysis showed that cytosolic dsDNAs accumulate in Er1^Cx/− microglia from brain cryosections (SI Appendix, Fig. S4E) and in cultured Er1^Cx/− microglial cells compared to the corresponding wt controls (SI Appendix, Fig. S4F; as indicated). Moreover, we find that DAPI stained foci (widely used in chromosome staining) colocalized with γ-H2A.X (also marking the presence of chromatin) in the cytoplasm of Er1^Cx/− cultured microglial cells (SI Appendix, Fig. S4G). Further work revealed that pATM and γ-H2A.X foci accumulate in the cytoplasm of Er1^Cx/− cells (SI Appendix, Fig. S4H; as indicated). Micronuclei and CCFs contain small chromosome fragments are often surrounded by nuclear envelope and associate with lamina disorganization and rupture (41). In our work, cytosolic chromatin fragments were positive for the DNA damage marker γ-H2A.X in Er1^Cx/− cells but, unlike in micronuclei, they stained negative for lamin A/C or lamin B1 (SI Appendix, Fig. S4I). Consistently, staining with antibodies raised against lamin A/C or lamin B1 revealed no major perturbations in the nuclear lamina of microglial cells either in culture or in brain cryosections (SI Appendix, Fig. S4 I and J, as indicated). Absence of ERCC1 can lead to recombination events between telomeres and interstitial telomeric sequences and finally to shorter telomeres and circular products containing telomeric DNA (25). TelC in situ FISH, however, revealed no striking difference in the nuclear TelC fluorescence signal between Er1^Cx/− and wt cultured microglial cells (SI Appendix, Fig. S5A). Finally, cytoplasmic DNA could also be a direct or indirect, due to replication fork collapse, result of R-loops and DNA lesions (42), (43). Replication fork stalling increases the exposure of ssDNA, allowing Replication Protein A (RPA)–coated ssDNA to trigger activation of ATR, which in turn drives activation of downstream targets, such as Chk1, MLL, and RPA itself (42), (44). To explore the possibility that Er1^Cx/− cells are under replication stress, we performed IF analysis with phosphorylated-RPA (RPA32-S33). We found enrichment of nuclear RPA32-S33 in Er1^Cx/− microglia compared to wt ones (SI Appendix, Fig. S5 B and C for cultured microglia, and brain cryosections respectively).

Genomic instability is often linked to the accumulation of cytoplasmic ssDNA (single-strand DNA) rather than of dsDNA as in Er1^Cx/− microglia (43), (45). The latter led us to hypothesize that secondary structures of repetitive ssDNA sequences generate the dsDNA sequences. Strikingly, we find that Er1^Cx/− cultured microglia accumulate lower levels of cytoplasmic ssDNA compared to wt microglia cells (SI Appendix, Fig. S5D). Moreover, cytoplasmic dsDNA fluorescence intensity in Er1^Cx/− microglia cells was reduced upon recombinant Mung Bean S1 nuclease transfection (SI Appendix, Fig. S5E). Previous reports have shown that insufficient DNA damage repair in Atm^−/− microglia results in active export of AT-rich repetitive elements in the cytoplasm (46). GSAT_MM, the major satellite repeat, was among the top-enriched AT-rich sequence in the cytosolic DNA of ATM-inhibited microglia. GSAT_MM levels were also higher in the cytosolic DNA of Er1^Cx/− microglia compared to the DNA of wt ones (SI Appendix, Fig. S5F). The decrease of cytosolic dsDNA signal in Er1^Cx/− cultured microglia upon S1 nuclease protein transfection and the accumulation of AT-rich sequences in the cytoplasm of Er1^Cx/− cells suggest that reannealed ssDNA could be the origin of the Er1^Cx/− related cytoplasmic dsDNA.

cGMP-AMP (cGAMP) synthase (cGAS) is a cytosolic sensor of microbial or self-dsDNA that was recently shown to sense cytoplasmic DNA due to nuclear DNA damage (41), (47). We find that cGAS accumulates in the cytoplasm of cultured Er1^Cx/− microglia and in microglia from Er1^Cx/− brain cryosections (SI Appendix, Fig. S6 A and B respectively). Moreover, cGAS colocalized with dsDNAs and with DAPI⁺/γ-H2AX⁺ foci in the cytosol of Er1^Cx/− microglia (SI Appendix, Fig. S6C). To test whether nuclear DNA damage in Er1^Cx/− microglia causally contributes to the accumulation of cytosolic dsDNAs in these cells, we next treated wt microglia with etoposide, a potent genotoxin that prevents religation of topoisomerase (TOP) II-mediated DNA double-strand breaks. Similar to Er1^Cx/− microglia, we found that cytosolic dsDNAs accumulate in etoposide-treated wt cells (Fig. 2A and respective graph). Positive cells for colocalized cGAS and dsDNA or DAPI increased in Er1^Cx/− cells and etoposide-treated wt cells (Fig. 2A and graph). However, in Er1^Cx/− and wt^ETO cells cytoplasmic dsDNA/cGAS aggregates were more abundant compared to DAPI/cGAS fragments (Fig. 2A and graph). The specificity of dsDNA staining was further validated through DNase I treatment of wt and wt^ETO cultured microglia postfixation (SI Appendix, Fig. S6E).

Fig. 2.

DNA damage triggers the secretion of type I interferon and EVs carrying dsDNA. (A) Immunofluorescence detection of cGAS and dsDNA in cultured untreated wt, etoposide-treated wt, and untreated Er1^Cx/− microglia. The graph depicts the percentage of cells with cytoplasmic cGAS⁺, cGAS⁺/DAPI⁺ or dsDNA⁺/cGAS⁺ structures and the dsDNA MFI in cultured untreated and etoposide-treated wt and Er1^Cx/− microglia. (B) The graph depicts the expression of pSTING (MFI) in pSTING⁺ microglia in the brains or spinal cords (SC) of 6-mo-old Er1^Cx/− and wt mice, purified with Percoll (n = 3 to 4). Microglia gating and representative histogram plots are shown in SI Appendix, Fig. S6F. (C) Quantitative PCR evaluation of the mRNA levels of interferon signature genes in the brain lysates of 6-mo-old wt and Er1^Cx/− mice (as indicated; RFU: relative fluorescent units). (D) Type I IFN bioactivity (B16 reporter assay OD fold change) in 6-mo-old Er1^Cx/− and age-matched wt brain lavages (n = 4). (E) Western blotting of IFN-β protein in CSF samples. Same volumes of CSF were used in each genotype. The graph represents IFN-β protein levels in wt and Er1^Cx/− samples. (F) EVs isolated from 15 μL wt and Er1^Cx/− CSF were subjected to phenol-chloroform DNA extraction, acrylamide gel electrophoresis, and EtBr staining. Experiment was repeated three times. (G) Scanning (i and ii) and Transmission (iii and iv) electron microscope images of circulating EVs purified from wt and Er1^Cx/− brain lavages. The energy-dispersive-X ray spectra from Er1^Cx/− EVs measured by the scanning electron microscope (ii) are shown in SI Appendix, Fig. S8E. (H) Flow cytometry analysis of purified EVs stained for CD11b and PicoGreen™. EVs were gated for CD11b (n = 3). A representative graph is presented in SI Appendix, Fig. S8F. (I) Immunofluorescence detection of PicoGreen in SH-SHY neurons incubated with wt and Er1^Cx/− EVs, prestained with ExoFlow and PicoGreen. The cytoplasm of SH-SHY cells was labeled with Nestin. The arrows indicate ExoFlow-stained and/or PicoGreen-stained EVs. The graph depicts the percentage of PicoGreen⁺ SH-SHY neuron cells. Error bars indicate SEM among n ≥ 3 replicates. Error bars indicate SEM among n ≥ 3 replicates. The asterisk indicates the significance set at P-value: *≤0.05, **≤0.01 (two-tailed Student’s t test). (Scale bar: 5 μm.)

The release of DNA into the cytosol stimulates STING (Stimulator of Interferon Genes) (48). Once phosphorylated by DNA sensors like cGAS, STING acts as a sensor of cyclic d-GMP and an adaptor protein mediating the IFN response. Type I IFNs execute antiviral functions (49) and under neuroinflammatory conditions, exert potent neurotoxic effects on the brain (50). Flow cytometry analysis revealed higher expression levels of pSTING in the brain and spinal cord microglia cells of 6-mo-old Er1^Cx/− mice (Fig. 2B and SI Appendix, Fig. S6F). Consistently, the mRNA levels of Ifnβ and several interferon signature genes, including the interferon activated (Ifi) gene 207, interferon regulatory factor (Irf)1 and interferon-induced GTP-binding protein Mx1 were higher in Er1^Cx/− whole brain lysates compared to wt controls (Fig. 2C). Using the B16-Blue™ IFN-α/β cell line, we also found that the bioactive murine type-I IFN levels were higher in the extracellular milieu (brain lavage) of 6-mo-old Er1^Cx/− brains compared to age-matched littermate controls (Fig. 2D). An ELISA showed comparable IFN-α levels between Er1^Cx/− and wt brain lavages (SI Appendix, Fig. S6G). However, IFN-β western blotting revealed higher IFN-β protein levels in the Er1^Cx/− cerebrospinal fluid (CSF) compared to wt controls (Fig. 2E). In line, the IFN-β levels were higher when we treated Er1^Cx/− cells with brefeldin A, a protein transport inhibitor typically used to increase intracellular cytokine staining signal by blocking transport processes during cell activation (SI Appendix, Fig. S7A) (51). Likewise, IFN-β levels were higher in brefeldin A–treated wt microglia that were exposed to etoposide compared to untreated wt controls (SI Appendix, Fig. S7A). Ercc1^−/− progeroid animals manifest several pathological features that closely resemble those seen in physiological aging (7), (27), (52). Staining of brain cryosections derived from young (2-mo-old) and old mice (24-mo-old) with MAC1 and γ-H2A.X showed that aged microglia present a higher number of nuclear γ-Η2Α.X foci compared to young microglia, indicating an age-linked genome instability (SI Appendix, Fig. S7B). In line with the cytosolic dsDNA phenotype of Er1^Cx/− microglia, we also found that cytosolic dsDNA and cGAS accumulate in microglial cells of 24-mo-old, naturally aged mice compared to 2-mo-old young adult animals (SI Appendix, Fig. S7C). Consistently, the percentages of pSTING⁺/CD11b⁺ cells were higher in naturally aged brains (SI Appendix, Fig. S7D). Thus, nuclear DNA damage triggers the accumulation of cytosolic dsDNAs and chromatin fragments in Er1^Cx/− microglia, leading to the activation of the type-I IFN response in DNA repair-deficient mice and likely also with aging.

Er1^Cx/− Microglia Secrete Extracellular Vesicles Loaded with Nucleic Acids.

We previously showed that persistent DNA damage accumulation in Ercc1^−/− tissue-infiltrating macrophages triggers the release of extracellular vesicles (EVs), both in vivo and ex vivo (53). Microglia, when activated are also known to efficiently produce and release EVs efficiently (54). In our current study, Er1^Cx/− microglia exhibited an increase in the expression of the EV marker CD9 and CD63 compared to wt control cells, indicating an EV-associated secretory phenotype (SI Appendix, Fig. S8 A and B). To isolate EVs from Er1^Cx/− brain lavages, we performed sucrose gradient ultracentrifugation followed by flow cytometry analysis for the microglia maker CD11b. We found a higher abundance of microglia-derived (CD11b⁺) EVs in the 6-mo-old Er1^Cx/− brains compared to corresponding controls (SI Appendix, Fig. S8C). Importantly, western blot analysis showed a modest enrichment of CD11b, and a more prominent enrichment of the tetraspanin CD81, commonly used as a small EV marker (55) and of the DNA damage marker γ-H2A.X in EVs derived from Er1^Cx/− brain lavages compared to corresponding wt controls (SI Appendix, Fig. S8D). Instead, the expression levels of the nonexosomal tethering protein EEA1, functioning in early endosome fusion, (56) were comparable between Er1^Cx/− and wt control EVs. Together, these findings indicate that Er1^Cx/− microglial cells secrete EVs loaded with cytosolic γH2AX⁺-associated chromatin fragments. The DNA associated with the outer membrane of EVs is typically larger and mostly double-stranded, whereas both single-stranded (ss) and dsDNAs are abundant inside EVs (57). Phenol-chloroform DNA extraction from CSF-derived EVs followed by acrylamide gel electrophoresis and staining of nucleic acids with EtBr indicated the higher abundance of <400 bp DNA species in Er1^Cx/− EVs compared to wt controls (Fig. 2F). Scanning and transmission electron microscopy of the vesicles derived from sucrose gradient fractionation of brain lavages revealed that microglia-derived EVs had a typical size of ~100 nm (Fig. 2 G, i–iv). Furthermore, Er1^Cx/− vesicles appeared to carry a higher density cargo, evidenced by their darker center and the enriched phosphorus marking the presence of positively charged DNA moieties surrounded by a donut-shaped membrane (Fig. 2 G, ii and iv and SI Appendix, Fig. S8E). To further corroborate this finding, EVs isolated from the 6-mo-old Er1^Cx/− and wt brain lavages were stained for CD11b and the fluorescent benzothiazole probe (PicoGreen) that specifically binds dsDNA. Flow cytometry analysis indicated that there is a higher percentage of microglia (CD11b⁺)-derived EVs loaded with dsDNAs in Er1^Cx/− brain lavages compared to wt controls (Fig. 2H and SI Appendix, Fig. S8F).

Recent evidence suggests that EVs can transport proteins, RNA species, and DNA molecules between cells (58, 59)–(60). These findings prompted us to investigate whether Er1^Cx/− microglia-derived EVs mediate the transport of cytosolic dsDNAs in neuronal cells. To test this, we incubated the SH-SHY neuronal cell line, marked with NESTIN-1, a known marker for neuronal progenitor cells (61), with ExoFlow- and PicoGreen-stained EVs derived from Er1^Cx/− and wt brain lavages. Importantly, we detected a higher PicoGreen signal in neuronal cells incubated with Er1^Cx/− EVs compared to wt control EVs for 16 h. Thus, Er1^Cx/− microglia-derived EVs are able to target and efficiently discharge their dsDNA payload to recipient neurons (Fig. 2I).

Er1^Cx/−Microglia Elicit an Antiviral-Like Response That Triggers Neuronal Cell Death.

Confocal microscopy of fluoromyelin and western blotting of Myelin Basic Protein (MBP)1 indicated that myelination is unaffected in 8-mo-old Er1^Cx/−mice (SI Appendix, Fig. S9 A and B). Confocal microscopy studies in wt and Er1^Cx/− brain cryosections showed no difference in the number of NeuN⁺ nuclei bearing γ-H2A.X foci between the two genotypes, further confirming the specificity of Cx3cr1-Cre (SI Appendix, Fig. S9C). However, we observed increased cell death in the Purkinje and granule cell layers of the cerebellum and in the dorsal root of the spinal cord in 6-mo-old Er1^Cx/−mice (Fig. 3A and SI Appendix, Fig. S9D; as indicated). TUNEL fluorescence intensity was similar in the cortex of wt and Er1^Cx/− animals of the same age (Fig. 3A and SI Appendix, Fig. S9D; as indicated). TUNEL positive and negative controls were performed on cryosections of wt cortices (SI Appendix, Fig. S9E). Consistently, flow cytometry analysis of 8-mo-old brain single-cell suspensions stained with Annexin V and Propidium Iodide revealed cell death in ~20 percent of Er1^Cx/−cells (SI Appendix, Fig. S10A).

Fig. 3.

Aged microglia elicit an antiviral-like response that leads to neuronal cell death. (A) Immunofluorescence detection of TUNEL⁺ cells in the cerebellum (CER), cortex (CTX), and the periphery of spinal cord (SC). Arrows indicate only TUNEL⁺ nuclei in the corresponding CNS regions of Er1^Cx/− mice. The graphs depict the percentage of TUNEL⁺CALB⁺ or TUNEL⁺PAX6⁺ cells against the total number of CALB⁺ or PAX6⁺ cells (CER) and the percentage of TUNEL⁺DAPI⁺ cells against the total number of nuclei (CTX and SC) (n = 3 animals, n > 7 optical fields per mouse). Single channel images are shown in SI Appendix, Fig. S9D. (B) Western blotting of IFNAR protein in whole-cell extracts of CNS areas. TUBULIN (TUB) was used as a loading control. The graph represents the IFNAR densitometry analysis normalized to TUB. (C) Immunofluorescence detection of IFNAR in PAX6⁺ and CALBINDIN⁺ cells from cerebellar cryosections and in cryosections of cortices and spinal cords of Er1^Cx/− and wt mice. Arrows indicate the cytoplasmic and membranous localization of IFNAR signal. The graphs depict the IFNAR MFI in indicated cell populations and areas (n > 7 optical fields per genotype). Single channel images are shown in SI Appendix, Fig. S10C. (D) Immunofluorescence detection of IFN-β in cryosections of different CNS regions from Er1^Cx/− and wt mice injected intraperitoneally with Brefeldin A. Arrows indicate IFN-β-positive cells in the corresponding CNS regions. The graph depicts the IFN-β MFI in indicated areas. Immunofluorescence detection of IFN-β in MAC1⁺ cells after Brefeldin A treatment is shown in SI Appendix, Fig. S10D. (E) Immunofluorescence detection of dsDNA in cryosections from different CNS regions of Er1^Cx/− and wt mice. Arrows indicate dsDNA⁺ cells in the corresponding regions. The graph depicts the dsDNA MFI in indicated areas. Single channel images are shown in SI Appendix, Fig. S11A. Error bars indicate SEM among n ≥ 3 replicates. The asterisk indicates the significance set at P-value: *≤0.05, **≤0.01 (two-tailed Student’s t test). (Scale bar: 10 μm.)

Once secreted, type I IFNs signal through type I-IFN receptor (IFNAR) in a paracrine and autocrine manner. The higher type I IFN protein levels and the increased expression of downstream effectors in Er1^Cx/−brains (SI Appendix, Fig. S6F and Fig. 2 C–E) prompted us to test the IFNAR protein levels. Western blot analysis in whole cell extracts showed higher type I-IFN receptor (IFNAR) protein levels in Er1^Cx/−cerebella and spinal cords compared to corresponding wt controls (Fig. 3B). Likewise, flow cytometry analysis indicated higher IFNAR protein levels in cells isolated from Er1^Cx/− cerebella compared to wt controls (SI Appendix, Fig. S10B). Immunofluorescence studies further confirmed the pronounced expression of IFNAR in Purkinje and granule cells of the cerebellum and in cells localized at the dorsal and lateral spinal cord of Er1^Cx/− mice (Fig. 3C and SI Appendix, Fig. S10C). The levels of IFNAR expression in wt and Er1^Cx/− cortices were similar. Together, these data indicate that a fraction of neurons in the CNS of Er1^Cx/− mice is sensitive to type I IFN stimuli. Consistently, immunofluorescence studies in different CNS areas of Er1^Cx/− and wt mice that were intraperitoneally injected with brefeldin A indicated the accumulation of IFN-β in neuron cells located in the cortex, in Purkinje cells, in cells of the granular cerebellar layers (Fig. 3D) and in microglial (MAC1⁺) cells across all CNS areas examined in Er1^Cx/− mice (SI Appendix, Fig. S10D). These findings support recent observations demonstrating that, in addition to brain microglial cells, neurons can produce IFN-β as a result of extracellular IFN-β stimuli and/or intracellular signaling (62). In line with the IFN response in non-microglial cell populations, further analysis revealed the pronounced accumulation of cytosolic dsDNAs in Purkinje, granule cells, and cells of the molecular layer of the cerebellum, as well as in the dorsal and lateral spinal cord of Er1^Cx/− mice compared to corresponding wt controls (Fig. 3E and SI Appendix, Fig. S11A). Postfixation DNase I treatment of Er1^Cx/− cryosections reduced the dsDNA levels in MAC1⁺ cells as well as in PAX6⁺/CALB⁺ cells, further arguing for the specificity of the dsDNA signal (SI Appendix, Fig. S11 B and C). Consistently, we found a higher percentage of CD11b⁻ pSTING⁺ cells in the 6-mo-old Er1^Cx/− cerebella and spinal cords compared to age-matched wt controls (SI Appendix, Fig. S11D).

Microglia-Derived Er1^Cx/− EVs Target IFN-α-Responsive Purkinje Cells Triggering Apoptosis.

Our finding that Er1^Cx/− microglia-derived EVs can target and discharge their dsDNA cargo in neuronal cells in vitro (Fig. 2 F–I) prompted us to hypothesize a similar scenario for Er1^Cx/−animals. We reasoned that Er1^Cx/−microglia package cytosolic dsDNAs into EVs, which are then released and target IFN-responsive, DNA repair-proficient neurons ultimately leading to neuronal cell death and progressive neurodegeneration in Er1^Cx/− animals. The latter would also explain the higher percentage of pSTING cells in the 6-mo-old Er1^Cx/−brains and spinal cords (SI Appendix, Fig. S11D), the higher IFNAR and IFN-β protein levels (Figs. 2E and 3 B–D) and the detection of cytosolic dsDNAs in distinct neuronal cell populations of the cerebellum (Purkinje cells) and the spinal cord (Fig. 3E and SI Appendix, Fig. S11 A–C). To test whether neurons are targeted by microglia-derived Er1^Cx/− EVs, we focused our studies in Purkinje cells, a class of GABAergic inhibitory neurons that play pivotal roles in coordination, control, and locomotor learning. First, we treated cultures of acute wt brain slices with Er1^Cx/− and wt EVs derived from 6-mo-old brains that were previously labeled with the lipophilic green fluorescent dye PKH67. Next, we subjected wt brain slices to simultaneous 2 to 3 multiphoton microscope scanning (SI Appendix, Fig. S12 A–C) allowing us to monitor the selective uptake of microglia-derived Er1^Cx/− EVs by CALBINDIN⁺ cells in at least 200 μm detection depths. Our analysis showed that Er1^Cx/− EVs efficiently target Purkinje cells compared to wt EVs; interestingly, the selective uptake of Er1^Cx/− EVs by Purkinje cells was higher when brain slices were further treated with type I IFN (SI Appendix, Fig. S12D). When dsDNAs in EVs were prestained with PicoGreen, the colocalization of PicoGreen signal with CALBINDIN was more profound in those brain slices treated with Er1^Cx/−microglia-derived EVs compared to the ones treated with wt control EVs. Intriguingly, the higher uptake of Er1^Cx/− EVs by Purkinje cells upon exposure to type I IFN was also followed by a higher PicoGreen uptake in these cells, indicating that the preferential targeting of Er1^Cx/−EVs is followed by the release of the Er1^Cx/−EV dsDNA cargo in recipient Purkinje cells (SI Appendix, Fig. S12E). Importantly, when we delivered ExoFlow-labeled EVs purified from Er1^Cx/−mice intranasally in wt mice and stained the wt cerebella with markers for neuronal cells (class III beta-tubulin—TuJ1), astrocytes (Glial fibrillary acidic protein—GFAP), and oligodendrocytes (adenomatous polyposis coli—CC1), we observed a high localization of ExoFlow puncta in neuronal cells of the cerebellum, primarily in the granule cell area. This does not exclude the possibility of Er1^Cx/−EV uptake in cell types adjacent to neurons. Additionally, a small percentage of ExoFlow puncta were observed in astrocytes and oligodendrocytes (SI Appendix, Fig. S13A). These findings align with our multiphoton microscopy results, further indicating that various neuronal cell types in the cerebellum beyond Purkinje cells can uptake Er1^Cx/−EVs. Finally, staining of acute brain slices with caspase-3 revealed that the exposure of type I IFN–treated brain slices to Er1^Cx/− EVs for 6 h is sufficient to induce Purkinje cell death (SI Appendix, Fig. S13B). Thus, microglia-derived Er1^Cx/− EVs preferentially target and release their dsDNA cargo to IFNAR⁺ neurons, which are responsive to type I interferon, ultimately leading to apoptosis. To further explore the role of IFNAR in the type I interferon and EV-mediated neuronal cell death, we treated wt brain slices with Er1^Cx/− EVs obtained from 6-mo-old brains. We utilized a blocking antibody against IFNAR or an isotype control antibody. Western blot analysis of protein extracts from the treated brain slices showed a reduction of cleaved caspase-3 protein levels after IFNAR blockade (SI Appendix, Fig. S13C). This aligns with our earlier findings, indicating a synergistic effect between type I interferon and Er1^Cx/− EVs for the induction of neuronal cell death.

Intranasal Delivery of DNase I-Loaded EVs Reduces the DNA Damage–Driven Antiviral-Like Response and Neuronal Cell Death in Er1^Cx/−Mice.

To remove cytosolic dsDNAs from microglial cells and reduce the inflammatory load in Er1^Cx/− brains, we next sought to develop an EV-based strategy to deliver recombinant DNase I nuclease and alleviate the dsDNA-mediated antiviral-like response in Er1^Cx/− brains. To do so, we first used the NIH/3T3 cell line to generate EVs loaded with recombinant (pH-independent) DNase I. To selectively target DNase I EVs to microglia cells, the NIH3T3-derived EVs were coated with a custom anti-CD11b peptide derived from a combination of a CD63 binding sequence, i.e., CRHSQMTVTSRL (63) and the αMI-domain binding peptide CP05, i.e., RKLRSLWRR (64). Next, we used an intranasal delivery method as a noninvasive method to bypass the blood–brain barrier and deliver CD11b-ligand EVs to the brain and the spinal cord. Prior to the treatment, the CD11b-ligand EVs were labeled with the exosome-specific dye Exoflow. Immunofluorescence studies in mice treated intranasally with CD11b-ligand EVs revealed that targeting (CD11b ligand-coated) EVs show higher colocalization with microglial (MAC1⁺) cells than the control (noncoated) EVs. Consistently, the ExoFlow dye was detected at similar levels in nonmicroglial (MAC1⁻) cells across all animal groups tested (Fig. 4A). In an effort to identify the subset of microglial cells capable of uptaking EVs loaded with a protein cargo in vivo, we administered ExoFlow-stained, CD11b-ligand coated EVs containing a recombinant protein conjugated with YFP. Our observations indicate that approximately 30% of MAC1+ cells colocalized with labeled EVs, with the fluorescence from the exosome dye coinciding with the YFP fluorescence in all measured microglial cells (SI Appendix, Fig. S14A). Notably, the delivery of targeting DNase I EVs to etoposide-treated microglia ex vivo eliminated cytosolic dsDNAs and cGAS compared to empty (naïve) EVs (Fig. 4B). Having established that CD11b-ligand EVs loaded with DNase I can efficiently target microglia and remove cytosolic dsDNAs ex vivo, we next tested the in vivo efficacy of engineered targeting DNase I EVs in reducing neuroinflammation and neuronal cell death in the Er1^Cx/− mice. To do so, targeting DNase I EVs were administered intranasally in 12-wk-old Er1^Cx/− animals, twice a week and for 6 to 15 wk after treatment with a vasoconstrictor to prevent drainage of EVs from blood vessels into the tissues lining the nasal passages. We performed immunofluorescence studies in brain cryosections of Er1^Cx/−mice that had received targeting DNase I EVs intranasally, and observed a reduction in the levels of dsDNA in the cytoplasm of brain microglia in vivo (Fig. 4C). In agreement with this, an attenuation in the type I interferon response was also detected at 6 wk posttreatment, as measured by the type I IFN protein levels in Er1^Cx/− brain lavages (Fig. 4D). Moreover, in vivo targeting of cytoplasmic DNA resulted in a lower percentage of MHC-II+CD86+ activated Er1^Cx/− microglial cells 6 wk after the treatment, indicating a substantial reduction in neuroinflammation (Fig. 4E and SI Appendix, Fig. S14B). Importantly, the administration of targeting empty EVs kept the percentage of MHC-II+CD86+ wt microglial cells at low levels compared to the activated Er1^Cx/− microglial cells (Fig. 4E). Besides Er1^Cx/− microglia, treatment of Er1^Cx/− mice with targeting DNase I EVs also reduced the PicoGreen-stained dsDNA signal of Er1^Cx/− microglia-derived (CD11b⁺) EVs (SI Appendix, Fig. S14C). Further work revealed that unlike with targeting naïve EVs, the intranasal administration of targeting DNase I EVs reduced the percentage of Annexin⁺ PI^- and Annexin⁺ PI⁺ cells in Er1^Cx/− mice (Fig. 4F and SI Appendix, Fig. S14D). Strikingly, rotarod assessment of motor coordination in wt and Er1^Cx/− mice treated with DNase I-loaded EVs showed an attenuation of the neurodegenerative symptoms in Er1^Cx/− animals. Naïve and DNase I-loaded EVs treatment started at 12 wk of age and continued for 15 wk until the animals were 27 wk old. Animals receiving DNase I-loaded EVs exhibited the most prominent latency differences compared to animals receiving targeting naïve EVs between 21 and 27 wk of age (i.e., 9 and 15 wk of EV treatment respectively; Fig. 4G). Notably, wt mice receiving DNase I or naïve EVs showed no particular latency changes in the span of 15 wk of treatment (SI Appendix, Fig. S14E). Thus, the use of targeting DNase I EVs can efficiently remove cytosolic dsDNAs and reduce the antiviral-like response and neuronal cell death in Er1^Cx/− mice, thereby providing a rationalized therapeutic strategy against age-related neuroinflammatory disorders.

Fig. 4.

CD11b-ligand decorated NIH-derived EVs loaded with DNase I preferentially target microglia cells ameliorating the antiviral-like response and neuronal cell death in Er1^Cx/− mice. (A) Immunofluorescence detection of ExoFlow prestained EVs with or without CD11b-ligand decoration after intranasal administration in wt mice. Arrows indicate CD11b⁺ExoFlow⁺ or CD11b⁻ ExoFlow⁺ cells. Graphs depict the percentage of CD11b⁺ExoFlow⁺ or CD11b⁻ExoFlow⁺ cells. (n > 1,000 cells counted in at least four optical fields each derived from three mice). (B) Immunofluorescence detection of MAC1, cGAS, and dsDNA in etoposide-treated wt microglia. Microglia was cultured in the presence of CD11b-ligand decorated EVs loaded with or without DNase (naïve EVs). The graph depicts cytoplasmic cGAS and dsDNA MFI in ETO-treated microglia cultured in the presence of naïve or DNase I-loaded EVs. (C) Immunofluorescence detection of dsDNA and MAC1 in brain cryosections of Er1^Cx/− mice treated with DNase I-loaded or naïve EVs (30 Units of DNase I/administration, 12 to 30 intranasal instillations, once every 3 d). (n = 3) (D) Type I IFN bioactivity in the lavage of Er1^Cx/−mouse brains after intranasal administration of DNase I-loaded or naive EVs (12 intranasal instillations) (n = 4) (E) Flow cytometry analysis of brain single-cell suspensions stained for CD11b, MHC-II, and CD86. The graph depicts the percentage of MHC-II⁺CD86⁺ microglial cells of 18-wk-old wt mice treated with naïve EVs and Er1^Cx/− mice treated with DNase I-loaded or naive EVs (12 intranasal instillations). Gating strategy for microglia is shown in SI Appendix, Fig. S14B. (F) The graph depicts flow cytometry analysis of brain single-cell suspensions isolated from 18-wk-old Er1^Cx/− or wt mice treated with DNase I-loaded or naive EVs stained for Annexin V and PI (as indicated, 12 intranasal instillations). Representative graphs are shown in SI Appendix, Fig. S14D. (G) Line graph depicting the motor coordination ability (latency to fall from the rod during rotarod assessment) of 3-mo-old wt mice receiving naïve EVs and Er1^Cx/− mice receiving DNase I-loaded or naïve EVs for a time period of 15 wk (30 intranasal instillations). The respective graph showing wt mice treated with DNase I EVs is shown in SI Appendix, Fig. S14E. (n = 4). Error bars indicate SEM among n ≥ 3 replicates. The asterisk indicates the significance set at P-value: *≤0.05, **≤0.01 (two-tailed Student’s t test).

Discussion

Until recently, endogenous DNA damage in postmitotic neurons was thought to be the primary cause of age-related neurodegenerative disorders seen in DNA repair-deficient patients and respective animal models (65). Besides neurons, brain degenerative changes involve the dysfunction of astrocytes, microglia, and oligodendrocytes, leading to neuronal cell death (13, 14)–(15). Activated microglia exacerbate the inflammatory response observed in the course of Alzheimer’s or Parkinson’s disease and with aging (66–72). Moreover, microglia cells derived from Ataxia telangiectasia patients secrete neurotoxic factors that promote synaptic loss and neuronal apoptosis (73), (74). Consistently, the aberrant neuromotor impairment seen in mice deficient in nucleotide excision repair coincide with microgliosis, a reaction of CNS microglia to pathogenic stimuli (75), (76). Thus, microglial cells have a prominent role in neurodegeneration, irrespective of other cell-autonomous neuronal deficits.

Er1^Cx/− mice are born at the expected Mendelian frequency, grow normally, are fertile, and show no visible pathological signs until adulthood. Αt 6 mo of age however, Er1^Cx/− mice exhibit marked signs of ataxia, a common feature of several DNA damage–driven progerias (77) associated with neuronal cell death. This outcome is unexpected because in Er1^Cx/− animals, neurons are proficient in DNA repair and do not accumulate DNA damage. Importantly, we find that γ-H2A.X-associated chromatin and dsDNAs build up in the cytosol of Er1^Cx/− and naturally aged microglia, stimulating the cGAS–STING signaling pathway and a type I IFN response. Like Er1^Cx/− microglia, wt microglia exposed to an exogenous genotoxin show the accumulation of cytosolic dsDNAs and cGAS aggregates. Similarly, Atm^−/− microglial cells amass cytoplasmic DNAs, activating the cGAS–STING pathway and inducing a proinflammatory response in vitro. (74), (78). Cytosolic DNAs could originate from e.g., DNA damage–driven R-loops (79), (80), chromosome segregation defects (micronuclei) (43), rupture of the nuclear envelope (CCFs) (43) or defects in the regulation of DNA end resection during repair events and/or replication fork stalling due to DNA lesions (81) or non-B DNA structures (26), (46), (82).

DNA species translocate into the cytosol passively by mitotic-driven nuclear envelope breakdown (83) or actively by a CRM1-dependent mechanism (46), (79). The latter scenario may signify a physiological response of the nucleus to eliminate irreversibly damaged DNA fragments or byproducts of DNA damage repair. Our findings indicate that the cytosolic dsDNA identified in Er1^Cx/− microglia may arise from the annealing of repetitive-sequence-rich ssDNA, a consequence of the accumulation of DNA lesions and replication fork stalling. Importantly, Er1^Cx/− microglia secrete EVs carrying H2A.X-associated chromatin and cytosolic dsDNAs that target and deliver their dsDNA cargo to recipient neurons ex vivo triggering cell death. In agreement, we demonstrate that dsDNA fragments accumulate in distinct Er1^Cx/− brain regions and the spinal cord, associated with increased IFNAR, p-STING, and IFN-β levels in these areas. Chronic exposure to increased type I IFN levels can trigger neuroinflammation and neurodegeneration (84). In our work, blocking the type I IFN receptor reduces the EV-mediated cell death in acute brain slices. Notably, the uptake of microglia-derived EVs by neurons is further pronounced when acute brain slices are treated with type I IFN. These findings indicate that the presence of microglial DNA in the cytosol of neuronal cells acts as an alarming, viral-like signal triggering widespread neuroinflammatory responses and ultimately leading to neurodegeneration.

EVs are nonimmunogenic carriers allowing their therapeutic cargo to circulate for extended periods within the body (85). To prevent neuroinflammation and potentially delay the premature onset of neurodegeneration in mice, we developed an EV-based strategy to deliver recombinant DNase I in primed Er1^Cx/− microglia cells in vivo. Intranasal administration of targeting EVs decreased the accumulation of cytoplasmic microglial dsDNA but also of dsDNA content from microglia-derived Er1^Cx/−EVs further maximizing the beneficial outcome of our treatment. Removal of cytosolic DNAs by targeting EVs decreased the percentage of activated Er1^Cx/−microglia cells, reduced the IFN-β levels in brain lavages, lessened neuronal cell death, and rescued the motor deficit of Er1^Cx/− animals. As DNase I-loaded EVs eliminate cytoplasmic dsDNA fragments from Er1^Cx/− microglia, the EV DNase I treatment would also prevent microglia from entering the IFN response microglia state and being neurotoxic. Importantly, removal of cytosolic dsDNAs by targeted delivery of DNase I to microglia averts downstream immunogenic stimuli without interfering with antiviral responses or the homeostatic functions of the cGAS–STING pathway. Thus, as DNA damage–associated dsDNAs accumulate over time in the cytosol of microglia, an EV-based therapeutic scheme could offer a promising therapeutic strategy to combat age-related neuroinflammation and improve the outcome of neurodegenerative disorders associated with aging (16), (86).

Methods Details

Detailed materials and methods are provided in SI Appendix, including all materials and instruments, imaging procedures and all in vitro and in vivo biological assays. Key techniques and protocols used in this manuscript are summarized below.

Cell Culture.

Microglial cells isolated from brains or spinal cords after collagenase dissociation and Percoll gradient purification were seeded on 24-well plates and cultured for 3 to 5 d in full DMEM.

Animal Studies.

Animals homozygous for the floxed Ercc1 allele (Ercc1^F/F) were intercrossed with mice carrying the Cx3cr1-Cre transgene to obtain inactivation of the Ercc1 gene Er1^Cx/− animals. Mice lacking the Cx3cr1-Cre transgene were used as wt controls (Er1^F/+, denotedas wt). All animal experiments received ethical approval by independent Animal Ethical Committee at IMBB-FORTH.

EV Isolation, Labeling, and Loading.

EVs were purified from brain lavages or NIH/3T3 cells with the differential ultracentrifugation protocol and the sucrose gradient protocol when necessary. Purified EVs were then labeled and/or incubated with a CD11b-ligand and a non-NLS bearing signal peptide DNase I.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

The video file depicts the 6-month-old Er1^Cx/- mice that keep their hind limbs in a clasped position and walk with a wide gait compared to age-matched littermate control animals when they are suspended by their tails.

Movie S2.

The video file depicts the 6-month-old wt animals that extend and shake their hind limbs to maintain balance when they are suspended by their tails.

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.