RNase H2 Loss in Murine Astrocytes Results in Cellular Defects Reminiscent of Nucleic Acid-Mediated Autoinflammation

Aicardi–Goutières syndrome (AGS) is a rare early onset childhood encephalopathy caused by persistent neuroinflammation of autoimmune origin. AGS is a genetic disorder and >50% of affected individuals bear hypomorphic mutations in ribonuclease H2 (RNase H2). All available RNase H2 mouse models so far fail to mimic the prominent CNS involvement seen in AGS. To establish a mouse model recapitulating the human disease, we deleted RNase H2 specifically in the brain, the most severely affected organ in AGS. Although RNase H2ΔGFAP mice lacked the nuclease in astrocytes and a majority of neurons, no disease signs were apparent in these animals. We additionally confirmed these results in a second, neuron-specific RNase H2 knockout mouse line. However, when astrocytes were isolated from brains of RNase H2ΔGFAP mice and cultured under mitogenic conditions, they showed signs of DNA damage and premature senescence. Enhanced expression of interferon-stimulated genes (ISGs) represents the most reliable AGS biomarker. Importantly, primary RNase H2ΔGFAP astrocytes displayed significantly increased ISG transcript levels, which we failed to detect in in vivo in brains of RNase H2ΔGFAP mice. Isolated astrocytes primed by DNA damage, including RNase H2-deficiency, exhibited a heightened innate immune response when exposed to bacterial or viral antigens. Taken together, we established a valid cellular AGS model that utilizes the very cell type responsible for disease pathology, the astrocyte, and phenocopies major molecular defects observed in AGS patient cells.

In the affected neonates, clinical signs of AGS mimic the sequelae of congenital virus infections. AGS is caused by mutations in the genes TREX1, RNASEH2, SAMHD1, ADAR1, or IFIH1, with RNase H2 mutations accounting for >50% of all cases (2). RNase H2 is a trimeric enzyme consisting of the subunits A (catalytic), B, and C, whereby all three subunits can be mutated, although individually, in AGS patients. With the exception of IFIH1, which encodes the cytosolic RNA sensor MDA5, all proteins associated with AGS are involved in nucleic acid metabolism. It has therefore been proposed that in AGS endogenous nucleic acids accumulate, thereby stimulating the innate immune system to produce increased amounts of the antiviral cytokine interferon α (3). While elevated interferon α levels are found in AGS patients in the cerebrospinal fluid and to a lesser extent also in serum, mRNA expression of interferon-stimulated genes (ISGs) in peripheral blood has turned out to be the most reliable AGS biomarker to date (4). This so-called interferon signature is also a hallmark of systemic lupus erythematosus (SLE), which shares many clinical features with AGS (3,5). Post-mortem analysis of diseased brains has established astrocytes as the cellular source of interferon α in AGS (6). In line with this, transgenic mice that constitutively express interferon α in astrocytes develop neurological symptoms similar to AGS, suggesting that AGS is driven by chronic cerebral presence of interferon α (7). A striking feature of all previous AGS mouse models is that none seems to recapitulate the strong CNS involvement associated with the human disease. TREX1 knockout (KO) mice for instance exhibit multi-organ autoimmunity and eventually succumb to sterile myocarditis, which is accompanied by strong stimulation of the type I interferon axis mainly in the heart but also in other tissues. The brain, however, is not affected in TREX1 KO mice (8,9). Both ADAR1 and RNase H2 KO mice die during embryogenesis, with only ADAR1 KO embryos displaying marked type I interferon upregulation (10)(11)(12). SAMHD1deficient mice on the other hand have no obvious phenotype, but exhibit a weak interferon signature in selected organs (13,14). Mice bearing a gain-of-function IFIH1 mutation show lupus-like systemic autoimmunity, again without CNS involvement (15). Two hypomorphic RNase H2 mouse models have been generated recently and depending on the introduced patient mutation, the animals die perinatally (G37S mutation in the catalytic A subunit) or display no overt phenotype (A177T mutation in the accessory B subunit) (16,17). Although both mouse models exhibit ISG upregulation in some tissues, no signs of neuroinflammation were apparent. The differences in disease presentation in terms of CNS involvement between AGS patients and the corresponding mouse models are striking, yet the underlying reasons are unclear. Given the absence of a CNS-related phenotype in hypomorphic RNase H2 mice and the early embryonic lethality of RNase H2 null mice, we generated conditional knockout mice devoid of RNase H2 in large parts of the brain (RNase H2 ΔGFAP mice) in an attempt to establish an RNase H2-based mouse model of AGS. In contrast to the severe phenotype associated with human disease, but in line with other AGS models, mice lacking RNase H2 in astrocytes and neurons displayed no clinical signs of disease. However, astrocytes isolated from RNase H2 ΔGFAP mice exhibited cellular defects reminiscent of AGS. We believe that the RNase H2-deficient astrocytes generated in this study can serve as a valid cellular model for AGS, especially as they represent the very cell type considered to govern the pathogenic interferon α expression in human patients.

Mice
RNase H2Bflox mice, the generation of which was described in Ref. (11), were kindly provided by Axel Roers (Institute of Immunology, Medical Faculty Carl Gustav Carus, University of Technology Dresden, Dresden, Germany). Floxed RNase H2B mice were crossed to Cre-hGFAP mice (18) for gene inactivation in astrocytes as well as cortical and hippocampal neurons and resulting RNase H2B flox/flox ; hGFAP-Cre mice were termed RNase H2 ΔGFAP mice. For RNase H2 ablation in neurons of the neocortex and hippocampus floxed RNase H2B mice were intercrossed with Cre-Emx1 mice (19) and RNase H2B flox/flox ; Emx1-Cre mice were termed RNase H2 ΔEmx1 mice. RNase H2B flox/flox littermates w/o Cre served as control animals. All mice were provided with food and water ad libitum and maintained in a 12-h light-dark cycle under standard conditions at Kiel University. Genotyping PCR was performed on processed tail clips (Direct PCR Lysis Reagent Tail, Peqlab), whereby respective PCR protocols as well as primer sequences were obtained from www.jax.org. Procedures involving animal care were conducted in agreement with national and international laws and policies with appropriate permission. All experiments were carried out in accordance with the guidelines for Animal Care of the Christian-Albrechts-University of Kiel.

generation and Treatment of Primary astrocytes
For astrocyte isolation, P3/4 pups were decapitated, their brains taken out, and meninges carefully removed under a dissecting microscope. Cerebral hemispheres were minced with a scalpel and digested in HBSS complemented with 2.5% trypsin (v/v) and 1% DNase I (m/v) for 30 min at 37°C (water bath). During digestion, the suspension was disintegrated every 10 min using a 1 ml pipet. Afterwards the cell suspension was filtrated through a 40 µm cell strainer and the filtrate topped up with two volumes minimum essential medium (MEM). Following centrifugation for 3 min at 1,300 g, cells were resuspended in glia medium 1 (MEM complemented with 10% horse serum, 0.6% glucose, and 1% penicillin/ streptomycin) and dispensed into three 6-wells coated with 0.1 mg/ ml poly-l-lysine. After 48 h, culture medium was changed to glia medium 2 (Dulbecco's modified Eagle's Medium, high Glucose (4.5 g/l), with GlutaMAX™, supplemented with 10% FCS, N2 Supplement, 1 μg/50 ml epidermal growth factor, and 1% penicillin/streptomycin) in which cells were cultured at 37°C, 5% CO2 and 3% O2 until experimental use. Upon confluency, control and RNase H2 ΔGFAP astrocytes were splitted 1:6 and 1:2, respectively, using Accutase and Trypsin/EDTA (2:3, Sigma Aldrich). A day prior to experiments microglia were removed from cell suspensions by multiple thorough PBS washing steps. The primary astrocyte culture consistently contained >90% GFAP-positive cells. Astrocytes were treated with Doxorubicin (1 µM for 18 h) and LPS (E. coli 0111:B4, 500 ng/ml, 6 h, both Sigma Aldrich) or transfected with polyI:C (5 µg, 6 h, Invivogen). To drive astrocytes into oncogeneinduced senescence (OIS), cells were retrovirally transduced with H-Ras G12V 4 days post isolation, as described in Ref. (20). Successful transfectants were enriched by 500 µg/ml puromycin (Sigma Aldrich) treatment for another 6 days and senescent astrocytes were harvested and analyzed 10 days post isolation.

immunofluorescence and immunohistochemistry
For immunofluorescence analysis, primary astrocytes were fixed with ice-cold acetone/methanol (1:2) for 10 min and blocked with 1% (w/v) BSA in PBS overnight at 4°C. Cells were incubated with primary antibodies for 4 h and secondary antibodies for 2 h in a humidified chamber and subsequently embedded in Immu-Mount (Thermo Fisher Scientific) complemented with 5 µg/ml 4′,6-diamino-2-phenylindole (DAPI, Sigma Aldrich). Antihis tone γH2AX (Cell Signaling, #2577) and rabbit anti-mouse RNase H2 antiserum [kindly provided by Andrew P. Jackson, MRC Human Genetics Unit, Institute of Genetics and Molecular Medicine, University of Edinburgh, UK, described in Ref. (12)] were used as primary antibodies, while anti-rabbit Cy3 (GE Healthcare) served as secondary Ab. For immunohistochemistry, mice were perfused transcardially via the left ventricle with PBS followed by 4% PFA (in PBS). Brains were removed and post-fixed additional 4 h by immersion at room temperature. Subsequently, brains were transferred to 30% sucrose (in PBS) overnight. 35 µm thick free floating sections were prepared on a Leica 9,000 s microtome (Leica, Wetzlar, Germany). Sections for immunohistochemistry were blocked with 1.6% H2O2 solution and washed in PBS. Sections for immunohistochemistry and immunofluorescence were rinsed in PBS and subsequently blocked with 4% normal goat serum in PBS containing 0.2% bovine serum albumin and 0.25% Triton X-100 (blocking solution) at room temperature for 2 h. Primary antibodies (anti-GFAP, clone G-A-5, Sigma; Ki-67, SolA15, eBioscience; anti-Iba-1, GTX100042, Genetex; anti-NeuN, MAB377, Millipore) were diluted in blocking solution and sections were incubated overnight at 4°C with gentle agitation. After washing 3× with washing buffer (PBS containing 0.25% TX-100), sections were incubated for 2 h at RT with appropriate secondary antibodies [fluorophore conjugated; AlexaFluor 488 and 594 (Molecular Probes) or biotinylated second antibodies (Vector Laboratories)] diluted in washing buffer. Antibody dilutions are provided in Supplementary Table B. After additional three washes with washing buffer, sections were coverslipped with MOWIOL/DABCO (fluorophore-labeled sections) containing DAPI or incubated with Streptavidin-HRP (ABC Elite Kit, Vector Labo ratories) for another 1 h at room temperature followed by another three washes, development with 3,3′-Diaminobenzidin substrate solution, and coverslipped in MOWIOL. A FV1000D Laser Scanning Confocal Microscope (Olympus) equipped with FV10-ASW 4.2 Viewer software was used for image acquisition.

qPcr, Western Blotting, and Facs
Total RNA was isolated using the Gene Jet RNA Purification Kit (Thermo Fisher Scientific), while cDNA was generated with Revert Aid Reverse Transcriptase (Thermo Fisher Scientific) according to the manufacturer's instructions, except a mixture of oligo(dT)18 and random hexamer primers (1:2, Thermo Fisher Scientific) was used. The qPCR reaction contained SYBR Green qPCR Master Mix (Thermo Fisher Scientific), 0.5 µM target specific primer pairs ( Table  A in Supplementary Material), and 10 ng cDNA template. SYBR Green incorporation was analyzed on a Lightcycler 480 II (Roche Applied Science) according to the following program: initial denaturation at 95°C for 10 min, 45 cycles of denaturation (10 s at 95°C), annealing (20 s at 60°C), and elongation (20 s at 72°C). Relative mRNA expression was calculated based on the ΔΔCp method with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) serving as house keeping control. For flow cytometry, detached astrocytes were fixed in 4% PFA (in PBS) for 20 min, saponin-permeabilized for 15 min (using saponin-based permeabilization and wash reagent, Invitrogen), and blocked in 1% BSA (in PBS) for 15 min Afterward, cells were incubated with FITC-coupled anti-histone γH2AX (Cell Signaling, #9719) for 1 h on ice and resulting fluorescence was detected using a FACSCanto flow cytometry system (BD biosciences). Western Blot analysis was performed as described in Ref. (21) and the following primary antibodies were used: anti-LAMP1 (DSHB, 1D4Bc), anti-RNase H2A (Origene, TA306706), anti-H-Ras (Santa Cruz, sc-29), and anti-PARP1 (Cell Signaling, #9542). To ensure equal loading amounts, membranes were stripped and re-probed with anti β-actin (Cell Signaling Technology, #4967). Antibody dilutions are provided in Supplementary Table B.

neurological and Behavioral analyses
The neurological as well as behavioral tests were performed at the German Mouse Clinic (22,23). The test pipeline started with 39 mice (12 male mutants, 7 male controls, 10 female mutants, and 10 female controls). Mice were maintained in IVC cages with water and standard mouse chow (Altromin no. 1314) according to the GMC housing conditions and German laws (for details, check https://www.mouseclinic.de/about-gmc/mouse-husbandry/ index.html). All tests performed at the GMC were approved by the responsible authority of the Regierung von Oberbayern (animal license no. 46-16). Welfare-related monitoring of experimental animals was performed according to the animal license. Sample size of the experiments was calculated and documented in the approval 46-16.

Open Field
The Open Field analysis was carried out as described previously (24,25). It consisted of a transparent and infra-red light permeable acrylic test arena with a smooth floor (internal measurements: 45.5 cm × 45.5 cm × 39.5 cm). Illumination levels were set at approx. 150 lux in the corners and 200 lux in the middle of the test arena. Data were recorded and analyzed using the ActiMot system (TSE, Bad Homburg, Germany).

Prepulse Inhibition (PPI) of the Acoustic Startle Response
Animals were separated based on sex, but not genotype. PPI was assessed automatically using an automated startle apparatus setup (Med Associates Inc., VT, USA) including four identical sound-attenuating cubicles. The protocols were written using the Med Associates "Advanced Startle" software. Experiments were carried out between 08:30 h and 17:00 h. Background noise was 65 dB, and startle pulses were bursts of white noise (40 ms). A session was initiated with a 5-min-acclimation period followed by five presentations of leader startle pulses (110 dB) that were excluded from statistical analysis. Trial types for the PPI included four different prepulse intensities (67, 69, 73, 81 dB); each prepulse preceded the startle pulse (110 dB) by a 50 ms inter-stimulus interval. Each trial type was presented 10 times in random order, organized in 10 blocks, each trial type occurring once per block. Inter-trial intervals varied from 20 to 30 s. This protocol is based on the protocol used in IMPRESS from the International Mouse Phenotyping Consortium (IMPC, see www. mousephenotype.org/impress), adapted to the specifications of our startle equipment.

Y Maze
Spontaneous alternations were assessed using the Y-Maze, which was made of opaque light gray PVC and had three identical arms (30 cm × 5 cm × 15 cm) placed at 120° from each other; illumination in the center of the maze was 100 lux. Each mouse was placed at the end of one arm and allowed to move freely through the maze during a 5-min session. Spontaneous alternations (defined as consecutive entries into all three arms without repetitions in overlapping triplet sets) were scored. Total numbers of arm entries were collected cumulatively over the 5 min. Spontaneous alternation performance percentage is defined as the ratio of actual (total alternations) to possible alternations (total number of triplets) × 100. When placed in the Y-Maze, normal mice prefer to explore the least recently visited arm, and thus tend to alternate visits between the three arms. To explore the three arms successively the mouse must maintain an ongoing record of the most recently visited arms, and continuously update such records. Therefore, alternation behavior is a measure of spatial working memory.

Social Discrimination
The Social Discrimination procedure consisted of two 4-min exposures of stimulus animals (ovariectomized 129Sv females) to the test animal in a fresh cage to which the test animal had been moved 2 h prior to testing. All stimulus animals are identified using colored non-toxic non-permanent paint markers on the tail. After a retention interval of 2 h, this stimulus animal was re-exposed to the test animal together with an additional, previously not presented stimulus animal. A separate "familiar" and "unfamiliar" stimulus animal was assigned to each test animal. The duration of investigatory behavior of the test animal toward the stimulus animals (familiar and unfamiliar) during this test phase was again recorded by a trained observer with a hand-held computer. A social recognition index was calculated as time spent investigating the unfamiliar stimulus mouse/time spent investigating both the familiar and unfamiliar stimulus mouse.

Beam Walk
Motor skills were tested on beams with different diameters (20 and 12 mm square and 22 and 15 mm round). The traversing time and numbers of falls, foot slips, and stops were recorded. The mouse performed three trials each consecutively, and the average time of these three trials was calculated.

Beam Ladder
The beam ladder consists of two Plexiglas screens connected with several metal beams of variable distance. The test is used to evaluate skilled walking of the mice. Mice traverse the ladder and foot slips of fore paws and hind paws are counted separately as well as traversing time.

experimental autoimmune encephalomyelitis (eae)
Mice were immunized subcutaneously at the tail basis with 50 µg MOG35-55 peptide (Gene Script) emulsified in Complete Freund's adjuvant (Difco) supplemented with 1.1 mg of heat-inactivated Mycobacterium tuberculosis (Difco). At the day of immunization and again 2 days later, mice received an intraperitoneal injection of 200 ng pertussis toxin (List Biological Laboratories). Clinical signs of disease were assessed daily and graded on a scale from 0 to 6 as follows: 0, no disease; 0.5, partial loss of tail tonicity; 1, complete loss of tail tonicity; 1.5, partially impaired righting reflex on attempt to roll over (within 3 s); 2, impaired righting reflex; 2.5 partial hind limb paresis resulted in staggering gait; 3, complete hind limb paresis; 3.5, complete hind limb paralysis; 4, unilateral forelimb paresis; 4.5, complete forelimb paresis; 5, moribund, and 6, dead animal. Experiments were performed in accordance with the guidelines of the central animal facility institution (TARC, University of Mainz) and approved by the local authorities under the license number G 13-1-098.
statistics All data are shown as mean ± SEM. Statistical analysis was performed with GraphPad Prism 5 using two-tailed Student's t-test unless indicated otherwise in the figure legends. For all tests, p-values less than 0.05 were considered statistically significant. The phenotyping screen at the GMC was mainly designed as a high throughput screen for new phenotypic alterations, a correction for multiple testing of the various parameters was not performed. If not stated otherwise, data were analyzed using R software (Version 3.0.2; Foundation of Statistical Computing, Vienna, Austria). resUlTs generation of conditional Knockout Mice lacking rnase h2 in the Brain As information regarding RNase H2 expression in the mammalian brain is scarce, we first tracked RNase H2 protein expression in the mouse brain using a rabbit antiserum raised against the RNase H2 holoenzyme. RNase H2 was expressed throughout the entire murine brain with highest expression in the hindbrain, the cerebellum, the hippocampus, and the cerebral cortex ( Figure 1A). Notably, particularly high expression was observed in the subgranular zone (SGZ) of the hippocampal dentate gyrus (dg), a region typically associated with adult hippocampal neurogenesis and proliferation. To further narrow down RNase H2 expression, we applied double-immunofluorescence staining of RNase H2 with NeuN, a neuronal marker, and the astrocytic marker GFAP (Figure 1B). RNase H2 extensively co-localized with NeuN in both the cerebral cortex and the dg ( Figure 1B).
Remarkably, those cells with the highest expression in the SGZ were negative for GFAP and NeuN, a marker for more mature neurons, indicating that these cells were immature neuronal precursor cells (Figure 1B). Accordingly, these abundantly RNase H2 expressing cells were positive for Ki67, a marker for dividing cells (Figure 1C). Immunohistochemistry of diseased brain sections revealed that astrocytes secrete the pathogenic interferon α in AGS patients (6), and transgenic mice overexpressing interferon α from the supposedly astrocyte-specific GFAP-promoter exhibit clinical signs reminiscent of AGS (7). To establish a mouse model of AGS, we conditionally deleted RNase H2 in the mouse brain by intercrossing RNase H2B flox mice with GFAP-Cre transgenic mice (Figure 2A). The resulting animals carrying homozygous RNase H2B deletions were designated RNase H2 ΔGFAP mice. RNase H2 ΔGFAP mice were viable, fertile, and exhibited no obvious phenotype. Offspring was generated in the expected Mendelian ratio ( Figure 2B). Western Blot and immunofluorescence analysis revealed robust expression of RNase H2 in wild-type astrocytes, while astrocytes isolated from RNase H2 ΔGFAP mice completely lacked the nuclease (Figures 2C,D). Importantly, genetic ablation of RNase H2B has shown to be sufficient to disrupt expression of the entire holoenzyme (12). Absence of RNase H2 was confirmed in vivo by examining different brain regions using immunohistochemistry. RNase H2 was absent from the majority of cortical and hippocampal neurons as well as astrocytes in RNase H2 ΔGFAP mice ( Figure 2F). In keeping with this, western blot analysis of whole brain lysates revealed a marked downregulation of RNase H2 protein expression ( Figure 2E). Although leakiness of the   recombinase is active predominantly in neurons of the neocortex and hippocampus. Ablation of neuronal RNase H2 was effective in RNase H2H2 ΔEmxl mice and largely mirrored the deletion pattern seen in RNase H2 ΔGFAP mice ( Figure S1A in Supplementary Material). However, because of the strong clinical link between astrocytes and AGS pathology, we focused in the present work on RNase H2 ΔGFAP animals.

rnase h2 ΔgFaP Mice Display no ags-like Phenotype
Aicardi-Goutières syndrome is characterized by fundamental morphological changes of the brain, including microcephaly, cerebral calcifications, and white matter abnormalities such as atrophy (1). However, baseline H&E histology revealed no apparent differences between RNase H2 ΔGFAP and control brains ( Figure S2A in Supplementary Material). Chronic neuroinflammation associated with activation of the type I interferon axis is a defining hallmark of AGS (2). Brains of 3 months old RNase H2 ΔGFAP mice did not show signs of astrogliosis or microgliosis compared to control littermates, as assessed by GFAP and Iba-1 immunostaining, respectively (Figures 3A,B). Importantly, qPCR analysis revealed similar expression of selected ISGs in RNase H2 ΔGFAP and control brains (Figure 3C). We additionally performed detailed neurological and behavioral tests with RNase H2 ΔGFAP animals (summarized in Figure S2B in Supplementary Material). However, behavioral and neurological examination of RNase H2 ΔGFAP mice only sporadically revealed subtle sexspecific effects of unclear relevance, which did not resemble the severe disabilities observed in AGS (summarized in Figure S2B, detailed in Figures S3A-F (28). Given the established link between DNA damage and type I interferon expression, it has been hypothesized that in fact DNA damage might represent the prime trigger of the pathogenic type I interferon production in AGS (12,28,29). We sought to investigate whether RNase H2 ΔGFAP astrocytes would recapitulate some of the features of AGS patient cells. We, therefore isolated astrocytes from neonatal RNase H2 ΔGFAP and control brains, with primary cell cultures consistently containing >90% GFAP-positive murine astrocytes (a representative astrocyte preparation is shown in Figure S4D in Supplementary Material). As shown above, cultured astrocytes from RNase H2 ΔGFAP mice were devoid of RNase H2 (Figures 2C,D). In contrast to RNase H2 ΔGFAP brains, which failed to show significant ISG upregulation (Figure 3C), we observed a robust interferon signature in cultured RNase H2 ΔGFAP astrocytes ( Figure 4A). Importantly, assessment of cerebral ISG expression was carried out at the same developmental stage as RNase H2 ΔGFAP astrocyte isolation (P3/4 pups). Using flow cytometry and immunofluorescence, we monitored phosphorylation of histone H2AX, a common marker for DNA doublestrand breaks, and detected extensive DNA damage in RNase H2 ΔGFAP astrocytes (Figures 4B,C). In line with increased DNA damage, RNase H2 ΔGFAP astrocytes showed a marked reduction in their proliferative capacity, as assessed by EdU incorporation ( Figure 4D). Western blot analysis revealed that RNase H2 ΔGFAP astrocytes did not exhibit caspase-mediated PARP-1 cleavage, indicating that RNase H2-deficient astrocytes did not undergo apoptosis ( Figure 4E). In contrast, RNase H2 ΔGFAP astrocytes developed an enlarged and flattened morphology, which is consistent with cellular senescence (Figure 5A). In keeping with this, transcript levels of several senescence-associated genes were upregulated in RNase H2 ΔGFAP astrocytes (Figure 5B). A hallmark of senescent cells is the enlargement of their lysosomal compartment, which is typically assessed by measuring activity of acidic β-galactosidase (30). However, high background levels of acidic β-galactosidase already in control astrocytes precluded a detailed analysis of this marker (data not shown). Instead, we observed increased mRNA and protein levels of the lysosomal proteins LAMP1 and Cathepsin A in RNase H2 ΔGFAP astrocytes, which likewise suggested elevated lysosome biogenesis in the absence of RNase H2 (Figures 5C,D). Importantly, control astrocytes that had been driven into oncogene-induced senescence (OIS) by retroviral transduction with oncogenic H-Ras showed comparible upregulation of senescence-associated genes (Figures 5E-G).
It has been hypothesized that AGS might require an environmental trigger for manifestation (see above). To test this hypothesis at the cellular level, we challenged isolated astrocytes "primed by DNA damage" with bacterial (LPS) and viral (poly I:C) antigens. DNA damage was inflicted either by treatment with the chemotherapeutic drug doxorubicin or RNase H2 loss (in RNase H2 ΔGFAP astrocytes). Importantly, when isolated astrocytes were primed by DNA damage prior to LPS or poly I:C treatment, enhanced activation of the type I interferon axis was observed (Figures 5H,I). Taken together, cultured murine astrocytes lacking RNase H2 exhibited increased ISG expression, extensive DNA damage, and signs of cellular senescence, all of which has previously been reported for mutant RNase H2 patient cells (28). Moreover, RNase H2-deficient astrocytes adapted a primed state that further amplified the innate immune response elicited by bacterial or viral antigens.

DiscUssiOn
Aicardi-Goutières syndrome is a severe autoimmune childhood disorder resulting in profound physical and intellectual disabilities. AGS is of genetic origin and mainly affects the brain, although extraneurological symptoms such as chilblain lesions of the skin are frequent (2). The disease trigger is most likely a chronic overactivation of the type I interferon system, which is supported by the AGS-like phenotype of transgenic mice overexpressing interferon α in the CNS (7). All genes associated with AGS are involved either in nucleic acid metabolism or detection, with mutations in the endoribonuclease RNase H2 accounting for >50% of cases (2). In addition, gene variants in the AGS genes TREX1 and RNASEH2 have also been associated with SLE, underlining the considerable phenotypic overlap between the two disease entities (28,31). RNase H2 has two distinct enzymatic activities. While it can cleave the RNA moiety of an RNA/ DNA hybrid (containing >4 consecutive ribonucleotides), it is also capable of initiating the excision of single ribonucleotides embedded in a DNA duplex (32). The latter enzyme activity appears to be the main physiological function of RNase H2, as mouse embryos deficient for RNase H2 strongly accumulate genome-embedded ribonucleotides leading to widespread DNA damage (12). In this context, replicative DNA polymerases have been shown to be much more promiscuous as previously anticipated, resulting in the establishment of misincorporated ribonucleotides as the most frequent naturally occurring DNA lesions (12,33). In line with this, RNase H2 expression is particularly strong in tissues with a high proliferative turnover, such as gut, skin, testis, and developing brain (12). Given this strong link between RNase H2 expression and proliferation, we were surprised by the high abundance of RNase H2 in postmitotic neurons. A possible explanation for the unexpectedly strong expression in postmitotic neurons could be a function of RNase H2 independent of DNA replication such as R-loop resolution during transcription (34) or RNA/DNA hybrid degradation during double strand break repair (35). However, this potential alternative function of RNase H2 in the brain is apparently not crucial for neuronal homeostasis, as neuronal RNase H2 ablation did not result in an obvious phenotype (this study). RNase H2 knockout mice die in utero as a consequence of a p53-mediated growth arrest starting around gastrulation (11,12). As due to their early embryonic lethality RNase H2 null mice preclude detailed analysis of AGS pathology, we generated conditional knockout mice lacking RNase H2 in the brain, the most affected organ in AGS. To that end, we intercrossed floxed RNase H2B mice to animals expressing Cre under the human GFAP promoter, thereby targeting essentially the same cell types as in the seminal study on brain-specific interferon α overexpression (7). In RNase H2 ΔGFAP mice, RNase H2 ablation was not only achieved in astrocytes  Increased expression of senescence-associated genes in RNase H2 ΔGFAP astrocytes compared to control astrocytes (x = 1) 10 days post isolation. Transcript levels were quantified by qPCR and normalized to GAPDH expression. Error bars are SEM, *p < 0.05, **p < 0.01, t-test (n = 3). (c,D) Elevated transcript (Lamp1, Ctsa) and protein (Lamp1) levels of lysosomal genes suggested enlargement of the lysosomal compartment in RNase H2 ΔGFAP astrocytes, a hallmark of cellular senescence. Expression was assessed by qPCR (c) or Western Blot (D), β-actin as loading control. Error bars are SEM, *p < 0.05, **p < 0.01, t-test (n = 3). (e) Control astrocytes were transduced with a retrovirus encoding H-Ras G12V and successful transduction was confirmed by Western Blot against H-Ras. 6 days after transduction with oncogenic H-Ras, astrocytes showed mRNA upregulation of senescence-associated (F) as well as lysosomal genes (g), as measured by qPCR. Error bars are SEM, *p < 0.05, ***p < 0.001, t-test (n = 3). (h,i) Isolated murine astrocytes were primed by DNA damage and then challenged with bacterial (LPS, left) or viral (polyI:C, right) antigens. DNA damage was provoked either by treatment with Doxorubicin or RNase H2-deficiency (H2 ΔGFAP ). Activation of the innate immune system was assessed by IFIT1 (h) and IFNβ (i) qPCR. Increase in mRNA expression is indicated relative to unchallenged control astrocytes (x = 1). Astrocytes were treated 7 days post isolation. Error bars represent SEM, *p < 0.05, **p < 0.01, ***p < 0.001, t-test (n ≥ 3).

11
Bartsch et al.  (2) the severe consequences of only hypomorphic RNase H2 mutations in AGS patients. This discrepancy between mouse model and human disease does not only hold true for RNase H2, but also for other AGS-associated genes. In a nutshell, all AGS mouse models described so far fail to recapitulate the strong CNS involvement of the human disease. The reason for this striking difference in disease course between the two species is not known at present and can only be speculated on. At least for RNase H2, however, our study might offer a possible explanation for this apparent conundrum. Whereas RNase H2 ΔGFAP mice did not exhibit AGS-related disease signs in vivo, RNase H2 ΔGFAP astrocytes isolated from the same animals showed cellular defects akin to AGS patient cells, including DNA damage, premature senescence, and increased ISG expression. In this context, it is important to note that RNase H2 ΔGFAP astrocytes were cultured under mitogenic conditions, which is in stark contrast to the largely postmitotic nature of resident neurons and astrocytes in RNase H2 ΔGFAP brains. It is therefore plausible that active proliferation is a prerequisite for the development of cellular features reminiscent of AGS. In support of this, recent reports have shed light on the role of micronuclei in activating the innate immune system upon DNA damage, which in two studies also included DNA damage caused by RNase H2 loss (36)(37)(38). Importantly, generation of these immunogenic micronuclei was strongly dependent on cellular proliferation (36)(37)(38). This reliance on proliferation could explain why humans with only hypomorphic RNase H2 mutations develop AGS, whereas mice carrying the same patient mutations (16,17) or even completely lack RNase H2 in large parts of the brain (this study) do not show AGS-like features: Due to the larger size of the human brain neurogenesis is faster and involves more cell divisions than in the murine brain (39). It is, therefore conceivable that the degree of neuronal proliferation is just not sufficient for the establishment of disease symptoms in the much smaller murine brain. However, further experiments are clearly needed to test this hypothesis. In aggregate, primary RNase H2 ΔGFAP astrocytes can serve as a valid cellular model of AGS, since they (1) display similar defects as RNase H2 mutant AGS patient cells and (2) constitute the very cell type responsible for pathogenic interferon α production in AGS individuals. In addition, DNA damage renders RNase H2-deficient astrocytes more sensitive toward foreign antigens, which could help explain the phenotypic variability among AGS patients bearing the same mutation and also why AGS sometimes seems to require an exogenous trigger in order to manifest.

eThics sTaTeMenT
All statements regarding the mouse work have been included in the Section "Materials and Methods" of the manuscript.