ADAR1 averts fatal type I interferon induction by ZBP1

Mutations of the ADAR1 gene encoding an RNA deaminase cause severe diseases associated with chronic activation of type I interferon (IFN) responses, including Aicardi–Goutières syndrome and bilateral striatal necrosis1–3. The IFN-inducible p150 isoform of ADAR1 contains a Zα domain that recognizes RNA with an alternative left-handed double-helix structure, termed Z-RNA4,5. Hemizygous ADAR1 mutations in the Zα domain cause type I IFN-mediated pathologies in humans2,3 and mice6–8; however, it remains unclear how the interaction of ADAR1 with Z-RNA prevents IFN activation. Here we show that Z-DNA-binding protein 1 (ZBP1), the only other protein in mammals known to harbour Zα domains9, promotes type I IFN activation and fatal pathology in mice with impaired ADAR1 function. ZBP1 deficiency or mutation of its Zα domains reduced the expression of IFN-stimulated genes and largely prevented early postnatal lethality in mice with hemizygous expression of ADAR1 with mutated Zα domain (Adar1mZα/– mice). Adar1mZα/– mice showed upregulation and impaired editing of endogenous retroelement-derived complementary RNA reads, which represent a likely source of Z-RNAs activating ZBP1. Notably, ZBP1 promoted IFN activation and severe pathology in Adar1mZα/– mice in a manner independent of RIPK1, RIPK3, MLKL-mediated necroptosis and caspase-8-dependent apoptosis, suggesting a novel mechanism of action. Thus, ADAR1 prevents endogenous Z-RNA-dependent activation of pathogenic type I IFN responses by ZBP1, suggesting that ZBP1 could contribute to type I interferonopathies caused by ADAR1 mutations.

Mutations of the ADAR1 gene encoding an RNA deaminase cause severe diseases associated with chronic activation of type I interferon (IFN) responses, including Aicardi-Goutières syndrome and bilateral striatal necrosis [1][2][3] . The IFN-inducible p150 isoform of ADAR1 contains a Zα domain that recognizes RNA with an alternative left-handed double-helix structure, termed Z-RNA 4,5 . Hemizygous ADAR1 mutations in the Zα domain cause type I IFN-mediated pathologies in humans 2,3 and mice [6][7][8] ; however, it remains unclear how the interaction of ADAR1 with Z-RNA prevents IFN activation. Here we show that Z-DNA-binding protein 1 (ZBP1), the only other protein in mammals known to harbour Zα domains 9 , promotes type I IFN activation and fatal pathology in mice with impaired ADAR1 function. ZBP1 deficiency or mutation of its Zα domains reduced the expression of IFN-stimulated genes and largely prevented early postnatal lethality in mice with hemizygous expression of ADAR1 with mutated Zα domain (Adar1 mZα/mice). Adar1 mZα/mice showed upregulation and impaired editing of endogenous retroelement-derived complementary RNA reads, which represent a likely source of Z-RNAs activating ZBP1. Notably, ZBP1 promoted IFN activation and severe pathology in Adar1 mZα/mice in a manner independent of RIPK1, RIPK3, MLKL-mediated necroptosis and caspase-8-dependent apoptosis, suggesting a novel mechanism of action. Thus, ADAR1 prevents endogenous Z-RNA-dependent activation of pathogenic type I IFN responses by ZBP1, suggesting that ZBP1 could contribute to type I interferonopathies caused by ADAR1 mutations.
Z-DNA and Z-RNA are nucleic acids with an alternative left-handed double-helix structure and have poorly understood biological function [10][11][12][13] . These Z-form nucleic acids are recognized by specific protein domains, termed Zα domains, which bind Z-DNA and Z-RNA in a conformation-specific manner 5,9,14,15 . Two proteins are known to harbour Zα domains in mammals, namely adenosine deaminase acting on RNA 1 (ADAR1) and Z-DNA-binding protein 1 (ZBP1, also known as DAI or DLM-1) 5,9,14 . ADAR1 is produced in two isoforms, the constitutively expressed nuclear p110 and the interferon (IFN)-inducible cytosolic p150 that contains a Zα domain 1 . ADAR1 p150 edits self-RNA derived predominantly from endogenous retroelements (EREs) to prevent its recognition by the cytosolic RNA sensor melanoma differentiation-associated gene 5 (MDA5) and the activation of mitochondrial antiviral signalling (MAVS)-dependent pathogenic type I IFN responses [16][17][18][19] . ADAR1 mutations mapping to the Zα domain combined with alleles resulting in loss of ADAR1 or specifically its p150 isoform were shown to cause Aicardi-Goutières syndrome (AGS) and bilateral striatal necrosis (BSN) in human patients 2,3 and severe MDA5-MAVS-mediated type I IFN-dependent pathology in mice [6][7][8] , indicating that the interaction of ADAR1 with Z-RNA is required to prevent activation of pathogenic IFN responses. ZBP1 is an IFN-inducible protein that senses viral and endogenous Z-form nucleic acids via its Zα domains and triggers cell death to induce antiviral immunity, but also causes tissue damage and inflammation [20][21][22][23][24][25][26][27] . Previous studies have shown that ZBP1 causes cell death in vivo and in vitro by activating receptor-interacting protein kinase 3 (RIPK3) in a RIP homotypic interaction motif (RHIM)-dependent manner, which then phosphorylates mixed-lineage kinase-like (MLKL) to induce necroptosis and can also engage RIPK1 to trigger caspase-8-dependent apoptosis 20,21,23,25,27 . We reasoned that ZBP1 may functionally interact with ADAR1 to regulate cellular responses to Z-RNA and assessed its role in the activation of type I IFN-dependent pathology in mice with Adar1 mutations.

ADAR1 Zα domain inactivation induces IFN responses
To address the role of the ADAR1 Zα domain, we generated knock-in mice expressing ADAR1 with two substitutions disrupting its interaction with Z-RNA (N175D/Y179A) 28,29 (Extended Data Fig. 1a), hereafter referred to as Adar1 mZα/mZα mice. Adar1 mZα/mZα mice were born at the expected Mendelian ratio, were viable and fertile and did not develop apparent pathology at least until the age of 1 year (Extended Data Fig. 1b-d). However, RNA sequencing (RNA-seq) showed upregulation of 57 genes in lung tissues from 4-to 5-month-old Adar1 mZα/mZα mice, all of which were functionally linked to type I IFN responses, compared with Adar1 mZα/WT and wild-type C57BL/6N animals (Extended Data Fig. 1e and Supplementary Table 1). Quantitative PCR with reverse transcription (qRT-PCR) analysis confirmed upregulation of a selected set of IFN-stimulated genes (ISGs) in lung, spleen and liver tissue from Adar1 mZα/mZα mice compared with control littermates (Extended Data Fig. 1f). Therefore, disruption of the ADAR1 Zα domain caused elevated expression of ISGs in the absence of overt tissue pathology, in line with recent reports 6,7,30 . Mutations affecting the ADAR1 Zα domain were found to cause AGS and BSN when combined with alleles resulting in loss of ADAR1 p150 expression 3,31 . To model this condition, we generated Adar1 mZα/mice and found that they developed a severe phenotype characterized by reduced body weight and early postnatal lethality (Fig. 1a,b). Haematological analysis at postnatal day (P) 1 showed reduced numbers of red blood cells (RBCs) as well as diminished haemoglobin (HGB) and haematocrit (HCT) levels in Adar1 mZα/mice compared with Adar1 mZα/WT mice (Fig. 1c), in line with the important role of ADAR1 in erythropoiesis 32 . Histological examination showed altered architecture with increased numbers of epithelial cells immunostained for cleaved caspase-3 (CC3) in the small intestine and colon of Adar1 mZα/pups, whereas other organs including the liver, lung, heart, kidney and brain did not show prominent pathological features ( Fig. 1d and Extended Data Fig. 2a-d). RNA-seq analysis of lung, brain and spleen showed increased expression of several genes in Adar1 mZα/mice compared with Adar1 mZα/WT littermates, the majority of which were linked to type I IFN responses (Extended Data Fig. 3 and Supplementary Table 2). Comparison of RNA-seq data from lung, brain and spleen identified a set of 93 genes, all ISGs, that were consistently upregulated in all three tissues from Adar1 mZα/compared with Adar1 mZα/WT mice (Supplementary Table 2). A smaller number of genes were downregulated in Adar1 mZα/mice, particularly in the spleen, most of which were functionally linked to erythrocyte development, in line with the impaired erythropoiesis observed (Extended Data Fig. 3). Crossing to Mavs −/− mice (Extended Data Fig. 2g) rescued the lethal phenotype of Adar1 mZα/mice, as Adar1 mZα/-Mavs −/− animals appeared healthy, did not show upregulation of ISGs and reached adulthood without displaying signs of pathology at least up to the age of 15 weeks (Fig. 1a-e and Extended Data Fig. 2). Therefore, in agreement with recent reports [6][7][8] , hemizygous expression of ADAR1 with a mutated Zα domain induced a strong MDA5-MAVS-dependent type I IFN response, causing severe early postnatally lethal pathology in mice.

ZBP1-RIPK3 function in Adar1 −/− mice
On the basis of our findings in Adar1 mZα/mice, we reasoned that ZBP1 might also contribute to the severe pathology caused by complete ADAR1 deficiency. However, we did not observe any live Adar1 −/− Zbp1 −/− mice born from crosses of heterozygous animals, showing that ZBP1 deficiency was not sufficient to rescue the embryonic lethality of Adar1 −/− mice (Extended Data Fig. 6a). Adar1 −/− Mavs −/− and Adar1 −/− Mda5 −/− animals develop to term but die during the first postnatal days 19,33 , showing that MDA5-MAVS-independent signalling causes postnatal pathology in Adar1 −/− mice. We therefore assessed whether ZBP1 deficiency might synergize with MAVS knockout to rescue the phenotype of Adar1 −/− mice. Indeed, we found that about 40% of Adar1 −/− Zbp1 −/− Mavs −/− mice survived to adulthood, in contrast to Adar1 −/− Mavs −/− mice that died shortly after birth ( Fig. 2f and Extended Data Fig. 6b). The Adar1 −/− Zbp1 −/− Mavs −/− mice that survived to adulthood showed reduced body weight but appeared healthy at least until the age of 15 weeks (Extended Data Fig. 6b-d). Histological analysis of different organs did not identify apparent pathology, with only mild hyperplasia and small numbers of dying cells observed in the intestine of Adar1 −/− Zbp1 −/− Mavs −/− mice (Extended Data Fig. 6e). We reasoned that ZBP1 may cause MAVS-independent pathology in Adar1 −/− mice by inducing RIPK3-mediated cell death. Indeed, RIPK3 deficiency alone or in combination with heterozygous knockout of the caspase-8 adaptor protein  Fig. 6i). ZBP1 deficiency prevented the phosphorylation of MLKL and reduced the cleavage of caspase-8 in these cells (Extended Data Fig. 6i), suggesting that ZBP1 induces necroptosis and, to a lesser extent, apoptosis. In line with this, combined pharmacological inhibition of RIPK3 and caspases strongly reduced the IFNγ-and CHX-induced death of Adar1 −/− Mavs −/− primary MEFs (Extended Data Fig. 6j). Although it remains unclear to what extent the ZBP1-RIPK3-dependent cell death induced in cells treated with CHX relates to the in vivo role of ZBP1-RIPK3 signalling in mediating the pathology of Adar1 −/− Mavs −/− mice, we reason that CHX treatment might mimic the activation of protein kinase R (PKR) and resulting inhibition of protein translation in ADAR1-deficient cells 7 . Collectively, these results showed that ZBP1-RIPK3-dependent signalling promoted the MAVS-independent pathology of Adar1 −/− Mavs −/− mice.

ZBP1 is not involved in Trex1 −/− mice
ZBP1 is expressed at very low levels in most tissues under steady-state conditions, but its expression is strongly induced by IFNs. Loss of ADAR1 function could trigger ZBP1-dependent pathology by promoting IFN-inducible upregulation of ZBP1 expression and/or by increasing the abundance of a ZBP1 ligand. To assess whether ZBP1 upregulation functions broadly to drive IFN-dependent pathology, we used another mouse model of type I interferonopathy caused by deficiency in TREX1, a cytosolic 3ʹ-5ʹ DNA exonuclease found to be mutated in people with AGS, familial chilblain lupus (FCL) and systemic lupus erythematosus (SLE) 34 . TREX1 deficiency in mice triggers cytosolic DNA-induced cyclic GMP-AMP synthase (cGAS)-stimulator of interferon genes (STING)-dependent type I IFN responses, resulting in systemic inflammation primarily manifesting in severe myocarditis [35][36][37][38] . Heart tissues from Trex1 −/− mice showed profound upregulation of ZBP1; however, Trex1 −/− Zbp1 −/− mice were indistinguishable from Trex1 −/− and Trex1 −/− Zbp1 WT/− littermates in terms of survival, body weight, splenomegaly, heart inflammation and fibrosis (Extended Data Fig. 7a-f). Moreover, ZBP1 deficiency did not inhibit the upregulation of ISGs and inflammatory cytokines and chemokines in heart tissues of Trex1 −/− mice (Extended Data Fig. 7c,g,h). Therefore, although its expression was strongly induced in tissues of Trex1 −/− mice, ZBP1 did not contribute to the cytosolic DNA-induced IFN response and pathology caused by TREX1 deficiency, in contrast to its important role in promoting the phenotype caused by impaired ADAR1 function.

Impaired ERE RNA editing in Adar1 mZα/mice
The specific role of ZBP1 in the pathology caused by ADAR1 deficiency could be explained by the accumulation of Z-RNA ligands in the absence of ADAR1-dependent RNA editing. In support of this hypothesis, previous studies have shown that the Zα domain of ADAR1 is required for editing of RNAs derived from EREs and particularly short interspersed nuclear elements (SINEs) in mouse cells 6,8,30 , which we previously identified as possible double-stranded RNA (dsRNA) ligands activating ZBP1 (ref. 21 ). Analysis of spleen RNA-seq data showed that expression of ERE groups previously shown to have the highest number of complementary reads with the potential to generate dsRNA 21 Mavs -/-(n = 9)

Article
with Adar1 mZα/WT pups (Fig. 3a). Interestingly, ZBP1 deficiency inhibited the upregulation of these ERE transcripts in Adar1 mZα/-Zbp1 −/− mice (Fig. 3a), reminiscent of its effect in suppressing ISG expression. These findings suggest that upregulation of EREs depends on IFN-mediated induction of gene expression, probably because these EREs reside near or within ISGs. We then assessed whether adenosine-to-inosine (A-to-I) RNA editing was affected in Adar1 mZα/and Adar1 mZα/mZα compared with Adar1 mZα/WT mice. Analysis of RNA-seq data from the spleen and brain showed a considerable loss of edited sites in Adar1 mZα/compared with Adar1 mZα/WT mice, the majority of which were found in SINE-derived RNAs (Fig. 3b,c and Extended Data Fig. 8). Adar1 mZα/mZα mice did not show an overall change in editing compared with Adar1 mZα/WT mice ( Fig. 3b and Extended Data Fig. 8). Notably, ZBP1 deficiency did not rescue the impaired RNA editing in spleen and brain tissues of Adar1 mZα/mice ( Fig. 3b and Extended Data Fig. 8c), in contrast to its strong effect in normalizing the IFN response. Similarly, MAVS deficiency did not rescue the editing defect in Adar1 mZα/mice (Fig. 3b). Together, these results indicate that increased expression of ERE-derived transcripts together with overall diminished editing of repeat RNAs in Adar1 mZα/mice could lead to the accumulation of dsRNAs with the capacity to generate Z-RNA ligands activating ZBP1.

Zα domain-dependent role of ZBP1 in Adar1 mZα/mice
To address the functional role of endogenous Z-RNA sensing by ZBP1, we crossed Adar1 mZα/mice with Zbp1 mZα/mZα mice expressing ZBP1 with both its Zα domains mutated 21 . ZBP1 Zα domain mutation substantially, albeit partially, rescued the early postnatal lethality of Adar1 mZα/mice, with about 50% of Adar1 mZα/-Zbp1 mZα/mZα mice surviving to the age of at least 15 weeks (Fig. 3d). Moreover, ZBP1 Zα domain mutation considerably restored body weight but did not substantially improve the impaired erythropoiesis in Adar1 mZα/pups (Fig. 3e,f). Comparison of lung RNA-seq data from newborn pups showed that ZBP1 Zα domain mutation suppressed the ISG response of Adar1 mZα/mice similarly to     ,k). In a, c, g, j and k, n = 5.

Fig. 3 | Endogenous Z-RNA likely derived from EREs triggers ZBP1-dependent IFN responses in
ZBP1 deficiency, with Adar1 mZα/-Zbp1 mZα/mZα mice clustering together with Adar1 mZα/-Zbp1 −/− mice (Fig. 3g). Moreover, qRT-PCR analysis showed reduced expression of a set of ISGs in spleen, lung and liver tissues from Adar1 mZα/-Zbp1 mZα/mZα compared with Adar1 mZα/mice at P1 (Extended Data Fig. 9a). Adult Adar1 mZα/-Zbp1 mZα/mZα mice appeared healthy but had reduced body weight and RBC, HGB and HCT levels compared with their littermate controls (Fig. 3h,i and Extended Data Fig. 9b). Histological analysis of tissues from 15-week-old mice showed signs of pericentral sinusoidal dilatation in the liver and mild hyperplasia with small numbers of dying cells in the intestine (Extended Data Fig. 9c). Furthermore, comparison of lung RNA-seq data from 15-week-old mice showed that ZBP1 Zα domain mutation suppressed ISG expression in Adar1 mZα/mice to the same extent as ZBP1 deficiency (Fig. 3j). Together, these results showed that Zα domain-dependent sensing of endogenous ligands, presumably Z-RNA, activates ZBP1-dependent signalling, promoting IFN responses and causing the severe postnatally lethal phenotype of Adar1 mZα/mice. However, Zα domain mutation conferred somewhat less protection in terms of mouse survival and could not substantially improve the impaired erythropoiesis of Adar1 mZα/mice compared with ZBP1 deficiency (Fig. 3d,f,i), suggesting that ZBP1 also exerts Zα domain-independent functions, as shown previously in viral infection and inflammation models 21 . The finding that ZBP1 deficiency or disruption of its Zα domains suppressed upregulation of ISG expression argues that ZBP1 promotes the IFN response in Adar1 mZα/mice. However, the reduction in ISG expression could also be secondary to rescue of tissue pathology. Therefore, to investigate whether ZBP1 regulates the IFN response independently of tissue pathology, we assessed the effect of ZBP1 deficiency or Zα domain disruption in Adar1 mZα/mZα mice, which are healthy but have elevated ISG expression (Extended Data Fig. 1). Notably, ZBP1 deficiency or disruption of its Zα domains strongly, albeit incompletely, suppressed the expression of ISGs in Adar1 mZα/mZα mice ( Fig. 3k and Extended Data Fig. 9d). Therefore, Zα domain-dependent ZBP1 signalling promotes the IFN response induced by disruption of the ADAR1 Zα domain independently of tissue damage.

RIPK1-and RIPK3-independent role of ZBP1
We reasoned that ZBP1 might engage RIPK3-dependent signalling to cause the severe pathology of Adar1 mZα/mice, as was the case in Adar1 −/− Mavs −/− mice. Unexpectedly however, Adar1 mZα/-Ripk3 −/− pups displayed early postnatal lethality as well as reduced body weight and RBC, HGB and HCT values, showing that RIPK3 knockout did not mimic the effect of ZBP1 deficiency (Fig. 4a-c). Furthermore, RNA-seq and qRT-PCR gene expression analysis showed that, in contrast to ZBP1 deficiency, RIPK3 knockout did not suppress ISG expression in lung, spleen and liver tissues from Adar1 mZα/mice ( Fig. 4d and Extended Data Fig. 10). To further address the role of necroptosis, we generated Adar1 mZα/-Mlkl −/− mice and found that MLKL deficiency also did not prevent early lethality and upregulation of ISG expression in lung tissues from Adar1 mZα/mice ( Fig. 4a-c,e). Thus, ZBP1 caused the pathology in Adar1 mZα/mice independently of RIPK3-MLKL-dependent necroptosis. We then reasoned that FADD-caspase-8-mediated apoptosis could contribute to the ZBP1-dependent pathology. However,

Discussion
Our results showed that ZBP1 promoted type I IFN responses and the associated pathology in Adar1 mZα/mice independently of RIPK1mediated signalling, RIPK3-MLKL-dependent necroptosis and FADD-caspase-8-mediated apoptosis. These findings are in contrast to the function of ZBP1 in Adar1 −/− Mavs −/− mice, where it causes early postnatal lethality by inducing RIPK3-dependent signalling (Fig. 2f). Therefore, ZBP1 has a dual role in mice with impaired ADAR1 function. On the one hand, it acts in a RIPK3-dependent manner to cause MAVS-independent pathology in Adar1 −/− mice, probably by inducing necroptosis. On the other hand, it acts in a RIPK1-and RIPK3-independent manner to promote a MAVS-dependent pathogenic type I IFN response, causing early postnatal lethality in Adar1 mZα/mice. MAVS deficiency nearly completely normalized whereas ZBP1 deficiency partially rescued ISG expression and the pathology of Adar1 mZα/mice, suggesting that ZBP1 is induced downstream of MAVS and contributes to type I IFN activation and the associated pathologies. The mechanisms by which ZBP1 promotes type I IFN activation in Adar1 mZα/mice remain elusive at present. TIR domain-containing adaptor-inducing interferon-β (TRIF), which induces IFN responses downstream of the Toll-like receptors TLR3 and TLR4 (ref. 41 ), also contains a RHIM and could be implicated in driving IFN activation downstream of ZBP1. ZBP1 was recently reported to contribute to TRIF-induced caspase-8 activation and interleukin (IL)-1β release 42 . However, this specific function was mediated via RHIM-dependent interaction with RIPK1 and therefore should be inhibited in a Ripk1 mR/mR Mlkl −/− genetic background, which did not rescue type I IFN activation in Adar1 mZα/mice, arguing against the involvement of this particular signalling pathway. It is also possible that RIPK1, RIPK3 and TRIF contribute to ZBP1-dependent IFN activation in Adar1 mZα/mice in a functionally redundant manner; if this is the case, inactivation of all three proteins may be required to mimic the effect of ZBP1 deficiency. Notably, in contrast to its previously suggested role as a cytosolic DNA sensor inducing IFN activation 43 , ZBP1 was not required for the cGAS-STING-dependent IFN response in Trex1 −/− mice, suggesting that it specifically functions to augment RNA-induced MDA5-MAVS-dependent IFN responses in Adar1 mZα/mice. Taken together, our results identified Zα domain-dependent cross-talk between ADAR1 and ZBP1 that critically controls IFN responses to endogenous RNA. Our findings identify suppression of endogenous Z-RNA formation by ADAR1 as a key mechanism preventing aberrant activation of pathogenic IFN responses by ZBP1. Although it remains technically challenging to directly assess Z-RNA formation in living cells and tissues, our results suggest that impaired ADAR1-dependent editing of RNAs primarily derived from SINEs causes the accumulation of Z-RNA ligands activating ZBP1. Strikingly, our genetic studies showed that ZBP1 induced pathogenic IFN responses in Adar1 mZα/mice in a manner independent of RIPK1, RIPK3-MLKL-dependent necroptosis and FADD-caspase-8-dependent apoptosis, suggesting a new mechanism of action. Collectively, while the specific downstream molecular mechanisms remain to be elucidated, our results identified ZBP1 as a key driver of pathogenic type I IFN responses triggered by impaired ADAR1 function and suggest that ZBP1-dependent signalling could contribute to the pathogenesis of type I interferonopathies caused by ADAR1 mutations in humans.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-04878-9.

Mice
Zbp1 −/− (ref. 21 ), Zbp1 mZα/mZα (ref. 21 ) and Ripk3 −/− (ref. 44 ) mice have been described previously. Adar1 −/− (Adar tm1b(EUCOMM)Wtsi ) mice were generated from the EUCOMM (https://academic.oup.com/bfg/ article/6/3/180/237263) line Adar tm1a(EUCOMM)Wtsi using CMV:Cre deleter mice 45 . Adar1 mZα mice, in which amino acids N175 and Y179 of the ADAR1 Zα domain were mutated to aspartic acid and alanine, respectively, were generated using CRISPR-Cas12a technology with all components purchased from Integrated DNA Technologies. Cas12a guide RNA (4 μM; 5′-CAGGGAGTACAAAATACGATTGA-3′; AsCas12a crRNA) targeting exon 2 of the Adar1 gene and 10 μM single-stranded DNA repair oligonucleotide (5′-A*G*G*TTTCCCCCTTCCTCTGTGC AGCTTTCCCTTCTTcTCCAGGGAagcCAAAATACGgTcGATGTCCCTTT TGGGGATTCTGAGCTCTCTGGCTAGCACATGGGCAG*T*G*G-3′; IDT, custom-made ultramer) with three phosphorothioate bonds at both ends (indicated by an asterisk) were co-electroporated essentially as described previously 46 with 4 μM AsCas12a protein and 4 μM DNA-based Cas12a electroporation enhancer into C57BL/6N zygotes. Correct exchange of the nucleotides, represented in the repair oligonucleotide with lower-case letters, was assessed by Sanger sequencing in the resulting F 0 mice. Trex1 −/− and Mavs −/− mice were generated using CRISPR-Cas9 technology. Of note, Trex1 is a single-exon gene. For Trex1 −/− mice, two sgRNAs (5′-TTCCAGGTCTAAGAAGATGA-3′ and 5′-CCTGGGCAGTAAGTCAAGAG-3′), each at 4 μM, in complex with 4 μM Cas9 protein (IDT) were co-electroporated into fertilized oocytes. Deletion of the Trex1 exon between the two sgRNAs was confirmed using primers 5′-ATCCCACTAGAACAACCCTGCC-3′ and 5′-TTCAGACTCCGCACCCTCATTT-3′ as well as by immunoblot analysis. For Mavs −/− mice, two sgRNAs (5′-CCGGTTCCCGATCTGCCTGT-3′ and 5′-ATACTGTGACCCCAGACAAG-3′) targeting exons 3 and 6, respectively, were co-injected into fertilized C57BL/6N oocytes at 50 ng μl -1 together with 100 ng μl -1 Cas9 mRNA (Trilink). Successful deletion of the critical exons was confirmed using PCR primers 5′-TTGATCC TCACACCGTACTTG-3′ and 5′-GTATTGTGTTGGCAGGTGCTT-3′. Mice used in this study were maintained in the animal facility of the CECAD Research Center, University of Cologne, in individually ventilated cages (Tecniplast, Greenline GM500) at 22 °C (±2 °C) and a relative humidity of 55% (±5%) under a 12-h light/12-h dark cycle on sterilized bedding (Aspen wood, Abedd) with access to a sterilized commercial pelleted diet (Ssniff Spezialdiäten) and acidified water ad libitum. The microbiological status of the mice was examined as recommended by the Federation of European Laboratory Animal Science Associations (FELASA), and the mice were free of all listed pathogens. All animal procedures were conducted in accordance with European, national and institutional guidelines, and protocols were approved by local government authorities (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen). Animals requiring medical attention were provided with appropriate care and were killed when they reached predetermined criteria of disease severity. No other exclusion criteria existed. Experimental groups were not randomized as mice were assigned to groups on the basis of genotype. Sample size was estimated on the basis of previous experience. Female and male mice of the indicated genotypes were assigned to groups at random. Mice were analysed at the age stated in the respective figure legends. Mouse studies as well as immunohistochemical assessment of pathology were performed in a blinded fashion. Whole blood samples of the mice were analysed using Abacus Junior Vet (Diatron).
Histological analysis of brains was conducted following standardized protocols at the Institute of Neuropathology, University Medical Center of Freiburg. Sections (4 μm thick) were processed and stained with H&E by standard laboratory procedures.

Histopathological analysis
Histopathological evaluation of intestinal tissues was performed on 3-μm-thick H&E-stained sections of paraffin-embedded Swiss rolls of intestinal tissues, using a modified version of a previously described scoring system 47 . In brief, histopathology scores were composed of four parameters: epithelial hyperplasia, quantity and localization of tissue inflammation, epithelial cell death and epithelial injury. An 'area factor' for the fraction of affected tissue was assigned and multiplied by the respective parameter score (1, 0-25%; 2, 25-50%; 3, 50-75%; 4, 75-100%). If different severities for the same parameter were observed in the same sample, each area was judged individually and multiplied by the corresponding area factor. Area factors for a given sample always added up to 4. The histology score was calculated as the sum of all parameter scores multiplied by their area factors. The maximum score was 64. Scores and ulcer quantification were based on one Swiss roll section per mouse and were determined in a blinded fashion. Quantification of CC3 + cells was performed on histological sections immunostained with antibodies against CC3. The total number of CC3 + cells was divided by the number of crypts to show the average number in one crypt. Two hundred crypts for small intestine and at least 74 crypts for colon were analysed per mouse. Counting was performed in a blinded fashion.
Histological analysis of kidneys was performed by applying standardized protocols at the Institute of Surgical Pathology, University Medical Center of Freiburg. In brief, 2-μm microtome sections of formalin-fixed, paraffin-embedded tissue were used for PAS reaction staining with standardized diagnostic procedures. Whole kidney slides (WSI) were digitalized using a Ventana DP 200 slide scanner (Roche Diagnostics Deutschland) equipped with a ×40 objective. The MS index was assessed by applying a four-tiered scoring system (0-3; 0, <5%; 1, 6-25%; 2, 26-50%; 3, >50%). Further quantitative analysis of tuft areas and cells (nuclei) per tuft area was performed using QuPath v0.3.2 image analysis software 48,49 . At least 50 glomerular tufts per mouse were manually segmented by random sampling. Nucleus segmentation of individual tufts was performed by applying the built-in nucleus segmentation tool (QuPath). Histopathological evaluation and quantitative analysis were performed in a blinded fashion by an expert renal pathologist. For analysis of histology, an inverted Zeiss Axio Imager microscope equipped with an Axiocam colour camera, ×10, ×40 and ×100 objectives and a Ventana DP 200 slide scanner was used.

Gene expression analysis by qRT-PCR
Total RNA was extracted with either TRIzol reagent (Life Technologies, 15596018) and chloroform or the NucleoSpin RNA kit (Macherey-Nagel, 740955.50), according to the manufacturer's instructions, followed by cDNA synthesis using the SuperScript III First-Strand Synthesis System (Life Technologies, 18080051) with subsequent treatment with RNase H (Invitrogen, AM2293) or the LunaScript RT SuperMix kit (New England Biolabs, E3010L). qRT-PCR was performed using TaqMan probes and TaqMan Gene Expression Master Mix (ThermoScientific, 4369016) or Luna Universal Probe qPCR Master Mix (New England Biolabs, M3004X) in a QuantStudio 5 Real-Time PCR System (ABI). Reactions were run in technical duplicates with Tbp as a reference gene. Relative expression of gene transcripts is shown using the 2 −ΔCt method and is represented in dot plot graphs as mRNA expression values relative to the reference gene. The TaqMan probes used were as follows: Mm00457981_m1).  50 with stringent settings to remove error-containing reads ('-q 20 --max-n 0 --max-ee 1'). Remaining reads were passed to HISAT2 (v2.1.0) 51 for strand-aware alignment, and strand-specific counts of uniquely mapping reads were prepared using featureCounts (within Subread v1.6.4; ref. 52 ) against Ensembl GRCh38.100 annotations. Additional unstranded counts were obtained with featureCounts against a database of repetitive elements previously prepared for GRCh38 (ref. 53 ) using reads unassigned to features during the previous step.

Differential expression analyses
DESeq2 (v1.22.1) 54 within R was used for read count normalization, and downstream differential expression analysis and visualization were performed within Qlucore Omics Explorer v3.3 (Qlucore). Repeat region annotation, RNA-seq read mapping and counting were carried out as previously described 21 .

Gene functional annotation
Pathway analyses were performed using g:Profiler (https://biit.cs.ut. ee/gprofiler) with genes ordered by the degree of differential expression. P values were estimated by hypergeometric distribution tests and adjusted by multiple-testing correction using the g:SCS (set counts and sizes) algorithm, integral to the g:Profiler server 55 . ISGs were defined according to the Interferome v2.01 database 56 .

Detection and analysis of A-to-I editing
Read alignments were processed with samtools markdup 57 to identify likely PCR duplicates within the sequenced libraries, and A-to-I editing was assessed using JACUSA2 (ref. 58 ) with settings to flag and exclude from analysis potential editing sites in close proximity to the start and end of reads, indel positions and splice sites, as well as sites within homopolymer runs of more than seven bases, using only primary alignments of properly paired, non-duplicate reads. Detected and differential A>G editing sites were filtered for Z score > 1.96, a depth of ≥10 reads, a minimum editing fraction of ≥1%, ≥2 replicates to display editing and exclude potential SNPs, and a maximum editing fraction of <50%. Additionally, for differential sites, a >2-fold difference was required between test and control sample groups. Sites obtained were assigned to genomic features using annotatr v1.16 within R 59 . Assessments of editing enrichment within repetitive elements were Extended Data Fig. 1 | Generation and characterization of Adar1 mZα/mZα mice. a, Schematic depicting the generation of the Adar1 mZα allele using CRISPR-Cas12a-mediated gene targeting. The indicated nucleotide substitutions were introduced in exon 1 of the Adar1 gene to substitute amino acids N175 with D and Y179 with A in the Zα domain. Sequencing traces of the desired mutations are shown in homozygous mice. b, RBC, platelet (PLT), white blood cell (WBC), neutrophil (NEU), monocyte (MON), lymphocyte (LYM) counts and HGB and HCT levels in the blood of 10-to-13-month-old mice with the indicated genotypes. c, Spleen-to-body weight ratio of mice with the indicated genotypes. d, Representative H&E-stained sections from spleen, lung, liver, heart and kidney of mice with the indicated genotypes. Adar1 mZα/mZα (n = 7), Adar1 WT/mZα (n = 10), Adar1 WT/WT (n = 3). Scale bar, 100 μm. e, Transcriptional comparison of lung tissues from Adar1 mZα/mZα , Adar1 WT/mZα and wild type C57BL/6N mice (n = 5 for all genotypes) at 4-5 months of age. The heatmap includes 58 genes differentially expressed between Adar1 mZα/mZα mice and the two control groups combined (p ≤ 0.05, q ≤ 0.05, ≥ 2-fold-change), with 57 of these upregulated in the former (Supplementary Data Table 1). Columns represent individual mice, hierarchically clustered according to differential gene expression. Table shows gene ontology (GO) functional annotation of the 57 genes upregulated in Adar1 mZα/mZα mice, performed with g:Profiler (https:// biit.cs.ut.ee/gprofiler). P values for differential expression analyses were calculated with Qlucore Omics Explorer using two-sided t-tests and with the q value for false discovery rates (FDR) set to 0.05. Calculation of q values was adjusted for multiple hypothesis testing using the Benjamini-Hochberg method. P values for pathway analyses were calculated with g:Profiler using hypergeometric distribution tests and adjusted for multiple hypothesis testing using the g:SCS (set counts and sizes) algorithm, integral to the g:Profiler server (https://biit.cs.ut.ee/gprofiler). f, qRT-PCR analysis of mRNA expression of the indicated genes in indicates tissues from 10-to-13-month-old mice with the indicated genotypes. In b, c and f, dots represent individual mice, bar graphs show mean ± s.e.m and P values were calculated by two-sided nonparametric Mann-Whitney test. Fig. 2 | Histological analysis of tissues from Adar1 WT/mZα , Adar1 −/mZα , Adar1 −/mZα Mavs −/− and Adar1 −/mZα Zbp1 −/− mice. a, Representative H&E-stained sections from liver, lung, heart, small intestine and colon from Adar1 WT/mZα (n = 5), Adar1 −/mZα (n = 7), Adar1 −/mZα Mavs −/− (n = 5) and Adar1 −/mZα Zbp1 −/− (n = 8) mice at P1. Scale bar, 200 μm. Graphs depict histological ileitis and colitis scores. Dots represent individual mice, bar graphs show mean ± s.e.m and P values were calculated by two-sided nonparametric Mann-Whitney test. b, Representative images of PAS-stained kidney sections from Adar1 WT/mZα (n = 6), Adar1 −/mZα (n = 6), Adar1 WT/mZα Zbp1 −/− (n = 6) and Adar1 −/mZα Zbp1 −/− (n = 6) mice at P1. Scale bar, 200 μm (top), 100 μm (middle) or 20 μm (bottom). Mice of all genotypes presented with normal zonal architecture of postnatal kidneys, in particular the cortical nephrogenic zone (white dotted line) appeared unaffected. Black arrows indicate representative protein absorption droplets in proximal tubuli; red arrowheads highlight intact mesangial compartments of respective early mature glomeruli. c-d, Representative images of brain sections from Adar1 WT/mZα (n = 8), Adar1 −/mZα (n = 11), Adar1 −/mZα Mavs −/− (n = 5) and Adar1 −/mZα Zbp1 −/− (n = 6) mice at P1 stained with H&E (c) or immunolabelled for CD3, B220, IBA1 or MAC3 (d). Scale bar, 200 μm. e, Schematic depicting the generation of Mavs −/− mice using CRISPR-Cas9-mediated gene targeting. Two gRNAs were used to cut in exon 3 and exon 6 resulting in deletion of the respective sequence. DNA sequencing in homozygous mice confirmed the resulting fusion between exons 3 and 6 and the generation of a premature stop codon. Fig. 3 | Comparison of RNAseq expression profiles in spleen, lung and brain tissues from Adar1 −/mZα and Adar1 WT/mZα mice. a-c, Transcriptional comparison of spleen (a), lung (b) and brain (c) tissues from Adar1 −/mZα and Adar1 WT/mZα littermate mice at P1. Heatmaps show differentially expressed genes between Adar1 −/mZα and Adar1 WT/mZα control mice. These were 1,594 (p ≤ 0.05, q ≤ 0.05, ≥ 2-fold-change), 657 (p ≤ 0.05, q ≤ 0.05, ≥ 2-foldchange), and 379 (p ≤ 0.05, ≥ 2-fold-change) genes for the spleen, lung and brain, respectively (Supplementary Data Table 2). Columns represent individual mice, hierarchically clustered according to differential gene expression. Tables show GO functional annotation of the genes found upregulated and downregulated in the comparison between the two genotypes in each organ, performed with g:Profiler (https://biit.cs.ut.ee/ gprofiler). P values for differential expression analyses were calculated with Qlucore Omics Explorer using two-sided t-tests and with the q value for false discovery rates (FDR) set to 0.05. Calculation of q values was adjusted for multiple hypothesis testing using the Benjamini-Hochberg method. P values for pathway analyses were calculated with g:Profiler using hypergeometric distribution tests and adjusted for multiple hypothesis testing using the g:SCS (set counts and sizes) algorithm, integral to the g:Profiler server (https://biit. cs.ut.ee/gprofiler). For all genotypes, n = 5 for spleen and lung, n = 4 for brain. , Adar1 WT/mZα (n = 5) and wild type C57BL/6N mice (n = 5) at 15 weeks of age. Heatmap shows 678 genes differentially expressed between Adar1 −/mZα Mavs WT/− mice and the two control groups combined (p ≤ 0.05, q ≤ 0.05, ≥ 2-fold-change), with 399 of these upregulated in the former (Supplementary Data Table 3). Columns represent individual mice, hierarchically clustered according to differential gene expression. Tables show GO functional annotation of the genes found upregulated and downregulated in this comparison, performed with g:Profiler (https://biit.cs.ut.ee/gprofiler). P values for differential expression analyses were calculated with Qlucore Omics Explorer using two-sided t-tests and with the q value for false discovery rates (FDR) set to 0.05. Calculation of q values was adjusted for multiple hypothesis testing using the Benjamini-Hochberg method. P values for pathway analyses were calculated with g:Profiler using hypergeometric distribution tests and adjusted for multiple hypothesis testing using the g:SCS (set counts and sizes) algorithm, integral to the g:Profiler server (https://biit.cs.ut.ee/gprofiler).  [15][16][17][18][19][20] week-old mice with the indicated genotypes. In c, d, control mice include littermates with Adar1 WT/− or Adar1 WT/WT genotypes. Dots represent individual mice, bar graphs show mean ± s.e.m and P values were calculated by two-sided nonparametric Mann-Whitney test. e, Representative H&E-stained sections from the spleen, liver, lung, heart, small intestine and colon of 15-20 week-old Adar1 WT/WT Mavs −/− Zbp1 −/− (n = 5) and Adar1 −/− Mavs −/− Zbp1 −/− (n = 5) mice. Black arrows indicate dying epithelial cells in small intestine and colon sections. Scale bars, 200 μm. f, Immunoblot analysis of total lysates from murine embryonic fibroblasts (MEFs) of the indicated genotypes, stimulated with IFNγ (1,000 U ml −1 ) for 24 h. One representative out of two independent experiments is shown. GAPDH was used as a loading control. g, h, j, Cell death measured by DRAQ7 uptake in MEFs of the indicated genotypes treated with IFNγ (1,000 U ml −1 ) alone or with combinations of IFNγ (1,000 U ml −1 ) (24-h pretreatment), cycloheximide (CHX) (1 μg ml −1 ), QVD-OPh (QVD) (10 μM) and GSK'872 (3 μM). Graph shows mean values from technical triplicates (n = 3), from one representative out of three independent experiments. i, Immunoblot analysis of total lysates from MEFs of the indicated genotypes pre-stimulated with IFNγ (1,000 U ml −1 ) for 24 h followed by treatment with CHX (1 μg ml −1 ) for the indicated time points. Data are representative of two independent experiments. GAPDH was used as loading control. For gel source data, see Supplementary Figure 1.