ADAR1 modifies endogenous RNAs to prevent activation of the innate immune RNA sensors MDA5, OAS–RNAse L and PKR (refs. 1,2,10,11). The interferon-inducible p150 isoform of ADAR1 includes an amino-terminal ZBD, and a naturally occurring point mutation in the sequence encoding this region causes a proline-to-alanine substitution at position 193 in human ADAR1 (ref. 4; Fig. 1a). This mutation is present at a remarkably high rate (about 1 in 360 in individuals of northern European descent), and if paired with a loss-of-function mutation on the second allele of ADAR1, it causes the severe autoinflammatory disease Aicardi–Goutières syndrome12 (AGS). Recently, a mouse model of this mutation was reported that recapitulates the genetic underpinnings and aspects of the pathology of AGS; mice carrying the mutation causing the P195A substitution (homologous to human P193A) on one or both alleles of Adar  (which encodes ADAR1) are phenotypically normal, but mice with this mutation combined with deletion of the p150 isoform in the second allele of Adar (AdarP195A/p150null) exhibit liver, kidney and spleen pathology, are runted, and have a median survival of 25 days (ref. 5).

Fig. 1: Immunopathology in ADAR1-mutant mice is driven by ZBP1.
figure 1

a, Schematic of ADAR1 and ZBP1. Zα and Zβ, Z-DNA binding domains; RBD, RNA-binding domain; RHIM, RIP homotypic interaction motif. b–d, Parental and expected offspring genotypes (b), observed genotypes (c) and percentage survival for each genotype (d) produced by the cross of Adarp150null/WT::Zbp1-a+/− mice to AdarP195A/P195A::Zbp1-a−/− mice. e,f, Observed weights for mice of the indicated genotypes at 3 (e), 6 (f, left) and 8 (f, right) weeks of age. Survival curves in d represent littermates across 18 litters, AdarP195A/WT::Zbp1−/− (n = 19), AdarP195A/WT::Zbp1+/− (n = 24), AdarP195A/p150null::Zbp1+/− (n = 14), AdarP195A/p150null::Zbp1−/−(n = 27). ****P ≤ 0.0001 (Mantel–Cox log-rank test). For litter weights: AdarP195A/WT::Zbp1+/+ (n = 17), AdarP195A/p150null::Zbp1+/+ (n = 18), AdarP195A/p150null::Zbp1+/+ (n = 5). ****P ≤ 0.0001, unpaired t-test, two-tailed. Weights: 6 weeks, from left to right, n = 15, 19, 18, 17, 2, 1; 8 weeks, from left to right, n = 10, 18, 17, 15, 2. The bars in e and f represent mean ± s.d.

Source data

ZBP1 loss rescues ADAR1 mutation

The AdarP195A/p150null model is driven by a point mutation in the sequence encoding the ZBD of ADAR1 (Fig. 1a). As ZBP1 contains the only other mammalian ZBD, we wondered whether these two proteins might functionally interact. Consistent with this possibility, our observations showed that ADAR1 co-immunoprecipitated with ZBP1, but that this interaction was abrogated when point mutations were introduced to the sequence encoding the ZBD of ZBP1 that prevented RNA binding (Extended Data Fig. 1a,b). We also observed that the interaction between ZBP1 and ADAR1 was strengthened by ultraviolet crosslinking, which covalently links protein to nucleic acid (Extended Data Fig. 1c). As these data suggested that ADAR1 and ZBP1 bind a common ligand through their ZBDs, we wondered whether the pathology caused by alteration of the ZBD of ADAR1 in the AdarP195A/p150null mouse model was driven by aberrant ZBP1 activation. To test this idea, we crossed AdarP195A/p150null mice to mice that lack ZBP1, using a widely used Zbp1-knockout mouse strain13, referred to here as Zbp1-a (Fig. 1b). We observed that AdarP195A/p150null::Zbp1-a−/− mice were born at expected frequencies and appeared phenotypically normal (Fig. 1c,d and Extended Data Fig. 2a,b). Following birth, AdarP195A/p150null::Zbp1-a−/− mice gained weight slightly more slowly than AdarP195A/WT littermates (Fig. 1e,f), but otherwise exhibited normal phenotype, fertility and survival. The kidney and liver pathology previously reported in AdarP195A/p150null mice5 was also largely normalized by ablation of Zbp1-a (Extended Data Fig. 2c).

As we were generating these data, it was reported that the Zbp1-a−/− mouse line was not fully congenic to the C57BL/6 genetic background14, a finding confirmed by our own single nucleotide polymorphism analysis (Extended Data Fig. 3a,b). We therefore generated a second cross of AdarP195A/p150null mice to a separately derived, fully congenic Zbp1−/− strain15 referred to here as Zbp1-g (Extended Data Fig. 3c). AdarP195A/p150null::Zbp1-g−/− mice also appeared phenotypically normal (Extended Data Fig. 3d). Notably, we also observed an extension of survival in AdarP195A/p150null::Zbp1-g+/− mice, an effect not observed in AdarP195A/p150null::Zbp1-a+/− mice (Extended Data Fig. 2a). Together, these data confirm that loss of ZBP1 reverses the immunopathology observed in AdarP195A/p150null mice.

Complete ablation of the p150 isoform of ADAR1 causes uniform lethality during embryonic development, but crossing these mice to mice that lack the double-stranded-RNA sensor MDA5 (encoded by Ifih1) allows Adarp150null/p150null mice to survive to birth1. As MDA5 is a potent inducer of type I interferon (IFN), and because ZBP1 expression is stimulated by IFN (Extended Data Fig. 4a), we reasoned that loss of MDA5 might rescue Adar-p150 knockout  in mice by preventing ZBP1 upregulation. However, in mouse embryonic fibroblasts (MEFs), we observed that although loss of MDA5 completely abrogated IFN production induced by ADAR1 depletion, ZBP1 upregulation was only modestly attenuated (Extended Data Fig. 4b). This suggested that loss of ADAR1 could lead to ZBP1 upregulation and activation even in the absence of MDA5. Consistent with this finding, our observations showed that the survival of Adarp150null/p150null::Ifih1−/− mice was modestly but significantly extended by concurrent deletion of ZBP1 (Extended Data Fig. 4c). However, deletion of ZBP1 alone did not allow Adarp150null/p150null mice to survive to birth (Extended Data Fig. 4d). Together, these data indicate that the developmental lethality induced by deletion of ADAR1-p150 is mediated by simultaneous activation of ZBP1, MDA5 and other pathways.

An inflammatory signature remains in rescued mice

AdarP195A/p150null mice exhibit aberrant activation of MDA5, which drives IFN-dependent inflammation. We reasoned that this IFN-dependent signature would still be present in AdarP195A/p150null::Zbp1-g−/− mice, despite their normal phenotype. Consistent with this possibility, RNA-sequencing analysis of spleens from 23-day-old pups revealed that many aspects of the aberrant inflammatory and interferon-stimulated gene (ISG) signature present in AdarP195A/p150null mice were also present in AdarP195A/p150null::Zbp1−/− mice, despite them having no alterations in splenic cellularity (Fig. 2a–c, Extended Data Fig. 5a,b and Extended Data Fig. 6). Gene ontology analysis confirmed that the antiviral gene signature induced by alteration of ADAR1 was conserved in AdarP195A/p150null::Zbp1-g−/− mice (Extended Data Fig. 5c). This analysis also identified gene signatures present in AdarP195A/p150null mice but absent in AdarP195A/p150null::Zbp1-g−/− mice (Extended Data Fig. 5d). These genes may indicate targets of ZBP1 signalling or may represent pathways upregulated as a result of the immunopathology present in AdarP195A/p150null mice. In carrying out these analyses, we noted that although cells from AdarP195A/p150null::Zbp1−/− mice maintain many aspects of the ISG signature observed in AdarP195A/p150null mice, its magnitude is reduced (Fig. 2c and Extended Data Fig. 5a). This was confirmed by direct comparison of MEFs isolated from AdarP195A/p150null or AdarP195A/p150null::Zbp1−/− mice (Fig. 2d), and suggested that ZBP1 may have a role in augmenting ISG upregulation, a function previously ascribed to ZBP1 in other contexts16.

Fig. 2: The inflammatory program initiated by alteration of ADAR1 remains intact following knockout of Zbp1.
figure 2

a,b, Volcano plots depicting differential expression of genes detected by RNA sequencing of whole spleen tissue derived from 23-day-old mice of the genotypes (a) AdarP195A/p150null (affected) n = 4 (a) and AdarP195A/p150null::Zbp1-g−/− (rescued) n = 5 (b), in each case compared to spleens derived from WT pups (n = 4). c, Raincloud plots of affected (n = 4) and ZBP1 (n = 5) rescued mice depicting the statistically significant upregulated and downregulated ADAR1 signature identified in a. The ADAR1 signature was identified from the differential analysis of ADAR1 versus WT from genes with a log2[fold change] > 1 and adjusted P value < 0.01. d, Heatmap analysis of differential gene expression from NanoString analysis of WT versus Zbp1−/− ADAR1-mutant MEFs. Colours denote a row z-score (−2 to 2) indicating differential expression of gene counts of indicated genes. e,f, Cell death, as measured by loss of plasma membrane integrity using an IncuCyte imager, of Adar-mutant MEFs after 8 h zVAD treatment on Adar::Zbp1-mutant MEFs (genetic knockout/mutants) (e) or Adar-mutant MEFs with CRISPR–Cas9 knockout of Zbp1 (f). GSK843 is an inhibitor of RIPK3. Each experimental group (bar) contains n = 8 biological replicates (8 wells). Statistical significance was determined by unpaired Student t-tests, two-tailed. Incucyte analyses (e,f) are single representatives of independently duplicated experiments. Whisker bars (e,f) represent mean ± s.d.

Source data

We next sought to understand the pathways downstream of ZBP1 that are activated by alteration of ADAR1. We observed that culturing MEFs from AdarP195A/p150null mice led to the ZBP1-dependent loss of RIPK3 expression (Extended Data Fig. 7a), suggesting that the ZBP1–RIPK3 pathway is constitutively active in these cells. Consistent with this, our observations showed that despite the reduced level of RIPK3 expression in these cells, treating them with the caspase inhibitor zVAD caused ZBP1-dependent RIPK3 phosphorylation and cell death (Fig. 2e,f and Extended Data Fig. 7a,b). Together, these results indicated that AdarP195A/p150null cells are sensitized to ZBP1-dependent necroptosis.

Cell death signalling in ADAR1 mutant mice

Given this finding, we next sought to address the role of the necroptotic pathway in the pathology of AdarP195A/p150null mice. To do this, we crossed AdarP195A/p150null mice to mice that lack different components of the necroptotic pathway. Ablation of the necroptotic effector MLKL did not alter the phenotype of AdarP195A/p150null mice (Fig. 3a and Extended Data Fig. 8a), nor did crossing them to mice in which the kinase activity of RIPK1 is absent (Fig. 3b and Extended Data Fig. 8b), indicating that prevention of canonical necroptosis alone was not sufficient to explain the effect of Zbp1 knockout in these mice.

Fig. 3: ZBP1-induced necroptosis does not underlie immunopathology induced by alteration of ADAR1.
figure 3

ac, Survival proportions for AdarP195A/p150null mice crossed to mice that lack MLKL (a), carrying a point mutation abrogating the kinase activity of RIPK1 (Ripk1kd) (b) or lacking RIPK3 (c). In each case, the result of a breeding scheme analogous to that depicted in Fig. 1b is shown. Survival proportions (ac) represent littermates from 9 (Mlkl; a), 10 (Ripk1kinase-dead (Ripk1kd); b) or 18 (Ripk3; c) litters; AdarP195A/WT::Mlkl+/− (n = 28), AdarP195A/WT::Mlkl−/− (n = 10), AdarP195A/p150null::Mlkl+/− (n = 14), AdarP195A/p150null::Mlkl−/− (n = 9); AdarP195A/WT::Ripk1kd/+ (n = 25), AdarP195A/WT::Ripk1kd/kd (n = 27), AdarP195A/p150null::Ripk1kd/+ (n = 16), AdarP195A/p150null::Ripk1kd/kd (n = 13); AdarP195A/WT::Ripk3+/− (n = 45), AdarP195A/WT::Ripk3−/− (n = 42), AdarP195A/p150null::Ripk3+/− (n = 26), AdarP195A/p150null::Ripk3−/− (n = 34). Survival statistics determined by log-rank (Mantel–Cox) test; exact P values are indicated on the curves. NS, not significant. d, Weights of male (left) and female (right) Ripk3−/− mice, observed at 21 days after birth. For litter weights: AdarP195A/WT::Ripk3+/− (male/female n = 13/12), AdarP195A/WT::Ripk3−/− (male/female n = 12/11), AdarP195A/p150null::Ripk3−/− (male/female n = 11/6), AdarP195A/p150null::Ripk3+/− (male/female n = 11/13). Statistical differences were determined by individual unpaired t-tests (two-tailed); exact P values are indicated on the plots. e,f, Cell death, as measured by loss of plasma membrane integrity, observed in MEFs from Mlkl−/− and Mlkl−/−::Casp8−/− (n = 3 for each group) (e) and Ripk3−/− and Ripk3−/−::Casp8−/− (n = 4 for each group) (f) genotypes stably expressing 2×FV-ZBP1, following treatment with the activating drug B/B. The data in e and f are representative of three independently replicated experiments. Whisker bars (df) represent mean ± s.d.

Source data

By contrast, ablation of RIPK3 led to a significant extension of survival in AdarP195A/p150null mice (Fig. 3c). However, this extension of survival was partial, with approximately one-third of AdarP195A/p150null::Ripk3−/− mice succumbing to death within 40 days of birth, followed by slower attrition of mice over the next 200 days. AdarP195A/p150null::Ripk3−/− were severely runted, in contrast to the overtly normal phenotype observed in AdarP195A/p150null::Zbp1−/− mice (Fig. 3d). These findings indicate that ZBP1 can drive pathology in AdarP195A/p150null through signalling that is independent of RIPK3. After influenza infection, ZBP1 was reported to induce RIPK3-dependent induction of both caspase-8-dependent apoptosis and MLKL-dependent necroptosis17. Nonetheless, our finding that in AdarP195A/p150null mice ablation of RIPK3 failed to fully recapitulate the reversal of pathology observed following ablation of ZBP1 implies RIPK3-independent functions of ZBP1 when activated downstream of ADAR1 deficiency.

ZBP1 contains three confirmed RIP homotypic interaction motifs, which bind to similar domains present in RIPK1 and RIPK3 (ref. 18). We reasoned that, in the absence of RIPK3, ZBP1 may interact with RIPK1 to scaffold and activate caspase-8-dependent apoptosis. To study this, we turned to a reductive system in which ZBP1 could be directly activated. By replacing the ZBD of ZBP1 with tandem inducible dimerization domains derived from the protein FK506, we created a form of ZBP1 that could be activated using the cell-permeable small molecule B/B (Extended Data Fig. 9a). In wild-type (WT) lung epithelial type 1 (LET1) cells, SV40-transformed endothelial cells (SVECs) or MEFs, this construct (termed 2×FV-ZBP1) triggered cell death that was blocked only with combined inhibition of RIPK3 and the caspases, consistent with induction of both apoptosis and necroptosis downstream of ZBP1 activation (Extended Data Fig. 9b–d). Furthermore, B/B activation of ZBP1 resulted in phosphorylation of RIPK3 and MLKL, and when ZBP1 was directly activated in Mlkl−/− MEFs, it induced robust cell death that was dependent on caspase 8 and involved cleavage of caspase 3, consistent with apoptosis (Fig. 3e and Extended Data Fig. 10a,b). ZBP1 activation in Ripk3−/− MEFs also induced caspase-8-dependent cell death and caspase 3 cleavage, albeit with slower kinetics than those observed in Mlkl−/− cells (Fig. 3e,f). Notably, ZBP1 activation did not trigger detectable cell death in MEFs lacking both MLKL and caspase 8, or both RIPK3 and caspase 8 (Fig. 3e,f). This finding indicates that although RIPK3 can contribute to ZBP1-dependent apoptosis, ZBP1 can still drive caspase-8-dependent cell death responses in the absence of RIPK3.

Caspase 8 suppresses lethal inflammation

We reasoned that cell death dependent on ZBP1, RIPK1 and caspase 8 could underlie the pathology observed in AdarP195A/p150null::Mlkl−/− and AdarP195A/p150null::Ripk3−/− mice. Deletion of caspase 8 causes embryonic lethality due to unrestrained necroptosis18, but this phenotype is reversed by co-ablation of RIPK3 or MLKL (refs. 19,20,21). We therefore generated AdarP195A/p150null::Mlkl−/−::Casp8−/− and AdarP195A/p150null::Ripk3−/−::Casp8−/− mice to test whether ablation of caspase 8 in addition to necroptotic signalling would recapitulate the phenotypic rescue observed following ablation of ZBP1. Unexpectedly, AdarP195A/p150null::Ripk3−/−::Casp8−/− mice exhibited reduced weight and survival compared to AdarP195A/p150null::Ripk3−/−::Casp8+/− littermates (Fig. 4a,b and Extended Data Fig. 10c), whereas AdarP195A/p150null::Mlkl−/−::Casp8−/− mice were born at expected frequencies but uniformly failed to survive to weaning (Fig 4c). Histological analysis of these mice at birth revealed broadly normal development of the liver and kidney (Extended Data Fig. 11a–c), but a significant increase in IBA1-positive activated microglia in the brains of AdarP195A/p150null::Mlkl−/−::Casp8−/− neonates, consistent with unrestrained inflammatory signalling at this site (Extended Data Fig. 11d,e). These findings imply that caspase 8 suppresses ZBP1 activation in AdarP195A/p150null mice, and that ablation of caspase 8 may allow unrestrained ZBP1 transcriptional signalling that is independent of canonical apoptosis or necroptosis. Consistent with this possibility, co-immunoprecipitation experiments revealed recruitment of RIPK1, or of both RIPK1 and RIPK3, following 2×FV-ZBP1 activation in Ripk3−/−::Casp8−/− or Mlkl−/−::Casp8−/− MEFs, respectively, despite the lack of cell death responses observed in these conditions (Fig 3e,f and Extended Data Fig. 12a,b). We also observed that in AdarP195A/p150null::Mlkl−/− MEFs, zVAD treatment stabilized the interaction between ZBP1 and RIPK3, again in the absence of cell death responses (Extended Data Fig. 12c). These findings indicate that when necroptotic effectors are absent, loss or inhibition of caspase 8 promotes interactions between ZBP1 and the RIP kinases.

Fig. 4: Caspase 8 suppresses lethal inflammatory signalling in ADAR1-mutant mice.
figure 4

a,b, Ripk3−/−::Casp8−/−::AdarP195A/p150null cross survival (a) and weight (b) at weaning (day 21). For a, the log-rank (Mantel–Cox) statistical test carried out on survival curves for survival to day 50 between Ripk3−/−::AdarP195A/p150null littermates (AdarP195A/WT mice excluded from comparison) that were Casp8+/+ (n = 26), Casp8+/− (n = 43) and Casp8−/− (n = 25) showed a significant decrease in the proportion of survival for Casp8−/− mice compared to that for Casp8+/+ (P = 0.0468) and Casp8+/− (P = 0.0139) mice. For b, the difference between Casp8+/+ and Casp8+/− was not significant. Individual Student t-tests carried out on affected mouse weights (AdarP195A/p150null) show a statistically significant decrease in Casp8−/− (n = 20) compared to Casp8+/+ (n = 25), P = 0.0284, and Casp8+/− (n = 35), P = 0.0179, mice. c, Observed and expected mice from the Mlkl−/−::Casp8−/−::AdarP195A/p150null cross at birth (day 0) and at weaning (day 21). Chi-square power analysis was carried out on observed/expected frequencies at birth (not significant) and weaning (***P = 0.000596). d, Volcano plots showing differential gene expression from NanoString analysis of MEFs derived from Mlkl−/− or Mlkl−/−::Casp8−/− embryos, following depletion of ADAR1 mediated by short interfering RNA. e, Survival proportions for Trex1−/− (n = 61) and Trex1−/−::Ripk3−/− (n = 37) mice. Not significant (Mantel–Cox log-rank test). Whisker bars (b) represent mean ± s.d.

Source data

We next assessed whether loss of caspase 8 potentiated transcriptional signalling induced by ADAR1 insufficiency. We observed that whereas depletion of ADAR1 mediated by short interfering RNA in Mlkl−/− cells induced the expected upregulation of ISGs, ADAR1 depletion in Mlkl−/−::Casp8−/− MEFs induced a distinct transcriptional response dominated by NF-κB targets (Fig. 4d and Extended Data Fig. 13a,b). As RIPK1 is a key signalling adapter upstream of NF-κB activation22, this finding implied that ZBP1–RIPK1 signalling may underlie these transcriptional effects. Consistent with this, our observations showed that 2×FV-ZBP1 activation in Mlkl−/− MEF s induced RIPK1-dependent upregulation of the NF-κB targets CCL2 and CCL7, but not of the canonical ISG IFIT1 (Extended Data Fig. 13c). Together, these data indicate that following alteration or depletion of ADAR1, caspase 8 acts to suppress a ZBP1- and RIPK1-dependent program of inflammatory transcription. The modest extension of survival observed in AdarP195A/p150null::Ripk3−/−::Casp8−/− mice relative to AdarP195A/p150null::Mlkl−/−::Casp8−/− mice suggests that RIPK3 can potentiate, but is not required for, ZBP1- and RIPK1-dependent inflammatory signalling.

This study does not address the identity of the ligand(s) responsible for activating ZBP1 in AdarP195A/p150null mice. As the P195A alteration in ADAR1 lies in its ZBD, we can speculate that ADAR1(P195A) may be attenuated in its ability to bind ZBD ligands. Interestingly, aligning the ZBD sequences of ADAR1 and ZBP1, along with those present in the fish PKR homologue PKZ (ref. 23) and the vaccinia virus effector E3L (ref. 24), reveals that ZBP1 naturally contains an alanine at position 64, the site homologous to P195 of ADAR1 (Extended Data Fig. 14a). As substitution of alanine for proline at this site in ADAR1 limits its function, we speculate that the presence of an alanine at the homologous site within ZBP1 may reflect a naturally lower affinity for ligand by ZBP1 relative to other ZBDs. This may represent a means to limit aberrant ZBP1 activation at steady state. We sought to test this idea by creating a mouse line with a 'revertant' ZBP1, in which A64 is mutated to proline. However, ZBP1(A64P) mice did not reveal evidence of increased ZBP1 activation, either when crossed to the AdarP195A/p150null model or in response to influenza infection, but rather seemed attenuated in their signalling in both settings (Extended Data Fig. 14b,c). Although ZBP1(A64P) protein was properly expressed (Extended Data Fig. 14d), this attenuation may reflect a disruption of protein structure induced by this alteration, and implies that further alterations or larger domain swaps would be needed to effectively test this hypothesis. Future studies will clarify the identity of ZBP1 ligands that emerge following alteration of ADAR1.

AGS refers to a family of IFN-driven congenital pathologies driven by alterations in proteins involved in nucleotide sensing and regulation12. As ZBP1 is strongly induced by IFN, and has been described to bind DNA as well as RNA16, we wondered whether ZBP1 and RIPK3 signalling might have a role in AGS-like pathology driven by endogenous DNA ligands. To test this, we assessed mice lacking TREX1, a DNA exonuclease whose ablation causes aberrant activation of the cGAS–STING pathway25,26. However, we did not observe significant amelioration of pathology or extension of survival in Trex1−/− mice when RIPK3 was ablated, unlike the partial rescue observed in AdarP195A/p150null mice following knockout of Ripk3 (Figs. 4e and 3c). This suggests that engagement of the ZBP1–RIP kinase pathway is a feature of dysregulated endogenous RNA, but not DNA, sensing.

The unexpected susceptibility of Ripk3−/−::Casp8−/− and Mlkl−/−::Casp8−/− mice to alteration of ADAR1 indicates that although these mice develop normally, they are poised for hyperactive inflammatory signalling in response to ZBP1 activation. Indeed, previous studies have found that Mlkl−/−::Fadd−/− mice (comparable to Mlkl−/−::Casp8−/−) are highly susceptible to influenza infection8, a setting in which ZBP1 is strongly activated. Although this was interpreted as indicating a requirement for functional cell death pathways for antiviral defence, our data raise the possibility that these mice succumb to uninhibited inflammatory signalling triggered by ZBP1. Notably, we did not observe engagement of pyroptotic cell death following ZBP1 activation in cells lacking caspase 8 in combination with RIPK3 or MLKL, although our data do not rule out contribution of pyroptotic signalling to the immunopathology we observe.

Our findings identify ZBP1 as a key effector of autoinflammatory pathology induced by alteration of the Z-DNA-binding domain of ADAR1. Our data also highlight the pleiotropic nature of ZBP1 signalling: whereas ZBP1 ablation fully rescued the pathology of the AdarP195A/p150null model, individual deletion of the necroptotic signalling molecules MLKL or RIPK3 did not, and ablation of caspase 8 unleashed a lethal inflammatory program, apparently in the absence of ZBP1-dependent programmed cell death. These findings are consistent with the dual functions of caspase 8 as both an inducer of cell death and a suppressor of ZBP1-dependent necroptosis and inflammation. Indeed, both the presence and the absence of caspase 8 may drive pathology in ADAR1-mutant mice. Our in vitro data indicate that ZBP1 activation in the absence of RIPK3 can induce caspase-8-dependent cell death, and this pathway may contribute to the pathology of AdarP195A/p150null::Ripk3−/− mice; conversely, further ablation of caspase 8 in these mice exacerbates the observed pathology by instead unleashing unrestrained inflammatory signalling. Ultimately, apoptosis, necroptosis and inflammatory transcription may all contribute to the pathology of AdarP195A/p150null mice, and which of these pathways is engaged following alteration of ADAR1 probably varies between tissues and cell types depending on the abundance of the pathway components as well as of regulatory proteins such as cFLIP and the IAPs. This pleiotropy also suggests that even the combined targeting of apoptosis and necroptosis using small-molecule inhibitors is unlikely to reverse AGS pathology driven by alteration of ADAR1.



Mouse strains with modifications to Zbp1-a (ref. 13), Zbp1-g (ref. 15), AdarP195A/p150null (ref. 5), Mlkl (ref. 27), Ripk1kd (ref. 28), Ripk1mutRHIM (ref. 29), Ripk3 (ref. 30) and Casp8 (refs. 19,31), Trex1 (ref. 26) and Ifih1 (MDA5) (ref. 1) have been described previously. Zbp1A64P mice were generated as previously described32, using the single-guide RNA (gRNA) target sequence CCGCCTATGCTCCATGTTGCAGG and the repair-template sequence5′-AAAACCCTCAATCAAGTCCTTTACCGCCTGAAGAAGGAGGACAGAGTGTCCTCCCCA GAGCCTCCAACATGGAGCATAGGCGGGGCTGCTTCTGGAGATGGGGCTCCTGCAATCCCTGAGAACTCCAGT-3′. In brief, C57BL6/J oocytes were microinjected with Cas9 complexed with a single gRNA and single-stranded-DNA donor template as described, and then implanted into pseudopregnant female mice. Founder pups were screened using the Surveyor assay, and resulting mice were bred to homozygosity and genotyped using a Taqman probe system to detect the Zbp1A64P G>C nucleotide change using the primer sequences 5′-CCTCAATCAAGTCCTTTACC-3′ (sense) and 5′-GACAGATTACCAAGGCTAGG-3′ (antisense), 5′-CAGAGCCTGCAACATGGAG-3′ (WT probe) and 5′-CAGAGCCTCCAACATGGAG-3′ (mutant probe). All mice were housed in pathogen-free facilities at the University of Washington under 12-h light–dark cycles with access to food and water ad libitum. Temperatures were set to 74 ± 2 °F with humidity of 30–70%. All animals used were cared for and used in experiments approved by the University of Washington Institutional Animal Care and Use Committee (under protocols 4298-01 and 4190-01) in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care in accordance with the Guide for the Care and Use of Laboratory Animals and applicable laws and regulations. In breeding experiments, no specific criteria were used to determine final sample size. These experiments were not randomized or blinded.

Analysis through single nucleotide polymorphism typing of Zbp1-a and Zbp1-g mouse strains was carried out by Taconic Biosciences using the Mouse Genome Scan Panel.

Cell lines

MEFs were generated from pups at embryonic day 15 and immortalized by retroviral transduction of the SV40 large T antigen. HEK293T cells, LET1 cells and MEFs were maintained in standard conditions: Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum, glutamine, penicillin and streptomycin. Following isolation and immortalization, MEF lines were tested for cell death competence by stimulation with TNF/zVAD (RIPK3/RIPK1-dependent cell death) or infection with influenza A virus (ZBP1/RIPK3-dependent cell death).

Plasmids, lentiviral vectors and short interfering RNA

2×FV-ZBP1 was generated by replacing the first 146 amino acids of mouse ZBP1 (corresponding to the ZBD) with tandem copies of FKBP(F36V), from Clontech. The resulting fusion gene was cloned into the Tet-based inducible expression pSLIK vector33, and this construct was used to create lentiviral particles for transduction of target lines using standard protocols. Expression and activation of this construct were achieved by inducing 2×FV-ZBP1 expression with 1 μg ml−1 doxycycline for 12 h, and then treating with 100 mM B/B homodimerizer (Clontech). The full ADAR1-p150 isoform was cloned from mouse cDNA directly into a pRRL lentiviral backbone and subsequently sequenced to confirm identity. WT and mutant Zαβ (mZαβ) ZBP1 were subcloned from constructs previously obtained from the laboratory of Jason Upton into the pRRL backbone, at which point a 3×Flag tag was added to the carboxy terminus.

CRISPR–Cas9-mediated deletion was achieved using a lenti-CRISPR construct created by D. Stetson5, into which guide sequences listed below were inserted. These were used to create VSV-G-pseudotyped lentivirus particles, which were used to transduce target cells. Following 10–14 days of antibiotic selection, deletion of target proteins was confirmed by western blot. The sequences of the guide RNA (gRNA) target sites are as follows: non-targeting control gRNA: scramble gRNA: 5′-GACGGAGGCTAAGCGTCGCAA-3′, Zbp1 gRNA: 5′-GAGCCTGCAACATGGAGCAT-3′, Ifih1 (MDA5) gRNA: 5′-GTGTGGGTTTGACATAGCGCG-3′.

Short interfering RNA (siRNA) experiments were carried out by transfecting cells with SMARTpool siRNA cocktails (Dharmacon Horizon Discovery) targeting ADAR1 (siGenome mouse Adar, Entrez Gene 56417), RIPK1 (siGenome mouse Ripk1, Entrez Gene 19766) or a non-targeting scramble control (siGenome Non-targeting siRNA Pool 1). Transfection of siRNA was carried out using the dharmaFECT 1 transfection reagent (catalogue number T-2001-03, Horizon Discovery) according to the manufacturer’s protocols.

Antibodies and inhibitors

Where indicated, the following drugs were used at the listed concentrations: 50 μM zVAD (SM Biochemicals); 100 nM GSK843 (GlaxoSmithKline).

The following antibodies were used for western blots and immunoprecipitations: ADAR1 (15.8.6, SantaCruz), ZBP1 (Zippy-1, AdipoGen), actin (13E5, Cell Signaling Technology), MDA5 (D74E4, Cell Signaling Technologies), pRIPK3 (GEN135-35-9, Genentech), RIPK3 (1G6.1.4, Genentech or 2283, ProSci), pMLKL (D6E3G, Cell Signaling Technologies), MLKL (MABC604, Millipore), caspase 3 (9662, Cell Signaling Technologies), cleaved caspase 3 (9661, Cell Signaling Technologies), RIPK1 (38/RIP, BD Biosciences), anti-Flag (M2, Sigma) and anti-FKBP12 (PA1-026A, Thermo Fisher). IBA1 (catalogue number 019-19741, Wako-Chem) and cleaved caspase 3 (clone D3E9, Cell Signaling Technologies) were used in immunohistochemical analysis.

The following antibodies were used for flow cytometry analysis of splenocytes: FITC anti-CD19 (clone 1D3, BD Biosciences), PerCP–Cy5.5 anti-CD3e (clone 145-2C11, BD Biosciences), PE–Cy7 anti-Ly6C (clone HK1.4, Biolegend), APC anti-F4/80 (clone BM8, eBioscience), AF700 anti-Ly6G (clone 1A8, Biolegend), APC–Cy7 anti-NK1.1 (clone PK136, BD Biosciences), BV510 anti-CD8a (clone 53-6.7, BD Biosciences), BV605 anti-CD4 (clone RM4-5, BD Biosciences), BV650 anti-CD11b (clone M1/70, Biolegend) and BUV395 anti-CD45.2 (clone 104, BD Biosciences).

Western blot, immunoprecipitation and infrared crosslinkingand immunoprecipitation

WT HEK293T cells or HEK293T cells expressing Flag–ZBP1 or Flag–ZBP1 mZαβ were cultured in 6-well plates and were transfected with pRRL vector expressing ADAR1-p150 or empty pRRL vector using ×2 transfection reagent (Mirus). At about 24 h post-transfection, cells were washed in PBS and crosslinked with 254 nm ultraviolet C light (0.3 J cm−2) or left uncrosslinked, and lysed in 200 μl immunoprecipitation lysis buffer (25 mM Tris pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.5 mM EDTA, 5 mM MgCl2 + 1× protease inhibitor). Lysates were clarified by centrifugation at 5,000g for 10 min at 4 °C, and quantified by Bradford assay. A fraction of each sample was stored for input controls. A 200 μg quantity of each lysate was incubated with Protein G Dynabeads (Thermo Fisher) pre-conjugated with 4 μg anti-Flag M2 antibody (Sigma) in a final volume of 500 μl immunoprecipitation lysis buffer for 2 h at 4 °C with rotation. The beads were then washed four times with 1 ml immunoprecipitation lysis buffer before boiling them in 2× Laemmli buffer (Bio-Rad) + 5% beta-mercaptoethanol to elute protein complexes. Eluates and input controls were resolved on a 4–15% TGX gel (Bio-Rad) for SDS–polyacrylamide gel electrophoresis, and transferred to polyvinylidene difluoride membranes. Immunoblotting was carried out using horseradish-peroxidase-conjugated anti-Flag and anti-β-actin antibodies, and with anti-ADAR1 primary antibody and horseradish-peroxidase-conjugated anti-mouse secondary antibody (Jackson Immunolabs).

For RIPK3 and RIPK1 immunoprecipitations, Cells were lysed in ice-cold lysis buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% v/v Triton X-100, 10% v/v glycerol and 0.01% w/v SDS) supplemented with 1× complete Protease Inhibitor (Roche) for 30 min, followed by centrifugation at 1,000g for 10 min. Antibodies to ZBP1 or FKBP12 (Invitrogen, PA1-026A) were immobilized to Dynabeads Protein G (Invitrogen) as per the manufacturer’s instructions, and then incubated with total cell lysates overnight at 4 °C. Immunoprecipitates were eluted in Laemmli sample buffer (63 mM Tris-HCl, pH 8.0, 10% v/v glycerol, 2% w/v SDS, 0.01% w/v bromophenol blue, 2.5% v/v 2-mercaptoethanol) at 95 °C for 10 min.

We carried out infrared crosslinking and immunoprecipitation (irCLIP) as described previously34 with slight modifications. WT HEK293T cells or HEK293T cells stably expressing Flag–ZBP1 or Flag–ZBP1 mZαβ were cultured in 6-well plates. Cells were washed with PBS, crosslinked with 254 nm ultraviolet C light (0.3 J cm−2) and lysed in 200 μl irCLIP lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 1× protease inhibitor). After sonication in ice slurry, lysates were clarified by centrifugation at 5,000g for 10 min at 4 °C, and quantified by Bradford assay (Bio-Rad). A fraction of each sample was stored for input controls. A 200 μg quantity of each lysate was incubated with Protein G Dynabeads (Thermo Fisher) pre-conjugated with 4 μg anti-Flag M2 antibody (Sigma) in a final volume of 500 μl irCLIP lysis buffer for 2 h at 4 °C with rotation. The beads were then sequentially washed with the following ice-cold buffers: once with 1 ml irCLIP lysis buffer, once with 1 ml high-stringency buffer (20 mM Tris pH 7.5, 120 mM NaCl, 25 mM KCl, 5 mM EDTA, 1% Triton X-100, 1% NaDOC, 0.1% SDS), once with 1 ml high-salt buffer (20 mM Tris pH 7.5, 500 mM NaCl, 5 mM EDTA, 1% Triton X-100, 1% NaDOC), once with 1 ml low-salt buffer (20 mM Tris pH 7.5, 5 mM NaCl, 5 mM EDTA, 1% Triton X-100) and twice with 0.5 ml NT2 buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1 mM MgCl2, 0.05% NP-40). Beads were then resuspended in 30 μl NT2 buffer containing 25 ng ml−1 RNase A (Thermo Fisher) and 15% PEG400 (Sigma) for on-bead RNase digestion at 30 °C for 15 min with shaking (1,200 r.p.m.) in a Thermomixer. RNase digestion was quenched by the addition of 0.5 ml high-stringency buffer. Beads were washed twice with 0.3 ml PNK wash buffer (50 mM Tris pH 7.0, 10 mM MgCl2), and then resuspended in 30 μl PNK dephosphorylation mix (1× PNK buffer (Promega), 0.5 μl RNaseIN (Promega), 1 μl T4 PNK (Promega), 4 μl PEG400). Dephosphorylation reactions were conducted at 37 °C for 60 min with shaking (1,200 r.p.m.) in a Thermomixer. Dephosphorylation mix was removed and beads were washed with 0.25 ml PNK wash buffer. For ligation of oligonucleotide conjugated with infrared dye to protein-crosslinked RNA, beads were resuspended in 30 μl RNA ligation mix (1× RNA ligase I buffer (NEB), 1 μl RNA ligase I (NEB), 1 μl infrared-dye-labelled oligonucleotide34, 5 μl PEG400 and 0.5 μl RNaseIN) and incubated for 16 h at 16 °C with shaking in a Thermomixer (1,200 r.p.m.). The ligation mix was then removed, and the beads were washed twice with 0.25 ml PNK wash buffer, before elution of RNA–protein complexes in 20 μl 1× LDS Buffer (Thermo Fisher) + 10% beta-mercaptoethanol at 80 °C for 10 min. A 5 μl volume of eluates, as well as input controls, was then resolved by SDS–polyacrylamide gel electrophoresis on 4–12% Bis-Tris NuPAGE gels (Thermo Fisher), and transferred to nitrocellulose membranes. Fluorescent RNA–protein complexes in the eluates were visualized on a LiCOR Odyssey FC imager.

Flow cytometry

Splenocytes were blocked with Fc block (BD Biosciences) in PBS + 2% heat-inactivated fetal bovine serum for 10 min at 4 °C before cell surface staining by subsequent addition of a pre-mixed antibody cocktail. Cells were incubated with fluorescently labelled antibodies, each at a dilution of 1:200 for 30 min at 4 °C, washed and fixed with 2% paraformaldehyde in PBS for 10 min. Data were acquired on a BD FACSymphony A3 Cell Analyzer and using the BD Diva acquisition software (version 9.0) analysed using FlowJo (Tree Star, version 10.8.1).

Cell death analysis

Cell death was measured using an IncuCyte imaging system, as described previously35. In brief, cells were imaged in the presence of the cell-impermeable DNA intercalator Sytox Green (Thermo Fisher, R37168), and Sytox-positive cells were quantified at each time point using custom processing definitions, available on request. In parallel, separate cells plated in identical numbers were treated with the cell-permeable dye Syto Green (Thermo Fisher, S34854) and quantified using the same approach, and percentage cell death was calculated as Sytox+/Syto+ at each time point. For siRNA-knockdown cell death assays, to avoid excessive non-specific toxicity, cells were transfected with siRNA for 8 h, at which point cells (adherent) were washed and dye/inhibitors were added at final concentrations immediately before IncuCyte imaging.


Pathology analysis for the AdarP195A/p150null::Zbp1 experiments (Supplementary Fig. 2) was carried out by the same personnel and using a similar scoring system recently described for analysis of AdarP195A/p150null mice5. In brief, littermate mice 21 days of age were euthanized through CO2 asphyxiation and livers and kidneys were collected and washed in PBS and fixed in 10% neutral-buffered formalin. Tissues were embedded in paraffin and cut into sections of about 4 mm in thickness for haematoxylin and eosin staining. Liver and kidney sections were also stained with periodic acid–Schiff (PAS) stain. Slides were evaluated by a board-certified veterinary pathologist, who was blinded to genotype and experimental setup. For kidney, expansion of the glomerular mesangial matrix was scored from 0–4, with 0, normal; 1, minimal; 2, mild; 3, moderate; 4, severe. For the liver, microvesicular and lesser macrovesicular cytoplasmic vacuolation were scored from 0–5, with 0, normal; 1, minimal changes affecting only a small region (<5%) of the liver; 2, mild changes throughout the liver but without enlargement of hepatocytes, coalescing lesions or necrosis; 3, mild to moderate cytoplasmic vacuolation throughout the liver with enlargement of hepatocytes but no necrosis or loss of parenchyma; 4, moderate, coalescing throughout the liver with multifocal mild regions of loss of parenchyma or necrosis; 5, severe with moderate multifocal regions of cavitation and necrosis.

Tissues for the AdarP195A/p150null::Mlkl−/−::Casp8−/− experiments (Supplementary Fig. 11) were collected from pups euthanized by decapitation on the day of birth, and spleen, kidney, liver, heart, head with brain, and gastrointestinal tract were routinely embedded in paraffin and stained with haematoxylin and eosin. PAS stains were also obtained for liver and kidney. These slides were reviewed blindly, with the exception of gastrointestinal tissues, for which the pathologist was not blinded to genotype.

Representative images were captured from scanned slides or from glass slides taken using NIS-Elements BR 3.2 64-bit and image plates were created in Adobe Photoshop. Image white balance, lighting and contrast were adjusted using autocorrections applied to the entire image. Original magnification is stated in the figure captions.

Analysis of IBA1 and cleaved caspase 3 for Mlkl::Casp8::Adarp150null/P195A pups was carried out through the University of Washington Histology and Imaging Core utilizing the Leica Bond Rx Automated Immunostainer (Leica Microsystems). Slides were deparaffinized with Leica Dewax solution at 72 °C for 30 s. Antigen retrieval was carried out on all slides with EDTA, pH 9, at 100 °C for 20 min. All subsequent steps were carried out at room temperature. Initial blocking consisted of 10% normal goat serum (Jackson ImmunoResearch, catalogue number 005-000-121) in Tris-buffered saline for 20 min and further blocking with Leica Bond peroxide block for 5 min. Slides were incubated with IBA1 (1:1,000) or CC3 (1:250) primary antibodies in Leica primary antibody diluent. Slides were scanned in brightfield mode with a 20× objective using a NanoZoomer Digital Pathology System. Quantification of CC3 and IBA1 was carried out using the Visiopharm Image Analysis module.

Quantitative PCR analysis

RNA was isolated from primary MEFs or LET1 cells using Trizol extraction and first-strand cDNA synthesis was carried out with SuperScript III Reverse Transcriptase (Invitrogen, catalogue number 18080044). Quantitative PCR was carried out using a ViiA 7 Real Time PCR System (Thermo Fischer Scientific) using SYBR reagents (Thermo Fisher). The following primers were used for quantitative PCR: Zbp1, sense (S): 5′-AAGAGTCCCCTGCGATTATTTG-3′, antisense (AS): 5′-TCTGGATGGCGTTTGAATTGG-3′; Ripk1, S: 5′-GAAGACAGACCTAGACAGCGG-3′, AS: 5′-CCAGTAGCTTCACCACTCGAC-3′; Ccl2, S: 5′-TGGCTCAGCCAGATGCAGT-3′, AS: 5′-TTGGGATCATCTTGCTGGTG-3′; Ccl7, S: 5′-CCACATGCTGCTATGTCAAGA-3′, AS: 5′-ACACCGACTACTGGTGATCCT-3′; Ifit1, S: 5′-GCCATTCAACTGTCTCCTG-3′, AS: 5′-GCTCTGTCTGTGTCATATACC-3′; Ifnb, S: 5′-CTGGAGCAGCTGAATGGAAAG-3′, AS: 5′-CTTCTCCGTCATCTCCATAGGG-3′; Gapdh, S: 5′-GGCAAATTCAACGGCACAGT-3′, AS: 5′-AGATGGTGATGGGCTTCCC-3′.

RNA-sequencing and NanoString analysis

RNA was isolated from day-23 spleens (for RNA-seq) or treated cells (for NanoString experiments) using Trizol. An on-column DNAse treatment was included for RNA-seq experiments. Total RNA was added directly to lysis buffer from the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Takara), and reverse transcription was carried out followed by PCR amplification to generate full-length amplified cDNA. Sequencing libraries were constructed using the NexteraXT DNA sample preparation kit (Illumina) to generate Illumina-compatible barcoded libraries. Libraries were pooled and quantified using a Qubit Fluorometer (Life Technologies). Dual-index, single-read sequencing of pooled libraries was carried out on a HiSeq2500 sequencer (Illumina) with 58-base reads, using HiSeq v4 Cluster and SBS kits (Illumina) with a target depth of 10 million reads per sample.

For NanoString, 254 transcripts were quantified from total RNA using the mouse nCounter Inflammation V2 panel (NanoString). The nSolver Analysis Software 4.0 with the nCounter Advanced Analyses package (version 2.0.134) was used to normalize the data and perform differential gene expression analysis to generate log2[fold change] values and P values. Data visualizations were carried out in R (version 4.1.1). Differentially expressed genes were visualized as heatmaps and volcano plots using the packages pheatmap (version 1.0.12) and ggplot2 (version 3.3.5).

For RNA-seq analysis of day-23 spleens, reads were aligned using kallisto36 to the mouse reference genome (GRCm39) using default parameters. Quality control was carried out on raw reads using fastqc, and raw reads were then combined with aligned reads using multiqc (ref. 37), with no samples removed from the final dataset owing to quality control checks. Analysis of aligned reads was carried out with R using DESeq2 (ref. 38) using standard parameters to generate differential gene expressions for each of the conditions against the WT, and significant differential expression was defined by adjusted P value < 0.01 and absolute log2[fold change] > 1. The differential expression data were annotated using the bioMaRt package39,40. Fold changes against WT mice for ADAR1-deficient (AdarP195A/p150null) mice, with and without Zbp1-aknockout, were compared, defining genes with a >50% regression to a fold change of 0 after Zbp1-a knockout as high-recovery genes and all others as low-recovery genes. Gene ontology analysis was carried out using the clusterProfiler package41 using standard parameters comparing both high- and low-recovery genes against background separately.

Statistical analysis

Comparison of survival curves was carried out using a log-rank (Mantel–Cox) test. P values less than 0.0001 were a result of the statistical analysis package performed by GraphPad Prism and are represented as P < 0.0001. Data shown in graphs are mean or mean ± s.d. If the data fulfilled the criteria for Gaussian distribution tested by column statistics, an unpaired parametric t-test with Welch’s correction was carried out for statistical analysis. All statistical tests listed in the figure legends were two-sided and were carried out using Graphpad Prism, or Microsoft Excel (chi-square power values for Mendelian distributions). P values are presented in the figures or figure legends, generally, we used the following conventions: NS (P > 0.05), *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. All in vitro experiments, unless otherwise stated, were independently replicated a minimum of two times, and details on replication of displayed data are stated in the figure legends.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.