Sensing of viral and endogenous RNA by ZBP1/DAI induces necroptosis

1Nucleic acids form double‐stranded helices. Double‐stranded (ds) DNA typically adopts the so‐called B‐conformation, while dsRNA is usually in the A‐conformation. Both DNA and RNA can also adopt a Z‐form double helix that is characterised by a left‐handed helical arrangement, a zigzag pattern of the phosphodiester backbone and the absence of major grooves (Wang et al, 1979; Hall et al, 1984; Rich & Zhang, 2003). Antibodies raised against Z‐form RNA or DNA stain mammalian cells and chromosomes, respectively, suggesting that nucleic acids in this unusual conformation occur naturally in cells (Viegas‐Péquignot et al, 1983; Zarling et al, 1990). However, biological functions of the Z‐conformation are only partially understood (Rich & Zhang, 2003; Wang & Vasquez, 2007). Z‐DNA is believed to play a role in transcription by relieving torsional strain induced within the DNA template by the movement of RNA polymerases, and Z‐DNA may also promote genetic instability (Rich & Zhang, 2003; Wang & Vasquez, 2007). Physiological functions of Z‐RNA remain unknown.

2Early biochemical and structural studies identified a protein fold called Z‐binding domain (ZBD) that specifically binds to Z‐DNA but not to B‐DNA (Herbert et al, 1993; Schwartz et al, 1999). This 78‐amino‐acid domain is also referred to as Zα or Zβ, and subsequent work showed that it not only binds to Z‐DNA but also to Z‐RNA (Brown et al, 2000; Placido et al, 2007). ZBDs have been identified in four proteins: ADAR1, ZBP1, E3L and PKZ (Athanasiadis, 2012). ADAR1 and ZBP1 are mammalian proteins implicated in antiviral innate immune responses. E3L is a poxvirus protein known to antagonise this host response. Lastly, PKZ is a fish protein related to mammalian PKR, also involved in antiviral immunity. This suggests an important role of ZBDs in innate antiviral immune responses (Athanasiadis, 2012) and may imply that Z‐DNA and/or Z‐RNA trigger such a host defence response. Z‐form nucleic acids are further implicated in immune responses by the observation that sera from some systemic lupus erythematosus patients contain anti‐Z‐DNA autoantibodies (Lafer et al, 1983).

3Z‐DNA binding protein 1 (ZBP1, also known as DAI or DLM‐1) has two N‐terminal ZBDs (Zα1 and Zα2) and two RIP homotypic interaction motifs (RHIMs) (Ha et al, 2008; Kaiser et al, 2008; Rebsamen et al, 2009; Kim et al, 2011a,b; Yang et al, 2014). ZBP1 had initially been suggested to be a cytosolic B‐DNA sensor that induces type I interferons (IFNs) (Takaoka et al, 2007). However, this observation could not be confirmed in ZBP1‐deficient mice (Ishii et al, 2008). Instead, ZBP1 activates NF‐κB and triggers necroptosis (Kaiser et al, 2008; Rebsamen et al, 2009; Upton et al, 2012), an inflammatory form of programmed cell death (Wallach et al, 2016). NF‐κB activation is mediated by RIPK1 and RIPK3 kinases, which bind to ZBP1 via RHIM‐RHIM interactions (Kaiser et al, 2008; Rebsamen et al, 2009). In contrast, ZBP1‐dependent necroptosis depends on RIPK3 recruitment and does not require RIPK1 (Upton et al, 2012). RIPK3 phosphorylates and activates the pseudokinase MLKL, which upon oligomerisation ruptures the plasma membrane, leading to cell death (Wallach et al, 2016).

4Necroptosis mediates host defence by eliminating virus‐infected cells (Mocarski et al, 2015). Viruses in turn suppress necroptosis. A well‐studied example is mouse cytomegalovirus (MCMV) infection (Mocarski et al, 2015): MCMV prevents ZBP1‐ and RIPK3‐dependent necroptosis by expression of M45 (Upton et al, 2010, 2012). M45 has a RHIM that disrupts the interaction of ZBP1 and RIPK3. MCMV expressing M45 with a mutated RHIM fails to inhibit necroptosis, resulting in death of infected cells (Upton et al, 2012). Necroptotic cell death is normally held in check by the proteolytic activity of the pro‐apoptotic caspase‐8. Cells can be sensitised to RIPK3‐dependent necroptosis by genetic loss (Kaiser et al, 2011; Oberst et al, 2011) or chemical blockade of caspase‐8 (Vandenabeele et al, 2006). During M45‐RHIM‐mutant MCMV infection, the viral M36 protein—also known as viral inhibitor of caspase‐8‐induced apoptosis (vICA)—is believed to predispose cells to necroptosis by blocking caspase‐8 (McCormick et al, 2003; Ménard et al, 2003; Guo et al, 2015).

5Here, we show that necroptosis triggered by MCMV‐M45mutRHIM or by ectopic ZBP1 expression and caspase blockade required binding of nucleic acids to the tandem ZBDs of ZBP1. We generated Zbp1 knock‐in mice carrying four amino acid substitutions that abrogate binding to Z‐form nucleic acids. Cells from Zbp1‐mutant mice are resistant to MCMV‐M45mutRHIM‐induced necroptosis. During infection, inhibition of RNA transcription but not of viral DNA replication prevented ZBP1‐dependent necroptosis and we found that ZBP1 bound to newly synthesised RNA. Furthermore, ZBP1 cross‐linked to endogenous RNA and stained cells in an RNase‐sensitive manner. In sum, these results show that ZBP1 is an RNA sensor.

6We initially wished to confirm the role of ZBP1 in cell death upon virus infection. Expression of ZBP1 was induced by IFN in Zbp1 +/− but not in Zbp1 −/− primary mouse embryonic fibroblasts (MEFs; Fig 1A). We then assessed cell viability upon virus infection by determining intracellular ATP levels using CellTiter‐Glo reagent. Infection with MCMV‐M45mutRHIM, a virus that fails to inhibit ZBP1‐RIPK3‐dependent necroptosis (Upton et al, 2012), induced cell death in Zbp1 +/− primary MEFs following IFN pre‐treatment (Fig 1B). In contrast, viability upon MCMV‐M45mutRHIM infection was much higher in IFN‐treated Zbp1 −/− primary MEFs (Fig 1B), despite comparable expression of RIPK3 and MLKL (Fig 1A). We also infected primary MEFs with MCMV‐M45wt, which did not cause much cell death in either Zbp1 +/− or Zbp1 −/− cells (Fig 1B). As a control, the induction of necroptosis by TNF stimulation in the presence of the caspase inhibitor zVAD (Vercammen et al, 1998) (hereafter simply TZ) was independent of ZBP1 (Fig 1B). Similar results were obtained in immortalised MEFs (Fig EV1A and B). These observations are consistent with an earlier report (Upton et al, 2012) and demonstrate that death of MCMV‐infected cells required ZBP1 and was antagonised by the viral M45 protein.

7To test if binding of Z‐form nucleic acids to ZBP1 is involved in the induction of cell death during virus infection, we mutated key conserved residues involved in Z‐RNA/DNA binding in mouse ZBP1. Based on the structures of the ZBP1 Zα1 and Zα2 domains (Schwartz et al, 2001; Ha et al, 2008; Kim et al, 2011a,b; Yang et al, 2014) and on sequence alignments (Fig 1C and D), we introduced N46A and Y50A mutations into Zα1 (Zα1mut), N122A and Y126A substitutions into Zα2 (Zα2mut), and all four mutations together (Zα1α2mut). Mutation of the corresponding residues in ADAR1's Zα ZBD completely abolishes Z‐DNA binding without altering protein stability (Schade et al, 1999); accordingly, we predicted diminished binding of our ZBP1 mutants to Z‐RNA/DNA. Wild‐type and Zα1α2mut ZBP1 proteins were equally able to induce an NF‐κB reporter upon ectopic expression together with RIPK3 in HEK293T cells, suggesting that the mutant protein folds correctly and retains signalling via its RHIMs (Figs 1E and EV1C).

8We reconstituted immortalised Zbp1 −/− MEFs with ZBP1 3xFLAG‐tagged ZBD mutants (Fig 1F). Cell viability following TZ treatment or MCMV‐M45wt infection was comparable between parental Zbp1 −/− and ZBP1‐reconstituted cells (Fig 1G). MCMV‐M45mutRHIM infection induced pronounced cell death in ZBP1 wild‐type reconstituted cells (Fig 1G and H). In contrast, ZBP1‐Zα1α2mut‐expressing cells displayed higher viability after infection and behaved similar to the parental Zbp1 −/− cells, while ZBP1‐Zα1mut and ZBP1‐Zα2mut cells showed intermediate phenotypes (Fig 1G and H), with a greater contribution to cell death of the Zα2 domain. Cell death induced by MCMV‐M45mutRHIM in wild‐type ZBP1‐expressing cells was also observed by live‐cell microscopy using the dye SytoxGreen, which stains cells that have lost membrane integrity, and cell death was detected from 8 h post‐infection onwards (Fig 1I). We then used mouse NIH3T3 fibroblasts, a transformed cell line. Endogenous ZBP1 expression was undetectable by Western blot, but could be induced by IFN treatment (Fig EV1D). We stably transduced NIH3T3 cells to express either 3xFLAG‐tagged wild‐type or mutant ZBP1 (Fig EV1D and F). Cell viability was comparable between these cells after MCMV‐M45wt infection (Fig EV1E) and upon TZ treatment (Fig EV1G and H). Consistent with our results in MEFs, cell viability was decreased in wild‐type ZBP1 but not in ZBP1‐Zα1α2mut‐expressing cells following infection with MCMV‐M45mutRHIM (Fig EV1E–H). Taken together, these observations suggest that induction of cell death by ZBP1 requires intact ZBDs and may imply that nucleic acids in Z‐conformation trigger this response.

9We next wished to test whether restriction of MCMV infection by ZBP1 depends on the protein's ZBDs. Infection of cells with MCMV results in lytic cell death after several days of culture due to virus replication. Expression of wild‐type ZBP1 but not ZBP1‐Zα1α2mut protected Zbp1 −/− MEFs or NIH3T3 cells against MCMV‐M45mutRHIM‐induced cytopathic effects 5 days post‐infection (Figs 2A and EV2A). Furthermore, replication of the virus genome was blunted by wild‐type but not by mutant ZBP1 (Fig 2B). These observations show that ZBP1 defended cells against MCMV and that intact ZBP1 ZBDs were required for this.

10This block in virus replication may depend on ZBP1‐dependent cell death or engagement of an IFN response in infected cells. We first explored how mutations in ZBP1's ZBDs affected MCMV‐M45mutRHIM‐induced cell death. Previous work suggested that the type of cell death triggered by ZBP1 upon MCMV‐M45mutRHIM infection is necroptosis (Upton et al, 2012). Upon phosphorylation by RIPK3, MLKL forms trimers and large multimeric complexes, which execute necroptosis by causing plasma membrane permeabilisation (Wallach et al, 2016). Phosphorylation of MLKL was only observed in MCMV‐M45mutRHIM‐infected NIH3T3 cells reconstituted with wild‐type ZBP1 and not in ZBP1‐Zα1α2mut ZBP1‐expressing cells (Fig 2C). Concomitantly, reduced levels of the viral IE1 protein accumulated in cells expressing wild‐type ZBP1 (Fig 2C). Furthermore, we observed the formation of MLKL‐trimers (Cai et al, 2014) in non‐reducing SDS–PAGE gels after induction of necroptosis in NIH3T3 cells with TZ (Fig EV2B). Large MLKL oligomers, which did not migrate through the gel, were also observed in the gel‐loading pocket (Dondelinger et al, 2014; Fig EV2B). MCMV‐M45wt infection did not change MLKL electrophoretic mobility (Fig EV2B), which is in line with our earlier observation that this virus does not cause cell death (Fig 1). MCMV‐M45mutRHIM infection induced MLKL oligomerisation at 16 h post‐infection (Figs 2C and EV2B). These effects were enhanced and accelerated in cells expressing wild‐type ZBP1 but not in cells reconstituted with ZBP1‐Zα1α2mut (Fig EV2B). Furthermore, the viability of MCMV‐M45mutRHIM‐infected NIH3T3 cells expressing wild‐type ZBP1 was rescued upon treatment with the RIPK3 kinase inhibitor GSK'872 (Fig 2D). As described previously (Upton et al, 2010), blocking RIPK1 activity with Nec‐1 had no effect on virus‐induced cell death (Fig 2D). These observations show that ZBP1 indeed activates necroptosis in this setting.

11MCMV‐M45mutRHIM infection also induced an IFN response. Expression of the interferon‐stimulated genes (ISGs) Ifit1, Ifi44 and Oasl1 was increased upon infection and was not altered in cells expressing ZBP1 (Fig 2E). Consistent with this observation, the expression and secretion of CXCL10, a chemokine that signifies the induction of an IFN response, were independent of ZBP1 or MCMV M45 protein (Figs 2F and EV2C). Instead, CXCL10 induction was reduced to background levels in Tmem173 −/− cells that lack STING, a protein involved in cytosolic DNA sensing (Figs 2G and EV2C; Chen et al, 2016). Furthermore, overexpression of ZBP1 did not induce an IFNβ reporter in HEK293T cells (Fig EV2D). In sum, these results demonstrate that ZBP1 and its ZBDs were required specifically for the induction of necroptosis upon virus infection and not for the induction of an IFN response.

12To substantiate our conclusion that intact ZBDs are required for ZBP1‐dependent necroptosis, we developed a knock‐in mouse in which the endogenous Zbp1 allele is replaced by Zbp1‐Zα1α2 mut (Zbp1 Zα1α2) carrying the four point mutations shown in Fig 1C. The generation and validation of this model are detailed in Fig EV3A and B. We obtained primary MEFs by crossing of Zbp1 +/Zα1α2 animals. Similar levels of Zbp1 mRNA and ZBP1 protein were expressed at baseline and after IFN induction in cells expressing only wild‐type ZBP1 (Zbp1 +/+), only mutant ZBP1 (Zbp1 Zα1α2/Zα1α2) or both (Zbp1 +/Zα1α2) (Figs 3A and EV3C). Consistent with our experiments based on ectopic ZBP1 expression (Figs 1 and 2), intact ZBDs of endogenous ZBP1 were required for cell death induced by MCMV‐M45mutRHIM (Fig 3B and C). Cell death induced by TZ treatment was not affected by mutations in Zbp1 (Figs 3B and EV3D). Furthermore, the levels of phosphorylated MLKL and MLKL oligomerisation were reduced in Zbp1 Zα1α2/Zα1α2 primary MEFs upon MCMV‐M45mutRHIM infection (Figs 3D and EV3E), but not after TZ treatment (Fig 3E). Finally, virus growth and accumulation of the viral IE1 protein were enhanced in Zbp1 Zα1α2/Zα1α2 primary MEFs (Fig 3D and F). To test whether intact ZBDs are required to restrict virus replication in vivo, we infected Zbp1 Zα1α2/Zα1α2 knock‐in mice with MCMV‐M45mutRHIM. After 5 days, we were able to recover infectious virus from the spleens of eight of 13 infected Zbp1 Zα1α2/Zα1α2 animals, while the spleens of all wild‐type and heterozygous mice remained free of virus (Fig 3G). As expected, no differences in splenic virus titres were observed between the genotypes when mice were infected with wild‐type MCMV (Fig 3G). These observations provide further evidence that recognition of nucleic acids, potentially in Z‐conformation, by ZBP1 is required for the induction of necroptosis and virus restriction.

13To define the molecular requirements of ZBP1 and necroptosis activation in MCMV‐infected cells, we co‐cultured NIH3T3 cells expressing wild‐type and Zα1α2mut ZBP1 at a 2:1 ratio. The latter cells were labelled with a dye prior to co‐culture. We infected co‐cultures with MCMV‐M45wt or MCMV‐M45mutRHIM and, after 16 h, determined the ratio of wild‐type ZBP1 and ZBP1‐Zα1α2mut cells by flow cytometry (Fig 4A). Infection with MCMV‐M45wt did not affect the ratio of these cells; however, MCMV‐M45mutRHIM infection resulted in selective loss of wild‐type ZBP1‐expressing cells (Fig 4A). This observation shows that ZBP1‐dependent cell death is cell autonomous and further indicates that infection generates an intracellular ZBP1 agonist, conceivably a nucleic acid in Z‐conformation.

14To determine whether viral DNA replication was required for ZBP1 activation, we treated infected cells with 50 μM of the nucleotide analogue ganciclovir (Smee et al, 1995). Although this effectively prevented MCMV replication in NIH3T3 cells and immortalised MEFs (Fig EV4A), virus‐induced cell death in wild‐type ZBP1‐reconstituted cells remained unaffected (Fig 4B). UV inactivation of MCMV‐M45mutRHIM abrogated viral replication and expression of viral genes (Figs 4C, and EV4A and B). At the same time, CXCL10 protein and ISG mRNA induction, which are dependent on STING (Figs 2G and EV2C), remained normal (Figs 4C and EV4B). Importantly, UV‐treated MCMV did not induce cell death (Fig 4B). Together with the result using the DNA replication inhibitor ganciclovir, this suggested that viral RNA transcription was required for ZBP1‐dependent necroptosis. Indeed, inhibition of de novo RNA synthesis using actinomycin D (ActD) prevented cell death induced by MCMV‐M45mutRHIM in NIH3T3 cells expressing wild‐type ZBP1 (Fig 4D). Similarly, ActD but not the translation inhibitor cycloheximide (CHX) blocked MCMV‐M45mutRHIM induced cell death in primary MEFs (Fig 4E). We also interfered with expression of the viral immediate early protein IE3 that is required for expression of early and late genes (Angulo et al, 2000). siRNA‐mediated depletion of IE3 was confirmed using MCMV‐IE1/3‐GFP infection (Martínez et al, 2010; Fig EV4C and D). IE3 depletion led to diminished cell death upon infection of NIH3T3 cells expressing wild‐type ZBP1 with MCMV‐M45mutRHIM (Fig 4F). MCMV‐M45mutRHIM infection predisposes NIH3T3 cells to TNF‐induced cell death independent of ZBP1 expression (Fig EV4E). This predisposition is likely due to expression of the MCMV protein M36, which blocks caspase‐8 and in combination with TNF treatment triggers necroptosis (Guo et al, 2015; Omoto et al, 2015). IE3 depletion did not influence cell death triggered by MCMV‐M45mutRHIM and TNF, indicating that IE3 depletion did not change sensitisation to necroptosis (Fig EV4F). MCMV induces a potent IFN response (Fig 2E and F), and IFN‐induced endogenous transcripts could contribute to ZBP1 activation. To test this, we cultured wild‐type ZBP1‐reconstituted MEFs in the presence of neutralising antibodies against the type I IFN receptor (anti‐IFNAR1). Blockade of IFNAR1 signalling did not affect MCMV‐M45mutRHIM‐induced cell death, while the expression of the ISGs Ifit1 and Ifi44, but not of Ifnb, was reduced upon anti‐IFNAR1 treatment (Fig EV4G and H). Taken together, these data show that MCMV RNA transcripts generated in infected cells contribute to the activation of ZBP1 and necroptosis.

15To further substantiate this conclusion, we employed the PAR‐CLIP protocol (Fig 4G; Hafner et al, 2010). In brief, NIH3T3 cells expressing ZBP1‐3xFLAG were pulsed with photo‐activatable nucleoside analogues (4SU or 6SG) and were then infected with MCMV‐M45mutRHIM. Eight hours post‐infection, cells were treated with UV‐A light (365 nm), which only cross‐links newly synthesised RNA that contains 4SU or 6SG to proteins. Cells were lysed and treated with RNase A. In this setting, small fragments of RNA are protected from degradation by cross‐linking to proteins. ZBP1 was then immunoprecipitated using α‐FLAG antibody and stringent conditions (see Materials and Methods). The RNA‐ZBP1 complex was visualised by 32P labelling of the RNA, followed by gel electrophoresis, blotting and autoradiography (Fig 4G). Indeed, we observed bands corresponding to the molecular weight of ZBP1 (60 kDa) when MCMV‐infected cells were treated with 4SU or 6SG (Fig 4G). To investigate whether cross‐linking of newly synthetised RNA required intact ZBDs, NIH3T3 cells expressing wild‐type or mutant ZBP1 were infected with MCMV‐M45WT in the presence of 6SG. Wild‐type M45‐expressing virus was used to prevent loss of wild‐type ZBP1‐expressing cells due to necroptosis. After cross‐linking, we observed enhanced ZBP1‐RNA complex formation with wild‐type ZBP1 compared to mutant ZBP1, confirming that ZBDs are important for RNA binding (Fig 4H). These finding supports the notion that ZBP1 binds newly synthesised RNA in infected cells.

16Surprisingly, we observed that expression of ZBP1 in NIH3T3 cells triggered spontaneous cell death upon caspase inhibition with zVAD (Figs 5A and EV5A). zVAD had a similar, albeit less pronounced, effect in immortalised MEFs (Fig 5A). Cell death in zVAD‐treated cells expressing ZBP1 required intact ZBDs (Figs 5A and EV5A). Phosphorylation of MLKL at 16 h post‐treatment confirmed that cells died by necroptosis (Fig EV5B). Furthermore, treatment with the RIPK3 inhibitor GSK'872 prevented cell death in zVAD‐treated cells (Fig 5B). Viability was also restored by Nec‐1 treatment, suggesting that spontaneous ZBP1‐dependent cell death induced by caspase inhibition additionally requires the activity of RIPK1 (Fig 5B).

17Next, using a polyclonal cell population, we generated NIH3T3 clones expressing wild‐type ZBP1 at different levels by limiting dilution (Figs 5C, and EV5C and D). Expression of ZBP1 at both low and high levels sensitised cells to killing by MCMV‐M45mutRHIM (Fig 5D). In contrast, cell death triggered by ectopic ZBP1 and zVAD required high levels of expression (Fig 5D). To further explore this observation, we transduced cells expressing wild‐type or Zα1α2mut ZBP1 with M36 that blocks caspase‐8‐mediated apoptosis. This led to selective loss of cells expressing wild‐type ZBP1 and M36 (Fig 5E). Western blot analysis further indicated that expression of wild‐type ZBP1 and M36 was mutually exclusive, indicating death of cells expressing these two proteins together (Fig 5F). Taken together, these observations suggested that cells contain endogenous nucleic acids that engage ZBP1 if present at sufficiently high levels. This then leads to necroptosis in conditions where caspase‐8 activity is blocked.

18To further explore the observation that ZBP1 triggers necroptosis in response to endogenous nucleic acids, we used recombinant ZBP1 protein as a staining reagent in uninfected cells. We expressed 3xFLAG‐tagged ZBP1 and ZBP1‐Zα1α2mut in HEK293T cells and purified the proteins by immunoprecipitation (Fig EV6A). We then incubated uninfected, fixed and permeabilised HEK293T cells with purified ZBP1 protein, followed by staining with a secondary α‐FLAG antibody. Cells were analysed by flow cytometry. Recombinant wild‐type ZBP1 stained cells, whereas ZBP1‐Zα1α2mut showed only background staining (Fig 6A, left panel). Interestingly, the labelling of cells with wild‐type ZBP1 was sensitive to RNase but not to DNase treatment (Fig 6A, middle and right panel). These observations suggest that ZBP1, via its ZBDs, binds endogenous RNA, potentially in the Z‐conformation.

19Next, we used cross‐link immunoprecipitation (CLIP, Fig 6B) to further explore the binding of endogenous nucleic acids to ZBP1 (Darnell, 2010). In brief, we cross‐linked nucleic acids and proteins using UV‐C light (254 nm) in uninfected NIH3T3 cells expressing ectopic ZBP1‐3xFLAG. Cell extracts were then treated with DNase and/or RNase. This degrades free nucleic acids leaving intact fragments of DNA or RNA cross‐linked to proteins. Next, we immunoprecipitated ZBP1 using α‐FLAG antibodies (Fig EV6B). Nucleic acids present in the precipitate were then end‐labelled using 32P, followed by Western blot and autoradiography (Fig 6B). Using high amounts of RNase A, we observed a major band in the autoradiogram corresponding in size to ZBP1 (sample 3 in Fig 6B). In samples without RNase A, this band was not visible; instead, a smear of higher molecular weight was apparent (sample 1). These patterns were not changed in samples containing DNase I (samples 2 and 4). These findings suggest that ZBP1 binds RNA and not DNA. At the high RNase A concentrations used here, a very short stretch of RNA is likely to be protected by ZBP1, explaining the sharp band in lane 3 in Fig 6B. In the absence of RNase, this band shifted to a slower migrating smear, indicating the presence of longer RNA fragments bound to ZBP1. Indeed, when we used a range of decreasing RNase A concentrations, we observed a concomitant increase in the amount of 32P‐labelled material migrating more slowly (Figs 6C and EV6C). Importantly, this shift was markedly reduced when we used uninfected cells expressing ZBP1‐Zα1α2mut despite equal efficiency of protein precipitation (Figs 6D and EV6D). Taken together, these data show that ZBP1 binds endogenous RNA but not DNA in uninfected cells.

20Mammalian cells autonomously detect virus infection (Pichlmair & Reis e Sousa, 2007). Pattern recognition receptors (PRRs) sense virus invasion and recognise pathogen‐ or danger‐associated molecular patterns (Medzhitov, 2009). Upon activation, these receptors induce immune responses, including the production of cytokines such as IFNs, chemokines and inflammatory mediators, as well as the triggering of programmed cell death (Pichlmair & Reis e Sousa, 2007; Mocarski et al, 2012, 2015; Lupfer et al, 2015). In many instances, nucleic acids such as viral RNA or DNA genomes trigger PRRs (Goubau et al, 2013). This raises an important question: How can PRRs distinguish between the RNAs and DNAs abundantly found in healthy, uninfected cells and nucleic acids introduced during infection? Several “solutions” to this problem have been described and include detection of nucleic acids in cellular compartments usually devoid of RNA or DNA, detection of covalent chemical modifications unique to viral RNAs and DNAs, and detection of specific sequences (Goubau et al, 2013; Roers et al, 2016).

21Here, we propose that RNA in the unusual Z‐conformation is a molecular pattern recognised by ZBP1. Indeed, we found that intact ZBDs were required for ZBP1‐dependent induction of necroptosis in MCMV‐infected cells. Several lines of evidence showed that ZBP1 recognises RNA, and not DNA, and that this RNA likely corresponds to viral transcripts newly synthesised after infection. Firstly, the induction of cell death could not be blocked by an inhibitor of the viral DNA polymerase (GCV). Secondly, UV‐inactivated virus that failed to generate viral RNA transcripts also did not induce necroptosis. Thirdly, necroptosis induction was sensitive to ActD, which potently inhibits RNA transcription. Fourthly, knock‐down of the viral IE3 gene—that drives expression of many other viral genes—reduced cell death. Lastly, PAR‐CLIP analysis showed that ZBP1 cross‐linked with newly synthesised RNA in infected cells.

22Two recent studies published while this manuscript was in preparation suggest that ZBP1 senses influenza A virus (IAV) infection and induces cell death of IAV‐infected cells (Kuriakose et al, 2016; Thapa et al, 2016). Our results confirm these two reports in a different infection model and resolve an important discordance: Thapa et al reported that ZBP1 senses IAV RNA (Thapa et al, 2016) while Kuriakose et al concluded that IAV proteins, particularly PB1 and NP, trigger ZBP1 (Kuriakose et al, 2016). Our results strongly suggest that ZBP1 is an RNA sensor. We used stringent cross‐link and immunoprecipitation protocols and provide a detailed analysis of ZBP1 mutants unable to bind Z nucleic acids, including a knock‐in model. It is conceivable that ZBP1 binds IAV PB1 and NP indirectly via viral RNA as both IAV proteins tightly associate with IAV RNA genomes and transcripts (Eisfeld et al, 2015). In the future, it will be interesting to infect ZBP1‐Zα1α2mut knock‐in mice with IAV to analyse the role of Z‐RNA sensing and necroptosis in in vivo settings. Moreover, it will be worthwhile to explore other viral infections. Indeed, necroptosis is also important during infection with human cytomegalovirus (HCMV) and herpes simplex viruses (HSV) 1 and 2 (Mocarski et al, 2015) and, in HIV‐1 infection, necroptosis of infected T cells may contribute to pathogenesis (Gaiha et al, 2014; Pan et al, 2014).

23Surprisingly, we found that cells expressing high levels of ZBP1 undergo necroptosis when caspase‐8 is blocked and that ZBP1 cross‐links to endogenous RNA in uninfected cells. This suggests a model in which ZBP1 continuously senses self‐RNA, but only engages the necroptosis machinery when sufficient levels of downstream signalling components such as RIPK3 become available, as is the case when caspase‐8 is blocked. Indeed, loss of caspase‐8 activity predisposes to RIPK3‐dependent necroptosis (Kaiser et al, 2011; Oberst et al, 2011). Interestingly, ZBP1‐dependent cell death in this setting required the activity of both RIPK1 and RIPK3. This is in contrast to MCMV‐induced necroptosis, which occurs independently of RIPK1 (Upton et al, 2010). This may be important during development and in sterile inflammation. RIPK1‐deficient mice display severe developmental defects and die around birth, and genetic ablation of RIPK1 in the skin leads to inflammatory phenotypes (Silke et al, 2015). These observations have been linked to excessive RIPK3‐MLKL‐driven necroptosis (Silke et al, 2015). Two recent manuscripts showed that the defects of germline or keratinocyte‐specific RIPK1‐deficient mice are rescued by additional knock‐out of ZBP1 (Lin et al, 2016; Newton et al, 2016). It is tempting to speculate that ZBP1 engages an endogenous Z‐RNA in these settings and activates RIPK3 if inhibition by RIPK1 is lost. Our ZBP1‐Zα1α2mut knock‐in model will be an important tool to address this question.

24It will also be interesting to extend our CLIP analysis by deep sequencing of ZBP1‐associated nucleic acids. It is possible that ZBP1 binds GC‐rich sequences given that GC‐rich nucleic acids adopt the Z‐conformation in the test tube under high salt conditions (Wang et al, 1979; Hall et al, 1984). It is noteworthy that in cells Z‐RNA may not exist as a pre‐formed structure but that this conformation is induced upon binding of ZBP1 and other ZBD‐containing proteins to nucleic acids that are prone to adopt this structure, for example, due to their sequence content (Kim et al, 2011a). We would also like to point out that the RNA recognised by ZBP1 may correspond to both double‐stranded and single‐stranded molecules. The former may be the case in MCMV‐infected cells, given that herpesviruses produce overlapping transcripts from both DNA strands (Rajcáni et al, 2004; Marcinowski et al, 2012; Juranic Lisnic et al, 2013). The latter may be the case when endogenous RNA is sensed; for example, ribosomal RNAs fold back and form base‐paired hairpin structures that may be recognised by ZBDs (Feng et al, 2011).

25Taken together, we demonstrate that binding of viral RNA to ZBP1 triggers necroptosis in infected cells. Whether these RNA molecules are indeed arranged in the Z‐conformation requires further experimental testing. Additionally, endogenous nucleic acids may also adopt this unusual conformation and engage ZBP1. These data indicate that Z‐RNA may be molecular pattern and that ZBP1 functions as its receptor, which induces inflammatory cell death.

26All mice were on the C57Bl/6 background. Tmem173 −/− (STING‐deficient) mice were a gift from J Cambier (Jin et al, 2011). The Zbp1 gene is located on chromosome 2 (GenBank NM_021394.2; Ensembl ENSMUSG00000027514). ZBP1‐Zα1α2mut transgenic mice were generated by Cyagen (see Fig EV3A) and have been designated as Zbp1 < tm1.1Jreh> (MGI:5775152). See Appendix Supplementary Methods for further details. This work was performed in accordance with the UK Animals (Scientific Procedures) Act 1986 and institutional guidelines for animal care. This work was approved by a project license granted by the UK Home Office (PPL No. 40/3583) and was also approved by the Institutional Animal Ethics Committee Review Board at the University of Oxford.

27TLA HEK293T (Open Biosystems), NIH3T3 cells (kind gift from C. Reis e Sousa), and immortalised MEFs (generated in this study) were grown in DMEM (Sigma‐Aldrich) supplemented with 10% heat‐inactivated FCS and 2 mM l‐Gln. Cells were grown at 37°C and 5% CO2. Primary MEFs (kind gift from K. Ishii or generated in this study) were grown in DMEM 10% FCS, 2 mM l‐Gln, 100 U/ml penicillin and 0.1 mg/ml streptomycin under low oxygen conditions (3% O2) at 37°C and 5% CO2. The absence of mycoplasma contamination was confirmed monthly for all cells.

28Mouse Zbp1 was cloned from pCMV‐SPORT6‐Zbp1 (PlasmID clone: MmCD00314236, Harvard Medical School). A 3xFLAG‐tag was added C‐terminally to ZBP1 and TOPO‐TA cloned into the pCR8/GW/TOPO entry vector (ThermoFisher). Zα1 (N46A, Y50A) and Zα2 (N122A, Y126A) mutant ZBP1 were cloned using overlap PCR. MCMV M36 was cloned from cDNA isolated from MCMV‐infected NIH3T3 cells. ZBP1 and M36 were then Gateway cloned into pLenti6.3/TO/V5‐DEST or pcDNA3.2/V5‐DEST using the LR clonase II enzyme mix (ThermoFisher). The oligos used for cloning are listed in Appendix Table S1. The plasmids for human MDA5 (pcDNA3‐3xFLAG‐hMDA5), the IFNβ firefly luciferase reporter (p125Luc) and Renilla luciferase (Rluc) plasmids have been described before (Rehwinkel et al, 2010). The mouse RIPK3 expression plasmid (pCMV‐SPORT6‐mRIPK3) and the NF‐κB firefly luciferase reporter plasmid (pConLuc) were a gift from M. Bertrand and G. van Loo, respectively. The expression vector for mouse STING (pcDNA3‐HA‐mSTING) was a kind gift from R. Vance (Burdette et al, 2011). pGEMTeasy‐MCMV‐gB was a kind gift from E. Beuken. IFN‐A/D was purchased from Sigma‐Aldrich or R&D Systems. Recombinant mouse TNF was from PeproTech. zVAD (Z‐VAD‐FMK) was from Enzo Life Sciences. Actinomycin D was from Life Technologies and CHX was from Sigma‐Aldrich. Nec‐1 was from Enzo Life Sciences, and GSK'872 was from Calbiochem. Neutralising anti‐IFNAR1 (clone MAR1‐5A3) was from Biolegend.

29MCMV‐M45WT (K181 strain) and MCMV‐M45mutRHIM were a kind gift from E. Mocarski (Upton et al, 2010). MCMV was propagated in NIH3T3 cells or immortalised Zbp1 −/− MEFs. See Appendix Supplementary Methods for further details. Lentivirus for ZBP1 and M36 reconstitution was produced in TLA HEK293T cells using pLenti6.3/TO/V5‐DEST transducing vectors and the Virapower packaging mix (ThermoFisher).

30A total of 5,000 NIH3T3 cells or MEFs were seeded per well in 96‐well plates. Twenty‐four hours later, cells were infected with MCMV or treated with TNF and/or zVAD. Endpoint viability (generally 16 h post‐infection/treatment) was determined by CellTiter‐Glo (Promega) using a luminometer (GloMax, Promega). For kinetic analysis of cell death, an in‐incubator imaging platform was used (IncuCyte ZOOM, Essen Bioscience; 10× objective, 1 scan per well in 96‐well plates per hour; samples were imaged in triplicate wells). To measure viability, the cell‐impermeant dyes YOYO‐3 Iodide (0.25 μM) or Sytox Green (25 nM) (ThermoFisher) were added to the medium prior to imaging. At the end of the experiment, cells in control wells were lysed with 0.0625% Triton X‐100 and total number of dye‐positive cells were set to 100%. For NIH3T3 co‐culture experiments, mutant ZBP1‐expressing cells were labelled with 5 μM CellTrace Violet (Molecular Probes) for 20 min at 37°C. Cells were washed with medium and plated together with non‐labelled wild‐type ZBP1‐expressing cells in 24 wells. Cells were then infected with MCMV. Sixteen hours post‐infection, cells were harvested and the ratio of labelled vs. non‐labelled cells was analysed by flow cytometry.

31A total of 150,000 TLA HEK293T cells were seeded in 24‐well plates. Cells were transfected with NF‐κB (50 ng) or IFNβ firefly (125 ng) and 25 ng of Renilla luciferase reporter constructs, and the indicated amounts of expression vectors using FuGENE 6 (Promega). Luciferase activity was measured 24 h after transfection using a luminometer (GloMax, Promega) and Dual‐Glo reagents (Promega). CXCL10 ELISA was from eBioscience.

32Cells were lysed in RLT buffer and homogenised using a Qiashredder (Qiagen), and RNA was purified using RNeasy columns with DNase I digestion (Qiagen). cDNA synthesis was performed with SuperScript II reverse transcriptase (Thermo Fisher) and oligo(dT) primers (Ambion). For DNA extraction, cells were lysed in AL‐buffer and incubated with proteinase K at 56°C for 10 min, and DNA was purified using DNeasy blood and tissue columns (Qiagen). 15 ng DNA was amplified using TaqMan universal PCR (Applied Biosystems) or EXPRESS SYBR GreenER (ThermoFisher) master mixes on a 7,800 real‐time PCR system (Applied Biosystems). mRNA expression data were normalised to GAPDH and analysed by the comparative Ct method. A plasmid carrying a fragment of the MCMV gB gene (pGEMTeasy‐MCMV‐gB) was used as a standard for determining the viral genome copy numbers (Vliegen et al, 2003). The lower detection limit was 30 molecules. The qPCR probes and primers used in this study are detailed in Appendix Table S2.

33Cells were lysed in 25 mM Tris.HCl pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% Igepal CA‐630 (Sigma‐Aldrich) and 5% glycerol, protease inhibitors (Cell Signalling) and phosphatase inhibitor cocktail 3 (Sigma). Protein lysates were cleared by centrifugation at 16,000 g for 10 min. Alternatively, cells were directly lysed in NuPAGE SDS–PAGE sample loading buffer (ThermoFisher). 50 mM of the reducing agent DTT was added to the lysates, except for non‐reducing gels. Protein lysates were run on Bis‐Tris SDS–PAGE gels (NuPAGE, ThermoFisher). Primary antibodies used for Western blotting are listed in Appendix Table S3. Anti‐rabbit, anti‐rat or anti‐mouse secondary antibodies coupled to HRP were from GE Healthcare.

34For intracellular ZBP1 staining, cells were fixed and permeabilised in Cytofix/Cytoperm (BD biosciences) according to manufacturer's instructions. When recombinant ZBP1 was used (Fig 6A), cells were incubated with 0.3 μg ZBP1‐3xFLAG for 1 h in Perm/Wash buffer. In some instances, cells were pre‐incubated for 30 min at 37°C with 30 U Purelink RNase A (Invitrogen) or Turbo DNase I (Ambion). Anti‐FLAG‐APC (Abcam, clone M2) was used to detect ZBP1‐FLAG. Samples were analysed in FACS buffer (PBS, 2 mM EDTA, 1% FCS). 1 μg/ml DAPI (Sigma‐Aldrich) or Live/Dead Fixable Violet (Molecular Probes) was used to exclude dead cells. Data were acquired on Beckmann Coulter CyAn or BD Biosciences LSRFortessa flow cytometers. FCS files were analysed using FlowJo V.10.0.8. For fluorescence microscopy, polyclonal ZBP1‐expressing NIH3T3 cells were seeded on poly‐d‐lysine (Sigma‐Aldrich)‐coated coverslips (Lab‐Tek). Cells were fixed using 4% paraformaldehyde (ThermoFisher) and blocked and permeabilised for 1 h in PBS + 5% BSA (Sigma‐Aldrich) + 0.1% saponin (Sigma‐Aldrich). Primary antibody (anti‐FLAG‐HRP, clone M2, Sigma‐Aldrich) and detection antibodies (anti‐HRP‐Alexa Fluor 594, Jackson Immunoresearch) were incubated for 1 h in PBS + 5% BSA + 0.1% saponin (Sigma‐Aldrich). Images were acquired using a Zeiss Axiovert S100 inverted fluorescence microscope.

35siRNAs targeting MCMV IE3 were directed against exon 5 of the IE1/3 gene and were designed using siDESIGN Center from Dharmacon (GE Healthcare). siRNAs were ordered from Dharmacon, and the sequences are detailed in Appendix Table S4. A total of 5,000 NIH3T3 cells were seeded per well in 96‐well plates. 25 nM siRNAs were transfected using DharmaFECT 1 (Dharmacon, GE Healthcare) according to manufacturer's instructions. One hour later, transfection mix was washed away, and cells were infected with MCMV.

36PAR‐CLIP and CLIP protocols were adapted from (Hafner et al, 2010; Huppertz et al, 2014). Further details are provided in Appendix Supplementary Methods.

37Wild‐type and Zα1α2mut mouse ZBP1 were purified based on the protocol published in Sun et al (2013). See Appendix Supplementary Methods for further details.

38Statistical analysis was performed in GraphPad Prism v7.00 as detailed in the figure legends.

39JM and JR conceived the study, designed experiments, analysed data and wrote the manuscript. JM performed all experiments. LL and AB provided technical help and analysed data. KBR and JWU provided MCMV‐M45mutRHIM virus for the experiment shown in Fig 3G. All authors read and approved the final manuscript.

40The authors declare that they have no conflict of interest.