Ring finger protein 213 assembles into a sensor for ISGylated proteins with antimicrobial activity

ISG15 is an interferon-stimulated, ubiquitin-like protein that can conjugate to substrate proteins (ISGylation) to counteract microbial infection, but the underlying mechanisms remain elusive. Here, we use a virus-like particle trapping technology to identify ISG15-binding proteins and discover Ring Finger Protein 213 (RNF213) as an ISG15 interactor and cellular sensor of ISGylated proteins. RNF213 is a poorly characterized, interferon-induced megaprotein that is frequently mutated in Moyamoya disease, a rare cerebrovascular disorder. We report that interferon induces ISGylation and oligomerization of RNF213 on lipid droplets, where it acts as a sensor for ISGylated proteins. We show that RNF213 has broad antimicrobial activity in vitro and in vivo, counteracting infection with Listeria monocytogenes, herpes simplex virus 1, human respiratory syncytial virus and coxsackievirus B3, and we observe a striking co-localization of RNF213 with intracellular bacteria. Together, our findings provide molecular insights into the ISGylation pathway and reveal RNF213 as a key antimicrobial effector.

I SG15 is a ubiquitin-like (UBL) protein with antimicrobial activity. Similar to ubiquitin, ISG15 conjugates via its C-terminus to lysine residues of substrate proteins in a process called ISGylation, which is mediated by an E1 enzyme, UBE1L, an E2 enzyme, UBCH8, and three known E3 enzymes, HHARI, TRIM25, and HERC5 [1][2][3][4][5] . ISG15 and its conjugation machinery are strongly upregulated by Type I and III interferon, viral nucleic acids 1 , bacterial DNA 6 , and lipopolysaccharide (LPS) 7 . ISG15 has potent antiviral effects both in vitro and in vivo 8 . In fact, cells or mice which lack ISG15 are unable to control various viral pathogens including clinically relevant etiologic agents such as Influenza 9 , human respiratory syncytial virus 10 , and coxsackievirus 11,12 . ISG15's crucial antiviral role is bolstered by the variety of viral evasion strategies targeting the ISGylation pathway, either by interfering with ISG15 conjugation 1,13 or by expressing ISG15-specific proteases that lead to reversible [14][15][16][17] or irreversible deconjugation 18 . In addition to its antiviral role, ISG15 can also act as an antibacterial effector against intracellular bacterial pathogens such as Listeria monocytogenes 6 and Mycobacterium tuberculosis 19 and it can restrict the intracellular eukaryotic pathogen Toxoplasma gondii 20 . Despite its broad antimicrobial role, the molecular mechanisms by which ISG15 modification is sensed and how it protects against microbial infections remain largely undefined. One model posits that the antiviral function of ISG15 is based on the localization of the E3 ligase HERC5 at the ribosome. According to this model, proteins are co-translationally modified by ISG15 during infection, thereby interfering with the function of newly translated viral proteins in a nonspecific manner 21 . However, the model does not predict what subsequently happens to ISGylated proteins in the cell and how they are recognized and trafficked.
Unlike ubiquitin, ISG15 does not solely exert its antimicrobial function by covalently conjugating to target proteins. It can also be secreted and act as a cytokine [22][23][24] and can noncovalently interact with viral and host proteins 25 . Relatively little is known about what interaction partners bind ISG15 and how the functions of these interactions contribute to host responses to pathogens. Indeed, only a few host proteins have been reported to noncovalently bind to ISG15, mainly in its free form. One of the bestcharacterized ISG15-binding proteins is USP18, the predominant ISG15 deconjugating protease in human and mice 26,27 . In addition to its enzymatic activity, USP18 functions as a major negative regulator of type I interferon (IFN-I) signaling by binding to one of the subunits of the interferon α/β receptor (IFNAR2) [28][29][30] . In order to prevent IFN-I over-amplification in humans, USP18 needs to be stabilized by binding to free ISG15 in a noncovalent manner. Hence, patients who lack ISG15 expression due to a frameshift mutation have a strong upregulation of the IFN-I pathway leading to an Aicardi-Goutières-like interferonopathy 31 . In contrast to Isg15-deficient mice, these patients do not display enhanced susceptibility to viral infection, instead they are susceptible to bacterial infection including clearance of the attenuated tuberculosis vaccine, Bacille Calmette-Guerin (BCG) 32 . In addition to USP18, a few other proteins have been reported to interact with ISG15 in a noncovalent manner. Free intracellular ISG15 can bind to leucine-rich repeat-containing protein 25 (LRRC25) and mediate the autophagic degradation of retinoic acid-inducible gene I protein (RIG-I/DDX58) 33 . Upon forced overexpression, p62 and HDAC6 were further shown to interact with the C-terminal LRLRGG sequence of ISG15 (and ISGylated proteins) 34 . Free ISG15 also interacts with the E3 ligase NEDD4 and ISG15 overexpression disrupts NEDD4 ubiquitination activity thus blocking the budding of Ebola virus-like particles 35,36 . Likewise, free ISG15 was reported to interact in a noncovalent manner with the Hypoxia inducible factor 1α (HIF1α), preventing its dimerization and downstream signaling 37 .
ISG15 has not been studied as extensively as ubiquitin or SUMO for which interacting domains or motifs have been reported, thus to date no such domains or motifs have been described for ISG15. Although affinity purification mass spectrometry (AP-MS) and yeast two-hybrid (Y2H) screens were performed with ISG15 as bait, these interactome screens were hampered by the inherently limited binding surface of such a small protein and the stringency of yeast two-hybrid, which would preclude low affinity interactions [38][39][40][41] . Therefore, we endeavored to identify noncovalent interaction partners of ISG15 using a recently developed approach named Virotrap, which is based on capturing protein complexes within virus-like particles (VLPs) that bud from mammalian cells 42 . Virotrap is a cell-lysisfree method and thus preserves existing protein complexes, and, importantly, this method is uniquely suited to capture weak and transient cytosolic interactions that otherwise do not survive protein complex purification. Using this technology, we systematically mapped the noncovalent interactome of ISG15 in human cells. We identified a rich interactome but for further study focused on a very large protein, called Ring Finger Protein 213 (RNF213), mutations in which predisposes patients to early cerebrovascular events. Subsequent biochemical evaluation confirmed the selective binding of RNF213 to ISG15 and showed that IFN-I signaling induces RNF213 oligomerization into a sensor for ISGylated proteins associated with lipid droplets. Finally, in vitro and in vivo loss and gain of function experiments showed that RNF213 is a pivotal, broadly acting microbial restriction factor that colocalizes to the surface of intracellular bacteria.

Results
Identification of RNF213 as an ISG15-binding protein by Virotrap. To investigate the noncovalent interactome of ISG15 and cytosolic sensors for ISG15 modification, we used Virotrap, a recently developed MS-based approach to map protein-protein interactions 42 . Briefly, the sequence of mature ISG15 (ending on -LRLRGG) was genetically fused via its N-terminus to the GAG protein of HIV-1 and expressed in HEK293T cells treated with IFN-I. Expression of the GAG-ISG15 fusion protein led to selfassembly of VLPs internally coated with ISG15. After budding from the cells and capturing ISG15 interaction partners, VLPs were collected from the cell culture supernatants and used for MS-based protein identification (Fig. 1a). To distinguish specific ISG15-binding proteins from nonspecific (GAG-binding) proteins, the experiment was performed in a quantitative fashion comparing ISG15 VLPs with VLPs containing dihydrofolate reductase from Escherichia coli (eDHFR), a nonspecific bait protein with no obvious interaction partners in human cells. In addition to ISG15, we identified RNF213 as the most enriched protein in the ISG15 VLPs, associated with high spectral and peptide counts (Fig. 1b, Supplementary Data 1). Of note, among several less enriched proteins we also identified USP18, which is an ISG15-specific isopeptidase and thus a well-known interaction partner of ISG15 26,27 . To rule out any type of covalent interaction between GAG-ISG15 and RNF213, we repeated the Virotrap experiment using the nonconjugatable ISG15AA (with the c-terminal glycines mutated to alanines: LRLRAA) and precursor variants of ISG15, both with and without IFN-I treatment ( Supplementary Fig. 1, Supplementary Data 1). Without exception, in all conditions we identified RNF213 as the most enriched protein in the ISG15 VLPs. We also confirmed the presence of RNF213 in the ISG15 VLPs by western blotting (Fig. 1c).
In addition to RNF213, 28 other proteins were enriched in ISG15 VLPs under all conditions ( Supplementary Fig. 1, Supplementary Data 1). Following RNF213, the most notable among these identifications was the valosin-containing protein (VCP/p97).
VCP is a hexameric AAA + ATPase with a central role in the endoplasmic-reticulum-associated protein degradation (ERAD) pathway, where it associates with cofactor proteins to facilitate the proteasomal degradation of misfolded proteins 43 . In addition, several cofactors of VCP were found to interact with ISG15 such as the VCP-UFD1-NPL4 protein complex as well as PLAA, UBXN6 and YOD1 44 , FAF1, NSFL1C, UBXN1, and RNF115 [45][46][47][48] . We also detected several components of the proteasome complex enriched in ISG15 VLPs, including ADRM1 and several PSMA and PSMB subunits of the proteasome. Moreover, we identified TAB1 and TAK1 (MAP3K7), which are involved in innate immune signaling pathways as well as TRAFD1 49,50 .
We then performed an additional series of Virotrap-MS experiments in which only the N-or C-terminal domain of   Fig. 3a-c). This suggested that affinity between a single ISG15 molecule and RNF213 is too weak to detect, similar to a ubiquitin or SUMO molecule to their respective binding proteins 51,52 . Hence, we applied a Glutathione-S-Transferase (GST) pull-down assay in which we coated glutathione beads with GST-ISG15 and then used these ISG15-decorated beads to search for ISG15-binding proteins in a lysate of HEK293T, HeLa, and THP-1 cells treated with IFN-I. GST-coated beads were used as control and interacting proteins were identified by mass spectrometry. Interestingly, this approach yielded RNF213 as a strongly enriched partner of ISG15 along with several other interactors ( Fig. 1d-f); however, only RNF213 was identified in both the GST enrichment and the Virotrap experiments (Supplementary Fig. 3d). Among the other interactors, we detected UBE1L (UBA7) and HDAC6, which are known ISG15 binders in THP-1 and HEK293T cells, respectively 1,34 . To further confirm the specificity of the ISG15/RNF213 interaction, we next combined the pull-down of GST-ISG15 with western blotting. In these experiments, a lysate of HEK293T cells expressing FLAG-RNF213 was mixed with glutathione beads bound to GST-ISG15, GST-SUMO, or GST-ubiquitin. While RNF213 clearly binds to GST-ISG15, only background binding could be observed to GST-SUMO and GST-ubiquitin (Fig. 1g). As a positive control, we confirmed binding of RNF4 and RNF31, known binding partners of GST-SUMO and GST-ubiquitin, respectively 51,53 (Fig. 1h, i).
Together, these data show that RNF213 specifically binds ISG15, rather than ubiquitin or SUMO.
RNF213 binds ISGylated proteins on lipid droplets. RNF213, also known as mysterin, is a multidomain megaprotein of unknown function. Polymorphisms in RNF213 predispose human patients to Moyamoya disease (MMD), a rare vascular brain disease leading to stroke, but the underlying molecular mechanisms of this disease remain elusive 54 . The most abundant RNF213 isoform is 5207 amino acids long and is comprised of two adjacent AAA + ATPase modules with a RING domain 55 . RNF213 is expressed in most cells and tissues 56 , but levels can be upregulated by immune and inflammatory signaling 55 . AAA + proteins typically assemble into hexameric toroidal complexes that generate mechanical force through ATP binding/hydrolysis cycles. In the case of RNF213, complex formation is dynamic and driven by ATP binding while ATP hydrolysis mediates disassembly 55 .
Interestingly, a recent microscopy study showed colocalization of RNF213 and lipid droplets (LDs), suggesting a model in which monomeric RNF213 is directly recruited from the cytosol and ATP binding drives its oligomerization on the surface of lipid droplets 57 . This model, together with the observation that ISG15 binds RNF213 in Virotrap and GST pull-down experimentswhich both present multiple ISG15 bait molecules in close proximity on a concave or convex surface, respectively (Fig. 1a)led us to hypothesize that RNF213 oligomerizes on LDs into a binding platform for ISG15 and potentially multiply ISGylated proteins.
To test this hypothesis, we first verified whether RNF213 can bind ISGylated proteins. To this end, we pulled down FLAG-RNF213 from lysates of HEK293T cells expressing HA-tagged ISG15 or ISG15AA and its conjugation machinery to induce ISGylation. Western blotting against the HA-tag revealed a smear of ISGylated proteins after FLAG-RNF213 pull-down, but not with FLAG-eGFP pull-down or when ISG15AA was expressed (Fig. 2a). We repeated the experiment with eGFP-tagged RNF213 ( Supplementary Fig. 4a) and in both cases did not observe any free ISG15 or ISG15AA associated with RNF213, indicating that in cells RNF213 indeed preferentially senses and binds ISGylated proteins rather than free ISG15. In order to determine whether the ISGylated proteins detected were instead partially degraded RNF213 we included a 1% SDS wash, which reduced the ISGylated proteins ( Supplementary Fig. 4b) and even demonstrated that RNF213 could bind ISGylated proteins in trans, which originated from a different cell lysate ( Supplementary  Fig. 4c, d). These data indicate that RNF213 does bind ISGylated proteins. Finally, we also showed that this binding was Fig. 1 Identification of RNF213 as ISG15-binding protein. a Workflow to map ISG15 interaction partners. Virotrap and Glutathione-S-Transferase (GST) pull-down were employed as two orthogonal methods. In Virotrap, GAG-ISG15 was expressed in HEK293T cells. Twenty four hours post-transfection, cells were treated with interferon (IFN)-α for 24 h of left untreated. Budding virus-like particles (VLPs) containing ISG15 and its interaction partners were purified, lysed and digested into peptides prior to LC-MS/MS analysis. In GST pull-down, glutathione beads were decorated with GST-ISG15 and mixed with a cellular lysate from HEK293T, HeLa, or THP-1 cells. Prior to lysis cells were treated for 24 h with interferon-α (HEK293T) or interferon-β (HeLa and THP-1). Following on-bead digestion, the resulting peptides were analyzed by LC-MS/MS. b Volcano plot showing the result of a t-test to compare VLPs containing mature ISG15 versus VLPs containing dihydrofolate reductase from Escherichia coli (eDHFR) as negative control (n = 4 replicates). Proteins outside the curved lines represent specific ISG15 interaction partners. Proteins identified as ISG15 interaction partners in all virotrap screens are annotated (n = 29) and listed in Supplementary Data 1. c VLPs containing ISG15 or eDHFR were collected and analyzed by immunoblot (IB) against RNF213, ISG15, and GAG. FLAG-RNF213 purified from a lysate of HEK293T cells was loaded as a positive control and confirmed the presence of an RNF213 band in the ISG15 VLPs. d-f Volcano plots comparing GST pulldowns using GST-ISG15-coated beads with GST-coated beads as control in lysates of HEK293T, THP-1, and HeLa cells (n = 3). Proteins significantly enriched in PDs with ISG15-coated beads are annotated. g GST PDs with ISG15, ubiquitin, and SUMO followed by immunoblotting show binding of RNF213 to ISG15, but not to ubiquitin or SUMO. Beads coated with each UBL were mixed with a lysate of HEK293T cells expressing FLAG-RNF213 and bound proteins were analyzed by immunoblot against FLAG and GST. h, i Validation of GST PD assays with ubiquitin and SUMO using RNF31 and RNF4 as known binders of these UBLs, respectively. Assays were performed as in g. Source data are provided as a Source Data file.
ISG15 specific, since ubiquitinated proteins were not enriched ( Supplementary Fig. 4e), in line with the results from the GST pull-down (Fig. 1g).
We next tested whether the interaction between RNF213 and ISGylated proteins occurs on LDs. In HeLa cells treated with IFN-I we observed occasional colocalization of overexpressed eGFP-RNF213 with endogenous ISG15 by fluorescence microscopy; however, ubiquitous intracellular distribution of free ISG15 prevented reliable quantitation of these events. We therefore verified this question further biochemically using macrophages, a cell type which is more relevant for innate immune signaling. We isolated LDs from THP-1 cells and primary human monocytes by flotation on a sucrose gradient 58 . Western blotting for RNF213 and LD marker proteins such as PLIN1 and PLIN2 confirmed that in THP-1 cells the majority of RNF213 is associated with LDs, while a minor fraction is in the cytosol ( Supplementary  Fig. 5a). Upon IFN-I treatment, the fraction of RNF213 associated with LDs further increased, along with a striking appearance of a smear of ISGylated proteins, an observation that was corroborated in primary human monocytes (Fig. 2b). Finally, to show that ISGylated proteins associate with LDs via RNF213, we repeated the experiment in THP-1 cells in which we reduced the expression of RNF213 by siRNA. As expected, knockdown of RNF213 markedly reduced association of ISGylated proteins with LDs and this occurred without any apparent loss of lipid droplet stability as indicated by unchanged levels of LD markers PLIN1, PLIN2, and ATGL (Fig. 2c). Furthermore, in bone-marrowderived macrophages (BMDMs) isolated from RNF213 knockout mice 59 we found that the number and size of lipid droplets was similar to WT BMDMs ( Supplementary Fig. 5b, c), indicating that in macrophages RNF213 is not essential for lipid droplet stability, in contrast to other cell types 57 . Together, these data show that RNF213 binds ISGylated proteins and that RNF213 recruits ISGylated proteins to LDs.

IFN-I induces RNF213
ISGylation and oligomerization on lipid droplets. We subsequently investigated how oligomerization of RNF213 on LDs is regulated. Since both ISG15 and RNF213 are IFN-I induced genes ( Supplementary Fig. 6a) 60 , it seemed plausible that multimerization of RNF213 into a sensing platform for ISGylated proteins could be mediated by interferon signaling. We therefore evaluated the effect of IFN-I on the distribution of endogenous RNF213 between the soluble and membrane-associated fractions in HeLa cells and found that the level of membrane-associated RNF213 slightly increased upon IFN-I treatment ( Supplementary Fig. 6b). To further confirm RNF213 oligomerization in a relevant model, we separated lysates from control and IFN-I treated THP-1 cells by sedimentation over continuous glycerol gradients. Western blotting of the collected fractions clearly showed that IFN-I treatment gave rise to a pool of higher molecular weight RNF213 in fractions 14-20, as expected after oligomerization (Fig. 3a). Interestingly, this pool of slow migrating RNF213 was marked by the presence of higher molecular weight bands on the blots, most likely representing modified forms of RNF213. Since we recently identified RNF213 as the most ISGylated protein following Listeria monocytogenes infection with 22 ISGylation sites 61 , we tested whether oligomeric RNF213 was ISGylated. Immunoprecipitating RNF213 from each Lipid droplet fraction Lysate fraction followed by immunoblotting for ISG15 showed that this was indeed the case, while a ubiquitin-specific signal was only detected in the low-density fractions (Fig. 3a). The ISG15 blot revealed two major groups of ISGylated RNF213 with one group of bands running at virtually the same height as monomeric RNF213 and a second, hyper-ISGylated group of bands running well above (Fig. 3a). Knockdown of UBE1L, the E1 activating enzyme for ISG15, strongly reduced the intensity of both groups of ISGylated RNF213 while it increased the intensity of monomeric RNF213, indicating that ISGylation of RNF213 is required for its oligomerization ( Fig. 3b and Supplementary Fig. 6c). Finally, we repeated the glycerol sedimentation experiment after LD isolation and showed that the oligomeric, ISGylated forms of RNF213 are associated with LDs only after IFN-I treatment 11  The monomeric and oligomeric forms of RNF213 were separated by density gradient ultracentrifugation after interferon-β treatment as in a and knockdown of UBE1L by siRNA (siUBE1L) treatment for 48 h, using a nontargeting scrambled siRNA (siScramble) as control. Knockdown of UBE1L strongly reduced RNF213 oligomerization upon interferon treatment. c LDs were isolated by ultracentrifugation floatation assay on a sucrose step-gradient and associated proteins were further separated by density gradient ultracentrifugation to isolate the monomeric and oligomeric form of RNF213 after interferon-β treatment as in a. Fractions were concentrated by TCA precipitation and analyzed by immunoblotting against RNF213, showing association of oligomeric RNF213 with LDs upon interferon treatment. Source data are provided as a Source Data file.
( Fig. 3c). Together, these results show that IFN-I induces the oligomerization of RNF213 on LDs and that this process requires ISG15 modification of RNF213 itself.
RNF213 exerts broad antimicrobial activity in vitro. The discovery of RNF213 as a sensor for ISGylated proteins made us suspect that RNF213 could also have antiviral properties, similar to ISG15, which was in line with the recently reported increased susceptibility of Rnf213-deficient mice to Rift Valley fever virus infection 62 . To investigate this, we infected cells with reduced or enhanced expression levels of RNF213 with different viruses. RNF213 knockdown or overexpression did not affect ISG expression ( Supplementary Fig. 7a, b). We first infected HeLa cells with herpes simplex virus 1 (HSV-1), a widespread human pathogen that can cause cold sores and genital herpes that is ISG15 sensitive 63 . We used a GFP-expressing HSV-1 strain to monitor infection levels up to 3 days post infection by fluorescence intensity. Knockdown of RNF213 led to a small but significant increase in infection compared to scrambled siRNA control, both in the presence or absence of IFN-I (Fig. 4a-c). Conversely, overexpression of FLAG-RNF213 significantly lowered HSV-1 infection levels, although not to the same extent as overexpressed MXB, which was recently reported as a Herpesvirus restriction factor (Fig. 4d, Fig. 7c-h). We also determined the possible involvement of RNF213 in the control of replication of human respiratory syncytial virus (RSV), an important human respiratory pathogen that is susceptible to an ISG15-dependent antiviral effect 10 . For these experiments, A549 cells were infected with RSV-A2 at a low multiplicity of infection (MOI) and incubated for 6 days to allow multiple rounds of infection (Supplementary Fig. 8a). Knockdown of RNF213 led to a small but significant increase in the viral load in the cell culture supernatants 5 days post infection compared to scrambled siRNA control, an effect that disappeared at day 6 ( Fig. 5a). In addition, when the cells were treated with IFN-I, significantly higher viral loads were observed in siRNF213-treated cells compared to siScramble-treated cells at day 5 and 6 after infection (Fig. 5b).
We also assessed the effect of RNF213 knockdown on RSV replication in a plaque assay because the plaque size is a measure for the extent of cell-to-cell spread of the virus. On day 6 after RSV infection, we observed significantly larger plaques upon knockdown of RNF213 both with and without IFN-I, suggesting also increased cell-to-cell spread of RSV under these conditions (Fig. 5c, d). As a third ISG15-sensitive virus we tested CVB3, a Picornavirus that can lead to cardiomyopathy and that is known to be ISG15 sensitive 11,12 ( Supplementary Fig. 8b). Again, knockdown of RNF213 significantly increased the replication of CVB3 as indicated by higher viral genome (Fig. 6a) and enhanced viral protein expression levels ( Fig. 6b) as well as by the elevated We subsequently wondered whether RNF213 could be broadly antimicrobial and also target bacterial pathogens. Thus, we infected HeLa cells with Listeria monocytogenes (further referred to as Listeria), a facultative intracellular bacterium that was recently reported to be counteracted by ISG15 6 ( Supplementary  Fig. 9a). When we knocked down RNF213 with a pool of siRNAs, we measured a 30% increase in intracellular bacteria compared to a pool of control siRNAs, an increase that was similar to what we observed by knockdown of ISG15 (Fig. 7a). Individual siRNAs against RNF213 also augmented the bacterial load (Supplementary Fig. 9b) and we observed no effect of RNF213 knockdown or overexpression on bacterial entry ( Supplementary Fig. 9c, d).
Conversely, overexpression of FLAG-RNF213 reduced the bacterial load with 85% relative to mock transfected cells, a reduction that was similar to overexpression of HA-ISG15 (Fig. 7b). Importantly, the protective effect of overexpressed FLAG-RNF213 was lost in ISG15 knockout cells (Fig. 7c), indicating that, unlike for the antiviral activity, the antimicrobial activity of RNF213 requires ISG15. In contrast, overexpressed HA-ISG15 still counteracted Listeria in RNF213 knockout ( Fig. 7d) or knockdown cells ( Supplementary Fig. 9e), meaning that ISG15 also harbors antimicrobial activity independent of RNF213. Taken together, we showed that increasing the  expression levels of RNF213 led to lower infection levels of HSV-1, CVB3, and Listeria in cultured cells, while reducing the levels of RNF213 promoted in vitro infection with these pathogens as well as RSV. Compared to the antiviral activity, the antibacterial effect of RNF213 was more pronounced and dependent on ISG15.
These results show that RNF213 plays an important role in the innate cellular immune response against invading Listeria, which is in line with recent findings on Salmonella 65 and indicating a functional link between ISG15 and RNF213.
RNF213 decorates intracellular Listeria and is antibacterial in vivo. Given the protective function of RNF213 against Listeria, we wondered if RNF213 adopts a specific subcellular localization during infection. To this end, we transfected HeLa cells with eGFP-RNF213. Co-staining for LDs followed by fluorescence microscopy revealed that RNF213 colocalized with the majority (~70%) of LDs, both in uninfected and Listeria-infected cells (Fig. 8a, b). We noticed that infection with Listeria was associated with an increase in LDs ( Supplementary Fig. 10a), similar to previous results in human macrophages, which had phagocytosed E. coli or Salmonella 66 . Intriguingly, RNF213 also colocalized with a subset of intracellular Listeria, and seemed to decorate the bacterial surface (Fig. 8a). To quantify how many bacteria were targeted, we repeated the experiment with mCherry-expressing Listeria and found that on average approximately 40% of intracellular Listeria cocolocalized with RNF213 at 18 h post infection (Fig. 8c, d). In order to determine whether RNF213 was docking on bacteria or on membrane remnants near bacteria, we made use of a triple mutant, which lacks phospholipase A (PlcA), phospholipase B (PlcB), and the primary hemolysin Listeriolysin O (LLO) 6 . This mutant is unable to escape from the vacuole of epithelial cells and when compared with the parental strain at 6 h post infection, we did not observe colocalization with RNF213 ( Fig. 8e), indicating that RNF213 likely targets a subset of cytosolic bacteria, as was recently shown for Salmonella 65 . The latter study demonstrated that ubiquitination of cytosolic Salmonella by RNF213 initiates autophagic clearance. Unlike Salmonella, however, the majority of wild-type cytosolic Listeria evades autophagy through a variety of mechanisms including the use of ActA to sterically protect the surface of bacteria from being modified by ubiquitin 67 and of PlcA to alter phospholipid levels in the cell to thereby interfere with LC3-associated phagocytosis in macrophages or autophagy initiation in epithelial cells 68,69 . Furthermore, more recent work suggests that Listeria can evade host autophagy in a multistep process involving LLO, ActA, and PlcA 70 . As such, we visualized actin using phalloidin and mono or poly ubiquitin linkages using anti-FK2 in RNF213 overexpressing, Listeria-infected cells. Our data are in line with the published role of ActA as surfaces of individual bacteria marked by RNF213 and ubiquitin (sides of bacteria and sites of division) did not colocalize with surfaces marked by actin once ActA has relocalized to the bacterial poles (Fig. 8e). Furthermore, RNF213 and ubiquitin colocalize on the bacterial surface suggesting that similar to its modification of Salmonella, RNF213 also acts as a ubiquitin E3 ligase during Listeria infection ( Supplementary  Fig. 10b). Together, these data suggest that overexpression of RNF213 prior to infection can overcome the capacity of wild-type Listeria to evade the initial host-mediated targeting steps that mark the bacterial surface with ubiquitin and allow the host cell to ultimately dispose of bacteria in an autophagosome or autolysosome. Future work will assess the interplay of RNF213 and downstream autophagic targeting of Listeria in both phagocytic and epithelial cells.
In order to test whether this localization has functional consequences in vivo, we deleted RNF213 in mice using CRISPR/Cas9 targeting of mouse RNF213 exon 28 as previously described 59 (Supplementary Fig. 11). At 24, 48, and 72 h post infection with Listeria we observed a dramatic increase in bacterial burden in the spleen of Rnf213 −/− animals compared to WT littermate controls, which increased over time (Fig. 9a). At 72 h, a significant increase was also detected in the liver (Fig. 9b) and this observation was confirmed in two additional independent experiments ( Supplementary Fig. 12a-d). This profound difference in bacterial load, particularly in the spleen, reveals a central role of RNF213 in the host defense against bacterial infection. Compared to WT littermates, RNF213 KO animals were also more likely to lose more bodyweight and reach the ethical endpoint following challenge with a mouse-adapted strain of RSV even though the difference in lung virus loads determined on day 5 after infection did not reach statistical significance ( Supplementary Fig. 12e-g). Future work will assess which cell type is unable to control Listeria infection and whether RNF213 deletion also affects adaptive immune responses in addition to innate responses. This is of particular relevance to understanding the role of inflammation or prior infection in human patients with inborn errors in RNF213.
The RNF213 E3 module is required for its antibacterial activity. Ahel et al. recently published the cryo-EM structure of monomeric RNF213 71 showing that RNF213 does not depend on its RING domain to function as an E3 ubiquitin ligase, but instead classifies as a special type of E3 ligase employing a Cys-containing motif to promote ubiquitin transfer via transthiolation. While a ΔRING still exhibits ubiquitination activity 65,71 and without knowledge of the catalytic residues, we sought to use this structural data to test whether the E3 ligase activity of RNF213 is required for its antimicrobial function. We therefore deleted the C-terminal 1210 amino acids of RNF213 comprising the large E3 module and small C-terminal domain to generate a FLAG-RNF213ΔC construct that is expected to be devoid of any ubiquitination activity. GST pull-down showed that RNF213ΔC was still capable of binding ISG15, while the complementary C-terminal fragment was not (Fig. 10a). Similarly, FLAG-RNF213ΔC was capable of pulling down a smear of ISGylated proteins to the same extent as full-length FLAG-RNF213 (Fig. 10b). However, without any effects on bacterial entry ( Supplementary Fig. 9c), overexpression of RNF213ΔC no longer retained its notable antimicrobial effects when overexpressed as it could not reduce the amount of intracellular Listeria in infected HeLa cells (Fig. 10c). Furthermore, fluorescence microscopy confirmed that eGFP-RNF213ΔC no longer colocalizes with intracellular Listeria (Fig. 10d) or with LDs ( Supplementary  Fig. 13a-b). Taken together, these results indicate that the E3 module of RNF213, and likely its ubiquitination activity, is required for its bacterial targeting and antimicrobial activity. Of note, while the present paper was in review, Otten et al. reported a catalytic zinc-binding RZ domain adjacent to RNF213's RING domain 65 , something that is mechanistically supported by recent structural work 72 . In line with the data on Listeria presented here, catalytic activity of the RZ domain was shown to be required for targeting of RNF213 to cytosolic Salmonella 65 .

Discussion
We report the discovery of RNF213 as an intracellular sensor for proteins modified by ISG15, a ubiquitin-like, immunity-related modification. We show that IFN-I signaling induces the oligomerization and translocation of RNF213 on lipid droplets in macrophages and that this process requires ISGylation of RNF213 itself. Furthermore, in vitro and in vivo infection assays with four different pathogens revealed an as yet undescribed, broad antimicrobial function of RNF213. Thus, our unique Virotrap method revealed a functional link between ISG15 and RNF213, as well as an undiscovered role for RNF213 in host defense pathways to both bacterial and viral infections.
We identified RNF213 as an ISG15-binding protein by Virotrap and confirmed the interaction by GST pull-down assays, methods that both create a curved surface, either convex or concave, studded with ISG15 molecules which could be a surrogate for multiply modified proteins or larger binding surfaces such as organelles. It was of particular interest, that we identified a large AAA + ATPase as ISG15 interactor under these conditions, RNF213, which presumably docks on ISG15 surfaces or vice versa. Indeed, our data suggest that oligomeric RNF213 acts as a binding platform for ISGylated proteins on the surface of   57 , and lipotoxicity 76 . The latter study also assessed changes in the RNF213-dependent ubiquitylome and showed that RNF213 was required to activate the nuclear factor κB (NFκB) pathway upon palmitate treatment. This NFκB-inducing function of RNF213 would, however, be negatively regulated by its ubiquitin ligase activity 77 and requires further investigation, especially in light of the RNF213 antimicrobial activity reported here. The recently published cryo-EM structure of monomeric RNF213 revealed three major structural components comprising an N-terminal stalk, a dynein-like ATPase core and a C-terminal E3 module 71 . Previous studies showed that the ATPase activity of RNF213 is essential for its multimerization and localization to lipid droplets; however, in contrast to most AAA + ATPases that form stable hexamers 55 , multimerization of RNF213 is proposed to be a dynamic process, similar to the dynamic Fig. 8 RNF213 decorates intracellular Listeria. a Representative images of HeLa cells transfected with eGFP-RNF213 and counterstained for lipid droplets (LDs). Following transfection cells were left untreated or infected with Listeria monocytogenes EGD for 24 h at MOI 25. Scale bars in the pictures and insets are respectively 10 microns and 1 micron. White arrows indicate colocalization between RNF213 and lipid droplets or instances of RNF213 colocalization with intracellular bacteria (DAPI = 4′,6-diamidino-2-phenylindole). b LDs in uninfected (n = 66 cells) and Listeria-infected cells (n = 40 cells) from a were quantified with Fiji and the percentage of LDs per cell that colocalized with RNF213 was calculated (representative results from a single experiment, twotailed unpaired t-test, AVG uninfected = 68.29%, AVG Listeria-infected = 67.14%, n.s. = not significant). c Representative images of HeLa cells transfected with eGFP-RNF213 and infected for 18 h with Listeria monocytogenes EGD stably expressing mCherry. Scale bars in the pictures and insets are respectively 10 microns and 0.5 micron. d Intracellular Listeria from c were quantified with Imaris 9.6 and the percentage of Listeria that was decorated by eGFP-RNF213 was calculated for each field by mapping the cell surface, enumerating intracellular bacteria and quantifying bacteria that colocalized with GFP-RNF213 (defined as bacteria within 0.5 µm of RNF213, see methods for more detail). At least 200 cells were counted per experiment, and data were compiled from three independent experiments indicating that on average 37. assembly of the microtubule-severing protein katanin 55,57 . In this model RNF213 exists as a monomer in solution and forms a hexamer on the surface of lipid droplets upon ATP binding 57 . Contrary to this, the RNF213 structure demonstrated the presence of six AAA units within the monomeric molecule 71 , calling into question the mechanism of oligomerization and proposed hexameric nature of oligomer RNF213. Here we report that oligomerization of RNF213 on lipid droplets is induced by IFN-I and requires ISGylation of RNF213 itself, an observation that fits with our previous identification of RNF213 as the most ISGylated protein in liver of Listeria-infected mice 61 . Various inducers of RNF213 ISGylation could thus potentially facilitate future structural studies on oligomeric RNF213. However, the mechanism of IFN-I-induced oligomerization of RNF213 remains an open question. Relevant in this regard is that RNF213 was reported as a substrate of Protein tyrosine phosphatase 1B (PTP1B or PTPN1), which controls nonmitochondrial oxygen consumption in response to hypoxia 78 . Since PTP1B also affects JAK-STAT signaling 79,80 , it would be worthwhile to test whether PTPB1 also regulates RNF213 oligomerization in response to interferon.
It is likely that further processing of ISGylated proteins bound to RNF213 is closely linked to the emerging function of LDs beyond passive lipid storage and lipid homeostasis. LDs have a role in the coordination of immune responses as they participate in the production of pro-inflammatory mediators 81 . LDs have been associated with the IFN-I response and antigen cross presentation 82 . Moreover, an increasing number of studies show that viral and bacterial pathogens target LDs during infection either for nutritional purposes or as part of an anti-immunity strategy 81 . Recent work showed that host cells exploit LD targeting by pathogens by loading LDs with antimicrobial proteins as an intracellular first line of defense 66 . In this study, RNF213 was found to be enriched on LDs isolated from the liver of LPStreated mice, similar to our results on IFN-treated cells. This is particularly interesting since ISGylation is also induced by LPS 7 . The localization of RNF213 on the surface of LDs could represent a host-cell strategy for fighting intracellular pathogens that hijack LDs. Given that LDs are also sites of viral assembly, RNF213 could interrupt this process to decrease infectivity in the cell. The broad antimicrobial activity of RNF213 that we observed bolsters this hypothesis. GST-ISG15 beads: Fig. 10 The RNF213 E3 module is required for its antimicrobial activity. a Glutathione-S-Transferase (GST) pull-down with ISG15 followed by immunoblotting shows that 3xFLAG-RNF213ΔC binds to ISG15 similar to full-length 3xFLAG-RNF213. Beads coated with GST-ISG15 were mixed with a lysate of HEK293T cells expressing full-length 3xFLAG-RNF213, 3xFLAG-RNF213ΔC, or the complementary C-terminal fragment of RNF213 and bound proteins were analyzed by immunoblot (IB) against FLAG and GST. b FLAG immunoprecipitation (IP) was performed from lysates of HEK293T cells expressing 3xFLAG-RNF213, 3xFLAG-RNF213ΔC, or FLAG-eGFP as control. After immunoprecipitation, beads were mixed with a lysate of HEK293T cells expressing HA-ISG15 and the ISGylation machinery (E1, E2, E3). 3xFLAG-RNF213ΔC was capable of pulling down ISGylated proteins similar to 3xFLAG-RNF213, while FLAG-eGFP was not. c HeLa cells were infected with Listeria monocytogenes EGD for 4 h at MOI 25. Twenty four hours prior to infection, HeLa cells were transfected with plasmids encoding 3xFLAG-RNF213 or 3xFLAG-RNF213ΔC or with an empty vector (mock) as control. Intracellular Listeria were quantified after serial dilution by counting colony-forming units (CFUs) in a gentamycin assay. The percentage of intracellular bacteria relative to mock plasmid-transfected cells is shown (lower panel, AVG ± SEM, n = 3 independent experiments, two-tailed unpaired t-test, n.s. not significant). Immunoblots against FLAG with tubulin as loading control confirmed expression of FLAG-RNF213, FLAG-RNF213ΔC (upper panel). d Representative images of HeLa cells transfected with eGFP-RNF213ΔC and infected for 18 h with Listeria monocytogenes EGD stably expressing mCherry. Scale bars in the pictures and insets are respectively 10 microns and 0.5 micron. eGFP-RNF213ΔC showed a diffused cellular staining, not decorating intracellular Listeria (DAPI = 4′,6-diamidino-2-phenylindole). In c asterisks indicate p values with ****p < 0.0001. Source data are provided as a Source Data file. The localization of RNF213 as a general virus restriction factor on the surface of LDs could be further rationalized in the context of enteroviruses, such as coxsackievirus, which replicate at specialized membranous domains named replication organelles 83 . However, whether RNF213 also resides in such lipid-rich replication organelles remains to be determined. On the other hand, ISG15 contributes to the IFN-dependent anti-RSV effect, whereas no role for RNF213 has been reported in the control of RSV. In this context, it is important to note that RSV encodes two prominent nonstructural proteins that profoundly suppress type I IFN induction and signaling 84 . The latter is also true for HSV-1, which has evolved multiple strategies to evade the host antiviral response and to establish latent infections, so far without any known involvement of RNF213 85 . With all three tested viral pathogens we observed a clear, but subtle antiviral effect of RNF213 in vitro, in line with the relatively modest phenotype observed in vivo in RSV infected RNF213 KO mice. Knockdown or overexpression of RNF213 changed viral infection levels two to five fold, in contrast to Listeria where we observed up to 100,000 fold differences in bacterial load, especially in vivo, indicating a more pronounced antibacterial effect. Interestingly, while the protective effect of RNF213 against Listeria was dependent on ISG15, this was not the case for HSV-1, suggesting that different mechanisms underlie the antiviral and antibacterial effects. Of note, a recent study also uncovered that loss of RNF213 increases susceptibility to Rift Valley fever virus infection 62 , indicating that the antiviral activity of RNF213 extends beyond the viruses tested here.
Genome-wide siRNA screens in Listeria-infected HeLa cells previously ranked RNF213 as a protective host protein in vitro, in line with our observations 86 . Importantly, the protective effect of RNF213 was lost in ISG15 KO cells, indicating that the antimicrobial activity of RNF213 requires ISG15. Inversely, ISG15 was still functional in RNF213-deficient cells, pointing towards other unknown conjugation-dependent or -independent mechanisms by which ISG15 counteracts Listeria, as shown for many other pathogens (Perng and Lenschow, 2018). Our infection experiments with Listeria in mice, which lack RNF213 corroborated the protective effect in vivo, showing a multi-log increase in the number of colony-forming units in liver and spleen over time compared to WT animals. This increase is larger than what was reported under the same conditions for Isg15deficient animals 6 and is reminiscent of Listeria infections in mice deficient in key immune signaling pathways necessary to control primary infection such as IFN-gamma 87,88 . Further characterization of the immune response in the Rnf213-deficient animals is both timely and relevant to assess whether a defect in cellautonomous immunity is underlying the increased susceptibility, as suggested by our in vitro data.
Colocalization of RNF213 with intracellular Listeria further supports a function of RNF213 as a restrictive host factor. Proteins that colocalize with intracellular bacteria are often involved in an antibacterial strategy known as xenophagy. Upon activation, specific E3 ubiquitin ligases, such as LRSAM1 89 , SMURF1 90 , LUBAC 91,92 , or Parkin 93 decorate invading pathogens with ubiquitin chains. Cytosolic autophagy receptors such as p62 or NDP52 recognize these chains and recruit LC3-II to capture the ubiquitinated bacteria into expanding autophagosomes 91 . Since RNF213 has an active E3 ubiquitin ligase module, it is tempting to speculate that RNF213 conjugates ubiquitin chains to invading Listeria and facilitates its capture into autophagosomes. While the present paper was in review, Otten et al. reported that RNF213 indeed acts as a bacterial E3 ligase that can ubiquitinate cytosolic Salmonella, directly on the lipid A moiety of LPS, and that this is required to attract ubiquitin-dependent autophagy receptors and induce antibacterial autophagy 65 . It remains to be determined if RNF213 also ubiquitinates Listeria and on which cell wall components (as a Gram-positive bacterium Listeria lacks LPS). A vacuole-contained mutant of Listeria did not colocalize with RNF213 and, together with occasional co-staining of RNF213 with ubiquitin on the surface of wild-type bacteria this suggests that RNF213 might target Listeria for autophagy similar to Salmonella. Future studies should investigate the fate of RNF213decorated Listeria in more depth, especially since in contrast to Salmonella, cytosolic Listeria can evade xenophagy through various mechanisms involving major virulence factors such as LLO, PlcA, and ActA [67][68][69][70] . ATPase-and ubiquitin-deficient mutants of RNF213 will allow to assess the implication of both enzymatic activities of RNF213 as well as the role of RNF213 oligomerization. In this regard, loss of bacterial and LD colocalization of the RNF213ΔC mutant could be analogous to what was shown for variants lacking a functional ATPase core, i.e. that ubiquitination activity seems to be required for oligomerization and organelle targeting of RNF213 57 . Taken together, the data presented here provides evidence for colocalization of RNF213 with intracellular Listeria, classifying RNF213 as a pathogen-directed E3 ubiquitin ligase.
Strain generation. Strains used in this study are listed in Supplementary Table 1. To generate mCherry EGD we codon optimized mCherry for Listeria monocytogenes using the JCAT website 103 , ordered the codon optimized geneblock from IDT and subcloned mCherry into pAD, electroporated competent Listeria monocytogenes EGD (BUG600) and selected for chromosomal integration of the plasmid.
SiRNA transfection. A commercially available siRNA pool was used to knockdown human RNF213 (#M-023324-02, GE Healthcare Dharmacon) in Figs. 2, 3, 4, 5, 6, and 7. As control siRNA treatment, a nontargeting scramble siRNA (#D-001210-01-05, GE Healthcare Dharmacon) was used in the experiments shown in Fig. 3 and Supplementary Fig. 6c while a pool of four scrambled siRNAs (#D-001206-13-05, GE Healthcare Dharmacon) was used in the experiments shown in Figs. 2, 4, 5, 6, and 7. In all experiments, siRNAs were transfected with DharmaFECT transfection reagent (#T-2001-02, GE Healthcare Dharmacon) according to the instructions of the manufacturer. For Supplementary Fig. 9b, the siRNA D-023324-05 (GE Healthcare Dharmacon) and Stealth RNAis (#HSS126645, #HSS184009, Thermo Fisher Scientific) were also transfected in HeLa cells to knockdown RNF213 with DharmaFECT and Lipofectamine 3000 transfection reagent (#L3000008, Thermo Fisher scientific), respectively. For Fig. 7, a commercially available siRNA pool was used to knockdown ISG15 (#M-004235-04-005, GE Healthcare Dharmacon). In these experiments, a reverse siRNA transfection protocol was adopted to knockdown the expression of ISG15 and RNF213 genes in HeLa cells prior to Listeria infection. A single siRNA, custom produced by Dharmacon, was used to knockdown RSV-N expression as described in ref. 104 . Immunoblotting assays were conducted to confirm reduction of protein expression levels. The aforementioned siRNAs are listed in Supplementary Table 3.
Lipid droplets isolation and cellular fractionation. THP-1 cells treated or not with interferon-β (100 U/mL; 8-10 h, #11343524, Immunotools) were washed and scraped in ice-cold hypotonic disruption buffer containing 20 mM tricine and 250 mM sucrose pH 7.8, 0.2 mM PMSF followed by homogenization in a chilled glass homogenizer. THP-1 lysates underwent a nitrogen bomb at a pressure of 35 bar for 15 min on ice to complete cellular disruption. Lysates were centrifuged at 3000 × g for 10 min at 4°C and the post-nuclear supernatants was loaded at the bottom of 13 mL centrifuge tubes and overlaid with ice-cold wash buffer containing 20 mM HEPES, 100 mM KCl and 2 mM MgCl2 at pH 7.4. Tubes were centrifuged in a swing-out rotor at 182,000 × g at 4°C for 1 h. Lipid droplets were collected from the top of the gradient. Lipid droplet-enriched samples were centrifuged at 20,000 × g for 5 min at 4°C to separate the floating lipid droplets from the aqueous fraction. Lipid droplets were washed three times by adding 200 µL ice-cold wash buffer followed by centrifugation at 20,000 × g for 5 min at 4°C. Lipid droplets were resuspended in 100 µL 2× Laëmmli buffer and further processed for immunoblot analysis as described above. To isolate the membrane fraction, THP-1 lysates were centrifuged at 3000 × g for 10 min at 4°C and the pellet was washed three times by resuspension with 1 mL of ice-cold wash buffer followed by centrifugation. The insoluble fractions were resuspended in 2× Laëmmli buffer and further processed for immunoblot analysis as described above. To isolate the cytosolic fraction, THP-1 lysates were centrifuged at 3000 × g for 10 min and 1 mL aliquot was sampled from the middle of the lysate. The 1 mL aliquot was further centrifuged at 270,000 × g for 1 h at 4°C in a microcentrifuge tube using a TLA100.3 rotor. The supernatants were collected and mixed to a final 2× Laëmmli buffer prior further processing for immunoblot analysis as described above. For cellular fractionation in HeLa cells, cells were grown on 100 cm 2 petri dishes and either untreated or treated for 24 h with interferon-β at 10 ng/ml (#11343524, Immunotools). Cells were processed as described in 106 . Briefly, cells were detached with trypsin, washed with PBS and lysed in a buffer containing 300 mM sucrose, 5 mM Tris-HCl, 0.1 mM EDTA, pH 7.4 and a protease inhibitor cocktail (Roche, cOmplete, Mini, EDTA-free tablet) followed by homogenization in a chilled glass homogenizer. Lysates were centrifuged at 800 × g for 8 min at 4°C followed by separation of the soluble and membrane fractions by a second centrifugation step at 20,000 × g for 120 min at 4°C. The membrane-associated fraction was redissolved in lysis buffer to a volume equal to the volume of the soluble fraction. Samples were mixed with Laëmmli buffer and processed for immunoblotting as described above.
Glycerol gradient analyses. Lysates prepared as described for the purification of lipid droplets were loaded onto a glycerol gradient (10-40% (w/v). The gradients were centrifuged at 237,000 × g for 20 h at 4°C, using a SW 55 Ti rotor and Beckman L-80 ultracentrifuge. Twenty fractions for each sample were collected in 2 mm increments using a BioComp Piston Gradient fractionator. Fractions were concentrated by TCA precipitation and analyzed by immunoblotting as described above. For immunoprecipitations, each gradient fraction was first desalted over Amicon filter to remove glycerol and resuspended in 1 ml of resuspension buffer containing 20 mM HEPES, 100 mM KCl, and 2 mM MgCl2 at pH 7.4. Resuspended fractions were precleared with Protein G-agarose beads by incubation for 1 h at 4°C. Immunoprecipitation of RNF213 was performed from precleared fractions by incubation with RNF213 antibody for 3 h at 4°C with end-over-end rotation. Beads were washed twice in 1 mL of resuspension buffer by centrifugation and samples were eluted into 1× Laëmmli buffer by heating at 60°C and processed for immunoblotting as described above.
In vitro infection with Listeria monocytogenes. Listeria monocytogenes (EGD BUG600 strain) was grown in brain heart infusion (BHI) medium at 37°C. Listeria were cultured overnight and then subcultured 1:10 in BHI medium for 2 h at 37°C. Bacteria were washed three times in PBS and resuspended in medium without FBS prior to infection. HeLa cells were grown in 6-well plates and infected with Listeria at a multiplicity of infection (MOI) of 25. Right after infection, plates were centrifuged at 1000 × g for 1 min followed by incubation for 1 h at 37°C to allow entry of the bacteria. Afterwards, cells were washed two times with PBS and then grown in MEM medium with 10% FBS, 1% glutamax, 1% nonessential amino acids, 1% sodium pyruvate, 1% hepes, supplemented with 40 µg/mL of gentamicin to kill extracellular bacteria. For immunoblotting, infected cells were washed with ice-cold PBS, lysed in 2× Laëmmli buffer and further processed as described above. To count the number of intracellular bacteria, HeLa cells were washed and lysed with miliQ water to release intracellular bacteria. Colony Forming Units (CFUs) were determined by serial dilution and plating on BHI agar.
In vivo infection with Listeria monocytogenes. Listeria monocytogenes (EGD BUG600 strain) was grown as described above. Female and Male C57BL/6 mice (Rnf213 +/+ or Rnf213 −/− ) between 8 and 12 weeks of age were infected intravenously by tail vein injection with 5 × 10 5 bacteria per animal. Mice were sacrificed 72 h following infection. CFUs per organ (liver or spleen) were enumerated by serial dilutions after tissue dissociation in sterile saline. The animals were housed in a temperature-(21°C) and humidity-(60%) controlled environment with 12 h light/dark cycles; food and water were provided ad libitum. The animal facility operates under the Flemish Government License Number LA1400536. All experiments were done under conditions specified by law and authorized by the UGent Institutional Ethical Committee on Experimental Animals.
In vitro infection with Herpes Simplex Virus Type 1 strain C12 (HSV-1 C12). HSV-1 C12 is a recombinant HSV-1 strain SC16 containing a cytomegalovirus (HCMV) IE1 enhancer/promoter-driven enhanced green fluorescent protein (EGFP) expression cassette in the US5 locus (kindly provided by Stacey Efstathiou (University of Cambridge, Cambridge, UK). HSV-1 C12 was propagated in Vero cells and titrated by plaque assay with Vero cells according to Grosche et al. 107 . HeLa cells were grown in 96-well plates and infected with HSV-1 C12 at MOI 0.1 (Fig. 4a-c) or MOI 0.05 (Fig. 4d, e). Cells were treated either without or with 1000 U/mL IFNα2a (#11343504, Immunotools). Cells were maintained in 2% FBS FluoroBrite DMEM Media. GFP fluorescence was monitored at 24 h intervals using the Infinite® 200 PRO Reader and Tecan iControl Software (Tecan Life Sciences). Each individual measurement (n = 4) was normalized to the mean value of the uninfected wells (n = 4). The relative area under the curve from 0 to 72 h post infection was calculated for each growth curve according to 64 . For immunoblot analysis, cells 24 h p.i were washed with ice-cold PBS, lysed in 2× Laëmmli buffer and further processed as described above.
In vitro infection with coxsackievirus B3 (CVB3). CVB3 Nancy strain was propagated in Green Monkey Kidney Cells and quantified by plaque assay in HeLa cells as described below. HeLa cells were grown in six-well plates and infected with CVB3 (MOI 0.01) in serum-free medium. After 1 h, the CVB3 inoculum was replaced and HeLa cells were cultured in MEM medium with 10% FBS, 1% glutamax, 1% nonessential amino acids, 1% sodium pyruvate, and 1% hepes. For immunoblotting, the cells were washed with ice-cold PBS, lysed in 2× Laëmmli buffer and further processed as described above to reveal VP1. For relative quantification of VP1, the intensity of the VP1 band was divided by the intensity of the GAPDH band, as loading control, and normalized to the intensity of the VP1 band in the control condition. For qRT-PCR, cells were lysed in 500 µl TRIzol™ Reagent. RNA was extracted from the aqueous phase after chloroform addition and precipitated by addition of isopropyl alcohol. Two hundred and fifty nanograms RNA per sample was used for reverse transcription to cDNA. For relative quantification, the ΔΔCt method was used with Hypoxanthine Phosphoribosyl transferase (HPRT) as housekeeping gene. Plaque-forming units (PFU)/mL were determined by serial dilution on confluent HeLa cells.
In vitro infection with respiratory syncytial virus (RSV). RSV-A2, an A subtype of RSV (ATCC, VR-1540, Rockville), was propagated in HEp-2 cells and quantified by plaque assay with A549 cells as described below. A549 cells were grown in 6well, 48-well, or 96-well plates, respectively, for immunoblot, titration or plaque assay. A549 cells were infected in serum-free medium with RSV-A2 at MOI 0.001, MOI 0.02 or MOI 0.005 for the respective experiment. After 4 h, the RSV inoculum was replaced and A549 cells were cultured in DMEM medium with 10% FBS. For immunoblot analysis, cells were washed with ice-cold PBS, lysed in 2× Laëmmli buffer and further processed as described above. In the plaque and titration assay, A549 cells were treated either without or with 10 ng/mL interferon-β (#11343524, Immunotools) for 48 h. For plaque assays, the DMEM medium with 10% FBS was supplemented with 0,6% (w:v) avicel RC-851 (FMC biopolymers) and the cells were fixed 6 days post RSV infection with 4% paraformaldehyde (PFA) in PBS and permeabilized with 0,2% Triton-X100 in PBS. RSV plaques were stained with a polyclonal goat anti-RSV serum and secondary anti-goat IgG. Plaques were visualized by TrueBlue TM Peroxidase substrate (5510, Sera Care). Each plaque was quantified in Fiji ImageJ 1.51n. In titration assays, Plaque Forming Units (PFU)/ mL were determined by serial dilution on confluent A549 cells. Supernatants were collected and mixed (1:1) with a 40% sucrose solution in HBSS, snap frozen and stored at −80°C.
In vivo infection with respiratory syncytial virus (RSV). An in-house generated mouse-adapted RSV-A2 108 was used for in vivo infection. RSV was propagated and quantified by plaque assay in HEp-2 cells as described above. Twelve-week-old C57Bl/6 J mice (Rnf213 +/+ and homozygous Rnf213 −/− littermates) were infected intranasally with 6 × 10 6 PFU after isoflurane-based sedation. Seventeen mice (9 wild-type (7 females, 2 males) and 8 knockout (5 females, 3 males) littermates were euthanized at day 5 post infection to determine the lung viral titers by serial dilution after tissue dissociation in sterile HBSS supplemented with 20% sucrose (Supplementary Fig. 12e). Ten mice (5 wild-type and 5 knockout littermates, 2 males and 3 females each) were used to follow-up bodyweight ( Supplementary  Fig. 12f) and survival rate ( Supplementary Fig. 12g) for 14 days post infection. The animals were housed in a temperature-(21°C) and humidity-(60%) controlled environment with 12 h light/dark cycles; food and water were provided ad libitum. The animal facility operates under the Flemish Government License Number LA1400536. All experiments were done under conditions specified by law and authorized by the UGent Institutional Ethical Committee on Experimental Animals.
Isolation and cell culture of BMDM's. Female and Male C57BL/6 mice (Rnf213 +/+ or Rnf213 −/− ) between 8 and 12 weeks of age were used to isolate bone marrow cells from femurs. Isolated bone marrow cells were cultured for 7 days at a density of 10 5 cells/mL in DMEM/F-12 medium, supplemented with 10% FBS, 10 units/ml penicillin, 10 μg/mL streptomycin, and 20 ng/mL murine M-CSF (#PMC2044, Thermo Fisher Scientific) at 37°C in a humidified atmosphere with 5% CO 2 . On day 3, fresh medium containing 40 ng/ml M-CSF was added and cells were further differentiated for 4 days.
Microscopy analysis of interferon-treated BMDM's. The day prior to an experiment, BMDMs were seeded in IBIDI μ-Slide eight-well chambers (#80826, Ibidi) and treated with oleic acid conjugated to BSA (molar ratio 5:1 in PBS) at a concentration of 200 µM, if mentioned. 24 h prior fixation, cells were treated with interferon-β (#11343524, Immunotools) at 10 ng/mL or left untreated. BMDMs were fixed in PBS with 4% PFA for 20 min at RT. LHCS LipidTOX™ Red Neutral Lipid Stain (#H34477, Thermo Fisher Scientific) and Hoechst (#H1399, Thermo Fisher Scientific) were used to visualize lipid droplets and nuclei respectively. Cells were mounted using Prolong Diamond Mounting Medium (#P36965, Thermo Fisher Scientific). Images were acquired using a confocal laser scanning microscope (LSM880 with Airyscan, Carl Zeiss Microscopy) and the software ZEN 2.3 Pro. Images were deconvoluted using Zen 3.1 (Blue Edition) for optimal resolution. The number and relative area of LDs were analyzed using Volocity 6.0 software.

Microscopy analysis of Listeria monocytogenes infected HeLa cells.
Overnight cultures of Listeria monocytogenes were diluted in Brain Heart Infusion media (BD) and grown to exponential phase (OD 0.8-1), washed three times in serum-free MEM (Life Technologies), and resuspended in serum-free MEM at corresponding MOIs. A fixed volume was then added to each well. Cells were centrifuged for 1 min at 201 × g to synchronize infection. The cells were then incubated with the bacteria for 1 h at 37°C, 5% CO 2 . Following this incubation, the cells were washed at room temperature with 1× DPBS, and then fed with full growth medium (with 10% fetal bovine serum) supplemented with 20 μg/ml gentamicin to kill extracellular bacteria.
HeLa cells were plated on coverslips in six-well dishes the day prior to an experiment and transfected with 2.0 µg of EGFP-RNF213 or eGFP-RNF213ΔC at the time of seeding with Fugene® HD (Promega) at a ratio of 3 uL reagent: 1 ug DNA. Listeria monocytogenes EGDe prfA* at MOI 1, EGDe prfA*ΔplcAΔplcBΔhly at MOI 75, or mCherry EGD at MOI 20 for 6 h, respectively. For experiments with eGFP-RNF213 and eGFP-RNF213ΔC, mCherry EGD infection of cells was performed at MOI 10 for 18 h (unless otherwise indicated). After infection, cells were subsequently fixed with 4% PFA (Electron Microscopy Sciences, Hatfield, Pennsylvania) in PBS for 20 min at room temperature and permeabilized with 0.5% Triton in PBS for 10 min at room temperature. Cells were then counterstained for Listeria (R11 anti-Listeria, raised in house) with goat anti-rabbit IgG (H + L), Superclonal™ Recombinant Secondary Antibody, Alexa Fluor 647 or ubiquitinated proteins (FK2, Millipore sigma) with goat anti-mouse IgG (H + L), Superclonal™ Recombinant Secondary Antibody, Alexa Fluor 647. Coverslips were mounted using ProLong Diamond Antifade Mountant with or without 4′,6-diamidino-2phenylindole (DAPI, Thermo Fisher Scientific). Images were acquired using an inverted wide-field fluorescence microscope (Axio Observer 7, Carl Zeiss Microscopy, Germany) equipped with an Axiocam 506 mono camera and the software ZEN 2.3 Pro. Where indicated images were deconvolved using Zen 3.1 (Blue Edition) for optimal resolution using default deconvolution settings (better, fast, regularized inverse filter).

Virotrap sample preparation for LC-MS/MS and immunoblotting analysis.
Virotrap experiments were performed as described in ref. 109 . Briefly, the day before transfection 10 6 Hek293T cells (authenticated) were seeded in four T75 flasks for each condition. Cells were then transfected with four different bait proteins, each fused via their N-terminus to HIV-1 GAG: mature ISG15 (ending on -LRLRGG), mature nonconjugatable ISG15 (ending on -LRLRAA), full-length ISG15 precursor, and eDHFR (dihydrofolate reductase from Escherichia coli) as control. VSV-G and FLAG-VSV-G were co-transfected to allow single-step purification of the produced particles. The day after transfection, these cells were treated with interferon-α for 24 h (10 ng/mL; #11343596, Immunotools) during the particle production phase. Supernatant was harvested 40 h after transfection followed by centrifugation and filtration (0.45 µM; SLHV033RS, Merck Millipore) to remove cellular debris. Virotrap particles containing protein complexes were purified in a single step using biotinylated anti-FLAG BioM2 antibody (#F9291, Merck) and Dynabeads MyOne Streptavidin T1 Beads (#65601, Thermo Fisher Scientific), and were consecutively eluted by competition using FLAG peptide (#F3290, Merck). For immunoblot analysis, samples containing the purified VLPs were mixed 1:1 with 4× Laëmmli buffer buffer complemented with 20 mM DTT further analyzed as described above. After FLAG elution, samples were processed with Amphipols A8-35 (#A835, Anatrace), digested using trypsin (V5111, Promega), and acidified. Each condition (4 baits, ±interferon-α) was analyzed in quadruplicate, leading to a total of 32 samples for LC-MS/MS analysis. Peptides were purified on Omix C18 tips (Agilent), dried and redissolved in 20 µl loading solvent A (0.1% trifluoroacetic acid in water/acetonitrile (ACN) (98:2, v/v)) of which 2.5 µl was injected for LC-MS/MS analysis.
GST pull-down for LC-MS/MS and immunoblotting. HEK293T, HeLa or differentiated THP-1 cells were grown at~80-90% confluence in 15 cm culture dishes (1 petri dish/sample). HEK293T cells were treated with interferon-α while HeLa and THP-1 cells were treated with interferon-β, at 10 ng/mL for 24 h (#11343596, #11343524, Immunotools). Each condition (2 baits, 3 cell lines) was analyzed in triplicate, leading to a total of 18 samples for LC-MS/MS analysis. Cells were washed three times with PBS with Ca 2+ and Mg 2+ and scraped in 1 mL lysis buffer containing 50 mM Tris.HCl (pH 8.0), 150 mM NaCl, 1% triton-x-100 (v/v), 1 mM PMSF, and a protease inhibitor cocktail (Roche, cOmplete, Mini, EDTA-free tablet, 4693159001). Samples were incubated 60 min on ice or 20 min under end-over-end agitation at 4°C. Lysates were cleared by centrifugation for 15 min at 16,000 × g at 4°C to remove insoluble components. Fifteen microliters of magnetic glutathione beads (Pierce Thermo Fisher Scientific) were incubated with 4 µg of purified GSTtagged ISG15 (R&D systems, UL-600-500) or GSTP1 (#G5663, Merck) under agitation. For immunoblotting, glutathione beads were also decorated with GSTtagged-SUMO1 (R&D systems, UL-710-500) or GST-tagged-Ubiquitin (R&D systems, U-540-01M). Incubation was performed overnight at 4°C in 500 µL 50 mM Tris.HCl, 150 mM NaCl, 1% triton-x-100 (v/v), 1 mM DTT after which decorated glutathione beads were blocked by addition of BSA to a final concentration of 2% (w/v) and incubation for an additional hour at 4°C. Then, the cleared lysates were incubated overnight at 4°C with the decorated glutathione beads under agitation to allow binding of proteins to ISG15. The following day, the beads were precipitated with a magnetic stand, washed once with 1 mL wash buffer containing 50 mM Tris.HCl (pH 8.0), 150 mM NaCl, 1% triton-x-100 (v/v). For immunoblotting, the beads were washed two extra times with 1 mL wash buffer for a total of three washes. Beads were mixed with 60 µL 2× Laëmmli buffer complemented with 20 mM DTT and boiled at 95°C for 10 min. Beads were precipitated with a magnet and the supernatants were further analyzed by immunoblotting as described above. In the LC-MS/MS analysis, beads were washed three extra times with 1 mL trypsin digestion buffer containing 20 mM Tris-HCl pH 8.0, 2 mM CaCl2. Washed beads were resuspended in 150 µl digestion buffer and incubated for 4 h with 1 µg trypsin (Promega) at 37°C. Beads were removed, another 1 µg of trypsin was added and proteins were further digested overnight at 37°C. Peptides were purified on Omix C18 tips (Agilent), dried and redissolved in 20 µl loading solvent A (0.1% trifluoroacetic acid in water/acetonitrile (ACN) (98:2, v/v)) injected for LC-MS/MS analysis. In the experiment with HeLa cells, 15 µL of samples was injected while 5 µL was injected for the experiment with THP-1 and Hek293T cells.
Immunoprecipitation (IP) for LC-MS/MS and immunoblotting. HEK293T and HeLa cells were grown at~80-90% confluence in 15 cm culture dishes (1 petri dish/sample). In the IP-immunoblotting experiments, cells were transfected either with FLAG-tagged-eGFP or FLAG-RNF213 plasmids, plus, the ISG15 conjugation machinery with HA-tagged-ISG15 matured, UBA7, UBCH8, and HERC5 plasmids. In the IP-MS experiment, cells were transfected either with plasmid coding for HAtagged-ISG15AA or a mock plasmid. HEK293T cells were treated with interferon-α at 24 h or 48 h while HeLa cells were treated with interferon-β at 24 h. Intereferonα/β treatment were performed at 10 ng/mL (#11343596, #11343524, Immunotools). Each condition (2 baits, 3 types of interferon-treated cells) was analyzed in triplicate, leading to a total of 18 samples for LC-MS/MS analysis. Cells were washed three times with PBS with Ca 2+ and Mg 2+ and scraped in 1.5 mL lysis buffer containing 50 mM Tris.HCl (pH 8.0), 150 mM NaCl, 1% triton-x-100 (v/v), 1 mM PMSF and a protease inhibitor cocktail (Roche, cOmplete, Mini, EDTA-free tablet, 4693159001). Samples were incubated 60 min on ice or 20 min under end-over-end agitation at 4°C. Lysates were cleared by centrifugation for 15 min at 16,000 × g at 4°C to remove insoluble components. Supernatants were incubated with Pierce anti-HA magnetic beads (#88836, Pierce Thermo Fisher Scientific) or anti-FLAG magnetic beads (#M8823, Merck). Lysates were incubated at 4°C under agitation overnight. The beads were precipitated with a magnetic stand and subsequently washed once with wash buffer containing 50 mM Tris.HCl (pH 8.0), 150 mM NaCl, 1% triton-x-100 (v/v). In the IP-immunoblotting experiments, the beads were washed two extra times with wash buffer. Beads were mixed with 60 µL 2× Laëmmli buffer and heated at 60°C for 10 min. Beads were precipitated with a magnet and the supernatants were further analyzed by immunoblot analysis as described above. In the IP-MS experiment, the beads were washed three extra times with trypsin buffer containing: 20 mM Tris-HCl pH 8.0, 2 mM CaCl 2 . The proteins were on-bead digested with 1 µg trypsin (Promega) for 4 h at 37°C under agitation. Then, the beads were precipitated and the supernatants were mixed with 1 µg of trypsin and further digested overnight at 37°C under agitation. Peptides were desalted on reversed phase C18 OMIX tips (Agilent), all according to the manufacturer's protocol. Purified peptides were dried, redissolved in 20 µl loading solvent A (0.1% trifluoroacetic acid in water/acetonitrile (ACN) (98:2, v/v)) and 5 µL were injected for LC-MS/MS analysis.
LC-MS/MS and data analysis. LC-MS/MS analysis was performed on an Ultimate 3000 RSLCnano system (Thermo Fisher Scientific) in line connected to a Q Exactive mass spectrometer or a Q Exactive HF (Thermo Fisher Scientific). Trapping was performed at 10 μl/min for 4 min in loading solvent A on a 20 mm trapping column (100 μm internal diameter (I.D.), 5 μm beads, C18 Reprosil-HD, Dr. Maisch, Germany) before the peptides were separated on a 150 mm analytical column packed in the needle (75 µm I.D., 1.9 µm beads, C18 Reprosil-HD, Dr. Maisch). Prior to packing of the column, the fused silica capillary had been equipped with a laser pulled electrospray tip using a P-2000 Laser Based Micropipette Peller (Sutter Instruments). Alternatively, peptides were separated on a 500 mm long micro pillar array column (µPAC™, PharmaFluidics) with C18-endcapped functionality. This column consists of 300 µm wide channels that are filled with 5 µm porous-shell pillars at an inter pillar distance of 2.5 µm. With a depth of 20 µm, this column has a cross section equivalent to an 85 µm inner diameter capillary column. Peptides were eluted from the analytical column by a nonlinear gradient from 2 to 55% solvent B (0.1% FA in water/ acetonitrile (2:8, v/v)) over 30, 60, or 90 min at a constant flow rate of 250 or 300 nL/min, followed by a 5 min in 99% solvent B. Then, peptides were eluted by a 5 min in 99% solvent B. The column was then re-equilibrated with 98% solvent A (0.1% FA in water) for 15 min. In the virotrap experiment, the mass spectrometer was operated in positive and data-dependent mode, automatically switching between MS and MS/MS acquisition for the 5, 10, or 12 most abundant ion peaks per MS spectrum. Full-scan MS spectra (400-2000 m/z) were acquired at a resolution of 70,000 (at 200 m/z) in the orbitrap analyzer after accumulation to a target value of 3E6 for a maximum of 80 ms. The 10 most intense ions above a threshold value of 1.7E4 were isolated in the trap with an isolation window of 2 Da for maximum 60 ms to a target AGC value of 5E4. Precursor ions with an unassigned, or with a charge state equal to 1, 5-8, or >8 were excluded. Peptide match was set on "preferred" and isotopes were excluded. Dynamic exclusion time was set to 50 s. Fragmentation were performed at a normalized collision energy of 25%. MS/MS spectra were acquired at fixed first mass 140 m/z at a resolution of 17,500 (at 200 m/z) in the Orbitrap analyzer. MS/ MS spectrum data type was set to centroid. The polydimethylcyclosiloxane background ion at 445.12002 was used for internal calibration (lock mass) in addition to the ion at 361.14660 corresponding to a tri-peptide of asparagine that was spike-in the MS solvents as described in 110 . Similar settings were used for the GST pull-down and immunoprecipitation experiments.
Data analysis was performed with MaxQuant (version 1.6.3.4) using the Andromeda search engine with default search settings including a false discovery rate set at 1% on the PSM, peptide and protein level. All spectral data files were searched with MaxQuant against all human proteins in the Uniprot/Swiss-Prot database (database release version of January 2019 containing 20,413 protein sequences (taxonomy ID 9606), downloaded from www.uniprot.org). For virotrap, the 32 recorded spectral data files were searched together and the search database was complemented with the GAG, VSV-G, and eDHFR protein sequences. For the GST pull-down experiments, three different searches were performed, one search for each cell line. For the IP experiments, three different searches were performed, two searches for HEK293T cells treated with interferon-α for 24 h and 48 h, and one search for HeLa cells treated with interferon-β for 24 h. For GST pull-down and IP experiment, each search comprised six spectral data files. The mass tolerance for precursor and fragment ions was set to 4.5 ppm and 20 ppm, respectively, during the main search. Enzyme specificity was set as C-terminal to arginine and lysine (trypsin), also allowing cleavage at arginine/lysine-proline bonds with a maximum of two missed cleavages. Variable modifications were set to oxidation of methionine (sulfoxides) and acetylation of protein N-termini. Matching between runs was enabled with an alignment time window of 20 min and a matching time window of 1 min. Only proteins with at least one peptide were retained to compile a list of identified proteins. In virotrap, 1214 proteins were identified in all 32 samples. In the GST pull-down, 599 proteins were identified in the experiment with THP-1 cells, 800 with HeLa cells, and 933 with HEK293T cells. In the IP experiments, 401 proteins were identified in the experiment with HEK293T cells treated with interferon-α for 24 h; 3795 proteins were identified in the experiment with HEK293T cells treated with interferon-α for 48 h; 1419 proteins were identified in the experiment with HeLa cells treated with interferon-β for 24 h. Proteins were quantified by the MaxLFQ algorithm integrated in the MaxQuant software. A minimum of two ratio counts from at least one unique peptide was required for quantification. Further data analysis was performed with the Perseus software (version 1.6.2.3) after loading the proteinGroups table from MaxQuant. Hits identified in the reverse database, only identified by modification site and contaminants were removed and protein LFQ intensities were log2 transformed. Replicate samples were grouped, proteins with less than three or four valid values in at least one group were removed and missing values were imputed from a normal distribution around the detection limit to compile a list of quantified proteins. In the virotrap experiment, 613 proteins were quantified. In the GST pull-down experiments, 595 quantified proteins in the experiment with THP-1 cells, 415 with HeLa cells, and 933 with Hek293T cells. In the IP experiments, 223 proteins were quantified in the experiment with HEK293T cells treated with interferon-α for 24 h; 2362 proteins were quantified in the experiment with HEK293T cells treated with interferon-α for 48 h; 547 proteins were quantified in the experiment with HeLa cells treated with interferon-β for 24 h. On the quantified proteins, for each ISG15 bait a t-test was performed for a pairwise comparison with the control condition to reveal specific ISG15 interaction partners. The results of these t-tests are shown in the volcano plot in Fig. 1 and Supplementary Figs. 1-3. For each protein, the log2 (ISG15/control) fold change value is indicated on the X-axis, while the statistical significance (−log p-value) is indicated on the Y-axis (Supplementary Data 1). Proteins outside the curved lines, set by an FDR value of 0.001 and an S0 value of 2 in the Perseus software, represent specific ISG15 interaction partners.
Statistics and reproducibility. Statistical tests, significant p values and number of replicates are indicated in the figure legends and are briefly described here. Western blot data are representative of three independent experiments. Proteomics experiments were performed with ≥3 biological replicates. HSV-1 infection experiments were performed with four biological replicates in three independent experiments unless otherwise noted. RSV infection experiments were performed with three biological repeats. CV and Listeria monocytogenes infection experiments were performed with three independent repeats, unless otherwise noted. Mouse experiments were performed with ≥4 animals. Nonparametric Mann-Whitney, one-tailed and two-tailed unpaired t-tests or ANOVA tests were performed using GraphPad Prism 9.2.0. P value thresholds used for the statistical tests corresponds to *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. The values of single datapoints and the exact p values are indicated in the source data.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The raw proteomics data have been deposited in the PRIDE database under accession codes PXD018345 and PXD018346. Data supporting the findings of this manuscript are available within the article, the Supplementary Information and the Source Data files or are available from the corresponding authors upon reasonable request. Source data are provided with this paper.