TRIM25 Enhances the Antiviral Action of Zinc-Finger Antiviral Protein (ZAP)

The host factor and interferon (IFN)-stimulated gene (ISG) product, zinc-finger antiviral protein (ZAP), inhibits a number of diverse viruses by usurping and intersecting with multiple cellular pathways. To elucidate its antiviral mechanism, we perform a loss-of-function genome-wide RNAi screen to identify cellular cofactors required for ZAP antiviral activity against the prototype alphavirus, Sindbis virus (SINV). In order to exclude off-target effects, we carry out stringent confirmatory assays to verify the top hits. Important ZAP-liaising partners identified include proteins involved in membrane ion permeability, type I IFN signaling, and post-translational protein modification. The factor contributing most to the antiviral function of ZAP is TRIM25, an E3 ubiquitin and ISG15 ligase. We demonstrate here that TRIM25 interacts with ZAP through the SPRY domain, and TRIM25 mutants lacking the RING or coiled coil domain fail to stimulate ZAP’s antiviral activity, suggesting that both TRIM25 ligase activity and its ability to form oligomers are critical for its cofactor function. TRIM25 increases the modification of both the short and long ZAP isoforms by K48- and K63-linked polyubiquitin, although ubiquitination of ZAP does not directly affect its antiviral activity. However, TRIM25 is critical for ZAP’s ability to inhibit translation of the incoming SINV genome. Taken together, these data uncover TRIM25 as a bona fide ZAP cofactor that leads to increased ZAP modification enhancing its translational inhibition activity.

The host factor and interferon (IFN)-stimulated gene (ISG) product, zinc-finger antiviral protein (ZAP), inhibits a number of diverse viruses by usurping and intersecting with multiple cellular pathways. To elucidate its antiviral mechanism, we perform a loss-of-function genome-wide RNAi screen to identify cellular cofactors required for ZAP antiviral activity against the prototype alphavirus, Sindbis virus (SINV). In order to exclude off-target effects, we carry out stringent confirmatory assays to verify the top hits. Important ZAP-liaising partners identified include proteins involved in membrane ion permeability, type I IFN signaling, and post-translational protein modification. The factor contributing most to the antiviral function of ZAP is TRIM25, an E3 ubiquitin and ISG15 ligase. We demonstrate here that TRIM25 interacts with ZAP through the SPRY domain, and TRIM25 mutants lacking the RING or coiled coil domain fail to stimulate ZAP's antiviral activity, suggesting that both TRIM25 ligase activity and its ability to form oligomers are critical for its cofactor function. TRIM25 increases the modification of both the short and long ZAP isoforms by K48-and K63-linked polyubiquitin, although ubiquitination of ZAP does not directly affect its antiviral activity. However, TRIM25 is critical for ZAP's ability to inhibit translation of the incoming SINV genome. Taken together, these data uncover TRIM25 as a bona fide ZAP cofactor that leads to increased ZAP modification enhancing its translational inhibition activity.

Author Summary
Organisms have evolved various innate strategies to defend against pathogens. During virus infection, the cell senses the viral nucleic acid and produces type I interferon, which alerts the neighboring cells. Signaling of type I interferon triggers expression of a wide a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Introduction
The recent re-emergence and spread of viruses beyond their normal geographic distribution have affected countries worldwide. Understanding the biology of host factors with broad antiviral activity is crucial to vaccine and drug development efforts to counteract existing and emerging viral infections.
As a first line of defense against viruses, the host produces type I interferons (IFNs), which signal through the JAK/STAT pathway to induce hundreds of IFN-stimulated genes (ISGs) that block various steps of the viral life cycle (reviewed in [1]). Among the ISGs, zinc finger antiviral protein (ZAP), encoded by the ZC3HAV1 gene, inhibits alphaviruses, filoviruses, hepatitis B virus, retroviruses, and the LINE-1 and Alu retroelements [2][3][4][5][6][7][8]. However, ZAP does not inhibit yellow fever virus, vesicular stomatitis virus, and herpes simplex virus 1 (HSV-1) [3]. It is not well understood what determines the broad yet specific antiviral activity of ZAP.
ZAP, also called PARP13, is a member of the poly(ADP-ribose) polymerase (PARP) family and is alternatively spliced. The long isoform of ZAP (ZAPL) contains a PARP-like domain on the C-terminus that is missing in the short isoform (ZAPS). This PARP-like domain is not enzymatically active [9], although exchange of the inactive catalytic triad in ZAPL to that of the active PARPs completely abolishes its antiviral activity [10], suggesting an important yet unknown role of the PARP-like domain in the antiviral function of ZAP. Several studies have demonstrated distinct activities for the two isoforms. ZAPL is more active against alphaviruses, such as SINV and Semliki Forest virus, than ZAPS, and carries signatures of positive selection [11,12]. While both isoforms are induced by IFN, ZAPS is upregulated more than ZAPL by virus infection and type I IFN [5,13,14].
Diverse cellular pathways have been implicated in ZAP's function (reviewed in [15]), but its precise mechanism is unknown. It is possible that ZAP interacts with multiple host factors, and the involvement of those factors in the viral life cycle is what provides the specificity. For example, ZAP binds RNA and recruits the exosome complex to target viral RNAs for degradation [5][6][7][16][17][18]. ZAP also directly inhibits translation of the incoming alphaviral genome [3], with interference in the interaction between eIF4A and eIF4G [19] implicated as one mechanism. In addition, ZAP synergizes with other ISGs for its maximal activity and upregulates RIG-I-mediated IFN-β production [14,20]. These studies support a model in which ZAP interacts with various host factors and cellular complexes to achieve an optimal antiviral state against diverse viruses.
In an attempt to unify the divergent pathways in which ZAP is involved and to uncover novel cofactors that are important for ZAP's inhibitory activity, we performed a genome-wide siRNA screen in a cell line inducible for ZAP expression. Large-scale RNAi screens allow us to take an unbiased approach to interrogate every gene in the genome. However, off-target effects lead to false positive hits and severely limit the value of genome-wide screens [21,22]. To address this we performed a rigorous set of confirmatory assays to verify the top hits and exclude off-target effects. We identified several genes that synergize with ZAP to target SINV or inhibit SINV independently of ZAP. Among the hits, TRIM25 was validated to be a cofactor of ZAP. TRIM25 is an E3 ubiquitin and ISG15 ligase, and is responsible for the polyubiquitination and activation of RIG-I [23][24][25]. We generated CRISPR clones in ZC3HAV1-knockout 293T cells where TRIM25 expression is significantly reduced and further confirmed that TRIM25 synergizes with both ZAPS and ZAPL to block SINV replication. Our data demonstrates that TRIM25 triggers ubiquitination of ZAP and enhances its antiviral activity through inhibition of viral translation, highlighting the importance of cofactors in the mechanisms of broadly antiviral proteins.

Results
A genome-wide siRNA screen reveals novel host factors that synergize with ZAP or exhibit ZAP-independent antiviral activity We performed a genome-wide screen with pooled siRNAs from Dharmacon to identify genes that are required for ZAP's antiviral activity (see Fig 1A for screen workflow). Viral replication is low in T-REx-hZAP cells when ZAP expression is induced, and silencing of ZC3HAV1 increases replication of a luciferase-encoding SINV, Toto1101/Luc, by 2 logs. The premise of the screen is as follows: should knockdown of an essential cofactor render ZAP nonfunctional, viral replication will be restored, resulting in increased luciferase activity (refer to "ZAP cofactor siRNA" column in Fig 1A). The screen was performed in triplicate to improve robustness, and we identified 480 genes, whose silencing significantly elevated SINV Toto1101/Luc replication with an average robust Z score of >3 ( Fig 1B). As expected, ZC3HAV1 was the top hit with an average robust Z score of 582.65; this was followed by BAI3 (165.56), TRIM25 (116.52) and RICS (100.42) (see S1 Table for the entire results).
Normalized percent activation (NPA) and robust Z score were utilized for hit selection [26]. The genes with the highest NPA and robust Z score >3 in all three replicate wells were chosen for validation in a secondary screen (91 genes). Since siRNAs can act as microRNAs and target genes non-specifically through their seed sequences [27], we included 4 genes that were potential off-target candidates from Haystack analysis (PDIK1L, SNAP25, FOXK1, DGAT2L3). In addition, we included 1 gene based on overlap with ISGs that we previously found as synergistic with ZAP in an overexpression screen (MAP3K14; [20]), and 6 genes from pathways that were significantly enriched but were not on the top 91 list (APC, FZD2, GFRA1, JAK1, SP1, and WNT8B).
We re-screened the candidate genes with a library of single siRNAs obtained from a different company (Ambion) to exclude hits that are mediated by off-target effects from further characterization. Knockdown of 5 genes by 6.25 nM siRNA (Fig 2A) and knockdown of 13 genes by 25 nM siRNA ( Fig 2B) significantly increased SINV Toto1101/Luc replication (see S2 Table for the entire results). Among them, ZC3HAV1 was identified as the top hit. TRIM25, KCNH5, GCS1 and JAK1 were also hits at both siRNA concentrations. In addition to the T-REx-hZAP cells used for the primary screen, a 293 clone that is inducible for the expression of rat ZAP C88R mutant, a dominant negative inhibitor of ZAP function [28], was also tested in parallel. Since endogenously expressed ZAP is antiviral, this cell line inhibited for ZAP function allowed us to identify hits with a ZAP-independent antiviral role. Silencing of GCS1 and GPRC5D by 6.25 nM ( Fig 2C) and 25 nM (Fig 2D) siRNAs significantly increased SINV Toto1101/Luc replication in these cells where ZAP is not functional (see S2 Table for the entire results), suggesting that they might inhibit SINV in a ZAP-independent manner. T-REx-hZAP cells transfected with control or gene-specific siRNA were treated with doxycycline to induce ZAPS overexpression one day post-transfection and infected with SINV Toto1101/Luc two days posttransfection. Cell lysates were harvested for measurement of luciferase activity at 24 h post-infection (p.i.). Relative luciferase units represent the level of SINV replication. Cells treated with the control non-targeting (NT) pooled siRNA have low SINV replication while ZAP knockdown by ZC3HAV1-specific pooled siRNA rescues viral replication by 2 logs. The large dynamic range in which hypothetical hits (ZAP cofactors) were identified is plotted on the right side of the graph. (B) Pooled siRNAs targeting the entire human genome (Dharmacon) were tested in triplicate and genes with an average robust Z score of greater than 3 are plotted. ZC3HAV1 is highlighted in red while the top hits immediately following ZC3HAV1 are highlighted in blue.  TRIM25 is a bona fide cofactor of ZAP Next, we validated the candidate ZAP cofactors in a lower throughput assay. Silencing of TRIM25, KCNH5, JAK1, and ZC3HAV1 led to increased virus replication for at least 2 of the 3 siRNAs (Fig 3A), which was consistent with reduced protein levels of TRIM25 (S1A Fig) and ZAP (S1B Fig), and reduced mRNA levels of KCNH5 and JAK1 (S1C Fig). Among the candidates, 3 independent siRNAs targeting TRIM25 significantly increased SINV Toto1101/Luc Candidate genes that significantly increased SINV replication in the secondary screen when silenced by individual siRNAs at both concentrations were validated in a larger scale 24-well plate format. Triplicate wells of T-REx-hZAP cells were transfected with the indicated siRNA, induced to express ZAPS, and infected with SINV Toto1101/Luc at a MOI of 10. Each symbol represents the value obtained from a single well after 24 h of infection. White circles represent results using pooled siRNA controls that were either NT or ZC3HAV1-specific. The data is representative of 3 independent experiments. Asterisks indicate statistically significant differences (Student's t-test, **, p<0.005; ***, p<0.0005; ****, p<0.0001). (B) Triplicate wells of T-REx-hZAP cells were transfected with the indicated siRNA, induced to express ZAPS, and infected with SINV Toto1101/Luc at a MOI of 10. Protein expression levels of TRIM25 and ZAP for the same transfections in a duplicate well were determined by immunoblotting. β-actin was used as a loading control. The data is representative of 4 independent experiments. The pvalue from Student's t-test is shown. (C) SINV replication in infected 293T cells in which ZC3HAV1 (left) or TRIM25 (middle) were silenced, and in ZC3HAV1-null 293T cells in which TRIM25 was silenced (right) is plotted. At 48 h post-transfection with siRNA, cells were infected with SINV Toto1101/Luc at a MOI of 0.01, and lysed at 6, 12, 24, and 40 h p.i. for measurement of luciferase activity. The data is representative of 3 independent experiments. Asterisks indicate statistically significant differences (two-way ANOVA, *, p<0.05; **, p<0.01; ****, p<0.0001).
replication by 10-to 26-fold compared to the NT control ( Fig 3A). Since TRIM25-targeting siRNA #3 was most efficient at rescuing SINV Toto1101/Luc replication, we designed and tested an additional C911 control. The 9 th to 11 th nucleotides of the siRNA were mutated to their complementary sequence, hence the designation C911, to rule out the possibility that the knockdown phenotype was due to the off-target effects of the siRNA [29,30]. The C911 control should lose its siRNA activity but still maintain its off-target effects as a microRNA. While TRIM25-targeting siRNA #3 rescued SINV Toto1101/Luc replication by about 1 log compared to the NT control, #3-C911 did not lead to increased viral replication and did not reduce the protein level of TRIM25 (Fig 3B), suggesting that the phenotype of siRNA #3 is due to specific silencing of TRIM25 and not inhibition of an off-target gene. Furthermore, TRIM25 was silenced in ZC3HAV1-knockout 293T cells [14], which were then infected by SINV Toto1101/ Luc to determine whether endogenously expressed TRIM25 has a ZAP-independent antiviral role. As controls, ZC3HAV1 and TRIM25 were silenced in 293T cells. While ZC3HAV1 ( Fig  3C; left) and TRIM25 (Fig 3C; middle) silencing in 293T cells restored SINV Toto1101/Luc replication by 1-3 logs by 40 h p.i., silencing of TRIM25 in ZC3HAV1-knockout cells did not further increase SINV Toto1101/Luc replication compared to the NT control (Fig 3C; right). These data suggest that TRIM25 requires ZAP for its anti-SINV activity.

TRIM25 interacts with ZAP through its SPRY domain
Next, we asked whether TRIM25 physically interacts with ZAP. We infected 293T with endogenous TRIM25 and ZAP expression with SINV, and immunoprecipitated TRIM25 to look for ZAP association at various time points following infection. We found that endogenous ZAP co-immunoprecipitated with endogenous TRIM25 over the course of SINV infection (Fig 4A). There were less TRIM25 and ZAP proteins present at 24 h p.i., which is consistent with the cytopathic effects observed at that time point, leading to less TRIM25 and ZAP pulldown. Previously, TRIM25 was found to interact with ZAP in the presence of RNA although only ZAPL was investigated [8]. To determine the interaction of TRIM25 and different ZAP isoforms, ZAPS or ZAPL, and/or TRIM25 were co-expressed in ZC3HAV1-knockout 293T cells, which were harvested for co-immunoprecipitation. Immunoprecipitation of both ZAPS and ZAPL by a monoclonal antibody recognizing the N-terminal portion of human ZAP (NZAP) pulled down TRIM25, although more TRIM25 is associated with ZAPL ( Fig 4B). TRIM25 was dramatically modified in the presence of ZAPS, evident by the presence in the whole cell lysates (WCL) of a ladder of bands larger than the molecular weight of TRIM25 (Fig 4B; WCL). TRIM25 consists of a RING domain, two B box domains, a coiled coil (CCD) domain, and a SPRY domain. The RING domain encodes the ubiquitin ligase activity while the CCD domain is required for oligomerization of TRIM proteins and the SPRY domain is important for mediating protein interactions and substrate specificity [31]. Next, we asked which domain in TRIM25 was responsible for interaction with ZAP. Since more ZAPL associates with TRIM25, we co-expressed similar levels of individual TRIM25 domains with ZAPL and immunoprecipitated ZAPL. We found that the SPRY domain of TRIM25 but not its RING and B box/CCD domains co-immunoprecipitated with ZAPL ( Fig 4C). These data suggest that both ZAP isoforms form a complex with TRIM25, likely through interaction with the SPRY domain of TRIM25.
also ubiquitinates ZAP and/or other proteins that complex with ZAP in order to stimulate ZAP's antiviral activity. To test this, we targeted TRIM25 by CRISPR in ZC3HAV1-knockout 293T cells to interrogate the functional interaction between ZAP isoforms and TRIM25. CRISPR targeting led to either in-frame deletions in all three chromosomal copies of TRIM25 (clone D) or frameshift insertions in two chromosomal copies and an in-frame deletion in one (clone F), consistent with the almost undetectable protein expression of TRIM25 (S2 Fig). Clones D and F are designated TRIM25 lo . Clone E is wild type and has similar TRIM25 protein expression as the parental ZC3HAV1-knockout 293T cells (S2 Fig). Both TRIM25 lo clones (D and F) were tested in the subsequent experiments and showed similar results. We transfected TRIM25 lo cells with ZAP-expressing constructs and infected them with SINV Toto1101/Luc. TRIM25 knockdown significantly enhanced SINV Toto1101/Luc replication in the presence of ZAPS and ZAPL overexpression ( Fig 5A). Furthermore, we co-transfected TRIM25 lo cells with TRIM25-and ZAP-expressing constructs to determine the antiviral effect of the different TRIM25 and ZAP combinations. FL TRIM25 alone inhibited SINV replication, likely due to overexpression ( Fig 5B). We found that FL TRIM25 enhanced the activity of ZAPS more than the RING-and CCD-deficient TRIM25 mutants at both high MOI ( Fig 5B)  This data suggests that both the E3 ligase activity and oligomerization of TRIM25 are important for the activation of ZAP. However, FL TRIM25 does not significantly enhance SINV inhibition by ZAPL (Fig 5B and S3A Fig), which hints at potential isoform-specific differences in the ZAP-TRIM25 synergy.

TRIM25 mediates ubiquitination of ZAPS and ZAPL
Since the E3 ligase-defective TRIM25 (ΔRING) failed to stimulate the activity of ZAPS ( Fig  5B), we hypothesized that TRIM25 might act by ubiquitinating ZAP and/or other host proteins. First, we asked whether ZAP is ubiquitinated. 293T cells transfected with a construct expressing HA-tagged ubiquitin were lysed under denaturing conditions to disrupt proteinprotein interactions, and endogenous ZAP was immunoprecipitated with an anti-ZAP polyclonal antibody and probed with an anti-HA antibody. A control anti-GFP antibody of the same species as the anti-ZAP antibody was used to check for non-specific pulldown. We found that ZAP was modified by ubiquitination at baseline and upon SINV infection ( Fig  6A). When ZAPS and ZAPL were overexpressed in the ZC3HAV1-knockout 293T cells in the presence of HA-tagged ubiquitin and subject to immunoprecipitation, both ZAP isoforms were found to be modified by ubiquitin, suggesting that the modified lysine(s) are likely shared by both isoforms and located in ZAPS ( Fig 6B). However, ZAPL is more polyubiquitinated than ZAPS, which hints to additional ubiquitination sites in the PARP-like domain. In addition, we asked whether TRIM25 is implicated in the ubiquitination of ZAP and determined the effect of both endogenous and overexpressed TRIM25 on ZAP ubiquitination level. Both ZAPS and ZAPL ubiquitination in the TRIM25 lo cells was reduced compared to that in the parental TRIM25 sufficient ZC3HAV1-knockout cells (Fig 6C). Consistent with that, overexpressed TRIM25 dramatically increased the level of modification of ZAP isoforms by both endogenous (Fig 6D) and overexpressed ubiquitin ( Fig 6E). Furthermore, we determined which linkage type of polyubiquitin TRIM25 induces on ZAP by overexpressing ubiquitin mutants, in which all the lysine residues are mutated except for K48 or K63. We immunoprecipitated ZAP isoforms that were co-expressed with TRIM25, and HA-tagged wild type or mutant (K48, K63) ubiquitin, and found that both ZAPS and ZAPL were ubiquitinated by both K48-and K63-linked polyubiquitin ( Fig 6F). Together, these data suggest that TRIM25 is responsible for ZAP modification.
Next, the lysine residues in ZAPS that were predicted to be ubiquitinated with medium to high confidence by UbPred and CKSAAP_UbSite were changed to arginine residues individually (K226R, K296R, K314R, K401R, K416R, K448R, and K629R) or in combination (K296R/ K448R, 7UbΔ) to determine whether potential ubiquitination at any of these sites affects ZAP's antiviral activity. Moreover, a previous study, in which a global approach was taken to identify all the ubiquitinated proteins in the cell, reported a tryptic peptide with a ubiquitinated lysine in ZAPL (EEGK 783 (glygly)LLFYATSR) [32]. We confirmed that this residue is ubiquitinated using a targeted mass spectrometry approach and the lysine was mutated to arginine in ZAPL (K783R). When we co-expressed the panel of ZAP ubiquitination site mutants with TRIM25 and immunoprecipitated ZAP to determine the level of modification, we found that the mutations diminished the level of ZAP ubiquitination to various degrees (S4A Fig). Most importantly, introduction of all 7 mutations at the same time almost completely abrogated ZAP ubiquitination (refer to "S 7UbΔ" in S4A Fig). However, the ZAPS 7UbΔ mutant was still It is possible that ZAP changes the interactome of TRIM25, which ubiquitinates and functionally modulates these interacting partners resulting in an antiviral state. To test this hypothesis, we co-immunoprecipitated TRIM25 in the absence or presence of ZAPS, of which the antiviral activity was more affected by the E3 ligase function of TRIM25 (Fig 5B), and identified by mass spectrometry proteins that interacted significantly more or less with TRIM25 upon ZAPS expression (S5 Fig). We found that most of these ZAP-mediated TRIM25 interacting partners are involved in mRNA metabolism and translation (S3 Table), which are cellular processes known to be affected by ZAP. However, the change in abundance of most of these proteins is not dramatic (S5 Fig), suggesting that the ZAP-mediated effects on TRIM25 targets might lie in their ubiquitination status. Taken together, we have shown that TRIM25 ubiquitinates ZAP, and that ZAP changes the interactions of TRIM25 with other host proteins that might contribute to the antiviral effects of the ZAP-TRIM25 synergy. Further work is required to determine whether these TRIM25 interacting partners are ubiquitinated and the functional consequences of their modification by TRIM25.

TRIM25 plays a critical role in ZAP inhibition of viral translation
In order to further elucidate the mechanism by which TRIM25 synergizes with ZAP, we investigated the effects of TRIM25 on different ZAP activities that were previously reported (reviewed in [15]). We knocked down TRIM25 in T-REx-hZAP cells and infected them with a temperature sensitive SINV that is unable to replicate at the non-permissive temperature. Since the antiviral mechanisms of ZAP include targeting of viral RNAs for degradation by the exosome complex and translational inhibition, we measured the level of SINV RNA and translation of the incoming viral genome over the course of infection. We found that the kinetics of viral RNA degradation was similar in the presence or absence of TRIM25 (Fig 7A). However, ZAP's ability to block SINV translation was significantly reduced by about 1 log in the absence of TRIM25 (Fig 7B; two-way ANOVA: p<0.0001), mirroring the magnitude of SINV inhibition upon TRIM25 knockdown in the same cells (Fig 3A). This data suggests that the mechanism by which TRIM25 enhances ZAP activity is through viral translational inhibition.

Discussion
We reported in this study that ZAP requires multiple host factors for its maximal antiviral activity. We identified novel partners of ZAP that are normally important for a range of cellular processes, such as membrane ion permeability, innate immune signaling, and post-translational protein modification. Among the hits, JAK1 is a kinase important for signaling of the type I IFN receptor and can potentially act by augmenting the stimulatory effects of ZAP on the RIG-I pathway. On the other hand, KCNH5 is an outward rectifying potassium channel and it is not clear how it can stimulate ZAP's function. It has been shown that reduction of the intracellular K+ concentration can activate the NLRP3 inflammasome, linking ion efflux to innate immunity [33]. It is interesting to note that known interacting partners of ZAP, such as RIG-I or components of the exosome complex, were not hits in the screen, although that is likely largely dependent on the type of assay used and the basal expression levels of genes that are knocked down. Moreover, although MAP3K14 was previously found to be synergistic with ZAP in an ISG overexpression screen [20], it was not a hit in our confirmatory screen. It is likely that synergistic effects are less pronounced in our screen where endogenous levels of proteins are being interrogated and the effect from silencing of one gene might be compensated by a homologous gene or another ZAP partner.
Our study shows for the first time that both ZAP isoforms are ubiquitinated, adding to the existing body of work on post-translational modifications of ZAP. Previous studies have shown that in addition to being post-translationally modified itself, ZAP is implicated in the modifications of other cellular and viral proteins. In some of these cases the modification regulates the function of ZAP or other proteins. For example, the C-terminal end of ZAPL is prenylated, and mutation of this prenylation site reduces the anti-SINV activity of ZAPL to a level similar to that of ZAPS, which is normally not prenylated [12]. It is possible that prenylation positions ZAPL in membrane compartments, allowing it to target viruses with an endocytic life cycle step. In addition, phosphorylation of ZAP by glycogen synthase kinase 3β positively modulates ZAP activity potentially by enhancing its ability to inhibit mRNA translation [34]. Moreover, when cells are stressed, ZAP localizes to cytoplasmic stress granules and is modified by poly-ADP-ribosylation [35]. Modified ZAP is implicated in the poly-ADP-ribosylation of Ago2, which correlates with derepression of miRNA-mediated translational silencing of cellular transcripts [13,35]. In particular, derepression of ISG transcripts results in viral inhibition, as demonstrated for HSV-1 and influenza A virus [13]. ZAP, specifically the long isoform, is also implicated in the poly-ADP-ribosylation and ubiquitination of influenza viral PB2 and PA polymerase proteins and their subsequent degradation [36]. Interestingly, another study reports that ZAPS inhibits influenza protein expression and is antagonized by the viral NS1 protein [37]. A NS1 mutant that lacks the ability to suppress the E3 ligase activity of TRIM25 [38] also loses the ability to antagonize ZAPS [37], suggesting that ZAPS could inhibit influenza virus through a mechanism that requires TRIM25-mediated ubiquitination, which is antagonized by NS1.
Although we clearly demonstrate here that TRIM25 is implicated in ZAP modification ( Fig  6C-6F), mutagenesis of ZAP ubiquitination sites does not impact its antiviral function (S4 Fig). One plausible explanation is that the E3 ligase activity of TRIM25 enhances ZAP function (Fig 5B) by ubiquitinating other host factors. It is likely that ZAP affects the interactions of TRIM25 with other proteins, resulting in changes of their modification and function and hence remodeling of the antiviral proteome in the cell. Since ZAP and TRIM25 synergize to block viral translation (Fig 7B), it is intriguing yet not unexpected to find that most of the TRIM25 interacting partners affected by ZAP are involved in mRNA metabolism and translation (S3 Table). Future work determining the ubiquitination status of these proteins will shed light on the impact of TRIM25 E3 ligase activity on the antiviral effects of ZAP.
Alternatively, TRIM25 and ZAPS can synergize to positively modulate the activity of RIG-I, leading to heightened innate immune signaling. About half of the TRIM family members have been shown to enhance innate immune responses and both ZAPS and TRIM25 physically interact with RIG-I to stimulate type I IFN response [14,23,39]. However, whether there is a requirement for RIG-I in ZAP function and vice versa are not settled. Inhibition of the RIG-I pathway and knockdown of RIG-I fail to abrogate ZAP-mediated HBV and XMRV inhibition, respectively [5,7]. Furthermore, RIG-I-dependent production of type I IFN and cytokines are not reduced in ZAP-deficient primary mouse cells [40]. Hence, the contribution of TRIM25 and ZAP synergy to innate immunity warrants further investigation.
Our data also suggest differences in the interaction between TRIM25 and ZAPS and ZAPL, consistent with previous studies demonstrating distinct activities for the ZAP isoforms. TRIM25 was dramatically modified in the presence of ZAPS, although these modified forms were not associated with ZAPS ( Fig 4B). On the other hand, ZAPL expression did not result in extensive modification of TRIM25 and more TRIM25 was found to associate with ZAPL ( Fig  4B). This suggests that in addition to the shared domains in ZAPS and ZAPL, ZAPL might interact with TRIM25 through its PARP-like domain. These observations argue that ZAP interacts primarily with unmodified TRIM25, although the expression of the ZAP isoforms differentially regulates TRIM25 post-translational modification. Based on these observations, we postulate that ZAP might affect the ubiquitination status of TRIM25 and as a result modulate the innate immune response. It has been shown that ubiquitin-specific peptidase 15 deubiquitinates TRIM25, leading to its stabilization and sustained RIG-I signaling [41]. However, when we performed mass spectrometry on the immunoprecipitates of TRIM25, we did not see a significant increase in ubiquitinated TRIM25 peptides in the presence of overexpressed ZAPS. It is possible that the increased modification of TRIM25 is due to something other than ubiquitin. Future studies are needed to determine the impact of ZAP isoforms on post-translational modification of TRIM25. In addition, even though endogenous TRIM25 is required for the function of both ZAPS and ZAPL (Fig 5A), overexpressed TRIM25 stimulates the antiviral activity of ZAPS to a greater extent than that of ZAPL (Fig 5B), which is likely due to ZAPL's greater baseline inhibitory effect compared to ZAPS in the absence of TRIM25.
In conclusion, our study has uncovered a novel requirement for TRIM25 in ZAP function and elucidated the mechanism of this synergy. Recent outbreaks, such as chikungunya virus (Alphavirus) in the Caribbean and U.S., Ebola virus (Filovirus) in West Africa, and Zika virus (Flavivirus) in South and Central America, prompts the development of therapeutic strategies for disruption of crucial virus-host interactions. Given that the replication of these viruses greatly depend on the host factor repertoire of the target cells, our study is highly relevant and advances our understanding of host factor contributions to innate immune responses.

Genome-wide siRNA screen
For the primary ZAP cofactor screen, siRNAs from the siGENOME library targeting the whole human genome (Dharmacon; pools of 4 siRNAs per gene) were pre-arrayed in fiftyeight 384-well plates, mixed with DharmaFECT 1 Transfection Reagent (diluted 1:100), and transferred to 384-well assay plates (5 μl/well; 25 nM final) with a Janus automated workstation (PerkinElmer). The entire screen was carried out in triplicate by reverse transfection. siGEN-OME NT and ZC3HAV1-targeting siRNA smartpools were used as negative and positive controls (5 wells per control per plate; see the position of the controls in S1 Table). A total of 7500 T-REx-hZAP cells were seeded in DMEM with 3% FBS, 5 μg/ml blasticidin, and 200 μg/ml Zeocin (25 μl/well) with a MultiDrop Combi liquid dispenser (Thermo Fisher Scientific). One day post-plating, ZAP expression was induced by addition of 1 μg/ml doxycycline (5 μl/well). Since removing the media from 384-well plates and adding virus inoculum in a minimal volume to allow adsorption was not feasible, cells were infected with a highly concentrated PEGprecipitated SINV stock at a MOI of~200 (5 μl/well) two days following siRNA treatment. The MOI was calculated based on titration of the viral stock under standard infection conditions, and due to dilution was likely much lower under the conditions used in the screen. Twentyfour hours p.i., cells were lysed with Steady-Glo Luciferase Assay Substrate (Promega; 20 μl/ well) and luminescence measured on an EnVision plate reader (PerkinElmer). For each well, the raw luciferase value reflects the level of SINV infection. The raw luciferase values and the plate median from each of the 384-well plates were used to calculate the median absolute deviation (MAD): 1.4826 x MEDIAN (each raw value − plate median). The robust Z score was then calculated by dividing the difference between each raw luciferase value and the plate median by MAD [26]. Normalized percent activation (NPA) was also calculated for each well as follows: 100 Ã (X i −μ c-)/(μ c+ −μ c-), where X i = raw luciferase value, μ c+ = mean positive controls and μ c-= mean negative controls. The entire screen was performed in triplicate and the Z' and strictly standardized mean difference (SSMD) were calculated for every plate [26]. A Z' factor of >0.5 and a SSMD of >3 were used as cutoffs for assay validation [26,48,49]. Those wells with the highest NPA and robust Z score >3 in all three replicates were considered 'hits'.
For the secondary screen, Silencer siRNAs (Ambion; 3 individual siRNAs per gene) were tested in 384-well plates in triplicate at two different concentrations: 6.25 nM (the concentration of each single siRNAs in the original Dharmacon pools) and 25 nM (the concentration of the original Dharmacon pools used in the primary screen). In order to rule out ZAP-independent effects of the top hits, the secondary screen was carried out in two different cells lines: T-REx-hZAP and T-REx-rZAPC88R. Genes with an average Z score of >3 for at least two out of the three siRNA sequences were considered 'hits'.

Bioinformatics and statistical analyses
Pathway analysis was performed using Enrichr [50]. All genes with an average robust Z score >3 were normalized using -5 as the minimum (essentially the value of the negative controls) and the mean value for ZC3HAV1 as the maximum, scaled between 0 and 1, and exported as a comma-separated table with the normalized score and gene symbol, a format recognized by Enrichr. Genes belonging to pathways that were significantly enriched and overlapping with the list of genes with the highest NPA and robust Z score >3 in at least 1 out of 3 replicates were validated in the secondary screen. Haystack analysis was performed to identify the most statistically significant genes whose predicted knockdown via off-target effects was correlated with the phenotypes observed in the primary screen [27]. Differences between experimental conditions during the course of infection were determined using two-way ANOVA. Differences between experimental conditions at a single time point were determined using an unpaired, two-tailed Student's t-test with a 95% confidence interval.

Plasmid transfection
Cells were transfected in different combinations with constructs expressing V5-tagged TRIM25 (FL) or derivatives (RING, B box/CCD, SPRY, ΔRING, and ΔCCD), ZAPS or ZAPL, and HA-tagged wild type ubiquitin or mutants with X-tremeGENE 9 DNA Transfection Reagent (Roche Life Science) at a ratio of 3 μl reagent to 1 μg DNA. Total plasmid amount in co-transfections was kept constant by transfecting cells with empty vectors (pIRES-puro, pTRIPZ and pcDNA3.1).

TRIM25 targeting by CRISPR
Guide RNAs (gRNAs) targeting exon 1 of the human TRIM25 gene were designed by the MIT Optimized CRISPR Design portal (http://crispr.mit.edu/), and two with the least predicted offtarget effects (gRNA #1: GTCGCGCCTGGTAGACGGCG; gRNA #3: GAGCCGGTCACCA CTCCGTG) were selected for cloning into the Cas9-expressing PX459 vector. Oligos containing the gRNA sequences (gRNA #1: 5'-CACCGGTCGCGCCTGGTAGACGGCG-3' and 5'-AAACCGCCGTCTACCAGGCGCGACC-3'; gRNA #3: 5'-CACCGGAGCCGGTCACCAC TCCGTG-3' and 5'-AAACCACGGAGTGGTGACCGGCTCC-3') were ligated and cloned into PX459 linearized with BbsI. ZC3HAV1-knockout 293T cells were transiently transfected with PX459 expressing gRNA #1 or 3, and one day after transfection selected under 1 μg/ml puromycin for 2 days to eliminate cells that were not transfected. Surviving cells were then counted, diluted to 0.5 cell/well in a 96-well plate and seeded in 10% FBS DMEM. Single cell clones were marked and allowed to expand. Several clones per gRNA #1 were treated with or without puromycin and the ones that were sensitive to puromycin, suggesting that the clones had not integrated the gRNA-expressing vector, were harvested for immunoblot analysis to evaluate TRIM25 expression. Two knockdown clones (D and F) and one wild type clone (E) were selected (see S2 Fig). Genomic DNA was isolated from these clones, and a 600bp sequence flanking the gRNA targeting site was amplified by PCR and cloned into TOPO vector (Thermo Fisher Scientific). Sequencing of TOPO clones confirmed that all three chromosomes were targeted in clones D and F resulting in insertions and/or deletions in exon 1 of TRIM25, while clone E is wild type.

Quantitative reverse transcription PCR (RT-qPCR)
Total RNA was isolated from siRNA-treated cells using the RNeasy mini kit (Qiagen). 1 μg of input RNA was used as a template for reverse transcription using SuperScript III (Invitrogen, Carlsbad, CA) and random hexamers. RT-qPCR was performed using 5 μl of 10-fold-diluted cDNA and primers targeting JAK1 (5'-CCACTACCGGATGAGGTTCTA-3' and 5'-GGGT CTCGAATAGGAGCCAG-3'), KCNH5 (5'-CCGTGTGGCTAGGAAACTGG-3' and 5'-CAATGACCTCGTAGTCTCCGA-3'), and RPS11 (5'-GCCGAGACTATCTGCACTAC-3' and 5'-ATGTCCAGCCTCAGAACTTC-3' [53]) in a SYBR Green qPCR assay on the LightCycler 480 Real-Time PCR System (Roche Applied Sciences, Indianapolis, IN). qPCR conditions were as follows; initial denaturation step at 50˚C for 2 min and 95˚C for 10 min, then 45 cycles of 95˚C for 15 sec, 56˚C for 15 sec, and 72˚C for 20 sec, and followed by a melting step of 95˚C for 10s, 65˚C for 10s and a 0.07˚C/s decrease from 95˚C, and a cooling step of 50˚C for 5s. Transcript levels of JAK1 and KCNH5 were determined by normalizing the target transcript CT value to the CT value of the endogenous housekeeping RPS11 transcript. This normalized value was used to calculate the fold change relative to the average of cells treated with the NT siRNA control (CT method).
To determine SINV RNA levels in TRIM25-targeting or NT siRNA treated T-REx-hZAP cells over the course of infection with Toto1101/Luc:ts6, 1 μg of total cellular RNA was used in a one-step quantitative real-time PCR assay using primers and a Taqman probe targeting the nsP2 region of SINV. Primer pairs for SINV Taqman  Clones were sequence verified and packaged into lentiviruses by co-transfection with Gag-Pol and VSV-G-expressing plasmids into 293T cells [20]. ZC3HAV1-knockout 293T were transduced with lentiviruses carrying RFP alone, ZAPS, ZAPL, and various ubiquitination site mutants, and 2 days later infected with TE/5'2J/GFP at a MOI of 10. Cells were harvested at 6-8 h p.i. and fixed in 1% paraformaldehyde for flow cytometry analysis. Data was acquired using a BD LSRII and FACS Diva software (version 8.0) and analyzed using FlowJo 8.8.7 (TreeStar Inc.). Percent infected (GFP+) cells in the transduced (RFP+ or RFP-tagged ZAP+) population was calculated for ZAPS, ZAPL and their mutants.

Mass spectrometry (MS)
For the ZAP ubiquitination experiment, based on alignment of the ZAP IP Western Blot, SDS-PAGE gel bands corresponding to ZAP were excised and trypsinized as described previously [54]. For the co-immunoprecipitation experiment, TRIM25 and associated proteins were co-immunoprecipitated and eluted from anti-V5 antibody-crosslinked M-270 Epoxy Dynabeads (Thermo Fisher Scientific) using 8M Urea (GE Healthcare Life Sciences), reduced, alkylated and digested with Endopeptidase Lys-C (>4M Urea) for 6 hours followed by overnight trypsinization (>2M Urea).
Peptides were analyzed by reversed phase nano LC-MS/MS (Ultimate 3000 nano-HPLC system coupled to a Q-Exactive Plusor a Fusion Lumos mass spectrometer, operated in high/ low mode, Thermo Fisher Scientific). Known (EEGK 783 (glygly)LLFYATSR [32]) and predicted ubiquitinated ZAP peptides were analyzed by parallel reaction monitoring (PRM) [55] while data dependent acquisition (DDA) was used for unknown ZAP ubiquitination sites.
Peptides generated from the co-immunoprecipitation experiment were analyzed in DDA mode. All peptides were separated on a C18 column (12 cm / 75 μm, 3 μm beads, Nikkyo Tecnologies) at 300 nl/min with a gradient increasing from 2% Buffer B/98% buffer A to 40% buffer B/60% Buffer A in 37 min or 80 min (buffer A: 0.1% formic acid, buffer B: 0.1% formic acid in acetonitrile).
For data analysis, ubiquitination focused DDA data was extracted and queried against Uni-Prots complete human proteome database (March 2016) concatenated with common contaminants [56] using Proteome Discoverer 1.4 (Thermo Fisher Scientific)/MASCOT 2.5.1 (Matrix Science). Protein N-terminal acetylation, oxidation of methionine, and di-glycine modification of lysine were allowed as variable modifications. In cases where proteins were reduced (DTT) and alkylated (iodoacetaminde), carbamidomethylation was included as a variable modification of cysteines and lysines. 10 ppm and 20 mDa were used as mass accuracy for precursors and fragment ions, respectively. Matched peptides were filtered using 1% False Discovery Rate calculated by Percolator [57] and in addition requiring that a peptide was matched as rank 1 and that precursor mass accuracy was better than 5 ppm. Data from the co-immunoprecipitation experiment, in biological triplicates, was analyzed by MaxQuant v.1.5.3.28 using match between runs. Search criteria similar to above were used. Fold differences were calculated based on label-free quantification (LFQ) values (http://www.ncbi.nlm.nih.gov/pubmed/ 24942700) while intensity-based absolute quantitation (iBAQ) values (https://www.ncbi.nlm. nih.gov/pubmed/21593866) were used to estimate abundance of different proteins. The statistical packet Perseus (https://www.ncbi.nlm.nih.gov/pubmed/27348712) was used for data analysis.
Supporting Information S1 Fig. Validation of TRIM25, ZC3HAV1, KCNH5 and JAK1 silencing. Knockdown of (A) TRIM25 and (B) ZC3HAV1 by 3 individual siRNAs in T-REx-hZAP cells was validated by immunoblot analyses using the indicated antibodies. β-actin was used as a loading control. Pooled siRNAs that were NT or targeted ZC3HAV1 were used as negative and positive controls for siRNA transfection, respectively. (C) Knockdown of KCNH5 and JAK1 by 3 individual siR-NAs in T-REx-hZAP cells was validated by RT-qPCR. Mean fold change relative to the NT control pool treated cells ± standard deviation is plotted. (TIF)

S2 Fig. Validation of TRIM25 knockout in CRISPR clones. (A)
CRISPR-targeting region in the genomic sequence of TRIM25 is shown in clones D, E and F. Clone E has the wild type sequence. The alignment shown is in the same reading frame of the wild type TRIM25 protein.
A red dash represents a deletion whereas a red nucleotide represents an insertion when compared to the wild type TRIM25 sequence. (B and C) Protein level of TRIM25 in the parental ZC3HAV1-knockout 293T, and CRISPR clones E, D and F. Clone E has similar TRIM25 expression as the parental cells whereas TRIM25 expression is significantly lower in clones D and F (designated TRIM25 lo ) that are mutated for wild type TRIM25. β-actin was used as a loading control. Short and long exposures are shown in (B) and (C), respectively.  Table. Results of the primary screen. The raw luciferase data collected from the T-REx-hZAP cell line, and the corresponding library gene list and controls are shown. Raw luciferase values are color-coded based on their magnitude (red: high; yellow: medium; green: low). Each column represents a different 384-well plate. A total of 58 plates were tested (H101-H158) in triplicates (A, B and C). (XLSX) S2 Table. Results of the secondary screen. The raw luciferase data collected from T-REx-hZAP and T-REx-rZAPC88R cell lines and the corresponding library gene list are shown. (XLSX) S3 Table. ZAP-mediated TRIM25 interactome. Host proteins that were significantly (Student's t-test, p<0.05) up-or down regulated by 1.5 linear fold or more in TRIM25 pulldown in the presence of ZAPS are shown. The proteins with the most significant p-value are listed first. LFQ and iBAQ values were used to calculate fold differences of proteins between samples with and without ZAPS, and abundance of different proteins, respectively (see the section on MS in Materials and Methods). (XLSX)