To investigate the role of ZAP and TRIM25 RNA binding in the context of viral translation, we generated a panel of constructs with mutations previously demonstrated to impact RNA binding. For ZAP, each of the following mutations was introduced in both ZAPS and ZAPL to probe potential isoform differences in RNA binding and antiviral function. These mutations fall into two general categories: 1) ZnF mutants that individually disrupt each CCCH motif and 2) CpG RNA binding cavity mutants. We made four individual mutations to disrupt each N-terminal CCCH ZnF: H86K, C88R, C168R, and H191R, which are located in ZnF 1, 2, 3, and 4, respectively (Guo et al., 2004). Two putative RNA binding cavities were previously identified based on a crystallized NZAP structure without bound RNA (Chen et al., 2012). Based on this study, we also made two triple mutations in the two predicted RNA binding cavities contained within the ZnFs: V72A/Y108A/F144A (abbreviated as VYF), found within ZnFs 2-3, and H176A/F184A/R189A (abbreviated as HFR), found within ZnF 4 (Chen et al., 2012). More recent studies have elucidated structures of NZAP bound to CpG-containing RNA (Meagher et al., 2019; Luo et al., 2020). Building on this work, we made two double mutations and one triple mutation within the ZnFs that mediate CpG dinucleotide-specific binding. These mutations are C96A/Y98A (abbreviated as CY) and K107A/Y108A (abbreviated as KY) within ZnF 2, and E148A/K151A/R170A (abbreviated as EKR) within ZnF 3, which demonstrate a range of RNA binding and antiviral activities (Meagher et al., 2019; Luo et al., 2020). NZAP mutants CY, KY, and EKR previously demonstrated loss of antiviral activity against a SINV NanoLuc luciferase reporter virus, with EKR exhibiting the least defect as compared to NZAP WT (Luo et al., 2020). Notably, the individual mutations Y108A/F and F144A/Y appear to be critical for ZAP recognition of CpG-rich RNA, wherein they completely or partially abolish ZAP’s ability to discriminate between CpG-rich and CpG-deficient strains of HIV-1, respectively (Meagher et al., 2019; Luo et al., 2020).

For TRIM25, we generated two constructs, one in which we replaced the lysine-rich motif in the L2 linker with alanines (TRIM25 7KA) and one in which we deleted the previously identified RNA binding domain in the PRY/SPRY domain (TRIM25 ΔRBD) (Choudhury et al., 2017; Sanchez et al., 2018).

We cloned each of these ZAP and TRIM25 WT or RNA binding mutants into a pcDNA3.1-3XFLAG or -myc plasmid, allowing for transient expression of each construct following transfection into ZAP or TRIM25 KO 293T cells. We titrated the amount of plasmid to transfect for each construct to ensure even expression across constructs for ZAPS, ZAPL, and TRIM25 WT ( Supplemental Figure 1 ). Because the ZAP CpG RNA binding cavity mutants generally express at higher levels than the ZnF mutants ( Supplemental Figure 1 ), we decided to transfect two amounts of ZAPS and ZAPL WT in our assays to match these two expression patterns. We then assessed the ability of the ZAP and TRIM25 mutants to bind SINV genomic RNA by an in vitro RNA pull-down assay. We incubated lysates of cells transfected with ZAP or TRIM25 mutants with biotin-labeled SINV (Toto1101 strain) genomic RNA, allowing for RNA and bound protein to be immunoprecipitated using streptavidin beads and probed for the presence of bound ZAP or TRIM25. As a negative control, we also assessed the ability of the WT constructs to bind firefly luciferase (Fluc) RNA. We quantified the resultant ZAP and TRIM25 bound to RNA and normalized to input ZAP and TRIM25 protein levels with ImageJ (Davarinejad, 2015). The ZnF mutations in both ZAPS and ZAPL drastically reduce all SINV RNA binding, with only the ZAPS and ZAPL ZnF 1 mutant (H86K) and the ZAPS ZnF 3 mutant (C168R) showing low levels of binding ( Figure 1A ). In contrast, mutations in the CpG RNA binding cavities result in a range of binding phenotypes. For both ZAPS and ZAPL, the VYF, CY, and KY mutants show complete to near complete loss of SINV RNA binding, while the HFR and EKR mutants show similar or increased RNA binding relative to ZAPS and ZAPL WT ( Figure 1B ).

The TRIM25 mutants demonstrate diverging RNA binding phenotypes. The 7KA mutant not only has increased binding to SINV RNA relative to WT, but also to the Fluc negative control ( Figure 1C ). Consistent with previous findings, the ΔRBD mutation abolishes binding to SINV RNA ( Figure 1C ). Together, our findings indicate that the different RNA binding residues of ZAP and TRIM25 contribute in varying degrees to viral RNA binding.

Given that TRIM25 RNA binding has been purported to stimulate its interaction with ZAP (Choudhury et al., 2017), and that the ZAP ZnFs responsible for RNA binding are also thought to mediate its interaction with TRIM25 (Gonçalves-Carneiro et al., 2021), we hypothesized that abolishing ZAP RNA binding would negatively impact ZAP association with TRIM25, and vice versa. We also aimed to capture any ZAP isoform-specific characteristics for association with TRIM25, given previous reports that TRIM25 preferentially interacts with ZAPL (Li et al., 2017; Kmiec et al., 2021). Therefore, we expected that ZAP RNA binding mutants would display decreased enrichment in the presence of TRIM25 co-immunoprecipitation (co-IP) compared to ZAP WT. In accordance with prior work and our observation of different RNA binding activities of the TRIM25 mutants ( Figure 1C ), we also expected that TRIM25 7KA and ΔRBD would behave differently.

To test this hypothesis, we transfected ZAP KO 293T cells with either FLAG-tagged ZAPS or ZAPL RNA binding mutants and myc-tagged TRIM25 WT. Given that the ZAP ZnFs 2-4 display approximately equal, near complete loss of RNA binding ( Figure 1A ), and that ZnFs 2 and 4 are more important for mediating ZAP antiviral activity against alphaviruses and for CpG specificity (Bick et al., 2003; Meagher et al., 2019; Luo et al., 2020), we proceeded with testing only ZnF mutants 2 and 4 in addition to the complete panel of CpG RNA binding cavity mutants. We then performed a myc IP to enrich for TRIM25 and probed for the presence of associated ZAPS or ZAPL, quantifying resultant ZAP pulldown with ImageJ (Davarinejad, 2015). Surprisingly, we found that both ZnF mutants (C88R and H191R) in both ZAPS and ZAPL display increased association with TRIM25 ( Figures 2A, B ). Of the remaining RNA binding mutants, VYF, CY, and KY with near complete loss of RNA binding ( Figure 1B ) display similar to increased TRIM25 association for both ZAPS and ZAPL, though to a lesser extent than the ZnF mutants ( Figures 2A, B ). Only HFR with unaffected or mildly reduced RNA binding ( Figure 1B ) exhibits diminished interaction with TRIM25 as compared to ZAPS and ZAPL WT ( Figures 2A, B ), while EKR which mostly retains or even gains binding to SINV RNA ( Figure 1B ) binds to TRIM25 to a similar degree as compared to ZAP WT ( Figures 2A, B ).

To test the ability of TRIM25 RNA binding mutants to interact with ZAP, we transfected TRIM25 KO 293T cells with myc-tagged TRIM25 RNA binding mutants and FLAG-tagged ZAPS or ZAPL WT. When we transfected amounts that would yield similar levels of TRIM25 expression ( Figure 1C ), we found that these levels are too low to visualize previously demonstrated interactions between TRIM25 WT and ZAPS or ZAPL (data not shown). In our hands, the TRIM25 ΔRBD mutant is markedly less stable than the other TRIM25 constructs. Therefore, we maximized the transfected amounts of all TRIM25 forms to better visualize any differences in TRIM25 association with ZAP, normalizing the TRIM25 co-IP to input lysates and quantifying with ImageJ (Davarinejad, 2015). As expected, we observed that TRIM25 7KA binds more robustly than TRIM25 WT to both ZAPS and ZAPL, while the TRIM25 ΔRBD levels are likely too low to be detected in the ZAP co-IP ( Figure 2C ).

Together, these data suggest that ZAP RNA binding may compete with ZAP-TRIM25 interaction likely because ZAP interacts with TRIM25 through its N-terminal ZnFs. Moreover, there do not appear to be ZAP isoform-specific differences for the RNA binding mutants as a whole, with trends of increased or decreased TRIM25 association holding true for each mutant in ZAPS and ZAPL.

Next, we asked how loss of RNA binding would affect the ability of ZAP and TRIM25 to inhibit SINV replication. Given the centrality of RNA binding to ZAP antiviral activity and to TRIM25 ligase activity, we hypothesized that mutations with near complete loss of RNA binding would abolish antiviral activity completely, while mutations with moderate loss of RNA binding would exhibit an intermediate phenotype. We transfected ZAP KO 293T cells with ZAP RNA binding mutants and infected with the SINV luciferase reporter virus Toto1101/luc. As expected, the mutant EKR which retains its ability to bind SINV RNA as compared to ZAPS and ZAPL WT ( Figure 1B ) also remains capable of inhibiting SINV replication ( Figures 3A, B ). Both ZAPS and ZAPL ZnF mutants display reduced antiviral activity, and all other CpG RNA binding cavity mutants display differing degrees of loss of antiviral activity ( Figures 3A, B ). We observed significant differences in viral replication between ZnF mutants and WT for ZAPL ( Figure 3B ) but not for ZAPS ( Figure 3A ), likely due to ZAPL’s greater viral inhibition. In ZAPS, mutants CY and KY have the most significant increase in viral replication compared to ZAPS WT ( Figure 3A ), while ZAPL HFR has the most significant increase in viral replication compared to ZAPL WT ( Figure 3B ).

Meanwhile, we transfected TRIM25 KO 293T cells with TRIM25 RNA binding mutants to assess their ability to inhibit SINV replication. TRIM25 7KA, which binds SINV RNA more robustly than TRIM25 WT ( Figure 1C ) exhibits a similar degree of SINV inhibition ( Figure 3C ). On the other hand, TRIM25 ΔRBD, which fails to bind SINV RNA at all, restores viral replication to TRIM25 KO levels ( Figure 3C ). Finally, we asked whether the observed loss of antiviral activity for any of the ZAP and TRIM25 RNA binding mutants tested was due to a decrease in protein expression. To test this, we assayed protein expression both before and after SINV infection and found that with the exception of the ZAPL ZnF mutants, protein expression for all ZAP and TRIM25 variants increases to varying degrees during infection ( Supplemental Figure 2 ). Still, almost all of the ZAP RNA binding mutants express more highly than ZAP-WT ( Supplemental Figures 2A, B ), supporting our hypothesis that it is mutation of these RNA binding residues and not overall protein levels that determines degree of antiviral activity. Together, these data point strongly to the primacy of RNA binding in both ZAP and TRIM25 inhibition of SINV replication.

Following our characterization of ZAP and TRIM25 RNA binding mutants’ ability to inhibit SINV replication, we asked whether this antiviral activity stems from a block in alphavirus translation, given that ZAP blocks SINV translation and that TRIM25 is absolutely required for this inhibition (Bick et al., 2003; Li et al., 2017). These prior studies readily measure SINV translation with a replication-deficient temperature-sensitive luciferase reporter virus, Toto1101/luc:ts6, such that any luciferase activity would reflect translation of the incoming viral genome (Rice et al., 1987). Here, we decided to omit the ZnF 4 mutant (H191R) due to its similarity in behavior to the ZnF 2 mutant (C88R), and the CpG RNA binding mutant EKR due to its lack of antiviral activity ( Figures 3A, B ). We transfected ZAP KO 293T cells with ZAPS or ZAPL ZnF 2 and CpG RNA binding cavity mutants and infected with Toto1101/luc:ts6. Here, we found that while transfection of mutants VYF and KY significantly restores SINV translation in both ZAPS and ZAPL ( Figures 4A, B ), HFR only significantly restores SINV translation in the context of ZAPL ( Figure 4B ), in line with its antiviral activity against SINV replication ( Figure 3B ). Given that the residue Y108 is mutated in both VYF and KY and has previously been implicated in determining ZAP CpG specificity (Meagher et al., 2019), our data suggest that Y108 is very important for ZAP inhibition of SINV translation. Moreover, given that HFR largely retains RNA binding in both ZAPS and ZAPL ( Figure 1B ), these data point suggest that mechanisms in addition to RNA binding may modulate ZAP translation inhibition. No dramatic restoration of viral translation is seen for the ZnF 2 mutant (C88R) in either isoform ( Figures 4A, B ), nor does any TRIM25 RNA binding mutant impact TRIM25 inhibition of viral translation when transfected into TRIM25 KO cells ( Figure 4C ), though the lack of effect could be attributed to low protein expression of TRIM25 ΔRBD. Interestingly, we observed that there appear to be two distinct translation phenotypes in the presence of the TRIM25 ΔRBD mutant, wherein half of the replicate wells exhibit inhibited translation similar to TRIM25 WT and the other half have partially restored SINV translation ( Figure 4C ).

We also asked here whether the observed loss of inhibition of viral translation for any of the ZAP and TRIM25 RNA binding mutants tested was due to a decrease in protein expression. Similar to the replication competent SINV Toto1101/luc, infection with the replication-deficient SINV Toto1101/luc:ts6 generally results in higher expression of ZAP and TRIM25 variants ( Supplemental Figure 3 ), though the phenotype for a 6 hour infection is not as robust as a 24 hour infection. Interestingly, we observed a mild decrease in expression for ZAPS CY, ZAPS VYF and ZAPL ZnF 2 mutant (C88R) ( Supplemental Figures 3A, B ). Still, most ZAP and TRIM25 RNA binding mutants exhibit similar, if not slightly higher protein expression as compared to ZAP and TRIM25 WT. Together, these data suggest that CpG recognition by ZAP is critical for its alphavirus translation inhibition and this is independent of ZAP isoforms.

Because we observed that specific residues involved in ZAP CpG recognition are more important for SINV translation inhibition, rather than RNA binding in general, we next asked if this trend holds true for ZAP translation inhibition of other viruses. JEV is another virus that is translationally inhibited by ZAP (Chiu et al., 2018). To assess if ZAP and TRIM25 RNA binding play a similar role in blocking JEV translation as they do in blocking alphavirus translation, we assessed the ability of our constructs to inhibit a replication-defective JEV replicon expressing Renilla luciferase (Rluc) (Li et al., 2016). Because ZAP can also mediate degradation of JEV RNA, we measured luciferase activity 4 hours after transfecting the JEV replicon reporter to capture a time point that is early enough to see ZAP primarily functioning through translation inhibition, but also long enough to see significant luciferase expression from the replicon. Previous work has shown that there is a minimal decrease in JEV replicon RNA at this time (Chiu et al., 2018). We attempted to co-transfect the JEV replicon with Fluc RNA as a transfection control, but found that ZAPL also restricts Fluc expression ( Supplemental Figure 4 ), as previously demonstrated (Li et al., 2019).

We found that the ZAP and TRIM25 RNA binding mutants inhibit JEV translation to varying degrees ( Figure 5 ). Surprisingly, we observed that the ZAPS and ZAPL ZnF 4 mutants (H191R) are significantly more inhibitory than their WT counterparts, while the ZAPL ZnF 2 mutant (C88R) is more inhibitory than ZAPL WT ( Figures 5A, B ). Given that these mutants show increased association with TRIM25 ( Figure 2B ), these data suggest a positive relationship between ZAP-TRIM25 interaction and JEV translation inhibition. The ZAP CpG RNA binding cavity mutants either show similar inhibition to ZAPS and ZAPL WT or, in the case of the ZAPS KY and ZAPL HFR and KY mutants, decreased inhibition ( Figures 5A, B ). Taken together with the SINV translation inhibition data, these data further point to the importance of the RNA binding cavities in ZAP inhibition of viral translation, including CpG-specific binding by the residue Y108.

Upon assaying TRIM25 variants, we found that TRIM25 ΔRBD inhibits JEV translation similarly to TRIM25 WT ( Figure 5C ), as we observed with SINV translation inhibition ( Figure 4C ). Surprisingly, the TRIM25 7KA mutant loses its ability to inhibit JEV translation ( Figure 5C ), in contrast to its ability to inhibit SINV translation similarly to TRIM25 WT ( Figure 4C ). Given the diverging phenotypes of the 7KA mutant, we were curious if the TRIM25 mutants show a different pattern of binding to SINV versus JEV replicon RNA that would explain the differences in translation inhibition. We found that the TRIM25 7KA mutant binds JEV replicon RNA to a greater degree than TRIM25 WT, while the TRIM25 ΔRBD mutant loses the ability to bind JEV replicon RNA ( Figure 5D ). These findings recapitulate the TRIM25 SINV RNA binding phenotypes ( Figure 1C ) and suggest that RNA binding by TRIM25 is not an important determinant for its ability to inhibit JEV translation.

Given our wide panel of ZAPS and ZAPL RNA binding mutants, we sought to look for significant correlations between any pairs of ZAP phenotypes. To facilitate this, we quantified the immunoblots for ZAP SINV RNA binding and ZAP-TRIM25 co-IPs and calculated fold inhibition relative to empty plasmid transfection for the viral inhibition assays ( Table 1 ). Using these values, we calculated Pearson correlation coefficients and p-values for each pairwise phenotype comparison. When analyzing our data for the ZnF 2 and 4 mutants and the complete panel of CpG RNA binding cavity mutants ( Figure 6A ), we found a significant negative correlation (r = -0.63, p<0.01) between ZAP SINV RNA binding and ZAP-TRIM25 interaction ( Figure 6B ), lending further support to the idea that ZAP RNA binding competes with its ability to associate with TRIM25. We also observed a significant positive correlation (r = 0.78, p<0.001) between ZAP SINV RNA binding and SINV replication inhibition ( Figure 6C ), but interestingly, no correlation between ZAP SINV RNA binding and SINV translation inhibition ( Table 1 ). This suggests that while general ZAP RNA binding is important for its antiviral activity, it is not as critical for the specific step of translation inhibition. Finally, we found a significant positive correlation (r = 0.64, p<0.05) between ZAP-TRIM25 interaction and JEV translation inhibition ( Figure 6D ), further suggesting that ZAP interaction with TRIM25 potentiates its ability to inhibit JEV translation. Because we had a limited number of mutants for TRIM25, we were not able to find significant correlations between any pair of TRIM25 phenotypes (data not shown).

Dr. Akinori Takaoka at Hokkaido University generously provided ZAP KO 293T cells (clone 89) and its parental 293T lines (Hayakawa et al., 2011). TRIM25 KO 293T cells were generated using CRISPR-Cas9 as previously described (Li et al., 2017; unpublished data). Cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, Thermo Fisher Scientific) with 10% fetal bovine serum (FBS) added.

Infections with SINV expressing firefly luciferase (Toto1101/Luc) and temperature-sensitive SINV (Toto1101/Luc:ts6) have been previously described (Rice et al., 1987; Bick et al., 2003). Each independent experiment included triplicate wells of biological replicates per condition. BHK-21 cells were used to generate viral stocks and titers for multiplicity of infection calculations (Bick et al., 2003).

The replication-defective JEV replicon plasmid was generously provided by Dr. Bo Zhang at the Wuhan Institute of Virology (Li et al., 2016). The plasmid pcDNA3.1-3XFLAG was kindly gifted to us by Dr. Oliver Fregoso at UCLA. To generate pcDNA3.1-myc, a myc tag was swapped in for the 3XFLAG tag using BamHI and HindIII restriction sites. ZAPS and ZAPL were cloned into pcDNA3.1-3XFLAG from pTRIP-TagRFP-hZAPS and pTRIP-TagRFP-hZAPL (Li et al., 2017), respectively, using NotI and XbaI restriction sites. Dr. Jae U. Jung at the University of Southern California generously provided full-length TRIM25 (Gack et al., 2007). TRIM25 was cloned into both pcDNA3.1-3XFLAG and pcDNA3.1-myc using XhoI and XbaI restriction sites. All ZAP and TRIM25 constructs are myc- or 3XFLAG-tagged on the N-terminal end.

Point mutations in ZAPS and ZAPL were generated using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs) and all plasmids were verified by sequencing (Genewiz). Primers for mutations were synthesized by Integrated DNA Technologies ( Supplementary Table 1 ). The TRIM25 ΔRBD mutant (Choudhury et al., 2017) was generated by overlapping PCR ( Supplementary Table 1 ). The TRIM25 7KA mutant was generated by ordering a gBlocks Gene Fragment from IDT with all lysines in ₃₈₁KKVSKEEKKSKK₃₉₂ mutated to alanines (Sanchez et al., 2018), and utilizing innate restriction sites in TRIM25 “BsrGI and BamHI (underlined)”, to replace wild-type sequence in TRIM25. The 7KA mutated sequence is bolded in the below gene block, and nonessential nucleotides on the 5’ and 3’ ends are written in lowercase.

5’gtttTGTACAGTCAGATCAACGGGGCGTCGAGAGCACTGGATGATGTGAGAAACAGGCAGCAGGATGTGCGGATGACTGCAAACAGAAAGGTGGAGCAGCTACAACAAGAATACACGGAAATGAAGGCTCTCTTGGACGCCTCAGAGACCACCTCGACAAGGAAGATAAAGGAAGAGGAGAAGAGGGTCAACAGCAAGTTTGACACCATTTATCAGATTCTCCTCAAGAAGAAGAGTGAGATCCAGACCTTGAAGGAGGAGATTGAACAGAGCCTGACCAAGAGGGATGAGTTCGAGTTTCTGGAGAAAGCATCAAAACTGCGAGGAATCTCAACAAAGCCAGTCTACATCCCCGAGGTGGAACTGAACCACAAGCTGATAAAAGGCATCCACCAGAGCACCATAGACCTCAAAAACGAGCTGAAGCAGTGCATCGGGCGGCTCCAGGAGCCCACCCCCAGTTCAGGTGACCCTGGAGAGCATGACCCAGCGTCCACACACAAATCCACACGCCCTGTGGCAGCAGTCTCCGCAGAGGAAGCAGCATCCGCAGCACCTCCCCCTGTCCCTGCCTTACCCAGCAAGCTTCCCACGTTTGGAGCCCCGGAACAGTTAGTGGATTTAAAACAAGCTGGCTTGGAGGCTGCAGCCAAAGCCACCAGCTCACATCCGAACTCAACATCTCTCAAGGCCAAGGTGCTGGAGACCTTCCTGGCCAAGTCCAGACCTGAGCTCCTGGAGTATTACATTAAAGTCATCCTGGACTACAACACCGCCCACAACAAAGTGGCTCTGTCAGAGTGCTATACAGTAGCTTCTGTGGCTGAGATGCCTCAGAACTACCGGCCGCATCCCCAGAGGTTCACATACTGCTCTCAGGTGCTGGGCCTGCACTGCTACAAGAAGGGGATCCgttt-3’

X-tremeGENE9 DNA Transfection Reagent (Roche Life Science) was used to transfect cells at a ratio of 3 μL to 1 μg DNA according to the manufacturer’s instructions. To keep the total plasmid amount in co-transfections constant, empty vectors pcDNA3.1-myc or 3XFLAG were transfected as necessary (6 well plate, 2 μg total input; 24 well plate, 250 ng total input).

SINV DNA templates for transcription were generated by XhoI linearization of pToto1101 (Rice et al., 1987). SINV RNA was transcribed in vitro by Sp6 RNA polymerase (New England Biolabs) in the presence of the cap analog [m7G(5’)ppp(5’)G] (New England Biolabs). Fluc DNA templates for transcription were amplified from the pGL3-Control plasmid (Promega). Fluc RNA was transcribed in vitro using the mMESSAGE mMACHINE T7 Transcription Kit (Invitrogen). Biotin-labeled RNAs were generated by adding 10mM biotin-16-UTP (Roche Life Science) to in vitro transcription reactions. JEV replicon DNA templates for transcription were generated by XhoI linearization and transcribed in vitro using the mMESSAGE mMACHINE T7 Transcription Kit (Invitrogen). Transcribed RNAs were purified using the Quick-RNA Miniprep Kit (Zymo Research) and biotinylation was confirmed by streptavidin dot blot (Chan et al., 2020).

0.4 pmol of biotin-labeled SINV or Fluc RNA probes were heated for 2 min at 90°C, chilled on ice for 2 min, and incubated with 50 μL 2x RNA structure buffer for 30 min at room temperature to ensure proper secondary RNA structure formation, as previously described (Bai et al., 2016). In vitro RNA pull-down was then performed as previously described (Chiu et al., 2018). In brief, RNA probes were incubated with 100 μg of lysates from ZAP or TRIM25 KO 293T cells transfected with ZAP or TRIM25 constructs for 48 hours. Cell extracts were lysed by CHAPS lysis buffer [10 mM Tris-HCl (pH 7.4), 1 mM MgCl₂, 1 mM EGTA, 0.5% CHAPS, 10% glycerol, and 5 mM 2-mercaptoethanol] with complete protease inhibitor cocktail (Roche Life Science) and incubated with RNA probes in a final volume of 100 μL RNA binding buffer supplemented with 1 unit/μL RNAseOUT (Thermo Fisher Scientific), 1 μg/μL heparin (Sigma-Aldrich), and 100 ng/μL yeast tRNA (ThermoFisher) for 30 min at 30°C. Lysate-RNA mixtures were then incubated with 300 μL of Dynabeads M-280 Streptavidin (Invitrogen) for 30 min at room temperature on a shaker. Protein-RNA complexes were washed three times by RNA binding buffer, and proteins were eluted by incubation with 30 μL of 4x Laemmli Sample Buffer (Bio-Rad) for 5 min at 95°C. The proteins were further analyzed by immunoblot and quantified by ImageJ as previously described (Davarinejad, 2015). Briefly, to calculate the quantity of ZAP or TRIM25 bound to RNA relative to input ZAP or TRIM25 protein, the net value of the FLAG (ZAP or TRIM25) band in the whole cell lysate was divided by the net value of the β-actin loading control band in the whole cell lysate, giving a normalized input value. The net value of the FLAG band in the RNA IP band was then divided by the normalized input value.

Proteins were resolved through SDS-PAGE using NuPAGE MOPS SDS running buffer (Thermo Fisher Scientific) and 4-15% precast Mini-PROTEAN TGX Gels (Bio-Rad) before transferring to a PVDF membrane (Bio-Rad). Immunodetection was achieved with 1:2,500 anti-myc (Cell Signaling Technology), 1:20,000 anti-FLAG (Sigma-Aldrich), and 1:20,000 anti-actin-HRP (Sigma-Aldrich). Primary antibodies were detected with 1:20,000 goat anti-mouse HRP (Jackson ImmunoResearch) or 1:20,000 goat anti-rabbit HRP (Thermo Fisher Scientific). Proteins were visualized on a ChemiDoc (Bio-Rad) using ProSignal Pico ECL Reagent (Genesee Scientific). ImageJ was used to quantify western blots as previously described (Davarinejad, 2015). Briefly, protein band intensities for each blot were measured by taking the net grey mean value of each band. The net grey mean value is defined as the inverted pixel density (255 – grey mean value) of a band with the inverted pixel density of the background (defined as an equivalent area of the blot above or below the band) subtracted.

To assess ZAP or TRIM25 co-immunoprecipitation (co-IP) with RNA binding mutants of TRIM25 or ZAP, respectively, cells were transfected in 6-well plates, collected, and then lysed by rotating in FLAG IP buffer (100 mM Tris-HCl 8.0, 150 mM NaCl, 5 mM EDTA, 1 mM DTT, 5% glycerol, 0.1% NP-40) supplemented with a complete protease inhibitor cocktail (Roche Life Science) at 4°C for 30 min, before spinning down at 14,000 rpm at 4°C for 15 min. To equilibrate beads prior to use, anti-FLAG beads (EZview™ Red ANTI-FLAG M2 Affinity Gel, Sigma-Aldrich) or anti-myc beads (EZview™ Red Anti-c-Myc Affinity Gel, Sigma-Aldrich) were washed 3 times in FLAG IP buffer. Three hundred μL of whole cell lysate (WCL) were incubated with 30 μL of anti-FLAG or -myc beads rotating at 4˚C for 45 minutes. FLAG IP buffer was used to wash immunoprecipitates 3 times before eluting bound proteins with SDS loading buffer, and boiling for 5 minutes for immunoblot analysis. Western blot ImageJ analysis was performed as previously described (Davarinejad, 2015). Briefly, to calculate the relative quantity of ZAP RNA binding mutants in the myc (TRIM25) co-IP, the net value of the FLAG (ZAP) IP band, defined as the net grey mean value of ZAP alone subtracted from each mutant band, was divided by the net value of the myc IP for each mutant. To calculate the relative quantity of TRIM25 RNA binding mutants in the FLAG (ZAP) co-IP, the net value of the myc (TRIM25) IP band, defined as the net grey mean value of TRIM25 alone subtracted from each mutant band, was divided by the net value of the FLAG IP for each mutant. To account for the different expression levels of TRIM25 mutants, the myc IP band was first divided by the net grey mean value of the myc WCL band for each mutant, normalized to the value of the band for TRIM25 alone.

Following transfection of ZAP or TRIM25 constructs into ZAP or TRIM25 KO 293T for 48 hours, JEV replicon RNA was transfected by TransIT-mRNA Transfection Kit (Mirus Bio). Cells were lysed 4 hours post-transfection of replicon RNA and luciferase activity was measured by Dual-Luciferase Reporter Assay (Promega). Each independent experiment included triplicate wells of biological replicates per condition.

Statistical analyses in Figures 3 – 5 were performed on biological replicates from triplicate wells using GraphPad Prism. Statistical analyses in Figure 6 were performed using the SciPy package in Python and visualized using the Matplotlib and Seaborn packages (Hunter, 2007; Virtanen et al., 2020; Waskom, 2021).