Substrate trapping approach identifies TRIM25 ubiquitination targets involved in diverse cellular and antiviral processes

The tripartite motif (TRIM) family of E3 ubiquitin ligases is well known for its roles in antiviral restriction and innate immunity regulation, in addition to many other cellular pathways. In particular, TRIM25-mediated ubiquitination affects both carcinogenesis and antiviral response. While individual substrates have been identified for TRIM25, it remains unclear how it regulates diverse processes. Here we characterized a mutation, R54P, critical for TRIM25 catalytic activity, which we successfully utilized to “trap” substrates. We demonstrated that TRIM25 targets proteins implicated in stress granule formation (G3BP1/2), nonsense-mediated mRNA decay (UPF1), and nucleoside synthesis (NME1). R54P abolishes TRIM25 inhibition of alphaviruses independently of the host interferon response, suggesting that this antiviral effect is a direct consequence of ubiquitination. Consistent with that, we observed diminished antiviral activity upon knockdown of several TRIM25-R54P specific interactors including NME1. Our findings highlight that multiple substrates mediate the cellular and antiviral activities of TRIM25, illustrating the multi-faceted role of this ubiquitination network in diverse biological processes.


INTRODUCTION
Many of the TRIM proteins are upregulated by interferon (IFN) and play significant roles in the host innate 48 immune response. 7 Upon detection of viral infection by the host cell, type I IFN is produced, inducing expression 49 of hundreds of IFN-stimulated genes (ISGs) to establish an antiviral environment. 17,18 TRIM25 is one such ISG 50 which not only stimulates innate immune signaling by ubiquitinating and activating the dsRNA sensor RIG-I, but 51 also functions as a critical co-factor of another ISG, zinc finger antiviral protein (ZAP). 19-21 While TRIM25 has 52 been shown to complex with ZAP in the context of several different viral infections, 22 its ligase activity has only 53 been tied to its participation in blocking translation of incoming RNA genomes of alphavirus (family 54 Togaviridae). 20 Given that ubiquitination of ZAP or lack thereof fails to affect its viral translation inhibition, 20 it is 55 likely that TRIM25 antiviral involvement depends on its ubiquitination of other cellular proteins. Interestingly, both 56 TRIM25 and ZAP not only bind viral RNA but also interact with other RNA binding proteins, implying that proteins 57 involved in RNA processes may feature prominently among TRIM25 substrates. 23-26 58 In light of this question, we set out to identify novel TRIM25 substrates that may play a role in translation and 59 RNA processes. Because identification of E3 ligase substrates is technically challenging due to the transient 60 nature of ligase-substrate interactions, we utilized a "substrate trapping" approach as previously reported 27 to 61 capture TRIM25 interactors in a co-immunoprecipitation (IP)/mass spectrometry (MS) experiment. We sought to 62 generate a TRIM25 mutant that would be unable to interact with the upstream E2 carrier enzyme, thus 63 simultaneously rendering it incapable of ubiquitination and prolonging its interactions with substrates. We 64 identified a point mutation, R54P, in the TRIM25 RING catalytic domain, which almost completely abolishes its 65 autoubiquitination in cells. 66 We found that TRIM25-R54P enriches for additional interactors as compared to TRIM25-WT, though almost all 67 of the more highly enriched interactors are shared by both TRIM25-WT and -R54P. Further characterization of 68 some of the most highly enriched interactors, Ras-GTPase-activating protein SH3-domain binding proteins 69 (G3BP) 1 and 2, RNA helicase up-frameshift protein 1 (UPF1), and nucleoside diphosphate kinase 1 (NME1), 70 has validated their identification as novel TRIM25 substrates. Moreover, upon characterization of its antiviral 71 activity, the TRIM25-R54P mutant demonstrates a complete loss of inhibition against a panel of Old World and 72 New World alphaviruses albeit higher IFN and ISG expression compared to WT, suggesting that ubiquitination 73 of TRIM25 substrates directly leads to activation of an antiviral state. Altogether, we have identified both known 74 and novel interactors as TRIM25 substrates, and demonstrated the validity of this "substrate trapping" approach 75 in identifying bona fide E3 ligase substrates. We have shed light on the ways that TRIM25-mediated 76 ubiquitination might target substrates to modulate translation, nucleic acid metabolism, and antiviral response, 77 paving the way for further work characterizing the critical role of TRIMs in diverse cellular and viral processes. 78

RESULTS 79
Point mutations in TRIM25 RING domain almost completely abolish TRIM25 autoubiquitination 80 It is technically challenging to identify E3 ligase-substrate interactions as they are often transient, resulting in 81 proteasomal degradation or a change in localization or activity of the substrates. In order to enrich for transient 82 E3 ligase-substrate interactions, we turned to a less conventional co-IP approach that makes use of E3 mutants 83 unable to interact with E2 conjugating enzymes. This prevents ubiquitin transfer to E3 substrates and their 84 subsequent targeting to other cellular pathways and as a result, "trapping" these substrates. This approach 85 successfully identified the cellular 'structural maintenance of chromosomes' (Smc) complex Smc5/6 as being 86 targeted by hepatitis B virus X protein for ligase-mediated degradation. 27 We hypothesized that a similar 87 approach would serve to identify TRIM25 substrates, which will be immunoprecipitated more robustly with a 88 TRIM25 E2 binding mutant than with TRIM25-WT, as the former is unable to mediate transfer of ubiquitin from 89 E2 to substrates. 90 Residues important for the RING-E2 interaction and thus necessary for ligase activity have already been 91 identified in the RING E3 ligase MDM2. 28 We aligned the structure of the TRIM25 RING domain complexed to 92 E2-ubiquitin (Ub) to the analogous MDM2-E2-Ub structure and identified two conserved critical E2 interaction 93 residues in TRIM25 RING, I15 and R54 (Fig. 1A). To assess loss of ligase activity, we transfected HA-tagged 94 Ub and FLAG-tagged TRIM25 into 293T cells and immunoprecipitated TRIM25 in denaturing conditions. We 95 then blotted for HA-Ub, wherein polyubiquitination manifests as a ladder of bands. These TRIM25 E2 binding 96 mutants (I15K and R54P), are deficient in autoubiquitination, suggesting successful crippling of ligase activity 97 (Fig. 1B). Individual E2 binding mutants retain a mono-Ub band (Fig. 1B), so we generated the double mutant 98 I15K/R54P, which did not display further reduction in ligase activity (Fig. 1B). Therefore, we selected the R54P 99 mutant for future co-IP/MS studies since this mutation has previously been shown to reduce TRIM25 catalytic 00 activity and polyubiquitin chain formation. 29 01 Substrate trapping approach enriches for novel TRIM25 interactors 02 Next, we asked what proteins are modified by TRIM25, as identification of these substrates will elucidate how 03 ubiquitination facilitates TRIM25-mediated cellular and antiviral activities. We first used CRISPR-Cas9 to 04 generate a TRIM25 KO 293T cell line (Fig. S1). We then stably integrated doxycycline (dox) inducible FLAG-05 tagged TRIM25 wild-type (WT) and mutant R54P using the ePiggyBac (ePB) transposon system, 30 where both 06 TRIM25-WT and TRIM25-R54P are similarly induced in a dose-dependent manner ( Fig. 2A). TRIM25 protein 07 levels are comparable upon detection using a FLAG or TRIM25-specific antibody ( Fig. 2A). 08 To capture TRIM25 substrates, we performed two independent co-IP/MS experiments using our reconstituted 09 TRIM25 KO 293T cell lines (Fig. 2B). We induced TRIM25-WT or -R54P expression in the presence or absence 10 of the prototype alphavirus Sindbis virus (SINV), performed a FLAG IP to enrich for TRIM25, and analyzed the 11 resultant protein mixture using MS. TRIM25 KO 293T cells were used as a control. We found that this "substrate 12 trapping" approach enriches for interactors specific to TRIM25-R54P under both mock and infected conditions 13 ( Fig. 2C-D, red circles). These TRIM25-R54P-specific interactors tend to have lower fold change in abundance 14 over background than interactors common to both TRIM25-WT and TRIM25-R54P ( Fig. 2C-D TRIM25 interactors drastically decreases during viral infection for both TRIM25-WT (29 to 7 interactors;  22  Tables 2 and 4) and TRIM25-R54P (38 to 23 interactors; Tables 1 and 3). We used DAVID bioinformatics 23 resources 31,32 to find that TRIM25 interactors are highly enriched in GO terms involved in translation, RNA 24 metabolism, and viral transcription (Fig. 2E). This is in line with our hypothesis that TRIM25 substrates mediate 25 diverse cellular and viral processes as a consequence of ubiquitination. 26 TRIM25 interacts with G3BP1 and 2 through a conserved binding motif and modifies them with 27 predominantly K63 polyubiquitin chains 28 Among the most enriched TRIM25-R54P interactors in the presence and/or absence of SINV infection (Tables  29  1, 3), we identified the core stress granule proteins G3BP1 and 2, RNA helicase UPF1 (Fig. 2C-D, blue arrows),  30  and the metastatic suppressor and nucleoside kinase NME1 as high priority candidates given our interest in RNA  31 metabolic and translation processes (G3BP1 and 2, UPF1) and TRIM25's role in regulating carcinogenesis 32 (NME1). Next, we asked whether any of these are TRIM25 ubiquitination substrates. 33 Both G3BP1 and G3BP2, hereafter collectively referred to as G3BP, associate very strongly with TRIM25 in the 34 co-IP/MS (Tables 1-4, G3BP1: log2FoldChange 2.5 -6.5; G3BP2: log2FoldChange 5.5 -9.5). G3BP normally 35 function in stress granule (SG) assembly, interacting with RNA and other cellular proteins to induce SG 36 formation. 33 is also necessary for TRIM25-G3BP2 interaction, we co-transfected myc-tagged G3BP into TRIM25 KO 293T 49 cells along with FLAG-tagged TRIM25-WT, -R54P, or -PTAA, and performed a FLAG IP to pull down TRIM25. 50 While both TRIM25-WT and -R54P robustly associate with both G3BP1 and -2, TRIM25-PTAA does not 51 associate with either G3BP1 or 2 (Fig. 3A), validating our co-IP/MS identification of G3BP as TRIM25 interactors. 52 We then used the ePiggyBac transposon system to reconstitute TRIM25 KO 293T cells with dox inducible 53 TRIM25-PTAA. To establish that TRIM25 ubiquitinates G3BP and that the TRIM25-G3BP interaction is 54 necessary for ubiquitination, we co-transfected myc-tagged G3BP with HA-Ub into TRIM25-WT, -R54P, and -55 PTAA inducible cell lines. After inducing TRIM25 expression, we performed a myc IP and probed for the presence 56 of ubiquitinated G3BP. We found that both G3BP1 and 2 are robustly polyubiquitinated only in the presence of 57 TRIM25-WT (Fig. 3B), again validating our co-IP/MS identification of G3BP as TRIM25 substrates. No 58 ubiquitination is detected in the presence of ligase-deficient TRIM25-R54P, whereas ubiquitination is 59 dramatically diminished in the presence of G3BP-interaction deficient TRIM25-PTAA (Fig. 3B). Interestingly, 60 TRIM25 appears to more robustly ubiquitinate G3BP2 as compared to G3BP1 (Fig. 3B). Given the TRIM25-61 mediated polyubiquitination of G3BP1 and 2, we then characterized G3BP ubiquitination linkage type. To do so, 62 we transfected our TRIM25-WT inducible cell line with myc-G3BP1 or -2 and different forms of HA-Ub; -WT, -63 K48, and -K63. Ub-K48 and -K63 have all lysines mutated to arginine except -K48 and -K63, respectively, such 64 that only K48 or K63 polyubiquitin chains are able to be formed. 44 We found that both G3BP1 and 2 are most 65 robustly ubiquitinated in the presence of Ub-K63, suggesting that TRIM25 primarily mediates K63-linked 66 ubiquitination of both proteins (Fig. 3C). Interestingly, while both G3BP1 and 2 exhibit a lower level of 67 ubiquitination in the presence of Ub-K48, G3BP1 possesses more K48-linked polyubiquitin chains as compared 68 to G3BP2 (Fig. 3C), indicating that TRIM25 is able to distinguish between and differentially ubiquitinate these 69 related proteins. 70

TRIM25 interacts with and mono-ubiquitinates UPF1 at K592 71
Moreover, UPF1 associates very strongly with TRIM25 in the co-IP/MS (Tables 1-4, log2FoldChange 3.9 -5.5), 72 supporting a role for UPF1 as a novel TRIM25 interactor. UPF1 is best known for its central role in nonsense-mediated mRNA decay (NMD), where it is recruited to premature termination codons to catalyze the NMD 74 pathway, inhibiting further translation and recruiting other RNA-degrading enzymes. 45 UPF1 has also been 75 implicated in serving an antiviral role in the context of alphaviral infection. 46 The authors of this study found that 76 depletion of UPF1 or two other NMD components promotes viral replication, and specifically depleting UPF1 77 likely stabilizes incoming viral RNA genomes. 46 78 We first validated that TRIM25 interacts with UPF1. To do so, we transfected V5-tagged UPF1 into TRIM25 79 inducible cell lines, then induced for TRIM25-WT or -R54P expression with dox, and performed a FLAG IP to 80 pull down TRIM25. We found that UPF1 is robustly detected only when TRIM25 is induced (Fig. 4A), validating 81 the TRIM25-UPF1 interaction identified in our co-IP/MS. To test the hypothesis that TRIM25 ubiquitinates UPF1, 82 we co-transfected V5-tagged UPF1 with HA-Ub into our TRIM25 inducible cell lines and induced TRIM25 83 expression. We then performed a V5 IP and probed for the presence of ubiquitinated UPF1. We found that UPF1 84 is more robustly mono-ubiquitinated only in the presence of TRIM25-WT and not ligase-deficient TRIM25-R54P 85 (~50% more by ImageJ quantification, Fig. 4B Given its well-characterized role as a metastatic suppressor, we decided to validate NME1 as a TRIM25 98 ubiquitination substrate. We first set out to validate TRIM25 interaction with NME1 as identified in our co-IP/MS 99 (Tables 1-3). To do so, we transfected myc-tagged NME1 or UPF1 to serve as a positive control in our TRIM25 00 inducible lines, induced for TRIM25-WT or -R54P expression, and performed a FLAG IP to pull down TRIM25. 01 We then probed for any associated UPF1 or NME1. While we saw robust association of UPF1 with both TRIM25-02 WT and -R54P in line with our previous results ( Fig. 4A, 5A, MW ~135 kDa), we did not identify NME1 (Fig. 5A, 03 MW 20-25 kDa). We also performed the reverse IP where we pulled down myc-tagged NME1, but were unable 04 to find any TRIM25 interacting with NME1 ( Fig. 5B). We hypothesized that this lack of TRIM25-NME1 interaction 05 could be due to functional differences between ectopically expressed myc-NME1 and endogenous NME1, given 06 our successful validation of the other robust TRIM25 interactors from our co-IP/MS, G3BP and UPF1 ( Fig. 3-4). 07 To test this hypothesis, we performed a FLAG IP using our TRIM25 inducible lines and probed for co-IP of 08 endogenous NME1 along with endogenous G3BP and UPF1 as positive controls. In line with our co-IP/MS 09 results, endogenous G3BP, UPF1, and NME1 enrich robustly with TRIM25 pulldown, despite a low level of non-10 specific binding of NME1 to the FLAG IP in TRIM25 KO 293T cells (Fig. 5C). 11 To test whether TRIM25 ubiquitinates NME1, we transfected myc-tagged NME1 into TRIM25-WT and -R54P 12 inducible cells in the presence or absence of dox and performed a myc IP. We found that NME1 is more robustly 13 polyubiquitinated in the presence of TRIM25-WT as compared to TRIM25-R54P, although we cannot yet rule 14 out the possibility that TRIM25 might mono-ubiquitinate NME1 at multiple sites (Fig. 5D). Taken together, these 15 results suggest that robust TRIM25 interactors, such as G3BP, UPF1, and NME1 identified in our TRIM25 co-16 IP/MS function as bona fide TRIM25 substrates. 17 TRIM25 antiviral activity is dependent on its ligase activity 18 Given our identification of diverse host factors as TRIM25 substrates ( Fig. 3-5), many of which function in 19 translational and RNA processes (Fig. 2E) and several of which have known roles in alphavirus replication, we 20 hypothesized that TRIM25 ligase activity is critical to orchestrating an antiviral response. 21 We used TRIM25 inducible cell lines in the KO background ( Fig. 2A) to characterize the requirement of ligase 22 activity in TRIM25-mediated viral inhibition. We found that TRIM25-WT, which retains ligase activity, represses 23 SINV replication, whereas ligase mutant TRIM25-R54P does not (Fig. 6A). Overexpression of TRIM25-WT ( alphaviruses. TRIM25-WT remains potently antiviral against all alphaviruses tested, while overexpression of 38 TRIM25-R54P either has no effect on or restores viral replication to levels higher than the TRIM25 KO 39 background (Fig. 6D). Taken together, these data clearly demonstrate that TRIM25-dependent ubiquitination is 40 required for inhibition of alphavirus replication, specifically through a block in viral translation. 41

TRIM25-mediated viral inhibition is independent of changes in the type I IFN response 42
To exclude the complementary possibility that TRIM25 is exerting antiviral effects through affecting type I IFN or 43 ISG production, we quantified the mRNA of IFN- and the prominent ISGs IFIT1, ISG15, and OAS2 in the 44 presence of poly(I:C), a dsRNA mimetic and stimulator of innate immune signaling. If TRIM25 antiviral activity is 45 mediated through a strengthened IFN response, we would expect that both IFN and ISG production to increase 46 when TRIM25-WT is induced and to be lower in the presence of TRIM25-R54P due to its defective antiviral 47 activity. Poly(I:C) stimulation works well, inducing IFN-β and ISGs robustly in the absence of TRIM25 (>10 fold 48 over absence of poly(I:C); data not shown). Surprisingly, we found that overexpression of either TRIM25-WT or 49 TRIM25-R54P significantly suppresses production of IFN-, IFIT1, ISG15, and OAS2 mRNA in the presence of 50 poly(I:C) (Fig. 6E). We also observed that induction of TRIM25-WT results in a more drastic suppression of the 51 ISGs as compared to TRIM25-R54P (Fig. 6E), which could be due to higher viral replication in the presence of 52 TRIM25-R54P ( Fig. 6A-B), leading to a higher type I IFN response in the TRIM25-R54P inducible cell line. Still, 53 these data support our hypothesis that TRIM25 antiviral activity is not mediated through the IFN response. 54

Identifying TRIM25-R54P specific interactors as critical for viral inhibition 55
As we showed that the loss of antiviral activity of TRIM25-R54P does not correlate with the levels of IFN and 56 ISG expression, suggesting a direct consequence of TRIM25-mediated ubiquitination of target proteins, we then 57 decided to examine TRIM25-R54P interactors identified in our co-IP/MS (Tables 1,3) that are not consistently 58 present in the TRIM25-WT enrichment. These candidate proteins likely exhibit weaker or more transient 59 interactions with TRIM25 and are ubiquitinated by TRIM25. We hypothesized that if any of these interactors are 60 critical for TRIM25 antiviral activity, loss of their expression would result in increased viral replication even in the 61 presence of overexpressed TRIM25-WT. While we initially also assessed a subset of ribosomal proteins 62 identified as TRIM25-R54P interactors, their knockdown results in high cytotoxicity and therefore are excluded 63 from subsequent analyses (data not shown). We validated most of the TRIM25-R54P interactors that are not 64 present on the TRIM25-WT list (Tables 1-4) in the absence (Fig. 7A) or presence of viral infection (Fig. 7B).
While knockdown of multiple interactors trends towards restoring SINV replication, only loss of RTRAF (Table 1,  66 log2FoldChange 1.6 -1.8) and NME1 (Table 3, log2FoldChange 3.4 -4.8) significantly restores SINV replication 67 ( Fig. 7A-B). Moreover, knockdown of MOV10 (Table 1, log2FoldChange 4.5 -4.9) approaches significant 68 restoration of SINV replication (Fig. 7A, p=0.0631). 69 Given that loss of NME1 results in the most significant restoration of SINV replication (Fig. 7B) and is 70 ubiquitinated by TRIM25 (Fig. 5D), we de-convoluted its siRNA pool in both our inducible TRIM25-WT cell line 71 and in the parental 293T cell line with endogenous TRIM25 and ZAP expression. There, we observed that the 72 degree of NME1 mRNA knockdown positively correlates with increase of viral replication (Fig. 7C-D R54P, which is predicted to abolish its interaction with E2 carrier enzymes and is sufficient to cripple TRIM25 83 ligase activity (Fig. 1). We reported identification of TRIM25 substrates involved in nucleic acid metabolism and 84 translation (Fig. 2E), in line with its role in blocking viral translation. 20 We characterized the ubiquitination of the 85 most enriched TRIM25-R54P interactors G3BP (Fig. 3), UPF1 (Fig. 4), and NME1 (Fig. 5), representing proteins 86 with essential cellular functions, some of which with prior involvement in alphavirus infection. 46,51 We also used 87 the TRIM25-R54P mutant to definitively show the critical role of ubiquitination in TRIM25 antiviral activity that is 88 independent of IFN production and signaling (Fig. 6). We then examined proteins that display a preference for 89 association with TRIM25-R54P under mock and viral infection conditions, and found that several of these are 90 necessary for TRIM25 antiviral activity (Fig. 7), identifying them as potential TRIM25 substrates mediating viral 91 inhibition. Our results suggest that targeting of any single substrate by TRIM25 is insufficient to mediate the 92 entirety of its cellular and antiviral activities, illustrating the powerful, multi-faceted role of this ubiquitination 93 network in diverse biological processes. 94 We propose that the success of this "substrate trapping" approach in identifying TRIM25 ubiquitination substrates 95 hinges on preservation of protein structure. For the first time, we identified G3BP1/2, UPF1, and NME1 as bona fide TRIM25 substrates (Fig. 3-5). 09 Furthermore, we were able to characterize TRIM25 polyubiquitination of G3BP as primarily utilizing K63 linkages 10 (Fig. 3C). This type of linkage is commonly used to build signaling scaffolds, as TRIM25 does to activate RIG-I, 11 and could potentially play a role in either SG assembly or disassembly by recruiting SG components in the former 12 or generating steric hindrance in the latter. Additionally, our validation of K592 as a monoubiquitination site on 13 UPF1 (Fig. 4C) overlaps with a predicted acetylation site on the same residue, and neighbors a predicted 14 phosphorylation site at T595, potentially modulating these other post-translational modifications of UPF1. 54 15 These residues lie within the AAA ATPase domain of UPF1, suggesting that ubiquitination of UPF1 by TRIM25 16 might affect its ATP hydrolysis, thus hindering UPF1 in its NMD target discrimination and efficient translation 17 termination. 55,56 Interestingly enough, G3BP1 and UPF1 cooperate to mediate structure-mediated RNA decay. 57 18 It is entirely possible that TRIM25-mediated ubiquitination could affect this process by modulating their interaction 19 with one another, though further experiments are required to explore this hypothesis. Furthermore, NME1 has 20 been previously demonstrated to be ubiquitinated and subsequently targeted for degradation by the E3 ligase 21 SCF-FBXO24 58 . Seeing as TRIM25 is able to modify G3BP with both proteolytic K48-and non-proteolytic K63-22 polyubiquitin linkages, TRIM25 may also be targeting NME1 for degradation, thereby hindering nucleotide 23 synthesis and general RNA metabolic processes. 24 We also utilized the TRIM25-R54P mutant to define the requirement for ligase activity in TRIM25 inhibition of 25 alphavirus replication. We found that TRIM25 ligase activity is absolutely required for its inhibition of diverse 26 alphaviruses through a block in viral translation. Interestingly, overexpression of both TRIM25-WT and -R54P 27 results in a dampened IFN response in our hands (Fig. 6E), contrasting with the previously established role of 28 TRIM25 in activating RIG-I signaling and implicating TRIM25 as a negative regulator of the type I IFN response. 59 29 Moreover, TRIM25-R54P with a complete loss of antiviral activity actually exhibits relatively more production of 30 IFN and a subset of ISG mRNAs (Fig. 6E), which may be indicative of higher viral replication overall (Fig. 6A-B). 31 Still, these data together strongly suggest that the robust TRIM25 antiviral activity against alphaviruses is not 32 mediated through an augmented IFN response, but through its ligase activity and subsequent ubiquitination 33 network. 34 Our examination of the contribution of a subset of TRIM25-R54P specific interactors to TRIM25 antiviral activity 35 has yielded several hits, namely RTRAF (Fig. 7A) and NME1 (Fig. 7B). RTRAF, also known as hCLE or 36 C14orf166, is an RNA binding protein involved in cellular transcription, translation, and RNA transport, and is 37 required for influenza virus replication. 60-62 Notably, RTRAF is a member of a cap-binding complex that activates 38 mRNA translation. 61 Given RTRAF's role in facilitating translation of mRNAs, it is therefore tempting to speculate 39 that RTRAF may be required for translation of alphavirus RNA, and TRIM25-mediated ubiquitination of RTRAF 40 may affect its ability to do so. The novel bona fide TRIM25 substrate NME1, which functions as a major 41 synthesizer of non-ATP nucleoside triphosphates, upon ubiquitination may inhibit alphavirus replication via a 42 similar mechanism as the potent restriction factor SAMHD1, which depletes deoxynucleotide pools, effectively 43 preventing replication of varied DNA viruses and reverse transcription of HIV-1. 63 On the other hand, TRIM25-44 mediated ubiquitination of NME1 may inhibit its metastatic suppressor activities, potentially serving as a novel 45 mechanism for TRIM25's previously described roles in carcinogenesis. Further studies need to be carried out to 46 elucidate the functional consequences of these TRIM25 substrates in blocking viral translation and other cellular 47 processes. 48 The novelty of this work lies within our innovative approach to uncover a multifaceted ubiquitination network 49 involved in mediating TRIM25 cellular and antiviral functions. Many questions remain unanswered as to how 50 TRIM25-mediated ubiquitination modulates the activity of these substrates. In contrast to the more binary 51 consequences of K48-linked dependent degradation, other types of ubiquitin linkage may effect more nuanced 52 cellular changes by modulating substrate activity and localization. 53 Given TRIM25 proclivity for K63 linkages in 53 the context of alphavirus infection and innate immunity, 19-21 we are tempted to speculate that TRIM25 eschews 54 a simple degradation approach in favor for a more nuanced modulation of substrate activity and localization. 55 Current therapeutics that harness E3 ligases focus on their degradative power, generating compounds that bring 56 ligases in close proximity to a target protein for degradation. 64 Further research is warranted to explore the utility 57 of alternate modes of ubiquitination in biological therapeutics. 58  to design guide RNAs (gRNAs) targeting exon 1 of the human TRIM25 gene (Fig. S1A). The guide with the 25 highest ranking in both scoring programs (5'-CGGCGCAACAGGTCGCGAACGGG-3') was selected for cloning 26 into the PX459 vector (Addgene, #62988), a non-lentiviral construct that also delivers Cas9. 74  TRIM25, leading to frameshift mutations and premature stop codons in both alleles (Fig. S1D). 43

Generation of TRIM25 inducible cell lines 44
To reconstitute TRIM25 expression (WT and R54P) in our TRIM25 KO 293T cell line (clone #8; see above for 45 details), we used the enhanced PiggyBac (ePB) transposable element system provided by the Brivanlou 46 laboratory at the Rockefeller University, as previously described. 75 Dried pellets were processed at the UCLA Proteomics Core. Protein samples were reduced and alkylated using 62 5mM Tris (2-carboxyethyl) phosphine and 10mM iodoacetamide, respectively, and then proteolyzed by the 63 sequential addition of trypsin and lys-C proteases at 37˚C as described. 77 Digested peptides were resuspended 64 in 5% formic acid and fractionated online using a 25cm long, 75 μM inner diameter fused silica capillary packed 65 in-house with bulk C18 reversed phase resin (length, 25 cm; inner diameter, 75 μM; particle size, 1.9 μm; pore 66 size, 100 Å; Dr. Maisch GmbH). 78 The 140-minute water-acetonitrile gradient was delivered using a Dionex 67 Ultimate 3000 UHPLC system (Thermo Fisher Scientific) at a flow rate of 300 nL/min (Buffer A: water with 3% 68 DMSO and 0.1% formic acid and Buffer B: acetonitrile with 3% DMSO and 0.1% formic acid). Fractionated 69 peptides were ionized and analyzed by tandem mass spectrometry (MS/MS) Orbitrap Fusion Lumos mass 70 spectrometer (Thermo Fisher Scientific). Label-free quantitation was performed using the MaxQuant software 71 package. 79 The EMBL Human reference proteome (UP000005640 9606) was utilized for all database searches. 72 Statistical analysis of MaxQuant output data was performed with the artMS Bioconductor 80 package which 73 performs the relative quantification of protein abundance using the MSstats Bioconductor package (default 74 parameters). Intensities were normalized across samples by median-centering the log2-transformed MS1 75 intensity distributions. The abundance of proteins missing from one condition but found in more than 2 biological 76 replicates of the other condition for any given comparison were estimated by imputing intensity values from the 77 lowest observed MS1-intensity across samples and p-values were randomly assigned to those between 0.05 78 and 0.01 for illustration purposes. Significant hits were defined as interactors that possessed a log2FoldChange 79 of >1.5 and a -log10Pvalue > 1.3. 80 rotating. Immunoprecipitates were washed 3 times with the FLAG IP buffer. Bound proteins were eluted with 89 SDS loading buffer and boiled for 5 minutes for immunoblot analysis. 90

Ubiquitination IP assay 91
To assess TRIM25 ubiquitination of putative substrates, immunoprecipitation was performed essentially as 92 previously described. 20 Briefly, cells were collected and lysed in 0.5% SDS buffer supplemented with complete 93 protease inhibitor cocktail. Three hundred μL of WCL were diluted into 1X TNA buffer (0.25% Triton, 50 mM Tris-94 HCl, pH 7.5; 200 mM NaCl, 1 mM EDTA) + 2 mg/mL BSA. WCL containing V5-tagged substrates were then 95 incubated with 1 μg of anti-V5 antibody overnight at 4˚C. The next morning, 40 μL Protein A Dynabeads 96 (Invitrogen, Waltham, MA) were added and incubated for 2 h at 4˚C. WCL containing myc-tagged substrates 97 were incubated directly with anti-myc beads for 45 minutes at 4˚C, rotating. Following incubation with beads, 98 both myc-tagged and V5-tagged immunoprecipitates were washed 3 times with 1X TNA buffer + 2 mg/mL BSA. 99 Myc-tagged NME1 underwent an additional two washes with 1X TNA buffer only. Bound proteins were eluted 00 with SDS loading buffer and boiled for 5 minutes for immunoblot analysis. 01