TRIM25 and its emerging RNAbinding roles in antiviral defense

The innate immune system is the body’s first line of defence against viruses, with pattern recognition receptors recognising molecules unique to viruses and triggering the expression of interferons and other anti-viral cytokines, leading to the formation of an anti-viral state. The Tripartite Motif Containing 25 (TRIM25) is an E3 ubiquitin ligase thought to be a key component in the activation of signalling by the pattern recognition receptor Retinoic Acid-Inducible Gene I Protein (RIG-I). TRIM25 has recently been identified as an RNA-binding protein, raising the question of whether its RNA-binding activity is important for its role in innate immunity. Here, we review TRIM25’s mechanisms and pathways in non-infected and infected cells. We also introduce models that explain how TRIM25 binding to RNA could modulate its functions and play part in the antiviral response. These findings have opened new lines of investigations into functional and molecular roles of TRIM25 and other E3 ubiquitin ligases in cell biology and control of pathogenic infections.


INTRODUCTION
Humans, as well as other organisms, must protect themselves from infection by viruses and other pathogens. To do this, they developed a robust means of distinguishing self from nonself and responding accordingly. Their primary means of doing so is through the immune system. The immune system is divided into two parts, the adaptive and innate immune systems. Adaptive immunity consists of antigen-specific receptor-mediated responses to specific pathogens, while innate immunity consists of pathways for the detection of factors common to many pathogens as well as physical barriers such as the skin. In general, the innate immune system is fast-acting, involving elements that are ubiquitously expressed in somatic cells and acts in early infection. In contrast, the adaptive immune response is slower, acting in late infection, as there are only small numbers of each antigen-specific receptor and the cells expressing these must undergo clonal expansion before an effective response can be mounted 1 . The innate immune system involves non-antigen-specific pattern recognition receptors (PRRs) that recognise pathogen associated molecular patterns (PAMPs) that are common to many pathogens, but not found in host cells. Upon detection of PAMPs by PRRs, a signalling cascade is initiated that results in the expression of various anti-pathogenic molecules such as interferons (IFNs) and other cytokines, resulting in activation of host defences such as inflammation and recruitment of the adaptive immune system. Importantly, no single pathogen is recognised by a single PRR and biologically unrelated pathogens can be recognised by the same PRR, allowing for a fast and efficient response to any pathogen 2 .
Different classes of pathogens are recognised by different PAMPs, for example viruses are recognised through glycoproteins and various DNA and RNA species 3 . Bacteria through lipoproteins 4 , peptidoglycan and derivatives 5 , CpG DNA 6 , lipopolysaccharides (LPS) 7 and proteins such as flagellin 8 . Fungi are generally recognised through cell wall or cell surface components such as phospholipomannan 9 or β-glycan 10 .
There are a wide variety of PRRs found in humans that recognise different classes of PAMPs.
PRRs are generally classed into the membrane-bound Toll-like receptors (TLRs) and C-type Here we will focus on the Tri-partite Motif containing protein 25 (TRIM25), which was thought to be involved in the RIG-I/INF pathway. Recent findings showed that TRIM25 is an RNAbinding protein 11,12 and put into question its major role in the RIG-I activation 13,14 . We will elaborate on what is known about TRIM25 and its RNA-binding activity and speculate how it can affect its antiviral properties. This protein represents a large group of proteins with newly identified RNA-binding potential and can serve as an exemplar towards critical evaluation of protein functions in biological systems.

TRIM25 is an E3 ubiquitin ligase
The Tri-partite Motif (TRIM) family is a large (>80 members in humans) group of E3 ubiquitin ligase proteins that share a common domain structure. Human TRIM25 is a 630 amino acid, 71kDa E3 ubiquitin ligase that is widely expressed across human cell types and is conserved among vertebrates including fish, birds and mammals [15][16][17][18][19] . Like other TRIM family proteins, it consists of an N-terminal zinc-finger Really Interesting New Gene (RING) domain, responsible for its E3 ubiquitin ligase activity, two B-box zinc finger domains of unknown function, a coiled-coil domain (CCD), responsible for homo and heterodimerization, with a linker domain leading to a C-terminal associated with SPRY/SPla and the RYanodine Receptor (PRY/SPRY) domain, responsible for protein-protein interactions.
The primary role of TRIM E3 ubiquitin ligases is to catalyse the addition of polyubiquitin chains or single ubiquitin monomers (monoubiquitination) to lysine residues on their target proteins.
Ubiquitin is a 76aa, 8.5kDa protein and polyubiquitin chains are made by the formation of isopeptide bonds between the ubiquitin C-terminal glycine and one of the 7 lysine residues present in the protein 20,21 . It has also been shown that polyubiquitin chains can form in a 'head-to-tail' manner via the N-terminal methionine residue 22 . The addition of ubiquitin monomers to a target protein or extension of a polyubiquitin chain involves three types of proteins; E1 activating enzyme, E2 conjugating and E3 ubiquitin ligase 23 . The multiple lysines present in ubiquitin allow the formation of several different types of polyubiquitin chains, each of which has different functionality. The most well studied are straight, homogenous K48-linked polyubiquitination, which leads to targeting of proteins for degradation via the proteasome, and K63-linked polyubiquitination which is used in many intracellular signalling pathways, such as activation of NF-κB and induction of RIG-I/IKK/Interferon type I 24 .
TRIM25 forms an antiparallel dimer mediated by its CCD, with the RING domain of each monomer at opposite ends of the dimer and the PRY/SPRY domains positioned at the centre via the CCD-PRY/SPRY linker 25 (Figure 1). Further work indicated that the RING domain of TRIM25 must dimerize in order to catalyse polyubiquitin chain formation, implying that higher-order assembly of TRIM25 dimers is required for its activity 26 . Two separate mechanisms of higher-order assembly were proposed. Firstly, an 'end-to-end' model in which RING domains on each end of the dimer interact with RING domains from separate dimers.
Secondly, a 'tetramer' model in which TRIM25 dimers effectively stack on top of each other with RING domains on either end of one dimer interacting with both RING domains from another dimer 26 . When human TRIM25 RING domain was crystallised with an ubiquitincharged E2 conjugating enzyme, UBE2D1, it was shown to form a dimer with both RING monomers contacting the ubiquitin molecule 27 . Crystal structures have also been generated for the PRY/SPRY domain of mouse TRIM25, showing that its overall structure is that of two anti-parallel β-sheets in a sandwich type conformation, similarly to PRY/SPRY domains found in other proteins 28 .

Miscellaneous roles of TRIM25 in the cell
TRIM25 has been found to perform many roles in the cell and has been implicated in several cancers [29][30][31][32][33][34][35][36][37][38] . TRIM25 was initially identified as a protein responsive to oestrogen in a screen for regions of DNA bound by the oestrogen receptor and was shown to be upregulated in oestrogen receptor-positive mammary cells 39 . Since then, TRIM25 has been found to have an important role in the defence against viruses and has been implicated in the activation of the RIG-I/INF pathway, which will be discussed further in the next section.
TRIM25 has been shown to play other functions in innate immunity. It was shown to positively regulate Melanoma Differentiation-Associated gene 5 (MDA5)-mediated signalling through tumour Necrosis factor receptor-associated factor 6 (TRAF6), an E3 ubiquitin protein ligase, leading to activation of NF-κB 40 . TRIM25 has also been shown to enhance the activity of Zincfinger Antiviral Protein (ZAP), which is a known antiviral protein, this interaction will be addressed later in this review 41,42 . In contrast, TRIM25 can also be involved in the dampening of RIG-I signalling. The ubiquitin-like FAT10 (also known as Ubiquitin D) forms a complex with TRIM25/RIG-I that sequesters RIG-I away from the mitochondria and causes it to form insoluble aggregates to prevent further signal transduction 43 . TRIM25 stabilises FAT10, which is usually unstable, by preventing its proteasome-mediated degradation 43 . Oligomerisation of RIG-I is also important for the formation of tetramers of the 2CARD (two Caspase Activation and Recruitment Domains), which is required for activation of downstream signalling via MAVS through interaction with MAVS' own CARD 52 . A critical process in the formation of the 2CARD tetramer and the activation of MAVS is the TRIM25mediated K63-linked polyubiquitination of the 2CARD, which has been shown to stabilise the 2CARD tetramer and enhance the formation of MAVS filaments that are required for signalling 53,54 (Figure 2). It has been showed that another E3 ubiquitin ligase Riplet is also necessary for release of the 2CARD and RIG-I activation 55,56 . Initially, TRIM25 was identified as the key E3 ubiquitin ligase for the ubiquitination of the RIG-I 2CARD in both mice and humans by Gack et al. 57 . In this report, 6 lysine residues on the RIG-I 2CARD that underwent K63-linked polyubiquitination were identified by mass spectrometry (MS); K99, K169, K172, K181, K190 and K193. Of these, only mutation of K172 to arginine resulted in a reduction in polyubiquitination of the 2CARD and a concomitant reduction in activation of the NF-κB and IFNβ promoters when the 2CARD was transfected into HEK293 cells. This suggested that polyubiquitination of this residue is key for RIG-I 2CARD-mediated signalling. However, later work suggested that a RIG-I K172R mutant was fully functional and could efficiently trigger innate immune signalling in response to Sendai virus infection, suggesting that K172 may not be required for signalling 58  Further work has underlined the role that TRIM25 plays in ubiquitination of the RIG-I 2CARD.
Mutation of T55 in the first RIG-I CARD was found to abolish the TRIM25-2CARD interaction and this was required for TRIM25-mediated polyubiquitination of the 2CARD 59  This is followed by ubiquitination at various sites on the 2CARD by Riplet, TRIM25, TRIM4 and MEX3C. This would fit with several recent observations 13,14,65 that Riplet is absolutely required for efficient RIG-I signalling activation, while TRIM25 is not, as well as explaining why there is seemingly redundancy between the other E3 ligases. In particular, recent work by Cadena et al. showed that deletion of TRIM25 from HEK293T and MEF cells results in an increase, not a reduction in RIG-I activation in response to 5'ppp-RNA. In contrast, deletion of Riplet from the same cell line results in complete abolition of RIG-I signalling, to the same level as deletion of RIG-I itself 14 . The same was found to be true for expression of IFNβ mRNA in response to infection with Sendai virus, with deletion of TRIM25 in this case resulting in a significant increase in IFNβ expression 14 . Interestingly this study also found that TRIM25 was not capable of ubiquitinating RIG-I in vitro, even in the presence of 5'ppp-dsRNA. This goes against previous work which has shown TRIM25 to be capable of ubiquitinating the RIG-I 2CARD in vitro, although these studies used purified 2CARD and not full-length RIG-I 26,67 In order to be able to replicate efficiently, most human viruses have evolved mechanisms for avoiding the triggering of innate immune signalling. Many RNA viruses that can be recognised by RIG-I have developed mechanisms of inhibiting RIG-I signalling at different stages of the pathway. The NS1 protein of IAV is known to block ubiquitination of RIG-I through interactions with TRIM25, Riplet and RIG-I itself [68][69][70] . IAV NS1 binds to the coiled-coil domain of TRIM25, and this was thought to prevent its multimerisation which it requires for its catalytic activity 68 .
However, more recent structural data indicates that the inhibitory effect of NS1 on TRIM25 activity is due to disruption of interactions between the TRIM25 PRY/SPRY and coiled-coil domains that are required for RIG-I ubiquitination activity, not a lack of TRIM25 multimerisation 71 . IAV NS1 is capable of inhibiting Riplet activity in both mouse and human cells and this is likely important for the overall inhibitory effect of NS1 on RIG-I signalling 70 .
Other viruses also target TRIM25-mediated RIG-I ubiquitination for inhibition in order to dampen the innate immune response. The V proteins of several paramyxoviruses (Nipah, measles, Sendai and parainfluenza viruses) were found to interact with both the RIG-I 2CARD and the SPRY domain of TRIM25, preventing TRIM25-mediated ubiquitination of RIG-I 72 .
Similarly, RSV NS1 and N protein from SARS-CoV were also found to interfere with TRIM25 activity 73,74 .
TRIM25 can be targeted for inhibition by the host cell in order to downregulate RIG-I signalling to prevent excessive inflammation and IFN responses. The Linear Ubiquitin Assembly Complex (LUBAC), composed of Heme-Oxidized IRP2 Ubiquitin Ligase 1 (HOIL-1) and HOIL-1 Interacting Protein (HOIP), competes with TRIM25 for binding to RIG-I and also targets TRIM25 for degradation via the proteasome 75 .

The partnership between TRIM25 and ZAP
ZAP is a well know IFN induced antiviral protein that has been shown to inhibit the replication of many different types of viruses, such as SINV 76 , Ebola 77 , HIV 78 and Hepatitis B 79 . However, this inhibition is specific as some viruses such as Herpes simplex 1 and vesicular stomatitis virus are not repressed by ZAP 76 . ZAP was found to bind CG dinucleotides in viral RNAs and inhibit virus replication 80 . The frequency of CG dinucleotides is suppressed in host RNAs allowing the distinction of self from non-self. However, RNA viruses have evolved to supress their CG dinucleotide content, thereby evading the host response. ZAP's ability to suppress a virus was shown to correlate with the viral RNA CG content 80 . This inhibition seems to be mainly through reducing the levels of viral mRNAs, however the mechanisms are still unknown, that said ZAP has been shown to bind viral RNAs directly 81 and to promote translational repression 82 . Additionally, ZAP can degrade viral RNAs by recruiting proteins to remove the poly-A tail and trigger de-capping of the mRNA and finally recruit the mRNA degradation machinery 78,83,84 . ZAP is targeted to stress granules during viral infection, which is important for its antiviral activity 85,86 . Tang et al. found that NS1 binds to and inhibits ZAP's ability to block IAV mRNA translation and degradation, by interfering with its RNA binding ability 87 . ZAP has also been shown to interact with and activate RIG-I by promoting its oligomerisation in human cells 88 . However, the RIG-I-dependent type I interferon response does not seem to important in the antiviral defence against all viruses and in all cell types such as primary mouse cells 85  ZAP does not have enzymatic activity 91 and is thought to require co-factors. TRIM25 has been shown to be required for ZAP's antiviral function 41,42 . TRIM25 was found to interact with ZAP through its SPRY domain, with both the ubiquitin ligase activity and multimerisation of TRIM25 being important in enhancing ZAP's antiviral activity (Figure 3). TRIM25 was also shown to be dependent on ZAP for inhibiting SINV replication 42 . TRIM25 was shown to mediate K48-and K63-linked polyubiquitination of ZAP, but only K68-linked ubiquitination enhanced its antiviral activity 41,42 . TRIM25 was found to do this by aiding the RNA binding ability of ZAP, leading to inhibition of viral translation 41,42 . However, although TRIM25's E3 ligase activity was needed to enhance ZAP-mediated inhibition of Sindbis virus RNA translation, ubiquitination of ZAP itself did not seem to directly affect antiviral activity 42 .
TRIM25 was also found to be required for ZAP's antiviral activity against a HIV-1 virus with clustered CpGs 80 . Another co-factor KHNYN (a cytoplasmic protein of previous unknown function) is thought to form a complex with ZAP and TRIM25 to inhibit the replication of HIV-1 with clustered CpGs. KHNYN required both TRIM25 and ZAP for its antiviral activity.
However, KHNYN was not needed in the inhibition of SINV replication 92 . The authors speculated that KHNYN could be needed for ZAP mediated RNA degradation, but not translation inhibition.

RNA-BINDING ROLES OF TRIM25
TRIM25 is an RNA-binding protein TRIM25 was initially discovered to be an RNA-binding protein (RBP) in a screen of mRNA binding proteins in HeLa cells 11 . Proteins were cross-linked to RNA via UV irradiation and mRNAs were isolated from the cell lysate with oligo(dT) probes before bound proteins were analysed by mass spectrometry 11 . A similar strategy also identified mouse Trim25 as an RBP in mouse embryonic stem cells 93 . Binding of human TRIM25 to RNA was further validated by immunoprecipitation (IP) of TRIM25 followed by radiolabelling of RNA. Signal from radiolabelled RNA after TRIM25 IP was reduced in cells in which TRIM25 had been knockeddown by RNAi, indicating that TRIM25 was binding to RNA 93 . This study also tested several truncation mutants of TRIM25 for their RNA binding activity. Truncations in which the Nterminal RING and B-box domains or the C-terminal PRY/SPRY domain were deleted were capable of binding RNA in this assay as seen by signal from radiolabelled RNA corresponding to the size of the constructs as visualised by western blot. Conversely, any mutants in which the CCD was deleted were incapable of binding RNA, suggesting that the CCD could play a role in TRIM25 RNA binding activity 93 .
We independently discovered TRIM25's RNA binding potential and showed the first RNAdependent function for TRIM25 in the Lin28-mediated degradation of pre-let-7a-1 94 (Figure 4). TRIM25ΔRBD (delta RNA binding domain), in which these amino acids were deleted, was unable to bind to pre-let-7a-1 in RNA pulldown or EMSA and exhibited loss of binding to target mRNAs and miRNAs in RIP experiments. Interestingly, TRIM25ΔRBD was capable of associating with TRIM25 WT, suggesting that dimerization with two intact PRY/SPRY domains is necessary for RNA binding (Figure 4). Aligning amino acids 470-508 with the crystal structure of the PRY/SPRY domain from mouse Trim25 (the composition of which is highly similar to humans) shows that this region comprises β sheet 1, RNAs were also reduced upon overexpression of TRIM25, again with gibbon TRIM25 showing a larger effect and this was also seen for TRIM25 mutants lacking ubiquitin ligase activity (TRIM25 C13A/C16A) 95 . Deletion of RIG-I and TRIM25 from human A549 lung cells resulted in increased viral titres, viral protein, and vRNA levels, which could be rescued by expression of human or gibbon TRIM25. Interestingly, deletion of RIG-I alone had no effect, suggesting this function of TRIM25 is RIG-I-independent 95 . Further evidence that RIG-I is not required for this activity was shown by the use of a viral minigenome assay in HEK293T ΔRIG-I cells.
Overexpression of both human and gibbon TRIM25 in these cells reduced expression of a luciferase reporter that could only be expressed in the context of IAV RNA polymerase activity, indicating that TRIM25 was inhibiting the activity of the IAV polymerase 95 . TRIM25 was subsequently shown to bind to IAV viral RNPs (vRNPs) in an RNA-dependent manner and that gibbon TRIM25 binds to vRNPs more efficiently than human TRIM25. In addition to this, purified human or gibbon TRIM25 was able to inhibit viral mRNA chain elongation in vitro, again with gibbon TRIM25 doing this more efficiently, reflecting its higher ability to restrict viral replication and protein production as well as its stronger binding to vRNPs 95 . The authors of this study proposed a mechanism whereby TRIM25 blocks the IAV RNA polymerase from moving down the vRNA template and prevents the onset of chain elongation, thus restricting the virus.

Dengue virus RNA inhibits TRIM25
Dengue virus has been shown to take advantage of TRIM25's RNA-binding by expressing a subgenomic flavivirus RNA (sfRNA) that binds to and inhibits TRIM25 preventing RIG-I activation and type I IFN induction. This lead to an increase in the epidemiological fitness of a new Dengue virus clade (PR-2B) compared to the previous one (PR-1) 96 . The authors found that the sfRNA from PR-2B bound more than that from PR-1 by EMSA and showed that TRIM25 bound to PR-2B sfRNA was more ubiquitinated than that bound by PR-1 sfRNA. The sfRNA did not affect TRIM25's ability to co-immunoprecipitate with RIG-I, but TRIM25 bound to RIG-I was still ubiquitinated more in the presence of PR-2B sfRNA than PR1 sfRNA.
However, it was unclear if this difference was because of the increased level of TRIM25 pulled down. It was concluded that the increased ubiquitination of TRIM25 was due to the inability of USP15 to de-ubiquitinate TRIM25 thereby preventing RIG-I activation and type I IFN induction. As mentioned previously, LUBAC binds to and ubiquitinates TRIM25 targeting it for degradation by the proteasome, preventing it from ubiquitinating and activating RIG-I 75 .
However, Manokaran et al. did not test the nature of TRIM25 ubiquitination. In fact, more TRIM25 was co-immunoprecipitated with RIG-I in the presence of PR-2B sfRNA. This ubiquitination band could be the reported auto-ubiquitination of TRIM25, which has as yet to be found a function for 12,26,75 .

TRIM25 requires RNA binding for its E3 ubiquitin ligase activity
We recently found that TRIM25 requires RNA binding to be able to efficiently autoubiquitinate itself and to ubiquitinate ZAP, one of its targets 12 67 . These results support the hypothesis that TRIM25 requires RNA binding to ubiquitinate its target proteins and further suggest that this is a general property of TRIM25 activity that is not restricted to a single target protein. Sanchez  Additionally, they showed that TRIM25 localises to stress granules in an RNA-binding dependent way. ZAP also localises to stress granules 85,86 and it is possible this helps these proteins to co-localise.
The mechanism by which TRIM25 RNA-binding facilitates its ubiquitination activity is unknown. It is known that the RING domain of TRIM25 is active as a dimer and it is likely that this requires higher order oligomerisation of TRIM25 dimers. It is therefore possible that RNA binding is important for this higher order organisation, for example by clustering the PRY/SPRY domains from separate dimers together to facilitate formation of a 'tetramer'-like structure. This is rendered less likely by our data showing that purified His-TRIM25ΔRBD and His-TRIM25 WT both form tetramers in vitro, although it remains possible that the proteins behave differently in vivo. It is also possible that binding to RNA causes TRIM25 to undergo a conformational change that allows ubiquitin ligase activity.
Although TRIM25 has been established as an RNA-binding protein there are many outstanding questions. How does the emerging RNA-binding roles of TRIM25 contribute to innate immunity? Which host and viral RNAs do TRIM25 bind during viral infection? Do viral proteins, such as IAV NS1, interfere with TRIM25 binding to host RNAs? Is the ubiquitination and PRY-SPRY mediated RNA-binding activity of TRIM25 necessary for antiviral defence?
Which other E3 ubiquitin ligases bind to RNA and are involved in the innate immune response? Answering these questions will result in a paradigm shift in our understanding of TRIM25-mediated control of RNA viruses and will pave the way towards novel, targeted, antiviral therapies.