TRIM

tri

m

TRIM5 RNA is multiply spliced, generating a family of isoforms, each shorter from the C terminus. The longest, TRIM5α, encodes a C terminal B30.2 domain that interacts directly with viral capsid and determines antiviral specificity 182122. The shorter isoforms, TRIM5γ and TRIM5δ, do not have B30.2 domains and act as dominant negatives to TRIM5α and rescue restricted infectivity when over-expressed 723. It is assumed that the shorter forms form heteromultimers via the coiled coil and titrate the viral binding B30.2 domains. It is therefore possible that TRIM5's antiviral activity is regulated by splicing.

The B30.2 domain comprises a combination of a PRY motif followed by a SPRY motif 24. Whilst SPRY domains are evolutionary ancient, B30.2 domains, found in butyrophilin and TRIM proteins, appeared more recently. There is unlikely to be a precise function for B30.2 domains, rather they are involved in protein-protein interactions such as substrate recognition. A series of TRIM5 mutagenesis studies demonstrated that the TRIM5 B30.2 domain determines antiviral specificity and defined the specific regions of the B30.2 responsible 1821222526. In vitro capsid/TRIM5 binding assays have been developed and these demonstrate that, at least in the case of wild type TRIM5α proteins, binding correlates well with the ability to restrict infection 2728.

The recent solution of the structure of several B30.2 domains allows us to interpret the conservation and variation between TRIM5 B30.2 domains 293031. The structures indicate that the B30.2 core is formed from a distorted 2-layer beta sandwich with the beta strands in an anti-parallel arrangement. Extending from the core are a series of loops and it is these surface loop structures that vary between the TRIM5 sequences from each primate and between different B30.2 domains of TRIM5 homologues. The loops form 3 or 4 variable regions, all of which appear to impact on antiviral specificity 32. The TRIM21 structure in complex with its ligand, IgG Fc indicates that there are 2 binding surfaces, one in the PRY (V1) and 1 in the SPRY (V2-V3) and this is likely to be true for TRIM5α.

B30.2 mutagenesis studies, as well as sequence analysis of TRIM5α from related primates, suggested that the differences defining anti-viral specificity are concentrated in patches in the B30.2 domain 33. The patches, which correspond to the surface loops, have been under very strong positive selection as evidenced by a high dN:dS ratio. dN:dS ratios have been calculated by comparing TRIM5 sequences from primates and comparing the number of differences that lead to a change in the protein sequence (non synonomous, dN) to the number of differences that do not (synonomous, dS). A high ratio indicates positive selection and is evidence of the host-pathogen arms race known as the Red Queen hypothesis 34. This phenomenon, named after Lewis Carroll's Red Queen who claimed 'It takes all the running you can do to keep in the same place', refers to the selection driven genetic change that occurs in both host and pathogen as each alternately gains the advantage. Whether selection pressure on TRIM5 has been from pathogenic retroviruses or from endogenous retroviruses and retrotransposons is unclear. The relative youth of lentiviruses, as compared to other retroviruses and endogenous elements, is thought to preclude them from impacting on TRIM5 selection, although the discovery of an endogenous lentivirus in rabbits 35 has recently extended their age from less than 1 million years to greater than 7 million years and it certainly seems possible that this age will extend further as we better understand lentiviral history.

The other side of the Red Queen's arms race is the change in the retroviral capsids to escape restriction by TRIM5. TRIM5 molecules can generally restrict widely divergent retroviruses including gamma retroviruses as well as lentiviruses. For example Agm and bovine TRIMs restrict MLV-N, HIV-1, HIV-2 and SIVmac 363738. It is now clear that retroviral capsid structures are conserved and capsid hexamers are found in both lentiviruses and gamma retroviruses 3940 so we imagine that the TRIMs recognise a conserved shape. Paradoxically, point mutants can often escape strong restriction. MLV-N CA R110E escapes human, simian and bovine TRIMs, SIVmac CA QQ89-90LPA escapes rhesus and squirrel monkey TRIM5s and HIV-1 G89V escapes owl monkey TRIMCyp 13841424344. It therefore remains unclear how TRIM5 can be effective if a small number of changes in CA can rescue infectivity, especially given that retroviral capsid sequences appear quite plastic.

We are beginning to understand TRIM5α 's antiviral mechanism. TRIM5α is trimeric 1945 and interacts with hexameric capsids 46. TRIM5α is ubiquitinylated within cells and is rapidly turned over by the proteasome in a RING domain dependent way suggesting that autoubiquitinylation might drive this process 1547. We imagine that the rapid turnover of TRIM5α and presumably TRIM5α-virus complexes leads to an early block to infection, before the virus has had the opportunity to reverse transcribe (Figure 1A). This notion is supported by the observation that inhibition of the proteasome during restricted infection allows the virus to reverse transcribe, when it is protected from degradation 4849 (Figure 1B). However, infection is not rescued by inhibition of the proteasome, indicating that the TRIMα-virus complex remains uninfectious, even when protected from degradation. How exactly TRIM5α renders the virus uninfectious remains unclear, but it may be that by simply coating the core with multivalent complexes TRIM5α trimers are able to disrupt the rearrangement/uncoating and or trafficking required to continue to the nucleus and to integrate. Other possibilities include TRIM5α rapidly uncoating incoming HIV-1 capsids. In fact, this has been observed using an assay of capsid density to measure uncoating 4650 and it will be interesting perform this assay in the presence and absence of proteasome inhibitors to address whether the proteasome has a role this process. Proteasome independent degradation of capsids by TRIM5α has also been described 51. Importantly, DNA circles remain inhibited, even in the absence of proteasome activity, suggesting that the restricted TRIM5α-virus complex cannot access the nucleus. 4849 (Fig 1). It is possible that these observations indicate several independent antiviral activities of TRIM5α but we prefer the interpretation that there are several possible fates for a restricted virion. It may be degraded by the proteasome, it may inappropriately uncoat, or it may remain intact, make DNA but not have access to the nucleus. The different fates are likely to be influenced by factors such as the particular virus, the particular TRIM5α as well as virus dose and TRIM5α expression levels and the cellular background. Understanding the contribution of these activities to restriction by TRIM5α will require further study but the field continues to make steady progress.

The relationship between Cyclophilin A (CypA) and HIV-1 has a long history. CypA is a peptidyl prolyl isomerase that performs cis/trans isomerisation of proline peptide bonds in sensitive proteins. CypA interacts with gag in infected cells leading to its recruitment into nascent HIV-1 virions 5253. Recent data has shown that CypA also interacts with incoming HIV-1 cores in newly infected cells and that this interaction is more important for infectivity than that occurring as cores assemble 42545556. This may be because only about 10% of the capsid molecules in the core recruit a CypA molecule into the virion 5253. CypA performs cis/trans isomerisation at CA G89-P90 on the outer surface of the capsid 5758 and this leads to changes in infectivity. In Old World monkey (OWM) cells CypA decreases HIV-1 infectivity, but only in the presence of TRIM5α 596061. Blocking CypA activity using the immunosuppressant competitive inhibitor of CypA cyclosporine A (CSA), or reducing CypA expression with small interfering RNA, reduces the susceptibility of HIV-1 to restriction by OWM TRIM5 and rescues HIV-1 infectivity.

In human cells the interaction between incoming HIV-1 cores and CypA is important for maximal infectivity. Preventing this interaction reduces HIV-1 infectivity independently of TRIM5 expression 5962. It is suspected that in the absence of CypA activity, HIV-1 gets restricted by a TRIM5 independent antiviral activity. This suspicion is borne from the fact that the requirement for CypA is both cell type, and species, specific, suggesting that CypA is not required simply to uncoat the core. This notion is further supported by the observation that CA point mutants close to the CypA binding site such as HIV-1 CA A92E or G94D appears to lead to restriction of HIV-1 in human cells 5556. A92E or G94D infectivity is reduced in some human cell lines but not others and strikingly, infectivity is rescued by inhibition of CypA. It is possible that these mutants become sensitive to human restriction factor(s) and that the interaction between the factor and the virion is sensitive to the activity of CypA on the peptide bond at P90.

How might CypA impact on recognition of CA by TRIM5α? One possibility is that in some cases, capsid with CypA attached may make a better target for TRIM5α. This possibilty has been discounted on the basis that HIV-1 mutated to prevent CypA binding (HIV-1 CA G89V) remains restricted by TRIM5α from Old World monkeys 5961. Importantly, this mutant is not restricted by TRIM-Cyp, which relies on the CypA domain to recruit it to the HIV-1 capsid 43. A second possibility is that recruitment of TRIM5α to capsid is improved by the prolyl isomerisation activity of CypA on HIV-1 capsid. Prolyl isomerisation has been shown to regulate protein-protein interaction in diverse biological systems including the control of cell division by cdc25C and signalling by the Itk receptor. The prolyl isomerase Pin1 catalyses the cis/trans isomerisation of a proline peptide bond in cdc25C. Cdc25C activity is regulated by phosphorylation and since its phosphatase PP2A only recognises the cdc25C trans isomer, Pin1 activity leads to dephosphorylation and cdc25C activation 63. A similar molecular switch has been described for Itk signalling and CypA. CypA catalyses cis/trans isomerisation of proline 287 in the Itk SH2 domain impacting on interaction with phosphorylated signalling partners and regulating Itk activity 6465. NMR measurements have shown that HIV-1 CA contains around 86% trans and 14% cis at G89-P90 in both the presence and absence of CypA 57. However, in the presence of CypA, CA is rapidly isomerised between the two states 57. It is therefore possible that OWM TRIM5α binds preferentially to CA containing G89-P90 in the cis conformation 59. In this case, in the presence of TRIM5α, CypA maintains the percentage of cis at 14% even as TRIM5α sequesters it from the equilibrium. In this way the trans form is isomerised to cis and becomes bound by TRIM5α. Blocking CypA activity would limit the availability of the cis conformation and therefore TRIM5α's ability to see the CA, resulting in rescued infectivity. This model is summarised in Fig 2. CypA also appears to impact on replication of feline immunodeficiency virus in feline and human cells although whether TRIM5 is required for this remains unclear 66.

Surprisingly in the New World species owl monkey a CypA pseudogene has been inserted into the TRIM5 coding region, replacing the viral binding B30.2 domain with CypA, leading to a molecule called TRIMCyp 4344. This restriction factor strongly restricts HIV-1, SIVagm and FIV by recruitment of the incoming capsid to the RBCC domain facilitated by interaction between the CypA domain and the capsid 206667. Viral infectivity is rescued by inhibition of CypA-CA interactions with CSA indicating the dependence on CypA binding to capsid for robust restriction. We assume that at some point in owl monkey evolution the modification of TRIM5 to TRIMCyp provided a significant selective advantage. We can only speculate on what might have provided the selection pressure but a pathogenic virus that recruited CypA is a possibility. It is worth noting that a TRIMCyp in the human genome would be a useful antiviral as we face the current AIDS pandemic.

The role of CypA in sensitivity to TRIM5, its fusion to TRIM5 in owl monkeys and its role as a target for immunosuppression implies that CypA might have a general role in immunity. Viruses are likely to be under considerable pressure to alter their shape and become invisible to antiviral shape recognition systems such as TRIMs. Molecules, such as CypA, that induce shape changing, may have an important role in making escape difficult. For example, HIV-1 appears to be invisible to OWM TRIM5 in the absence of CypA, but in its presence HIV-1 is strongly restricted 596061. Conversely, HIV-1 is highly infectious in human cells in the presence of CypA but appears to become restricted in its absence 42. It seems that HIV-1 is invisible to human TRIM5 whether CypA is active or not but becomes restricted by something else in the absence of CypA activity 5962. HIV-1 appears to have adapted to tolerate CypA activity and this adaptation has made it dependent on CypA. Why can't HIV-1 simply avoid recruiting CypA? The answer to that is not clear but a clue can be found in alignment of the CypA binding region of lentiviruses (Figure 3). All primate lentiviruses have conserved the proline rich CypA binding loop and many encode glycine proline motifs within it. This suggests that the motifs that recruit CypA are important, conserved and cannot easily be mutated. The loops and glycine proline motifs are also conserved in the equine lentivirus EIAV and the feline FIV 67. Their purpose however remains unclear and this loop is not conserved in MLV 40 (Figure 4).

The fact that TRIM5 restricts retroviral infection so potently, at least in monkeys, has suggested that polymorphism in human TRIM5 might impact on HIV-1 transmission and/or pathogenesis in vivo. Several studies have addressed this issue and shown at best, only weak association of any particular TRIM5α allele with disease progression 68697071. Importantly, human TRIM5α is not polymorphic in the regions of the B30.2 domain known to impact on viral recognition, and its over expression does not reduce HIV-1 infectivity by more than a few fold 791072. Furthermore under in vitro conditions where rhesus TRIM5 efficiently binds the HIV-1 capsid, the human protein binds only poorly 46. It therefore seems likely that TRIM5 doesn't significantly impact on HIV-1 replication and pathogenesis in humans. Indeed, we imagine that HIV-1's insensitivity to TRIM5 has been an important factor in its success as a pathogen in humans. Conversely the TRIM5 gene in rhesus macaques and sooty mangabeys is relatively polymorphic with a number of polymorphisms occurring in the variable loops that dictate antiviral specificity. Indeed, expression of these alleles in permissive feline cells followed by challenge with retroviral vectors derived from HIV-1, SIVmac MPMV or MLV-N demonstrated that the different alleles have slightly different antiviral specificities 72.

The antiviral activity of TRIMs in mammals other than primates remains less well characterised. A bovine TRIM (BoLv1) with broad anti retroviral activity suggests that TRIM-mediated restriction of retroviruses is widespread amongst mammals 3738. BoLv1 is closely related to primate TRIM5 genes suggesting that they are orthologs derived from an ancestral antiviral TRIM. Cattle encode at least 4 genes closely related to TRIM5, in addition to homologs of TRIM34 and TRIM6. The fact that one of these proteins has antiviral activity supports the notion that these genes are derived from an ancestral sequence with antiviral activity. It is likely that antiviral TRIMs will be identified in more mammals soon. Indeed, antiviral TRIMs are probably responsible for the poor infectivity of cells from pigs and bats to MLV-N and those of rabbits to HIV-1 13573.

Protein families arise through the duplication of ancestral gene sequences and therefore members of a family share common ancestry. Human TRIM5 lies on chromosome 11 within a group of closely related TRIMs, comprising TRIMs 5, 6, 34 and 22, which have presumably arisen by gene duplication. These TRIMs, as well as TRIMs 1, 18, 19 and 21 have no, or relatively weak, antiviral activity against a panel of distantly related retroviruses including HIV-1, HIV-2, SIVmac, EIAV and MLV 20. Whether this is because they have an alternate function or whether they are simply not active against this selection of viruses is difficult to say. It is worth noting however that comparison of the sequences of these TRIMs from primates shows that unlike TRIM5, TRIMs 6, 22 and 34 do not have strongly selected B30.2 domains, suggesting that they have not been under the same selection pressures as TRIM5 74.

There is an increasing body of evidence, gathered over many years suggesting that TRIM19, otherwise known as PML, may have antiviral activity. PML exists in sub-nuclear structures called PODs, ND10 or PML bodies and are of unclear function. It has long been known that a number of diverse viruses including influenza, SV40 and papilloma virus form replication complexes in close association with PML bodies, reviewed in 7576. Infection by other viruses, including herpes viruses and adenoviruses, causes degradation of PML protein and dispersal of the body components. The molecular details of PML degradation by herpes simplex type 1 (HSV-1) have been partially solved. The HSV-1 protein ICP0 is responsible for inducing proteasome dependent degradation of PML, and HSV-1 deleted for this protein replicates poorly, leaving PML bodies intact 77787980. Importantly, mutant HSV-1 (ICP0-) becomes almost fully infectious if PML expression is reduced using RNA interference, indicating that an important function of ICP0 is to eliminate PML 81. An antiviral role for PML is also suggested by a real time microscopy study demonstrating that PML is recruited to incoming HSV-1 (ICP0-) replication complexes 82. Such active recruitment is strongly suggestive of an antiviral response. Furthermore, reduction of PML expression increases permissivity of human cells to human cytomegalovirus infection 83, and over-expression of PML reduces permissivity to vesicular stomatitis virus and influenza A 8485. These data, along with the observation that PML expression is stimulated by type 1 interferon, strongly support an antiviral role for TRIM19 (PML). Interestingly, PML does not have a B30.2 domain suggesting that it interacts with target viruses in a different way to TRIM5α interacting with retroviruses.

Further data supporting an antiviral role for TRIM proteins comes from expression studies in which TRIMs are expressed in permissive cells and the modified cells tested for permissivity to infection by retroviral vectors. Such studies have demonstrated weak anti-retroviral activity of TRIM1 from African green monkeys and Owl monkeys against MLV-N 9. It is also worth noting that a particular TRIM protein can impact on viral infectivity by influencing the activity of another antiviral TRIM protein. For example, expression of TRIM34 can reduce the antiviral activity of TRIM5 presumably via heteromutimerisation mediated via the coiled coil 20. This observation suggests a complex mechanism of regulation and generation of alternate antiviral specificities through heteromultimerisation. Whether further TRIMs have antiviral activity remains largely untested. The fact that TRIMs 10, 15, 26, 27, 31, 38, 39, 40 are associated with the major histocompatibility complex on chromosome 6 86 and the observation that the expression of most of these genes is up-regulated by influenza infection 87 suggests that they might have a role in immunity.

TRIM20, otherwise known as pyrin, presents as an intriguing antiviral possibility. Polymorphism in the TRIM20 B30.2 domain can cause familial Mediterranean fever, a disease characterised by recurrent attacks of fever and inflammation. Sequencing TRIM20 from a variety of primates revealed that many encode the disease causing mutations as wild type sequence 88. Furthermore, phylogenetic analysis suggested episodic selection in the B30.2 domain, similar to that seen for TRIM5, suggesting the intriguing possibility that viral infection underlies this disease. Rather strikingly in 2001 these authors suggested that the B30.2 domain of pyrin might interact directly with pathogens and that the mutations are counter evolutionary changes selected to cope with a changing pathogen 88. Such a model is remarkably close to what we believe to be true for TRIM5, retroviruses and the Red Queen 6 years later.

Just as we considered that the important aspects of TRIM5 biology had been largely described, the Ikeda lab described tantalising findings that make a complicated subject significantly more complicated 89. They show that rhesus TRIM5 causes degradation of gag in infected cells. Importantly this activity is independent of the C-terminal B30.2 domain suggesting that it acts via an alternative specificity determinant, perhaps the coiled coil. It is worth noting that APOBEC3G has also been described as being able to restrict infection of both incoming as well as outgoing HIV-1 9091. It may be therefore that such dually active restriction factors are not uncommon.

Whether the study of host factors influencing viral infection will translate into improvements in antiviral therapy in the foreseeable future remains uncertain. However, it is likely to allow the improvement of animal models for HIV/AIDS as we enhance our understanding of the viral and cellular determinants for viral replication and disease 92. This work is also likely to improve our ability to transduce cells, therapeutically and experimentally, with viral gene delivery vectors, particularly poorly permissive primary cells and stem cells. It certainly promises to remain an active and exciting field in infectious disease research.