Restriction of HIV-1 and retroviruses by TRIM5

Abstract

Mammalian cells express a variety of innate immune proteins called restriction factors that defend against invading retroviruses such as HIV-1. Two members of the tripartite motif protein family – TRIM5α and TRIMCyp – were identified in 2004 as restriction factors that recognize and inactivate the capsid shell that surrounds and protects the incoming retroviral core. Research on these TRIM5 proteins have uncovered a novel mode of non-self recognition that protects against cross-species transmission of retroviruses. Our developing understanding of the TRIM5 restriction mechanism underscores the concept that capsid uncoating and reverse transcription of the viral genome are coordinated processes rather than discrete steps of the post-entry pathway of retrovirus replication. In this Review, we provide an overview of the current state of knowledge on the molecular mechanism of TRIM5-mediated restriction, highlight recent advances, and discuss implications for the development of capsid-targeted antiviral therapeutics.

Introduction

The restriction activity of TRIM5α and TRIMCyp – collectively termed TRIM5 – was first described in studies comparing the growth of HIV and SIV in primate cell lines, in which it was found that HIV is better than SIV at infecting cells from New World monkeys, whereas the reverse is true for Old World monkeys^1,2. A similar activity was described that restricts replication of N-tropic but not B-tropic murine leukemia virus (MLV) in various mammalian cells³. TRIM5 plays an important role in protecting cells against cross-species transmission of retroviruses, and its detailed characterization was spurred by efforts to understand intrinsic cellular barriers to HIV-1 infection^4,5. Research on TRIM5 has now uncovered an unprecedented mechanism of non-self recognition that recognizes the entire viral capsid as a single unit, and has provided an experimental window to elucidate the post-entry stages of retroviral replication. Understanding its detailed mechanism of action may also inspire new ways to inhibit HIV-1.

It is now established that TRIM5 recognizes and intercepts the incoming retroviruses by binding to the capsid that surrounds and protects the viral core. The molecular basis of recognition and how this activates TRIM5-mediated ubiquitination are now reasonably well-understood, although a number of details remain to be elucidated. On the other hand, how binding leads to core inactivation is not yet definitively established, with models invoking direct core destabilization upon TRIM5 binding and/or ubiquitin-dependent recruitment of cellular degradation machinery. In this Review, we summarize the current state of understanding of how TRIM5 proteins recognize incoming retroviral cores, inhibit virus replication, and activate innate immune pathways that signal the presence of an invader. We use HIV-1 as the primary model system, as the discovery of TRIM5α and TRIMCyp arose from efforts to elucidate barriers against HIV-1 and the bulk of literature on TRIM5-mediated restriction focus on this retrovirus.

Organization, trafficking, and uncoating of the HIV-1 core

For a retrovirus to successfully infect a host cell, the viral genomic RNA must be reverse-transcribed into double-stranded DNA and then integrated into the host’s chromosomes (Fig. 1). Retroviral genomes initially enter the cell’s cytoplasm as part of the viral core, which is protected by the viral capsid. The capsid facilitates reverse transcription, shields the viral nucleic acids from cytoplasmic host defense sensors, and – in the case of lentiviruses such as HIV-1 – recruits host factors required for nuclear import of the viral DNA. Prior to integration, the capsid uncoats and releases the viral genome from its protective shell. This uncoating event appears to be tightly regulated and is coordinated with the other post-entry functions of the capsid.

Figure 1 |. The early stages of HIV-1 replication.

The mature HIV-1 virus particle consists of the viral core surrounded by the viral membrane (envelope). The membrane contains embedded envelope protein Env (composed of gp120 and gp41 subunits) and MA proteins that make up the matrix layer. The viral core consists of the capsid and its contents. The capsid is a protein shell comprising around 1,500 copies of CA. Inside the capsid are the viral RNA genome, bound nucleocapsid (NC) proteins and the viral enzymes reverse transcriptase (RT) and integrase (IN). To initiate infection, Env binds to the CD4 receptor and co-receptor (CCR5 or CXCR4) on the host cell surface. Fusion of the viral and cellular membranes releases the viral core into the cytoplasm. The viral core then traffics to the cell nucleus where it docks at the nuclear pore. Reverse transcription converts the viral RNA genome into double-stranded DNA, forming the pre-integration complex (PIC) that is imported into the nucleus. The viral genome is then integrated into the host cell chromosome. Prior to or during nuclear import, the capsid disassembles in a process termed uncoating. In a restriction-competent cell, TRIM5 intercepts and disables the core in a manner that induces failure of reverse transcription.

The CA protein, which makes up the viral capsid, is the key genetic determinant of TRIM5 recognition, and early biochemical studies have established that it is the capsid itself – the assembled form of CA – that is bound by TRIM5 rather than soluble forms of CA^6,7. The HIV-1 capsid consists of around 1,500 CA molecules, which are assembled following the geometric principles of a fullerene cone; a typical capsid contains around 250 hexamers and exactly 12 pentamers in a completely closed shell^8,9 (Box 1).

Box 1 |. The structure of retroviral capsids.

A retrovirus core consists of the capsid – a protein shell comprising 1,500 or more CA molecules – and its contents. The HIV-1 capsid has the architecture of a fullerene cone with around 250 hexamers and exactly 12 pentamers in a completely closed shell^8,9. The CA protein itself consists of two domains, called NTD and CTD (N-terminal and C-terminal domains)^154,155. In both the hexamer and pentamer, the NTDs make a central ring surrounded by a ‘belt’ of CTDs^156–158. The CTDs connect the hexamers and pentamers into a hexagonal lattice with a characteristic spacing of around 100 Å. In the assembled capsid, the centers of mass of the NTDs and CTDs are shifted such that adjacent rings are separated by a deep ‘canyon’ around 30 Å wide and 40 Å deep, with the outer edges of the NTD hexamers and pentamers forming the walls of the canyon and the CTD forming the floor. The outer surface of the capsid therefore has a complex topology that arrays a number of protrusions, pockets, and pores that can mediate interactions with binding partners. Protruding bulges and loops are found primarily on top of the NTD rings; the most prominent of these is the flexible cyclophilin-binding loop that inserts into the active site pocket of the cellular prolyl rotamase CypA and TRIMCyp^91,95. Along the ridges and walls of the canyon are pockets and crevices, some of which are known to act as binding sites, including the ‘NTD-CTD interface’ pocket that binds the cellular factors CPSF6 and NUP153 as well as experimental capsid-targeting drugs like PF74^115,159,160. A more recently discovered binding site is the central pore of the hexamer, which is surrounded by the N-terminal β-hairpin of CA and lined by a ring of arginine residues that coordinate ionic interactions with nucleoside triphosphates and the capsid stability factor, inositol hexakisphosphate (IP6)^20,21,161.

Post-entry, the HIV-1 core travels to the cell nucleus, reverse transcription of the genome is completed, and the resulting pre-integration complex (PIC) docks at the nuclear pore. At some point during these steps, the capsid disassembles in a process called uncoating. TRIM5-mediated restriction is associated with defects in uncoating, and so it is important to consider the current understanding of how this process occurs, especially under non-restricted infection conditions. As this topic has already been comprehensively reviewed^10–12, we highlight below the aspects of uncoating that are particularly relevant to TRIM5.

Two models currently exist for core uncoating (refs. ^10–12 and references therein). In one model, the capsid shell remains fully intact until it docks at the nuclear pore, and its disassembly occurs as the viral DNA is imported into the nucleus. In a second model, the capsid undergoes some form of structural remodeling in the cytoplasm that results in capsid rupture or loss of integrity in the assembled CA shell, but a portion of the hexagonal CA lattice remains intact until disassembly is completed at the nuclear pore. In either case, the key point is that some form of an assembled CA lattice remains associated with the PIC until it docks at the nuclear pore.

A growing number of studies point to an important relationship between uncoating and the progression of reverse transcription. CA mutations that either increase or decrease the intrinsic stability of the capsid cause loss of infectivity, which is associated with defects in production of reverse transcripts¹³. These observations imply that capsid stability is optimized to promote reverse transcription. In addition, inhibiting reverse transcription delays the onset of uncoating^14,15. Atomic force microscopy analysis of purified HIV-1 cores directly demonstrated that reverse transcription increases pressure inside the capsid until the fullerene structure ruptures and eventually disassembles¹⁶, although it is still unclear at what step in DNA synthesis loss of capsid stability occurs^14,16–19. Interestingly, recent studies have shown that inositol hexakisphosphate (IP6) binds and potently stabilizes the mature HIV-1 capsid^20,21. IP6 appears likely to oppose the structural effects of reverse transcription and allows the capsid lattice to persist until it docks at the nuclear pore. As TRIM5 induces aberrant uncoating and inhibits reverse transcription, studies that further clarify the relationship between these post-entry processes will be key to a full understanding the mechanism of TRIM5-mediated restriction.

Structure of TRIM5

Being members of the TRIM protein family, TRIM5 proteins contain an N-terminal RBCC or tripartite motif, which consists of RING, B-box 2, and coiled-coil domains²² (Fig. 2a). This motif is followed by a C-terminal domain that is required for capsid recognition: SPRY (also known as PRY-SPRY or B30.2) in TRIM5α or cyclophilin A (CypA) in TRIMCyp. The fundamental oligomeric state of TRIM5 is a dimer^23–25, in which two coiled-coil domains pack in an anti-parallel manner to form an elongated rod that is capped at each end by a B-box 2 domain²⁶ (Fig. 2b). The B-box also independently mediates self-association^27,28, having a plastic oligomerization interface that can form both dimers and trimers^29–32 (Fig. 2c). The combined oligomerization properties of the coiled-coil and B-box 2 domains confer upon TRIM5 the propensity for higher-order assembly^{27,28,33–35}. In vitro, purified TRIM5 protein dimers spontaneously assemble into a hexagonal lattice³⁵. In cells, TRIM5 forms cytoplasmic assemblages or ‘bodies’ when overexpressed³⁶, and these are reasonably believed to be made of the same hexagonal lattice observed in vitro (Fig. 2e). By live cell imaging, TRIM5 cytoplasmic bodies appear as discrete puncta that rapidly exchange with the more diffuse cytoplasmic fraction³⁷. Pre-formed cytoplasmic bodies are not required for retroviral restriction^38–40, but higher-order assembly is required for core recognition and cytoplasmic bodies can form de novo around an incoming retroviral core⁴¹ (Fig. 2e).

Figure 2 |. Structural and functional properties of TRIM5.

a | Schematics of the primary sequences of TRIM5α and TRIMCyp. These proteins share a common tripartite motif (TRIM, also known as RBCC motif), composed of RING, B-box 2 and coiled-coil domains. The RING and B-box 2 domains are connected by a flexible segment, Linker 1 (L1). Linker 2 (L2) connects the RBCC scaffold to the capsid-binding domains, SPRY in TRIM5α and CypA in TRIMCyp. b | The B-box 2 and coiled-coil domains make up the basal dimer scaffold structure of TRIM5²⁶ (PDB code 4TN3), and defines the spatial dispositions of the N-terminal RING and C-terminal capsid binding domains. c | The B-box 2 domain forms trimers²⁹ (PDB code 5IEA) that can connect three dimers. The combination of coiled-coil mediated dimerization and B-box mediated trimerization generates a TRIM hexagonal lattice. d | The N-terminal RING domain dimerizes in order to bind a ubiquitin-conjugated E2 enzyme⁴² (PDB code 4TKP) and catalyze ubiquitination. c | In cells, TRIM5 undergoes higher-order assembly to form cytoplasmic bodies. These bodies can assemble spontaneously or form around an incoming retrovirus core.

Both the basal coiled-coil mediated dimer and the B-box mediated hexagonal lattice assemblies function as scaffolds that regulate the spatiotemporal organization and functional interactions of the N-terminal RING and C-terminal capsid-binding domains. The RING domain is attached to the B-box 2 domain by a short linker (linker 1 or L1), which constrains its ability to function as an E3 ligase enzyme. This is because the RING domain engages a ubiquitin-loaded E2 conjugating protein as a dimer; the two RINGs cooperate during catalysis by configuring both the E2 and ubiquitin moieties to maximize the reactivity of the thioester-linked ubiquitin tail for acceptor lysine residues or primary N-terminal amines⁴² (Fig. 2d). As the L1 linker that connects the RING and B-box 2 domains does not have sufficient reach to allow dimerization of two RING domains from opposite ends of the elongated coiled-coil dimer, RING activation requires (and promotes^28,32,43,44) higher-order association of TRIM5^29,42. The C-terminal SPRY or CypA domains are connected to the coiled-coil by an extended linker region called linker 2 or L2, part of which folds back along the coiled-coil rod to position two capsid-binding domains near the middle of the dimer^25,26. In TRIM5α, L2 is proposed to facilitate direct packing of the SPRY domain against the coiled-coil scaffold^25,45 (Fig. 3a). In contrast, the CypA domains of TRIMCyp are tethered more flexibly by an additional length of disordered residues²⁶ (Fig. 3b).

Figure 3 |. Mechanism of core recognition by TRIM5.

a | In TRIM5α, the two SPRY domains are thought to function as a single unit to bind the assembled CA subunits on the capsid. The structure of the SPRY domain is known^48,59 (PDB code 2LM3), and binding requires the flexible V1 loop. However, it is not yet known precisely how the SPRY domain contacts CA. b | In TRIMCyp, the two CypA domains are connected flexibly to the TRIM scaffold and binds to cyclophilin-binding loops displayed on the surface of the HIV-1 capsid⁹¹ (PDB code 1AK4). c | Model of a TRIM-coated capsid, assembled from crystal structures of the HIV-1 CA hexamer¹⁵⁷ (PDB code 3H47) and pentamer¹⁵⁸ (PDB code 3P05), and rhesus TRIM5α coiled-coil dimer²⁶, B-box 2 trimer²⁹, and SPRY domain⁵⁹ (PDB codes 3H47, 3P05, 4TN3, 5EIA, and 2LM3).

Mechanism of capsid recognition

All retroviral capsids are organized as fullerene structures with hexagonal lattice symmetry (Box 1), and it is this shared architecture that underlies the mechanism of capsid recognition by TRIM5. Recognition is described by a model (Fig. 3) in which: (1) the SPRY and CypA domains each bind to some epitope on CA, (2) the individual pair-wise affinities are weak, and (3) higher-order assembly mediated by the TRIM5 coiled-coil and B-box 2 domains amplifies the weak interactions by positioning arrays of the capsid-binding domains to match the pattern of CA binding epitopes on the capsid surface. Here, we describe the development of this ‘pattern recognition’ model and highlight remaining unresolved questions.

Avid capsid recognition by TRIM5α.

Early studies established that TRIM5α does not bind the isolated CA subunit, and that binding can only be detected in vitro when the CA proteins are assembled as capsid-like tubes^{6,7,23,24,46,47}. Measurable binding can be observed between the isolated SPRY domain of rhesus TRIM5α and a soluble form of the HIV-1 CA hexamer, but even with this hexavalent CA assembly the affinity remains very weak^48,49. Other studies showed that the self-association properties of the TRIM5α coiled-coil and B-box 2 domains are critical for restriction activity^27,50,51. Taken together, these indicated that higher-order assembly of both TRIM5α and CA amplifies very weak pair-wise affinities between SPRY and CA.

The molecular basis for avidity-driven binding arose from biochemical analysis of TRIM5α proteins, which were found to assemble spontaneously into hexagonal lattices in vitro³⁵. This led to the realization that TRIM5α can match both the symmetry and spacing of the capsid lattice and generate powerful avidity effects. In support of this model, pre-assembled mimics of the HIV-1 capsid were shown to act as templates for TRIM5α assembly in vitro^35,52. Subsequently, it was shown that TRIM5α proteins also assemble on the surface of authentic HIV-1 cores⁵². This recognition mechanism explains why TRIM5α binding to a capsid is highly cooperative⁵³ and why the restriction capacity is saturable^54–57. Only about 25% of the CA subunits in a capsid need to be susceptible for restriction to occur⁵³, because once assembly is initiated it will proceed to completion.

Structural and biochemical studies have now revealed the atomic details of the TRIM5α hexagonal lattice^{25,26,29,30,32,35,52}. The lattice has the appearance of a hexagonal wire mesh or a tessellation of hexagons, in which each hexagon edge corresponds to a TRIM5 dimer and each vertex corresponds to a B-box 2 domain trimer. All of the SPRY domains are located on one face of this lattice, enabling simultaneous binding to multiple sites on the capsid surface. The RING domains are located on the opposite face, where they are activated for self-ubiquitination at the B-box vertices.

An important unanswered question is precisely how the SPRY domain contacts the CA subunits. Comparative sequence analyses have revealed that the SPRY domain contains variable loops, termed V1-V4⁵⁸. These loops show strong evidence of positive selection, a phenomenon that reflects the co-evolution of two proteins in conflict, in this case a restriction factor and its viral protein target. Structural studies indicate that the SPRY domain has a β-sandwich fold that displays all four loops on one side^48,59, where they can act independently or in concert to bind CA. The V1 loop (~20 amino-acid residues) harbors the principal determinants for capsid binding specificity^60–70 and is conformationally flexible^48,59,71, suggesting that binding may involve insertion of loop residues in a pocket or crevice on the capsid surface. However, a large number of studies map susceptibility and resistance determinants to extensive surface regions of both the HIV-1 and MLV capsids^{64,70,72–86}, suggesting that the SPRY domain may contact CA in multiple different ways. Other studies suggest that a single SPRY domain bridges adjacent CA hexamers on the capsid surface, perhaps by binding the canyon walls (Box 1) that surround either the two-fold or three-fold symmetry axes of the CA lattice^48,49,71,87. Finally, it was also suggested that the two SPRY domains in a TRIM5α dimer act as a single unit and not independent of each other^25,45. Detailed molecular insights on how SPRY contacts CA still await experimentally determined structures that can be used to integrate all of these observations.

Target specificity.

Genetic analyses of rhesus macaque TRIM5α have identified two allelic lineages whose relative specificities for HIV and SIV strains are dictated by either a ‘TFP’ (threonine-phenylalanine-proline) or ‘Q’ (glutamine) motif in the V1 loop⁶⁷. Pair-wise tests of these TRIM5α alleles and their targets have revealed a variety of specificity determinants on the surface of CA, including a conserved patch on the NTD⁸⁴. Another case is human TRIM5α, which restricts N-tropic but not B-tropic MLV^74,75. Here, susceptibility seems dictated primarily by a single amino acid residue on the NTD, which is located near the three-fold symmetry axis of the MLV capsid lattice^75,77. While these and related studies provide insight on how TRIM5α can discriminate between closely related retroviruses, the actual specificity range for a given TRIM5α protein typically spans different retroviral genera. For example, most Old World monkey TRIM5α proteins restrict HIV-1 (a lentivirus) and N-MLV (a gammaretrovirus)⁸⁸. The CA proteins of these retroviruses have very little sequence homology and their capsids display dramatically different patterns of amino-acid residues on their surfaces, with little evidence of a common binding epitope. An individual SPRY domain therefore must have significant binding plasticity. It was suggested that the SPRY V1 loop behaves like so-called ‘intrinsically disordered proteins’, which do not fold into single rigid structures but rather into ensembles of transient conformations that can mediate binding to different partners^71,89. Such a mechanism could explain how the SPRY domain can bind to different capsids or even different epitopes on the same capsid, but has not yet been explored fully. Furthermore, factors outside the SPRY domain may come into play; for example, the coiled-coil domain has also been suggested to modulate restriction specificity^45,90.

Capsid recognition by TRIMCyp.

The capsid-binding domain of TRIMCyp is homologous to the cellular prolyl rotamase enzyme CypA, whose interaction with the HIV-1 CA monomer has been well-characterized^91–95. CypA binds to the cyclophilin-binding loop that is prominently exposed on the NTD of HIV-1 CA with a dissociation constant of 10 µM. A key proline residue (Pro90) in this loop binds to the enzyme active site and mutation of this residue essentially abolishes the interaction⁹³. When bound to the assembled CA lattice, CypA is reported to also have a secondary contact with another NTD in the adjacent CA hexamer⁹⁶, and so the TRIMCyp dimer can potentially contact four CA subunits simultaneously. Unlike the SPRY domains, the two CypA domains are more flexibly tethered to the coiled-coil domain, and so each TRIMCyp dimer presumably has a wider reach and can sample a larger area to optimize local binding²⁶. Some studies have suggested that TRIMCyp might have dispensed with the need for higher-order assembly to efficiently recognize retroviral capsids, consistent with the significantly higher affinity of the CypA domain for the CA subunit. In support of this idea, B-box 2 mutations that abrogate both the higher-order assembly and restriction activities of TRIM5α were reported to have no effect on TRIMCyp⁶⁵. However, more recent studies have shown that TRIMCyp assembles into the same hexagonal lattice as TRIM5α⁵², and that the dependence of TRIMCyp recognition on higher-order assembly is unmasked by cyclosporine, which is a compound that inhibits CypA binding to CA³². Therefore, both forms of TRIM5 use the same mechanism of avidity-driven capsid recognition.

Mechanism of restriction

The precise mechanism by which TRIM5 proteins inhibit virus replication has been challenging to pinpoint, in part because the natural progression of infection in between the entry and integration steps is not yet fully understood. The simplest scenario is that binding in itself arrests the ordered progression of post-entry steps or perhaps diverts the core towards a non-productive or ‘off-pathway’ trajectory. This view is founded upon early studies showing that TRIM5 induces premature or non-productive uncoating that then disrupts reverse transcription^7,50. These and many subsequent studies can now be integrated in a multi-step mechanism in which: (1) the virus core is recognized through multivalent TRIM5 binding, which may directly inactivate the virus, (2) effector functions of the RING domain increase the efficiency of inactivation and recruit cellular degradation machinery that clear the debris, and (3) ubiquitination-dependent pathways activate innate immune pathways to signal the presence of an invader. Here, we summarize studies that describe how TRIM5 affects the biochemical and functional behavior of the core and note how the available data inform mechanistic understanding of TRIM5-mediated restriction (Fig. 4).

Figure 4 |. Mechanism of TRIM5-mediated restriction.

TRIM5 recognizes an incoming retrovirus core by binding to the capsid shell that coats and protects the core. The assembled TRIM5 proteins make a cage that surrounds and traps the core. a | Some studies indicate that binding of TRIM5 to the capsid is sufficient to induce premature and non-productive uncoating^{7,24,39,50,101}. This leads to failure of reverse transcription. b | Other studies indicate that accelerated core dissociation induced by TRIM5 requires ubiquitination and recruitment of proteasomes^104,110,111. Proteasome recruitment can also lead to virus clearance after the core has been inactivated. A very recent study shows that immunoproteasomes, a specific type of proteasome, allows HIV-1 restriction by human TRIM5α¹²⁸. c | Other studies invoke virus clearance through a mechanism of precision autophagy termed virophagy, in which TRIM5 directs the bound retrovirus core to isolation membranes and, eventually, lysosomes for degradation¹²⁰.

Restriction by capsid binding.

The notion that core stability is optimized to promote reverse transcription and infectivity implies that capsid-binding molecules that perturb this delicate balance may be inhibitory in certain contexts. This is supported by studies that generate artificial restriction factors by linking two or more monomers of CypA with unrelated scaffolding proteins^97,98. Other studies show that TRIMCyp proteins containing deleterious mutations in the RING or B-box 2 domains still retain significant restriction activity, in some cases at apparently the same levels as wildtype TRIMCyp^32,46. These studies provide proof of principle that a bivalent or trivalent capsid-binding protein is intrinsically inhibitory, perhaps by interfering capsid stability or with critical post-entry interactions of the core. In vitro, the TRIM lattice can completely cage purified cores and recombinant capsid mimics^29,52, which is likely to be much more effectively inhibitory than simpler oligomers. However, many other studies indicate that the effects of TRIM5 on the core and its capsid shell are more nuanced than a model of generalized functional disruption would suggest.

Binding-induced non-productive uncoating.

The idea that uncoating is spatially and temporally coupled with reverse transcription also implies that premature dissociation of the core can have negative consequences for the virus. This was first tested by the ‘fate of capsid’ assay, a gradient centrifugation protocol which revealed that yields of pelletable and presumed core-associated HIV-1 CA protein are significantly reduced in cells that express TRIM5^7,99. In some cases, loss of particulate CA is accompanied by corresponding increase in soluble CA, which is taken to indicate disassembly^50,65 (Fig. 4a). Subsequent studies showed that both the core-associated viral RNA and IN protein are also reduced under conditions of TRIM5 restriction¹⁰⁰. Core-associated reverse transcription products are likewise reduced, demonstrating that the loss of reverse transcripts caused by TRIM5 directly correlates with accelerated core dissociation. However, for some of these components (IN in particular) a corresponding increase in soluble form is not observed, suggesting that TRIM5-mediated restriction involves both disassembly and degradation, with various core components having differing fates¹⁰⁰.

Biochemical studies indicate that purified TRIM5 proteins can induce breakage of in vitro capsid mimics, supporting the idea that TRIM5 binding directly destabilizes the capsid lattice^24,47,101. Proposed mechanisms for binding-induced capsid destabilization include local mismatches between the TRIM and CA lattices that can create discontinuities in the capsid shell³⁵, conformational changes in the L2 linker connecting the TRIM5α SPRY and coiled-coil domains that can mechanically disassemble the capsid lattice¹⁰², and global structural perturbations of the capsid¹⁰³. However, other studies show that stable complexes of HIV-1 CA tubes and TRIM5α proteins can be generated in vitro^29,52. Likewise, stable TRIM5/capsid complexes persist in the cell when ubiquitination or the proteasome is inhibited¹⁰⁴. These can be interpreted to mean that TRIM5 binding does not intrinsically induce dissociation of the CA lattice. A recent study shows that TRIM5α actually induces global rigidification of the CA lattice, while at the same time inducing structural and dynamic changes that map to large portions of the CA tertiary structure¹⁰³. Thus, there are apparently contradictory data on the effect of TRIM5 on capsid stability, and this needs to be resolved. Furthermore, the biochemical data now need to be reviewed in light of the recent discovery that IP6 is a powerful modulator of HIV-1 capsid assembly and stability^20,21. Since this small molecule has not been employed in previous studies, its influence on capsid stability in context of TRIM5 binding is yet unknown.

Ubiquitin-dependent non-productive uncoating.

A number of other studies indicate that the accelerated core dissociation induced by TRIM5 is not a simple consequence of the binding interaction, but rather a more complex phenomenon (Fig. 4b). TRIM5 cytoplasmic bodies (both in the presence or absence of virus cores) are linked with ubiquitination and ubiquitin-dependent cellular degradation machinery such as proteasomes^105–107. In context of restriction, the ‘fate of capsid’ assay shows that in the presence of proteasome inhibitors, the levels of particulate CA and core components resemble those observed under non-restricting conditions^65,100. TRIM5α undergoes self-ubiquitination and inhibiting this modification abrogates accelerated uncoating and restores reverse transcription^108–110, presumably because the self-attached ubiquitin chains recruit proteasomes¹⁰⁵. Collectively, these data demonstrate that both in vitro and in cells TRIM5 and HIV-1 cores initially form stable complexes, and that the ubiquitin-proteasome system is required to non-productively accelerate disassembly of the capsid coat. It is important to note, however, that the ubiquitin-proteasome system is not required to inhibit virus replication³⁹, and even under conditions where reverse transcription is restored by proteasome inhibitors or RING domain mutations, infectivity is not restored^{44,104,108,111}. Thus, either TRIM5 imposes sequential blocks to the post-entry pathway, or TRIM5 binding diverts the core into an ‘off-pathway’ trajectory that cannot be corrected by inhibiting the proteasome or disrupting TRIM5 self-ubiquitination. Integration-competent PICs are generated by HIV-1 when restriction assays are performed in the presence of proteasome inhibitors¹¹¹ or with RING domain TRIM5 mutants¹⁰⁸, which appears to support the first model.

The effect of proteasome inhibitors and RING domain mutations on TRIM5 restriction might be explained by analogy to recent studies showing that reverse transcription causes rupture of the HIV-1 capsid shell in vitro and the effects of the small molecule inhibitor PF74 on this process^16,112. TRIM5 may have the same effect as PF74, in that the assembled TRIM cage stabilizes the capsid lattice against rising pressure from inside the core. Concomitant self-ubiquitination of TRIM5 would then recruit proteasomes which induce disassembly of the TRIM-coated capsid and degrade key core components, explaining failure of reverse transcription and phenotypes observed in ‘fate of capsid’ and other assays. Under conditions where ubiquitination is inhibited, capsid rupture may still occur but the TRIM5 cage would stabilize the remaining lattice. This may then permit reverse transcription but the resulting PIC remains disabled, perhaps because the TRIM5 cage also interferes with other functions of the capsid such as engagement of nuclear import and integration machinery. In this regard, it is interesting to note that PF74 shares the key phenotypic effects of TRIM5, including apparently contradictory observations of destabilization or stabilization of the CA lattice^113–116, mechanical rigidification of the capsid^112,117, and induction of accelerated uncoating in cells with concomitant disruption of reverse transcription^114,118.

Virophagy.

One model proposes that TRIM5 is a selective autophagy receptor that directs degradation of the retrovirus core through a mechanism termed ‘virophagy’ (Fig. 4c). This is supported by studies showing that TRIM5α binds and activates key autophagy effectors, including p62, ULK1, and Beclin 1^119,120. TRIM5 also directly binds LC3 and other Atg8 homologs^120,121, which are ubiquitin-like proteins that are conjugated to phosphatidylethanolamine and marks autophagosomal membranes¹²². However, autophagy is not required for TRIM5-mediated inhibition of virus infectivity^121,123, which parallels observations on the proteasome^39,104,111. It may be that the proteasome and autophagy pathways are redundant or cell type-dependent mechanisms of virus clearance, which operate after the core has already been disabled.

TRIM5-mediated signaling

In addition to directly inhibiting retrovirus replication, TRIM5 also activates ubiquitin-dependent innate immune signaling upon capsid recognition and induces the so-called ‘anti-viral state’¹²⁴. TRIM5 proteins therefore constitute a class of ‘pattern recognition receptor’ (PRR) that senses a retroviral capsid as a ‘pathogen-associated molecular pattern’ (PAMP). Like other PRR, both TRIM5α and TRIMCyp expression levels are significantly upregulated by interferon^125,126, and TRIM5-mediated pattern recognition of retroviral capsids may contribute to virus control during the natural course of infection in vivo^127–129.

An important question is how TRIM5 self-ubiquitination is regulated to activate signaling in the presence of virus and yet avoid spontaneous signaling in the absence of viral infection (Fig. 5). As discussed above, the anti-parallel coiled-coil dimer positions two RING domains on opposite ends of a 20-nm long rod^25,26. Since enzymatic activity requires RING dimerization⁴², ubiquitination reactions are enabled when the B-box 2 domain self-associates and brings more than one RING domain into close proximity. According to one model, B-box dependent association of the TRIM5α dimers recruits the E2 conjugating enzyme Ube2W, which selectively attaches a single ubiquitin moiety at the TRIM5α N-terminus^44,109,130. In the presence of retroviral cores, higher-order assembly of the TRIM5α hexagonal lattice generates B-box domain trimers, which clusters three RING domains. The three RINGs then undergo a dynamic ‘two-plus-one’ ubiquitin chain elongation mechanism, in which another E2 – Ube2N/Ube2V2 (also known as Ubc13/Mms2) – attaches K63-linked polyubiquitin chains to the N-terminal monoubiquitin⁴⁴. The K63-linked chains then activate interferon through the AP1 and NF-κB pathways^44,124. The cascade of capsid-induced assembly, assembly-dependent self-ubiquitination, and polyubiquitin extension by the trivalent RING clustering is proposed to constitute a ‘proofreading’ mechanism that tests for the presence of an invading capsid and TRIM5α assembly before producing an inflammatory signal¹³⁰.

Figure 5 |. TRIM5-mediated signaling.

TRIM5 protein levels are normally kept low by rapid turnover, either through proteasomal or autophagosomal degradation. Cage formation around a restriction-susceptible capsid triggers production of K63-linked polyubiquitin chains, which activate interferon signaling through the AP1 and/or NF-kB pathways^44,124. The signaling K63-linked polyubiquitin chains are reported to be either unanchored¹²⁴ or attached to the TRIM5 N-terminus⁴⁴. Interferon induction is not required for virus inhibition in tissue culture systems, but likely contributes to effective virus control in vivo.

In the absence of virus, signaling is prevented by rapid TRIM5α protein turnover. This is proposed to be facilitated by the ubiquitin fusion degradation (UFD) machinery, which efficiently degrades N-ubiquitinated substrates via the proteasome⁴⁴. However, other studies have found that in non-infected cells, inhibition of the proteasome^104,131,132 or disruption of ubiquitination⁴⁰ only mildly affect TRIM5α protein levels, whereas treatment with autophagy inhibitors increases protein TRIM5α levels and promotes cytoplasmic body accumulation^123,131. Thus, it appears that whether constitutive TRIM5α degradation occurs through autophagy or proteasomes remains to be resolved.

Human TRIM5α and HIV-1

Humans express a TRIM5α ortholog (but not TRIMCyp), which early studies showed to only weakly restrict HIV-1^4,72,74,133. This lack of potency was explained to arise primarily from inefficient core recognition, because a single point mutation (R332P) in the SPRY domain enhances restriction^60,61 and capsid binding in vitro⁵². Subsequent studies have shown that various factors outside of the SPRY/CA interaction can modulate restriction efficiency. For example, human TRIM5α’s restriction activity is proposed to be modulated by shorter protein variants (termed TRIM5γ, δ, ε, ι, κ, and ζ) that arise from alternative messenger RNA splicing^4,36,134. Most of these variants lack the SPRY domain and consequently cannot bind the HIV-1 capsid; nevertheless, they contain an intact coiled-coil dimerization domain and strongly inhibit TRIM5α activity as dominant negatives^4,134–136. Other studies have uncovered virus strain-specific and/or cell type-specific modes of restriction^136–139. Of particular note, HIV-1 variants that arise from selection pressure by cytotoxic T lymphocyte (CTL) responses have been shown to be highly sensitive to human TRIM5α^136,137,139.

Recent studies now indicate that human TRIM5α is a major determinant of type 1 interferon-mediated suppression of HIV-1 replication^127,128,140. This observation provides a significant boost to the idea that human TRIM5α contributes to virus control during the acute and chronic phases of a natural HIV-1 infection, which are characterized by high interferon levels¹⁴¹. Mechanistically, human TRIM5α restriction of HIV-1 is licensed by interferon-mediated stimulation of the immunoproteasome¹²⁸, which is distinguished from the constitutive proteasome by different catalytic subunits and its dependence on the PA28 regulatory complex^142,143. The immunoproteasome induces more rapid TRIM5α turnover¹²⁸, which dovetails nicely with the proteasome-dependent mechanism of virus clearance suggested by previous studies^{44,104,110,111}. Interestingly, this mechanism of restriction is necessarily RING and ubiquitin-dependent¹²⁸, which is distinct from rhesus TRIM5α-mediated restriction of HIV-1. This can be interpreted to mean that more efficient routing of human TRIM5α to a cellular degradation pathway may compensate for its weak initial binding relative to rhesus TRIM5α. This highlights a fascinating property inherent to a multi-step mechanism, in that modulating any individual step fine tunes the overall restriction efficiency.

Relevance to anti-HIV strategies

A number of studies have started to explore the possibility of improving the potency of human TRIM5α against HIV-1 for the purpose of anti-viral gene therapy^68,144–147. In principle, editing the human TRIM5 gene may be used to enhance capsid-binding efficiency, for example by mutating the SPRY domain or inserting a CypA domain as found in TRIMCyp. Such a strategy will need to be balanced against potential effects on other functions of TRIM5α in the cell (which are currently unknown) and on the ability to protect against other extant or undiscovered human retroviruses. Another possible approach is to identify compounds that bridge binding of the human TRIM5α SPRY domain and the HIV-1 CA protein, in effect mimicking the effect of the R332P mutation that enhances human TRIM5α restriction of HIV-1⁸⁸. Small molecules may also be identified that promote assembly or stability of the TRIM hexagonal lattice, as has been found for other systems^20,148,149. Because the TRIM lattice acts as an affinity amplification mechanism for the SPRY domain, enhancing its assembly would in principle also boost binding. Further assessment of this approach should be facilitated by a more quantitative understanding of the degree to which the hexagonal lattice actually amplifies the basal SPRY affinity for CA, which is still unknown. As described above, more efficient routing of TRIM5α-core complexes to a cellular degradation pathway¹²⁸ may also be a viable strategy of enhancing restriction. More generally, the existence of TRIM5α and other naturally evolved capsid-binding restriction factors demonstrates that the HIV-1 core and its surrounding capsid are attractive anti-viral targets¹⁵⁰. As mentioned above, the key phenotypic effects of TRIM5α on HIV-1 replication – inducing non-productive uncoating and inhibiting reverse transcription – are shared by known small molecule inhibitors such as PF74¹¹⁴. Given the developing understanding of TRIM5-mediated restriction and the relationships between capsid stability, uncoating, and reverse transcription, it can now be appreciated that a counterintuitive strategy of screening for molecules that promote mature capsid assembly and/or stabilize the CA lattice is likely to yield compounds that elicit the same effects as TRIM5 and PF74, regardless of the actual binding mechanism. A recently described compound, GS-CA1 from Gilead Sciences, binds in the same position as PF74 and exhibits high anti-viral potency^150,151. Although its detailed mechanism of action is still under investigation, the drug shows great promise as a long-acting injectable.

Concluding remarks

The study of TRIM5 restriction factors has proven to be a fascinating source of novel insights on the biology of host-virus interactions. The ability of TRIM5 to recognize the large-scale architecture of an entire retroviral capsid – ~40 megadaltons in size – provides fresh appreciation for the cell’s ability to distinguish self from non-self, which is the bedrock of immunity. Elucidating its detailed mechanism of action has been and remains an important and interesting challenge. As described in this review, TRIM5α and TRIMCyp use a lattice-matching mechanism to form a cage which binds the molecular patterns displayed on the surface of the capsid that coats the incoming retrovirus core. However, a consensus model that describes precisely how a TRIM5 cage disables a retrovirus core remains to be developed. Further studies are also required to elucidate precisely how the arrayed SPRY domains of TRIM5α contact their bound CA subunits. Another important unresolved issue is the mechanism of binding specificity. Whereas susceptibility to TRIMCyp is explained by the presence or absence of a compatible cyclophilin-binding loop in CA (but also see refs. ^{94,95,152,153}), the molecular basis of TRIM5α binding specificity largely remains a mystery. Given the apparent importance of capsid stability for productive uncoating, it seems possible that at least some of the CA mutations that result in gain or loss in restriction susceptibility to TRIM5α may not necessarily reflect effects on binding but rather allosteric effects on the network of protein-protein interactions that form the capsid lattice. Understanding how TRIM5α recognizes some retroviruses but not others should be particularly informative for efforts to efficiently target human TRIM5α against HIV-1.

One theme that seems to be arising is that TRIM5 proteins may route retrovirus cores to different clearance or degradation pathways in different contexts, as exemplified by recent studies of human TRIM5α^128,138. Further exploration of this idea may resolve current disagreements as to the relative roles of proteasomes and autophagosomes in TRIM5-mediated restriction.

Finally, this is an opportune time for studies of HIV-1 uncoating. The recent discovery of IP6 as a capsid stability factor answers the previous conundrum of how the apparent instability of the capsid in vitro can be reconciled with its behavior in cells^20,21. IP6 should now facilitate purification and biochemical manipulation of authentic HIV-1 cores and direct examination of the structural transformations that occur during reverse transcription. Continued advancements in live cell imaging and correlative light and electron microscopy also promise direct visualization of these processes in the cell. Our developing knowledge of uncoating during the natural course of HIV-1 infection will undoubtedly bolster understanding of TRIM5-mediated restriction, and perhaps uncover novel vulnerabilities that may be targeted with therapeutics.