Cyclophilin A, TRIM5, and Resistance to Human Immunodeficiency Virus Type 1 Infection

Upon entry of a retrovirus into the cytoplasm of a permissive cell, a viral protein, reverse transcriptase, generates cDNA using viral genomic RNA as template. In a subsequent but also essential step in the retroviral replication cycle, the nascent viral cDNA is ligated to chromosomal DNA by another viral protein, integrase, establishing the provirus. Retroviral infection is thus mutagenic to the host. One highly publicized illustration of the potential for retroviruses to disrupt host gene expression involved activation of the LMO2 proto-oncogene by proximal integration of a vector in a gene therapy trial for X-linked severe combined immunodeficiency (33).

Aside from altering host gene expression, retroviruses such as human immunodeficiency virus type 1 (HIV-1) may cause pathology by transducing toxic genes (24). In addition to the genes encoding structural proteins and enzymes, which are themselves toxic to host cells, HIV-1 possesses six accessory genes that exhibit wide-ranging effects on cell physiology. Despite immune responses that decrease HIV-1 viremia after acute infection and appear to limit viral pathology for a decade, HIV-1 ultimately kills the host, most likely via complex effects involving all nine of the transduced viral genes.

Given the potent effects of retrotransposition on biology and evolution, it is not hard to imagine that susceptible host organisms would elaborate factors that block reverse transcription or any of the steps that lead to integration. Several such factors have been identified and have been discussed previously in a comprehensive fashion (7, 31). This review will focus on how, in the course of efforts to identify retroviral inhibitory factors, cyclophilin A (CypA) was discovered to be an HIV-1 CA-binding protein. It will then describe how the characterization of the HIV-1 CA-CypA interaction revealed TRIM5 to be a potent antiretroviral restriction factor. Finally, it will present our current understanding of the mechanism of retroviral inhibition by TRIM5 and explain how CypA modulates retroviral restriction activity.

Fv1 AND CA-SPECIFIC RESTRICTION

Fifty years ago, Charlotte Friend described a transmissible, leukemia-like illness in mice (28). Her concomitant observation that the ability of the etiologic “agent” to cause disease was dependent on the strain of the mouse sparked a search for the genetic basis of host resistance to retroviral infection. Frank Lilly, among others, identified several such resistance loci, including Fv1 (50, 80).

The Fv1 barrier presents itself early in retroviral infection, after membrane fusion, but the precise step that is inhibited has been difficult to pinpoint (Fig. 1). Fv1 is generally said to act after the preintegration complex enters the nucleus (64), though effects of Fv1 on reverse transcription have also been detected (92). Retroviral determinants for sensitivity to restriction by Fv1 map to particular amino acid residues in the gag-encoded CA (8, 19, 48, 67). Curiously, when Fv1 was finally cloned in 1996, it was found to resemble an endogenous gag (6). Though an understanding of the Fv1 restriction mechanism still eludes us, the extraordinary studies of Fv1 established CA as a determinant of retroviral tropism and as a target of cellular factors that regulate retroviral infectivity.

FIG. 1.

CA-specific restriction mediated by Fv1. Idealized virions bearing two RNA genomes (two vertical lines) coated with nucleocapsid (vertical oval) within a core made of CA (color-coded octagons). Virions are shown entering target cells (plasma membrane is indicated by the double horizontal line) and, when infection is successful, gaining access to chromatin DNA in the nucleus. Mouse strains bearing the Fv1^b allele (A) restrict infection by N-tropic MLV, which bears an arginine at CA residue 110 (core shown in red). B-tropic MLV has a glutamine at CA residue 110 (core shown in green) and is not restricted by Fv1^b. Mouse strains bearing the Fv1ⁿ allele (B) restrict infection by B-MLV but not by N-MLV. Though some studies indicate that Fv1 restriction activity exerts its effect after reverse transcription, the diagram attempts to accommodate all studies by showing restriction occurring after virion fusion with the plasma membrane and before integration within the nucleus of the target cell.

CypA SPRINGS FROM EFFORTS TO IDENTIFY CA-SPECIFIC RESTRICTION FACTORS

The genetically dominant restriction activity encoded by the different Fv1 alleles, along with the CA determinants for susceptibility to restriction, suggests that Fv1 directs the synthesis of a CA-binding protein (Fig. 1). Prior to the cloning of Fv1, the protein that it encodes was sought unsuccessfully in a yeast two-hybrid screen using CA from susceptible virus as bait. Even now that Fv1 has been cloned, an interaction between its protein product and the CA of a restricted retrovirus has not been detected. Instead, in a concurrent two-hybrid screen, it was discovered that HIV-1 CA binds the ubiquitous, cytoplasmic protein CypA (51).

Since the discovery of the HIV-1 CA-CypA interaction, it has been established that CypA promotes an early step in the infection of human cells by HIV-1 (10, 11, 13, 26, 27, 36, 77, 85). Via interaction with the CA domain of the Gag polyprotein, CypA from the virion producer cell is efficiently incorporated into virions (27, 58, 85). Nonetheless, in terms of clarifying the function of CypA in HIV-1 replication, this tantalizing observation has borne no fruit. Examination of CypA-deficient HIV-1 virions failed to detect abnormalities in virion protein content, viral precursor protein processing kinetics, viral genomic RNA packaging, endogenous reverse transcriptase activity, or virion ultrastructure (11, 13, 27, 46, 85, 88). Analysis of the retroviral replication cycle, using RNA interference to disrupt CypA in the virion producer cell—from which CypA is incorporated into virions—or in the target cell, indicated that target cell CypA alone promotes HIV-1 infectivity (36, 77). Perhaps a function for virion-associated CypA will one day be discovered.

HIV-1 CA DETERMINANTS FOR INTERACTION WITH CypA CORRELATE WITH THOSE FOR SPECIES-SPECIFIC TROPISM

Initial surveys indicated that HIV-1 and closely related chimpanzee viruses are the only retroviruses which encode a CA that binds CypA and, correspondingly, only these viruses are dependent on CypA for replication (12, 13, 26, 27, 51, 85). These observations raised the possibility that CA determinants for CypA binding, or for the genetically separable property of CypA dependence (10), might constitute tropism determinants for HIV-1. Indeed, compared to well-studied simian immunodeficiency virus (SIV) isolates, HIV-1 has a relatively limited host range (1, 5, 18, 38, 39, 49, 54, 73) that is attributable to a dominant and saturable anti-HIV-1 activity in target cells from nonpermissive species (5, 18, 54). The block to HIV-1 infection in nonhuman primate cells generally occurs prior to reverse transcription (5, 18, 34, 38, 39, 54, 74, 87), which is the same point in the retrovirus life cycle where CypA acts (11). The replication phenotype of chimeric viruses generated from HIV-1 and SIV sequences confirmed that the CypA-binding region of HIV-1 CA overlaps with viral determinants for species-specific tropism (4, 15, 18, 22, 35, 41, 47, 59, 60, 70, 73, 74, 81, 87).

CypA AND THE DISCOVERY OF TRIM5

Competitive inhibitors and genetic modifications that block the HIV-1 CA-CypA interaction were shown to decrease HIV-1 sensitivity to restriction activity in simian cells (3, 4, 47, 70, 87). This raised the possibility that CypA was somehow modulating restriction of HIV-1, but analysis of CypA coding sequences failed to identify species-specific polymorphisms that account for the block to HIV-1 infection in nonhuman primates, and simian CypA cDNA was insufficient to transfer the HIV-1 restriction activity to otherwise permissive human cells (70, 96).

It was then found that the anti-HIV-1 activity in owl monkey cells was eliminated by any of three different short hairpin RNAs (shRNAs) specific for CypA (70). However, restriction activity in the CypA knockdown cells was not restored when CypA protein was reintroduced. This instructive experiment was doggedly repeated over the course of a year by expressing CypA cDNA bearing silent mutations in the shRNA target sequence (70). The explanation for the failure to restore restriction activity by reintroduction of CypA protein was revealed by a screen of owl monkey cDNA for an alternative target of the CypA-specific shRNAs: owl monkeys express a remarkable mRNA fusion consisting of CypA and TRIM5 (Fig. 2) that had been targeted by the CypA-specific shRNA and was sufficient to account for the potent anti-HIV-1 restriction activity in owl monkey cells (70).

FIG. 2.

Schematic diagram showing the structural motifs in TRIM5α from rhesus macaques or most primates, including humans (A) and in the extremely unusual TRIMCyp from owl monkeys (B). The RING finger (RF), B-box zinc finger motif, and coiled-coil domains that define a protein as a tripartite motif family member are common to these two proteins. The two proteins are distinguished from each other by their very different CA-specific recognition domains: SPRY, SPRY or B30.2 domain; CypA, cyclophilin A domain.

The significance of TRIM5 for HIV-1 restriction had been made evident just one week prior to the discovery of owl monkey TRIMCyp (81). A rhesus macaque library had been screened for cDNAs that confer HIV-1 resistance to otherwise permissive human cells. Macaque TRIM5α cDNA was found among the rare cells that remained uninfected after challenge with HIV-1 at a high multiplicity of infection. The rhesus macaque TRIM5 gene gives rise to several isoforms via differential splicing, but TRIM5α was the only isoform that conferred HIV-1 restriction activity (Fig. 2).

TRIM5 is a member of the very large tripartite motif family of proteins (∼70 family members) that is defined by the presence of RING finger, B-box, and coiled-coil domains (57, 65). The alpha isoform is the only TRIM5 isoform that possesses a SPRY domain at the C terminus. Though the structure of TRIM5α has yet to be solved, reasonable models for the TRIM5α SPRY domain have been proposed based on the structures of other molecules that contain SPRY domains (32, 52, 89). It appears to fold as a β sandwich specialized for protein-protein interactions.

TRIM5α orthologues from humans, nonhuman primates, and cows were isolated subsequently and shown to account for restriction activity against a broad range of retroviruses, including HIV-1, HIV-2, SIVs from several nonhuman primate species, N-tropic murine leukemia virus (N-MLV), and equine infectious anemia virus (37, 45, 63, 75, 79, 94, 97). Among the TRIM family members that have been tested, modest retroviral restriction activity has also been detected with TRIM1 and TRIM34 (75, 78, 93, 94, 98).

TRIMCyp, then, is an unusual variant of TRIM5 that has been found only in the Aotus (owl monkey) genus and is the only TRIM5 allele found in all 10 species of owl monkeys (56, 66, 70, 79). It arose via retrotransposition of a complete CypA cDNA into TRIM5 intron 7 and has all the hallmarks of a cDNA integration event mediated by the enzymatic machinery of a LINE-1 element. TRIMCyp presumably conferred a potent survival advantage to the animal in which it arose, though these South American primates are not known to have been exposed to HIV-1. Recently, two additional TRIMCyp-sensitive retroviruses were identified, an SIV from Cercopithecus tantalus and a feline immunodeficiency virus (21, 98). The sequences of these two retroviruses suggest that both encode a CA that binds CypA. Consistent with this assessment, both viruses were rescued from TRIMCyp restriction by the competitive inhibitor cyclosporine.

PATTERN RECOGNITION RECEPTOR FOR THE MULTIMERIC PROTEIN LATTICE OF RETROVIRAL CAPSIDS

The carboxy-terminal CypA domain of owl monkey TRIMCyp and the similarly placed SPRY domain of TRIM5α (Fig. 2) are required for CA-specific binding (72, 82, 94), for CA-specific restriction, and for species-specific differences in restriction activity (2, 37, 45, 55, 61, 63, 68, 70, 78, 79, 81, 83, 94, 95). CypA binding is enhanced by CA multimerization (14, 17), though binding to monomeric CA can be detected at high protein concentrations (29). CA-specific interaction with TRIM5α cannot be visualized by conventional biochemical means. Detection of this interaction has required relatively elaborate binding assays, indicating that CA recognition by the TRIM5α SPRY domain is more complex than a simple protein-protein interaction. In the first report, glutathione S-transferase-human TRIM5α fusion protein was attached to a glutathione matrix and used to pull out the CA provided by detergent-stripped N-MLV virions (72). The second report showed cosedimentation of TRIM5α with HIV-1 cores assembled in vitro from recombinant CA-NC (82). TRIM5α forms trimers which have been modeled to make multiple contacts with the hexameric lattice of CA that constitutes the surface of the retrovirion core (43, 53, 72, 82). One might imagine the SPRY domain of TRIM5α to be pattern recognition receptors of the innate immune system, specialized to recognize the multimeric protein lattice of invading virion cores.

E3 UBIQUITIN LIGASE ACTIVITY AND A TWO-STEP MECHANISM FOR RESTRICTION

TRIM5 deletion mutants that retain the coiled-coil multimerization domain along with the SPRY or CypA CA-binding domains are sufficient for virion core binding (82) but not for restriction activity (42, 61, 81). Deletion analysis indicates that all protein motifs in TRIM5α probably contribute in some way to retroviral restriction activity (42, 61); these motifs include, in particular, the B box, the function of which is a mystery.

The presence of the RING finger in TRIM5 suggests that E3 ubiquitin ligase activity is required for retroviral restriction activity. Ubiquitin is indeed transferred to TRIM5 by an E2 ubiquitin ligase in vitro (91); mutation of RING finger sequences decreases restriction activity (42, 61, 62, 81); and proteasome inhibitors increase viral titer, albeit to a modest extent at best (16, 62, 71). Several leading investigators in the field, however, have argued convincingly that ubiquitin is dispensable for restriction (42, 62, 82). The interpretation of the ubiquitin/RING finger experiments is complicated further by the fact that TRIM5 itself is polyubiquinated and its half-life is increased by proteasome inhibitors (20).

One way to reconcile these findings is to propose that TRIM5 blocks more than one step in the retrovirus life cycle. In addition to blocking viral cDNA accumulation in restrictive cells, TRIM5 poses a block to the nuclear import of viral cDNA (4, 90). Some interventions that suppress HIV-1 restriction in macaque cells, such as arsenic trioxide or proteasome inhibitors, increase HIV-1 cDNA; others, such as cyclosporine, increase the proportion of viral cDNA in the nucleus. Thus, the disruption of cellular ubiquitin by proteasome inhibitors rescues reverse transcription but rescues infectivity to a lesser extent, because there are additional TRIM5-mediated blocks that occur after reverse transcription (90). Additionally, fusions of CypA with the RING finger, B-box, and coiled-coil domains taken from different TRIM family members indicate that TRIM-mediated restriction can occur with blocks apparent either before or after reverse transcription, depending on the TRIM family member (93).

EFFECTS OF CypA ON HIV-1 RESTRICTION BY NONHUMAN PRIMATE ORTHOLOGUES OF TRIM5

The discovery of a fusion protein in owl monkeys consisting of TRIM5 and CypA suggested that the function of the two unfused proteins, as they are found in all other primate species, might be related (Fig. 3). HIV-1 restriction by TRIM5α orthologues from rhesus macaques and African green monkeys is, in fact, dependent on CypA (3, 44). In these species, CypA seems to promote HIV-1 restriction via effects on HIV-1 CA, rather than on TRIM5α, since CypA has no influence on the restriction of TRIM5α-sensitive retroviruses that encode a CA which does not bind CypA.

FIG. 3.

Model to explain the seemingly paradoxical effects of CypA on HIV-1 restriction in different primate species. (A) In owl monkeys, CypA is expressed as a fusion to TRIM5 and, by binding to HIV-1 CA, it acts in cis to restrict HIV-1. (B) In African green monkeys (AGM) or macaques, CypA acts in trans to alter the conformation of HIV-1 CA so that the HIV-1 virion core is more readily recognized by the SPRY domain of TRIM5α. (C) In human cells, CypA acts in trans to alter HIV-1 CA conformation such that the HIV-1 virion core is less readily recognized by an unknown restriction factor. Yellow wedges, cyclophilin A; gray circular structures, CA (p24) monomers.

How might CypA influence HIV-1 CA recognition by restrictive TRIM5α orthologues? CypA catalyzes the cis-trans isomerization of peptidyl-prolyl bonds (25). This activity is a rate-limiting step in the refolding kinetics of model enzymes in vitro, but its significance for CypA function in vivo is not established. HIV-1 CA residues G89 and P90 fit within the active site of CypA (27, 29), and the peptidyl-prolyl bond linking these CA residues undergoes catalytic isomerization (9). In the absence of CypA, the vast majority of the covalent bonds between G89 and P90 are in the trans conformation (30). If TRIM5α recognizes only the trans conformation of CA, then CypA might protect HIV-1 by catalyzing the rate of conversion into the restriction-resistant cis isoform. Data supporting this model have been elusive, perhaps because the changes in CA conformation associated with CypA-catalyzed isomerization are localized to a small area on the surface of the virion core (9).

Genetic approaches to clarify the significance of CypA catalytic activity for HIV-1 CA function have also been inconclusive. For example, it has not been possible to engineer CypA mutants with reduced catalytic activity that do not also reduce substrate binding affinity (23). This result raises the point that CypA binding to HIV-1 CA, in the absence of catalysis, might be sufficient to trigger TRIM5α recognition. Binding to CypA might induce an allosteric change in CA, leading to the exposure of an otherwise hidden TRIM5α-binding surface. Alternatively, the CypA-CA complex might form a new composite binding surface, analogous to the calcineurin-binding surface that is generated when the immunosuppressive drug cyclosporine fills the hydrophobic pocket of CypA (40).

EFFECTS OF CypA ON HIV-1 REPLICATION IN HUMAN CELLS

Several experiments suggested that the effect of CypA on TRIM5α-mediated restriction in human cells is opposite to the effect that CypA has in cells from macaques and African green monkeys. That is, the interaction of CypA with CA seemed to protect HIV-1 from restriction by human TRIM5α. Retroviral restriction is overcome by infection at high multiplicity or by flooding target cells with nonreplicating virus-like particles (5, 18, 34, 86). This phenomenon has been explained by postulating the existence of a saturable factor necessary for restriction. Cross-saturation experiments with human cells suggest that, in the absence of CypA, HIV-1 is particularly sensitive to the same factor that restricts N-MLV (87). Since N-MLV is potently restricted by human TRIM5α (37, 45, 63, 94), it was reasonable to guess that CypA binding to HIV-1 CA shields incoming HIV-1 cores from restriction by TRIM5α. Additional support for this hypothesis was provided by the finding that HIV-1 infection is CypA independent in a human cell clone that was selected for loss of N-MLV restriction activity (69).

After further analysis, it was found that neither the TRIM5α sequence nor TRIM5α expression levels were altered in the clone selected for loss of N-MLV restriction activity (69). And though endogenous human TRIM5α modestly restricts HIV-1, as demonstrated by the knockdown of TRIM5 expression in HeLa cells by RNA interference (44, 76, 84), the magnitude of this antiviral activity is not altered by disruption of the CA-CypA interaction or by elimination of CypA protein (44, 76, 84). Taken together, the results of these experiments suggest that CypA protects HIV-1 from an unknown antiviral activity in human cells (Fig. 3 and 4). One important goal for the future will be to identify this hypothetical factor.

FIG. 4.

Model to explain HIV-1 restriction pathways in human cells. In a wild-type (WT) cell (A), the HIV-1 CA lattice is recognized to a modest extent by TRIM5α (green), which brings the HIV-1 virion core into a restriction complex with an unknown but necessary cofactor (blue) that is saturable and common to both restriction pathways. When CypA is disrupted (B), a second unknown factor (red) is capable of recognizing the HIV-1 CA lattice and bringing the HIV-1 virion core into the restriction complex with the same saturable cofactor (blue). Note that recognition of the HIV-1 CA by TRIM5α (green) is completely independent of CypA (yellow) and of the unknown CypA-regulated factor (red). See the text for a discussion of the experimental data that suggested this model.

Additionally, the increased infectivity that is observed after the loading of target cells with heterologous virus-like particles, and the fact that TRIM5α appears unaltered in the clone selected for loss of N-MLV restriction, suggests that there is another, unknown factor of great importance for understanding the mechanism of retroviral restriction (Fig. 4). This second hypothetical factor appears to be required for the retroviral restriction activity of both TRIM5α and the putative CypA-regulated restriction factor, and it is presumably the factor that is saturated when target cells are loaded with large quantities of restriction-sensitive virion cores.