Independent genesis of chimeric TRIM5-cyclophilin proteins in two primate species

Abstract

The host range of retroviruses is influenced by antiviral proteins such as TRIM5, a restriction factor that recognizes and inactivates incoming retroviral capsids. Remarkably, in Owl monkeys (omk), a cyclophilin A (CypA) cDNA has been transposed into the TRIM5 locus, resulting in the expression of a TRIM5-CypA fusion protein (TRIMCyp) that restricts retroviral infection based on the retroviral capsid-binding specificity of CypA. Here, we report that the seemingly improbable genesis of TRIMCyp has, in fact, occurred twice, and pigtailed macaques (pgt) express an independently generated TRIMCyp protein. The omkTRIMCyp and pgtTRIMCyp proteins restrict infection by several lentiviruses, but their specificities are distinguishable. Surprisingly, pgtTRIMCyp cannot bind to or restrict HIV-1 capsids as a consequence of a point mutation close to the Cyp:capsid-binding interface that was acquired during or after transposition of pgtCypA. However, the same mutation confers on pgtTRIMCyp the ability to restrict FIV in the presence of cyclosporin A, a drug that normally abolishes the interaction between pgtTRIMCyp or omkTRIMCyp and lentiviral capsids. Overall, an intuitively unlikely evolutionary event has, in fact, occurred at least twice in primates and represents a striking example of convergent evolution in divergent species.

The host range of retroviruses is determined in part by antiretroviral restriction factors (1, 2). These factors inhibit particular processes that enable retroviral replication, and one important restriction to HIV-1 infection involves recognition and inactivation of incoming viral capsids (3–7). This restriction is usually mediated by the tripartite motif 5α (TRIM5α) protein (8), whose antiretroviral specificity varies. For instance, rhesus macaque (rh) TRIM5α potently inhibits HIV-1 but not SIV_MAC infection, whereas the human (hu) TRIM5α is inactive against HIV-1 but inhibits other retroviruses. TRIM5α proteins from other primate species have varying spectra of activity (8–13).

TRIM5α is functionally organized into three domains: (i) an N-terminal region (comprising RING and B-box) is not absolutely required for antiviral activity but provides an unknown function that dramatically increases potency (8, 14, 15), (ii) a central coiled-coil domain mediates TRIM5α multimerization that is required for activity (14, 16), and (iii) a C-terminal B30.2/SPRY domain governs antiviral specificity (14, 17–19). The mechanism by which TRIM5α inhibits infection is not well understood but likely involves accelerated capsid disassembly or partial capsid degradation (20, 21). As a consequence of TRIM5α-mediated restriction, nascent viral DNA accumulation is often but not always blocked, in a proteasome-dependent manner, but this latter event is not required for inhibition of infection (22, 23).

TRIM5α-mediated restriction likely involves direct recognition of the incoming capsid by the C-terminal B30.2/SPRY domain (20, 24). TRIM5 sequences encoding this domain exhibit evidence of diversifying selection, consistent with the idea that varying challenges by ancient retroviral epidemics have placed evolutionary pressure on TRIM5 (19, 25). Moreover, a remarkable evolutionary event has occurred at the TRIM5 locus in Owl monkeys (Aotus trivirgatus), a New World monkey species. Specifically, a long interspersed nuclear element (LINE)-mediated retrotransposition event has placed a cyclophilin A (CypA) cDNA within TRIM5, between exons 7 and 8. Thus, a fusion protein, termed TRIMCyp, is expressed, whereby the TRIM5α SPRY domain is replaced by a CypA domain (26, 27). Because HIV-1 and a few other lentiviral capsid proteins bind CypA, omkTRIMCyp inhibits their infection (15, 28, 29). Additionally, because the interaction between CypA and HIV-1 capsids can be inhibited by cyclosporin A (CsA), omkTRIMCyp-mediated restriction is abolished by CsA, providing a useful tool to probe parameters of restriction (30).

Heretofore, the expression of TRIMCyp was thought to be an anomaly, unique to Owl monkeys. Here, however, we report the existence of a similar TRIMCyp protein in pigtailed macaques (pgt, Macaca nemestrina). The CypA domain in pgtTRIMCyp appears to have arisen via a separate LINE-mediated retrotransposition event; has evolved a distinct capsid-binding specificity; and, surprisingly, differs from other CypA proteins, in that it does not recognize the HIV-1 capsid.

Results

Identification of a TRIMCyp Fusion Protein Expressed in Pigtailed Macaque Cells.

In a survey of primate cell line sensitivities to retroviral infection, we found that fibroblasts derived from pigtailed macaques (pgtF) exhibited surprising properties (Fig. 1). In particular, a VSV-G pseudotyped HIV-1/GFP reporter virus was quite infectious in pgtF cells (Fig. 1A), whereas SIV_AGMTan and HIV-2 infection was dramatically enhanced by 5 μM CsA (Fig. 1 B and C). Additionally, FIV infection of pgtF cells was also enhanced by 5 μM CsA (Fig. 1D), although, even under these conditions, FIV infectivity in pgtF cells remained relatively poor.

Fig. 1.

TRIM5-Cyp fusion protein (pgtTRIMCyp) expressed by pigtailed macaques. (A–D) Infection of pgtF cells by varying amounts of GFP-expressing VSV-G-pseudotyped vectors or reporter viruses based on HIV-1(A), SIV_AGMTan (B), HIV-2_ROD (C), or FIV (D), in the presence or absence of 5 μM CsA. (E) PCR products generated using pgtF cDNA template along with primers derived from the 5′ and 3′ ends of the rhTRIM5α coding sequence (lane 1) or from the 5′ end of rhTRIM5α and 3′ end of CypA (lane 2). Arrows indicate bands from which pgtTRIM5η (lane 1) and pgtTRIMCyp (lane 2) cDNAs were cloned. (F) Configuration of the TRIM5 and CypA domains and the linker peptide derived from the CypA mRNA 5′ noncoding sequence in pgtTRIMCyp, as compared with omkTRIMCyp. (G) Protein sequence differences in the Cyp domains of pgtTRIMCyp and omkTRIMCyp, as compared with human and pgt CypA proteins, amino acid postions are numbered relative to the methionine at the N terminus of each Cyp domain or CypA protein.

To understand the basis for this unusual pattern of infection sensitivity, we attempted to clone TRIM5α from pgtF cells. Additionally, because HIV-1, HIV-2, SIV_AGMTan, and FIV share the unusual property of being inhibited, in a CsA reversible manner, by omkTRIMCyp, we entertained the unlikely possibility that perhaps a TRIMCyp fusion protein was expressed in pgtF cells. Using pgtF cDNA and PCR primers based on the ≈1.5-kb rhTRIM5α sequence, it proved difficult to generate PCR products that corresponded to pgtTRIM5α (Fig. 1E). Nonetheless, we did clone sequences from a 1.5-kb product that potentially encoded a protein similar to rhTRIM5α, except for the absence of exon 7 [supporting information (SI) Fig. 6]. This corresponds to a TRIM5 spliced isoform (TRIM5η) that was recently discovered (31).

Remarkably, however, a robust product of ≈1.4 kb was obtained when primers corresponding to the 5′ end of rhTRIM5 and the 3′ end of CypA coding sequences were used (Fig. 1E). Similar results were obtained by using cDNA prepared from pgt lymphocytes from a different individual (data not shown). Of the 19 cDNA clones derived from this PCR product, nine contained TRIM5 coding exons 2–4, fused to a complete CypA-like cDNA, and one contained TRIM5 exons 2–6 and part of exon 8 fused to a short 3′ portion of CypA-like cDNA sequence. Because these cDNAs encoded truncated SPRY domains or CypA sequences that were out of frame with the TRIM5 exons, it was unlikely that they would generate functional proteins, and they were not analyzed further.

However, 9 of the 19 clones had the capacity to encode a TRIMCyp fusion protein. They contained TRIM5 exons 2–6, linked to a full-length CypA-like cDNA (SI Fig. 6). The short intervening sequence, which placed the 3′ CypA sequence in frame with the 5′ TRIM5 sequence, was identical to 5′ noncoding sequence immediately proximal to the start codon in rhesus macaque CypA mRNA (Fig. 1F). The Cyp sequences in pgtTRIMCyp and omkTRIMCyp were more closely related to orthologous CypA sequences than they were to each other (Fig. 1G and data not shown). In fact, the Cyp sequence in pgtTRIMCyp differed from both pgtCypA at only two amino acid positions (Fig. 1G). Moreover, analysis of genomic DNA from two pgt fibroblast cell lines revealed the presence of a CypA-like cDNA, inserted into the 3′ noncoding region of the TRIM5 locus, clearly at a different position within TRIM5 to that which gave rise to TRIMCyp in owl monkeys (SI Fig. 7) (26, 27). The insertion exhibited characteristics of a LINE-mediated retrotransposition event and, notably, the CypA sequence derived from the 3′ noncoding region of the pgt TRIM5 genomic locus was identical to that in the CypA portion of the pgtTRIMCyp cDNA (SI Fig. 7C). Thus, TRIM5 CypA fusion genes and corresponding TRIMCyp proteins have evidently arisen on two occasions via independent retrotransposition events in divergent primate species.

pgtTRIMCyp Does Not Inhibit HIV-1 Infection but Is Active Against Other Lentiviruses.

We generated Chinese hamster ovary (CHO) cell lines expressing C-terminally HA-tagged forms of the two candidate restriction factors recovered from pgtF cDNA. Analysis of several single CHO cell clones expressing varying levels of the pgtTRIM5η protein (examples of three such clones are shown in SI Fig. 8) revealed that it was inactive against HIV-1 SIV_AGMTan, and N-MLV, in contrast to the control rhTRIM5α and huTRIM5α proteins. (SI Fig. 8 A and B). Thus, we could find no evidence for antiretroviral activity in pgtTRIM5η.

We next compared the properties of pgtTRIMCyp with those of omkTRIMCyp using the same approach. Notably, although omkTRIMCyp inhibited HIV-1 infection by >200-fold in a manner reversible by 5 μM CsA, pgtTRIMCyp was virtually inactive against HIV-1 (Fig. 2A). In contrast, both omkTRIMCyp and pgtTRIMCyp restricted SIV_AGMTan infection in a CsA-reversible manner, by >200- and ≈60-fold, respectively (Fig. 2B). Similarly, HIV-2 infection was inhibited, in a CsA reversible manner, >10-fold by both TRIMCyp proteins (Fig. 2C). Interestingly, although feline immunodeficiency virus (FIV) infection was inhibited by both TRIMCyp proteins, omkTRIMCyp-mediated inhibition was almost fully reversed by 5 μM CsA, whereas pgtTRIMCyp was unaffected (Fig. 2D). Importantly, the differences in omkTRIMCyp versus pgtTRIMCyp activity were not due to differences in their expression level in the cell clones (Fig. 2E).

Fig. 2.

Pigtailed macaque TRIMCyp restricts infection by some lentiviruses, but not HIV-1. (A–D) Infection of CHO cells expressing C-terminally HA-tagged omkTRIMCyp or pgtTRIMCyp or by HIV-1(A), SIV_AGMTan (B), HIV-2_ROD (C) or FIV (D), in the presence or absence of 5 μM CsA. (E) Western blot analysis (αHA) of omkTRIMCyp or pgtTRIMCyp expression in the CHO cells used throughout the manuscript. (F and G) Infection of unmanipulated CHO cells, or CHO cells expressing pgtTRIMCyp or omkTRIMCyp by SIV_AGMTan (F) or FIV (G) in the presence of varying (0–20 μM) concentrations of CsA. (H) Same as F and G, except that pgtF cells were used as targets.

These results suggested that pgtTRIMCyp might interact with the FIV capsid in an unconventional way, independent of the CsA-binding site within the Cyp domain. To investigate this in more detail, we challenged the same omkTRIMCyp and pgtTRIMCyp expressing CHO cells (Fig. 2E) with two pgtTRIMCyp-sensitive retroviruses (SIV_AGMTan or FIV) in the presence of varying concentrations of CsA. Notably, SIV_AGMTan restriction by both omkTRIMCyp and pgtTRIMCyp proteins was reversed by low concentrations of CsA (2.5–5 μM) (Fig. 2F), arguing against the notion that the two TRIMCyp proteins differ in their affinity for CsA. FIV restriction by omkTRIMCyp was completely reversed by low concentrations (2.5–5 μM) of CsA (Fig. 2G), but these low concentrations of CsA had little effect on FIV restriction by pgtTRIMCyp, and much higher levels of CsA (20 μM) only partly relieved pgtTRIMCyp-mediated FIV restriction. These results suggest that either (i) the interaction between pgtTRIMCyp and FIV capsid includes additional contacts outside the CsA binding site, or (ii) the FIV capsid binds pgtTRIMCyp with significantly higher affinity than it does omkTRIMCyp.

Presumably because pgtF cells express lower levels of pgtTRIMCyp than do the CHO cell lines, CsA was able to at least partly reverse FIV restriction in pgtF cells (Figs. 1D and 2H). Nevertheless, the levels of CsA required to enhance FIV infection were significantly greater than those required to enhance SIV_AGMTan infection in pgtF cells or the CHO cell lines (compare Fig. 2 H with F).

Variation in the Cyclophilin Portion of TRIMCyp Proteins Governs Restriction Specificity.

To map determinants governing the differential properties of omkTRIMCyp and pgtTRIMCyp, we constructed chimeric proteins in which the CypA domains of the two proteins, which differ at 6-aa positions (Fig. 1G), were precisely exchanged. Specifically, pgtT/omkC encoded the pgt-derived TRIM5 domain and linker peptide, coupled to the omk-derived Cyp domain, whereas the reverse was true in the omkT/pgtC protein (Fig. 3A). Thereafter, we derived CHO cell clones expressing the chimeric proteins at the same level as the two parental TRIMCyp proteins (Fig. 3B). The distinct properties of omkTRMCyp and pgtTRIMCyp were clearly governed by the Cyp domain (Fig. 3 C–F). Specifically, the pgtT/omkC protein, like omkTRIMCyp, restricted HIV-1 infection, whereas the omkT/pgtC protein, like pgtTRIMCyp, did not (Fig. 3C). Similarly, FIV restriction by the pgtT/omkC protein was reversible by 5 μM CsA, whereas restriction by the omkT/pgtC protein was not (Fig. 3F). Both chimeric proteins, like the parental proteins, inhibited SIV_AGMTan and HIV-2 infection in a manner that was reversible by CsA (Fig. 3 D and E).

Fig. 3.

Differential restriction specificity of pgtTRIMCyp and omkTRIMCyp is governed by the Cyp domain. (A) Nomenclature and schematic representation of chimeric TRIMCyp proteins used herein. (B) Western blot analysis (αHA) of intact and chimeric TRIMCyp protein expression in CHO cells. (C–F) Infection of CHO cells expressing intact and chimeric TRIMCyp proteins by HIV-1 (C), SIV_AGMTan (D), HIV-2 (E) or FIV (F), in the presence (white bars) or absence (black bars) of 5 μM CsA.

The Cyp Domains of omkTRIMCyp and pgtTRIMCyp Differ in Their Ability to Bind HIV-1 Gag.

The aforementioned data suggested that the omkTRIMCyp and pgtTRIMCyp proteins differ in their recognition of retroviral capsids, particularly that of HIV-1. Therefore, we next configured yeast two-hybrid assays to directly test interaction between HIV-1 Gag and CypA or TRIMCyp proteins. Notably, HIV-1 Gag interacted strongly with omkTRIMCyp, but not with pgtTRIMCyp (Fig. 4A). When isolated Cyp domains were analyzed by using the same approach, HIV-1 Gag bound, as expected, to huCypA, to pgtCypA, and to the isolated Cyp domain from omkTRIMCyp (which differs from huCypA and pgtCypA at four amino acid positions) (Fig. 4B). However, HIV-1 Gag did not bind to the Cyp domain from pgtTRIMCyp (Fig. 4B), which differs in sequence from pgtCypA at only two amino acid positions (Fig. 1G). The differential ability of the TRIMCyp proteins and Cyp domains to bind HIV-1 Gag was not due to variable expression in yeast (Fig. 4C). Thus, the inability of pgtTRIMCyp to restrict HIV-1 infection appears strictly due to the inability of the Cyp domain of the protein to bind to the HIV-1 capsid, presumably as a consequence of mutations acquired during or following retrotransposition of the CypA cDNA into the pgtTRIM5 locus.

Fig. 4.

Yeast two-hybrid analysis of interactions between HIV-1 Gag and TRIMCyp or CypA proteins. (A) β-Galactosidase activity [expressed in optical density units at 590 nm (OD₅₉₀)] in lysates of Y190 yeast cells expressing GAL4 DNA-binding domains either alone or fused to HIV-1 Gag, along with HA-tagged VP16 activation domains fused to omkTRIMCyp or pgtTRIMCyp. (B) Same as A, except that the VP16 activation domain was fused to the indicated CypA proteins or the Cyp domains from omkTRIMCyp [omkTC(Cyp)] or pgtTRIMCyp [pgtTC(Cyp)]. (C) Western blot analysis (αHA) of VP16-TRIMCyp and VP16-CypA protein expression in yeast lysates.

A Single Amino Acid Mutation Acquired During or After CypA Transposition Alters the Restriction Specificity of pgtTRIMCyp.

The aforementioned data implicated two amino acids in the CypA domain as determinants of differential specificity exhibited by omkTRIMCyp and pgtTRIMCyp. Specifically, residues D66 and R69 are invariant in huCypA, pgtCypA, and omkTRIMCyp proteins, each of which bind HIV-1 capsid and are the only residues in the Cyp domain of pgtTRIMCyp (N66 and H69) that differ from the huCypA and pgtCypA sequence (Fig. 1G).

We generated four mutant TRIMCyp proteins. One was based on omkTRIMCyp in which D66/R69 were mutated to their pgtTRIMCyp counterparts (N66/H69) and termed omkTC(NH), whereas a second, reciprocal pgtTRIMCyp mutant (N66/H69 to D66/R69) was termed pgtTC(DR) (Fig. 5A). Two additional pgtTRIMCyp mutants contained single amino acid substitutions, N66D in pgtTC(N) and H69R in pgtTC(R) (Fig. 5A). Notably, the two amino acid substitution in omkTC(NH) abolished HIV-1 restriction (Fig. 5B). Conversely, the reciprocal substitution, in pgtTC(DR), conferred the ability to potently restrict HIV-1 infection. In fact, a single amino acid substitution (H69R) in pgtTC(R) was sufficient to confer HIV-1 restriction activity on pgtTRIMCyp, whereas the N66D mutation, in pgtTC(D) had no effect (Fig. 5B). Analysis of the ability of these mutant TRIMCyp proteins to bind HIV-1 Gag in yeast two-hybrid assays revealed an essentially perfect correlation between HIV-1 restriction activity and HIV-1 Gag binding (Fig. 5 B and C). Specifically, proteins encoding R69 [omkTRIMCyp, pgtTC(DR) and pgtTC(R)] bound HIV-1 Gag, whereas those encoding H69 [pgtTRIMCyp, omkTC(NH) and pgtTC(D)] did not (Fig. 5C). Thus, a single amino acid difference (H/R69) between pgtTRIMCyp and omkTRIMCyp governs their differential ability to recognize HIV-1 Gag and restrict HIV-1 infection. Although there were minor differences in the level of expression of the mutant proteins in CHO cells (Fig. 5D) and in yeast (Fig. 5E), those that failed to recognize HIV-1 were, if anything, generally expressed at marginally higher levels than the active proteins.

Fig. 5.

A single amino acid mutation in pgtTRIMCyp proteins alters recognition of lentiviral capsids. (A) Nomenclature and schematic representation of the mutant TRIMCyp proteins used herein, position numbering is relative to the methionine at the N terminus of the Cyp domain of each protein. (B) Infection of CHO cell clones expressing wild-type or mutant TRIMCyp proteins by HIV-1, in the presence (white bars) or absence (black bars) of 5 μM CsA. (C) β-Galactosidase activity (expressed in optical density units at 590 nm [OD₅₉₀)] in lysates of Y190 yeast cells expressing GAL4 DNA-binding domains either alone or fused to HIV-1 Gag, along with HA-tagged VP16 activation domains fused to wild-type or mutant TRIMCyp proteins. (D) Western analysis (αHA) of wild-type and mutant TRIMCyp protein expression in CHO cells. (E) Western analysis (αHA) of wild-type and mutant VP16-TRIMCyp fusion protein expression in yeast cells. (F–H) Infection of CHO cells expressing C-terminally HA-tagged wild-type or mutant TRIMCyp proteins by SIV_AGMTan (F), HIV-2_ROD (G), or FIV (H), in the presence (white bars) or absence (black bars) of 5 μM CsA. (I) Structure of an HIV-1 capsid monomer in complex with huCypA (identical in sequence to pgtCypA), indicating the amino acid resides (white) that differ in pgtTRIMCyp and influence capsid recognition. Also indicated is an HIV -1 capsid residue (P85, yellow) that is closely opposed to the R69 residue in CypA.

Analysis of the mutant proteins for inhibition of the other TRIMCyp-sensitive lentiviruses revealed that the omkTC(NH) protein was inactive against SIV_AGMTan, despite the fact that both parental proteins and each of the other mutants were active against the same virus (Fig. 5F). The same omkTC(NH) protein was active against both HIV-2 and FIV (Fig. 5 G and H), indicating that it does have intrinsic antiretroviral activity, but likely fails to recognize the SIV_AGMTan capsid. In fact, all of the mutant proteins were, like the wild-type parental proteins, active against HIV-2 and FIV (Fig. 5 G and H). Notably, however, the same amino acid difference (H/R69) that governed HIV-1 Gag binding and restriction also determined whether FIV restriction was CsA-sensitive. Specifically, restriction by proteins that encoded H69 [pgtTRIMCyp, omkTC(NH), and pgtTC(D)] could not be reversed by 5 μM CsA, whereas restriction by proteins that encoded R69 [omkTRIMCyp, pgtTC(DR) and pgtTC(R)] was reversed by this treatment (Fig. 5H). Thus, CypA residues that are mutated in pgtTRIMCyp relative to pgtCypA, particularly R69H, alter interaction with retroviral capsids. Consistent with this notion, inspection of the published crystal structure of the huCypA:HIV capsid complex (32) revealed that D66 and R69 are positioned close to the CypA:capsid interaction interface (Fig. 5I). In particular, CypA residue R69 has the potential to directly influence the contacts between CypA and the base of the CypA-binding loop, at around HIV-1 capsid residue P85, a previously demonstrated determinant of HIV-1 capsid:CypA-binding (33). Alternatively, R/H69 mutations could potentially affect local Cyp structure and more drastically change the configuration of the capsid-binding site in pgtTRIMCyp.

Discussion

Previously, the generation of a chimeric gene encoding a TRIM5-CypA fusion protein was thought to be an improbable event that occurred uniquely in owl monkeys (26, 27). However, essentially the same protein has been generated in pigtailed macaques via an independent retrotransposition event. Several lentiviruses are known to encode capsids that are sensitive to omkTRIMCyp (15, 28, 29), presumably because they bind to CypA. Although the purpose of lentiviral capsid:cyclophilin interactions remains somewhat mysterious and may originally have evolved as a viral defense against restriction factors (34), CypA binding by retroviral capsids likely provides the evolutionary impetus for the occurrence of new restriction factors that exploit this interaction.

That the same fusion protein has arisen twice, independently, during primate evolution seems intuitively astonishing. However, several factors may contribute to two independent observations of a seemingly unlikely evolutionary event. One factor is the very potent selection pressure that could potentially be placed on the TRIM5 gene. Because TRIM5 can provide a very high degree of resistance to lethal, and common, diseases that are sometimes characteristic of lentivirus infections, then even a very rarely generated TRIM5 variant could provide a sufficiently strong evolutionary advantage that it rapidly predominates within a population. Another factor is the apparently robust propensity of Cyp domains to retrotranspose within mammalian genomes; human and mouse genomes contain numerous CypA-derived pseudogenes, and there are several examples of larger proteins that contain Cyp domains inserted within them (35, 36).

Even so, it is surprising that cyclophilin fusion proteins based specifically on TRIM5 have arisen twice. There are dozens of TRIM genes in the human genome and fusion of a Cyp domain to several TRIM motifs, Fv1 or other multimerizing proteins can generate active antiretroviral factors, albeit with decreased potency as compared with the naturally occurring omkTRIMCyp (28, 37–39). That TRIM5 has twice served as a target for the insertion of CypA and the genesis of a restriction factor suggests it is uniquely suited to this function. Alternatively, it is conceivable that TRIM5 is a “hotspot” for LINE-mediated insertions.

These considerations beg the question of whether TRIMCyp fusion proteins (based on TRIM5 or any other TRIM gene) have arisen in other species. Previously, CsA-dependent sensitivity to HIV-1 would be expected to be reasonable marker to uncover such activities. However, such assays are clearly unreliable indicators of TRIMCyp protein activity, and divergence from the original CypA sequence in transposed Cyp sequences can apparently lead to the evolution of novel capsid-binding specificities. Indeed, this seems to have occurred in pgtTRIMCyp. Specifically, a single amino acid change (R69H) in pgtTRIMCyp as compared with pgtCypA or omkTRIMCyp results in the loss of interaction with the HIV-1 capsid, but a “gain” of interaction with the FIV capsid, insofar as it became more difficult to abolish FIV restriction with CsA. Either the configuration of the FIV capsid:Cyp interaction is changed by the R69H substitution, such that it is less dependent on CypA residues that form the CsA-binding site, or the FIV capsid has significantly higher affinity for TRIMCyp proteins that encode H69 rather than R69. Unfortunately, we were not able to configure FIV Gag:TRIMCyp-binding assays to investigate this, and the potentially close proximity of R/H69 to the capsid:Cyp interface is consistent with either possibility. Nonetheless, these observations invite the speculation that some retroviral infection, perhaps by an FIV-like virus, led to the selection of H69 in pgtTRIMCyp, because H69 confers some benefit in terms of restriction. Remarkably, precisely the same R69H substitution has arisen in TRIMCyp variants found in some Aotus species (40). Thus, it is even conceivable that a single ancient retroviral pandemic led to the selection and convergent evolution of the TRIMCyp protein in both Old and New World monkeys.

Overall, the de novo generation of a gene expressing a TRIMCyp fusion protein is made even more surprising by the fact that it has occurred on at least two separate occasions. Thus, the potent evolutionary pressure placed on organisms by infectious agents can illuminate understanding of how new gene products and protein-binding specificities arise, and these findings illustrate how apparently remarkable evolutionary events can be less improbable than they might intuitively appear.

Materials and Methods

Molecular Construction.

The origin and derivation of plasmid vectors expressing the various TRIM5, TRIMCyp, and CypA proteins are described in SI Text. Plasmid vectors for generating VSV-G pseudotyped, GFP-expressing HIV-1, HIV-2, SIV_AGMTan, and FIV reporter virus stocks have been described (4, 7, 41, 42).

Cells and Infection Assays.

Fibroblasts derived from a pigtailed macaque (pgtF cells) were obtained from Coriell cell repositories. A derivative of hamster CHOK1 cells was used to generate single cell clones expressing the TRIMCyp proteins. This was done by transducing cells with LNCX2-based HA-tagged TRIM protein-expressing retroviral vector, as described (9, 14), and cell clones were derived by limiting dilution. In all cases, multiple cells clones (5–10), expressing similar levels of each TRIMCyp protein underwent a preliminary analysis, and representative clones were used in more detailed experiments. For infection assays, cells (2 or 4 × 10⁴ per well in 48-well trays) were inoculated with a single dose of each VSV-G pseudotyped GFP-expressing reporter virus, generated by transient transfection in 293T cells, that was sufficient to infect 10–30% of the unmanipulated CHO cells. Alternatively, in some experiments (Fig. 1), the dose of virus was varied. CsA was added at the time of infection, and both virus and drug were removed after 24 h. The number of infected cells was enumerated by FACS, 48 h after infection.

Yeast Two-Hybrid Analyses.

Y190 yeast cells were cotransformed with plasmids expressing the GAL4 DNA-binding domain either alone or fused to HIV-1 Gag along with plasmids expressing a VP16 activation domain fused to an HA epitope tag and a TRIMCyp or CypA protein. Transformants were selected on appropriate media, and β-galactosidase activity therein was determined as described (14).

Western Blot Analyses.

Whole-cell lysates of CHO cells expressing C-terminally HA-tagged TRIM proteins or extracts of yeast cells expressing VP16-HA- fusion proteins were separated on polyacrylamide gels and transferred to nitrocellulose membranes. Thereafter, the membranes were sequentially probed with an αHA monoclonal antibody (Covance) and a mouse IgG peroxidase conjugate.