Unique Spectrum of Activity of Prosimian TRIM5α against Exogenous and Endogenous Retroviruses^,

Abstract

Lentiviruses, the genus of retrovirus that includes HIV-1, rarely endogenize. Some lemurs uniquely possess an endogenous lentivirus called PSIV (“prosimian immunodeficiency virus”). Thus, lemurs provide the opportunity to study the activity of host defense factors, such as TRIM5α, in the setting of germ line invasion. We characterized the activities of TRIM5α proteins from two distant lemurs against exogenous retroviruses and a chimeric PSIV. TRIM5α from gray mouse lemur, which carries PSIV in its genome, exhibited the narrowest restriction activity. One allelic variant of gray mouse lemur TRIM5α restricted only N-tropic murine leukemia virus (N-MLV), while a second variant restricted N-MLV and, uniquely, B-tropic MLV (B-MLV); both variants poorly blocked PSIV. In contrast, TRIM5α from ring-tailed lemur, which does not contain PSIV in its genome, revealed one of the broadest antiviral activities reported to date against lentiviruses, including PSIV. Investigation into the antiviral specificity of ring-tailed lemur TRIM5α demonstrated a major contribution of a 32-amino-acid expansion in variable region 2 (v2) of the B30.2/SPRY domain to the breadth of restriction. Data on lemur TRIM5α and the prediction of ancestral simian sequences hint at an evolutionary scenario where antiretroviral specificity is prominently defined by the lineage-specific expansion of the variable loops of B30.2/SPRY.

INTRODUCTION

TRIM5α is a host restriction factor providing a species-specific barrier against retroviruses. TRIM5α belongs to the tripartite motif (TRIM) family, and its TRIM domain, also referred to as RBCC, consists of an N-terminal RING, a B-box 2, and a coiled-coil domain (45). The RING domain is important for restriction potency, the coiled coil promotes protein oligomerization, and B-box 2 mediates the cooperative association of TRIM5α dimers with the retroviral capsid (6, 18, 33, 40). In addition, TRIM5α comprises a C-terminal B30.2/SPRY domain responsible for the specific recognition of the capsid (57, 64). Upon viral entry, TRIM5α intercepts incoming cores and induces their premature disassembly, impairing the next steps of viral infection (56). TRIM5α-mediated restriction targets lentiviruses, gammaretroviruses, spumaviruses, and betaretroviruses, indicating a role for TRIM5α in innate immunity against retroviral infections (7, 15, 21, 46, 55, 62, 63, 66). TRIM5α is transcriptionally regulated by interferon and contributes to the cellular antiviral state (1, 3, 4, 47).

TRIM5 genes have evolved under strong positive selection since the separation of Old World and New World monkeys (28, 35, 37, 38, 49, 50, 53). The amino acids under selection are located mostly in the coiled-coil domain and in the four variable regions (v1 to v4) of the capsid-interacting B30.2/SPRY domain. Exposure to ancient retroviruses is likely to have been the driving force in TRIM5α evolution, and the interspecies variation of TRIM5α sequences underlies the species-specific patterns of restriction. TRIM5α from Old World monkeys potently restrict human immunodeficiency virus type 1 (HIV-1) (54, 55). Similarly, New World monkey TRIM5α variants broadly inhibit simian immunodeficiency viruses (SIVs) (54). In contrast, human TRIM5α is weakly active against primate lentiviruses while efficiently blocking N-MLV (15, 21, 41, 63). Whereas most primate TRIM5α proteins restrict N-tropic murine leukemia virus (N-MLV), no natural TRIM5α variant has been found to block B-tropic MLV (B-MLV) so far (5, 30).

We have previously reconstructed ancestral TRIM5α sequences to describe the very dynamic process of the evolution of antiviral specificity (13). This analysis allowed an estimation of the activity of a 25-million-year-old TRIM5α protein based on New World and Old World primate sequences. To further characterize TRIM5α evolution in primates, we investigated prosimian TRIM5α. Lemurs are the only living nonhuman primates in Madagascar. They have colonized the island and speciated separately from other primates 60 to 50 million years ago (16). Recent studies showed that two Malagasy lemur species, i.e., fat-tailed dwarf lemur and gray mouse lemur, possess nearly intact copies of a 4-million-year-old endogenous lentivirus referred as PSIV (prosimian immunodeficiency virus) (11, 12). Thus, Malagasy lemurs were in contact with an ancient exogenous lentivirus several million years ago, and unknown SIV strains might still circulate among those species (52). Therefore, we hypothesized that TRIM5 may have diversified in Malagasy lemurs to counteract a unique set of ancient and modern lentiviruses. For this reason, we cloned TRIM5α from two lemur species, i.e., the gray mouse lemur, which harbors an endogenous PSIV, and the ring-tailed lemur, which does not carry PSIV. We characterized the spectrum of TRIM5α activity against modern retroviruses and against a chimeric lentivirus bearing an ancient PSIV capsid (14).

MATERIALS AND METHODS

Prosimian material.

Total RNA was extracted from ring-tailed lemur (Lemur catta) untransformed skin fibroblast AG07099 cells (Coriell Institute for Medical Research) and from the liver of a gray mouse lemur (Microcebus murinus), euthanized for veterinary reasons (Institute of Zoology TiHo), using a Qiagen RNeasy Plus minikit. cDNA was prepared by reverse transcription using Expand reverse transcriptase (Roche). Putative TRIM5 genes for tarsier (Tarsius syrichta), bushbaby (Otolemur garnettii), gray mouse lemur (Microcebus murinus), and tree shrew (Tupaia belangeri) were identified with a BLAST-Like Alignment Tool (BLAT) search on ongoing genome sequencing projects (http://www.ensembl.org/info/about/species.html). The predicted gray mouse lemur TRIM5α (gmlTRIM5α) sequence was aligned with known primate TRIM5α proteins and used to design specific primers (5′-GAAACCTCATCAGCGAGGACAG-3′ and 5′-GCGCTGCACAGAAGTGGC-3′) in conserved regions of the 5′ and 3′ untranslated regions (UTRs) for PCR amplification. Sequence similarity facilitated the amplification and sequencing of TRIM5α from ring-tailed lemur skin fibroblast cell lines (rtlTRIM5α). Multiple independent PCRs (n = 5) were performed on cDNA with Pfx Supermix (Life Technologies) and sequenced to verify the absence of sequence errors. The PCR products were cloned into the pCR-BluntII-TOPO vector (Life Technologies), and several individual clones (n = 5 for rtlTRIM5α; n = 10 for gmlTRIM5α) were sequenced to identify allelic variants. Genomic DNA was prepared from the liver of the gray mouse lemur by using a Qiagen genomic DNA kit. Primers 5′-AAGGCGCTGGCCACCAAAGG-3′ and 5′-CCGATTGTCTCCCTACCTTCAGAG-3′ were used for PCR amplification using Pfx Supermix (Life Technologies) and subsequent sequencing of the PSIV capsid.

Sequence analysis, ancestral reconstruction, and structural modeling.

For phylogenetic analysis, TRIM coding sequences from primates and mammals (see Table S1 in the supplemental material) were aligned with ClustalW and curated manually, and the alignment was used to build a neighbor-joining tree (Jukes-Cantor model) using Geneious software (v5.1; A. J. Drummond et al., Biomatters Ltd.). Bootstrap support (1,000 iterations) was shown on the tree branches. Maximum likelihood and maximum parsimony trees were also constructed (using PHYML and MEGA4.0, respectively). For the reconstruction of the TRIM5α ancestral sequence, nucleotide sequences from all primates except the tarsier (Table S1) were aligned based on the protein alignment using the EMBOSS program tranalign. Variable regions in the B30.2/SPRY domain, less informative for ancestral reconstruction, were clipped based on a structural three-dimensional (3D) model. The conserved structural backbone of the protein was kept, while the variable loops (v1, v2, and v3) were removed from the alignment. Ancestral reconstruction was performed by using the codeml program in the PAML package.

For structural model building, the B30.2/SPRY domains of human TRIM5α (Uniprot accession number Q05CU3) (60), rtlTRIM5α, and gml₁TRIM5α were defined as target sequences for homology modeling. Experimental crystal structures of the B30.2/SPRY domains of human TRIM21 (Protein Data Bank [PDB] [2] accession number 2IWG [17]) and pyrin (PDB accession number 2WL1 [61]) as well as the B30.2/SPRY domain of murine TRIM21 (PDB accession numbers 2VOK and 2VOL [23]) were used as structural templates. The multiple-sequence alignment of the target and template sequences was realized by using the MUSCLE algorithm (8, 9). The sequence identity between the targeted proteins and the structural templates ranged from 34 to 46%. Based on the target-template sequence alignments, model structures of the human, rtlTRIM5α, and gml₁TRIM5α B30.2/SPRY domains were built by the satisfaction of spatial restraints using the MODELLER program, v9.1 (48). One thousand models were built for each protein, and the final model was selected based on the modeler objective function. The structural and sequence alignments were performed by using the MatchMaker command of the UCSF Chimera program (42). The weights applied to the residue similarity and secondary structure during the computation of the alignments were 0.7 and 0.3, respectively. The PSIPRED server (http://bioinf.cs.ucl.ac.uk/psipred/) was used for the analysis of the v1, v2, v3, and v4 regions, for which no template was available. PSIPRED is known to provide excellent predictions of secondary structure elements, with a success rate close to 80% (32).

Cell lines and vectors.

Human embryonic kidney 293T cells and Crandell feline kidney (CrFK) cells were maintained in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal calf serum, 2 mM glutamine, and antibiotics (100 U/ml penicillin and 100 μg/ml streptomycin).

pLPCX vectors encoding human, rhesus macaque, African green monkey, and tamarin TRIM5α proteins were previously described (13). Expressed TRIM5α proteins contain an in-frame hemagglutinin (HA) tag at the C terminus. HA-tagged coding sequences of rtlTRIM5α, gml₁TRIM5α, and gml₂TRIM5α were cloned into the EcoRI and ClaI restriction sites of the pLPCX vector. The gml₁TRIM5α-L418H and gml₂TRIM5α-H418L mutants were obtained by exchanging BsmI-ClaI fragments, corresponding to the SPRY domain, between pLPCX-gml₁TRIM5α and pLPCX-gml₂TRIM5α. BsmI-ClaI fragments corresponding to chimeric SPRY domains were generated by multiple PCRs and inserted into BsmI and ClaI sites in pLPCX-gml₁TRIM5α and pLPCX-rtlTRIM5α to replace the original SPRY domains. BsmI-ClaI fragments bearing deletions and insertions of the 32-residue sequence were generated by PCR and cloned into BsmI and ClaI sites in pLPCX-gml₁TRIM5α and pLPCX-rtlTRIM5α. All inserts, mutants, and chimeras were confirmed by sequencing.

The vesicular stomatitis virus glycoprotein (VSV-G) plasmid is pMD.G (a gift from D. Trono). The following genomic and packaging plasmids were used for the production of green fluorescent protein (GFP) reporter retroviruses: genomic plasmid pSIN.cPPT.EF1.GFP.WPRE, packaging plasmid pCMVR8.92, and pRSV-Rev plasmids for HIV-1-based viruses (a gift from D. Trono); the packaging HIV-2-pack and genomic HIV-2-GFP plasmids (a gift from A. Lever); the SIVmac vectors pSIV-GFP and pSIV3+ (a gift from F. L. Cosset); the genomic vector SIVagm tan-GFP (a gift from J. Luban) (54); the feline immunodeficiency virus (FIV) packaging vector pFP93 and genomic vector pGINSIN encoding enhanced GFP (a gift from E. Poeschla) (47); and the packaging vectors encoding N-MLV, B-MLV, and Moloney murine leukemia virus (Mo-MLV) Gag-Pol (pCIG3-N, pCIG3-B, and pCIGPB, respectively) and genomic MLV plasmid pCNCG (gifts from D. Trono). Equine infectious anemia virus (EIAV) Gag-Pol plasmid pONY3.1 and genomic EIAV plasmid pONY8.4ZCG harboring the lacZ and GFP reporter genes were provided by K. Mitrophanous. Chimeric Gag-Pol plasmid pEIAV-PSIV was described previously (14).

Stable lines.

293T cells were cotransfected with each pLPCX-TRIM5α plasmid, the packaging MLV Gag-Pol plasmid, and the VSV-G plasmid using polyethylenimine (Polysciences). As controls, we also used pLPCX and pLPCX-eGFP. Supernatants were concentrated and used to transduce 10⁶ CrFK cells in the presence of 10 μg of Polybrene/ml. After 72 h, CrFK cells were selected in the presence of 5 μg of puromycin/ml for at least 10 days before testing. All clones from stable cell lines were confirmed by sequencing.

Viral production.

The GFP reporter viruses were produced by transfecting 10-cm dishes of 293T cells with 20 μg of total DNA using polyethylenimine (PolyScience). Two-part SIVagm vectors were produced by the transfection of genomic and VSV-G plasmids at a ratio of 7:1. Three-part vectors (HIV-2, SIVmac, FIV, EIAV, N-MLV, B-MLV, and Mo-MLV) were produced by the transfection of genomic, Gag-Pol, and VSV-G plasmids at a ratio of 5:4:1. Four-part HIV-1 vectors were produced by the transfection of genomic, packaging, Rev, and VSV-G plasmids at a ratio of 5:3:1:1. The transfection mixture was replaced with fresh medium 24 h later, and viral supernatants were harvested at 48 h posttransfection and filtered (0.45 μm).

EIAV and EIAV-PSIV viral stocks were made by the transfection of 293T cells with pONY8.4ZCG, VSV-G, and either pONY3.1 or pEIAV-PSIV plasmids at a ratio of 5:1:4 using polyethylenimine. Sodium butyrate (10 mM) was added at 8 h posttransfection, and the transfection mixture was replaced by fresh medium containing 10 mM sodium butyrate at 24 h posttransfection. Viral supernatants were harvested 48 h after transfection and concentrated by using Centricon-100 filters (Millipore).

Infectivity assays.

Infections with GFP reporter viruses were performed by adding virus-containing supernatants to 10⁵ CrFK cells in 24-well plates. Two days after infection, GFP-positive target cells were quantified by fluorescence-activated cell sorter (FACS) analysis. Due to cytotoxicity, SIVagm supernatants were spinoculated onto CrFK cells in the presence of Polybrene (10 μg/ml), and FACS analysis was performed at 36 h postinfection.

EIAV and EIAV-PSIV stocks were titrated by the infection of 2 × 10⁵ CrFK cells in a 12-well plate by spinoculation for 2 h at 1,500 × g in the presence of 10 μg/ml of Polybrene. Cells were fixed and stained at 3 days postinfection. Briefly, cells were fixed in 0.5% formaldehyde–0.05% glutaraldehyde for 5 min at room temperature, followed by overnight staining with an X-gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) solution [5 mM potassium hexacyanoferrate(II), 5 mM potassium hexacyanoferrate(III), 2 mM MgCl₂, 0.5 mg/ml X-gal]. Titers were determined by counting lacZ-positive loci, in wells containing 20 to 200 loci, for two successive dilutions.

Western blots.

Cells were lysed with radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 1% NP-40, 0.1% SDS, 0.5% sodium deoxycholate) supplemented with Complete Mini protease inhibitor cocktail (Roche) and centrifuged at 12,000 × g for 10 min. The total protein concentration was measured with a BCA protein assay kit (Pierce) for sample loading normalization. The equivalent of 30 μg of proteins was subjected to SDS-PAGE and immunoblotted with rat monoclonal anti-HA antibody (Roche) and mouse monoclonal anti-tubulin (Sigma).

Nucleotide sequence accession numbers.

The sequences of rtlTRIM5α, gml₁TRIM5α, and gml₂TRIM5α have been deposited in GenBank under accession numbers HQ413344, HQ413345, and HQ413346, respectively.

RESULTS

Identification of prosimian TRIM5α sequences.

A putative TRIM5α sequence was identified by using the BLAST-Like Alignment Tool (BLAT) on the low-coverage genome of the gray mouse lemur. On the basis of the sequence information, two allelic variants, gml₁TRIM5α and gml₂TRIM5α, were cloned and sequenced from a liver sample of one unique animal. The variant gml₁ is identical to the sequence retrieved by BLAT and was used for the various constructions. The variant gml₂ differed by 4 amino acids (see Fig. S1 in the supplemental material) and was the subject of the specific analysis of the sole difference in the B30.2/SPRY domain. Sequence similarity facilitated the amplification and sequencing of TRIM5α (rtlTRIM5α) from ring-tailed lemur skin fibroblast cells. Additional TRIM5α sequences were identified by using BLAT on the low-coverage genomes of the tarsier, bushbaby, and tree shrew using gray mouse lemur and ring-tailed lemur TRIM5α as bait sequences. The alignment of coding sequences was used to build a phylogenetic tree encompassing known TRIM sequences in primates and other mammals (Fig. 1A). Neighbor-joining trees, as well as maximum likelihood and maximum parsimony trees, served to confirm that the amplified sequences corresponded to TRIM5α. Lemur TRIM5α and the predicted bushbaby TRIM5α formed a distinct prosimian branch. The complete alignment of TRIM5α from 13 representative species is shown in Fig. S1 in the supplemental material. The RBCC domains of lemur and simian TRIM5α sequences were highly similar. In contrast, the B30.2/SPRY domains diverged in the variable regions (Fig. 1B). rtlTRIM5α comprised an extended v2 region of 51 residues that contrasts with the 17- to 19-residue-long v2 that characterizes primate orthologues. Ring-tailed lemur v2 was constituted from repetitive motifs (Fig. 1B, boxes) as described previously for the v1 and v3 regions of African green monkey TRIM5α (agmTRIM5α) and spider monkey TRIM5α, respectively (53). One of the four amino acids differentiating gml₁TRIM5α and gml₂TRIM5α was localized in the v3 region (Fig. 1B, arrow).

Fig. 1.

Identification of prosimian TRIM5α. (A) Phylogenetic tree showing the relationship between the studied TRIM5α proteins and representative TRIM5 and TRIM-related proteins in selected primates and mammals. The alignment of coding sequences was generated with ClustalW and was used to build a neighbor-joining tree. Bootstrap values (1,000 replicates) are shown on the branches. The scale bar corresponds to 0.05 nucleotide substitutions per site. TRIM5α sequences used in this study are in boldface type. Predicted sequences are in italic type. (B) Alignment of the B30.2/SPRY domains of the TRIM5α proteins expressed in recombinant cell lines in this study: human, rhesus macaque, African green monkey (agm), cotton-top tamarin, ring-tailed lemur, and gray mouse lemur (alleles 1 and 2). Black, identity; gray, similarity. Variable regions (v1 to v4) are indicated according to previously reported descriptions (36, 53). Repeated motifs in v2 of ring-tailed lemur TRIM5α are delimited in boxes. The arrow designates amino acid 418 differing between gml₁TRIM5α and gml₂TRIM5α. Stars delineate the regions exchanged in the chimeric TRIM5α proteins, and black triangles mark the 32-residue-long insertion in v2 of ring-tailed lemur TRIM5α.

Differential retroviral restriction by two lemur TRIM5α proteins.

As inter- and intraspecies polymorphisms in TRIM5α lead to distinct antiviral activities (15, 35, 63), we assessed the restriction potential of HA-tagged rtlTRIM5α, gml₁TRIM5α, and gml₂TRIM5α in comparison with known TRIM5α orthologues. Recombinant proteins were stably expressed in CrFK cells, characterized by the absence of functional endogenous TRIM5α (31). Western blot analysis showed that rtlTRIM5α was efficiently expressed in CrFK cells, whereas gml₁TRIM5α and gml₂TRIM5α were weakly expressed (Fig. 2A). Infections were performed by titrating VSV-G-pseudotyped retroviruses encoding GFP on TRIM5α-expressing cells and control cells (Fig. 2B). The restriction patterns of human, rhesus macaque, African green monkey, and tamarin TRIM5α proteins corresponded to those reported previously. rtlTRIM5α restricted FIV more than 10-fold; HIV-1, HIV-2, SIVmac, SIVagm, and EIAV about 100-fold; and N-MLV more than 1,000-fold compared to the controls. In contrast, gml₁TRIM5α restricted only N-MLV, and strikingly, gml₂TRIM5α efficiently restricted N-MLV and B-MLV, the first description of a natural TRIM5α variant being able to restrict the latter retrovirus (30). To increase gml₁TRIM5α and gml₂TRIM5α expression levels, we transduced CrFK cells multiple times, but expression levels of these proteins remained significantly lower than those of other orthologues (data not shown). However, as will be illustrated later in this paper, the low level of expression of gmlTRIM5α did not limit its ability to restrict B-MLV and N-MLV. These results provide a unique pattern of the antiviral restriction of TRIM5α from the two lemur species: a broad inhibition of lentiviruses and N-MLV by rtlTRIM5α and a narrow spectrum accompanied by unique activity against B-MLV by gmlTRIM5α.

Fig. 2.

Restriction activity of lemur TRIM5α. (A) Expression analysis of HA-tagged TRIM5α in stable CrFK cell lines expressing TRIM5α from human (hu), rhesus macaque (rh), African green monkey (agm), cotton-top tamarin (tam), ring-tailed lemur (rtl), and gray mouse lemur (gml₁ and gml₂ alleles) by Western blotting using antibodies directed against the HA tag or tubulin. Control cells contain an empty pLPCX vector. (B) Infections of TRIM5α-expressing CrFK cell lines with serial dilutions of GFP reporter retroviruses pseudotyped with VSV-G. The percentage of infected cells was measured by flow cytometry 48 h following the infection, except for SIVagm infections, which, due to viral cytotoxicity, were performed by spinoculation, allowing the quantification of GFP-positive cells at 36 h postinfection. These results are representative of results from two independent experiments.

A single amino acid, H418, is the major determinant of gmlTRIM5α-mediated B-MLV restriction.

gmlTRIM5α variants differ in 4 amino acids: position 20 in the RING domain, positions 86 and 90 between the RING and B-box 2 domains, and position 418 in the v3 region of B30.2/SPRY (see Fig. S1 in the supplemental material). We tested whether amino acid 418 was responsible for the differential activity against B-MLV by generating mutants at this position. Wild-type sequences and the gml₁TRIM5α-L418H and gml₂TRIM5α-H418L mutants were similarly expressed in CrFK cells (Fig. 3). The gml₁TRIM5α-L418H mutant restricted B-MLV similarly to gml₂TRIM5α. In contrast, gml₂TRIM5α-H418L lost B-MLV restriction (Fig. 3A, left). Both mutants were functional, as they restricted N-MLV as efficiently as their wild-type TRIM5α counterparts (Fig. 3A, right). Thus, the swap of the amino acid 418 between the two grey mouse lemur alleles reversed the restriction phenotype against B-MLV without affecting N-MLV restriction, demonstrating that a single amino acid in the v3 region of gml₂TRIM5α was a major determinant for B-MLV restriction.

Fig. 3.

A single amino acid in the v3 region of gray mouse lemur TRIM5α is responsible for the restriction of B-MLV. (A) Infections of CrFK cells expressing gray mouse lemur TRIM5α alleles (gml₁ and gml₂) or their reciprocal mutants with GFP reporter B-MLV and N-MLV. Control cells contain an empty pLPCX vector. Infections were performed in duplicates, and results are representative of data from two independent experiments. Error bars represent the range values of duplicates. (B) Expression of HA-tagged TRIM5α in the corresponding CrFK stable cell lines, assessed by Western blotting using anti-HA and anti-tubulin antibodies.

v1/v2 regions determine retroviral restriction mediated by ring-tailed lemur TRIM5α.

Next, we characterized the domains responsible for the differences between restriction by rtlTRIM5α, which broadly inhibits lentiviruses, and restriction by gmlTRIM5α variants, which display a narrow spectrum of activity. The variable regions in the B30.2/SPRY domain have been shown to determine the spectrum of activity of TRIM5α (36). Therefore, we constructed chimeric TRIM5α by exchanging regions of the B30.2/SPRY domain between rtlTRIM5α and gml₁TRIM5α and generated CrFK stable cell lines expressing those constructs (Fig. 4A, left, and B). Cells expressing chimeric and wild-type TRIM5α were infected with HIV-1, N-MLV, and B-MLV GFP viruses (Fig. 4A). Infections with HIV-1 showed that gml₁TRIM5α and GR1 were similarly inactive, while GR2, GR3, and GR4 chimeras restricted HIV-1, suggesting that v1-v2 or v2 carries determinants of the rtlTRIM5α-mediated restriction of HIV-1. The GR4 chimera, comprising the whole rtlSPRY domain, was less efficient than wild-type rtlTRIM5α, possibly due to the contribution of the RING, B-box 2, and coiled-coil domains (18, 26, 29, 40). RG1, RG2, and RG3 were inactive against HIV-1, N-MLV, and B-MLV (Fig. 4A) and against an expanded panel of viruses (results not shown). The general lack of efficacy against multiple viruses may reflect the reassortment of the variable regions of B30.2/SPRY or may result from an intrinsic defect of the chimeric constructs. Of note, RG1 to RG3 were efficiently expressed in CrFK cells (Fig. 4B), and their correct sequence was confirmed by sequencing. RG4 TRIM5α bearing gml₁SPRY restricted HIV-1, N-MLV, and B-MLV. Previous studies reported an unexpected gain of activity by some TRIM5α chimeras (36, 57, 62). These data highlight, by using the two lemur TRIM5α proteins as representatives of broad or narrow restriction capacities, the complex participation of multiple domains in TRIM5α in the inhibitory pattern.

Fig. 4.

Role of the variable regions of lemur TRIM5α proteins in restriction activity. (A) Chimeric TRIM5α constructs and infection with selected GFP reporter retroviruses of the various recombinant CrFK stable cell lines. GR, gray mouse lemur–ring-tailed lemur chimeras; RG, ring-tailed lemur–gray mouse lemur chimeras. Control cells contain an empty pLPCX vector. Infections were performed in duplicates, and data are representative of data from two independent experiments. Error bars represent the range values of duplicates. (B) Analysis of HA-tagged TRIM5α expression in the corresponding CrFK cell lines by Western blotting using anti-HA and anti-tubulin antibodies. The molecular weights of chimeric TRIM5α varied in size due to v2 length variation (rtlTRIM5α v2 is 32 amino acids longer than gml₁TRIM5α v2).

The 32-amino-acid insertion in v2 of ring-tailed lemur TRIM5α is a major determinant of its broad restriction activity.

To better understand the protein domains contributing to the marked differences in the antiviral activities of the two lemur TRIM5α proteins, we modeled the structure of B30.2/SPRY (Fig. 5A). The observation of profound differences in the length of the v2 loop triggered its functional analysis. We deleted the corresponding 32 amino acids in rtlTRIM5α and inserted them into gml₁TRIM5α. A new set of CrFK cell lines stably expressing recombinant and wild-type TRIM5α proteins was infected with a large panel of viruses (Fig. 5B). rtlTRIM5αΔ32aa displayed a reduction in its restriction activity against HIV-2, SIVmac, FIV, and EIAV, but the activity against HIV-1 and N-MLV was not significantly affected. The insertion of the 32 residues into gml₁TRIM5α resulted in a gain of restriction against multiple lentiviruses, including HIV-1, HIV-2, SIVmac, and EIAV; no effect on FIV restriction; and a loss of activity against N-MLV. These results indicated a major role for the 32-residue-long insertion in the v2 region in the broad antilentiviral activity of rtlTRIM5α.

Fig. 5.

Role of the 32-residue-long region in v2 of ring-tailed lemur TRIM5α. (A) Model structure of B30.2/SPRY domains of gray mouse lemur₁ TRIM5α (gml₁TRIM5α) and ring-tailed lemur TRIM5α (rtlTRIM5α) highlighting the v2 region of gml₁TRIM5α (green) and the extended v2 (red) that characterizes rtlTRIM5α. Loops corresponding to the four variable regions (v1 to v4) are indicated. Based on an analysis of their sequences, the v1, v2, v3 and v4 regions were predicted by PSIPRED (32) to have a coil conformation (discontinuous lines). (B) Infection of recombinant CrFK cell lines expressing gray mouse lemur₁- or ring-tailed lemur-derived TRIM5α constructs with selected GFP reporter retroviruses. rtlTRIM5αΔ32aa corresponds to rtlTRIM5α deleted of the 32-residue-long region in v2. gml₁TRIM5α+32aa corresponds to gml₁TRIM5α with those 32 residues inserted into its v2 region. Control cells contain an empty pLPCX vector. Infections were performed in duplicates, and data are representative of data from two independent experiments. Error bars represent the range values of duplicates. (C) Assessment of HA-tagged TRIM5α expression in the corresponding CrFK cell lines by Western blotting using anti-HA and anti-tubulin antibodies. The molecular weights of TRIM5α variants varied according to the presence of the 32-residue-long region in v2.

Restriction of an ancient PSIV by TRIM5α variants.

Finally, we evaluated the ability of various TRIM5α proteins to restrict the ancient prosimian immunodeficiency virus (PSIV), which was endogenized in the genome of the gray mouse lemur but not of the ring-tailed lemur (11, 12). The absence of PSIV sequences in the genome of the ring-tailed lemur was verified by BLAT analysis of the available whole-genome shotgun assembly (http://blast.ncbi.nlm.nih.gov/Blast.cgi) and by PCR using ring-tailed lemur genomic DNA with specific primers for the PSIV long terminal repeat (LTR), capsid, and envelope (data not shown). As the viral determinant for TRIM5α-mediated restriction is the capsid, we used chimeric EIAV-PSIV bearing an ancestral PSIV capsid, as recently reported (14). The PSIV capsid sequence was synthesized according to the consensus PSIV provirus inferred from PSIV copies found in the genome of gray mouse lemur and fat-tailed dwarf lemur (12), predicted to reflect the capsid sequence of an exogenous PSIV that circulated 4 million years ago. The PSIV capsid was inserted into an EIAV Gag-Pol vector to produce EIAV-PSIV viral particles by coexpression with the VSV-G envelope and an EIAV genome harboring lacZ. The viral titers were determined by transducing TRIM5α-expressing CrFK cell lines with serial dilutions of virus. Infections were performed with EIAV-PSIV and EIAV as control. Human, rhesus macaque, African green monkey, tamarin, and ring-tailed lemur TRIM5α proteins resulted in a strong restriction activity against EIAV-PSIV (Fig. 6A). gmlTRIM5α weakly restricted EIAV-PSIV. An analysis of the reciprocal swap of the 32 amino acids of v2 showed that this extended loop determines the activity of rtlTRIM5α against EIAV-PSIV (Fig. 6B). The infectivity of parental EIAV confirmed the expected pattern of restriction of the various TRIM5α proteins.

Fig. 6.

Restriction activity against chimeric PSIV. (A) Infection of CrFK cell lines expressing primate TRIM5α with LacZ reporter EIAV-PSIV and EIAV. (B) Infection of CrFK cell lines expressing recombinant prosimian TRIM5α with lacZ reporter EIAV-PSIV and EIAV. Control cells contain an empty pLPCX vector. Infections were performed with serial dilutions of virus in duplicates. TRIM5α-expressing CrFK and control cell lines were stained for β-galactosidase activity 3 days following infection. Titers were determined as lacZ-forming units per milliliter of virus by considering dilutions resulting in 20 to 200 lacZ-forming loci. Error bars represent the range values of duplicates. Results are representative of data from two independent experiments.

De novo sequencing of genomic DNA of the gray mouse lemur indicated that the endogenous PSIV capsid differed in 6 amino acids from the consensus gray mouse/fat-tailed dwarf lemur PSIV capsid (Fig. 7). The insertion of the corresponding mutations in the chimeric EIAV-PSIV Gag-Pol vector resulted in a defective PSIV. These experiments indicated that all tested TRIM5α proteins restricted endogenous PSIV. However, the lowest level of restriction was observed for its host primate, the gray mouse lemur. Analysis of the predicted 40-million-year-old ancestral sequence of TRIM5α (see Fig. S1 in the supplemental material) suggests that the acquisition of the expanded v2 probably endowed the ring-tailed lemur with a superior restriction of PSIV and of multiple exogenous lentiviruses. The pattern of lineage-specific expansion of variable regions (Fig. S1) limits the possibility of reconstructing and testing a parsimonious ancestral TRIM5α because there are no identifiable, shared ancestral loops across lineages.

Fig. 7.

Alignment of PSIV capsids. Six different amino acids are found between the consensus PSIV capsid, based on gray mouse and fat-tailed dwarf lemur PSIV proviruses, and the capsid specifically sequenced in the gray mouse lemur genome. Black, identity; gray, similarity. Triangles indicate the regions cloned into the EIAV Gag-Pol plasmid. The C terminus of the capsid is truncated in gray mouse lemur PSIV as previously described (11, 12).

DISCUSSION

In the present study, we identified TRIM5α proteins that display profoundly different antiretroviral activities in two lemur species. TRIM5α from the ring-tailed lemur strongly inhibited a representative set of retroviruses. In contrast, TRIM5α from the gray mouse lemur had a very limited spectrum of activity but a unique capacity to restrict B-MLV. TRIM5α proteins from both species inhibited PSIV, the only endogenous lentivirus identified in primates; however, the least active TRIM5α protein corresponded to the PSIV host primate. Analysis through chimeras and site-directed mutagenesis characterized the role of variable regions of the B30.2/SPRY domain and identified specific residues determining the spectrum of activity of the lemur TRIM5α proteins.

The study of prosimian TRIM5α may shed light on the development of primate antiretroviral specificities. Characterization of TRIM5 genes in hominoids, Old World monkeys, and New World monkeys reveals a molecular evolution strewn with frequent episodes of positive selection over the past 40 million years (13). Paleoviruses, i.e., ancient extinct viruses, likely shaped the evolution of TRIM5 to counteract past infections (10), resulting in the current TRIM5α sequences, which, while ineffective against established host retroviruses, are important in limiting cross-species transmission. We have shown previously that ancestral TRIM5α restricted HIV-1 better than did the hominoid TRIM5α ancestor and human TRIM5α (13). Likewise, SIV strains in hominoids and Old World monkeys are resistant to host TRIM5α but sensitive to other simian TRIM5α proteins in a species-specific manner (15, 25, 54). In New World monkeys, TRIM5α mediates species-specific restriction against foamy viruses naturally found in those species (39). More recently, TRIM5α was proven to modulate SIVmac replication in vivo (27) and to prevent the cross-species transmission of SIVsm strains in rhesus macaques (24).

Ring-tailed lemur TRIM5α has activity against a broad set of retroviruses, comparable to those of rhesus macaque TRIM5α and agmTRIM5α. v1 and v2 were both essential to rtlTRIM5α for restricting HIV-1, suggesting a combined role of these domains in capsid recognition, as described previously for orangutan TRIM5α toward HIV-1 and SIVmac (36). However, it is the acquisition of the 32-amino-acid insertion in v2 that may have brought a large advantage to rtlTRIM5α for the broad recognition of lentiviral capsids. The size of the 32-amino-acid extension (∼50 Å) in the v2 loop, comparable to the size of the core structure of B30.2/SPRY (40 to 50 Å), matches the dimensions of an HIV-1 capsid monomer (45 by 35 by 22 Å). The unique breadth of restriction conferred by v2 could be attractive in human gene therapy applications (43). Another remarkable aspect of lemur TRIM5α is the pattern of activity against gammaretroviruses. N-MLV, restricted by most primate TRIM5α proteins, was also sensitive to lemur TRIM5α. B-MLV, which is resistant to all natural TRIM5α proteins previously characterized, could be targeted by one allele of gray mouse lemur, gml₂TRIM5α. By comparing gml₂TRIM5α with the other allele, gml₁TRIM5α, we mapped the activity to residue H418 in v3. To date, only human TRIM5α mutants of residues R335 and Y336 in v1 were found to restrict B-MLV (30, 43); however, these residues are not observed in nature. The characterization of the degree of allelic diversity in lemurs would be very informative in the search for functional diversity; however, this type of analysis is limited by their status as protected species.

There are no reports of exogenous lentiviral infections in Malagasy prosimians, although the detection of cross-reactive antibodies against SIV and FIV in the wild-range ring-tailed lemur may suggest contemporary exposure (52). It is in this setting that the differences between gray mouse lemur and ring-tailed lemur TRIM5α proteins are surprising and possibly a reflection of differences in ecological niches, exposure, or susceptibility. The ring-tailed lemur is a large diurnal group-living and partially terrestrial lemur species that occurs exclusively in the very dry regions of southern Madagascar (34). The gray mouse lemur is a small nocturnal solitary forager, which is strictly arboreal, with a very broad omnivorous diet that includes arthropods, gum, fruits, insect secretions, and nectar (19). It has a large distribution from southern to northwestern Madagascar. Both species are known to coexist in some parts of southern Madagascar (e.g., Berenty and Andohahela) (34, 44); however, due to their profoundly different life-styles, they definitely occupy different niches.

This work uniquely addressed the role of various TRIM5α proteins, including the cognate gmlTRIM5α, in restricting PSIV, the only known example of an endogenous lentivirus in primates. Lentiviruses were thought not to exist as endogenous retroviruses until the discovery of rabbit endogenous lentivirus type K (RELIK) in lagomorphs and PSIV in lemurs, pushing back the estimated origin of lentiviruses to a minimum of 12 million years ago (20, 22). The reasons why PSIV invaded independently the genomes of gray mouse lemur and fat-tailed dwarf lemur 4 million years ago, but not those of other primates, remain unclear. Several primate TRIM5α proteins efficiently restricted an EIAV carrying an ancestral PSIV capsid. In contrast, gmlTRIM5α weakly restricted the PSIV chimeric construct. This pattern of limited activity against the cognate endogenous lentivirus is reminiscent of the lack of activity against exogenous SIV in the host primates, as discussed above. The fact that the long association of PSIV and the gray mouse lemur has not endowed this primate with a more active gmlTRIM5α may reflect that endogenization may lead to mutagenesis and to various forms of transcriptional silencing of the provirus and suppression of expression or viral structures. Thus, the current gmlTRIM5α protein would not have been positively selected to counteract the endogenous lentivirus.

The data on lemur contribute new information to a model of TRIM5α evolution under retroviral pressure. We estimated the possible 40-million-year-old ancestral state of TRIM5α in primates in an attempt to understand the diversification of its restriction activity. Ancestral reconstruction uses recent advances in phylogenetics to infer the ancestral states of genes while taking into consideration various uncertainties about the evolutionary process (59). The most striking process in TRIM5α evolution is the lineage-specific expansion of the v1 to v3 loops in the B30.2/SPRY domain. v1 expansion characterizes hominoid and Old World monkey TRIM5α proteins, and v2 expansion is observed for the cow and rat TRIM5 proteins and mouse TRIM30 (51, 58, 65), while v3 expansion characterizes New World monkey TRIM5α proteins. Here, the broad activity of rtlTRIM5α and the narrow activity of gmlTRIM5α depend on differences in v2 length. The lack of a consensus in variable regions limits the possibility of constructing a parsimonious ancestral TRIM5α protein. In this context, the absence of an expansion of variable regions in the gray mouse lemur TRIM5α protein could reflect the activity afforded by a “minimal” B30.2/SPRY domain.