|Home | About | Journals | Submit | Contact Us | Français|
The antiretroviral protein TRIM5α is known to have evolved different restriction capacities against various retroviruses, driven by positive Darwinian selection. However, how these different specificities have evolved in the primate lineages is not fully understood. Here we used ancestral protein resurrection to estimate the evolution of antiviral restriction specificities of TRIM5α on the primate lineage leading to humans. We used TRIM5α coding sequences from 24 primates for the reconstruction of ancestral TRIM5α sequences using maximum-likelihood and Bayesian approaches. Ancestral sequences were transduced into HeLa and CRFK cells. Stable cell lines were generated and used to test restriction of a panel of extant retroviruses (human immunodeficiency virus type 1 [HIV-1] and HIV-2, simian immunodeficiency virus [SIV] variants SIVmac and SIVagm, and murine leukemia virus [MLV] variants N-MLV and B-MLV). The resurrected TRIM5α variant from the common ancestor of Old World primates (Old World monkeys and apes, ~25 million years before present) was effective against present day HIV-1. In contrast to the HIV-1 restriction pattern, we show that the restriction efficacy against other retroviruses, such as a murine oncoretrovirus (N-MLV), is higher for more recent resurrected hominoid variants. Ancestral TRIM5α variants have generally limited efficacy against HIV-2, SIVagm, and SIVmac. Our study sheds new light on the evolution of the intrinsic antiviral defense machinery and illustrates the utility of functional evolutionary reconstruction for characterizing recently emerged protein differences.
A newly described form of innate immunity, coined “intrinsic immunity,” provides a constitutive line of defense, which relies on intracellular obstacles to hinder the replication of pathogens (1). This component of the immune system has gained much attention as a cornerstone of the resistance of mammals against several classes of retroelements and retroviruses (43).
Representative components of this cellular defense system include members of the tripartite motif (TRIM) family (21). The best-studied family member, TRIM5α (31), restricts retroviral infection by specifically recognizing the viral capsid and promoting its premature disassembly (3, 20, 32), and, as recently reported, by blocking viral production at a posttranslational stage (23). Human TRIM5α has limited efficacy against human immunodeficiency virus type 1 (HIV-1), while proteins encoded by some primate TRIM5α orthologs can potently restrict this particular lentivirus (18, 28, 29, 33). Longstanding selective pressures exerted by retroviruses and retroelements may have contributed to the generation of diverse patterns of antiretroviral specificity of TRIM5α and other host defense genes (18, 28).
To better understand the evolution of antiretroviral specificity patterns in primates, in particular along the lineage leading to humans, we utilized a functional evolutionary genomics approach (34). In the present study, we reconstructed ancestral primate TRIM5α sequences and tested their specificity in vitro against HIV-1 and five other retroviruses. We used these six present day viruses as extant markers to evaluate the functional differences over evolutionary time.
TRIM5α coding sequences from primates were obtained by amplification and sequencing of genomic DNA or cDNA or downloaded from the National Center for Biotechnology Information database (6, 12, 19, 25, 31, 38, 40): human (Homo sapiens AY625000), bonobo (Pan paniscus DQ229282), chimpanzee (Pan troglodytes AY923177), gorilla (Gorilla gorilla AY923178), Bornean orang-utan (Pongo pygmaeus AY923179), lar gibbon (Hylobates lar AY923180), nomascus (Hylobates leucogenys DQ229283), siamang (Hylobates syndactylus DQ229284), rhesus monkey (Macaca mulatta AY625001), olive baboon (Papio anubis AY843505), red guenon (Erythrocebus patas AY843514), African green monkey (Cercopithecus [chlorocebus] aethiops AY669399, Cercopithecus [chlorocebus] tantalus AY593973), eastern black-and-white colobus (Colobus guereza AY843507), douc langur (Pygathrix nemaeus AY843508), Bolivian titi (Callicebus donacophilus AY843519), Bolivian squirrel monkey (Saimiri boliviensis boliviensis AY928202), pygmy marmoset (Callithrix pygmaea AY843512), red-chested mustached tamarin (Saguinus labiatus AY843518), cotton-top tamarin (Saguinus Oedipus DQ229285), white-faced saki (Pithecia pithecia AY843515), Bolivian red howler monkey (Alouatta sara AY843511), common woolly monkey (Lagothrix lagotricha AY843520), and black-handed spider monkey (Ateles geoffroyi AY843516).
TRIM5α sequences were aligned by using CLUSTAL W. Coding regions were aligned according to their corresponding amino acid sequences by using the EMBOSS package (22). For the reconstruction of TRIM5α ancestral sequences (and for the calculation of posterior probabilities of reconstructed amino acids), we used a maximum-likelihood approach as implemented in the codeml tool (parameter settings according to model M0, i.e., model = 0 and NSsites = 0) of the PAML program package (37), in the framework of the accepted primate phylogeny (5). An alternative model (the free-ratio model, where each branch of the phylogeny may have a different KA/KS values) reconstructs the same amino acid variants at all sites and nodes, except for the single site 9 (I instead of V). However, since this site is located in the N terminus, which we demonstrate not to affect anti-HIV-1 specificities of TRIM5α (see below), this ambiguity is not relevant for our conclusions. To validate the predictions from the maximum-likelihood approach, we used a Bayesian approach as implemented in the MrBayes program (7) using the Jones amino acid model and the known species topology as a prior (the topology around node 3 was constrained, i.e., it was fixed to reflect the divergence of the orangutan from the other great apes, as the Bayesian TRIM5α gene tree groups the orangutan sequence with gibbons—against the well-established primate phylogeny). Posterior probabilities were calculated on 100 samples representing the last 10,000 generations (sampling frequency 100) of a total of 100,000 generations. We verified that the log likelihoods of the ancestral reconstruction had converged before the last 10,000 generations.
pLPCX-TRIM5αhu-HA (NIH AIDS Reagent Program, donated by J. Sodroski), an oncoretroviral vector encoding the human TRIM5α (31), was used to generate the ancestral genes of TRIM5α by consecutive rounds of site-directed mutagenesis using the QuikChange protocol (Stratagene). TRIM5α from African green monkey was cloned from COS-7 (European Collection of Cell Cultures, ECACC no. 87021302), and TRIM5α tamarin was cloned from cells derived from cotton-top tamarin (ECACC no. 85011419). pLPCX-TRIM5αrh-HA (J. Sodroski) encodes the rhesus monkey TRIM5α. Expression of hemagglutinin-tagged TRIM5α proteins was determined by Western blotting.
Oncoretroviral vectors were packaged in 293T cells by cotransfecting the various pLPCX-TRIM5α constructs with the pNB-tropic murine leukemia virus (MLV) Gag-Pol and pVSV-G packaging plasmids (a gift from D. Trono) using the calcium phosphate method. As controls, we also used pLPCX and pLPCX-GFP instead of pLPCX-TRIM5α. Supernatants were concentrated and used to transduce 105 HeLa and CRFK (feline renal fibroblasts) cells in the presence of 5 μg of Polybrene/ml. At 72 h after transduction, cells were selected in the presence of 0.5 μg (HeLa cells) and 3 μg (CRFK cells) of puromycin/ml for at least 12 days before testing.
To produce HIV-1-based reporter vector particles, 293T cells (3 × 106 cells) were cotransfected with four plasmids by the calcium phosphate method (16). Plasmids encoded the vesicular stomatitis virus G protein pantropic envelope (pMD.G), the Gag and Pol proteins (pCMVΔR8.92), and Rev (pRSV-Rev), and the fourth plasmid encoded the HIV vector segment carrying green fluorescent protein (GFP) as the reporter transgene (pSIN.cPPT.EF1.GFP.WPRE). Additional constructs represent N-MLV and B-MLV (a gift from D. Trono), HIV-2 (a gift from A. Lever), and the simian immunodeficiency virus (SIV) variants SIVmac (a gift from F.L. Cosset) and SIVagm (a gift from J. Luban).
The infectivity of recombinant viruses was determined by titration on HeLa and CRFK cells. Single-round infectivity assays with the different recombinant viruses in HeLa and CRFK cells were performed at various multiplicities of infection (MOIs). At 48 h after transduction, cells were analyzed by using a fluorescence-activated cell sorter.
To trace the evolutionary history of TRIM5α, we aligned the TRIM5α coding region from 24 primate species representing 25 million years of primate evolution. Evolutionary analysis allowed the estimation of the most likely sequence at each ancestral node in the framework of the accepted primate phylogeny (5) (Fig. (Fig.1A).1A). We then introduced, by site-directed mutagenesis, all predicted amino acid changes at each ancestral node in a stepwise fashion starting from a cloned human TRIM5α (Fig. (Fig.1B).1B). The oldest reconstructed sequence represents TRIM5α from the last common ancestor of Old World monkeys and apes, ~25 million years ago (mya), and differs by 29 amino acids from human TRIM5α. In addition, we tested TRIM5α from selected terminal taxa: Old World (rhesus monkey and African green monkey) and New World (cotton-top tamarin) monkeys. Attempts by our laboratory and other groups (28) to identify a TRIM5α ortholog in prosimians (e.g., lemurs and galagos) have been unsuccessful. Thus, a suitable outgroup that would permit us to infer the sequence of the common simian ancestor (i.e., Old World primate-catarrhine/New World monkey-platyrhine, 40 mya, Fig. Fig.1A)1A) is not available.
To test the restriction capacity of the different gene variants, we first stably transduced HeLa cells (a human epithelial carcinoma cell line) with oncoretroviral vectors expressing the various ancestral and modern TRIM5α sequences. Multiple independently transduced cell lines were examined for each sequence in order to identify those showing comparable levels of transgene expression (Fig. (Fig.2A).2A). We then infected the different cell lines with recombinant viruses—HIV-1, HIV-2, SIVagm, SIVmac, and the N- and B-tropic variants of the murine leukemia virus (N-MLV and B-MLV)—expressing the green fluorescent protein (GFP). The proportion of cells expressing GFP (relative infectivity) was then used as a proxy for the capacity of a particular TRIM5α sequence to restrict infection of cells by the respective recombinant virus.
We observed a statistically significant reduction of HIV-1 restriction by reconstructed TRIM5α variants (representing 25 million years of evolution) from the common catarrhine ancestral sequence toward the human variant. Ancestral TRIM5α variants are generally poorly restrictive of HIV-2, SIVagm and SIVmac (Fig. (Fig.2B).2B). The pattern of HIV-1 restriction by successive ancestral variants was unique among the various retroviruses tested (Fig. 2C to G). Similar to the reconstructed node 1 sequence, TRIM5α from Old World monkeys displays a high capacity to restrict HIV-1. Some Old World monkey lineages appear to have acquired restriction capacity against other SIVs since the common catarrhine ancestor.
We carefully assessed the amino acid variants included in the reconstruction of each ancestral protein in the context of published experimental data of single and complex mutagenesis of TRIM5α (9, 14, 18, 31, 33, 39). The high HIV-1 restriction capacity of the 25 mya ancestral TRIM5α could not be deduced from the reconstructed sequence based on previous evidence from the literature. In particular, no ancestral construct carries the critical residue proline 332, associated with potent restriction of HIV-1 in previous studies (14, 39). The present study reconstructs position 332 with a glutamine (Q) in nodes 1 to 4 with high probability (P >0.95%; a substitution from glutamine to arginine occurred in the human-chimpanzee ancestor, between nodes 4 and 5, Fig. Fig.1B).1B). A glutamine at position 332 plays a role in the restriction activity of the gorilla B30.2 domain (18), and the change in restriction of HIV-1 between nodes 4 and 5 may be explained by a substitution from Q to R at this site (10). In general, the ability of human TRIM5α to bind the HIV-1 capsid is modulated by the presence of any charged residue at position 332 (14). However, despite the maintained presence of 332Q between nodes 1 and 4, HIV-1 restriction diminishes between these nodes. It should also be emphasized that the change in restriction occurred without any increase in length of the variable regions of the B30.2 domain (28, 33), since these regions were kept stable in size for the reconstruction of ancestral variants. Residue K389 was built as Q in nodes 4 to 1; however, recent analysis of extant sequences indicates that only nodes 3 to 1 should carry 389Q. Functional analysis indicates that this error is unlikely to modify restriction capacity (see below).
Reconstructed ancestral sequences also reveal amino acid substitutions in the other protein domains of TRIM5α. However, single-amino-acid mutagenesis data from published functional studies (9) do not include any of the positions mutated here, with the exception of the R136Q substitution in the coiled-coil domain. This variant was shown not to modify TRIM5α restriction of HIV-1 (4, 24, 30), while in a study by Javanbahkt et al. the 136Q variant of human TRIM5α was associated with protection from HIV-1 infection in African Americans and with a weak restriction phenotype in vitro (8).
In contrast to the pattern seen for HIV-1, restriction of the murine oncoretrovirus N-MLV was stronger for reconstructed TRIM5α of extant taxa than for more ancestral TRIM5α constructs (Fig. (Fig.2F).2F). The functional change was observed without changes in the RING domain, previously reported to be relevant for N-MLV restriction (24) (Fig. (Fig.1B).1B). Specifically, ancestral nodes do not carry the substitution H43Y, a frequent human TRIM5α allele that may negatively affect its putative E3 ubiquitin ligase activity and is associated with impaired restriction of N-MLV (4, 24). The detrimental allele dates back to before the emergence of the African diaspora and is found at a frequency of 43% in indigenous Central and South Americans (24).
Ancestral TRIM5α showed various patterns of restriction capacities against other retroviruses tested (HIV-2, SIVagm, SIVmac, B-tropic variants of the MLV, B-MLV) (Fig. 2C to E and G). Overall, the 25-mya ancestral TRIM5α (node 1) exhibited limited efficacy against HIV-2, SIVagm, and SIVmac. Extant TRIM5α variants from Old World monkeys display higher restriction capacity against HIV-2 than the reconstructed TRIM5α from the common catarrhine ancestor (suggesting a gain in restriction capacity), whereas ancestral hominoid TRIM5αs all show little restriction efficiency with respect to this virus.
To rule out cell line-specific effects, we assessed the antiretroviral activities of selected ancestral and modern TRIM5α in another cell line (Fig. (Fig.3),3), CRFK (feline renal fibroblasts, no intrinsic retroviral restriction ). The CRFK series was less complete than that for HeLa cells, because we failed to generate stable cell lines carrying node 3 or cotton-top tamarin TRIM5α variants, whereas all tested clones carrying node 2 TRIM5α were massively overexpressing the protein. However, the overall pattern was confirmed: (i) decreasing restriction of modern HIV-1 from nodes 1 to 5, (ii) more restriction of N-MLV in humans and the human-chimpanzee ancestral TRIM5α, and (iii) greater restriction of HIV-2 by TRIM5α from Old World monkeys. Consistent with previous reports (6, 42), we found that the expression of human TRIM5α in CFRK is associated with partial restriction of HIV-1 and HIV-2, which explains the different results for the human TRIM5α between the two cell lines tested. We note that we used vesicular stomatitis virus-pseudotyped particles for our challenge experiments. While the characteristic of the envelope protein is reported to not significantly change restriction by TRIM5α, human TRIM5α restricts primary HIV-1 isolates to variable degrees (11).
A potential concern with respect to the functional reconstructions in our study is the accuracy of the inferred ancestral TRIM5α sequences. The sequences studied are from closely related species, an approach which, in principle, facilitates ancestral inference. However, some regions of TRIM5α may have been affected by multiple (parallel and convergent) substitutions at the same site due to positive selection that has been driving its evolution (17), which may render reconstruction of ancestral variants potentially less reliable. Nevertheless, we believe that the ancestral proteins tested here are of adequate quality, for the reasons detailed below.
The posterior probabilities obtained for the reconstruction of key residues using the maximum-likelihood procedure are generally high (the proportions of reconstructed amino acids with a P of >0.95 were 92% [node 1], 97% [node 2], 98% [node 3], 98% [node 4], and 99% [node 5]), and the reconstructions are robust with respect to different models and parameters used. Furthermore, J. W. Thornton (34) indicates that errors in ancestral sequence reconstruction are generally expected to bias resurrected genes toward less functional variants. However, in our study, ancestral sequences exhibit greater efficacy against HIV-1 than modern TRIM5α. In contrast to J. W. Thornton, Williams et al. (35) suggest that the assumption that reconstruction errors can be considered similar to random (i.e., generally deleterious) mutations is questionable. In their view, extant variants used for reconstructions are necessarily acceptable in the functional context, which cannot be assumed for random mutations (35).
Williams et al. (35) compared the properties of ancestral proteins inferred by using maximum-parsimony, maximum-likelihood, and Bayesian methods. These researchers concluded that these approaches offered various benefits. To address the possibility of systematic error and to accommodate reasonable alternative scenarios in ancestral reconstruction, we used an alternative Bayesian approach for reconstruction, in addition to the maximum-likelihood approach (see Materials and Methods for details). Key amino acid positions, for which the maximum-likelihood procedure yielded potentially ambiguous reconstructions (i.e., relatively low posterior probabilities for reconstructed variants) and for which a different amino acid was reconstructed by the Bayesian approach, were incorporated into the functional analyses. For the 29 residues that were found to differ between human TRIM5α and node 1 based on the maximum-likelihood procedure, 5 showed low codeml probabilities (<95%), and an alternative amino acid was supported by Bayesian analysis (Table (Table1).1). We successfully tested four of the alternative amino acids (49Y/H, 131 M/V, 340 M/V, and 483T/E) by introducing them into the original node 1 sequence and failed to build 253V. None of the ambiguous-position variants modified the restriction of HIV-1 (Fig. (Fig.4),4), which ensures that potentially erroneously reconstructed residues likely do not account for the efficient restriction of HIV-1 by the node 1 TRIM5α sequence.
Another issue relates to the recent finding of long-term persistence of multiple TRIM5α alleles at critical positions due to balancing selection (17). Thus, reconstruction of specific nodes may be complicated by the simultaneous presence of key alleles in multiple lineages. Humans present features of balancing selection at position R136Q (4), while sooty mangabeys and rhesus monkeys (17) present multiple alleles at positions 182, 194, 213, 331, 332, indel 337-338, and 339 (human numbering). However, we analyzed the accuracy of reconstruction for these positions (except for indel 337-338, which is not present in all species and is thus not included in the analysis) and found that the ancestral inference at these sites appears to generally be unambiguous (with high posterior probabilities, >98%, for the reconstructed amino acid variant) and not affected by the putative trans-species polymorphisms. In addition, we searched novel TRIM5α submissions to GenBank, aligned available entries to the sequences from our study, and screened critical residues under positive selective pressure for the presence of polymorphism within the species. We re-estimated node 1 under the premise of balancing selection and found the original prediction unchallenged on the basis of currently available data on intraspecies polymorphism. Only residue 253 could have been built as 253V in addition to 253I at node 1, due to the fact that gorilla, orangutan, and nomascus primates are now known to be polymorphic at this position.
After confirming the reliability of reconstruction of node 1, we sought to describe domains and residues that could contribute to node 1 restriction. For this, we generated reciprocal chimeric constructs of the N1-284- and C285-493-terminal portions of the node 1 and node 4 sequences. We then assessed their restriction capacities using CRFK cells (Fig. (Fig.4).4). This analysis demonstrates that only substitutions in the C-terminal domain appear to have affected restriction of HIV-1 by reconstructed TRIM5α on the human lineage (Fig. (Fig.4),4), since it is the hybrid construct with the node 1 C terminus that displays high restriction capacity (as high as that of the intact node 1 sequence).
We then selected residues in the C-terminal region of node 1 that (i) differed from human TRIM5α and (ii) were under positive selective pressure (Table (Table1).1). Using site-directed mutagenesis experiments, we demonstrate that changing the ancestral 324E to the more recent 324K (a substitution occurring on the lineage leading to the last common African ape ancestor, node 4) or changing the ancestral 335T residue to 335R (originating in the common human/chimpanzee ancestor on the lineage leading to node 5) strongly reduces the ability of the 25-mya ancestral TRIM5α to restrict HIV-1. These positions have been previously investigated as 324N and 335L and found to be associated with restriction of HIV-1 (14, 33). In addition, a sequence context that includes 324E/332Q/335T in variable region v1 and 389Q in the v2 region was associated with effective restriction in a recent study (18). For amino acid residue 335, we tested another amino acid (335N) found in Old World monkeys. The resulting sequence variant maintains the high restriction capacity seen for the original node 1 sequence (Fig. (Fig.4),4), which underscores that the loss of restriction capacity is specific to reversion to 335R. Thus, the present study reveals novel functionally relevant amino acid variants at positions 324 and 335 and generally confirms the importance of these sites for virus restriction by TRIM5α. It is noteworthy, however, that although this sequence context is maintained from nodes 1 to node 3, restriction efficacy is gradually diminishing between these nodes. This suggests that other substitutions and/or sites (potentially in combination) explain the diminishing restriction capacity of reconstructed TRIM5α compared to reconstructed TRIM5α of the last common Old World primate ancestor (node 1). We individually tested three additional residues (340V, 389K, and 483G) that were previously shown to have evolved under positive selection (Table (Table1)1) in this domain, but none of these substitutions appear to reduce the restriction of the original node 1 TRIM5α protein when changed individually (Fig. (Fig.4).4). We did not assess the ambiguous position C385, which corresponds to the V2 region of length polymorphism. Substitution of tyrosine with cysteine at position 385 does not modify the restriction of HIV-1 (18).
The novelty in the present study resides in the experimental approach used to characterize recently emerged protein differences (reconstruction of ancestral host sequences). The strategy was able to identify and experimentally test new variants despite the fact that this protein has been very thoroughly investigated over the past 3 years. Thus, evolutionary genetics and ancestral reconstruction could be of interest in future investigation of less-well-characterized proteins. Interestingly, as for TRIM5α, evolutionary analysis of a second antiretroviral protein, APOBEC3G (27), would also predict shifts in antiretroviral specificity on the human lineage since the catarrhine ancestor, because the ancestral APOBEC3G likely carried the amino acid variant K128 that governs sensitivity of this protein to modern HIV-1 Vif-mediated inhibition (2, 15, 26, 36). There is clearly a great level of complexity in imputing an evolutionary direction of restriction of modern viruses by reconstructed TRIM5α variants. However, ancestral proteins can in the future be tested against resurrected infectious agents (10, 13).
Such various restriction capacities might be explained by lineage-specific pandemics that shape and redirect the intrinsic defense mechanisms. For example, the recent retroviral infection of two great ape lineages—chimpanzees and gorillas—by horizontal transmission from an exogenous source of PtERV1 has not affected the human genome (41). Recent data indicate that the R332 mutation in human and chimpanzee TRIM5α improves the ability of this protein to restrict PtERV1 while resulting in increased susceptibility to HIV-1 (10). Thus, as a result of trade-offs in the ability to restrict different retroviruses, humans might have been exposed to SIVcpz and/or HIV-1 at a time when one or several intrinsic defense proteins lacked the appropriate specificity to avert its transmission from chimpanzees.
We thank D. Trono, A. Lever, F. L. Cosset, J. Luban, and G. Towers for reagents.
A.T. and H.K. conceived the study. A.T., V.G., and A.C. designed the experiments. V.G., A.C., M.O., and M.M. performed the experiments. MO., D.B., and H.K. performed the evolutionary analysis. A.T., H.K., V.G., and A.C. wrote the report.
This study was funded by the Swiss National Science Foundation and a grant for interdisciplinary research from the Faculty of Biology and Medicine of the University of Lausanne.
Published ahead of print on 12 December 2007.