Homology-based Identification of Capsid Determinants That Protect HIV1 from Human TRIM5α Restriction*

The tropism of retroviruses relies on their ability to exploit cellular factors for their replication as well as to avoid host-encoded inhibitory activities such as TRIM5α. N-tropic murine leukemia virus is sensitive to human TRIM5α (huTRIM5α) restriction, whereas human immunodeficiency virus type 1 (HIV1) escapes this antiviral factor. We previously revealed that mutation of four critical amino acid residues within the capsid can render murine leukemia virus resistant to huTRIM5α. Here, we exploit the high degree of conservation in the tertiary structure of retroviral capsids to map the corresponding positions on the HIV1 capsid. We then demonstrated that, when changes were introduced at some of these positions, HIV1 becomes sensitive to huTRIM5α restriction, a phenomenon reinforced by additionally mutating the nearby cyclophilin A binding loop of the viral protein. These results indicate that retroviruses have evolved similar mechanisms to escape TRIM5α restriction via the interference of structurally homologous determinants in the viral capsid.

The tropism of retroviruses relies on their ability to exploit cellular factors for their replication as well as to avoid hostencoded inhibitory activities such as TRIM5␣. N-tropic murine leukemia virus is sensitive to human TRIM5␣ (huTRIM5␣) restriction, whereas human immunodeficiency virus type 1 (HIV1) escapes this antiviral factor. We previously revealed that mutation of four critical amino acid residues within the capsid can render murine leukemia virus resistant to huTRIM5␣. Here, we exploit the high degree of conservation in the tertiary structure of retroviral capsids to map the corresponding positions on the HIV1 capsid. We then demonstrated that, when changes were introduced at some of these positions, HIV1 becomes sensitive to huTRIM5␣ restriction, a phenomenon reinforced by additionally mutating the nearby cyclophilin A binding loop of the viral protein. These results indicate that retroviruses have evolved similar mechanisms to escape TRIM5␣ restriction via the interference of structurally homologous determinants in the viral capsid.
Viruses are obligatory parasites that require specific components of the host cell machinery to replicate. However, their tropism also depends on their ability to overcome hostencoded intracellular antiviral activities termed restriction factors (1,2). TRIM5␣ is one such activity that restricts retroviruses by targeting their capsid early after viral entry, thus provoking the premature uncoating of incoming virions (3)(4)(5).
TRIM5␣ is part of the tripartite motif (TRIM) 2 protein family, characterized by the presence of a RING finger, a B-Box II, and a coiled-coil domain collectively forming the so-called RBCC motif (6,7). TRIM5␣ also contains a C-terminal PRYSPRY domain responsible for direct binding to the retroviral capsid (CA) (8 -12). Species-specific amino acid variations, notably in the PRYSPRY domain, dictate the spectrum of retroviruses restricted by different TRIM5␣ orthologues (13)(14)(15). Remarkably, a given TRIM5␣ species can target a wide range of retroviruses. For example, human TRIM5␣ (huTRIM5␣) blocks gammaretroviruses such as N-tropic murine leukemia virus (N-MLV) as well as lentiviruses such as feline immunodeficiency virus (FIV) and equine infectious anemia virus, but it cannot efficiently block human immunodeficiency viruses types 1 (HIV1) and 2 (HIV2) (16 -20). In contrast, rhesus macaque TRIM5␣ (rhTRIM5␣) can inhibit all of theses viruses albeit with variable efficiencies (21,22).
The retroviral capsid is derived from the proteolytic cleavage of the Gag (group-specific antigen) precursor, which provides all the structural elements responsible for organizing the virion (23). Gag is synthesized as a polyprotein of 55 kDa that assembles at the plasma membrane together with the viral genome, leading to the release of immature virions. Gag is then cleaved by the viral protease to generate matrix, CA, and nucleocapsid homomultimers that reassemble to form the three inner shells of the mature virion. Although matrix lies beneath the envelope-bearing viral outer lipid bilayer, CA forms the virion inner core, wrapping the nucleocapsid-associated viral genome and enzymes (23).
Mature retroviral capsids can adopt spherical (e.g. MLV), cylindrical (e.g. Mason-Pfizer Monkey virus) or conical (e.g. HIV) morphologies, but despite these differences, they exhibit a remarkable degree of structural conservation, all deriving from a hexagonal array of hexameric CA monomers (24 -28). Assembly is stabilized by homophilic interactions between the N-and C-terminal CA domains. The C-terminal domain, which is buried inside the capsid, mediates both interhexameric and intrahexameric contacts, whereas the N-terminal domain (NTD), exposed at the surface, forms the hexameric rings. Remarkably, although CA proteins from different retroviruses display little sequence identity, their tertiary structures are highly conserved (27, 29 -38). The NTD is composed of six (MLV) or seven (HIV) ␣-helices, and for lentiviruses only, it includes an extended loop between helix 4 and 5. In the case of HIV1 and FIV, this loop is known to bind the cellular peptidyl prolyl cis-trans isomerase cyclophilin A (CypA) ( The precise role of CypA during HIV1 infection is still not well understood, yet many studies suggest that it facilitates viral disassembly (39,(42)(43)(44)(45)(46). CypA is packaged into the virions during viral assembly through binding with the CA moiety of the Gag polyprotein (47)(48)(49). CypA is recruited by CA proteins from HIV1, FIV, simian immunodeficiency virus (SIV) from African green monkeys (SIVagm) and chimpanzee but not from other lentiviruses such as equine infectious anemia virus and SIV from macaque (SIVmac) (41,(47)(48)(49)(50)(51)(52).
The CA-CypA interaction can be prevented by mutating CA at positions 89 (CA89) and 90 (CA90) or by using the competitive inhibitor cyclosporine A (CsA). This leads to a decrease in HIV1 infectivity in human cells through mechanisms that are still unclear (44,47,49). Although CypA can both bind to and catalyze the cis-trans isomerization of the peptide bond between HIV1 CA89 and CA90, it is not known which of these two properties is required for its effect on viral infectivity (42). However, CsA-based experiments have revealed that CypA recruitment in target rather than producer cells is responsible for boosting HIV1 infectivity (53).
CypA has also been demonstrated to impact upon TRIM5␣-mediated restriction. In particular, HIV-1 restriction by TRIM5␣ from Old World monkeys such as rhesus macaque is impaired when target cells are treated with CsA or CypA-directed RNA interference or, to some extent, by mutating the cyclophilin binding loop of CA (54 -56). In contrast, the positive effect of CypA on HIV1 replication in human cells is independent of TRIM5␣ (55,57). Remarkably, retrotransposition of the CypA coding sequence into the TRIM5 gene has occurred on two independent occasions during primate evolution (58 -64). As a result, owl monkeys (part of the New World monkeys) as well as two macaque species (part of the Old World monkeys) express a chimeric protein, TRIMCyp, in which the PRYSPRY domain is replaced by CypA, allowing the capture of incoming retroviral capsids through a CypA-CA interaction. HIV1 successfully infects human cells partly due to its ability to escape huTRIM5␣, a feat that N-MLV cannot accomplish. However, if the N-MLV capsid is mutated at four key positions (CA82, CA109, CA110, and CA117) situated at the edge of a cavity formed by helices 4 -6, it is no longer inhibited by huTRIM5␣ (19,65,66).
Here, we take advantage of the high conservation of retroviral capsids to identify the corresponding positions of the HIV1 CA. Through a combination of genetic and functional studies, we then demonstrate that these positions can be engi- neered to render HIV1 susceptible to huTRIM5␣. We further reveal that this effect is amplified by preventing CA-CypA binding. Together, these results shed interesting light on the mechanisms of TRIM5␣-mediated restriction and escape there from and highlight structural determinants of the virus that could serve as attractive drug targets.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture-Mus dunni tail fibroblasts (MDTF) were purchased from the American Type Culture Collection and cultured in Dulbecco's modified Eagle's medium as described (66).
Plasmids-MLV-based particles were produced using packaging constructs N and B-tropic MLV (pCIG3-N and pCIG3-B) and the GFP-encoding construct pCNCG as de-scribed (66). HIV1-based vectors were produced with the packaging construct pCMV⌬R8.74 and the vector pRRLsin PGK GFP. The env construct for all viral productions was pMD2G, expressing vesicular stomatitis virus G protein.
The MLV plasmids encoding human, rhesus macaque TRIM5␣, or owl monkey TRIMCyp (pLPCX-huTRIM5␣-HA, pLPCX-rhTRIM5␣-HA, pLPCX-TRIMcyp-HA) were kind gifts from J. Sodroski. The squirrel monkey TRIM5␣ sequence was amplified from pCXCR-sqTRIM5␣ (kindly provided by G. Towers) and fused in 3Ј by PCR with the sequence encoding the influenza virus hemagglutinin (HA). The PCR product was inserted downstream of the CMV promoter of the MLV vector construct pLPCX (Clontech) allowing for puromycin selection of transduced cells. HuTRIM5␣ and MLV capsid mutants were already de-FIGURE 2. Identification of positions on the HIV1 capsid that are structurally homologous to MLV CA82, CA109, CA110, and CA117. A, shown is structural sequence alignment using known (HIV1 and N-and B-MLV) and predicted (HIV-2, SIVmac, and SIVagm) structures of retroviral capsid N-terminal domains (CA NTD) based on a combination of residue identity/similarity and secondary structure correspondence. Residues at the four positions on the HIV1 capsid (CA83, CA120, CA122, and CA129) homologous to MLV CA82, CA109, CA110, and CA117 are depicted in the superimposition CA NTDs. The glycine at HIV1 CA89 on the cyclophilin binding loop is also illustrated. Distinct colors are used to represent individual CA NTDs structures. B, the same structural sequence alignment as in A is shown but depicting only the N-MLV (violet) and HIV1 (blue) CA NTD (central panel) with the four HIV1 residues CA83, CA120, CA122, and CA129 shown. Highlighting these four positions shows the proximity and similar orientation of HIV1 CA120 with MLV CA110 (top left), HIV1 CA122 with MLV CA109 (bottom left), HIV1 CA129 with MLV CA117 (top right), and HIV1 CA83 with MLV CA82 (bottom right). C, shown are residues present at the four homologous positions from MLV in N-and B-MLV, HIV1, HIV2, SIVmac, and SIVagm. Of note, the previously unknown structure of the N-terminal domain of the HIV2 capsid has recently been resolved, whereas this present study was ongoing (64). Superimposition of the predicted model of the HIV2 capsid with its defined structure showed a low averaged root mean square deviation (0.9 Å) (supplemental Fig. S1). scribed (66). All HIV1 capsid mutants were generated by site-directed mutagenesis of pCMV⌬R8.74 using the XL QuickChange mutagenesis kit (Stratagene). All primers used in this study are listed in supplemental Table S2.
Homology Modeling and Structural Sequence Alignment of Retroviral Capsid Proteins-The HIV2 (Uniprot entry P04584 (67)), SIVagm (UniProt entry Q02836), and SIVmac (UniProt entry P05897) capsid protein sequences were defined as target sequences for homology modeling. Experimental crystal structures of HIV1 (PDB (68) ID 1AFV (37)), N-MLV (PDB ID 1U7K (27)), and B-MLV (PDB ID 3BP9 (38)) were used as structural templates. Multiple sequence alignment of the target and template sequences was performed using the MUSCLE algorithm (69,70). HIV and SIV belong to the lentivirus retroviral genus (71), and the sequence identity between their capsid proteins ranges from 60 to 86%. MLV belongs to the gammaretroviral genus (71), and the sequence identity between the N-MLV and B-MLV capsid proteins is 97%. The sequence identity for the capsid proteins between the lentiviruses and gammaretroviruses is only around 13%. Based on the target-template sequence alignments, model structures of HIV-2, SIVagm, and SIVmac capsid proteins were built by satisfaction of spatial restraints using the MODELLER program Version 9.1 (72). 2000 models were built for each protein, and the final model was selected based on the modeler objective function.
The structural sequence alignments were performed using the MatchMaker command of the UCSF Chimera program (73). This alignment is based on a combination of residue identity/similarity and secondary structure correspondence. The former is obtained using the Needleman-Wunsch algorithm (74) with the BLOSUM-62 (75) residue similarity matrix, whereas the secondary structure assignment is computed with the Kabsch and Sander algorithm (76). The weights applied to the residue similarity and secondary structure during the computation of the structural alignments were 0.7 and 0.3, respectively.
Vector Production and Determination of Titers-HIV1based GFP vectors were produced by CaPO 4 -mediated transient co-transfection of the retroviral vector, gag-pol, and env encoding constructs at a ratio of 2.8:1.8:1 as described (66). All vector-containing supernatants were centrifuged, filtered, and concentrated by ultracentrifugation and resuspended in PBS, whereas producer cells were harvested to determine gag expression levels (see below). The physical titers were determined by monitoring the reverse transcriptase enzymatic activity of a certain volume of PBS-containing vectors as described (77). Infectious titer was determined by a single-round infectivity assay on Fv1-null MDTF cells. Virion infectivity was calculated by dividing the infectious titer by the physical titer.
Gag Expression and Processing-To test Gag expression, total proteins were extracted from producer cells in radioimmune precipitation assay buffer as described (66). Equal amounts of proteins were resolved on SDS-PAGE followed by immunoblot. Gag polyproteins were detected using a mouse FIGURE 3. Western blot analysis of HIV1 particles production. A, shown is a Western blot of extracts from 293T cells transfected with the retroviral vector, the envelope plasmid, and the packaging construct harboring either a WT or a mutated CA and incubated with either anti-p24 (top)-or anti-actin (bottom)-specific antibodies. Of note, the structure of the HIV1 capsid used for the superimposition analysis (Fig. 2) displayed a leucine at CA83, yet the wild type HIV1-based vector used in this study harbored a valine at that position. B, shown are Western blots of concentrated virions harboring a CA with the indicated mutation produced in the presence (ϩ) or absence (Ϫ) of CsA and analyzed using the anti-p24 (top) and anti-CypA antibodies. monoclonal antibody against the HIV1 capsid p24. 3 To ensure equal loading, the level of actin was assessed using a mouse monoclonal antibody (Chemicon) followed by a secondary sheep anti-mouse antibody conjugated to horseradish peroxidase (HRP). Viral proteins were analyzed by loading equal amounts of virions (based on reverse transcriptase activity as described (77)) on SDS-PAGE followed by immunoblot. Gag protein in virions was detected using the same antibody as described above. A polyclonal CypA-specific antibody followed by a donkey anti-rabbit HRP-conjugated antibody was used to determine CypA incorporation into virions.
Generation of Cell-stable Cell Lines-Retroviral vectors encoding TRIM5␣ variants were produced using the pLPCXderived plasmids as described above. 5 ϫ 10 4 MDTF were transduced with viral supernatants containing recombinant retroviral vectors and selected 3 days later by the addition of puromycin (Sigma) at a concentration of 5 g/ml and maintained continuously in the presence of the drug. TRIM5␣ expression levels were determined by immunoblot as described (66). The HA-tagged proteins were detected using a peroxidase-conjugated rat monoclonal (clone 3F10, Roche Applied Science). Actin was used as a protein loading control and detected as described above.
Infection with GFP Reporter Vectors-MDTF cell lines were seeded at 2.5 ϫ 10 4 in a 24-well plate and transduced 24 h later with, in the case of MLV-based vectors, 6 doses of 2-fold serial dilutions of GFP reporter vectors. Infections with HIV1based GFP vectors were performed with 3 doses (in duplicate) or 6 doses of 2-fold serial dilutions starting with equal amounts of virions particles normalized by reverse transcriptase activity as described (77). In experiments using CsA, the drug was added at the time of infection at a concentration of 5 M. Cells were harvested 3 days post-transduction, fixed in 1% formaldehyde-containing PBS, and resuspended in PBS 1% fetal calf serum. The percentage of GFP-positive cells was determined by flow cytometry using the BD Biosciences FACScan. Results were analyzed with FlowJo 8.1.1 software. The -fold restrictions were calculated as described (66).

Four Residues of the MLV Capsid Influence Restriction by
Different Alleles of TRIM5␣-Each orthologue of TRIM5␣ blocks a different range of retroviruses, yet the targeted capsids present a high degree of structural homology. We focused our attention on four MLV CA residues (CA82, CA109, CA110 and CA117) that were previously demonstrated to condition susceptibility to huTRIM5␣, asking whether these positions also influenced MLV restriction by other TRIM5␣ species (19,65,66).
HuTRIM5␣ and rhTRIM5␣ both restrict N-MLV, albeit with different efficiencies, yet only the rhesus orthologue blocks HIV1 (16 -19, 21, 78). However, huTRIM5␣, with a change at arginine 332, gains the ability to restrict HIV1, at least to a certain extent (15,79,80). We, hence, engineered cell lines stably expressing a series of TRIM5␣ variants by retroviral vector-mediated transduction of permissive Fv1-null MDTF (Fig. 1A). We then challenged these cells with N-or B-MLV GFP vectors harboring either wild type capsid or a capsid mutated at one of the four critical positions and monitored infection by flow cytometry (Fig. 1B). Both wild type and R332A huTRIM5␣ could restrict N-MLV, as could rhTRIM5␣, although to a lesser extent as expected. Single mutations at position 82 (Asn to Asp) or 117 (Leu to His) of the N-MLV capsid were sufficient to confer it with complete protection from rhTRIM5␣, whereas combined mutations at these positions were required for the virus to significantly escape the block by R332A but not by wild type huTRIM5␣ (Fig. 1B). Conversely, although B-MLV was resistant to all three TRIM5␣ variants, replacing the CA glutamate 110 with an alanine was sufficient to render it susceptible to both huTRIM5␣ derivatives but not to rhTRIM5␣, although we did find that the presence of an arginine at this position could render B-MLV more sensitive to the rhesus allele (Fig. 1B, supplemental Table S1). Additional changes at MLV CA82, CA109, CA110, and CA117 further indicated that these positions exert an influence on restriction by the three TRIM5␣ variants (supplemental Table S1).
Homology-based Identification of HIV1 Capsid Determinants Modulating huTRIM5␣-mediated Restriction-The finding that some positions of the MLV capsid modulate re- The physical and functional titers were measured by the amount of reverse transcriptase activity in cpm/ml or the transducing units (TU) per ml, respectively. The relative infectivity was calculated by the ratio between the functional and physical titers. NI, non-infectious; ND, not done.

HIV1 strain
Physical titer Functional titer Relative infectivity -Fold decrease in relative infectivity compared to WT 10  striction of different TRIM5␣ variants suggests conservation in the modalities of TRIM5␣-capsid recognition. However, HIV1 is protected from restriction by huTRIM5␣, but not by huTRIM5␣ R332A and rhTRIM5␣. Furthermore, a change at tyrosine 336 of huTRIM5␣ (huTRIM5␣ Y336K) was previously demonstrated to expand its activity toward B-MLV, Moloney MLV (Mo-MLV), SIVmac, and HIV-2, whereas HIV-1 and SIVagm were still able to escape restriction (66,81). We, hence, sought to identify positions in the HIV1 capsid that were responsible for interfering with viral capture by both wild type and Y336K huTRIM5␣. We used the structural data available for the N-terminal domain of N-and B-MLV and HIV1 capsids to predict the three-dimensional structure of HIV2, SIVmac, and SIVagm capsids by assuming similar protein folding (27,37,38). We then performed a superimposition of these tertiary structures taking into account both primary sequences and structure alignments ( Fig. 2A). Positions CA83, CA122, CA120, and CA129 on HIV-1 were found to be homologous to the four MLV capsid sites modulating TRIM5␣-mediated restriction (Fig. 2, B and C). Interestingly, with the exception of CA129, residues found at these positions in HIV1 differ markedly from the amino acids present at homologous positions of other capsids (Fig. 2C). To test whether these identified HIV1 capsid positions interfere with restriction mediated by wild type or Y336K huTRIM5␣, we substituted residues at HIV1 CA83 (V83Q), CA122 (P122Q), and CA120 (H120R) with the corresponding amino acids found on viruses restricted by either or both TRIM5␣ variants, e.g. N-, B-MLV, HIV2, and SIVmac (Fig. 2C). Because HIV1, in contrast to these latter viruses, binds the cellular protein CypA, we decided to include also a mutation of CA89 (G89V) that was previously demonstrated to disrupt this interaction (Fig. 2, A and B) (47)(48)(49).
Physical and Functional Characterization of HIV1 Capsid Mutants-We produced HIV1-based GFP vectors harboring single or combined mutations at positions CA83, CA89, CA122, and CA120. To investigate any effect of these mutations on Gag polyprotein expression, processing, and particle production, we analyzed Gag proteins in the producer cell as well as in the virions by immunoblot. No major change in steady-state Gag level was noted except for a slight reduction with the V83Q capsid mutation (Fig. 3A). Western blot analysis of purified virions revealed aberrant processing for most P122Q-containing capsids, (Fig. 3B), with corresponding decreases in viral infectivity (Table 1). However, combining the P122Q mutation with changes at CA83 and CA120 completely restored processing and infectivity ( Table 1).
As the interaction between HIV1 capsid and CypA has been demonstrated to influence sensitivity to TRIM5␣-mediated restriction, we also assessed the incorporation of CypA into all produced HIV-1 capsid mutants. For comparison,  MARCH 11, 2011 • VOLUME 286 • NUMBER 10 each mutant was also produced in the presence of CsA, a drug that inhibits the capsid-CypA interaction. CypA could be detected in wild type HIV1 virions, whereas this incorporation was prevented by the presence of CsA during viral particle production or by an amino acid change at CA89 as expected. However, none of the single or combined mutations at positions CA83, CA120, and CA122 affected the incorporation of CypA into virions (Fig. 3B).

HIV1 Capsid Determinants to Escape TRIM5␣
HIV1 CA83, CA89, CA120, and CA122 Modulate Sensitivity to huTRIM5␣-We tested whether the identified HIV1 capsid positions could exert an influence on TRIM5␣-mediated restriction as observed for their MLV counterparts. For this, we used our panel of viruses to challenge MDTF cells stably expressing various TRIM5␣ variants that had distinct activities against HIV1 and MLV (Figs. 1A and 4A). In addition to wild type huTRIM5␣ and rhTRIM5␣, we included cells expressing the huTRIM5␣ Y336K mutant, which has an expanded restriction activity toward MLV strains yet remains inactive against HIV1 (66,81). Conversely we generated cells expressing huTRIM5␣ with various changes at position 332 (R332P, R332A, R332H), all of which have been demonstrated to confer restriction activity against HIV-1, whereas their spectrum of MLV restriction is limited to N-MLV (15,79,80). Finally, two TRIM5␣ alleles from New World monkeys, neither of which blocks N-MLV, were included: owl monkey TRIMCyp, which inhibits HIV1 through a CA-CypA interaction, and squirrel monkey TRIM5␣ (sqTRIM5␣), which has no effect on HIV1 (20,60,61,82).
As expected, HIV1 could escape wild type and Y336K huTRIM5␣ as well as sqTRIM5␣ but was susceptible to all Arg-332 huTRIM5␣ mutants, to rhTRIM5␣, and to TRIM-Cyp (Fig. 4A). Then, when a single point mutation was introduced at CA89 (G89V), HIV1 became partly susceptible to wild type and Y336K huTRIM5␣ (3-4-fold restriction, Fig.  4A and Table 2) but remained resistant to sqTRIM5␣ restriction. In contrast, this change in the HIV1 cyclophilin binding loop conferred either partial or complete protection toward rhTRIM5␣ and TRIMCyp restriction, respectively, as previously reported (54 -56, 60, 61). Interestingly, the Arg-332 huTRIM5␣ mutants blocked wild type and CA G89V HIV1 with the same efficiency (Fig. 4A, Table 2). The same was observed with an HIV1 CA mutant (P90A) similarly reported not to bind CypA (data not shown) or when using the drug CsA, a known competitive inhibitor of the CypA-CA interaction (Fig. 4B). In the presence of CsA, HIV1 became more sensitive to wild type (3-fold) and Y336K huTRIM5␣ (10-fold) yet remained equally sensitive to Arg-332 mutants (Fig. 4B).
In the case of MLV, the canonical 110 position of the capsid plays a primary role as a determinant of huTRIM5␣-mediated restriction (18,66,83). An arginine at the homologous position in HIV1, CA120, confers sensitivity to Y336K huTRIM5␣ but only has a slight effect on susceptibility to wild type huTRIM5␣ ( Fig. 5 and Table 2). Further sensitivity toward both forms of TRIM5␣ was observed when a change at CA120 was combined with either a mutation at CA122 and/or at CA89, whereas protection against sqTRIM5␣ was maintained for all HIV1 capsid mutants ( Fig. 5 and Table 2). These changes in HIV1 capsid led to a similar increase in restriction by rhTRIM5␣ and huTRIM5␣ Arg-332 mutants. In contrast, altering the amino acid at CA83 in the context of either a wild type or a mutated capsid allowed a slight escape of restriction from all TRIM5␣ variants (Table 2).
As mentioned above, a change of amino acid at CA89 disrupts CypA binding, but alternatively, the mutation per se may increase the sensitivity of HIV1 to TRIM5␣. We, therefore, tested, in the presence of CsA, single and combined mutations at CA83, CA120, and CA122 for their effect on TRIM5␣ susceptibility. CsA systematically increases the sensitivity of HIV1 mutants to both wild type and Y336K huTRIM5␣ (Fig. 5B). Yet, CsA treatment in some cases was less potent than the G89V mutation in increasing this susceptibility to TRIM5␣, suggesting a CypA-independent contribution when a mutation is performed at CA89 (Fig. 5B).

Aromatic Residues Located within the Cavity Surrounded by Key Residues Are Essential for Gag Processing and HIV1
Infectivity-This study together with previous reports demonstrates that HIV1 and MLV escape TRIM5␣ restriction activity by a combination of residues located at the edge of a cavity (19,66,84). The positions of these amino acids at the surface of the capsid likely offer possibilities for viral escape by allowing mutations without altering capsid stability significantly. Yet TRIM5␣ may target residues located deeper within the cavity that are essential for virus infectivity. Their crucial role in capsid stability would mean that any mutations would not  be tolerated by the virus, causing it to escape from TRIM5␣ by mounting obstacles in the way of this pocket through amino acid changes at the nearby positions CA83, CA89, CA120, CA122. Accordingly, three aromatic residues (phenylalanine 97, tryptophan 104, and tyrosine 118 in the MLV CA and tryptophan 133, tyrosine 130, and tryptophan 117 in the HIV1 CA) are located within this pocket and interact with each other (Fig. 6, A and B). These aromatic residues are conserved among the known and predicted structures of retroviral capsids that we analyzed (Fig. 6, B and C). To test their role in virus infectivity, we replaced these aromatic residues with alanine in both MLV and HIV1. All the resulting mutants were non-infectious in MDTF cells (Fig. 7A). Correlating this phenotype, Western blot analyses revealed particles with aberrantly processed Gag protein even though virion production itself was not dramatically altered (Fig. 7B).

DISCUSSION
TRIM5␣-mediated retroviral restriction proceeds through direct binding of its PRYSPRY domain to the capsid of incoming viruses. Amino acids at four positions (CA82, CA109, CA110, and CA117) of the MLV capsid were previously demonstrated to shelter some MLV strains from huTRIM5␣ restriction (66). Here, we first demonstrate that several TRIM5␣ variants or orthologues (wild type and R332A huTRIM5␣ as well as rhTRIM5␣), despite exhibiting different spectrums of anti-retroviral specificity, are all influenced in their action by the same four positions of the MLV capsid. In a second set of experiments, we demonstrate that structurally homologous positions on the HIV1 capsid similarly condition the recognition of this virus by the restriction factor. Together, these results attest to a high degree of conservation in the modalities of TRIM5␣-CA recognition.
Comparing the influence of MLV CA82, CA109, CA110, and CA117 on huTRIM5␣ and rhTRIM5␣ restriction revealed that changes at several of these positions were required for the virus to escape the human version of the antiviral, whereas single changes were sufficient to decrease or suppress blockade by its rhesus counterpart. It is likely that these differences reflect selective pressures exerted by lineage-specific retroviral infections during the course of primate evolution, as does the inverse pattern of activities of human and rhesus TRIM5␣ on N-MLV and HIV1 (85). Along the same line, a single change at arginine 332 in huTRIM5␣ increases its ability to block HIV1, but we found that this is at the cost of a decreased activity against MLV. Consistent with this result, an amino acid substitution at that position was previously demonstrated to induce a loss of activity of huTRIM5␣ against an extinct Gammaretrovirus derived from Pan troglodytes endogenous retrovirus (PtERV) as well as toward B-and Mo-MLV (66,86).
The present study also reveals striking similarities between the modes of recognition of both MLV and HIV1 capsids by TRIM5␣. Capitalizing on our previous observation that amino acids at specific positions of the MLV capsid can interfere with TRIM5␣-mediated blockade, we identified their structural homologues in the HIV capsid. Remarkably, we found that three of these HIV1 determinants (CA83, CA120, and CA122, corresponding to MLV CA82, CA110, and CA109, respectively) are major modulators of susceptibility to TRIM5␣. Interestingly, HIV1 CA122 had previously been designated as homologous to MLV CA110 based on primary sequences alignment of both capsids, yet structural homology between them revealed that a small hinge at position 120 of HIV1 CA orients proline 122 of this protein similarly to glutamine 109 of the MLV capsid (87).
Replacing proline 122 of HIV1 CA by glutamine led to a loss of infectivity coincident with abnormal Gag processing. Interestingly, these defects could be suppressed by further mutating CA120. These data are consistent with an important role for these residues in ensuring normal capsid folding and as a consequence its proper subjection to cleavage by the viral protease during maturation, as suggested within the context of similar processing defects (88).
The CA H120R mutation increased HIV1 susceptibility more to Y336K huTRIM5␣ than to its huTRIM5␣ wild type counterpart. Combining mutations at both CA120 and CA122 further increased HIV1 sensitivity toward both huTRIM5␣ variants. These two positions lie in a loop linking helices 6 and 7 of CA, consistent with previous studies pointing to a role for this loop in determining susceptibility to TRIM5␣mediated restriction. Replacing this loop on SIVagm with the corresponding loop of HIV1 yields a virus with increased sensitivity to rhesus and African Green monkey TRIM5␣ (89). As well, specific positions within this domain (CA116 and CA119) were found to influence HIV1 replication in various primate cells, although their impact on sensitivity to the corresponding TRIM5␣ orthologues was not specifically tested (84). Furthermore, residues present in the homologous loops based on a combination of residue identity/similarity and secondary structure correspondence (see "Experimental Procedures"). Residues located in ␣ helices or ␤ strands are framed with yellow-or green-filled rectangles, respectively. In the consensus sequence line, red capital letters correspond to residues that are identical in all sequences, whereas magenta capital letters correspond to residues that are highly conserved. In the sequence lines, colored residues correspond to highly conserved or even identical residues between sequences. These residues are colored according to their nature; red for positively charged residues, magenta for negatively charged residues, green for polar and neutral residues, blue for hydrophobic residues, and orange for glycine. The three aromatic residues that have similar positions after structure superimposition are highlighted; phenylalanine 97 in N-MLV with tryptophan 133 in HIV1 (framed in violet), tyrosine 118 in N-MLV with tyrosine 130 in HIV1 (framed in red), and tryptophan 104 in N-MLV with tryptophan 117 in HIV1 (framed in black). of other retroviral capsids were also shown to modulate susceptibility to TRIM5␣; capsid positions 119 and 120 in SIVmac and HIV-2, respectively, affect the sensitivity to both human and cynomolgus monkey TRIM5␣ proteins (90).
In contrast to mutations at CA120 and CA122, substitution of valine 83 by glutamine reduced the susceptibility of several HIV1 capsid derivatives to TRIM5␣-mediated restriction. As CA83 lies at the base of the CypA binding loop, it is tempting to speculate that it influences the orientation of this loop in a functionally relevant fashion.
Disruption of the CypA-CA interaction by the amino acid change G89V or by the use of the drug CsA augmented HIV1 sensitivity to both wild type and Y336K huTRIM5␣. This is consistent with previous studies performed in cells overexpressing huTRIM5␣ and with the observation that virions with capsids deprived of their interaction with CypA but not wild type capsids can abrogate N-MLV restriction in human cells (52,56,87). This seems at odds with the finding that HIV1 sensitivity to disrupting CypA recruitment is independent of huTRIM5␣ (55,57). TRIM5␣ overexpression, thus, appears to allow the detection of differential effects not apparent at physiological levels. Interestingly, CsA treatment had a greater impact on HIV susceptibility to Y336K huTRIM5␣ (10-fold) than to wild type huTRIM5␣ (3-fold). It has previously been demonstrated that Y336K mutation expands the spectrum of antiretroviral activity of huTRIM5␣ toward B-, Mo-MLV, HIV2, and SIVmac (66,81). In contrast, HIV1 and SIVagm fully escape this block, whereas FIV is only partially restricted (81). Interestingly, HIV1, SIV, and FIV are the only retroviruses whose capsids have been described to bind CypA (41,47,49,51,52). The mutation Y336K may induce a conformational change such that huTRIM5␣ docking with the capsid is now more sensitive to the presence of CypA.
When arginine 332 of huTRIM5␣ was replaced by proline, alanine, or histidine, CypA recruitment by the HIV1 capsid lost its protective effect, as illustrated by the HIV CA G89V mutant or infections performed in the presence of CsA. No structural data are available yet for the TRIM5␣ PRYSPRY domain, but extrapolating from the crystal structure of related proteins, it is predicted that position 332 of TRIM5␣ lies in an exposed loop that constitutes the so-called variable region 1 (91)(92)(93). Therefore, a mutation at arginine332 may modify the conformation of the variable region 1 (such that TRIM5␣ alleviates the interference caused by CypA binding and/or catalytic activity, which ultimately improves capsid recognition). Interestingly, consistent with a high degree of conservation in the modalities of TRIM5␣-CA recognition, sqTRIM5␣, which is inactive against N-MLV, did not exert significant restriction activity against an HIV1 derivative where critical positions of the capsid were substituted by their N-MLV homologues. Yet sqTRIM5␣-mediated restriction of SIVmac was shown to be dependent on residues located in the region homologous to the CypA binding loop of HIV1 (20). Therefore, further modifications of the HIV1 capsid may render it susceptible to this restriction factor. Reciprocally, mutating specific residues of the PRYSPRY or coiled-coil domains of sqTRIM5␣ may con- fer it with the ability to block N-MLV as well as wild type or mutant HIV1 (94).
The CA83, CA120, and CA122 positions modulate restriction by TRIM5␣, and the structurally homologous positions in MLV (CA82, CA109, and CA110) were previously demonstrated to influence sensitivity not only to TRIM5␣ but also to the mouse restriction factor Fv1. It seems unlikely that mutating these residues has global effects on CA structure, which would indirectly affect distantly located restriction factor binding sites. Rather, we favor a model in which the virus uses residues at these key positions to shelter the TRIM5␣ and Fv1 target sites, which themselves cannot be mutated without loss of capsid function. HIV1 CA83, -120, and -122 and MLV CA82, -109, and -110 are indeed situated on the edges of a pit, at the bottom of which reside three aromatic residues (phenylalanine 97, tryptophan 104, tyrosine 118 in the MLV CA and tryptophan 133, tyrosine 130, and tryptophan 117 in the HIV1 CA) that cannot be mutated without inducing aberrant Gag processing and loss of viral infectivity. These residues or others deeply situated in the pit might represent the TRIM5␣ recognition site, and viral escape is, thus, accordingly achieved by mounting obstacles to TRIM5␣ penetration into this pocket through changes at the more lenient yet critically placed CA82, CA109, CA110, and CA117 for MLV and, for HIV1, CA83, CA120, and CA122 acting in concert with the overhanging CypA binding loop. As such, resistance to TRIM5␣ is reminiscent of escape from neutralizing antibodies by glycosylation of the viral envelope, where sugar moieties hide critical protein epitopes (95,96). Whether this apparent Achilles' heel in the HIV1 capsid can be targeted by antiviral therapies warrants investigation.