Distinct Determinants in HIV-1 Vif and Human APOBEC3 Proteins Are Required for the Suppression of Diverse Host Anti-Viral Proteins

Background APOBEC3G (A3G) and related cytidine deaminases of the APOBEC3 family of proteins are potent inhibitors of many retroviruses, including HIV-1. Formation of infectious HIV-1 requires the suppression of multiple cytidine deaminases by Vif. HIV-1 Vif suppresses various APOBEC3 proteins through the common mechanism of recruiting the Cullin5-ElonginB-ElonginC E3 ubiquitin ligase to induce target protein polyubiquitination and proteasome-mediated degradation. The domains in Vif and various APOBEC3 proteins required for APOBEC3 recognition and degradation have not been fully characterized. Methods and Findings In the present study, we have demonstrated that the regions of APOBEC3F (A3F) that are required for its HIV-1-mediated binding and degradation are distinct from those reported for A3G. We found that the C-terminal cytidine deaminase domain (C-CDD) of A3F alone is sufficient for its interaction with HIV-1 Vif and its Vif-mediated degradation. We also observed that the domains of HIV-1 Vif that are uniquely required for its functional interaction with full-length A3F are also required for the degradation of the C-CDD of A3F; in contrast, those Vif domains that are uniquely required for functional interaction with A3G are not required for the degradation of the C-CDD of A3F. Interestingly, the HIV-1 Vif domains required for the degradation of A3F are also required for the degradation of A3C and A3DE. On the other hand, the Vif domains uniquely required for the degradation of A3G are dispensable for the degradation of cytidine deaminases A3C and A3DE. Conclusions Our data suggest that distinct regions of A3F and A3G are targeted by HIV-1 Vif molecules. However, HIV-1 Vif suppresses A3F, A3C, and A3DE through similar recognition determinants, which are conserved among Vif molecules from diverse HIV-1 strains. Mapping these determinants may be useful for the design of novel anti-HIV inhibitors.

In this study, we demonstrate that A3C and A3DE are recognized by HIV-1 Vif in a fashion similar to that observed for A3F, distinct from that seen for A3G. The carboxyl-terminal cytidine deamination domain of A3F alone is sufficient for its interaction with Vif and Vif-mediated degradation of A3F, and the requirements for the degradation of full-length A3F are the same as for its carboxyl-terminal cytidine deamination domain. Thus, the single cytidine deamination domain of A3C and carboxylterminal cytidine deamination domain of A3F are sufficient for Vif binding and targeted degradation, in sharp contrast to the requirement for both the amino-and the carboxyl-terminal cytidine deamination domains in the case of A3G.

Results
Distinct HIV-1 Vif regions are involved in the suppression of single-domain cytidine deaminase A3C and doubledomain A3G Distinct regions of HIV-1 Vif have been found to mediate A3G or A3F suppression (Fig. 1A); however, the regions of HIV-1 Vif that are involved in the suppression of other human cytidine deaminases such as A3C have not been determined. Although A3C has been shown to have only weak anti-HIV-1 activity in vitro [16,18], it is efficiently degraded by HIV-1 Vif through the usage of Cul5-ElonginB-ElonginC E3 ubiquitin ligase. It is possible that A3C has anti-HIV-1 function in vivo that has to be neutralized by Vif to allow viral replication. Alternatively, A3C is recognized by HIV-1 Vif through a similar mechanism as other potent anti-HIV-1 cytidine deaminases such as A3G or A3F. To determine whether the previously identified regions of the Vif protein that are required for A3G or A3F inhibition are also important for its activity against A3C, we generated a series of HIV-1 Vif mutant constructs in which critical residues known to be important for A3G or A3F suppression were mutated (Fig. 1A).
It is well established that A3C is a potent inhibitor [18] of Vifdeificient simian immunodeficiency virus from African green monkeys (SIVagm) and is degraded by both SIVagm and HIV-1 Vif [85]. We therefore examined the ability of HIV-1 Vif and the various Vif mutants to suppress the anti-viral activity of A3C against SIVagmgVif. HEK293T cells were transfected with SIVagmgVif and with an A3C expression vector plus a control vector, an expression vector for wild-type (WT) Vif, or a Vif mutant, as indicated in Fig. 1B. Viruses were produced from the transfected cells, and viral infectivity was tested in a standard Magi assay as previously described [85,86]. WT Vif suppressed A3C and maintained the infectivity of SIVagmgVif (Fig. 1B, column 2); this level of viral infectivity in the presence of WT Vif was considered to be 100% for comparison purposes. As expected, A3C dramatically reduced the infectivity of SIVagmgVif in the absence of Vif (Fig. 1B, column 1). The Vif DR14/15AA and VifW79A mutants, which have already been reported to be ineffective against A3F [79,81], were unable to efficiently suppress the anti-viral activity of A3C (Fig. 1B, columns 3 and 6). In contrast, the Vif K22E and VifRH41/ 42AA mutants were able to suppress the anti-viral activity of A3C (Fig. 1B, columns 4 and 5); as expected, these two mutant proteins were ineffective in suppressing the anti-viral activity of A3G (Fig. 1C, columns 4 and 5). Collectively, these results indicate that W79 and D14R15 of HIV-1 Vif are important for Vif-mediated degradation of A3C, its exclusion from virions, and the suppression of its anti-viral activity. However, these residues were not important for Vif-mediated degradation of A3G (  . Vif-myc proteins were immunoprecipitated from the cell lysates, and coprecipitation of A3C-HA was detected by immunoblotting. WT Vifmyc efficiently co-immunoprecipitated A3C-HA (Fig. 3B, lane 2); this interaction was specific, since A3C-HA was not detected in the absence of Vif (Fig. 3B, lane 1). Less A3C-HA was co-precipitated with VifDR14/15AA-myc (Fig. 3B, lane 3) and VifW79A-myc (Fig. 3B, lane 6) than with WT Vif-myc (Fig. 3B, lane 2), despite the fact that A3C-HA levels were higher in cells expressing VifDR14/ 15AA-myc (Fig. 3A, lane 3) and VifW79A-myc (Fig. 3A, lane 6) than in those expressing WT Vif-myc (Fig. 3A, lane 2). These data suggest that the impaired ability of the VifDR14/15AA and VifW79A mutants to degrade and suppress A3C is due at least in part to their reduced interaction with A3C. VifRH41/42AA-myc and VifK22Emyc were able to efficiently interact with A3C-HA (

Identification of the HIV-1 Vif regions required for A3DE suppression
A3DE also has anti-HIV-1 activity and is neutralized by HIV-1 Vif. We next examined the ability of WT Vif and Vif mutant molecules to affect A3DE expression (Fig. 4A).To this end, we transfected HEK293T with the A3DE expression vector plus a control vector (Fig. 4A, lane 1), the expression vector for WT Vifmyc (lane 2), or one of the Vif mutant molecules (lanes 3, 4, 5, and 6). The intracellular level of A3DE was efficiently reduced by WT Vif (Fig. 4A, lane 2) when compared to the vector control (Fig. 4A, lane 1). As compared to WT Vif, VifDR14/15AA and VifW79A were less effective in reducing A3DE expression (Fig. 4A, lanes 3 and 6). However, VifK22E and VifRH41/42AA preserved the ability to reduce A3DE expression (Fig. 4A, lanes 4 and 5).
As expected, VifDR14/15AA and VifW79A (Fig. 4B, lanes 3 and 6) were less effective than WT (Fig. 4B, lane 2) in excluding A3DE from HIV-1 virions. Mutant VifDR14/15AA and VifW79A also showed a reduced ability to neutralize the anti-viral activity of A3DE (Fig. 4C, columns 3 and 6) when compared to the WT Vif (Fig. 4C,  column 2). Mutant VifK22E and VifRH41/42AA were as effective as WT Vif (Fig. 4B) in excluding A3DE from virions, and they could efficiently counteract the anti-viral activity of A3DE (Fig. 4C). These data indicate that the residues of HIV-1 Vif that are important for A3F and A3C suppression are also important for A3DE suppression. Conversely, the residues of HIV-1 Vif that are important for A3G suppression are dispensable for A3DE suppression.
We also evaluated the interaction of WT and Vif mutant molecules with A3DE by co-immunoprecipitation analysis. HEK293T cells were transfected with an A3DE expression vector plus a control vector (Fig. 5A, lane 1) or the expression vector for   The myc-tagged Vif proteins were immunoprecipitated from cell lysates and the co-precipitation of A3DE was assessed by immunoblotting. A3DE-HA was efficiently co-immunoprecipitated with WT Vif-myc (Fig. 5B, lane 2), and this interaction was specific, since A3DE-HA was not precipitated in the absence of Vif (Fig. 5B, lane 1). Even though higher levels of A3DE were detected in cells expressing VifDR14/15AA (Fig. 5B, lane 3) and VifW79A (Fig. 5B, lane 6) than in those expressing WT Vif, significantly less A3DE was immunoprecipitated with VifDR14/15AA and VifW79A than with WT Vif. When intracellular level of A3DE was considered, VifDR14/15AA and VifW79A had 5-10 fold reduced ability to interact with A3DE compared to the WT Vif. Thus, the impaired ability of VifDR14/15AA and VifW79A to degrade A3DE could be attributed, at least in part, to their reduced recognition of A3DE.
The carboxyl-terminal deamination domain of A3F behaves like the full-length A3F in terms of Vif sensitivity HIV-1 Vif utilizes similar regions to interact with A3C, A3DE, and A3F. It is interesting to note that HIV-1 Vif suppresses the single-domain cytidine deaminase A3C and the double-domain enzyme A3F through similar means. This result raises the question of whether HIV-1 Vif recognizes the amino-or carboxyl-terminal domain of A3F. Alignment analysis of A3C and A3F showed that there is a strikingly higher degree of homology (77%) between A3C and the carboxyl-terminal domain of A3F (Fig. 6A) than with the amino-terminal domain of A3F (47%, Fig. 6B).
To determine whether the carboxyl-terminal domain of A3F alone can interact with HIV-1 Vif, we constructed an expression vector for the HA epitope-tagged carboxyl-terminal domain of A3F (amino acids 190-373). HEK293T cells were co-transfected with expression vectors for HIV-1 Vif-myc plus a control vector or either full-length A3F-HA or the carboxyl terminal A3F-C-HA. The A3F-HA and A3F-C-HA proteins were immunoprecipitated from transfected cell lysates with anti-HA antibody conjugated to agarose beads. Both full-length A3F-HA and A3F-C-HA (Fig. 6C, lanes 5 and 6) efficiently co-immunoprecipitated Vif-myc. Vif-myc was not co-immunoprecipitated in the absence of any A3F proteins (Fig. 6B,  lane 4), indicating the specificity of the assay system.
Since the carboxyl-terminal domain of A3F alone could interact with HIV-1 Vif, we asked whether the A3F-C alone could be degraded by HIV-1 Vif. HEK293T cells were transfected with an expression vector for A3F-C-V5 plus a control vector (Fig. 6D, lane  1), the expression vector for WT Vif-myc (lane 2), or one of various Vif mutants, as indicated (lane 3, 4, 5, and 6). The intracellular level of A3F-C-V5 was efficiently reduced by WT Vif (Fig. 6D, lane 2) when compared to the vector control (Fig. 6D, lane 1). However, VifDR14/15AA and VifW79A showed an impaired ability to reduce A3F-C-V5 expression (Fig. 6D, lanes 3 and 6) when compared to WT Vif. VifK22E and VifRH41/42AA retained their ability to degrade A3F-C-V5 (Fig. 6D, lanes 4 and 5). These data indicate that the requirement for the degradation of full-length A3F is the same as for the degradation of the carboxyl-terminal domain of A3F.

Discussion
Human APOBEC3 cytidine deaminases are either singledomain or double-domain proteins. Regions in the double-domain A3G protein that are important for HIV-1 Vif-mediated degradation have been shown to span both the amino-and the carboxyl-terminal domains of A3G. Conticello et al. has demonstrated that amino acids 54-124 of A3G alone can interact with HIV-1 Vif [68]. Residues D128 and D130 were later reported to be important for HIV-1 Vif binding as well [87][88][89][90][91]. More recently, Zhang et al. have demonstrated that although amino acids 1-156 of A3G are sufficient for the interaction with HIV-1 Vif, additional regions spanning amino acids 105-245 are required for HIV-1 Vif-mediated polyubiquitination and degradation [85]. Thus, the amino-terminal domain of A3G mediates its interaction with HIV-1 Vif, but both cytidine deamination domains of A3G are involved in Vif-mediated degradation.
We have now found that, unlike the case for A3G, the carboxylterminal domain alone of another double-domain protein, A3F, is sufficient for the interaction with HIV-1 Vif, and, more importantly, is all that is required for A3F to undergo Vifmediated degradation. These new data regarding the degradation of A3F are consistent with our previous observations that the carboxyl-but not the amino-terminal domain of A3F is important for its functional interaction with HIV-1 Vif [71,79]. Mutation of the aspartate residue, D128, in the amino-terminal domain of A3G has been shown to influence its recognition by HIV-1 Vif [87][88][89]. The analogous residue in A3F, E127, is not important for HIV-1 Vif binding [71], and modifications of this amino acid in human A3F do not change its recognition by HIV-1 Vif or SIVagm Vif [71]. Furthermore, while C-terminal tag modifications of A3F (HA-tag vs V5-tag) can significantly influence its ability to be degraded by HIV-1 Vif, the same modifications of A3G do not affect its ability to be recognized by HIV-1 Vif [79].
Here we also provide evidence that Vif-mediated degradation of the carboxyl-terminal domain of A3F is similar to that of full-length A3F. We and others have identified unique regions in HIV-1 Vif that are critical for A3G, but not A3F, degradation and vice versa [13,67,[78][79][80][81][82][83]. For example, W11, D14, R15, and W79 of HIV-1 Vif are required for A3F, but not A3G, binding and degradation. On the other hand, K22, R41, and H42 of HIV-1 Vif are required for A3G but not A3F binding and degradation. Residues such as D14R15 and W79 in HIV-1 Vif that are important for full-length A3F binding and degradation are also important for Vif-mediated degradation of the carboxyl-terminal domain of A3F. On the other hand, residues such as K22 and R41H42 that are important for the Vif-mediated degradation of A3G are dispensable for Vif-mediated degradation of both full-length A3F and the carboxyl-terminal domain of A3F. These data indicate the carboxyl-terminal domain of A3F is the main target of HIV-1 Vif against A3F.
Our data also demonstrate that the N-terminal region of HIV-1 Vif mediates its binding not only to the target molecules A3G and A3F but also to A3C and A3DE. Interestingly, we found that the single-domain cytidine deaminase A3C is also recognized and degraded by HIV-1 Vif through a mechanism similar to that for A3F. The residues in HIV-1 Vif that are important for A3F interaction and degradation were also important for Vif-mediated degradation of A3C. In contrast, those that are mainly important for the Vif-mediated degradation of A3G were found to be dispensable for Vif-mediated degradation of A3C. A3C has a high degree of amino acid homology to the carboxyl-terminal domain of A3F. However, unlike A3F, A3C has only weak anti-HIV-1 function [16,18]. It is not clear whether HIV-1 Vif has evolved to disable A3F, and its ability to suppress A3C is merely an incidental consequence of the high degree of homology between A3C and the carboxyl-terminal domain of A3F. Whether A3C has anti-HIV-1 activity in certain HIV-1 natural target cells in vivo is an open question.
Consistent with a previous report [72], we found that A3DE has substantial anti-HIV-1 activity (Fig. 4). This protein is expressed in peripheral blood mononuclear cells [72] and macrophages (data not shown). Sequence alignment of A3DE and A3F shows a high degree of homology between these two molecules, and HIV-1 Vif residues required for A3F binding and suppression are also essential for A3DE binding and suppression (Figs. 4 and 5).
Collectively, our data suggest that Vif employs distinct protein interfaces to recognize various human APOBEC3 proteins (Fig. 7). HIV-1 Vif recognizes A3F, A3C, and A3DE through a similar mechanism that is distinct from that for A3G (Fig. 7). The W 11 xxDRMR 17 and T 74 GERxW 79 motifs of HIV-1 Vif are dispensable for the suppression of the potent anti-HIV-1 cytidine deaminase A3G yet are highly conserved among diverse HIV-1 strains. It is an open question whether these motifs have been conserved only for the suppression of A3F or whether they have been evolutionarily selected to target other human cytidine deaminases, such A3C and A3DE, as well. The highly conserved nature of these residues in diverse HIV-1 Vif molecules indicates that A3F, A3C, and/or A3DE represent a selection force against HIV-1 in vivo. This argument would be consistent with the in vivo observation of HIV-1 G-to-A mutation patterns. GA-to-AA mutations (a pattern generated by A3F, A3C, and/or A3DE) are frequently detected in viral sequences recovered from HIV-1-infected individuals [20]. The highly conserved residues in Vif that are required for the suppression of multiple human cellular anti-HIV-1 factors represent another potential drug target against HIV-1.

Plasmid construction
The infectious molecular clone of the Vif mutant pNL4-3DVif construct was obtained from the AIDS Research Reagents Program, Division of AIDS, National Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH).
Cell culture, transfection, viral infectivity (MAGI) assays, APOBEC3 degradation, and antibodies HEK293T and MAGI-CCR5 (AIDS Research Reagents Program) cells were maintained and transfected or infected as previously described [62]. Transfection was performed with Lipofectamine2000 (Invitrogen) as instructed by the manufacturer. Viral infectivity (MAGI assays) were performed as described [62]:Virus was produced by transfecting HEK293T cells in a sixwell plate with 1 mg of NL4-3gVif or SIVagmTanDVif, 1 mg of wild-type or mutant Vif, and 0.3 mg of APOBEC3 as indicated. Virus was harvested from the supernatant for viral infectivity assays, and cell lysates were prepared for immunoblotting. Infectivity was assessed at 48 h post-infection and normalized to the input CAp24 or CAp27. The antibodies used in this study have been previously described [84]: anti-HA antibody, anti-myc antibody, anti-Vif (AIDS Research Reagents Program), and antihuman ribosomal P antigens. The anti-Vif antibody was obtained from the AIDS Research Reagents Program (2221), the mouse anti-V5 antibody from Invitrogen (R96025), and the mouse antitubulin antibody from Covance (MMS-410P).

Immunoprecipitation and immunoblot analysis
A T-25 flask of HEK293T cells was transfected with 3 mg APOBEC3 plus 3 mg of wild-type or mutant Vif expression vectors as indicated. Cells were harvested, washed twice with cold PBS and lysed in lysis buffer (50 mM Tris-HCl [pH 7.5] with 150 mM NaCl, 1% [v/v] Triton X-100, and complete protease inhibitor cocktail tablets) at 4uC for 1 h, then centrifuged at 10,000g for 30 min. For myc-tag immunoprecipitation, pre-cleared cell lysates were mixed with anti-myc antibody (Upstate) and incubated with protein G beads at 4uC for 3 h. For HA tag immunoprecipitation, pre-cleared cell lysates were mixed with anti-HA antibodyconjugated agarose beads (Roche) and incubated at 4uC for 3 h. Samples were then washed three times with washing buffer (20 mM Tris-HCl [pH 7.5], with 100 mM NaCl, 0.1 mM EDTA, and 0.05% [v/v] Tween-20). The beads were eluted with elution buffer (0.1 M glycine-HCl, pH 2.0), and 26 loading buffer was added. The eluted materials were then analyzed by SDS-PAGE and immunoblotting with the anti-myc antibody or the anti-HA antibody as described.