ABSTRACT
Vif counteracts the host restriction factor APOBEC3G (A3G) and other APOBEC3s by preventing the incorporation of A3G into progeny virions. We previously identified Vif mutants with a dominant-negative (D/N) phenotype that interfered with the function of wild-type Vif, inhibited the degradation of A3G, and reduced the infectivity of viral particles by increased packaging of A3G. However, the mechanism of interference remained unclear, in particular since all D/N Vif mutants were unable to bind Cul5 and some mutants additionally failed to bind A3G, ruling out competitive binding to A3G or the E3 ubiquitin ligase complex as the sole mechanism. The goal of the current study was to revisit the mechanism of D/N interference by Vif mutants and analyze the possible involvement of core binding factor beta (CBFβ) in this process. We found a clear correlation of D/N properties of Vif mutants with their ability to engage CBFβ. Only mutants that retained the ability to bind CBFβ exhibited the D/N phenotype. Competition studies revealed that D/N Vif mutants directly interfered with the association of CBFβ and wild-type Vif. Furthermore, overexpression of CBFβ counteracted the interference of D/N Vif mutants with A3G degradation by wild-type Vif. Finally, overexpression of Runx1 mimicked the effect of D/N Vif mutants and inhibited the degradation of A3G by wild-type Vif. Taken together, we identified CBFβ as the key player involved in D/N interference by Vif.
IMPORTANCE Of all the accessory proteins encoded by HIV-1 and other primate lentiviruses, Vif has arguably the strongest potential as a target for antiviral therapy. This conclusion is based on the observation that replication of HIV-1 in vivo is critically dependent on Vif. Thus, inhibiting the function of Vif via small-molecule inhibitors or other approaches has significant therapeutic potential. We previously identified dominant-negative (D/N) Vif variants whose expression interferes with the function of virus-encoded wild-type Vif. We now show that D/N interference involves competitive binding of D/N Vif variants to the transcriptional cofactor core binding factor beta (CBFβ), which is expressed in cells in limiting quantities. Overexpression of CBFβ neutralized the D/N phenotype of Vif. In contrast, overexpression of Runx1, a cellular binding partner of CBFβ, phenocopied the D/N Vif phenotype by sequestering endogenous CBFβ. Thus, our results provide proof of principle that D/N Vif variants could have therapeutic potential.
INTRODUCTION
The viral infectivity factor (Vif) is an accessory protein encoded by all primate lentiviruses except equine infectious anemia virus (EIAV). Vif is critical for the production of infectious virus by counteracting host restriction factors of the APOBEC3 family of cytidine deaminases (for reviews, see references 1 and 2). Vif interacts with cellular proteins Cul5, Elongin B (ELOB) and C, CBFβ, and Rbx (3–8). In the absence of Vif, cytidine deamination of single-stranded viral cDNA by APOBEC3G (A3G) renders the virus noninfectious (reviewed in reference 9). While natural Vif variants can differ in their ability to target A3G, A3F, or A3H (10, 11), there are no known primary replication competent viruses that completely lack expression of Vif. This suggests that vif-defective viruses are replication incompetent in vivo (12). Therefore, Vif offers itself as a target for antiviral therapy. Yet, there are currently no drugs in clinical use that specifically target Vif even though several small-molecule compounds with the potential to disrupt Vif function have been reported. These include VEC-5, which inhibits Vif function by preventing the interaction of Vif with Elongin C (13) and the zinc chelator TPEN [(N,N,N′,N′-tetrakis(2-pyridinylmethyl)-1,2-ethanediamine] that affects A3G degradation by inhibiting Cullin5 recruitment to Vif (14). In addition, high-throughput analysis identified RN-18 as a small molecule affecting the stability of Vif (15), as well as MM-1 and MM-2, two small molecules with similar backbone structure, both of which inhibit degradation of A3G (16). Another small-molecule inhibitor, N.41, was found to inhibit the Vif-A3G interaction, resulting in increased packaging of A3G and inhibition of viral infectivity (17). Yet another small-molecule inhibitor, IMB-301, was found to inhibit HIV-1 replication by binding to A3G and inhibiting Vif-mediated A3G degradation (18). Finally, screening for host factors with potential antiviral properties identified the mitogen-activated protein kinase ASK1, which was reported to inhibit A3G degradation by binding to the BC-box of Vif, thereby disrupting the assembly of the Vif-ubiquitin ligase complex (19).
We previously reported that Vif mutants with a dominant-negative (D/N) phenotype could inhibit the ability of wild-type Vif to induce degradation of A3G, presumably by competitive binding to factor(s) required for this process (20). In that context, we defined the following four groups of functionally defective Vif mutants: group 1 mutants included N-terminal deletion mutants that affected the ability of Vif to bind A3G, group 2 mutants contained mutations in the central region of Vif that also rendered them unable to bind A3G, group 3 mutants retained the ability to bind A3G but were thought to lack the ability to bind Cul5 and to assemble an E3-ubiquitin ligase complex, and, finally, group 4 mutants encompassed mutations in the conserved C-terminal PPLP motif. These mutants had lost the ability to bind A3G (20). Somewhat unexpectedly, only members of groups 3 and 4 had D/N properties and interfered with the function of wild-type Vif. This resulted in reduced A3G degradation in virus-producing cells and increased packaging of A3G into wild type HIV-1 virions (20). In contrast, coexpression of group 1 and group 2 Vif mutants, although functionally defective, had little or no effect on the function of wild-type Vif (20).
Vif-induced degradation of A3G involves the assembly of an E3-ubiquitin ligase complex on Vif that results in the ubiquitination of A3G bound to Vif (21–27). Recent studies showed that cofactor core binding factor beta (CBFβ) is a critical cofactor for the Vif-induced A3G degradation (4–6), and interaction of Vif and CBFβ is important for the formation and stabilization of the Cul5-E3-ubiquitin ligase complex (7, 28–30). In the absence of CBFβ, antagonism of A3G antiviral activity by Vif is impaired (4, 5). Subsequent studies revealed that CBFβ assumes a chaperone-like function to stabilize Vif and to facilitate the assembly of a Cul5-E3 ubiquitin ligase complex on Vif (29–37).
Our previous study on the D/N properties of certain Vif mutants was done prior to the identification of the role of CBFβ in Vif-dependent A3G degradation. Therefore, the goal of the current study was to revisit the issue of D/N interference by Vif mutants and to investigate the possible involvement of CBFβ in this process. Indeed, we found a correlation between D/N properties of Vif mutants and their ability to engage CBFβ. Only Vif mutants able to bind CBFβ exhibited D/N phenotypes. Of note, all of the Vif mutants analyzed, including those lacking D/N properties, had lost their ability to engage Cul5, ruling out competitive Cul5 binding as a possible mechanism underlying the D/N phenotype. Indeed, competition studies revealed that D/N Vif mutants directly interfered with the association of CBFβ with wild-type Vif. Furthermore, overexpression of CBFβ counteracted the interference of D/N Vif mutants with A3G degradation by wild-type Vif. Finally, overexpression of Runx1, known to bind CBFβ to form a functional transcription complex, mimicked the effect of D/N Vif mutants and inhibited the degradation of A3G by wild-type Vif. Taken together, we identified the molecular mechanism of D/N interference by Vif and identified CBFβ as the key player involved in this process.
RESULTS
The dominant-negative phenotype of Vif mutants is not defined by competitive binding to A3G, Cul5, or ELOB.We first tested the ability of Vif mutants to associate with components of the Cul5-E3-ubiquitin ligase complex. We chose four representative Vif mutants, one from each of the four groups of mutants defined in our previous study (20). We chose Vif-ΔK (group 1) and Vif-ΔH (group 2) as representative mutants that do not exhibit dominant-negative (D/N) effects. Vif-ΔC (group 3) and Vif-ΔN (group 4) were chosen as representatives of mutants with D/N properties. All four mutants carry in-frame deletions, as indicated in Fig. 1A. We confirmed that all four Vif mutants are expressed (Fig. 1B, Vif) but are inherently unable to induce A3G degradation (Fig. 1B, A3G). Of the four Vif mutants, only Vif-ΔC retained the ability to interact with A3G (Fig. 1B, bottom panel, lane 6). The amounts of wild-type Vif coimmunoprecipitating with A3G-myc (Fig. 1B, lane 3) were small compared with Vif-ΔC, which is presumably due to the Vif-induced degradation of A3G in the experiment.
The dominant-negative phenotype of Vif mutants is not defined by competitive binding to A3G, Cul5, or ELOB. (A) Vif mutants used in this study are schematically shown. The positioning of the deletions is roughly to scale. Numbers indicate amino acids deleted. All mutants are cloned in the backbone of partially a codon-optimized pcDNA-hVif vector (44). To C-terminally express myc-tagged Vif, the Vif stop codon in pcDNA-hVif was removed by site-directed mutagenesis to extend the coding sequence into the vector-encoded C-terminal epitope tag. Constructs are numbered according to their previous group assignment (20). (B) 293T cells were transfected with 2.5 μg each of the indicated Vif constructs together with 0.5 μg of pcDNA-A3G-myc expression vector (lanes 3 to 7). Lane 2 shows A3G-myc in the absence of Vif. Lane 1 is a mock control. In all cases, total amounts of transfected DNA were adjusted to 5 μg with empty vector DNA. Cells were collected 24 h after transfection, and detergent extracts were processed for coimmunoprecipitation analysis as described in Materials and Methods. A fraction from each extract was used as input control. The remaining samples were immunoprecipitated using anti-c-myc agarose affinity gel as described in Materials and Methods. Precipitated samples were separated by SDS-PAGE and transferred to PVDF membranes. Membranes were probed sequentially with antibodies to A3G and Vif. (C) 293T cells were transfected with 2.5 μg each of the indicated myc-tagged Vif constructs (lanes 2 to 6) together with 2.5 μg of empty vector. Cells transfected with 5 μg of empty vector DNA served as a mock control (lane 1). Detergent extracts were prepared 24 h later. A portion of the extracts was used directly as the input control and probed sequentially with the indicated antibodies (input). Tubulin served as a loading control. The remainder of each extract was immunoprecipitated with an anti-c-myc agarose affinity gel. Precipitated samples were separated by SDS-PAGE, and the membrane was probed sequentially with antibodies to Cul5, ELOB, or Vif. Proteins are identified on the right.
Next, we tested the ability of our Vif variants to interact with components of the Cul5-E3-ubiquitin ligase complex. Based on previous reports defining the zinc-finger domain in Vif as the Cul5 binding domain (38–41), we speculated that both Vif-ΔC and Vif-ΔN would retain their ability to interact with Cul5. To test this, we employed C-terminally myc-tagged Vif variants as bait for the pulldown of endogenously expressed E3-ubiquitin ligase components. Surprisingly, only wild-type Vif was able to pull down Cul5 (Fig. 1C, lane 2) (immunoprecipitated via the myc epitope tag [IP-myc]). None of the Vif variants, including Vif-ΔK carrying a deletion at the very N terminus of Vif far away from the reported Cul5 binding domain (see Fig. 1A), was able to coimmunoprecipitate endogenous Cul5 (Fig. 1C, lanes 3 to 6). In contrast, Vif-ΔK and Vif-ΔN, as well as wild-type Vif, solidly interacted with ELOB (Fig. 1C, lanes 2, 3, and 6), while Vif-ΔH only weakly interacted with ELOB (Fig. 1D, lane 4) despite the presence of the ELOB/C binding site (Fig. 1A, BC-box). As expected, Vif-ΔC, which lacks the BC box (Fig. 1A), interacted poorly with ELOB (Fig. 1C, lane 5). The ability of Vif-ΔC to interact with A3G (Fig. 1B) and the observed interaction of Vif-ΔN with ELOB (Fig. 1C) could suggest that these mutants exert their D/N properties through competitive binding of A3G and ELOB, respectively. Importantly, Vif-ΔK and Vif-ΔN both fail to interact with A3G and Cul5 but retain the ability to interact with ELOB. Yet, only Vif-ΔN exhibits D/N properties. These results, therefore, suggest that there is at least one additional Cul5- and A3G-independent mechanism involved in the dominant-negative interference of Vif mutants with the function of wild-type Vif.
The dominant-negative phenotype of Vif mutants correlates with the ability to bind CBFβ.Core binding factor beta (CBFβ) was recently identified as an essential cofactor for Vif-induced A3G degradation (4–7, 28). Indeed, the expression of CBFβ was found to both stabilize the expression of Vif and, in addition, increase the ability of Vif to target A3G (31, 36). The effect of CBFβ on Vif stabilization is mediated by a chaperone-like effect that involves a direct protein-protein interaction with the N terminus of the Vif protein during translation (36). This led us to speculate about a possible involvement of CBFβ in the dominant-negative inhibition of Vif by group 3 and 4 Vif mutants.
We first tested the interaction of our Vif mutants with endogenous CBFβ. Since our Vif-specific antibody reacts with all Vif mutants used here (see Fig. 1B), we employed untagged Vif to avoid possible interference of epitope tags with CBFβ interaction. In addition to CBFβ, Cul5, and ELOB interactions were assessed as well (Fig. 2). The expression of Vif variants, as well as endogenous Cul5, ELOB, and CBFβ, is shown in the top section of Fig. 2 (input). Coimmunoprecipitation of cellular factors with Vif is shown in the bottom part of the figure (Fig. 2, IP-Vif). Consistent with the results from Fig. 1C (which involved myc-tagged Vif variants), the pulldown analysis by untagged Vif revealed the same interaction pattern for Cul5 and ELOB (Fig. 2, bottom). Only wild-type Vif was able to interact with Cul5; furthermore, wild-type Vif, Vif-ΔK, Vif-ΔN, and (weakly) Vif-ΔH, but not Vif-ΔC, interacted with ELOB. Importantly, both mutants exhibiting the D/N phenotype (i.e., Vif-ΔC and Vif-ΔN) interacted with CBFβ (Fig. 2, bottom, lanes 5 and 6), whereas Vif-ΔK and Vif-ΔH failed to bind CBFβ (Fig. 2, bottom, lanes 3 and 4). The observation that VifΔC can interact with CBFβ in the absence of ELOB interaction contrast a previous observation by Wang et al. who found that ELOB is critical for the interaction between Vif and CBFβ (32). These authors employed a myc-tagged SLQ-AAA Vif mutant rather than our five-residue deletion of the SLQ motif in untagged Vif. Our own analysis of a myc-tagged SLQ-AAA mutant revealed a lack of interaction with ELOB without a loss of interaction with CBFβ (data not shown). These results indicate that the Vif/CBFβ interaction is not critically dependent on ELOB interaction.
The dominant-negative phenotype of Vif mutants correlates with the ability to bind CBFβ. 293T cells were transfected with 2.5 μg each of the indicated untagged Vif constructs (lanes 2 to 6) together with 2.5 μg of empty vector. As a mock control, cells were transfected with 5 μg of empty vector (lane 1). Cells were collected 24 h later and processed for immunoprecipitation as in Fig. 1B, except that samples were immunoprecipitated with a rabbit polyclonal antibody to Vif (Vif93). Input controls and immunoprecipitated samples were subjected to immunoblot analyses to probe for coimmunoprecipitation (co-IP) of Cul5, ELOB, or CBFβ. Input samples were also probed for tubulin to control for sample loading.
Vif mutants with the D/N phenotype compete with wild-type Vif for CBFβ binding.CBFβ is a cofactor of the transcription factor Runx1. It promotes the nuclear transport of Runx1 and increases the affinity of Runx1 to DNA (reviewed in reference 42). We previously reported that the expression of Vif can lead to inhibition of Runx1-mediated gene expression by sequestering CBFβ in the cytoplasm (43). These results imply that levels of endogenous CBFβ are limiting and suggest that the D/N effect of group 3 and 4 Vif mutants may be exerted by the adsorption of CBFβ to D/N Vif, thus limiting the amounts of CBFβ available for binding to wild-type Vif. Indeed, an analysis of endogenous levels of CBFβ indicates that most cell types tested, including peripheral blood mononuclear cells (PBMC), express levels of CBFβ that are similar to or lower than those observer in HEK293T cells studied here (data not shown).
To directly assess the possible competition of D/N Vif mutants for binding to CBFβ by wild-type Vif, constant amounts of myc-tagged wild-type Vif were coexpressed with untagged Vif mutants (Fig. 3A, input). Cell extracts were immunoprecipitated with anti-myc antibody (Fig. 3A, IP-myc) and binding of endogenous CBFβ to wild-type (myc-tagged) Vif in the presence or absence of untagged Vif variants was studied. As shown in Fig. 3A (input), all samples expressed comparable amounts of wild-type Vif (Vif-myc) as well as constant amounts of endogenous CBFβ. Expression of untagged Vif mutants was confirmed as well. Immunoprecipitation with myc-specific antibody yielded comparable amounts of myc-tagged wild-type Vif in all reactions. Interestingly, levels of coimmunoprecipitated CBFβ were reduced in samples expressing untagged wild-type Vif as well as Vif-ΔC and Vif-ΔN (Fig. 3A, bottom; lanes 3, 6, and 7) but not in samples expressing Vif-ΔK or Vif-ΔH (Fig. 3A, bottom; lanes 4 and 5). Indeed, quantitation of the results from five independent experiments demonstrates statistical significance for the interference of D/N Vif mutants with the interaction of CBFβ and wild-type Vif (Fig. 3B). These results indicate that levels of endogenous CBFβ are indeed limiting and that coexpression of CBFβ binding-competent Vif mutants effectively limits the levels of CBFβ available for binding to functional wild-type Vif.
Vif mutants with dominant-negative phenotype compete with wild-type Vif for CBFβ binding. (A) 293T cells were transfected with 2.5 μg of myc-tagged wild-type Vif (lanes 2 to 7) together with 2.5 μg of empty vector (lane 2) or 2.5 μg each of the indicated untagged Vif vectors (lanes 3 to 7). Cells transfected with 5 μg of empty vector served as the mock control (lane 1). Cells were harvested 24 h after transfection and processed for IP Western as in Fig. 1B. (B) Band intensities were measured in the CBFβ and Vif blots from the IP-myc samples to quantify the amounts of CBFβ bound to WT Vif-myc. The signal for CBFβ in the control lane (lane 2) served as a reference and was defined as 1. Error bars represent the SEM calculated from five independent experiments. Statistical significance was determined using a one-tailed, paired Student’s t test versus the control experiments. **, P < 0.01; n.s., not significant.
Overexpression of Vif mutants with the D/N phenotype inhibits the expression of provirus-expressed wild-type Vif.We previously reported that CBFβ exerts a chaperone-like activity that results in the cotranslational stabilization of Vif during de novo protein biosynthesis by physical interaction with nascent Vif protein (36). Thus, we hypothesized that if endogenous CBFβ levels are indeed limiting, expression of D/N Vif mutants should affect the expression and/or stability of wild-type Vif. The experiment shown in Fig. 3 did not reveal an effect of D/N Vif expression on the expression of wild-type Vif-myc, possibly because all proteins were expressed from a strong cytomegalovirus (CMV) promoter. We, therefore, tested the effect of D/N Vif variants on the expression of Vif in the context of the full-length virus NL4-3. For that purpose, 293T cells were transfected with the full-length molecular clone pNL4-3 together with a vector encoding A3G, as well as partially codon-optimized vectors for the overexpression of myc-tagged Vif variants (Fig. 4A, lanes 5 to 9). Of note, levels of Vif-myc expressed from our partially codon-optimized vectors (per microgram transfected plasmid DNA) are comparable with those expressed from pNL4-3 (44). As a control for the nonspecific secretion of A3G, A3G was transfected in the absence of proviral DNA (Fig. 4A, lane 2). Other controls included mock-transfected cells (Fig. 4A, lane 1), as well as Vif-defective (lane 3) or wild-type (lane 4) pNL4-3 in the presence of A3G but absence of competitor Vif. Cells and virus-containing supernatants were harvested 48 h after transfection and processed for immunoblot analysis as described in Materials and Methods. Digital images from multiple replicate experiments were quantified as described in Materials and Methods and used to calculate the relative levels of untagged Vif expressed from pNL4-3 (Fig. 4A, wild-type [WT] Vif; Fig. 4B) as well as cellular and virus-associated A3G (Fig. 4C and D, respectively). We confirmed that A3G was not secreted nonspecifically into the culture supernatants in the absence of virus production (Fig. 4A, lane 2) and that A3G was not degraded and was packaged efficiently in the absence of Vif (Fig. 4A, lane 3). As expected, the expression of Vif from pNL4-3 reduced cellular levels of A3G and inhibited A3G packaging (Fig. 4A, C, and D, lane 4). Expression of additional wild-type myc-tagged Vif from the codon-optimized vector (Fig. 4A, lane 5) did not further reduce cellular levels of A3G (Fig. 4C, lane 5), indicating that the levels of virus-encoded Vif are sufficient to fully control A3G in this experiment. However, coexpression of wild-type myc-tagged Vif resulted in reduced levels of Vif expressed from the proviral DNA (Fig. 4B, lane 5). Coexpression of Vif-ΔK and Vif-ΔH, both of which lack D/N properties, did not interfere with the reduction of cellular A3G or A3G packaging by NL4-3-encoded Vif (Fig. 4A, C, and D, lanes 6 and 7), nor did it affect the expression of NL4-3-encoded wild-type Vif (Fig. 4B, compare lanes 6 and 7 with lane 4). In contrast, coexpression of Vif-ΔC or Vif-ΔN with NL4-3-encoded wild-type Vif reduced the expression of NL4-3-encoded Vif (Fig. 4B, compare lanes 8 and 9 to lane 4) and, at the same time, interfered with the Vif-induced degradation and packaging of A3G (Fig. 4C and D, lanes 8 and 9). The results from this experiment indicate that CBFβ binding-competent Vif variants can indeed inhibit the expression of virus-encoded wild-type Vif and interfere with its ability to target A3G.
Overexpression of Vif mutants with D/N phenotype inhibits expression of provirus-expressed wild-type Vif. (A) 293T cells were transfected with 3 μg of pNL4-3/vif (−) together with 0.5 μg of pcDNA-A3G (lane 3), 3 μg of pNL4-3 and 0.5 μg of pcDNA-A3G (lane 4), 3 μg of pNL4-3, and 0.5 μg of pcDNA-A3G plus 2.5 μg of myc-tagged wild-type Vif (lane 5) or indicated untagged Vif mutants (lanes 6 to 9). myc-tagged wild-type Vif was used here to allow for the discrimination of Vif expressed from pNL4-3 and Vif expressed from the codon-optimized vector on the gel. A total 6 μg of empty vector for the mock sample (lane 1) and 0.5 μg of pcDNA-A3G (lane 2) for the control were transfected. Total amounts of transfected DNA were adjusted in all samples to 6 μg with empty vector DNA. Transfected cells and virus-containing supernatants were harvested 48 h posttransfection. Whole-cell lysates and concentrated viral extracts were prepared as described in Materials and Methods and were subjected to immunoblot analysis using antibodies to A3G, Vif, or tubulin (Tub) or an HIV-positive patient serum for the detection of viral capsid proteins (CA). (B) Band intensities of wild-type-Vif in cells were quantified. The Vif signal in lane 4 was defined as 1. (C) Band intensities of intracellular A3G were quantified. The A3G signal in lane 3 was defined as 1. (D) Band intensities of A3G in concentrated virus pellets (vps) were quantified. The A3G level in lane 3 was defined as 1. All data for B, C, and D are derived from at least 3 independent experiments (n = 3 to 8). Error bars represent the SEM. Statistical significance was determined using a one-tailed, paired Student’s t test versus lane 4 (transfected NL4-3 with A3G). *, P < 0.05; **, P < 0.01; ***, P < 0.001; **** P < 0.0001; n.s., not significant.
Overexpression of exogenous CBFβ counteracts the dominant-negative effect of Vif mutants.To further support our model of a CBFβ-dependent interference mechanism, we assessed the effect of exogenously overexpressed CBFβ. We hypothesized that if endogenous CBFβ was the factor responsible for the D/N phenotype of our Vif mutants, exogenous overexpression of CBFβ should alleviate the interference observed by D/N Vif mutants in the previous experiments. To directly test this hypothesis, 293T cells were cotransfected with pNL4-3 proviral DNA, together with a vector encoding A3G, and codon-optimized vectors for expression of untagged wild-type and mutant Vifs (Fig. 5A, lanes 3 to 12) in the presence or absence of a vector-encoding myc-tagged CBFβ. Controls included the expression of A3G in the absence (Fig. 5A, lane 1) or presence (lane 2) of Vif and ectopic CBFβ. Cells and virus-containing supernatants were harvested 48 h after transfection and were processed for immunoblotting, as described in Materials and Methods. Digital images from the A3G blots of three independent experiments were quantified as described for Fig. 4 (Fig. 5B and C). Consistent with our previous study (36), we found that overexpression of CBFβ indeed increased the expression of wild-type and mutant Vif proteins, except Vif-ΔK. As predicted, Vif mutants unable to bind CBFβ (i.e., Vif-ΔK and Vif-ΔH) did not interfere with the ability of wild-type Vif to reduce expression and inhibit packaging of A3G (Fig. 5, compare lane 1 to lanes 2 to 8). Importantly, exogenous overexpression of CBFβ neutralized the D/N phenotype of Vif-ΔC and Vif-ΔN and restored the ability of wild-type Vif to inhibit the packaging of A3G (Fig. 5, compare lanes 9 and 10 and lanes 11 and 12). This effect was dose dependent (data not shown). These results clearly demonstrate that the D/N effect of group 3 and 4 Vif mutants is exerted through competitive binding to endogenous CBFβ.
Overexpression of exogenous CBFβ counteracts the dominant-negative effect of Vif mutants. (A) 293T cells were transfected with 3 μg of pNL4-3/vif (−) together with 0.5 μg of pcDNA-A3G (lane 1), 3 μg of pNL4-3 and 0.5 μg of pcDNA-A3G (lane 2), or 3 μg of pNL4-3 together with 0.5 μg of pcDNA-A3G and 2.5 μg of untagged codon-optimized Vif vectors as indicated (lanes 3 to 12). Myc-tagged CBFβ (0.5 μg) was included as indicated (lanes 4, 6, 8, 10, and 12). Total amounts of transfected DNA were adjusted to 6.5 μg with empty vector DNA. Transfected cells and virus-containing supernatants were harvested 48 h posttransfection. Whole-cell extracts and concentrated virus samples were prepared as described in Materials and Methods and were subjected to immunoblot analysis with antibodies to A3G, Vif, CBFβ, or tubulin (Tub) or an HIV-positive patient serum for the detection of viral capsid proteins (CA). (B and C) Band intensities of A3G in cells (B) and A3G in the virus (C) were measured and expressed relative to the amounts detected in the absence of Vif (lane 1). Error bars reflect the SEM calculated from 3 independent experiments. Statistical significance was determined using a one-tailed, paired Student’s t test. **, P < 0.01. Lane numbers in B and C correspond to the numbering in panel A.
Runx1 overexpression mimics the effect of dominant-negative Vif.As noted above, CBFβ is a cofactor for the cellular transcription factor Runx1 (reviewed in reference 42). Indeed, interaction of Vif with CBFβ was found to inhibit Runx1 activity by sequestering CBFβ in the cytoplasm (43). Because Runx1 can bind CBFβ, we hypothesized that overexpression of Runx1 should induce a phenotype similar to the D/N Vif mutants. To test this, 293T cells were transfected with pNL4-3 and A3G together with human influenza virus hemagglutinin (HA)-tagged Runx1 (Fig. 6A, lanes 5 and 7) or a CBFβ binding-defective Runx1 mutant (Fig. 6A, lanes 6 and 8). Additional controls included pNL4-3 and pNL4-3/vif(−) with A3G in the absence of exogenous Runx1 (Fig. 6A, lanes 3 and 4) as well as mock-transfected or A3G only-transfected cells (Fig. 6A, lane 1 and 2, respectively). Cells and virus-containing supernatants were collected 48 h after transfection and were processed for immunoblot analysis, as described in Materials and Methods. Results from seven independent experiments were quantified (Fig. 6B to D). As expected, the expression of Vif from pNL4-3 reduced cellular and virus-associated levels of A3G (Fig. 6A, B, and D; compare lanes 3 and 4). Interestingly, overexpression of wild-type Runx1 dramatically reduced the levels of Vif expressed from pNL4-3 (Fig. 6C, compare lanes 4 and 5) and partially restored expression and packaging of A3G (Fig. 6A, B, and D; lane 5). Overexpression of a CBFβ binding-deficient mutant of Runx1 failed to affect Vif expression (Fig. 6A and C, compare lanes 4 and 6) and only insignificantly impacted the expression of A3G (Fig. 6A, B, and D, compare lanes 4 and 6). Exogenous expression of CBFβ significantly stabilized the expression of WT Runx1 (Fig. 6A, lane 7) but not of the CBFβ binding-deficient mutant (Fig. 6A, lane 8). At the same time, exogenous expression of CBFβ neutralized the effect of Runx1 on Vif expression and restored the ability of Vif to inhibit A3G expression and packaging (Fig. 6A, B, and D; compare lanes 4 and 5 to lane 7). Finally, expression of CBFβ in the absence of CBFβ binding-competent Runx1 significantly enhanced the expression of Vif and restored its ability to target expression and packaging of A3G (Fig. 6, compare lanes 4 and 5 to lane 8). Overall, these data confirm that competitive binding of Runx1 to endogenous CBFβ affects the CBFβ-dependent stabilization of Vif and severely inhibits its ability to target A3G. These effects of Runx1 depend on the ability to bind CBFβ and can be counteracted by overexpression of CBFβ. We, therefore, conclude that in these experiments Runx1 mimics the effect of D/N Vif mutants.
Runx1 overexpression mimics the effect of dominant-negative Vif. (A) 293T cells were transfected with 3 μg of pNL4-3/vif (−) together with 0.5 μg of pcDNA-A3G expression vector (lane 3), 3 μg of pNL4-3 and 0.5 μg of pcDNA-A3G expression vector (lane 4), 3 μg of pNL4-3 and 0.5 μg of pcDNA-A3G expression vector with 0.5 μg of HA-tagged Runx1 (lanes 5 and 7), or 0.5 μg of an HA-tagged CBFβ-binding-defective Runx1 mutant (Runx1 mut; lanes 6 to 8). In addition, 0.5 μg CBFβ was cotransfected in lanes 7 and 8. As a mock sample, 5 μg of empty vector was transfected (lane 1). Expression of A3G alone (0.5 μg of pcDNA-A3G expression vector) is shown in lane 2. Total DNA amounts in all samples were adjusted to 5 μg with empty vector DNA. Transfected cells and virus-containing supernatants were harvested 48 h posttransfection. Whole-cell lysates and concentrated viral lysates were prepared as described in Materials and Methods and subjected to immunoblot analysis using antibodies to A3G, Vif, HA (to detect Runx1 and Runx1 mut), CBFβ, and tubulin (tub) or an HIV-positive patient serum for the detection of viral capsid proteins (CA). (B) Band intensity of A3G in cell extracts were quantified and lane 3 was defined as 1. (C) Band intensities of wild-type-Vif in cells were quantified; the signal for lane 4 was defined as 1. (D) Band intensity of A3G in viral pellets (vps) were quantified; lane 3 was defined as 1. Error bars in panels B to D reflect SEM from seven independent experiments. Statistical significance was determined using a two-tailed, paired Student’s t test versus lane 4 (transfected NL4-3 with A3G). **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; n.s., not significant.
DISCUSSION
Of all the accessory proteins encoded by HIV-1 and other primate lentiviruses, Vif has arguably the strongest potential as a target for antiviral therapy. Several observations support this contention. First, while natural Vif variants can differ in their ability to target A3G, A3F, or A3H (45–49), there are no known primary replication-competent viruses that completely lack expression of a Vif protein, suggesting that Vif is absolutely critical for virus replication in vivo. Second, we previously attempted to force the selection vif-defective HIV-1 variants capable of replicating in naturally nonpermissive H9 cells in vitro by passaging vif-defective HIV-1 through semipermissive A3.01 cells. However, despite passaging the virus for 5 months in a mixture of semipermissive A3.01 cells and nonpermissive H9 cells, we were unable to select a variant/revertant capable of replicating in H9 cells (12). Thus, inhibiting the function of Vif via small-molecule inhibitors or other approaches, as mentioned in the introduction, has significant therapeutic potential. On the flipside, when targeting Vif, it will be important to achieve a near total inhibition since a partial inhibition could lead to sublethal mutagenesis by A3G and could aid in the selection for drug resistance, including drug-resistant Vif variants. (50).
The current study is an extension of previous work in which we identified Vif mutants with the D/N phenotype and where we presented a proof of principle that such Vif variants could have therapeutic potential by interfering with the function of wild-type Vif, resulting in increased packaging of A3G and, thus, inhibition of virus replication (20). However, the mechanism of action had remained elusive. We had shown that the D/N phenotype did not correlate with the ability of D/N Vif variants to compete with wild-type Vif for binding to A3G, and we had speculated that D/N Vif variants may act by depleting the pool of cellular Cul5 and/or ELOB/C, which are critical components of the E3 ubiquitin ligase complex involved in the ubiquitination and subsequent degradation of A3G.
We now show that none of our in-frame deletion mutants, including those with D/N properties, was capable of binding Cul5; only wild-type Vif was found to coimmunoprecipitate with Cul5 (Fig. 2). Of note, three of four deletions tested here do not impinge on the known Cul5 binding domain in Vif (38), suggesting that the in-frame deletions may induce conformational changes across the Vif protein that interfere with the binding of Vif to Cul5 and presumably to A3G as well. Thus, our initial model that D/N variants interfered with the function of wild-type Vif by binding to either A3G or Cul5, but not both, turned out to be incomplete. Indeed, we found that the common denominator for all Vif variants with the D/N phenotype was the ability to bind CBFβ. Our data clearly support the notion that CBFβ is expressed in cells in limiting quantities and that expression of functionally defective but CBFβ binding-competent Vif variants deplete the pool of free CBFβ, thereby inhibiting the production of properly folded wild-type Vif capable of inducing degradation of A3G. Our revised model of the mechanism of D/N interference, which involves competitive binding to CBFβ, is supported by our observation in the current study that overexpression of Runx1, a known cellular partner of CBFβ (51, 52), mimics the effect of D/N variants, by inhibiting the expression of Vif and increasing the packaging of A3G (Fig. 6). It is also supported by our previous observation that overexpression of Vif can inhibit the function of cellular Runx1 by competitive binding to CBFβ (43). Furthermore, our revised model is consistent with a study reporting that CBFβ stabilizes Vif to counteract A3G at the expense of Runx1-driven gene expression (31). Interestingly, one of the target genes of Runx1 is APOBEC3G (53). Thus, binding of CBFβ to Vif not only stabilizes Vif and potentiates its ability to induce degradation of A3G but also reduces the expression of A3G at the transcriptional level, thereby further augmenting the impact of Vif on A3G.
In conclusion, our data show that D/N variants of Vif are potentially useful as a tool for the intracellular immunization of cells against HIV-1. By competing with wild-type Vif for a cellular factor (i.e., CBFβ) that is critical for the production of biologically active wild-type Vif, stable expression of D/N Vif mutants could have powerful antiviral activity. This is especially true early in the infection cycle when infecting viruses are just starting to produce wild-type Vif and stably expressed D/N Vif variants are likely to outnumber the newly synthesized Vif, thus having a competitive advantage.
MATERIALS AND METHODS
Cell culture and transfection.The 293T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal bovine serum (FBS). Cells were seeded in 25-cm2 flasks 1 day before transfection. On the day of transfection, cells were at about 80% confluence (∼3 × 106 cells). Cells were transfected using TransIT (Mirus Bio LLC, Madison, WI) following the manufacturer’s recommendations. Where appropriate, empty vector DNA was used to adjust the total DNA amounts in each experiment.
Plasmids.Construction of pcDNA-hVif, a codon-optimized vector for the expression of wild-type HIV-1 Vif in the backbone of pcDNA3.1(−)myc-His was described previously (44). To express C-terminally myc-His-tagged Vif, the stop codon separating the vif gene from the epitope tag in pcDNA-hVif was removed by site-directed mutagenesis. Similarly, mutations in the vif gene were introduced into untagged or myc-His-tagged codon-optimized vectors via site-directed mutagenesis. A vector for the expression of untagged or myc-tagged human A3G in the backbone of pcDNA3.1(−) has been reported previously (54). The full-length molecular clone pNL4-3 (55) was used to produce wild-type HIV-1 virus. A vif-defective variant, pNL4-3ΔVif, was reported previously (56). Vectors for the expression of untagged or myc-tagged human CBFβ or HA-tagged human Runx1 and human RUNX G108R (a CBFβ binding mutant) have been reported previously (36).
Antibodies.For immunoblot analyses, the following antibodies were purchased from commercial sources: a rabbit polyclonal antibody to Cullin 5 (Cul5; catalog number ab184177; Abcam), rabbit polyclonal antibody to Elongin B (anti-TCEB2; catalog number A304-008A; Bethyl Laboratories, Inc.), rabbit polyclonal antibody to CBFβ (catalog number PA1-317; Thermo Fisher Scientific, Rockford IL). A rabbit polyclonal antibody raised in-house against recombinant Vif (Vif93) was used for immunoblotting and immunoprecipitation of Vif (56). APOBEC3G was identified using an in-house rabbit polyclonal antibody (ApoC17) (57). This antibody is available through the NIH Research and Reference Reagent program (catalog number 10082). HIV-1 Gag was identified using pooled HIV-immune globulin (IG) (catalog number 3957; NIH Research and Reference Reagent program). For immunoblot analysis of Runx1-hemagglutinin (HA), mouse anti-HA antibodies were used (catalog number H3663; Sigma-Aldrich, Inc., St. Louis, MO). A mouse monoclonal antibody to alpha-tubulin (catalog number T-9026; Sigma-Aldrich, Inc.) was used as a loading control. Horseradish peroxidase-conjugated antibodies were used for chemiluminescence (GE Healthcare, Piscataway NJ). For immunoblot analysis of immunoprecipitated samples, mouse- or rabbit-specific TrueBlot Ultra-HRP conjugates were used (catalog numbers 18-8817-33 and 18-8816-33, respectively; Rockland Immunochemicals Inc. Limerick, PA).
Coimmunoprecipitation analysis.Cells were harvested 24 h posttransfection and washed twice with cold PBS. Cells were lysed in 450 μl of radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 1% [vol/vol] NP-40, 0.5% sodium deoxycholate, and 0.1% sodium dodecyl sulfate) at 4°C for 20 min and then clarified by centrifugation at 16,100 × g for 5 min. Five percent of each lysate was removed and used as input control; the remaining lysate was used for immunoprecipitation. Precleared cell lysates were mixed with anti-c-myc agarose affinity gel (catalog number A7470; Sigma-Aldrich, Inc. St. Louis MO;) or anti-Vif-conjugated protein A-Sepharose beads (catalog number P3391; Sigma-Aldrich, Inc.) and incubated at 4°C for 1 h on a rotator. Samples were then washed three times with RIPA buffer. Proteins were eluted by heating beads in sample buffer (2% sodium dodecyl sulfate, 5% 2-mercaptoethanol, 62.5 mM Tris-HCl [pH 6.8], 5% glycerol, and 0.001% bromophenol blue) for 10 to 15 min at 95°C with occasional vortexing. Eluted samples were subjected to gel electrophoresis and immunoblot analysis.
Cell and viral sample preparation and immunoblotting.Cells and virus-containing supernatants were harvested 48 h after transfection. Cell pellets were washed with cold phosphate-buffered saline (PBS), resuspended in PBS (200 μl/5 × 106 cells), and mixed with an equal volume of 2× sample buffer. For virus-containing supernatants, cellular debris was removed by centrifugation (5 min at 800 rpm) followed by filtration (0.45-μm pore size). For immunoblot analyses, virus-containing supernatants were concentrated by pelleting through a 20% sucrose cushion (60 min, 4°C at 35,000 rpm in an SW55 rotor). Viral pellets were solubilized with 1× sample buffer. Samples were subjected to SDS-PAGE. Proteins were transferred to polyvinylidene difluoride (PVDF) membranes and reacted with appropriate antibodies as described in the text. Membranes were then incubated with horseradish peroxidase-conjugated secondary antibodies. The proteins were visualized by Clarity Western ECL substrate (Bio-Rad Laboratories) or immobilon Western chemiluminescent HRP substrate (Millipore). Images were acquired using the ChemiDocTM imaging system (Bio-Rad Laboratories, Hercules CA) and were quantified using Image Lab (v 6.0) software.
ACKNOWLEDGMENTS
We acknowledge Helena Barber for technical assistance and Haruka Yoshii for helpful discussions. We thank Alicia Buckler-White and the LMM core facility for their support with sequence analyses. The following reagent was obtained through the NIH AIDS Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases (NIAID): HIV-1 immunoglobulin (catalog number 3957).
This work was supported by the Intramural Research Program of the NIH, NIAID (1 Z01 AI000669).
FOOTNOTES
- Received 3 October 2019.
- Accepted 27 December 2019.
- Accepted manuscript posted online 15 January 2020.
- Copyright © 2020 American Society for Microbiology.
