To test whether UBA2 can stabilize APOBEC3G protein, UBA2 was fused at the C-terminal end of APOBEC3G (Fig. 1A). The APOBEC3G without the UBA2 fusion (Fig. 1B) or fused with a mutant L392A UBA2 that is incapable of stabilizing proteins (Fig. 1C; [35]), was used as controls. The fusion products were cloned into a mammalian gene expression plasmid pCDNA3.1 and the resulting plasmids were designated as pcDNA3.1(-)-Apo-E/Hygromycin (E) for the untagged APOBEC3G, pcDNA3.1(-)-Apo-U/Hygromycin (U) for the APOBEC3G-UBA2 fusion, and pcDNA3.1(-)-Apo-M/Hygromycin (M) for the APOBEC3G-UBA2* mutant fusion. Protein stability of APOBEC3G was determined either by expression of these plasmids individually or by co-transfection of each individual APOBEC3G-carrying plasmid construct with a Vif-carrying plasmid (Vif-VR1012) in HEK293 cells. As shown in Fig. 2A, expression of untagged APOBEC3G produced a strong protein band at approx. 46 kD consistent with the size of APOBEC3G (Fig. 2A, lane 2). Slight increase in molecular weight was detected in the APOBEC3G-UBA2 and APOBEC3G-UBA2* fusion products (Fig. 2A, lanes 3–4).

Approximately equal amount of protein was produced in each of these plasmid constructs without vif gene expression (Fig. 2A–b). When vif is expressed in the APOBEC3G-producing HEK293 cells, a significant decrease of APOBEC3G with more than 10-fold reduction was noticed in the untagged APOBEC3G cells (Fig. 2A, lane 5). In contrast, a small with about 2-fold decease of APOBEC3G-UBA was detected when APOBEC3G was fused with the wild type UBA2 (Fig. 2A, lane 6). Consistent with the finding that a single point mutation of APOBEC3G (L392A) abolishes the ability of APOBEC3G to stabilize proteins [35], production of Vif in these cells reduced the APOBEC3G-UBA2* protein level to the level that is similar to the untagged APOBEC3G (Fig. 2A lane 7 vs. lane 5). Together, these data suggested that the wild type UBA2, when it is fused with APOBEC3G, is indeed able to stabilize APOBEC3G and renders it more resistant to Vif than the untagged APOBEC3G.

One possibility for the observed resistance of APOBEC3G-UBA2 to Vif could be explained by the reduced binding of APOBEC3G-UBA2 to Vif. To test this possibility, Myc-tagged Vif was pull-down by immunoprecipitation in the APOBEC3G-producing HEK293 cells. Western blot analyses were carried out to measure the bindings of different APOBEC3G constructs to Vif. As shown in Fig. 2B, no obvious reduction of the binding of APOBEC3G-UBA2 to Vif was observed (Fig. 2B, lane 5). In fact, binding of APOBEC3G-UBA2 to Vif appeared to be stronger than the untagged APOBEC3G or APOBEC3G with the mutated UBA2. This increase binding could potentially be due to presence of the excessive APOBEC3G-UBA2, which is clearly shown by the high level of APOBEC3G remained in the supernatant (Fig. 2B, lane 2). Nevertheless, these data suggest that the observed resistance of APOBEC3G to Vif is not caused by reduction binding.

Most cellular proteins are targeted for degradation by the proteasome. Prior to proteasome-mediated proteolysis, the proteins are covalently attached to ubiquitin. A poly-ubiquitin chain will be formed and function as a degradation signal. The poly-ubiquitinated protein can then be recognized by the 26S proteasome for degradation [38]. If the ubiquitin chain elongation is interrupted, this protein cannot be recognized by the 26 S proteasome and thus it cannot be degraded. UBA2 binds to ubiquitin directly and inhibits elongation of polyubiquitin chains by capping conjugated ubiquitin [30,39]. Since Vif mediates APOBEC3G degradation by promoting protein ubiquitination of APOBEC3G [3]via Cullin5-EloB/C E3 ligase to induce polyubiquitination of APOBEC3G [19,20], it is possible that UBA2 may either sequester ubiquitin from APOBEC3G or prevent polyubiquitin chain elongation. As results, the un-ubiquitinated APOBEC3G becomes resistant to proteasome-mediated proteolysis. To test this possibility, polyubiquitin was overproduced through a pcDNA3.1-HA-Ubiquitin plasmid [40,41] in the HEK293 cells co-producing Vif and various APOBEC3G products. As shown in Fig. 3A, APOBEC3G-UBA2 fusion protein showed relative strong intensity in comparison with the untagged APOBEC3G (Fig. 3A–a, lane 3 vs. lane 1). However, production of excessive polyubiquitin completely abolished the difference between the protein level of APOBEC3G-UBA2 and APOBEC3G (Fig. 3A–a, lane 5 vs. lane 4). Western protein blotting with anti-Vif and anti-HA for ubiquitin detection confirmed proper production of Vif and polyubiquitin in these cells. Therefore, over-production of polyubiquitin can diminish the ability of UBA2 for APOBEC3G stabilization.

To further verify whether fusion of APOBEC3G to UBA2 results in less binding to polyubiquitin, APOBEC3G in the presence or absence of Vif was collected by immunoprecipitation using anti-APOBEC3G monoclonal antibody. The pull-down protein products were subject to Western blot analyses as shown in Fig. 3B. Approximately equal amount of APOBEC3G was collected in all cells with the exception of the control cells (Fig. 3B–a, lane 4), in which only endogenous APOBEC3G was pull-down. Without Vif, minimal and background level of polyubiquitin was detected in all APOBEC3G-producing cells (Fig. 3B–a, lanes 1–3). In contrast, strong polyubiquitin was detected in the vif-expressing cells with untagged APOBEC3G or APOBEC3G-UBA2* (Fig. 3B–a, lanes 5 and 7). However, much reduced level of polyubiquitination was observed in vif-expressing cells carrying the APOBEC3G-UBA2 (Fig. 3B–a, lane 6). This observation provides direct support to the notion that UBA2 may prevent polyubiquitin chain elongation on APOBEC3G.

To test whether inhibition of the 26S proteasome activity has any impact on the ability of UBA2 to stabilize APOBEC3G against Vif, APOBEC3G-producing HEK293 cells were treated the proteasome inhibitor MG132 in the presence of Vif. APOBEC3G protein levels were measured and compared between cells with or without the MG132 treatment. Similar to what we have shown in Fig. 2A, the protein intensity of APOBEC3G-UBA2 was significantly higher than that without the UBA2 tag (Fig. 4A, lane 2 vs. lane 1), suggesting the protein stabilizing capacity of UBA2. APOBEC3G fusion with a mutant UBA2* reduced its ability to stabilize APOBEC3G (Fig. 4A, lane 3). Significantly, HEK293 cells treated with the proteasome inhibitor MG132 all showed much higher protein intensities than the APOBEC3G-UBA2 producing cells without MG132 treatment (Fig. 4A, lanes 4–6 vs. lane 2). These enhanced protein levels were observed in all of the APOBEC3G protein constructs regardless whether it is fused with UBA2 or not, suggesting UBA2 stabilizes APOBEC3G through resistance to proteasome-mediated proteolysis.

To test whether APOBEC3G stabilized by UBA2 can further enhance the suppressive effect of APOBEC3G on viral infectivity, the HIV-1 viral particles were produced from HEK293 cells that expressing different constructs of APOBEC3G as described. To minimize potential differences of production of each protein construct and viral packaging, HEK293 cells that stably express APOBEC3G, APOBEC3G-UBA2, and APOBEC3G-UBA2* fusion proteins were created by proper antibiotic selection. High level expression of these proteins was further verified by Western blot analysis (Figure 5A–a). To produce APOBEC3G-carrying viral particles, the pNL4-3 plasmid was expressed in HEK293 viral producing cells that stably expressing different APOBEC3G fusion proteins. The infectious viral particles were harvested 48 hrs after transfection as previously described [42]. Presence of different APOBEC3G constructs was detected with approx. equal amount within all three types of viral particles (Fig. 5A–b). To test whether the potential effect of the viral expressing Vif on the stability of APOBEC3G, levels of APOBEC3G in the viral particle producing HEK293 cells were further measured after viral gene expressions. Essentially the same Vif effect on APOBEC3G was seen between the viral expressing Vif and Vif expressed from a plasmid (Fig. 5A–c).

To test the potential impact of different APOBEC3G constructs packaged in the viral particles on viral infectivity and replication, concentration of the viral stocks were normalized by determination of the p24 antigen levels. The viral infectivity of viruses packaged with different APOBEC3G constructs were measured with the MAGI-CCR5 assay as previously described [43]. This assay measures viral infectivity in a single cycle of viral infection. About 50% reduction of viral infectivity was observed in viruses packed from cells producing high level of APOBEC3G than endogenous level of APOBEC3G (Fig. 5B–a, lane E vs. lane C). An additional 17% and significant reduction of viral infectivity (P < 0.01) was also observed in viruses packaged from cells expressing the APOBEC3G-UBA2 fusion protein (Fig. 5B–a, lane U). In contrast, no significant difference was detected between viruses carrying untagged APOBEC3G or APOBEC3G fused with a mutant UBA2, indicating the additional reduction of viral infectivity observed in the APOBEC3G-UBA2 fusion was indeed due to the stabilizing effect of UBA2 on APOBEC3G (Fig. 5B–a, lane M vs. lane E). These differences in viral infectivity were not observed in the Vif(-) viral infections suggesting the observed differences were caused by Vif (Fig. 5B–b).

To further evaluate the observed effects of APOBEC3G variants on spread viral infection, CEM-SS cells, a cell line derived from CD4-positive T-lymphocytes, were infected with the same Vif(+) and Vif(-) viral particles packaged with different APOBEC3G variants as described above. P24 antigenemia was measured from day 3 to day 21 post-viral infection. Similar suppressive effects of the APOBEC3G variant on viral infection as described above for the MAGI-CCR5 experiment were also seen in CEM-SS cells (Fig. 5C). The differences are however most pronounced in day 21 post-infection: while infection of CEM-SS with the control viral particles produced approximately 1,200 ng/ml of p24 antigen, about 400 ng/ml of p24 antigen was seen in CEM-SS cells infected with viral particles packed with either untagged APOBEC3G or the UBA2 mutant variant (Fig. 5C–a). Additional reduction of viral replication with approx. 200 ng/ml was observed in the same cells when they were infected with the viral particles packaged with the APOBEC3G-UBA2. All of the APOBEC3G variants conferred the same level of strong viral suppression against Vif(-) viral infection (Fig. 5C–b), suggesting that the observed differences as described in the Vif(+) viral infections were due to interaction between Vif and APOBEC3G.

HEK293 cell was maintained in Dulbecco's minimal essential medium (DMEM) supplemented with 10% fetal bovine serum. MAGI-CCR5 cell, a HeLa-CD4 cell derivative that expresses CCR5 and that has an integrated copy of the HIV-1 long terminal repeat (LTR)-driven β-D-galactosidase reporter gene [43], was maintained in Dulbecco's modified Eagle's medium (DMEM) supplied with 10% (vol/vol) fetal bovine serum (FBS) (Bio Whittaker), 200 μg/ml G418, 50 U/ml hygromycin (CalBiochem), and 1 μg/ml puromycin. CEM-SS cells were grown in RPMI 1640 medium. To produce APOBEC3G, APOBEC3G-UBA2, and APOBEC3G-mutant UBA2 fusion proteins, three plasmids including pcDNA3.1(-)-Apo-E/Hygromycin (E), pcDNA3.1(-)-Apo-U/Hygromycin (U), and pcDNA3.1(-)-Apo-M/Hygromycin (M) were constructed according to the strategy shown in Figure 1. To make these plasmid constructs, the APOBEC3G gene was amplified from the plasmid pcDNA3.1-HA-APOBEC3G by PCR. The 5' primer used for the construction of all three plasmids was 5'-GCGCGCGCGCCTCGAGACCATGAAGCCTCACTT-3'; The 3' primers used for the construction of the E plasmid was 5'-ATCCAAGACGGAATTCCTAGAACTCGTTTTCCTGATTCTGGAG-3' and the 3' primer used for the U and M plasmid was 5'-ATCCAAGACGGAATTCGTTTTCCTGATTCTGGAG-3'. The UBA2 gene fragment was amplified from plasmid pcDNA3.1-HHR23A by PCR. The 5' primer used was 5'-ATCCAAGACGGAATTCACGCCGCAGGAGAAAGAAGCTATAG-3'; the 3' primer for the APOPEC3G-UBA2 fusion was 5'-ATCGTACTCGAAGCTTCTAACTCAGGAGGAAGTTGGCAG-3'; and the 3' primer for the APOPEC3G-UBA2* fusion was 5'-ATCGTACTCGAAGCTTCTAACTCAGagcGAAGTTGGCAG-3'. Purified PCR products were first cloned into the mammalian expression plasmid pcDNA3.1(-)/neo, the gene fragments were then cut off and cloned into a pcDNA3.1(-)/Hygromycin plasmid. Correct insertion and nucleotide sequence of each gene fragment was verified by restriction enzyme digestions and was confirmed by nucleotide sequencing. The pcDNA3.1-HA-Ubiquitin plasmid was used to express polyubiquitin [40,41]. The pNL4-3 plasmid was used to packaged virus in HEK293 cells as described previously [42].

Transfected HEK293 cells were harvested, washed 2 times with cold PBS, and lysed in lysis buffer (50 mM Tris, pH 7.5, with 150 mM NaCl, 1% Triton X-100, and complete protease inhibitor cocktail tablets) at 4°C for 1 h, then centrifuged at 10,000 g for 30 min. Cell lysates were mixed with anti-APOBEC3G Ab (NIH reagents program) and incubated at 4°C for overnight. The mixture of antigen and antibody was incubated with protein A agarose beads (Sigma) and incubated at 4°C for 3 h. Samples were then washed three times with washing buffer (20 mM Tris, pH 7.5, with 100 mM NaCl, 0.1 mM EDTA, and 0.05% Tween-20). Beads were eluted with loading buffer. The eluted materials were then analyzed by SDS-PAGE.

For Western blot analysis, HEK293 cells were harvested and rinsed with ice-cold HEPES-buffered saline (pH 7.0), then lysed in an ice-cold cell lysis buffer [20 mM Tris-HCl, pH7.6, 150 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 1 mM DTT, 5 μM Trichostatin A, 1 mM sodium orthovanadate, 1 mM PMSF, 1 mM NaF and complete protease inhibitors (Roche Applied Science)]. Cellular lysates were prepared and the protein concentration was determined using the Pierce protein assay kit. For immunoblotting, an aliquot of total lysate (50 μg of proteins) in 2 × SDS-PAGE sample buffer (1:1 v/v) was electrophoresed and transferred to a nitrocellulose filter. Filters were incubated with appropriate primary antibody in Tris-buffered saline (TBS, pH 7.5) and 5% skim milk or 5% BSA overnight. The primary antibodies include anti-APOBEC3G antibody at a 1:500 dilution (NIH reagents program), anti-Vif antibody at a 1:200 dilution (NIH reagents program), anti-HA antibody at a 1:1000 dilution, and anti-β-actin (Sigma) antibody at a 1:3,000 dilution. After washing, the filter was further incubated with secondary antibody in TBS-Tween-20 (TBS-T) buffer for 1 h. Protein bands were visualized by an ECL detection system. Goat anti-mouse or anti-mouse IgG-HRP conjugate (dilution of 1:3,000) were used as secondary antibodies according to the corresponding primary antibodies.

Plasmid DNA was transfected into HEK293 or CEM-SS cells by using Lipofectamine 2000 transfection reagent (Invitrogen) according to the manufacturer's instructions. To create stable APOBEC3G-expressing cell lines, the plasmid DNA of pcDNA3.1(-)-Apo-E/Hygromycin, pcDNA3.1(-)-Apo-U/Hygromycin, pcDNA3.1(-)-Apo-M/Hygromycin was transfected into HEK293 cells. HEK293 cells that stably produce a high level of APOBEC3G, APOBEC3G-UBA2 or APOBEC3G-UBA2* were first established by selection of hygromycin resistant cells (300 μg/ml) for 2 weeks and verified by the Western blot analyses (Fig. 5A–a). To generate infectious viral particles, HEK293 was inoculated into 6-well plate one day before pNL4-3 plasmid transfection to ensure adequate (50–60%) cell confluency. The full-length molecular clone pNL4-3 and pNL4-3 ΔVif plasmid was then transfected into HEK293 cells to produce wild type Vif(+) or mutant Vif(-) HIV-1 viruses [42]. Forty-eight hours p.t., HIV-1 viral particles were harvested from the supernatants of HEK293 cells by centrifugation at 1,000 rpm/min for 5 min. The isolated viral particles were split into 2 ml aliquots and stored in -80°C. To ensure equal levels of viral infection, the viral stocks were normalized by determining levels of p24 antigens in each viral stock. The level of p24 antigen was determined by using a commercial p24 antigen kit from ZeptoMetrix Co. (Buffalo, NY) following the manufacture's instructions.

To evaluate the suppressive effect of APOBEC3G, APOBEC3G-UBA2 and APOBEC3G-UBA2* on viral replication in proliferating CD4+ T-lymphocytes, CEM-SS cells (APOBEC3G negative CD4+ T-lymphocytes) that stably express a plasmid control, APOBEC3G, APOBEC3G-UBA2 and APOBEC3G-UBA2* were established. 1 × 10⁶ CEM-SS cells were either mock infected or infected with 3000 TCID50 of HIV-1_NL4-3 and HIV-1_NL4-3ΔVif. The viral replication was measured by p24 antigenemia.

MAGI assay was used to determine the viral infectivity as previously described [43]. Briefly, MAGI-CCR5 cells were cultured in 6-well plates one day before infection so that cells can reach approximately 40–50% confluency on the day of infection. The medium of each well was removed before the viral supernatant was added to infect cells in a total volume of 300 μl of complete DMEM with 20 μg/ml of DEAE-dextran. Cells in virus-containing medium were incubated at 37°C in a 5% CO₂ incubator. After 2 hours incubation, 1.5 ml complete DMEM medium was added to each well. The cells were further incubated under the same condition for 48 hrs after incubation. The media were removed and 2 ml fixing solution (1% formaldehyde, 0.2% glutaraldehyde in PBS) was added to each well. Cells were washed twice with PBS and then fixed for 5 min. with 600 μl of staining solution (20 μl of 0.2 M potassium ferrocyanide, 20 μl of 0.2 M potassium ferricyanide, 2 μl of 1 M MgCl₂, 10 μl of 40 mg/ml X-gal in PBS). Cells were then incubated for 2 hrs at 37°C in a non-CO₂ incubator. Staining was stopped by removing the staining solution and washed twice with PBS. Blue dots were counted as infected cells as described previously [43] under the microscope. The viral infectivity was determined by comparing the total number of infected cells with uninfected cells in each well. Each experiment was performed in triplicate and results of these experiments were repeated at least three times. Statistic student t-test was used to determine potential significant difference among different treatment groups.