Analysis of HIV-1 viral infectivity factor-mediated proteasome-dependent depletion of APOBEC3G: correlating function and subcellular localization.

To study how HIV-1 viral infectivity factor (Vif) mediates proteasome-dependent depletion of host factor APOBEC3G, functional and nonfunctional Vif-APOBEC3G interactions were correlated with subcellular localization. APOBEC3G localized throughout the cytoplasm and co-localized with gamma-tubulin, 20 S proteasome subunit, and ubiquitin at punctate cytoplasmic bodies that can be used to monitor the Vif-APOBEC3G interaction in the cell. Through immunostaining and live imaging, we showed that a substantial fraction of Vif localized to the nucleus, and this localization was impaired by deletion of amino acids 12-23. When co-expressed, Vif exhibited more pronounced localization to the cytoplasm and reduced the total cellular levels of APOBEC3G but rarely co-localized with APOBEC3G at cytoplasmic bodies. On the contrary, Vif(C114S), which is inactive but continues to interact with APOBEC3G, stably associated with APOBEC3G in the cytoplasm, resulting in complete co-localization at cytoplasmic bodies and a dose-dependent exclusion of Vif(C114S) from the nucleus. Following proteasome inhibition, cytoplasmic APOBEC3G levels increased, and both proteins co-accumulated nonspecifically into a vimentin-encaged aggresome. Furthermore in the presence or absence of APOBEC3G, Vif localization was significantly altered by proteasome inhibition, suggesting that aberrant localization may also contribute to the loss of Vif function. Finally mutations at Vif Ile(9) disrupted the ability of Vif or Vif(C114S) to coimmunoprecipitate and to co-localize with APOBEC3G, suggesting that the N terminus of Vif mediates interactions with APOBEC3G. Taken together, these results demonstrate that cytoplasmic Vif-APOBEC3G interactions are required but are not sufficient for Vif to modulate APOBEC3G and can be monitored by co-localization in vivo.

Vif and APOBEC3G coimmunoprecipitate from human cell lysates, and recombinant APOBEC3G and Vif have been shown to interact in vitro (5,17,19,21,(23)(24)(25), suggesting that these proteins directly interact. In addition, mutational analyses have uncovered some of the regions required for Vif-APOBEC3G functional interactions. Vif mutants lacking the conserved SLQ(Y/F)LA motif can coimmunoprecipitate with APOBEC3G but are not functional (19,26). Mutations at conserved Cys 114 or Cys 133 that abolish Vif effector functions also render Vif inactive toward APOBEC3G in the context of virus (27), but whether these mutations affect interactions with APOBEC3G is unknown.
In this study, we correlated Vif-mediated, proteasome-dependent degradation of APOBEC3G with the subcellular localization of the Vif-APOBEC3G interaction. Through the analysis of functional and non-functional Vif-APOBEC3G interactions, we demonstrated that APOBEC3G influenced the subcellular distribution of Vif, that Vif-APOBEC3G interactions were required but were not sufficient for Vif to modulate APOBEC3G protein levels, and that Vif Ile 9 was critical for Vif to interact with APOBEC3G.
DNA Manipulations-To generate yellow fluorescent protein (YFP) epitope-tagged versions of Vif or Vif mutants, the Vif coding region was PCR amplified and cloned into the EcoRI and BamHI sites of pEYFP-C1 (BD Biosciences, Palo Alto, CA). The APOBEC3G coding region was amplified from cDNA derived from frozen human peripheral blood mononuclear cells (a generous gift from Dr. Mario Stevenson). To generate pCFP-APO, the APOBEC3G coding sequence was cloned into the HindIII and SacII sites of pECFP-C1 (BD Biosciences). For expression of APOBEC3G with a C-terminal epitope tag, a Kozak ribosome recognition sequence (ccacc) was placed directly upstream of the APOBEC3G start codon during PCR amplification. APOBEC3G with a C-terminal 3ϫ hemagglutinin (HA) tag (pAPO-HA) and pAPO-CFP were engineered by cloning APOBEC3G into the EcoRI and XhoI sites of pIRES-hrGFP-2a (Stratagene, La Jolla, CA) and the HindIII and SacII sites of pECFP-N1 (BD Biosciences), respectively. Tat-red fluorescent protein (RFP) was generated by cloning HIV-1 Tat into the HindIII and BamHI sites of pDsRED-N1 (BD Biosciences).
To induce random mutations within the Vif C114S coding sequence, pNL-A1Vif C114S was used as a template for low fidelity PCR, and the resulting products were cloned into pEYFP-C1 as described above. DNA from random colonies was prepared (Promega, Madison, WI) and used to transfect 293T cells.
Cell Culture and Transfection Procedures-293T cells were maintained in a humidified incubator (5% CO 2 ) at 37°C in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin (Invitrogen). Qiagen-purified plasmid DNA (Qiagen, Valencia, CA) was transfected into 293T cells using Lipofectamine 2000 (Invitrogen). For Western blot and immunoprecipitation experiments, 293T cells were transfected in either 6-or 12-well plates when the cells were ϳ60% confluent. The amount of vector used and the molar ratio of vector for co-transfection experiments are described under "Results" and in the figure legends. When necessary, final DNA amounts were made equal by the addition of pGEM (Promega). For live imaging and immunolocalization, 293T cells were seeded into poly-D-lysine-coated 35-mm glass bottom culture dishes and transfected when the cells were ϳ50% confluent. For proteasome inhibition studies, culture medium containing 100 M of N-acetyl-Leu-Leu-norleucine-CHO (ALLN, Calbiochem) or the equivalent volume of Me 2 SO was added to cells 12 h prior to harvesting or imaging 36 h post-transfection.
Preparation of Total Cell Lysates, Immunoprecipitation, and Western Blot Analysis-To prepare total protein lysates, each well of a 6-or 12-well plate was washed one time in phosphate-buffered saline (PBS, Invitrogen) and then lysed in either 400 or 200 l, respectively, of Mammalian Protein Extraction Reagent (M-PER, Pierce) supplemented with 0.5% (v/v) Triton X-100 (Pierce), 150 mM NaCl, 5 mM EDTA, and a 1:100 (v/v) dilution of a protease inhibitor mixture for mammalian tissue for 30 min at 4°C with gentle rotation. Lysates were harvested from the well, and insoluble material was removed by centrifugation for 5 min at full speed in a microcentrifuge. Protein concentration was determined by DC protein assay (Bio-Rad). For immunoprecipitation, 0.5 mg of lysate was diluted to 0.5 mg/ml in 1 ml of lysis buffer. APO-HA was precipitated by incubation with agarose-conjugated rabbit ␣-HA (20 g of IgG, Santa Cruz Biotechnology, Inc., Santa Cruz, CA). To immunoprecipitate CFP-APO, lysates were first precleared with a 50-l bed volume of Protein G-Sepharose (Amersham Biosciences) for 1 h at 4°C. CFP-APO was then precipitated from precleared lysates by incubation with 5 g of an ␣-GFP rabbit polyclonal (BD Biosciences) for 3 h at 4°C. Antibody was captured by incubation with a 50-l bed volume of Protein G-Sepharose for 1 h at 4°C followed by four washes in 1 ml of lysis buffer for 10 min each time. Protein was eluted by boiling for 5 min at 100°C in sample buffer (50 mM Tris-HCl, pH 6.8, 100 mM dithiothreitol, 2% (w/v) SDS, 0.1% (w/v) bromphenol blue, 10% (v/v) glycerol). For SDS-PAGE of protein lysates, samples were denatured and reduced by adding 4ϫ SDS-PAGE sample buffer followed by boiling at 100°C for 5 min. Protein was resolved by 12% SDS-PAGE and transferred onto a polyvinylidene difluoride membrane (Bio-Rad) using a Semi-Dry Electroblotter (Bio-Rad). Following transfer, the membrane was blocked overnight in 5% (w/v) nonfat dry milk in TBS-T (20 mM Tris, pH 7.4, 150 mM NaCl, 0.1% (v/v) Tween 20) and washed three times for 10 min each in TBS-T before and after the addition of antibody. All antibodies were diluted in 2.5% (w/v) nonfat dry milk in TBS-T. CFP, YFP, and GFP were detected using a mouse monoclonal antibody (mAb) against GFP diluted to 1 g/ml (BD Biosciences). RFP was detected with a rabbit polyclonal (BD Biosciences) diluted to 0.1 g/ml. Human CycT1 was detected with a goat anti-CycT1 polyclonal antibody (Santa Cruz Biotechnology, Inc.) diluted to 0.1 g/ml. HA was detected with a rabbit polyclonal antibody (Santa Cruz Biotechnology, Inc.) diluted to 0.02 g/ml. Vif and Vif mutants were detected using a Vif mAb diluted 1:5000 (This reagent was obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, National Institutes of Health: HIV-1 Vif monoclonal antibody (no. 319) from Dr. Michael H. Malim (28 -30).). All horseradish peroxidase-conjugated secondary antibodies were used at a dilution of 0.05 g/ml (Santa Cruz Biotechnology, Inc.). Blots were developed with the BM Chemiluminescence Blotting kit (Roche Applied Science) and exposed to Kodak BioMax MR x-ray film (Eastman Kodak Co.).
Immunolocalization and Live Imaging-Cells were fixed in 4% (v/v) paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA) in PBS for 20 min followed by permeabilization with 0.1% (v/v) Triton X-100 (Pierce) for 20 min. Alternatively, cells were fixed and permeabilized using Ϫ20°C methanol for 10 min. Samples were blocked for 30 min in PBS containing 2% (w/v) bovine serum albumin (BSA/PBS). All primary and secondary antibodies were diluted in BSA/PBS. Primary antibodies ␣-Vif mAb (1:500 dilution), ␣-HA rabbit polyclonal (0.2 g/ml), ␣-␥tubulin mAb (1:500 dilution), ␣-vimentin mAb (0.4 g/ml, Santa Cruz Biotechnology, Inc.), ␣-20 S proteasome subunit rabbit polyclonal (1 g/ml, Calbiochem), and ␣-ubiquitin mAb (2 g/ml, Santa Cruz Biotechnology, Inc.) were diluted in BSA/PBS and incubated with sample for 1 h. ␣-Mouse and ␣-rabbit secondary antibodies directly conjugated to Alexa Fluor 488 or Alexa Fluor 568 (Molecular Probes, Inc., Eugene, OR) were diluted to 5 g/ml in BSA/PBS and incubated with sample for 45 min. To stain nuclei, Hoechst 33258 was used at a concentration of 0.5 g/ml. For live imaging, cells were visualized without changing the culture medium. Samples were visualized with a Leica confocal imaging spectrophotometer system (TCS-SP2) (Leica, Exton, PA) attached to a Leica DMIRE inverted fluorescence microscope and equipped with an argon laser (458, 476, 488, and 514 nm lines), two HeNe lasers (543 and 633 nm lines), an acousto-optic tunable filter to attenuate individual visible laser lines, and a tunable acousto-optical beam splitter. All images were acquired from a 63ϫ, 1.32 numerical aperture oil immersion objective, and image analysis was performed using Leica Confocal Software and Adobe Photoshop 7.0 (Adobe Systems, Mountain View, CA).

RESULTS
Analysis of the Proteasome-dependent Depletion of APOBEC3G by Vif-The level of Vif-mediated depletion of APOBEC3G was determined over a broad range of Vif (pNL-A1):APOBEC3G (pCFP-APO) vector ratios ranging from 1:1 to 1:0.0625 where 1 refers to 130 fmol of vector. The level of depletion was determined by comparing the steady-state levels of CFP-APO when co-expressed with Vif to the equivalent ⌬vif control (pNL-A1⌬vif). CFP-APO depletion by Vif was most significant between the 1:0.25 and 1:0.0625 ratios (Fig. 1A). Recovery of CFP-APO levels by proteasome inhibition (ALLN) was most dramatic at the 1:0.25 ratio (Fig. 1A). Despite significant depletion by Vif, only modest recovery of CFP-APO was observed at the 1:0.125 and 1:0.0625 ratios following proteasome inhibition (Fig. 1A). In the absence of Vif, CFP-APO levels were not significantly affected by proteasome inhibition except at the lowest level of pCFP-APO input (1:0.0625 ratio) for which a modest reduction in expression was observed (Fig. 1A). Identical expression profiles were also observed with pAPO-CFP or pAPO-HA in the presence and absence of Vif (data not shown). In the presence (Fig. 1A) or absence (data not shown) of CFP-APO, proteasome inhibition elevated the steady-state levels of Vif and resulted in the appearance of higher molecular weight species (Fig. 1A, arrow). The pNL-A1:pCFP-APO ratio of 1:0.25 was also visualized by confocal microscopy. In cells expressing Vif, CFP-APO was either not visible or was detected at a significantly reduced level relative to the ⌬vif control ( Fig.  1B, DMSO, arrows). Consistent with our immunoblot analysis, CFP-APO levels were elevated by proteasome inhibition, and the number of cells visibly co-expressing the two proteins increased (Fig. 1B, ALLN). Subcellular Localization of APOBEC3G-To examine the subcellular localization of APOBEC3G in 293T cells, APO-HA was immunolocalized and CFP-APO and APO-CFP were visu-alized directly in live cells. APOBEC3G localized throughout the cytoplasm and was largely excluded from the nucleus ( Fig.  2A). APOBEC3G was also concentrated at punctate bodies that were often in close proximity to the nucleus but were not necessarily restricted from other regions in the cytoplasm ( Fig.  2A). Titration of pCFP-APO revealed that transfection of as little as 4 fmol (31 ng) was sufficient to visualize the protein by live imaging. At this level of expression, CFP-APO continued to localize to cytoplasmic bodies in ϳ30% of cells. When co-expressed, APO-HA and CFP-APO, which also coimmunoprecipi-tated (data not shown), always co-localized to the same cytoplasmic bodies (Fig. 2B). Fluorescence resonance energy transfer analysis of APO-CFP and APO-YFP co-localized at cytoplasmic bodies showed that donor (APO-CFP) signal intensity typically increased 10 -20% following acceptor (APO-YFP) photobleaching. Taken together, these results indicated that cytoplasmic bodies contained APOBEC3G multimers that may represent the functional form of the protein (31, 32).
Cytoplasmic bodies did not co-localize with markers for the endoplasmic reticulum, Golgi apparatus, or endosomes (data not shown). Since they were often located in the vicinity of the nucleus, we also analyzed whether cytoplasmic bodies were associated with the centrosome. In the absence of APOBEC3G, ␥-tubulin immunolocalizes diffusely throughout the cytoplasm and in the majority of cells to a single centrosome (date not shown). In the presence of APOBEC3G, ␥-tubulin was always concentrated at cytoplasmic bodies (Fig. 2C, circles), but CFP-APO did not co-localize with the centrosome (Fig. 2C, arrows). ␥-Tubulin could not be coimmunoprecipitated with CFP-APO using standard protocols for protein solubilization (Fig. 4A), suggesting that these proteins may be part of an insoluble complex. Immunolocalization of 20 S proteasome subunit and ubiquitin revealed that both proteins were sometimes associated with cytoplasmic bodies (Fig. 2, D and E, circles); however, significant detection of either protein at cytoplasmic bodies was infrequent and limited to the relatively larger bodies. Cytoplasmic bodies were also not surrounded by a cage of the intermediate filament vimentin (Fig. 2F, circles), suggesting that these structures were different from an aggresome (Ref. 33

and see below).
Subcellular Localization of Vif-To establish where Vif localizes in 293T cells, Vif was immunolocalized in fixed cells using two different fixation methods and was localized in live cells using YFP-Vif. After fixation with paraformaldehyde, Vif localized predominantly to the nucleus in ϳ40% of cells and predominantly to the cytoplasm in ϳ60% of cells (Fig. 3A, Pf).
In contrast, after fixation with methanol, Vif localized almost exclusively to the cytoplasm in Ͼ98% of cells (Fig. 3A, MeOH). In live cells, a YFP-tagged version of Vif (YFP-Vif) exhibited a localization pattern more closely resembling Vif immunolocalization following paraformaldehyde fixation. However, their respective patterns were also different in that YFP-Vif exhibited more predominant nuclear localization in Ͼ90% of cells (Fig. 3A, Live, YFP-Vif). YFP-Vif imaged in living cells showed a similar localization pattern in HeLa cells and in non-permissive Hut78 T-cells, and this pattern was the same for either HXB2 or NL4-3 strain Vif (data not shown).
To analyze the effects of fixation on YFP-Vif localization, YFP-Vif visualized directly by YFP fluorescence was compared with YFP-Vif visualized by indirect immunostaining with ␣-Vif and ␣-GFP. Following methanol fixation, YFP-Vif localized almost exclusively to the cytoplasm, and the localization patterns observed using either YFP fluorescence or immunolocalization (␣-Vif and ␣-GFP) were indistinguishable (Fig. 3B, Methanol). Following paraformaldehyde fixation, YFP fluorescence showed YFP-Vif localized largely to the nucleus, but after immunolocalization with either ␣-Vif or ␣-GFP, more YFP-Vif appeared to be in the cytoplasm (Fig. 3B, Paraformaldehyde). These results demonstrated that the method of fixation can significantly affect the localization of Vif and suggested that cytoplasmic Vif may be more amenable to immunostaining.
To determine whether a particular region of Vif influences  Fig. 5, B, E, and C, respectively) with the exception of YFP-Vif ⌬12-23 , which was largely restricted to the cytoplasm (Fig. 5D). This analysis indicated that a specific region of Vif targets YFP-Vif to the nucleus.
Analysis of the Vif-APOBEC3G Interaction in Live Cells-To avoid the variability associated with immunolocalization, the Vif-APOBEC3G relationship was characterized using live imaging with YFP-Vif C114S serving as a non-functional control. As previously reported (27), Vif C114S was inactive against CFP-APO (Fig. 4A, Total lysate); however, Vif C114S coimmunoprecipitated with CFP-APO (Fig. 4A, IP: ␣-GFP), indicating that the loss of function associated with the C114S mutation was not due to an inability to interact with APOBEC3G. Interestingly Vif also coimmunoprecipitated with CFP-APO even after CFP-APO levels had been significantly depleted by Vif (Fig.  4A).
Having established that YFP-Vif functions against APOBEC3G, the subcellular localization of YFP-Vif and APO-CFP was examined in live cells. When pYFP-Vif and pAPO-CFP were co-transfected at a ratio of 130 to 16 fmol, respectively, a significant reduction in APO-CFP was observed relative to co-transfection with pYFP-Vif C114S . In cells where both proteins were detected, the overall APO-CFP signal was reduced, cytoplasmic bodies were rarely observed, and significant co-localization in the cytoplasm was not observed (data not shown). However, increasing the amount of pAPO-CFP could increase the number of cells visibly co-expressing YFP-Vif and APO-CFP. Upon increasing APO-CFP expression, a significant fraction of YFP-Vif was redistributed to the cytoplasm from the nucleus (Fig. 5 A). Although more YFP-Vif was present in the cytoplasm, significant co-localization of YFP-Vif and APO-CFP at cytoplasmic bodies was rarely observed (Fig. 5A, arrows). Using the same ratio of vectors, even more YFP-Vif C114S was redistributed to the cytoplasm, and frequently, YFP-Vif C114S was completely excluded from the nucleus (Fig. 5B). In further contrast to YFP-Vif, YFP-Vif C114S always co-localized with cytoplasmic bodies (Fig. 5B, circles). Upon reducing the amount of pAPO-CFP back to 16 fmol, the majority of YFP-Vif C114S remained in the nucleus, but YFP-Vif C114S and APO-CFP still co-localized at cytoplasmic bodies (data not shown), suggesting that the redistribution of YFP-Vif C114S was dependent on the cellular levels of APO-CFP. Similar to YFP-Vif C114S , YFP-Vif ⌬140 -148 co-localized with APO-CFP at cytoplasmic bodies (Fig. 5C, circles) but was not as dramatically redistributed to the cytoplasm (Fig. 5, compare B with C), which was consistent with the reduced levels of YFP-Vif ⌬140 -148 that coimmunoprecipitated with APO-HA (Fig. 4C). The localization patterns of YFP-Vif ⌬12-23 and YFP-Vif ⌬43-59 , both of which did not coimmunoprecipitate with APO-HA (data not shown, Fig. 4C), did not change in the presence of APO-CFP, and co-localization at cytoplasmic bodies was not detected (Fig. 5, D and E, arrows). Collectively, these results indicate that an interaction is required for Vif and APOBEC3G to co-localize and that the frequency of Vif-APOBEC3G co-localization is limited by the effector functions of Vif.
We used the stable nature of interactions between YFP-Vif C114S and APO-CFP in vivo as the basis of a screen to identify random mutations that disrupted Vif-APOBEC3G interactions (see "Experimental Procedures") and found the Vif I9T,C114S double mutant. I9T is encoded by a missense mutation that arose from a single base substitution in the Vif C114S coding sequence. Unlike YFP-Vif C114S , the localization pattern of YFP-Vif I9T,C114S was not altered in the presence of APO-CFP, and co-localization at cytoplasmic bodies was not detected (Fig. 6a, A and B). Furthermore YFP-Vif I9T,C114S did not coimmunoprecipitate with APO-HA (Fig. 6c). YFP-Vif harboring either the I9T or I9A single mutation did not significantly coimmunoprecipitate with APO-HA (Fig. 6c) and did not detectably affect APO-HA levels in the cell (Fig. 6b). In the presence of APO-CFP, the nuclear localization of YFP-Vif I9T and YFP-Vif I9A was not altered, and co-localization at cytoplasmic bodies was not detected (Fig. 6a, C and D). These results indicated that mutations at Vif Ile 9 disrupted Vif-APOBEC3G interactions and that these interactions were necessary for APOBEC3G to influence the localization of Vif. Subcellular Localization of Vif and APOBEC3G during Proteasome Inhibition-The subcellular localization of Vif, YFP-Vif, CFP-APO, and Vif-APOBEC3G interactions were also an- alyzed following proteasome inhibition. As observed with Vif, proteasome inhibition elevated the steady-state levels of YFP-Vif, demonstrating that the YFP tag did not interfere with the regulation of Vif through the proteasome (data not shown). Following proteasome inhibition, YFP-Vif (Fig. 7A, top panels,  arrow) and Vif (Fig. 7A, bottom panels, arrow) accumulated into a vimentin-encaged aggresome (Fig. 7B, top panels, ␣-vimentin, arrow). In addition, the majority of cells exhibited elevated levels of nuclear YFP-Vif and large nuclear aggregates (Fig. 7A, top panels, arrowheads), and a similar localization pattern was seen for Vif immunostained with ␣-Vif after paraformaldehyde fixation (Fig. 7A, bottom panels, arrowhead). Although proteasome inhibition did not significantly

HIV-1 Vif-mediated Depletion of APOBEC3G 8393
increase CFP-APO levels (Fig. 1A), CFP-APO frequently concentrated in subregions of the nucleus (Fig. 7B, left lower panel,  arrowhead) and accumulated into a vimentin-encaged aggresome (Fig. 7B, lower panels, arrows). When pNL-A1 and pCFP-APO were introduced into cells at a 1:0.25 molar ratio, respectively, co-expression of Vif and CFP-APO was rarely detected in the same cell (Fig. 1B). However, in cells where both proteins could be detected, Vif and CFP-APO exhibited diffuse localization throughout the cytoplasm (Fig.  7C, top panels, DMSO). Proteasome inhibition led to an in-crease in cytoplasmic CFP-APO levels and in the number of cells co-expressing both proteins (Figs. 1B and Fig. 7C, ALLN). Furthermore, three distinct localization patterns could be discerned after proteasome inhibition. First, in some cells, proteasome inhibition resulted in significant co-localization of CFP-APO and Vif (Fig. 7C, ALLN, arrows) or YFP-Vif within a vimentin-encaged aggresome (Fig. 7D, top panels). CFP-APO and YFP-Vif I9T also co-aggregated following proteasome inhibition (Fig. 7D, bottom panels), suggesting that co-localization to an aggresome did not require a specific interaction. Second, Vif and CFP-APO co-localized to subregions within the nucleus (Fig. 7C, ALLN), which were also observed when either protein was expressed alone (Fig. 7, A and B, arrowheads). Third, Vif and CFP-APO did not exhibit strong co-localization in the cytoplasm of some cells, and Vif remained largely localized in the nucleus (Fig. 7C, ALLN, arrowheads). DISCUSSION Analysis of Proteasome-dependent Vif-mediated Depletion of APOBEC3G-In this investigation, we studied the Vif-APOBEC3G relationship by correlating Vif function with the subcellular localization of Vif-APOBEC3G interactions. The observed effect of Vif on APOBEC3G was reported previously to be sensitive to the relative levels of Vif and APOBEC3G in the cell (22), and the extent to which Vif affects the stability of APOBEC3G has been a subject of debate (5,27). We found that the ability to observe both the depletion of APOBEC3G by Vif and the subsequent recovery of APOBEC3G expression following proteasome inhibition depended on the ratio of Vif: APOBEC3G vector used. Interestingly, at the ratio where the lowest level of recovery was observed, which also corresponded to the lowest amount of input vector used, proteasome inhibition had a negative effect on APOBEC3G expression in the absence of Vif. Considering proteasome inhibition has an array of pleiotropic effects on the cell (34), one of which is a defect in receptor-mediated endocytosis (35,36), this decrease in APOBEC3G expression when low levels of input vector are used may be linked to the effect proteasome inhibition has on liposome-mediated transfection. Taken together, these results indicated that transient expression systems can be used to study the relationship between Vif and APOBEC3G effectively, but establishing the appropriate levels of expression is crucial for studying the effects of Vif function on APOBEC3G.
Subcellular Localization of APOBEC3G-APOBEC3G localizes predominantly throughout the cytoplasm in living cells, which is consistent with the immunolocalization pattern of APOBEC3G observed here and previously (19). Our localization studies also showed that APOBEC3G concentrated with ␥-tubulin and to a lesser extent with ubiquitin and the 20 S proteasome subunit at punctate cytoplasmic bodies that formed regardless of epitope tag or expression level. Despite co-localizing with ␥-tubulin, cytoplasmic bodies were distinct from the centrosome, were not surrounded by a vimentin cage, and therefore were not aggresomes, which form in response to protein overexpression, misfolding, and proteasome inhibition (for a review, see Ref. 37). Interestingly we observed that APOBEC3G could form multimers in vivo, and APOBEC3G multimers, which presumably represent the functional form of the protein (31,32), localized to cytoplasmic bodies. The deposition of APOBEC3G into cytoplasmic bodies could serve a cytoprotective role (37), sequestering the potentially toxic cytidine deaminase away from the rest of the cytoplasm. Our proteasome inhibition studies revealed that the majority of APOBEC3G was largely resistant to ubiquitin-dependent proteasome degradation, which may explain the tendency for APOBEC3G to aggregate. Similar sequestration into cytoplasmic inclusions was also documented for a proteasome-resistant form of tumor suppressor p53 (38). Although the specific function(s) of these cytoplasmic bodies requires further investigation, these bodies can serve as an intracellular marker of the Vif-APOBEC3G interaction. Furthermore co-localization of APOBEC3G and ␥-tubulin raises the intriguing possibility that the microtubule network links APOBEC3G to the HIV-1 life cycle (39).
Subcellular Localization of HIV-1 Vif-Our live imaging studies revealed that while clearly localized to both compartments, YFP-Vif localized predominantly to the nucleus of 293T cells. Nuclear localization occurred regardless of the source of Vif (HXB2 or NL4-3 strains) and in both permissive (293T and HeLa) and non-permissive (Hut78) cell lines. Furthermore our mutant studies showed that amino acids 12-23 of Vif were important for nuclear localization. Using immunostaining, Vif was shown previously to localize predominantly in the cytoplasm and was only modestly observed in the nucleus, which was a phenotype that was sometimes attributed to overexpression (29,40,41). However, similar to feline immunodeficiency virus Vif (42), we found that cell fixation with the protein cross-linking agent paraformaldehyde did not alter the nuclear localization of YFP-Vif observed in living cells, but rapid fixation with methanol redistributed nuclear YFP-Vif to the cytoplasm. NL-A1 Vif immunolocalization after fixation was similar to YFP-Vif localization except that significant nuclear localization was observed less frequently following paraformaldehyde fixation. This discrepancy was explained when more YFP-Vif appeared to localize to the cytoplasm when immunolocalized with either ␣-Vif or ␣-GFP, which suggests that cytoplasmic Vif is more amenable to immunostaining. Altogether these results demonstrated that a more substantial fraction of Vif localizes to the nucleus than described previously and are consistent with previous findings that show Vif interacts with nuclear proteins Sp140 (43) and ZIN (44).
Correlating Vif Function with Subcellular Localization of Vif-APOBEC3G Interactions-We discovered that inactive Vif C114S interacted with APOBEC3G but did not function. Similarly, the SLQ(Y/F)LA domain is required for function but not for Vif-APOBEC3G interactions, which was shown here and by others previously (19,(21)(22)(23). The characterization of these two mutants suggested that interactions with APOBEC3G were not sufficient for Vif to function against APOBEC3G and that a specific effector function was mediated through Cys 114 after Vif and APOBEC3G interacted. Since Vif is already a target for proteasome degradation in the absence of APOBEC3G (this study and Ref. 45), an attractive model for Vif function is that a direct interaction with APOBEC3G leads to the co-degradation of both proteins; however, data supporting this type of mechanism have been controversial (23,45). Importantly, similar to wild type Vif, the steady-state levels of Vif C114S were elevated after proteasome inhibition (data not shown), suggesting that this mutant was a target for proteasome degradation. Therefore, Vif C114S interacted with APOBEC3G and was a target for proteasome degradation but did not affect APOBEC3G levels, arguing against a co-degradation mechanism and supporting a role for Vif as an adaptor molecule.
Functional and non-functional Vif-APOBEC3G interactions were correlated with subcellular localization in live cells using several Vif mutants. Even under conditions that resulted in significant depletion of APOBEC3G by Vif, co-expression of the proteins could be visualized within the same cell, and the number of cells exhibiting co-expression could be increased by altering the vector ratio of Vif:APOBEC3G. Live imaging showed that increasing the amount of APO-CFP increased the amount of YFP-Vif that localized to the cytoplasm, indicating that expression of APOBEC3G influenced the localization of YFP-Vif. These results also suggested that Vif and APOBEC3G interacted in the cytoplasm and that Vif localization was influenced by interactions with other proteins. Strikingly, YFP-Vif and APO-CFP rarely co-localized to cytoplasmic bodies even when more YFP-Vif appeared to co-localize with APOBEC3G in the cytoplasm. However, both YFP-Vif C114S and YFP-Vif ⌬140 -148 completely co-localized with APO-CFP throughout the cytoplasm and most notably to cytoplasmic bodies, suggesting that effector function may preclude Vif from co-localizing to cytoplasmic bodies. In the presence of APO-CFP, YFP-Vif C114S in particular was often excluded from the nucleus, indicating that in the absence of function Vif and APOBEC3G formed stable complexes in the cytoplasm that could prevent Vif from localizing to the nucleus. Co-localization of YFP-Vif C114S and APO-CFP in the cytoplasm and at cytoplasmic bodies was used as the basis for a screen in which we isolated the Vif I9T,C114S double mutant, which did not interact with APOBEC3G. We further discerned that mutations of Ile 9 alone were sufficient to impair coimmunoprecipitation and co-localization with APOBEC3G, suggesting that the N terminus of Vif mediates interactions with APOBEC3G.
The reduced level of co-localization observed between YFP-Vif and APO-CFP at cytoplasmic bodies suggested that Vifmediated depletion of APOBEC3G may not be limited to a particular subregion of the cell but rather took place throughout the cytoplasm. The lack of co-localization at cytoplasmic bodies may also reflect the transient nature of functional Vif-APOBEC3G interactions, and Vif may mediate the depletion of APOBEC3G before APOBEC3G concentrates at cytoplasmic bodies. Since a loss in Vif effector function led to the formation of stable cytoplasmic complexes, we postulated that the functional Vif-APOBEC3G interaction could be visualized by using proteasome inhibition to block Vif function. In the presence of Vif, a clear rise in the overall cytoplasmic levels of APOBEC3G was observed following proteasome inhibition, and the two proteins co-accumulated into a vimentin-encaged aggresome. However, both Vif and APOBEC3G accumulated into aggresomes when expressed alone, and YFP-Vif I9T and CFP-APO also accumulated into aggresomes following proteasome inhibition. Therefore, co-localization at aggresomes did not necessarily require or result from a specific interaction and may be an indirect consequence of proteasome inhibition. The formation of an aggresome is a documented cellular response to proteasome inhibition (33), and the significant effects of proteasome inhibition on the architecture of the cytoplasm indicate that co-localization and interactions observed under these conditions may not be specific. Independently of APOBEC3G, Vif levels and nuclear localization increased in response to proteasome inhibition, and Vif was sequestered into both nuclear and cytoplasmic aggregates, suggesting that Vif may be targeted for degradation in the nucleus, cytoplasm, or both. This result raises the possibility that proteasome inhibition may serve to inhibit Vif function through aberrant localization or sequestration. Overall these studies demonstrated that proteasome inhibition has profound effects on the localization of Vif that must be taken into account when using proteasome inhibition to study Vif functions against APOBEC3G.
In conclusion, this work delineated the relationship between Vif-APOBEC3G interactions, co-localization, and Vif functions. How Vif functions to accelerate the degradation of APOBEC3G still remains to be elucidated, but further studies of the role of Cys 114 in this process should provide significant insight into Vif effector functions. Further studies are also needed to determine whether Ile 9 is directly involved in binding to APOBEC3G, but this novel finding provides a potential first step toward the therapeutic disruption of the Vif-APOBEC3G interaction.