Inducible APOBEC3G-Vif Double Stable Cell Line as a High-Throughput Screening Platform To Identify Antiviral Compounds

Abstract

Inhibition of the interaction of the human cytidine-deaminase APOBEC3G (A3G) with the human immunodeficiency virus (HIV) type 1-specific viral infectivity factor (Vif) represents a novel therapeutic approach in which a cellular factor with potent antiviral activity (A3G) plays a key role. In HIV-infected cells, the interaction of Vif with A3G leads to the subsequent degradation of A3G by the 26S proteasome via the ubiquitin pathway and to the loss of antiviral activity. To establish a stable and convenient cellular testing platform for the high-throughput screening of potential antiviral compound libraries, we engineered a double transgenic cell line constitutively expressing an enhanced yellow fluorescent protein expressor (EYFP-A3G) fusion as well as a Tet-Off controllable Vif protein. With this cell line, we were able to measure precisely the Vif-induced degradation of A3G in the presence of potential antiviral compounds in an easy-to-handle, robust, and practical high-throughput multiwell plate format with an excellent screening window coefficient (Z factor) of 0.67.

Despite the availability of highly active antiretroviral therapy with the 25 anti-human immunodeficiency virus (anti-HIV) drugs that have been approved over the last two and a half decades (12), HIV infection or AIDS accounts for more than 3 million deaths worldwide every year. Several classes of anti-HIV compounds targeting essential steps of the viral replication cycle are currently in use, e.g., viral entry and coreceptor binding, reverse transcription, provirus integration, as well as protease-mediated maturation (12, 16). However, despite the availability of these potent anti-HIV compounds, the development of new and improved antivirals remains a key feature in controlling the disease due to the emergence of resistant HIV strains (16).

Besides targeting the HIV replication cycle, the natural cellular resistance factors that are part of the intrinsic immunity and their interaction with viral proteins play important roles in controlling the infection. For instance, cellular protein TRIM5α acts as a natural defense factor, in that it hinders the viral uncoating process (8, 25, 34, 37, 43) before reverse transcription takes place. In hematopoietic stem cells, which are resistant to HIV infection, cyclin-dependent kinase inhibitor p21, which is responsible for regulation of the size of the stem cell pool, protects cells by aborting the chromosomal integration of the provirus (51). Another important antiviral defense factor is the human apolipoprotein B mRNA-editing enzyme catalytic polypeptide (APOBEC) (39), which inhibits the replication of retroviruses and hepatitis B virus (10, 44). Currently identified are eight members of the APOBEC superfamily of RNA/DNA cytitidin deaminases with antiretroviral activity (3, 4, 10, 22), termed APOBEC3A to APOBEC3H (20). During HIV particle assembly, APOBEC3G (A3G) is specifically incorporated into virions via interaction with the Gag nucleocapsid protein (1, 6, 42). When reverse transcription takes place, A3G mediates the deamination of cytidine (C) to uridine (U) residues on the first minus-strand viral cDNA. This activity leads to a guanosine (G)-to-adenosine (A) hypermutation in the viral plus-strand DNA (4, 18, 29, 33, 50) and to the loss of viral replicating activity. Approximately seven copies of A3G are incorporated into a viral particle and are enough to block replication in the following round of infection (47). However, hypermutating the viral genome by cytidine deamination might not be the sole antiviral activity of A3G. It was also found that A3G and A3F interfere with reverse transcription and proviral DNA formation in a so far not fully characterized and cytidine deamination-independent mechanism (3, 17, 19, 28, 30, 36).

A counteracting protein to A3G is the HIV-encoded accessory protein viral infectivity factor (Vif) (9, 11). During HIV infection, the Vif protein directly binds to A3G, subsequently leading to its polyubiquitination and degradation by the 26S proteasome (7, 34, 40, 42). Vif-dependent A3G degradation depends on the ability of Vif to interact with the cellular proteins Cullin5, elongins B and C, and Rbx1 to form an Skp1-Cullin5-F box complex, which then induces the ubiquitination and degradation of A3G (48). Upon the formation of the degradation complex, the N-terminal part of Vif binds to the N-terminal part of A3G, whereas the C-terminal part of Vif interacts with Cullin5 and the elongins (5, 7, 31, 45, 48). Besides inducing proteasomal degradation, Vif also impairs the translation of A3G mRNA (42). This dual neutralizing activity leads to the successful depletion of A3G in HIV-infected cells, allowing the virus to retain its full infectivity (16).

The A3G-Vif interaction has the potential to be a platform for the development of novel therapeutics, in which the stability of a cellular antiviral factor (A3G) during viral infection plays a key role. It was shown by flow cytometric analysis that a transiently expressed N-terminal fusion between the green fluorescent protein (GFP) and A3G (GFP-A3G) can be used as a testing platform to quantify the degradation of A3G in the presence of coexpressed Vif (7). Recently, the transient coexpression of a yellow fluorescent protein (YFP) A3G variant (YFP-A3G) together with Vif was used as a screening platform to identify from a compound library a low-molecular-weight compound that inhibited the Vif-induced proteasomal degradation of YFP-A3G (35). However, we developed a significantly improved, stable, and controllable screening platform for the identification of antivirals that inhibit the Vif-induced proteasomal degradation of a EYFP-A3G fusion protein. For this purpose, we engineered a double stable cell line expressing EYFP-A3G as well as a Vif protein under the control of the bacterium-derived Tet-Off system (15).

The rationale for the use of a controllable Vif was twofold. First, with a Vif-inducible screening platform, we were able to precisely demonstrate the functionability of the system (the induction of Vif leads to a low level of expression of the EYFP-A3G expressor, whereas the repression of Vif leads to a high level of the EYFP-A3G expressor). Second, we were able to define the signal dynamic range, that is, the screening window of our system without a potential Vif antagonist. This enabled us to precisely evaluate the high-throughput screening potential of our cell line. This cell line features great advantages over any transient screening platform, in that (i) it can be used right away (no transfection of the interaction partners is necessary), (ii) it is a highly reproducible system, (iii) the system is exactly controllable, and (iv) the use of the system is cost-efficient.

MATERIALS AND METHODS

Cell lines and media.

HEK293T cells (ATCC CRL-11268) were used to generate all mammalian cell-derived cell lines. These cells were cultivated in Dulbecco's modified Eagle medium supplemented with 10% fetal calf serum and a standard tissue culture penicillin-streptomycin mixture in a humidified incubator under 5% CO₂ at 37°C. For clonal selection experiments and gene induction, the medium was supplemented with 400 ng/ml of puromycin (InvivoGen) or doxycycline (Clontech). The 26S proteasomal inhibitor MG132 (Sigma/Aldrich) was used for the proteasome inhibition experiments. For experiments with MG132, doxycycline, or RN-18 (see below), cells were cultivated without puromycin. Cells were grown on Nunc (Thermo Fisher Scientific) tissue culture plates and flasks.

Construction of plasmid vectors.

To create the Tet-Off controllable Vif expressor, plasmid pWPXL (plasmid 12257; Addgene) was used as the HIV-based retroviral backbone. In order to integrate the complete Tet-Off regulon into the viral backbone, two intermediate vectors were made. First, we created the intermediate vector pWPXL-Tet-Off-Advanced by PCR amplification of the tTA-Advanced region out of pTet-Off-Advanced (Clontech) with primers containing a 5′ BamHI restriction site and a 3′ MluI restriction site for subcloning into pWPXL via BamHI and MluI. Second, we created the intermediate vector pTRE-Tight-Vif by PCR amplification of Vif out of the HXB3 HIV molecular clone (bacteriophage λ HXB3 was from obtained from Beatrice Hahn and George Shaw through the AIDS Research and Reference Reagent Program) with primers containing a 5′ KpnI restriction site and a 3′ NotI restriction site for subcloning into pTRE-Tight (Clontech) via KpnI and NotI. This vector was used as the PCR template to amplify the pTRE-Tight-Vif cassette with primers containing a 5′ MluI restriction site and a 3′ SpeI restriction site for subcloning into the vector pWPXL-Tet-Off-Advanced via MluI and SpeI, which created the final vector, pWPXL-Tet-Off-Advanced-TRE-Tight-Vif (pWPXL-TTT-Vif).

To create EYFP-A3G, the vector pIRES-puro (Clontech) was used as the backbone. The EYFP coding region was amplified by PCR with pEYFP (Clontech) as the template and with primers containing a 5′ ClaI restriction site and a 3′ EcoRI restriction site. The PCR insert was subcloned into pIRES-puro via ClaI and EcoRI to create pIRES-EYFP. Subsequently, an N-terminal A3G PCR fragment containing 5′ EcoRI and AgeI restriction sites and a 3′ BamHI restriction site was inserted into pIRES-EYFP via EcoRI and BamHI to create pIRES-EYFP-A3GΔC. To make the full-length A3G, the C-terminal fragment containing 5′ and 3′ BamHI restriction sites was subcloned into pIRES-EYFP-A3GΔC via BamHI to create pIRES-EYFP-A3G. The A3G template DNA was a generous gift from K. Bishop (King's College, London School of Medicine, London, United Kingdom). To provide flexibility between the two neighboring proteins, we introduced a synthetic double-stranded oligonucleotide linker coding for 17 amino acids RIPCKIPNDLKQKVMTG (in accordance with the information provided previously [21]) into the EcoRI and BamHI sites of pIRES-EYFP-A3G. The sequences of all clones were verified by sequence analysis.

Transient transfections.

Linear polyethyleneimine with a molecular weight of 25,000 (Polyscience) was used as the transfection reagent, as described previously (23). Briefly, expression plasmids and polyethyleneimine (1 mg/ml) were mixed at a ratio of 1:2 (wt/vol) in Hanks balanced salt solution (Sigma/Aldrich) and added to HEK293T cells, which had been seeded at a density of 5.7 × 10⁴ cells/cm².

Pseudotyped lentiviral particles.

To produce viral particles, we used either a three-plasmid system consisting of the HIV-based retroviral vector backbone pWPXL, the packaging plasmid psPAX2 (plasmid 12260; Addgene), and the vesicular stomatitis virus envelope (VSV-G) expressor pMD2.G (plasmid 12259; Addgene) or a two-plasmid system consisting of the pGJ3-EGFP expressor retroviral vector (26) and pMD2.G. Transient transfections were used to express the two- or three-plasmid system in HEK239T cells for the production of viral particles. At 48 h and 72 h posttransfection, the viral supernatant was harvested by filtration through a 0.45-μm-pore-size cellulose acetate syringe filter and concentrated to 1/10 of the initial volume by using Amicom ultrafiltration concentrators with a pore size of 100 kDa (Millipore). The concentrated viral supernatant was used for the subsequent infection of target cells.

Cell lines.

The cell line expressing Vif under the control of the Tet-Off regulon as well as EYFP-A3G (293T^{Vif Tet-Off/EYFP-A3G} cells) was engineered in two steps. First, the vector pWPXL-TTT-Vif was used to produce viral particles for use for subsequent infection: 1 × 10⁶ HEK293T cells were seeded into a 10-cm culture dish and incubated overnight. The cells were then infected with 0.3 ml of concentrated viral supernatant in a total culture volume of 10 ml. On the following day, the virus-containing medium was exchanged for regular medium and the cells (293T^{Vif Tet-Off} cells) were cultivated under normal conditions. In the second step, we integrated the EYFP-A3G-IRES-puro expression cassette (Fig. 1A). To facilitate the stable integration of the transgene into 293T^{Vif Tet-Off} cells, the expression plasmid pIRES-puro-EYFP-A3G was linearized with the restriction enzyme Eam1105I (Fermentas). The linearized vector was then transfected into 293T^{Vif Tet-Off} cells (see above). After 24 h, the cells were cultivated under selective pressure with 400 ng/ml of puromycin (InvivoGen). The cells were kept under puromycin selection during normal cultivation. At 7 days posttransfection, the population with low levels of EYFP expression (EYFP^low) was isolated by fluorescence-activated cell sorting (FACS) (Fig. 1B) with a FACSVantage SE (BD) flow cytometer. To select single-cell clones, the EYFP^low population was seeded into a 24-well tissue culture plate at a clonal density with an average number of 0.25 cells per cm². When they were confluent, these clones (293T^{Vif Tet-Off/EYFP-A3G} single-cell clones) were further analyzed by flow cytometry for their ability to respond to doxycycline treatment with an increase in the level of EYFP-A3G fluorescence. One clone showing the strongest doxycycline-dependent upshift in EYFP-A3G fluorescence was selected and used for all subsequent experiments.

FIG. 1.

The HIV-1 Vif and EYFP-A3G expressors, selection of 293T^{Vif Tet-Off/EYFP-A3G} cells, and intracellular localization of EYFP-A3G. (A) Schematic drawing of the Vif and A3G expression cassettes. (Upper panel) Viral vector showing the bicistronic features of the Tet-Off-controlled Vif construct; (lower panel) pIRES-puro EYFP-A3G expression vector. P_CMV, cytomegalovirus promoter; P_EF-1α, elongation factor 1α promoter; tTA² and Ptight, Tet transactivator and promoter, respectively; IRES, internal ribosomal entry site; Puro, puromycin resistance gene; LTR, retroviral long-terminal-repeat sequences; dashed lines, EYFP low and high cutoff values. (B) Selection of 293T^{Vif Tet-Off/EYFP-A3G} cells responding to doxycycline treatment with an increase in EYFP fluorescence and flow cytometric analysis of doxycycline-treated cultures during selection. Rectangle, cells isolated by FACS. (C) Quantification of the mean fluorescence intensities from panel B. Dox, doxycycline; MFI, mean fluorescence intensity. (D) Intracellular localization of EYFP-A3G in 293T^{Vif Tet-Off/EYFP-A3G} cells. Cells were cultivated for 5 days in the presence of 10 ng/ml doxycycline. A live cell culture was then subjected to fluorescence microscopy. Arrowheads, localization of p bodies; N, nucleus.

Gel electrophoresis and immunoblotting.

Denaturing sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) and immunoblotting were done by standard procedures. Briefly, cells were harvested by treatment with 0.05% (wt/vol) EDTA in phosphate-buffered saline containing no MgCl₂ or CaCl₂ but containing 0.25% (wt/vol) trypsin. Approximately 6 × 10⁶ cells were lysed in 100 μl nondenaturing cell lysis buffer containing 20 mM HEPES (pH 7.4), 0.1% Triton X-100, and 1 mM dithiothreitol supplemented with a protease inhibitor cocktail (Roche). Cell lysates containing equal amounts of total protein were mixed with an equal volume of 2× SDS sample buffer, boiled, and separated on a 12% SDS-polyacrylamide gel. For immunoblot analysis of the viral particles, concentrated viral supernatant was layered over a cushion containing 20% sucrose in Hanks balanced salt solution (Sigma/Aldrich) and ultracentrifuged at 275,000 × g in a TST60.4 rotor (Sorvall) for 1.5 h at 4°C. Subsequently, the supernatant and the sucrose cushion were carefully aspirated. The sedimented viral particle fraction was dissolved in SDS sample buffer for subsequent SDS-PAGE. The separated proteins were electroblotted onto nitrocellulose, which was blocked for 1 h at room temperature with Tris-buffered saline containing 5% nonfat dry milk powder and 0.1% Tween 20. The membrane was incubated for 1 h at room temperature with primary antibodies diluted 1:1,000 to 1:3,000 in Tris-buffered saline containing 5% nonfat dry milk powder and 0.1% Tween 20. Detection was carried out with horseradish peroxidase-labeled immunoglobulin G and an enhanced chemiluminescent detection system (Santa Cruz Biotechnology).

Sequence analysis of proviral DNA.

HIV-based EGFP reporter gene viruses were produced in the presence of cotransfected wild-type (wt) A3G or EYFP-A3G. These viruses were used to infect HeLa cells. At 2 days postinfection, the HeLa cells were harvested for the subsequent isolation of genomic DNA. To amplify the transduced EGFP gene, genomic DNA samples were used as PCR templates with flanking primers containing a 5′ AgeI restriction site and a 3′ BamHI restriction site for subcloning into the vector pIRES. Sequence analysis was performed with pIRES clones (n = 10) harboring the transduced EGFP gene.

Quantification of fluorescence signals.

For flow cytometric quantification of GFP and EYFP fluorescence, cells were harvested by trypsin treatment. Single-cell suspensions were dissolved in phosphate-buffered saline and subjected to flow cytometric analysis with a FACS-Calibur (BD) flow cytometer. For quantification of the fluorescence on multiwell plates, cells grown on 96-well plates were harvested by removing the growth medium, followed by addition of cell lysis buffer. Fluorescence was quantified in a Safire (Tecan) plate reader (excitation wavelength, 510 nm; emission wavelength, 535 nm).

Synthesis of RN-18.

2-(4-Nitrophenylthio)benzoic acid (200 mg, 0.73 mmol) was suspended in 10 ml dichloromethane. Triethylamine (221 mg, 2.2 mmol) was added, and the solution was mixed with 2-methoxyaniline (179 mg, 1.45 mmol; Sigma/Aldrich). The reaction mixture was cooled to 0°C, and POCl₃ (334 mg, 2.2 mmol) was added dropwise. After the reaction mixture was stirred for 1 h at 0°C, it was stirred at room temperature for 3 days. The dichloromethane phase was then washed with saturated sodium bicarbonate solution, and the aqueous phase was extracted with dichloromethane. The combined organic phases were dried with sodium sulfate, and the solvent was removed in vacuo. The crude product was purified by column chromatography (silica gel; cyclohexane-ethylacetate, 4/1) to 94% purity. The analytical data for the purified product (yield, 72.4%; 200 mg) corresponded to the values in the literature (38). For the inhibition experiments, RN-18 was dissolved in dimethyl sulfoxide and applied to the tissue culture medium. The final dimethyl sulfoxide concentration in the medium was 0.5%.

RESULTS

Engineering a stable cell line expressing EYFP-A3G and Tet-Off-controlled Vif.

The 293T^{Vif Tet-Off/EYFP-A3G} stable cell line was engineered in two steps. First, we cloned a bicistronic HIV-based vector harboring the complete Tet-Off regulatory elements, which control the doxycycline-dependent expression of HIV type 1 (HIV-1) Vif (Fig. 1A). With this vector we produced viral particles, which we used to infect HEK293T cells. To verify the controllability of the system, infected cells were subjected to SDS-PAGE and immunoblotting to assess the expression of Vif in the presence and the absence of doxycycline (data not shown). In the second step, the 293T^{Vif Tet-Off} cells were used as target cells for the integration of the EYFP-A3G-IRES-puro expression cassette (Fig. 1A) via an Eam1105I-linearized plasmid to facilitate integration of the vector DNA into 293T^{Vif Tet-Off} cells. The cells were then subjected to FACS and seeded at a clonal density sufficient to obtain single-cell clones responding to a doxycycline-dependent increase of EYFP-A3G fluorescence (Fig. 1B and C). One clone showing the strongest upshift in EYFP-A3G fluorescence after addition of doxycycline was chosen for the subsequent analysis. This clone showed the typical cytoplasmic distribution of A3G with the appearance of large spots (Fig. 1D, arrowheads), most likely representing processing bodies (p bodies) (13). There was almost no visible nuclear localization of EYFP-A3G.

EYFP-A3G expression is Vif dependent and can be upregulated with doxycycline.

To assess the proper expression characteristics of Vif and EYFP-A3G, 293T^{Vif Tet-Off/EYFP-A3G} cells were analyzed by immunoblotting and flow cytometry. Immunoblot analysis showed that the expression of Vif was strictly doxycycline dependent. Addition of doxycycline led to an almost complete reduction in the Vif protein level after a 24-h expression period, whereas no detectable Vif was present in cell extracts from cultures cultivated for 48 h, 72 h, or 96 h (Fig. 2A). A doxycycline concentration between 0.1 and 10 ng/ml was sufficient to upregulate the expression of EYFP-A3G, whereas no further increase in fluorescence could be observed under the presence of higher concentrations (data not shown). We decided to use a doxycycline concentration of 10 ng/ml for all subsequent experiments. The doxycycline-dependent expression of Vif showed an inverse correlation with the levels of EYFP-A3G, in that the levels of the latter strongly declined when Vif was induced (Fig. 2A and B). In contrast to the almost complete downregulation of Vif upon doxycycline treatment after a 24-h incubation period (Fig. 2A), the level of EYFP-A3G declined after 24 h and 48 h and was still detectable after 72 h, even when Vif was present (Fig. 2B). No EYFP-A3G was detectable after 96 h when Vif was maximally induced. Flow cytometric quantification showed a clear upshift in EYFP-A3G fluorescence upon addition of doxycycline (inhibition of Vif expression) to the culture medium. In the presence of doxycycline, 293T^{Vif Tet-Off/EYFP-A3G} cells showed a clear upregulation of the EYFP-A3G fluorescence within a 24-h cultivation period. The signal peaked after 48 h, whereas longer cultivation periods (96 h) did not yield a higher fluorescence signal (Fig. 2C). Upon the removal of doxycycline, the EYFP-A3G level declined stepwise during the cultivation period and reached its lowest level after 96 h (Fig. 2D). This level corresponded to the signal in Fig. 2A and C (at 0 h) when doxycycline was added at the beginning of the cultivation period and reflected the maximum downregulation of EYFP-A3G by Vif. These results demonstrated that within our cell line (i) the EYFP-A3G fusion protein displayed the correct subcellular localization, (ii) the expression of Vif was tightly controlled by doxycycline, and (iii) EYFP-A3G was strongly downregulated when Vif was induced (in the absence of doxycycline).

FIG. 2.

Expression of Vif is tightly regulated by doxycycline and shows an inverse correlation with the level of EYFP-A3G. (A) 293T^{Vif Tet-Off/EYFP-A3G} cells were cultivated for the indicated times in the presence of 10 ng/ml doxycycline (Dox). Equal amounts of each cell lysate were subjected to SDS-PAGE and subsequent immunoblot analysis with specific antibodies against Vif, GFP (for the detection of EYFP-A3G), and the housekeeping protein glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Glyceraldehyde-3-phosphate dehydrogenase was included as a control for equal protein loading. (B) Cells were first cultivated for 96 h in the presence of doxycycline (10 ng/ml). The doxycycline was then removed and cells were further cultivated for the indicated times. Cells were treated as described in the legend for panel A. (C and D) Flow cytometric quantification of cells cultivated as described in the legends for panels A and B. Gray curves, value at 48 h marked for better visibility.

EYFP-A3G is specifically incorporated into virions and impairs expression of a retrovirally transduced GFP reporter through deamination which can be inhibited by induction of Vif.

To further characterize the proper function of the EYFP-A3G fusion protein, we assessed if it was properly packed into virions of HIV and if it could impair the expression of a reporter gene in infected target cells by proviral cDNA deamination. For this purpose, we generated HIV-based EGFP-reporter gene viruses using regular 293T cells as producer cells in the presence of increasing amounts of cotransfected wt A3G and EYFP-A3G. These viruses were used to infect HeLa cells to quantify the expression of the EGFP reporter by flow cytometric analysis. Figure 3A shows that the use of increasing amounts of the cotransfected expression plasmid gave correlating amounts of wt A3G and EYFP-A3G protein levels in 293T producer cells, whereas wt A3G showed a slightly enhanced overall level of expression. The protein levels of the EGFP reporter did not vary significantly. We then analyzed the viral preparation by immunoblotting. In this assay, only wt A3G and EYFP-A3G and not the EGFP reporter were packed into virions (Fig. 3A, virions). These virions were used to infect HeLa cells for quantification of the fluorescence of the transduced EGFP reporter by flow cytometry. The rates of infection by the virions produced in the absence of A3G or EYFP-A3G were used as the reference values (100% infectivity). Figure 3B shows that increasing amounts of cotransfected wt A3G or EYFP-A3G resulted in a decrease in viral infectivity, as displayed by the reduction in the level of EGFP-positive HeLa target cells. Compared with the results of immunoblot analysis shown in Fig. 3A, similar levels of wt A3G and EYFP-A3G (Fig. 3A, virions; compare lanes 2 and 6) gave almost identical levels of reduction in virion infectivity when infectivity was analyzed by flow cytometry (Fig. 3B, compare bars 2 and 6).

FIG. 3.

EYFP-A3G is specifically incorporated into virions and impairs the expression of a retrovirally transduced EGFP reporter gene, which can be rescued by induction of Vif. (A) 293T cells were used to produce a VSV-G pseudotype HIV-EGFP-reporter gene virus in the presence of increasing amounts of a cotransfected wt A3G or EYFP-A3G expression plasmid in the absence of Vif. Virus-producing 293T cells were harvested, lysed, and subjected to SDS-PAGE and immunoblotting with the indicated antibodies (293T producer cells). EGFP-reporter gene virions were harvested by ultracentrifugation and subjected to SDS-PAGE and immunoblotting as described above for the virions. (B) Virions were used to infect HeLa cells for subsequent flow cytometric analysis of EGFP reporter gene expression. Infection rates (percentage of EGFP-positive cells) of the control virions produced in the absence of A3G were set to equal 100%. (C) 293T^{Vif Tet-Off/EYFP-A3G} cells were used to produce a HIV-EGFP-reporter gene virus as described above in the absence and the presence of 10 ng/ml doxycycline (Dox), as indicated. Immunoblotting was performed as described in the legend to panel A. (D) Flow cytometric analysis of EGFP reporter gene expression in HeLa cells infected with virions from the assay whose results are shown in panel C. The infection rates of the control virions produced in regular 293T cells were set equal to 100%. The levels of glyceraldehyde-3-phosphate dehydrogenase and p24 indicate that equal amounts of sample were loaded. The values in panels B and D are the means of three independent samples. Error bars, SDs.

We also wanted to assess if the activity of EYFP-A3G described above could be inhibited by the induction of Vif. To test this, we generated EGFP-reporter gene viruses using cells of the 293T^{Vif Tet-Off/EYFP-A3G} line as producer cells in the presence and the absence of doxycycline. Viruses were used to infect HeLa cells to quantify the expression of the EGFP reporter by flow cytometric analysis. Figure 3C shows that all proteins were expressed at high levels in the 293T^{Vif Tet-Off/EYFP-A3G} producer cells. When virion samples were immunoblotted, however, we did detect a very faint EGFP signal in the preparation (Fig. 3C, virions). Compared to the EGFP signal of the producer cells, this signal most likely reflects negligible cross-contamination which occurred during preparation of the virion. These virions were used to infect HeLa cells for flow cytometric analysis of the transduced EGFP reporter, as described above. Virions produced in the absence of doxycycline (which induced Vif) displayed almost full infectivity, as shown by an EGFP reporter signal for approximately 80% of the target cells (Fig. 3D). On the other hand, HeLa cells infected with virions produced in the presence of doxycycline (which did not induce Vif) showed only low levels of infectivity, in that approximately only 20% of the target cells expressed the EGFP reporter (Fig. 3D). The low percentage of target cells infected with virions containing EYFP-A3G strongly correlated with the deaminase activity and the isolation of hypermutated proviral EGFP DNA from infected HeLa cells. Sequence analysis of the proviral EGFP reporter (n = 10) showed a total of 52 spots for wt A3G (Fig. 4A) and 55 spots for EYFP-A3G (Fig. 4B), in which a typical G → A mutation took place. In both cases, mutations accumulated within the middle of the EGFP sequence, between bases 253 and 442 (Fig. 4). These mutations were the result of wt A3G and EYFP-A3G deaminase activity and showed a great deal of similarity to those observed in a previous study (18).

FIG. 4.

The deaminase activity of EYFP-A3G introduces G → A mutations into the retroviral transduced EGFP reporter. 293T cells were used to produce a VSV-G pseudotype HIV-EGFP-reporter gene virus in the presence of a cotransfected wt A3G (A) or wt EYFP-A3G (A) expression plasmid and the absence of Vif. Virions were used to infect HeLa cells for analysis of the proviral DNA sequence (n = 10). Dots, positions of G → A mutations (in boldface with underscores).

Induction of Vif leads to ubiquitin-dependent degradation of EYFP-A3G.

To verify that the downregulation of EYFP-A3G through Vif is correlated with the ubiquitin-dependent proteasomal pathway, we checked for an induced polyubiquitination signal in the presence and the absence of doxycycline. Polyubiquitination was induced when Vif was expressed in the absence of doxycycline (Fig. 5A, lane −Dox [without doxycycline] and anti-Vif) and led to a subsequent decrease in the EYFP-A3G protein level and an increase in the polyubiquitination signal (antiubiquitin). The weak ubiquitin signal in the presence of doxycycline (Fig. 5A, lane +Dox [with doxycycline], antiubiquitin) most likely reflects the normal, Vif-independent ubiquitination/degradation of EYFP-A3G. In the following experiment, we tested the 26S proteasome inhibitor MG132, which counteracts the Vif-induced degradation of EYFP-A3G (7). We cultivated 293T^{Vif Tet-Off/EYFP-A3G} cells for 17 h in the presence of 0.625 μM MG132 and the absence of doxycycline (which induces Vif). Under these conditions, MG132 sufficiently inhibited the Vif-induced decrease of EYFP-A3G fluorescence (Fig. 5B), which normally correlates with ubiquitination (Fig. 5A) and the subsequent proteasomal degradation of the fusion protein.

FIG. 5.

Ubiquitination and degradation of EYFP-A3G is Vif dependent and can be blocked by MG132. (A) 293T^{Vif Tet-Off/EYFP-A3G} cells were cultivated for 72 h in the presence and the absence of doxycycline (Dox). Equal amounts of cell lysates were subjected to SDS-PAGE and subsequent immunoblot analysis with a ubiquitin-specific antibody. (B) Treatment with the proteasomal inhibitor MG132 blocks the Vif-induced degradation of EYFP-A3G. 293T^{Vif Tet-Off/EYFP-A3G} cells were cultivated for 17 h with MG132 at 0.625 μM in the absence of doxycycline and were subjected to flow cytometric analysis. Control (gray curve), cells not treated with MG132.

The Vif inhibitor RN-18 leads to upregulation of EYFP-A3G in the presence of Vif.

To assess the potential of our cell line to serve as a stable platform for the screening of antiviral compounds, we tested the effect of the newly discovered low-molecular-weight Vif inhibitor RN-18 (35) on the upregulation of EYFP-A3G fluorescence. For this purpose, we chose a 96-multiwell plate assay in order to demonstrate the high-throughput capabilities of our system. 293T^{Vif Tet-Off/EYFP-A3G} cells were seeded into 96-well plates and incubated for 48 h in the presence of doxycycline to switch off the expression of Vif. Doxycycline was then removed (which induced Vif), and RN-18 was added in serial dilutions over a concentration range of from 30 μM to 0.46 μM. The cells were incubated for another 24 h. This time period is sufficient to induce the Vif-dependent proteasomal degradation of EYFP-A3G (Fig. 2B). With this setup, we were able to determine that RN-18 had a half-maximal inhibitory concentration (50% effective concentration [EC₅₀]) of 5.8 μM for Vif (Fig. 6). We also wanted to validate the potential of the 293T^{Vif Tet-Off/EYFP-A3G} cell line to function as a high-throughput screening platform and calculated the Z factor as a measure of the particular screening window coefficient (52). The Z factor is a dimensionless statistical parameter used to evaluate the quality of high-throughput screening systems. It combines the signal dynamic range (the mean of the signal for the sample versus that for the control), as well as the variation in the data associated with signal measurements (the standard deviation [SD] for the sample versus that for the control) of a given system in a simple equation: Z = 1 − {[3·(SD of sample with doxycycline + SD of sample without doxycycline)]/(|mean for sample with doxycycline − mean for sample without doxycycline|)}, or Z = 1 − {[3·(3,069 + 731)]/(|38,962 − 4,615|)} = 0.67. A total of five categories were then used to describe the quality of the assay: (i) ideal, (ii) excellent, (iii) double, (iv) yes or no, and (v) impossible. For instance, an assay with a Z factor of 1 (ideal) represents an assay with a very large dynamic range (which approaches infinity) and the narrowest data variability (SD = 0). On the other hand, a Z factor of 0 (a yes-or-no type of assay) represents a very small or no separation band and overlapping SDs (52). To calculate the Z factor for our 293T^{Vif Tet-Off/EYFP-A3G} cell line, cells were seeded into a 96-well plate (n = 32) and cultivated for 96 h in the presence and the absence of doxycycline until the EYFP-A3G fluorescence was quantified. As shown in Fig. 7, 293T^{Vif Tet-Off/EYFP-A3G} cells displayed a Z factor of 0.67. On the basis of the ranking system described above, that Z factor for the assay with our cells corresponded to an excellent assay platform (excellent = 1 > Z ≥ 0.5).

FIG. 6.

Titration of the Vif antagonist RN-18 yields an EC₅₀ of ∼5 μM. 293T^{Vif Tet-Off/EYFP-A3G} cells were seeded onto a 96-well tissue culture plate and cultivated for 48 h in the presence of doxycycline. RN-18 was subsequently added at the indicated concentrations in the absence of doxycycline. After 24 h, the fluorescence was quantified with a Safire plate reader (excitation wavelength, 510 nm; emission wavelength, 535 nm). The EC₅₀ of RN-18 was calculated with the software SigmaPlot by using the four-parameter Hill equation. The data points are mean values for 12 parallel wells. Error bars, SDs; RFI, relative fluorescence intensity.

FIG. 7.

293T^{Vif Tet-Off/EYFP-A3G} cells display a screening assay quality factor (Z factor) of 0.67. 293T^{Vif Tet-Off/EYFP-A3G} cells were seeded onto a 96-well tissue culture plate and cultivated for 96 h in the presence and the absence of doxycycline (Dox), as indicated. Fluorescence was quantified with a Safire plate reader. RFI, relative fluorescence intensity; solid lines, mean values for 32 samples each; dashed lines, SDs.

DISCUSSION

In recent years, cellular host factors, especially the human APOBEC cytidine deaminases, have been thought to be attractive targets in the development of antiviral compound screening platforms (14, 16, 20, 27). In this report, we describe the engineering of a cell line stably expressing EYFP-A3G and a Tet-Off-controllable HIV-1 Vif protein as a high-throughput screening platform for the identification of low-molecular-weight anti-HIV compounds from compound libraries. Our system is based on an assay that uses Vif-triggered transient degradation of a GFP-tagged A3G fusion protein (7). A similar system was recently employed to identify a low-molecular-weight compound inhibiting the Vif-induced degradation of A3G from a compound library (35). However, the studies mentioned above used transient transfection to express both interaction partners in target cells. We, however, generated a stable HEK293T-based cell line constitutively expressing the EYFP-A3G fusion and an inducible Vif protein. Expression of the latter is regulated by the Tet-Off system (15) to precisely control the system.

The rationale for making a stable cell line was the observation that in a transient expression system, the levels of individual proteins, detected by immunoblotting or flow cytometric analysis, vary to a significant extent when other genes are coexpressed at the same time (M. Kirschner, unpublished observation). To overcome these fluctuations, we chose to create a cell line expressing both interaction partners in a stable fashion. Because our cell line is based on a single-cell clone, which was analyzed by FACS for the highest doxycycline-induced shift in the EYFP-A3G fluorescence signal, we worked with a population in which 100% of the cells were known to express both interaction partners. This ensures minimal fluctuations in the levels of the recombinant expressed genes from experiment to experiment and minimizes the generation of false-positive or -negative signals. With a Tet-Off-regulated Vif, we were able to precisely evaluate the dynamic range of the signal and the high-throughput capability of our system.

To assess the proper functioning of our 293T^{Vif Tet-Off/EYFP-A3G} cell line, we performed a panel of control experiments based on the findings of previous studies about the interaction of A3G and Vif and proved that our cell line, as expected, represents a robust, easy-to-handle, and cost-efficient screening platform. Since we worked with a fusion containing a flexible linker between EYFP and A3G, we demonstrated that the protein showed its typical cytosolic localization because of the formation of p bodies (2, 13, 46). We were further able to show that the EYFP-A3G fusion is properly packed into virions (41, 49) and that the A3G moiety possesses deaminase activity against an EGFP reporter gene transduced into a retrovirus (18). Furthermore, we verified the upregulation of ubiquitination in conjunction with the Vif-induced degradation of EYFP-A3G. This degradation could be inhibited by a 26S proteasomal inhibitor (7, 24, 32, 35).

In summary, we demonstrated that the 293T^{Vif Tet-Off/EYFP-A3G} cell line behaves as expected, in the sense that the critical features of the Vif-A3G interaction described above correspond well to the findings from other research groups; the method that uses our cell line as a screening platform is therefore significantly improved over the previously used method of transiently transfecting the interaction partners into target cells. When we tested the activity of the newly identified Vif antagonist RN-18 (35) against our 293T^{Vif Tet-Off/EYFP-A3G} cells, we found an EC₅₀ of 5.8 μM. This result is very close to the initially observed EC₅₀ of 3 μM detected when RN-18 was first discovered (35). However, Nathans et al. did not make titrations of their compounds on the initial screening platform (transiently transfected 293T cells) and used a replicating virus assay to prove the inhibitory effects of RN-18 (35). Our data therefore clearly demonstrate that the 293T^{Vif Tet-Off/EYFP-A3G} cells mimic very closely the effects of a potential HIV inhibitor on the viral replication cycle.

When we calculated the screening window coefficient as a statistical measure for characterization of the high-throughput capability of our system, we showed that the 293T^{Vif Tet-Off/EYFP-A3G} cell line represents an excellent assay platform, as it has a Z factor of 0.67 (52). However, each screening system can produce false-positive results. Since compound libraries usually consist of many chemically different structures, it is possible, for instance, that a particular compound inhibits the proteasome or acts in a nonspecific (not A3G Vif-related) fashion, both of which lead to false-positive EYFP-A3G fluorescence signals. To make sure that a newly identified compound is indeed a Vif antagonist, potential candidates need further testing in a viral infectivity assay with permissive and nonpermissive target cells (35).

Another key feature of a stable screening platform is cost-efficiency. For every transient transfection, one needs to consider the cost of highly purified plasmid DNA as well as the cost of the transfection reagent (e.g., Lipofectamine 2000). Since the cost of transient expression can add up and require a lot of spending on plasmid DNA purification and transfection reagents, the cost-efficiency of the expression system plays a key factor if procedures are to be scaled up in order to screen large libraries of low-molecular-weight compounds. For example, to make a transient transfection in a 96-well plate, the cost for the transfection reagent (e.g., Lipofectamine 2000) and the plasmid DNA is approximately €20 (≈$28.90) per plate. Approximately 26 different compounds can be screened in triplicate wells, including the control wells, of each 96-well plate. On the basis of this calculation, the reagent costs for the screening of one compound in a single round is €0.77 (≈$1.10). On the other hand, doxycycline is the only additional reagent needed for a screening with our 293T^{Vif Tet-Off/EYFP-A3G} cell line and can be obtained at a price of €34 (≈$39)/g. Approximately 100 ng of doxycycline is needed per 96-well plate, which translates to a cost factor of €1.3 × 10⁻⁷ ($1.88 × 10⁻⁷) per substance per screen.

Taken together, the use of the 293T^{Vif Tet-Off/EYFP-A3G} cell line represents a significant improvement over the use of a transient transfection screening platform, in that it is (i) a robust, stable, and highly reproducible system which (ii) is easy to handle (the cell line can be used right away), (iii) can be precisely controlled (because of the inducibility of Vif), (iii) is cost-efficient in use, and (vi) has excellent high-throughput capabilities.