Identification of a Novel HIV-1 Inhibitor Targeting Vif-dependent Degradation of Human APOBEC3G Protein*

Background: The interaction between HIV Vif protein and innate antiviral factor APOBEC3G represents a potential therapeutic target. Results: Screening for inhibitors of Vif-APOBEC3G interaction identified a small molecule, N.41, that protects APOBEC3G from Vif-mediated degradation and exhibits antiviral activity. Conclusion: N.41 is a lead for further development as an antiviral. Significance: These findings suggest new strategies for developing anti-HIV therapeutics. APOBEC3G (A3G) is a cellular cytidine deaminase that restricts HIV-1 replication by inducing G-to-A hypermutation in viral DNA and by deamination-independent mechanisms. HIV-1 Vif binds to A3G, resulting in its degradation via the 26 S proteasome. Therefore, this interaction represents a potential therapeutic target. To identify compounds that inhibit interaction between A3G and HIV-1 Vif in a high throughput format, we developed a homogeneous time-resolved fluorescence resonance energy transfer assay. A 307,520 compound library from the NIH Molecular Libraries Small Molecule Repository was screened. Secondary screens to evaluate dose-response performance and off-target effects, cell-based assays to identify compounds that attenuate Vif-dependent degradation of A3G, and assays testing antiviral activity in peripheral blood mononuclear cells and T cells were employed. One compound, N.41, showed potent antiviral activity in A3G(+) but not in A3G(−) T cells and had an IC50 as low as 8.4 μm and a TC50 of >100 μm when tested against HIV-1Ba-L replication in peripheral blood mononuclear cells. N.41 inhibited the Vif-A3G interaction and increased cellular A3G levels and incorporation of A3G into virions, thereby attenuating virus infectivity in a Vif-dependent manner. N.41 activity was also species- and Vif-dependent. Preliminary structure-activity relationship studies suggest that a hydroxyl moiety located at a phenylamino group is critical for N.41 anti-HIV activity and identified N.41 analogs with better potency (IC50 as low as 4.2 μm). These findings identify a new lead compound that attenuates HIV replication by liberating A3G from Vif regulation and increasing its innate antiviral activity.


APOBEC3G (A3G) is a cellular cytidine deaminase that restricts HIV-1 replication by inducing G-to-A hypermutation in viral DNA and by deamination-independent mechanisms.
HIV-1 Vif binds to A3G, resulting in its degradation via the 26 S proteasome. Therefore, this interaction represents a potential therapeutic target. To identify compounds that inhibit interaction between A3G and HIV-1 Vif in a high throughput format, we developed a homogeneous time-resolved fluorescence resonance energy transfer assay. A 307,520 compound library from the NIH Molecular Libraries Small Molecule Repository was screened. Secondary screens to evaluate dose-response performance and off-target effects, cell-based assays to identify compounds that attenuate Vif-dependent degradation of A3G, and assays testing antiviral activity in peripheral blood mononuclear cells and T cells were employed. One compound, N.41, showed potent antiviral activity in A3G(؉) but not in A3G(؊) T cells and had an IC 50  Retroviruses interact with cellular factors that can support or suppress viral replication; host cell factors that suppress viral replication are termed host restriction factors. The first restriction factors identified against human immunodeficiency virus type 1 (HIV-1) were members of the human cytidine deaminase apolipoprotein B mRNA-editing catalytic polypeptide-like 3 (APOBEC3) 2 family. These proteins inhibit not only HIV-1 but also a broad range of other viruses and endogenous retroelements (1)(2)(3)(4). Among the three members of the APOBEC family that exhibit the most potent anti-HIV-1 activity in vivo, APOBEC3D (A3D), ABOBEC3F (A3F), and APOBEC3G (A3G), A3G is the most well characterized and potent HIV-1 inhibitor (5).
The HIV-1 virion infectivity factor (Vif) is a 23-kDa viral accessory protein that counteracts the innate anti-HIV activity of A3G. In the absence of Vif, A3G is actively packaged into HIV-1 virions and deaminates cytidines in viral minus-strand DNA during reverse transcription, resulting in a G-to-A hyper-mutation and premature degradation of newly synthesized viral DNA (3, 4, 6 -8). A3G also inhibits viral replication via deamination-independent mechanisms that include inhibition of processive reverse transcription and proviral DNA formation (9 -15).
Vif preferentially suppresses APOBEC3 proteins of its host species. HIV-1 Vif binds to and inactivates human A3G (huA3G) but not A3G expressed in African green monkeys (AGM) or rhesus macaques (34 -38). Conversely, AGM simian immunodeficiency virus (SIVagm) Vif inactivates AGM and rhesus macaque A3G but not huA3G. This species specificity was demonstrated by altering a single amino acid in the Vifbinding site, 128 DPD 130 of huA3G. The D128K mutation controls species specificity (34,35,37,38). The Vif-binding site ( 128 DPD 130 ) is adjacent to residues 124 YYXW 127 , which have been implicated in A3G packaging into HIV-1 virions (39). Regions of Vif important for binding and neutralization of APOBEC proteins and species-specific recognition have been mapped to its N terminus (18, 40 -45). Residues 14 DRMR 17 play a role in the species specificity of Vif, whereas distinct regions in the N-terminal half of HIV-1 Vif were shown to be important for its interaction with A3G: residues 21 (40, 41, 46 -49).
Small molecules that inhibit HIV-1 Vif function in vitro have recently been identified, but these compounds do not inhibit the Vif-A3G interaction (50 -53). Another study identified two compounds, IMB-26 and IMB-35, as specific inhibitors of Vifdependent degradation of huA3G via stabilization of A3G (54). Although this study demonstrated a Vif-dependent effect on inhibition, a mechanistic explanation for the specific inhibition was unknown, and compound activity was not characterized in physiologically relevant target cells. Here, we used a high throughput screen for inhibitors of Vif-A3G binding to identify a novel lead compound that specifically protects A3G from Vifmediated degradation, thereby increasing A3G antiviral activity against HIV-1 replication.
Cell Transfection, Western Blot Analysis, and Co-immunoprecipitation-HEK293T cells were cultured in DMEM with 10% FBS and transfected by Lipofectamine 2000 (Life Technologies) according to the manufacturer's instructions. At 40 -48 h post transfection, lysates were prepared in lysis buffer (50 mM Tris-HCl, pH 7.0, 150 mM NaCl, 0.5% Nonidet P-40, and 1% protease inhibitor mixture). Twenty-five g of protein normalized by Bradford protein assay (Bio-Rad) were separated by SDS-PAGE, transferred onto polyvinylidene difluoride membranes (Millipore), and detected by standard Western blotting. For coimmunoprecipitation experiments, identical amounts of lysate were subjected to immunoprecipitation followed by Western blotting. HA-tagged proteins were immunoprecipitated by EZview Red anti-HA affinity gel (Sigma). For GST pulldown, 2.5 g of recombinant protein was incubated with 10 l of glutathione-Sepharose 4B beads and 250 l of 293/A3G cells lysate for 1 h at 4°C, the beads were washed, and isolated proteins were subjected to SDS-PAGE and Western blotting.
Time-resolved Fluorescence Resonance Energy Transfer (TR-FRET) Assay-The interaction between GST-Vif residues 1-94, which contains the A3G-binding site (mapped to residues 40 -72), and a biotinylated peptide consisting of A3G residues 110 -148 (bio-A3G) was detected using TR-FRET (40,46). europium (EU-W1024-labeled anti-GST antibody, PerkinElmer Life Sciences) and Ulight (LANCE Ultra Ulightstreptavidin, PerkinElmer Life Sciences) served as the donoraccepter pair. Briefly, 15 nM GST-Vif 1-94 protein was added to assay plates containing test compounds (1536 well format). After 30 min of incubation, 500 nM bio-A3G peptide was added. 2 nM europium-labeled anti-GST and 50 nM streptavidin-Ulight detection reagents were added 30 min later and incubated for 1 h. Samples were analyzed using Envision multiplate reader (PerkinElmer Life Sciences) with excitation at 340 nm and emission at 615 and 665 nm. The emission at 615 nm from europium-labeled anti-GST induces emission at 665 nm from Ulight conjugated to Streptavidin when the two molecules are in close proximity, resulting in a FRET signal (see Fig. 1A). A 307,520 compound library from the NIH Molecular Libraries Small Molecule Repository (MLSMR) was screened. The MLSMR collection of Ͼ300,000 compounds generically grouped into one of the following five categories: (a) specialty sets, comprising bioactive compounds such as known drugs and toxins, (b) non-commercial compounds, mainly from academic laboratories, (c) targeted libraries, (d) natural products, and (e) diversity compounds. A description of the library can be found at nih.gov. Negative control wells included all assay components in the absence of an inhibitor, and positive control wells included all assay components with GST-Vif 1-94 protein replaced by GST protein. Percent inhibition was determined as: 100 ϫ (test compound signal Ϫ median negative control signal)/(median positive control signal Ϫ median negative control signal)). Statistical analysis for calculating the % inhibition cutoff for selecting active hits was performed; FRET signal mean and standard deviation (S.D.) were calculated for all tested compounds, and the cutoff was set as 3 S.D. above the mean. All of the screening data has been published on PubChem under AID 1117320.
Dose-response and Counter Screen Assays-Compounds identified as hits in the screen were evaluated for dose-response in the TR-FRET assay, TR-FRET counter screen assay, and HIV-1 Tat-TAR fluorescence polarization (FP) assay. Compounds identified as hits in the screen were also evaluated in a Vero cell or THP-1 cell toxicity assay. The TR-FRET counter screen assay used 7.5 nM biotinylated GST protein in place of the GST-Vif 1-94 protein and bio-A3G peptide and was used to identify nonspecific compounds. The HIV-1 Tat-TAR biochemical assay utilizes fluorescence-labeled TAR RNA ( FAM TAR; 5Ј-GGC CAG AUC UGA GCC UGG GAG CUC UCU GGC C-3Ј; 6-carboxyfluorescein attached at the 5Ј end) and full-length Tat protein (Diatheva). Binding of Tat to FAM TAR results in a higher level of anisotropy compared with the unbound FAM TAR. Therefore, inhibition of Tat binding to FAM TAR leads to a decreased level of anisotropy. This assay was selected as an additional fluorescence-based counter screen to determine the specificity of hits for the Vif-A3G interaction. A Vero cell or THP-1 cell toxicity assay quantified compound toxicity. Cells were plated in 20 l of growth medium at 2,500 (Vero cells, hits from the first 100,000 compounds screened) or 5,000 (THP-1 cells; hits for the remainder of compounds screened) cells per well in 384-well black plates with clear bottoms containing compound with a high test concentration of 40 M to yield a final volume of 25 l/well. After a 72-h incubation at 37°C and 5% CO 2 , toxicity was measured using CellTiter-Glo Luminescent cell viability assay (Promega).
YFP-APOBEC3G Degradation Assay-2.5 ϫ 10 4 HEK293T cells seeded in 96-well plates were transfected by Lipofectamine 2000 (Invitrogen) with pEYFP (counter screen), pEYFP-A3G, or pEYFP-A3G and HIV Vif plasmids. 24 h post-transfection, media were replaced with fresh media supplemented with DMSO (0.4%) or 40 M concentrations of tested compound for 20 h. At 44 h post transfection, cells were lysed by M-PER Mammalian protein extraction reagent (Thermo Scientific) supplemented with protease inhibitor mixture (Roche Applied Science). Cleared supernatants of cell lysates were transferred to BD Optilux 96-well plates (BD Biosciences), and YFP fluorescence intensity (FI) was measured by SpectraMax M5e (Molecular Devices) Multi-Mode Microplate Reader (excitation at 510 nm and emission at 530 nm). Small molecules increasing or decreasing YFP expression from pEYFP were excluded due to their nonspecific effects. The mean FI was calculated for each tested compound (2 replicates) and control (DMSO). To exclude molecules that increase YFP expression nonspecifically, a cutoff of 3 S.D. above the mean FI(YFP) measured for the DMSO control was established. A cutoff of 2 S.D. below the mean FI(YFP) measured for the DMSO control was established to exclude cytotoxic molecules. The remaining molecules were tested in cells expressing YFP-A3G alone or YFP-A3G and Vif. FI means were calculated for each tested compound (2 replicates) and control (DMSO) for treated cells expressing YFP-A3G ϩ Vif or YFP-A3G alone. The following formula was used to calculate the percentage increase in YFP-A3G levels in compound-treated cells relative to non-treated cells (DMSO): 100 ϫ (FI(YFP-A3G ϩ HVif) compound Ϫ FI(YFP-A3G ϩ HVif) DMSO )/(FI(YFP-A3G) DMSO Ϫ FI(YFP-A3Gϩ HVif) DMSO ). 100% represents the FI of DMSO-treated cells expressing YFP-A3G, whereas 0% represents the FI from DMSO-treated cells expressing YFP-A3GϩHVif. A cutoff of 2 S.D. above the % in YFP-A3G increase calculated for the DMSO control was used to prioritize compounds, and % increase in FI relative to DMSO-treated control cells was calculated for each compound (see Fig. 2A). Molecules stabilizing YFP-A3G levels independently of Vif were expected to increase YFP-A3G levels (measured as FI), whereas compounds inhibiting YFP-A3G degradation through targeting Vif were expected not to increase YFP-A3G levels in the absence of Vif. To classify compounds as predicted to target either Vif or A3G proteins, we calculated -fold increase of YFP-A3G levels for each tested compound using the formula: compound-mediated -fold increase of YFP-A3G levels ϭ FI(YFP-A3G)compound/FI-(YFP-A3G) DMSO. We set a cutoff of 1 S.D. above the -fold increase of YFP-A3G levels calculated for the DMSO control (4 replicates). The -fold increase of YFP-A3G FI calculated for each compound is represented in the y axis of a scatter plot.
HIV-1 Replication in Peripheral Blood Mononuclear Cells-Compounds were evaluated in dose-response assays using a 100 M high test concentration (9 total concentrations using halflog dilutions) as described (56). Test drug dilutions were prepared at a 2ϫ concentration in microtiter tubes, and 100 l of each concentration was placed in designated wells. Activated  (63); catalogue #3185 listed as HIV-1 BZ167, from Dr. John Mascola) was then added to each test well. Primary virus stocks were prepared by low passage replication in freshly obtained PBMCs. Parallel plates lacking virus were prepared and monitored for cell viability using an MTS assay (Promega). PBMC cultures were maintained for 7 days at 37°C, 5% CO 2 . Cell-free supernatant samples were then collected and analyzed for reverse transcriptase (RT) activity (64), whereas cytotoxicity was measured by MTS assay.
HIV-1 Replication in T Cells-NL4Ϫ3 (CXCR4-tropic) and NL4Ϫ3 ⌬Vif viruses were produced in HEK293T cells, and viral titers were measured by RT assay as described (65). Virus stock (1 ϫ 10 5 3 H cpm) was used to infect 1 ϫ 10 6 H9 and CEM T cells, and 2 ϫ 10 4 cpm viruses were used to infect SupT1 and CEM-SS T cells. 3 h post incubation at 37°C, 5% CO 2 , infected cells were washed 3ϫ to remove free viruses, and 100 l of infected cells were added to wells in 96-well plates. Compounds were evaluated in dose-response assays using a 40 M high test concentration (3-4 total concentrations using 1:2 dilutions). Test drug dilutions were prepared at a 2ϫ concentration in microtiter tubes, and 100 l of each concentration was placed in appropriate wells. Ritonavir (protease inhibitor) was included as a positive control antiviral compound. Separate plates without virus were prepared in parallel for drug cytotoxicity studies. Plates were incubated for 6 days, and virus production was measured using p24 ELISA (PerkinElmer Life Sciences) according to the manufacturer's instructions; compound cytotoxicity was measured by MTS assay. For testing antiviral activity against HIV-1 spreading infection, infected CEM and CEM-SS T cells were treated by increasing concentrations of 0, 2.5, 5, and 10 M N.41. Virus produced from infected cells was collected every 3 days, and infected cells were washed and incubated with fresh N.41 supplemented media. The titer of viruses collected 3, 6, and 9 days post infection was evaluated by RT assay.
Single-round Infectivity Assay, Virus Purification, and A3G Virion Packaging Levels-Viruses were produced by co-transfecting HEK293T cells with VSV-G (vesicular stomatitis virus-G envelope protein) envelope plasmid, AGM or human APOBEC3G, human APOBEC3F, or human APOBEC3C and Vif-deficient proviral plasmid pNLX⌬Env⌬Vif or pNLX⌬Env. Viruses were quantitated by RT assays, and normalized amounts were used to infect the reporter cell line TZM-bl. Infectivity was measured 48 h after infection by performing luciferase (Promega) or ␤-galactosidase assays (Applied Biosystems). To purify viruses, 150 ϫ 10 3 cpm virus-containing supernatants (8.5 ml) were concentrated by ultracentrifugation through 1.7 ml of 20% sucrose in phosphate-buffered saline (PBS). Purified viruses were resuspended in Laemmli sample buffer (Bio-Rad) supplemented with 5% 2-mercaptoethanol, and p24 Gag and A3G protein levels in the purified viruses were detected by Western blot. Statistical significance was evaluated by using Student's t test (p Ͻ 0.05).
Expression and Purification of GST and GST-Vif Fusion Proteins in Escherichia coli-Briefly, E. coli Rosetta (DE3)plysS competent cells (Novagen) were transformed with pGEX-6P-1 vectors expressing 1-94 GST-Vif and full-length GST-Vif constructs, and expression of recombinant proteins was induced by isopropyl 1-thio-␤-D-galactopyranoside. Next, the bacterial culture was lysed and sonicated, and soluble GST recombinant proteins were purified through glutathione-Sepharose 4B fast Flow beads (GE Healthcare).

RESULTS
To identify compounds that inhibit the interaction between HIV-1 Vif and A3G in a high throughput format, we developed a homogeneous TR-FRET assay using LANCE (Lanthanide Chelate Excite) reagents ( Fig. 1A) (40,46). In this assay interaction between purified GST-Vif residues 1-94 (1-94 GST-Vif), which includes A3G binding sites, and a synthetic biotinylated peptide containing A3G residues 110 -148 (bio-A3G), which includes the Vif-binding site, is detected by europium (europium-donor fluorophore)-labeled anti-GST antibodies and Streptavidin-Ulight (acceptor fluorophore). Interaction between GST-Vif and bio-A3G brings europium and Ulight into close proximity (ϳ9 nm or less), supporting energy transfer between these molecules measured as a FRET signal. The attenuation of GST-Vif-bio-A3G interaction is expected to result in FRET signal reduction.
The screening pipeline to identify inhibitors of the HIV-1 Vif-A3G interaction is outlined in Fig. 1B. A 307,520 compound library from the NIH MLSMR was screened at a concentration of 6.25 M. The compounds were screened in three different sets of ϳ100,000 compounds each. The median z-values calculated for each of the sets were 0.84, 0.70, and 0.77. Statistical analysis identified 63.0%, 65.1%, and 33.7% inhibition as cutoffs between inactive and hit compounds (see % inhibition formula under "Experimental Procedures"). Based on these statistical criteria, 3686 hits representing 3650 unique compounds were identified (36 hits overlapped between the three screens), and the overall hit rate for the screen was 1.2%. For initial follow-up testing of the compounds, hits were evaluated for dose response in the TR-FRET biochemical assay as well as in a PerkinElmer Life Sciences TR-FRET Counter Screen assay to identify compounds that quench the fluorescence signal or inhibit the GSTanti-GST antibody interaction instead of Vif-A3G binding (i.e. to eliminate nonspecific inhibitors). The compounds were also tested in an HIV-1 Tat-TAR FP assay, which was selected as an additional fluorescence-based counter screen to determine specificity for Vif-A3G interaction. Lastly, a cytotoxicity assay using Vero (hits from first 100,000 compound set) or THP-1 cells (hits from second and third 100,000 compound sets) was used to identify cytotoxic compounds. Of 3686 hits identified in the screen, 195 hits achieved an IC 50 value of Ͻ25 M in the TR-FRET assay and were not active in any of the counter screen secondary dose-response assays tested or were potent in the TR-FRET assay with an IC 50 Ͻ 0.195 M (the low test concentration used in the experiments) with minimal activity observed in various counter screens. An additional 116 hits were flagged as potential Vif inhibitors because they exhibited at least a 10-fold difference in IC 50 values when secondary counter screen dose-response assays were compared with the TR-FRET biochemical assay. These combined 311 hits representing 303 unique compounds were identified for additional follow-up testing. Compound N.41 inhibited the Vif-A3G interaction by 89% when tested at 6.25 M in the TR-FRET high throughput screening assay. For comparison, P15, a 15-mer Vif peptide (residues 57-71) used as a positive control (40,46), exhibited an average inhibition of 40% at 50 M and 93.5% at 100 M. The secondary TR-FRET-based assay and counter screen assays (Fig. 1C) validated N.41 as a promising compound for subsequent follow-up evaluations. Further details for the 3650 unique hits identified in the primary screen and results from counter screen assays are shown in Supplemental Table 1.
We developed a cell-based assay to identify compounds that attenuate Vif-dependent degradation of YFP-A3G to screen the 311 hits selected for additional follow-up testing based on the primary screen and secondary dose-response experiments.
Compounds that significantly increased or decreased YFP levels in control experiments were excluded from additional follow-up testing. Cells expressing both YFP-A3G and Vif proteins were treated with DMSO (negative control) and the compounds of interest for 20 h. Cell lysates were assessed for YFP-A3G FI. The relative increase in FI in compound-treated cells that expressed YFP-A3G and Vif relative to untreated cells (DMSO control) was plotted. Fig. 2A shows a representative experiment. Fifty-eight of the tested compounds were identified as active molecules based on the percent increase of the FI being above a calculated cutoff (red dashed line in Fig. 2A).
Compounds targeting Vif were not expected to increase YFP-A3G levels in the absence of Vif. To identify such compounds, the -fold increase of FI in compound-treated cells expressing YFP-A3G compared with DMSO-treated cells was determined, and a cutoff (blue dashed line in Fig. 2A) was set to classify hit compounds into two groups corresponding to those potentially targeting Vif (below the cutoff) or targeting A3G (above cutoff). Nineteen compounds that increased YFP-A3G levels in the presence of Vif also increased A3G levels in the absence of Vif (a subset of these compounds is shown in Fig. 2A,  right upper quadrant). One example is compound C.18, a molecule identified in the cell-based screen as a false positive that significantly increases YFP levels. Based on this phenotype, C.18 was included in subsequent experiments as a positive control for molecules targeting YFP-A3G. Compound N.41 fell just above the cutoff set to identify compounds potentially targeting A3G, suggesting it has an effect on A3G protein levels. To further investigate whether compound N.41 targeted Vif versus A3G, we examined the effect of N.41 treatment on endogenous A3G protein levels in CEM cells (A3Gϩ). In these cells N.41 increased endogenous A3G protein in a dose-dependent manner (Fig. 2B).
We next determined the antiviral activity of hit compounds against HIV-1 Ba-L replication in PBMCs. Only 255 of the 311 hits (250 of 303 unique compounds) identified were commercially available and obtainable in quantities necessary for followup studies. This assay identified 18 compounds in which the TC 50 values were at least 5-fold greater than the corresponding IC 50 values (therapeutic index values ϭ TC 50 /IC 50 ranging from 5.12 to Ͼ73). Three compounds (N.32, N.41, and N.114) demonstrated both antiviral activity against HIV-1 replication in PBMCs and attenuated Vif-dependent degradation of A3G in the cell-based assay described in Fig. 2. Further details for the priority hits tested in the YFP-A3G degradation assay and PBMC antiviral assay are shown in supplemental Table 1. To confirm A3G-dependent antiviral activity, these three compounds were tested against HIV-1 NL4-3 replication in H9 (A3Gϩ) T cells and in SupT1 (A3GϪ) T cells. N.41 had an IC 50 as low as 8.4 M and a TC 50 of Ͼ100 M when tested against HIV-1 Ba-L replication in PBMCs (Fig. 3A) and displayed significant antiviral activity in H9 T cells (A3Gϩ) but not in SupT1 T cells (A3GϪ) (Fig. 3B); the other two compounds did not dem-onstrate A3G-dependent antiviral activity in these experiments (data not shown). Additional testing of N.41 antiviral activity against 3 other HIV-1 isolates, 92RW016, 92UG021, and JV1083, resulted in IC 50 values of 16.8, 20, and 17.5 M, respectively, demonstrating it has broad antiviral activity (Fig. 3A). Subsequent repeat testing of N.41 against HIV-1 Ba-L in PBMCs on two separate occasions resulted in IC 50 values of 20.8 M (see Fig. 7C) and 22.6 M (data not shown). The lower potency observed in these repeat assays may be the result of using PBMCs from different donors and/or compound stability in subsequent experiments. To further examine the A3G dependence of compound N.41 antiviral activity, we also tested it against HIV-1NL4-3 in a spreading infection in CEM (A3Gϩ) versus CEM-SS (A3GϪ) T cells (Fig. 3C). N.41 demonstrated antiviral activity in a dose-dependent manner when tested against virus replication in CEM but not in CEM-SS T cells, suggesting its mechanism of action is A3G-dependent.
In addition to A3G, T cells express other APOBEC3 proteins shown to have anti-HIV activity, including A3F and A3C (5,17,41). In contrast to A3G and A3F, APOBEC3C (A3C) exerts only modest antiviral activity against HIV-1 but has potent antiviral activity against simian immunodeficiency virus (66). Similar to A3G, A3F and A3C are targeted for proteasomal degradation by HIV-1 Vif. To further characterize how N.41 attenuates HIV-1 replication in cells, we tested whether it could regulate A3G, A3F, or A3C antiviral activity in a single round infectivity assay. Vif(ϩ) and Vif(Ϫ) viruses were produced in A3G-, A3F-, or A3C-expressing cells in the presence or absence of N.41. Western analysis revealed that A3G protein levels were increased in N.41-treated cells regardless of HIV-1 Vif expression (Fig. 4A, left upper panel). A3F levels were increased by N.41 treatment only in cells co-expressing HIV-1 Vif, and A3C levels were unaffected by N.41 (Fig. 4A, right upper panels). No change in Vif protein levels was detected in N.41-treated cells. A3G levels were increased in virus particles produced from N.41-treated cells regardless of Vif expression (Fig. 4A, left lower panel), whereas A3F and A3C levels in virus particles were increased by N.41 treatment only in Vif-expressing cells (Fig. 4A, right lower  panels). Next, we infected TZM-bl reporter cells with equivalent amounts of virus and found that N.41-treated cells expressing A3G produced virus with lower infectivity compared with untreated cells (Fig. 4B, left panel). In contrast, N.41 treatment had minor effects on virus infectivity that were not statistically significant in virus-producing cells expressing A3C and enhanced virus infectivity in cells expressing A3F (Fig. 4B, right  panel).
Although cellular and virion-packaged A3G protein levels were increased in N.41-treated cells in the absence of Vif (Fig.   4A), the effect of N.41 on A3G protein levels was greater when Vif was present. HuA3G but not African green monkey A3G (agmA3G) is subject to HIV-1 Vif regulation. To examine if N.41 antiviral activity targeted the Vif-huA3G interaction, we tested whether N.41 could attenuate the production of infectious viruses from cells expressing agmA3G. VSV-G pseudotyped Vif(ϩ) HIV-1 viruses and corresponding viruses that lack Vif (⌬Vif) were produced in 293T cells ectopically expressing HA-tagged huA3G or agmA3G proteins (Fig. 5A). Notably, in this experiment lower levels of A3G protein were expressed compared with the levels used in the infectivity assay (75 ng of transfected plasmid DNA instead of 200 and 400 ng used in Fig.  4) in an effort to achieve A3G expression at levels closer to physiological levels. Producer cells were treated for 40 h with 25 M N.41 or DMSO (untreated). N.41 treatment significantly increased huA3G protein levels in the presence of HIV-1 Vif, whereas only a modest change was seen in its absence (Fig. 5A). N.41 had no effect on agmA3G protein levels, suggesting spe- AgmA3G protein levels showed only a modest increase in these virions. Both human and agmA3G protein levels increased in Vif-deficient HIV-1 viruses produced in N.41-treated cells. Finally, testing Vif(ϩ) HIV-1 viruses in infectivity assays showed that only viruses produced from cells that express human but not agmA3G protein were less infectious when produced from N.41-treated cells compared with viruses produced from untreated cells (Fig. 5B). Next, we determined whether N.41 could directly attenuate the Vif-A3G interaction. In vitro pulldown assays with purified GST-Vif proteins and cell lysates from 293-apo cells (HEK293 cell line that stably express A3G-HA) were performed. The pulldown assays were performed with DMSO or 40 M N.41. In these experiments, a decreased A3G-Vif interaction was detected in the presence of N.41 (Fig.  5C). We also immunoprecipitated A3G-HA proteins from N.41-treated and untreated virus producer cells that express both Vif and A3G-HA proteins (Fig. 5D). This experiment showed that N.41 treatment reduced HIV-1 Vif co-immunoprecipitating with A3G, suggesting that N.41 binds to A3G and interferes with its recruitment by Vif.
The results of the preceding experiments support the idea that N.41 targets A3G, resulting in resistance to Vif-dependent degradation. To further delineate the structural features of N.41 important for this antiviral activity, 16 structurally related N.41 analog molecules (details shown in supplemental Table 2) were tested for their ability to affect endogenous A3G protein levels in CEM cells (Fig. 6, A and B). We also obtained and tested de novo synthesized N.41 compound (purity Ͼ 95% based on mass spectrometry) to validate its antiviral activity (Figs. 6B and 7). Evaluation of ␤-tubulin protein levels was used as an indicator of cytotoxicity. Analogs 3, 4, and 12, when tested at 10 M, enhanced A3G protein levels, although analog 12 exhibited cytotoxicity at 20 M. The effects of analogs 3, 4 (tested at 40 M), and 12 (tested at 20 M) on the production of infectious virions from A3G-expressing cells were also examined. Western blotting to detect A3G protein levels in the producer cells showed that treatment with N.41 and the three analogs of interest dramatically increased A3G protein levels (Fig.  7A, upper panel). However, analysis of the resulting virus particles revealed that only the N.41 parental compound and analogs 3 and 12, but not analog 4, increased virion-packaged A3G (Fig. 7A, lower panel). Importantly, testing Vif(ϩ) HIV-1 viruses in the infectivity assay (Fig. 7B) showed that viruses produced from cells treated by analog 3 were markedly less infectious (ϳ50 -60% of untreated control levels) than viruses produced from cells treated by the parental compound (ϳ80% of untreated control levels). Analog 12 showed similar antiviral activity, whereas analog 4 exhibited only a modest effect on virus infectivity. Finally, testing the antiviral activity of N.41 analogs against virus replication of the HIV-1 isolates HIV-1 Ba-L and NL4Ϫ3 (Fig. 7C)

DISCUSSION
In this study the HIV-1 Vif-A3G interaction served as the primary target in a high throughput screen to detect inhibitors of this protein-protein interaction (Fig. 1B). We successfully identified a novel lead compound N.41 that disrupts this critical interaction. N.41 had a calculated IC 50 of 2.18 M for inhibiting the interaction between GST-Vif (amino acids 1-94) and bio-A3G (amino acids 110 -148) peptide in a TR-FRET-based assay (Fig. 1C). N.41 also attenuated Vif-dependent degradation of A3G in a cell-based assay ( Fig. 2A) and inhibited viral replication in PBMCs (IC 50 as low as 8.4 M and TC 50 Ͼ100 M). N.41 also inhibited virus replication in H9 (A3Gϩ) T cells and attenuated HIV-1 replication in a spreading infection assay in CEM (A3Gϩ) T cells. The difference in N.41 antiviral activity in H9 compared with CEM cells (Fig. 3, B and C) is likely attributable to differences in experimental design; the H9 experiment is a dose response at one time point, whereas the CEM experiment is a virus replication study over time. When the CEM experiment is plotted as a dose response for the same time point (day 6), N.41 shows similar antiviral activity in H9 and CEM cells (data not shown). N.41 exhibited no significant antiviral activity in the absence of A3G expression (SupT1 and CEM-SS T cells), suggesting an A3G-dependent mechanism. Further characterization of the mechanism of action of N.41 supported this con-clusion. The cell-based assay showed that N.41 treatment reduced the infectivity of viruses produced from A3G-expressing cells (Fig. 4B), and Western blotting revealed that N.41 increased A3G protein levels in both producer cells and newly produced virions (Fig. 4A). Furthermore, N.41 treatment of CEM T cell lines increased endogenous A3G protein levels (Fig.  2B). Although N.41 treatment increased both cellular and virion A3G protein levels in the absence of Vif, effects on A3G protein levels were even more significant when Vif was present (Fig. 4A, left panel). Together, these results support the idea that compound N.41 specifically targets the A3G protein.
Vif preferentially suppresses APOBEC3 proteins of its host species (34 -38). A single amino acid in A3G, aspartic acid at position 128 in huA3G versus lysine in agmA3G, controls this specificity by direct effects on Vif-A3G binding. As expected from the design of our screen, N.41 attenuated huA3G degradation by Vif and inhibited the production of infectious virions from cells expressing huA3G, but not agmA3G (Fig. 5, A and B). Given these findings, we predict that N.41 binds huA3G in a region close to this essential aspartic acid, thereby hindering the Vif-A3G interaction. A pulldown assay (Fig. 5C) and coimmunoprecipitation analysis (Fig. 5D) further supported this prediction. Interestingly, the packaging of both huA3G and agmA3G proteins was increased in Vif-deficient virions pro- duced from N.41-treated cells (Fig. 5A, lower panel). The huA3G and agmA3G protein packaging motif ( 124 YYXW 127 ) is adjacent to the aspartic acid and lysine at position 128 (huA3G and agmA3G) that controls interaction with Vif. We speculate that N.41 binds both huA3G and agmA3G in this region and thereby increases packaging of these A3G proteins into virions.
In this study we found that N.41 treatment attenuated A3F but not A3C degradation in virus producing cells in a Vif-dependent manner (Fig. 4A, right panels). A recent study (67) revealed new determinants of Vif binding in the A3F-CTD: Glu-316, Ser-320, and Glu-324. These residues were shown to be part of a negatively charged interface that is directly involved in Vif binding. These residues were not important for A3C-Vif binding (68,69), so the Vif binding sites in A3F and A3C may be different. N.41 may target the Vif-binding site in A3F, similar to its target site in A3G. A3G residues 110 -140, which are part of the bio-A3G peptide used in the primary screen, share high sequence identity (75%) with A3F-CTD residues (amino acids 293-323); hence this region of A3F could potentially be targeted by N.41 as well. Unexpectedly, N.41 treatment enhanced the infectivity of virus produced in A3F-expressing cells. An important difference between A3G and A3F is the relative posi-tions of the Vif-binding site and the single-strand DNA (ssDNA)-binding site; in A3G, the Vif-binding site maps to residues 126 -132, and the ssDNA-binding site maps to the C-terminal domain. In contrast, in A3F, the Vif-binding site and ssDNA-binding site are in much closer proximity (68,70). Thus, potential effects of N.41 on the Vif-binding site in A3F might also impede its interaction with ssDNA, which is required for A3F antiviral activity. These findings provide further support for our hypothesis that N.41 targets the Vif-binding site in A3G and suggest that future studies evaluating N.41 analogs as potential antivirals should include assessment of their effects on other APOBEC3 family members.
To further our understanding of the N.41 mechanism of action and potentially improve its antiviral activity, we performed structure-activity relationship studies. N.41, 3-(4-Hydroxyphenylamino)-1-pyridin-3-yl-but-2-en-1-one, is a 254. 28-Da compound that joins two bulky groups at its ends: a pyridine ring and a hydroxyphenylamino group. We obtained 16 structurally related molecules (N.41 analogs) and assayed the analog compounds for antiviral activity. Analogs 3, 4, and 12 increased endogenous A3G levels in CEM T cells more effectively than the parental compound at comparable concentrations. These results suggest that the hydroxyl moiety (analogs 3 and 12) or the amide hydrogen of the N-methyl-amide moiety (analog 4) located at the para position of the phenylamino group are essential for activity, potentially through donation of a hydrogen bond to its target. Furthermore, conformational stabilization of the pyridine ring (as in analog 12) appeared important for optimal potency, although the nitrogen itself was not required (as in analog 3). These three active analogs also increased A3G levels in virus producer cells, but only analogs 3 and 12 increased A3G levels in newly produced virions (Fig. 7A). Unlike treatment with analog 4, treatment with analogs 3 and 12 resulted in a less infectious virus produced in cells expressing A3G (Fig. 7B). Analog 3 and 12 treatment also inhibited HIV-1 replication in PBMCs more potently than the parental compound N.41, with analog 12 demonstrating an IC 50 as low as 4.2 M (Fig. 7C). The structure-activity relationship results imply that N.41 interacts with A3G and obstructs its recruitment and degradation by Vif, but due to the proximity of the Vif-binding site and A3G packaging motif, some molecules, such as analog 4, may also negatively affect A3G packaging into virions and weaken its antiviral capacity. N.41 and its analogs such as 3 and 4 are enaminoketones (enaminones); unfortunately, such enaminones can have problems with stability. Therefore, goals for future work are to design lead compounds with better potency along with modified structures designed to increase their stability.
In summary, a primary screen for inhibitors of HIV-1 Vif-A3G binding together with cell-based screens assays for antiviral activity in relevant primary cells, and preliminary structureactivity relationship studies identified a parental N.41 molecule as well as two analogs that can suppress HIV replication via A3G-mediated restriction. Further structure optimization studies may lead to the identification of additional potent HIV-1 inhibitors that could specifically shield A3G from degradation by Vif and unleash A3G innate antiviral activity.