APOBEC3G Inhibits DNA Strand Transfer during HIV-1 Reverse Transcription*

Human APOBEC3G (hA3G) has been identified as an anti-HIV-1 host factor. The presence of hA3G in HIV-1 strongly inhibits the ability of the virus to produce new viral DNA upon infection. In this report, we demonstrate that the reduction of late viral DNA synthesis is due to the inhibition by hA3G of the strand transfer steps that occur during reverse transcription. Analysis of viral cDNA intermediates in vivo reveals that hA3G causes an inhibition of the minus and plus strand transfers, without having a significant impact on DNA elongation. Using an in vitro system to measure minus strand transfer similarly shows a dose-dependent reduction of strand transfer by hA3G. This inhibition of strand transfer occurs independently the editing activity of hA3G and is correlated with its ability to prevent RNaseH degradation of the template RNA.

Human APOBEC3G (hA3G) 3 has been identified as an anti-HIV-1 host factor (1). hA3G belongs to an APOBEC superfamily containing at least 10 members, which share a cytidine deaminase motif (a conserved His-X-Glu and Cys-X-X-Cys zinc coordination motif) (2). The APOBEC family in humans includes APOBEC1 (hA1), APOBEC2 (hA2), APOBEC3A-H (hA3A-H), and activation-induced cytidine deaminase. The APOBEC proteins are capable of inhibiting the replication of a wide variety of retroviruses and non-retroviruses, suggesting that these proteins represent a novel component of innate immunity to viral infection (for review, see Refs. 3 and 4). The virus counters hA3G's anti-viral activity with the viral protein Vif (virion infectivity factor), which binds to hA3G, and targets hA3G for proteasomal degradation (5,6). Vif is thus required for HIV-1 replication in cell types that constitutively express hA3G (termed "non-permissive" cells), such as primary T lymphocytes, macrophages, and T-cell lines such as H9. Vif is not required for viral replication in cells not expressing hA3G ("permissive" cells) such as SupT1, Jurkat, 293, HeLa, and CEM-SS lines (7)(8)(9).
HIV-1-containing hA3G shows a reduced ability to produce new viral DNA upon infecting cells (10 -13). It has been suggested that this results from the degradation of newly synthesized viral DNA, rather than the inhibition of new DNA synthesis. Thus, because the small amount of minus strand cDNA that is made in newly infected cells (ϳ5% of wild type) contains 1-2% of the cytosines deaminated by hA3G to form uracil (12, 14 -16), it has been suggested that most newly synthesized viral DNA edited in this manner will be degraded by the DNA repair system. For example, DNA glycosylases such as UNG2, a uracil DNA glycosylase packaged into HIV-1 (17,18), can recognize an altered base and remove the base by apurinic/apyrimidinic endonuclease 1, resulting in either a 5Ј-deoxyribose phosphate group that is a substrate for DNA repair enzymes or in the degradation of the DNA (19). However, a reduction in UNG2 incorporation into HIV-1 containing hA3G, using either an UNG2 inhibitor or virus-producing cells lacking endogenous UNG2 activity, had no effect on the antiviral capacity of hA3G, i.e. in the presence of hA3G, viral infectivity and resulting synthesis of viral DNA are reduced equally with or without viral UNG2 (20). Furthermore, several reports have shown that mutant hA3G (21)(22)(23) and mutant hA3F (24) that have lost their cytidine deaminase activity still show strong anti-HIV-1 activity and a reduction in viral DNA synthesis, and hA3F and/or hA3G can also inhibit hepatitis B virus replication with little or no editing (25,26).
One should then consider the possibility that hA3G inhibits reverse transcription. We have recently reported that Vif-negative virions containing hA3G show a 55% reduction in the production of early viral DNA in newly infected cells and that this is correlated with a similar reduction in the initiation of reverse transcription (22). In this report, we provide further evidence that the greater (ϳ95%) reduction in late DNA synthesis can be accounted for by the hA3G-induced inhibition of minus and plus strand transfer steps in reverse transcription.

EXPERIMENTAL PROCEDURES
Plasmid Construction, Cell Transfections, and Virus Purification-BH10 is a simian virus 40-based vector that contains full-length wild-type HIV-1 proviral DNA. The construction of BH10.Vif-, as well as wild-type and mutant forms of hA3G has been previously described (27). The culture and transfection of HEK-293T cells with these plasmids using Lipofectamine 2000 (Invitrogen), and the isolation of virions 48 h post-transfection from the cell supernatant, were done as previously described (22,27). Unless stated otherwise, 293T cells were transfected with 2 g of HIV-1 proviral DNA and 1 g of plasmid coding for wild-type or mutant forms of hA3G. The total amount of plasmid DNA used for transfection was kept constant in controls by replacing plasmid coding for hA3G with the empty vector, pcDNA3.1. Viral p24 was measured with a commercial kit available for p24 antigen capture (Abbott Laboratories). Culture of SupT1 and its infection with the HIV-1 produced from 293T cells, were as previously described (22).
Quantitation of Newly Synthesized HIV-1 DNA-Real-time fluorescence-monitored PCR, using the LightCycler Instrument (Roche Diagnostics, GmbH), was used to monitor, over a 24-h period post-infection, the synthesis of viral cDNA intermediates in a permissive cell line, SupT1, that had been infected with Vif-negative HIV-1 containing or lacking hA3G. Equal amounts of DNase-treated virions (100 ng of p24) were used to infect 2.4 ϫ 10 7 SupT1 cells rotating at 4°C for 2 h. The cells with bound viruses were washed twice with phosphate-buffered saline, and aliquots of ϫ 10 6 cells were plated into 24-well plates containing complete RPMI 1640 medium pre-warmed to 37°C and incubated at 37°C. At different time points postinfection, equal aliquots of cells were collected and washed with phosphate-buffered saline, and cellular DNA was extracted using the DNeasy tissue kit (Qiagen). Equal amounts of cellular genomic DNA (determined spectrophotometrically at an optical density of 260 nm) were used to quantitate viral cDNA intermediates containing sequences for either R-U5, R-U3, pol, gag, or U5-gag, which represents, respectively, minus strand strong stop DNA, minus strand transfer DNA/early minus single strand DNA, middle minus single strand DNA, late minus single strand DNA, and plus strand-transfer DNA. Five pairs of primers were used: U5-R forward, 5Ј-TTAGACCAGATCTG-AGCCTGGGAG; U5-R reverse, 5Ј-GGGTCTGAGGGATCT-CAGTTACC; R-U3 forward, 5Ј-CGGGACTGGGGAGTGGC-GAGC; R-U3 reverse, 5Ј-CAGAGTCACACAACAGACG-GGC; pol forward, 5Ј-GGGAGCCACACAATGAATG; 5Јpol reverse, 5Ј-CCAGGGCTCTAGTCTAGGATC; gag forward, 5Ј-GGCGGCGACTGGTGAGTAC; gag reverse, 5Ј-CCCTGC-TTGCCCATACTATATG; U5-gag forward, 5Ј-TGTGTGCC-CGTCTGTTGTGTGA; and U5-gag reverse, 5Ј-GAGTCCTG-CGTCGAGAGAGCT. All analyses were done in triplicate, with triplicate samples in each experiment. The statistical analyses employed herein include column statistics and one-way analysis of variance. The lowest level of significance was set at p Ͻ 0.05.
Quantitation of Viral Genomic RNA in the Cell-SupT1 cells were infected with equal amounts of wild-type or Vif-negative HIV-1, and then were collected at the different time point of post-infection, as above described. Total cellular RNA was extracted with TRIzol (Invitrogen), and ␤-actin mRNA was quantitated using real-time reverser transcription (RT)-PCR as described previously (28). Samples of total cellular RNA containing equal amounts of ␤-actin mRNA were used to generate cDNA containing the 5Ј-long terminal repeat of the viral genome, using SuperScript TM II reverse transcriptase (Invitrogen) and the U5-R primer, 5Ј-GGGTCTGAGGGATCTCAG-TTACC. The amount of cDNA containing the 5Ј-long terminal repeat of the viral genome was determined using real-time fluorescence-monitored PCR with a pair of primers U5-R forward and reverse, as described above.
In Vitro Analysis of Minus Strand Transfer-Using the MEGAscript kit (Ambion), an RNA template of 384 nt was synthesized from the linearized Tp-2 DNA (29). The RNA template contained both the 5Ј and 3Ј R regions and is competent to support strand transfer. An 18-nt DNA oligonucleotide (5Ј-GTCCCTGTTCGGGCGCCA), complimentary to the PBS, was used as primer in the reactions described below. The DNA oligomer, 5Ј-end-labeled using [␥-32 P]ATP and T4 kinase, was annealed to the RNA template by heating at 85°C for 5 min and gradually cooling to room temperature. Final reactions contained 2.25 M NCp7, 50 nM RNA template, 100 nM primer DNA, 0.1 mM of each dNTP, 100 ng of RT, 50 mM Tris-Cl (pH 7.5), 75 mM KCl, 5 mM Mg 2 Cl, 1 mM dithiothreitol, and 10 units of RNase inhibitor (Ambion), in a volume of 20 l. Various concentrations of purified hA3G or hA3G-containing cell lysate were added to the reaction mixtures to test their effects on minus strand transfer. Reactions were incubated at 37°C for 40 min, followed by the addition of 1 l of proteinase K, and further incubated at 65°C for 20 min. After extraction by phenol/chloroform, the deproteinized reaction products were resolved by electrophoresis in a 5% denaturing polyacrylamide gel containing 7 M urea, and the different bands were quantitated by phosphorimaging (Molecular Dynamics). Minus strand transfer efficiency is determined by the ratio of T-DNA (transfer DNA, 296 nt): total DNA (minus strand strong stop (Ϫsss) DNA (200 nt) plus T-DNA).
Recombinant hA3G was produced in the Baculo Expression System as a T-tag fusion protein, and was provided by the NIH AIDS Research and Reference Reagent Program (Catalog number 10068). The 72-amino acid HIV-1 NCp7 peptide used in this analysis was prepared by solid-phase chemical synthesis as previously described (30). Recombinant HIV-1 RT (p66/51) was prepared as previously described (31). QKI-6 RNA-binding protein was a gift from Stephane Richard (McGill University). For analysis of the effect of 293T cell lysate-expressing hA3G upon the minus strand transfer in vitro, the transfected 293T cells were lysed using CytoBuster TM protein extraction reagent (Novagen) in the presence of SUPERase.In TM (Ambion), which is effective only against RNase A-type enzymes.
hA3G Deamination of Ϫsss cDNA during in Vitro Minus Strand Transfer-The sequencing of Ϫsss DNA was performed after the minus strand transfer reaction. The total DNA in the reaction sample was extracted using the DNeasy tissue kit (Qiagen Inc.). PCR was performed with Platinum Taq polymerase (Invitrogen). The primers were as follows: forward (469 -492), 5Ј-CCAGATCTGAGCCTGGGAGCTC; reverse (764 -789), 5Ј-CTCCTTCTAGCCTCCGCTAGTC. The PCR products were cloned into pCR4-TOPO vector (Invitrogen), and individual clones were sequenced.
In Vitro RNA Template Degradation Assay-The minus strand transfer assay described above was used with a synthetic RNA template labeled by the incorporation of [␣-32 P]UTP and an unlabeled DNA oligonucleotide primer in the presence or absence of NCp7. The reaction was terminated at 20 min by adding 2 l of 0.5 M EDTA, and the labeled RNA content was monitored by one-dimensional PAGE. RNaseH-negative HIV-1 RT E478Q, a kind gift from Matthias Götte (32), was used as a control.

Characterization of the Inhibitory Effect of hA3G on Viral DNA Synthesis in Vivo-Upon infection of cells, Vif-negative
HIV-1-containing hA3G show a strong reduction in viral DNA synthesis compared with HIV-1 not containing hA3G (10 -13). Reports of the stage at which the initial viral DNA production is blocked have varied. Quantitative PCR analyses of endogenous RT transcripts have shown reduced production of both early and late reverse transcripts in some cases (10,13), whereas another report showed reduction of only the later RT transcripts (11). We have previously reported ϳ95% reduction in viral DNA production, with ϳ55% of this reduction due to a reduction in the synthesis of strong stop DNA due to the inhibition of tRNA Lys3 priming (22). The cause of the further reduction in late DNA production is addressed in this report.
The conversion of the HIV-1 genomic RNA into doublestranded proviral DNA requires two strand transfers, as shown in Fig. 1A (for recent review, see Ref. 35). The RTcatalyzed synthesis of minus strand strong-stop cDNA is initiated from a tRNA Lys3 bound to a sequence in the 5Ј-region of the viral RNA known as the primer binding site (PBS). Synthesis of this DNA is accompanied by degradation of the RNA template by RT-associated RNaseH, which allows the DNA to undergo the first strand transfer, annealing to the R sequences at the 3Ј terminus of the viral RNA. The synthesis of minus single strand cDNA then proceeds up through the PBS sequence and is accompanied by RNaseH degradation of the RNA template, except for a short, RNaseH-resistant polypurine tract, located at the 5Ј-region of U3. This poly-purine tract initiates plus strand DNA synthesis, which terminates after the 3Ј 18 nucleotides of tRNA Lys3 primer have been transcribed, i.e. a new PBS is generated. Removal of the tRNA Lys3 then allows for the plus strand DNA to undergo a second strand transfer, accompanied by annealing to complementary PBS sequences in the minus strand DNA. Plus and minus strand DNA synthesis are then completed, with the plus and minus strands of DNA each serving as a template for completion of the other strand.
In this work, we have investigated whether the reduction in the production of late DNA synthesis is due to the inhibition by hA3G of either DNA elongation, strand transfers, or of both these processes. Using real-time fluorescence-monitored PCR with equal amounts of cellular DNA, we monitored over a 24-h period post-infection the synthesis of viral cDNA intermediates in a permissive cell line, SupT1, that had been infected with Vif-negative HIV-1 containing or lacking hA3G. As shown in Fig. 1A, five pairs of primers were used to quantitate viral cDNA intermediates containing sequences for either R-U5, R-U3, pol, gag, or U5-gag, which represent, respectively, Ϫsss DNA, postminus strand transfer of early, middle, and late minus single strand DNA, and post-plus strand transfer late DNA. As shown in Fig. 1 (B-F), Ϫsss DNA (R-U5) and late viral DNA production after the second strand transfer (U5-gag) reached maximum concentrations at 8 and 12 h post-infection, respectively, with intermediate DNA species reaching peak concentrations at intermediate times.
A comparison of the synthesis of each DNA region in the absence or presence of hA3G, is presented graphically in Fig.  1G. In agreement with our previous report (22), the presence of hA3G in Vif-negative HIV-1 caused a 48% reduction in Ϫsss DNA and a 96% reduction in the full-length minus single strand DNA. A new observation is that the reduction in late DNA synthesis appears to reflect inhibition of the DNA strand transfer steps and not the rates of DNA elongation. Thus, the similar reductions in the three minus strand viral DNA intermediates produced after the first strand transfer (84%, 83%, and 85% for R-U3, pol, and gag, respectively) suggests that hA3G has no significant impact on the extension of minus strand viral DNA after the minus strand transfer. However, it can be seen that an additional drop in DNA production occurs after the second strand transfer (U5-gag). These results suggest that the reduction in late viral DNA synthesis by hA3G may be caused primarily by the inhibition of strand transfers during reverse transcription, and this is further examined below using an in vitro minus strand transfer system.
hA3G Inhibits the Minus Strand Transfer in Vitro-The ability of hA3G to inhibit minus strand transfer was measured using an in vitro assay that is described under "Experimental Procedures" and is depicted in Fig. 2A. A DNA primer, 5Ј-endlabeled with 32  A recent study has revealed that ϳ7 (Ϯ4) molecules of hA3G are incorporated into virions produced from activated human peripheral blood mononuclear cells (36). This result suggests that, during reverse transcription, the molar ratio of hA3G to HIV-1 genomic RNA is ϳ3.5:1. To avoid a biologically irrelevant effect resulting from a high concentration of hA3G, we have tested low concentrations of purified hA3G, from 0.05 to 12.5 nM in the reactions, in which the molar ratio of hA3G to the 50 nM RNA template never exceeds 0.25. Similarly, estimates of NCp7 molecules/ virion have ranged from 1400 (37) to 5000 (38). Thus the hA3G:NCp7 molar ratio in virions may be 0.001-0.005. The in vitro strand transfer reaction contains 2.25 M NCp7, and thus never exceeds an hA3G: NCp7 molar ratio of 0.006.
As shown in Fig. 2B, with increasing concentrations of purified hA3G added to the strand transfer reaction, the synthesis of T-DNA is gradually reduced (lanes 4 -8), and Fig. 2C plots the relative amounts of total DNA and DNA resulting from minus strand transfer (T-DNA/total DNA) obtained with increasing hA3G, whose concentrations are plotted on a logarithmic scale. Inhibition of minus strand transfer by hA3G increases from 11% to 76% as the hA3G concentrations increase from 0.2 nM to 3.2 nM. Experiments in which hA3G was replaced with bovine serum albumin did not result in any inhibition of minus strand transfer (data not shown). hA3G binds to both DNA and RNA (2, 39), but the replacement of hA3G with another RNA-binding protein, QKI-6 (lane 10) (40), has no effect upon strand transfer, making it less likely that hA3G inhibits strand transfer by sterically blocking the annealing of DNA to viral RNA.
Although hA3G strongly reduced the T-DNA product, it did not significantly affect the total DNA synthesis (Ϫsss DNA and T-DNA), because the amount of Ϫsss DNA increases when T-DNA synthesis decreases (Fig. 2, B and C). This observation suggests that the reduction in strand transfer DNA product induced by hA3G did not result from a general inhibition of reverse transcription by hA3G. This is shown further in Fig. 2D, in which increasing concentrations of hA3G are added in the absence of NCp7, and no changes in the synthesis of Ϫsss DNA are seen.

Inhibition of Minus Strand Transfer Occurs
Independently of hA3G Editing-We next investigated whether the hA3G inhibition of minus strand transfer depends upon the editing activity of this enzyme. The editing activity of hA3G is specific for single-stranded DNA (41), and only the Ϫsss DNA is a potential editing substrate for hA3G during minus strand transfer. As indicated in Fig. 1A, minus strand transfer is facilitated by annealing of sequences in the Ϫsss DNA (RЈ) to the complementary R sequence at the 3Ј terminus of viral RNA. C to U mutations introduced into RЈ could hinder first strand transfer. We therefore cloned and sequenced the R-U5 region of Ϫsss DNA produced from the in vitro strand transfer reaction. Ten independent nucleotide sequences were determined for Ϫsss DNAs that had been produced in the absence or presence of hA3G, and no C to T (U) mutations were detected. This confirms previous observations that DNA mutations induced by hA3G were least frequent in the 5Ј-long terminal repeat, probably representing less than 3% of total cytidines within the R-U5 region (41).
We also tested the ability of mutant hA3G lacking editing activity to inhibit minus strand transfer in vivo, determined by ratios of the early DNA intermediate synthesized immediately after the minus strand transfer in vivo (U3-R) to Ϫsss DNA, using real-time fluorescence-monitored PCR, as described above. The mutant forms of hA3G tested are shown in Fig. 3A. hA3G 1-156 and hA3G 105-384 contain the N-or C-terminal zinc coordination motifs, respectively, whereas hA3G 104 -245 contains the linker sequences between these two motifs. We have previously reported that all three hA3G fragments were incorporated efficiently into HIV-1, but none were found to deaminate viral DNA in vivo (22,27). 293T cells were transfected with plasmids coding for either wild-type or mutant forms of hA3G (Fig. 3A), and Western blots of viral lysates (Fig. 3B) show approximately equal viral content of these wild-type and mutant hA3G species. SupT1 cells were then infected with equal amounts of viruses containing similar amounts of wt or mutant hA3G species, and the effect of each mutant form of hA3G on minus strand transfer was quantitated. The results are graphically presented in Fig. 3B. Both hA3G 1-156 and hA3G 105-384, which show very little editing activity, inhibit the first strand transfer in vivo 60 -70% as efficiently as wild-type hA3G. hA3G 104 -245, missing both zinc coordination motifs, has almost no effect on the strand transfer. These results indicate that the inhibition of strand transfer can occur independently of the cytidine deamination editing activity of hA3G.
We have also analyzed the effect of the wild-type and mutant hA3G species upon minus strand transfer in vitro. Because we were unable to purify mutant hA3G fragments, we have tested the abilities of lysates of 293T cells expressing one or another of these species to inhibit minus strand transfer in vitro. An RNase inhibitor, SUPERase.In TM , was added during the preparation of cell lysates. This inhibitor is effective only against RNase A-type enzymes, but not RNaseH, and therefore does not interrupt the strand transfer assay. The Western blot in Fig. 4B shows the content of each hA3G species in the cell lysate per equal amounts of ␤-actin.
As shown in Fig. 4B, increasing amounts of 293T cell lysate (25-200 ng of cell protein) containing wild-type hA3G caused a gradual reduction in T-DNA products in a dose-dependent manner (lanes 3-6), whereas addition of cell lysate not expressing hA3G had no effect on minus strand transfer (lanes 7-10). In Fig. 4C, the ratios (T-DNA/total DNA) obtained relative to the ratio obtained for strand transfer in the absence of hA3G (panel A, lane 2) are plotted, and it can be seen that 1) minus strand transfer was inhibited from 12% to 74%, and 2) no significant reduction in total DNA synthesis was detected. These results demonstrate that, as with purified hA3G, hA3G present in cell lysates can be used to interrupt minus strand transfer in vitro, and that neither RNases nor proteases in the cell lysate modulate this process. We next investigated the inhibitory effect of mutant hA3G mutants expressed in 293T cells on in vitro minus strand transfer. After 293T cells were transfected with plasmids coding for either wild-type or mutant forms of hA3G, Western blots of 293T cell lysates showed that the expression of wild-type and mutant hA3G are similar (Fig. 4A). The effect of each mutant form of hA3G in the cell lysates (200 or 50 ng of total cellular protein) on minus strand transfer were quantitated and graphically presented in Fig. 4D. The general pattern of inhibition of minus strand transfer in vivo is similar to that obtained for the mutant forms of hA3G obtained in vitro, although hA3G105-384 and hA3G1-156 possess only ϳ50% the inhibitory activity as wild-type hA3G. These results indicate that the inhibition of strand transfer, either in vivo or in vitro, can occur independently of the cytidine deamination editing activity of hA3G.
Inhibition of the RNaseH-dependent Degradation of Viral RNA by hA3G-During the synthesis of Ϫsss DNA, the RNaseH activity associated with HIV RT degrades the RNA template, a process required for release of the Ϫsss DNA during minus strand transfer (42). We have therefore examined, in vivo and in vitro, whether the inhibition of minus strand transfer by hA3G is associated with a decrease in the degradation of the RNA template by RNaseH. 293T cells were cotransfected with a plasmid containing either BH10 or BH10VifϪ DNA, and with the pcDNA3.1 plasmid, either empty or containing DNA coding for hA3G. Thus, four types of viruses were produced: wildtype viruses (BH10) in the absence or presence of hA3G and Vif-negative viruses (BH10VifϪ) in the absence or presence of hA3G. We then monitored the content of viral genomic RNA in SupT1 cells infected with equal amounts of one of the four types of virions over 12 h post-infection, using real-time fluorescence-monitored RT/PCR on equal amounts of cellular nucleic acid first treated with DNase. One primer pair was used to reverse transcribe and amplify the U5 region at the 5Ј-end of viral RNA genome, which serves as part of the template for the synthesis of Ϫsss DNA, and is degraded by RNaseH to facilitate release of Ϫsss DNA for the minus strand transfer. As shown in Fig. 5A, at 8 h post-infection, the amount of the U5 RNA detected in SupT1 cells infected with virions containing little or no hA3G was reduced to Ͻ2% of the RNA present at time zero, whereas 76% of the U5 RNA still remained in the cells infected with Vif-negative HIV-containing hA3G. These data suggest that hA3G blocks the degradation of the RNA template during reverse transcription.
In addition, we have also performed a quantitative comparison of the effect of hA3G upon viral infectivity and U5 RNA degradation, using Vif-negative viruses produced from 293T cells transfected with 2 g of HIV-1 proviral DNA and different amounts of hA3G expressing plasmid. Viral infectivity was measured in CD4 ϩ HeLa cells, using the multinuclear-activation galactosidase indicator (MAGI) assay, as described under "Experimental Procedures." The same aliquot of virions were used to infect SupT1 cells, and viral U5 genomic RNA content was monitored by real-time RT-PCR. As shown in the right panel of Fig. 5A, it is clearly evident that the inhibition of U5 RNA degradation by hA3G directly correlates with its inhibition of viral infectivity.
We further investigated the effect of hA3G on the degradation of RNA template using the in vitro minus strand transfer assay described in Fig. 2, except that the synthetic RNA template was labeled by the incorporation of [␣-32 P]UTP, and an unlabeled DNA oligonucleotide primer was used. The fulllength (384 nucleotides) labeled RNA was resolved by one-dimensional PAGE. Under reverse transcriptase conditions, but in the absence of added RT and hA3G, the full-length RNA remains intact (Fig. 5B, lane 7), whereas incubation with wildtype RT for 20 min generates one major 200 nucleotide cleavage product, accompanied by a Ͼ95% reduction in full-length RNA (Fig. 5B, lane 1). The addition of increased concentrations of hA3G (0.8, 3.2, and 12.5 nM) was accompanied by increased amounts of full-length RNA in a dose-dependent manner,  The in vivo inhibition of strand transfer was measured using real-time RT-PCR, as described in Fig. 1. The cytidine deaminating activity of wild-type and mutant hA3G species was determined by sequencing viral DNA in the infected cell, as previously described (27). The bar graphs in panel C represent the means of results of experiments performed at least three times, and the error bars represent standard deviations.

Deamination Inhibition of minus strand transfer h A 3 G 1 0 5 -3 8 4 h A 3 G 1 -1 5 6 h A 3 G h A 3 G 1 0 5 -2 4 5 h A 3 G 1 0 5 -3 8 4 h A 3 G 1 -1 5 6 h A 3 G h
resulting in 11%, 24, and 37% of full-length RNA remaining (Fig.  5B, lanes 2-4). The use of RNaseH(Ϫ) RT for reverse transcription resulted in no significant reduction in the RNA template, with or without the treatment of 12.5 nM hA3G (Fig. 5B, lanes 5  and 6), indicating that the decrease in RNA template observed did result from RNaseH-mediated RNA degradation, and that the retention of RNA template by hA3G reflects its ability to interrupt the degradation process. HIV-1 NCp7 has been shown to facilitate the degradation of RNA template during DNA strand transfer (43). Because an interaction of hA3G with NCp7 may facilitate the incorporation of hA3G into virions, we investigated whether such an interaction might also be responsible for the inhibition of RNA template degradation, by performing the in vitro assay described above in the absence of NCp7. In Fig. 5C, we have plotted the amount of RNA template remaining during reverse transcription with or without NCp7 in the presence of increasing hA3G and compared this to a plot of the amount of first strand transfer occurring with increasing hA3G (obtained from Fig. 2C). Comparison of the three plots indicates that reduced RNA template degradation occurs independently of the pres-ence of NCp7 and is correlated with the inhibitory effect of hA3G on the minus strand transfer.

DISCUSSION
In this report and elsewhere (22), we have presented evidence that a strong reduction in reverse transcription plays an important role in the inhibition of HIV-1 replication by hA3G. We have previously reported that the reduction in production of Ϫsss DNA (ϳ55%) is correlated with a similar reduction in the ability to initiate reverse transcription (22), whereas in this report, evidence is presented that the reduction in late DNA synthesis (ϳ95%) is correlated with an hA3G-induced inhibition of minus and plus strand transfer. As shown in Fig. 1, hA3G inhibits the minus and plus strand transfers during reverse transcription in vivo. In the presence of viral hA3G, Ϫsss DNA synthesis decreases ϳ2-fold. Synthesis of viral DNA immediately after the minus strand transfer (U3-R) decreases another 3-to 4-fold, without further reduction in the synthesis of later minus strand DNA sequences (pol and gag). However, synthesis of plus strand DNA after the plus strand transfer (U5-gag) undergoes another 3-to 4-fold decrease. To test our interpre-

Inhibition of Reverse Transcription in HIV-1 by APOBEC3G
NOVEMBER 2, 2007 • VOLUME 282 • NUMBER 44 tation that these changes in vivo represent inhibition of minus and plus strand transfer, we have analyzed the ability of hA3G to inhibit minus strand transfer in vitro. In Fig. 2, we show that the inhibition of minus strand transfer by hA3G is dependent upon the concentration of hA3G. Furthermore, hA3G does not inhibit DNA-primed synthesis of strong stop DNA, thus supporting our in vivo results that hA3G does not effect DNA elongation. The fact that hA3G does inhibit the initiation of reverse transcription when tRNA Lys3 is used as a primer (22), but not when a DNA primer is used, may be explained by the fact that the tRNA Lys3 annealing to the viral RNA template in vivo is facilitated by nucleocapsid sequences (22), whereas the DNA primer is annealed in vitro using heat.
These data indicate that hA3G inhibits strand transfer during reverse transcription so as to reduce late viral DNA production. Our results also do not support a reduced viral DNA production due to the degradation of newly synthesized DNA. In particular, the reduction in late DNA production in vitro in the presence of purified hA3G cannot be ascribed to degradation by enzymes of the cellular DNA repair system, because none are present in the reaction. Furthermore, as shown in Figs. 3 and 4, fragments of hA3G containing either the N-or C-terminal zinc coordination unit could inhibit first strand transfer 60 -70% as well as wild type, even though neither fragment has cytidine-deamination activity. We have previously reported that these same N-and C-terminal hA3G fragments also inhibit viral DNA production in vivo at levels 60 -70% that of wild-type hA3G (22). This reduction in the efficacy of inhibition of viral DNA synthesis by these hA3G fragments could mean the conformation of either fragment is sub-optimal compared with the whole molecule or that both zinc-coordination units may be required for full inhibition of viral DNA synthesis. For the latter case, it is not clear if this reflects a need for either editing activity or some other function of these two domains.
It can also be noted that, in the small amount of viral DNA that is synthesized in the cell in the presence of hA3G, the frequency of deamination mutations is highest in coding regions of the genome that are reverse transcribed first (41). However, there is little if any difference in the inhibition of synthesis of minus strand DNA sequences representing gag, pol, or U3, implying that the mechanism for viral DNA reduction is not revealed in the mutational frequencies found in viral DNA that is synthesized. During minus strand transfer, the release of the annealed DNA from the 5Ј-region of template RNA requires the degradation of template RNA by the RNaseH activity associated with reverse transcriptase. Fig. 5 demonstrates that hA3G inhibits the degradation of the template RNA, both in vivo and in vitro. The mechanism for this remains unclear. Because hA3G has been reported to bind to RT (44), there might be a direct interaction between hA3G and RT that reduces RNaseH activity without effecting DNA elongation, because the polymerase activity of HIV-1 RT is not coupled to its RNaseH activity (45). Alternatively, because hA3G is able to bind efficiently to singlestranded DNA and RNA (39), the interaction of hA3G with the RNA/DNA hybrid may make the substrate less available to RNaseH.
hA3G may also disrupt NCp7 function in reverse transcription by binding to NCp7. hA3G inhibits both tRNA Lys3 priming (22) and DNA strand transfer, and both processes are facilitated by NCp7 (42), whose sequences in Gag are also required for the incorporation of hA3G into HIV-1 (27, 46 -49). NCp7 is required for strand transfer, but the experiment shown in Fig.  5B is not a strand transfer reaction, i.e. inhibition of cleavage of RNA template by hA3G occurs independently of strand transfer, and of NCp7. So, although hA3G could inhibit strand trans-

FIGURE 5. Inhibition of viral RNA template degradation by hA3G in vivo and in vitro.
A, RNA degradation measured in vivo. Left panel: total cellular RNA was extracted at different times post-infection from SupT1 cells infected with Vifϩ or VifϪ BH10, lacking or containing hA3G. Vifϩ and VifϪ virions containing hA3G were produced from 293T cells transfected with 2 g of HIV-1 proviral DNA and l g of hA3G plasmid (pAPOBEC3G). Viral U5 genomic RNA content was monitored by real-time RT-PCR. Right panel: relationship between viral U5 genomic RNA degradation and virus infectivity. VifϪ virions containing hA3G were produced from 293T cells transfected with 2 g of HIV-1 proviral DNA and different amounts of pAPOBEC3G and measured for infectivity in CD4 ϩ HeLa cells, using the multinuclear-activation galactosidase indicator (MAGI) assay, as described under "Experimental Procedures." The same aliquot of virions were used to infect SupT1 cells, and viral U5 genomic RNA content was monitored by real-time RT-PCR. B, RNA degradation measured in vitro. The in vitro minus strand transfer reaction conditions were used, and a 32 P-labeled synthetic RNA template was incubated alone (lane 1) or with increasing amounts of hA3G (lanes 2-4). The full-length 384-nt template RNA and the 200-nt cleavage product were resolved by one-dimensional PAGE. Lanes 5 and 6, RNaseH(Ϫ) RT was used with or without the presence of 12.5 nM hA3G, respectively. Lane 7, the reaction contained neither RT nor hA3G. C, plots of minus strand transfer (Fig. 2B) and the amount of RNA template remaining in the presence or absence of NCp7, with increasing hA3G concentrations. fer by inhibiting template RNA cleavage, hA3G may alternatively, or in addition to inhibiting RNA cleavage, inhibit strand transfer through its interaction with NCp7. NCp7 is known to be important in facilitating strand transfer by both facilitating removal of cleaved RNA from the DNA template (which was not measured in the experiments in Fig. 5) and through facilitating annealing of the DNA to the RNA acceptor (42). Whether hA3G has an affect on either of these two steps in strand transfer is not known, but this remains a possibility because it has been reported that hA3G inhibits tRNA Lys3 annealing to viral RNA through its interaction with NCp7 (50). Indeed, as shown in Fig. 5C, the addition of hA3G is able to inhibit ϳ80% of minus strand transfer in vitro but only 37% of RNA template degradation.
We have previously shown that, in HIV-1 produced from 293T cells transfected with 1 g of hA3G plasmid, the viral content of hA3G is ϳ10 times greater than in virions produced from the naturally non-permissive H9 cell line (22). The question arises as to whether the inhibition of strand transfer during viral DNA synthesis that occurs after viral infection of SupT1 cells is an artifact of this increased content of viral hA3G. We suggest that this is not likely, based upon studies on the effect of hA3G on tRNA Lys3 -primed initiation of reverse transcription (22). Virions produced from H9 cells show a reduction of Ͼ50% in initiation of reverse transcription upon infecting new cells. Virions produced from 293T cells expressing hA3G also show a similar reduction in tRNA Lys3 priming upon infection but require 10-fold more hA3G in the virions to achieve this same level of reduction. Thus, the reduction of tRNA Lys3 priming in H9 cell-produced virions indicates that this phenomenon in 293T cell-produced virions is not the result of overexpression of hA3G. Rather, it seems more likely that 293T cells expressing hA3G are not equivalent to naturally non-permissive cells such as H9, which probably contains other cellular factors contributing to the non-permissive phenotype that may be incorporated into the virus, and that may enhance the effects of hA3G. The study of the effect of viral hA3G upon strand transfer in Vif-negative HIV-1 produced from naturally non-permissive cells such as H9 cells is technically difficult because of the low titers of virions produced as a result of adsorption and nonproductive infection of surrounding H9 cells.
Some HIV-1 DNA is made in the presence of hA3G, and recent reports have indicated that hA3G inhibits integration of this DNA (51,52). Mibisa et al. (52) observed that, prior to plus strand transfer, hA3G interferes with correct cleavage of tRNA Lys3 attached to the 5Ј terminus of minus strand DNA. This results in abnormal termination of the 3Ј terminus of fulllength plus strand DNA, suggesting that this may be responsible for the inefficient integration of that viral DNA. The ability of hA3G to inhibit correct RNA processing is in agreement with our own observations reported in Fig. 5, which show that hA3G inhibits template RNA degradation occurring prior to minus strand transfer. On the other hand, while Mibisa et al. (52) found that hA3G inhibited plus strand transfer during HIV-1 reverse transcription, they did not detect an inhibition of minus strand transfer (52). There is, however, the possibility that the hA3G:viral RNA ratio in their virions is much lower than in our system due to the overproduction of HIV-1-derived vector par-ticles (36). As we described in the previous paragraph, this could result in a lack of effects that hA3G has upon reverse transcription in genuinely non-permissive cell types due to the absence of non-permissive cell-associated factors in the naturally permissive 293T cells. There are also other basic differences between the biological system they used, and the one we use in this report. In our report, 293T cells were transfected with DNA coding for Vif-positive or Vif-negative HIV-1, and DNA synthesis was studied in SupT1 cells infected with the HIV-1 obtained from these transfected cells. In their report, 293T cells were transfected with an HIV-1-derived vector containing a truncated viral RNA, expressing only Gag, GagPol, Tat, and Rev, and missing sequences coding for envelope and the accessory proteins Vif, Vpr, Vpu, and Nef. Nef was further replaced with sequences coding for green fluorescent protein, and viral incorporation of the accessory proteins was facilitated by a second helper virus containing non-packageable RNA that expressed these proteins. The resulting HIV-1 vector particles were also pseudotyped with vesicular stomatitis virus (VSV) envelope, and DNA synthesis was studied by infecting 293T cells with these particles. Thus, in studying reverse transcription in their system, the cell type used was different, the method of viral entry into the cell was different, and the RNA to be reverse-transcribed was truncated. Whether any of these properties can alter either the reverse transcription pattern that occurs upon native HIV-1 infection, or the effect of hA3G on reverse transcription, is not clear. It can be noted, however, that these authors monitored the DNA resulting from minus and plus strand transfers at 6 h post-infection, whereas our data ( Fig. 1) indicate very little viral DNA being produced at 6 h post-infection, with DNA species resulting from the minus and plus strand transfers peaking at 8 and 12 h, respectively.