APOBEC3F Can Inhibit the Accumulation of HIV-1 Reverse Transcription Products in the Absence of Hypermutation

APOBEC3F (apolipoprotein B mRNA-editing enzyme catalytic polypeptide 1-like protein 3F) is a cytidine deaminase that, like APOBEC3G, is able to restrict the replication of HIV-1/ Δvif. Initial studies revealed high numbers of mutations in the cDNA of viruses produced in the presence of these proteins, suggesting that cytidine deamination underpinned the inhibition of infection. However, we have recently shown that catalytically inactive APOBEC3G proteins, derived through mutation of the C-terminal cytidine deaminase motif, still exert a substantial antiviral effect. Here, we have generated a panel of APOBEC3F mutant proteins and show that the C-terminal cytidine deaminase motif is essential for catalytic activity and that catalytic activity is not necessary for the antiviral effect of APOBEC3F. Furthermore, we demonstrate that the antiviral activities of wild-type and catalytically inactive APOBEC3F and APOBEC3G proteins correspond well with reductions in the accumulation of viral reverse transcription products. Additional comparisons between APOBEC3F and APOBEC3G suggest that the loss of deaminase activity is more detrimental to APOBEC3G function than to APOBEC3F function, as reflected by perturbations to the suppression of reverse transcript accumulation as well as antiviral activity. Taken together, these data suggest that both APOBEC3F and APOBEC3G are able to function as antiviral factors in the absence of cytidine deamination, that this editing-independent activity is an important aspect of APOBEC protein-mediated antiviral phenotypes, but that APOBEC3F may be a better model in which to study it.

Initial reports indicated that inhibition of HIV-1 infection by APOBEC proteins corresponded to high G-to-A mutation frequencies (3,5,6,8,12,14). However, the following lines of evidence demonstrate that antiviral effects can be exhibited in the absence of measurable editing. First, it was shown that APOBEC3G, and more recently APOBEC3F, can inhibit the replication of hepatitis B virus, but with no or very little G-to-A hypermutation being evident (27)(28)(29)(30). Second, mutated APOBEC3G proteins carrying disruptions in the C-terminal deaminase motif that prevent editing still retain significant antiviral activity (31). Third, in a recent flurry of publications, it was reported that several APOBEC proteins are able to inhibit long terminal repeat (LTR) and non-LTR retrotransposons by nonediting mechanisms (32)(33)(34)(35)(36)(37). Fourth, APOBEC3G residing in resting human CD4ϩ T-lymphocytes has been shown to be responsible for the inability of these cells to support productive HIV-1 infection, again with no detectable hypermutation in the limited numbers of cDNAs that accumulate under these conditions (38).
The mutational potency of APOBEC3F appears to be less than that of APOBEC3G, and the antiviral effect is less pronounced (3, 6, 10, 13, 39 -41). In light of these findings, and the notion that APOBEC proteins can exert antiviral (or antiretrotransposon) effects independent of editing, we set out to determine whether nonediting variants of APOBEC3F could inhibit HIV-1 infection and to compare the phenotypes of such proteins with analogous mutants of APOBEC3G (31). Accordingly, we demonstrate that mutations introduced at the conserved Cys, His, or Glu residues of the C-terminal deaminase motif of APOBEC3F abolish cytidine deaminase activity, with little effect on antiviral potency. Corresponding mutations in the N-terminal deaminase motif do not alter editing function but do decrease packaging efficiency into virions, even though significant amounts of antiviral activity are retained. However, because doubly mutated proteins carrying matching alterations in both motifs fail to package, it seems that both domains play some role in the incorporation of APOBEC3F into virions. Through measurements of nascent reverse transcript levels in target cells, we show that the antiviral activity of wild-type and mutant APOBEC3F and APOBEC3G proteins correspond well with reductions in the accumulation of HIV-1 cDNAs in target cells, suggesting that this is an important additional attribute that contributes to the antiviral phenotypes of APOBEC proteins. Finally, dose-response comparisons of the antiviral activities of wild-type and mutant APOBEC3G/F proteins reveal that the loss of editing capability places a much higher cost on the function of APOBEC3G relative to APOBEC3F. This implies that APOBEC3F may be a better model with which to study the mechanism by which APOBEC proteins suppress the accumulation of viral cDNA replication intermediates

EXPERIMENTAL PROCEDURES
Plasmid Constructs-The human APOBEC3F cDNA was a kind gift from R. Harris (University of Minnesota) and was cloned into pcDNA3.1 as a KpnI to XhoI fragment. The sequence of this APOBEC3F protein (NM_145298) differs from that previously described by our group (3) in that residue 108 is alanine instead of serine, residue 123 is valine instead of isoleucine, residue 272 is threonine instead of alanine, and residue 370 is glutamic acid instead of glycine. APOBEC3F mutant constructs were generated by site-directed mutagenesis (Stratagene) and verified by DNA sequencing. Wild-type and mutant APOBEC3G constructs have been described previously (31). Provirus expression vectors for wild-type and ⌬vif HIV-1 (pIIIB and pIIIB/⌬vif, respectively), were modified by site-directed mutagenesis at nucleotide 567 to create a G-to-A substitution in the U5 region of the 5Ј-LTR that is copied to the 3Ј-LTR during reverse transcription (3). This modification was introduced to ensure that all DNA sequencing data reflected the products of reverse transcription and not DNA from contaminating transfection cocktails.
Virus Production and Single Cycle Infectivity Assays-All of the cell lines were maintained under standard conditions. HIV-1 stocks were prepared by co-transfection of 293T cells with proviruses and wild-type or mutant APOBEC expression vectors (or control vectors) at a 1:1 ratio, unless otherwise indicated. All of the viruses were pseudotyped with the G protein of vesicular stomatitis virus (42). The medium was changed after ϳ6 h, and the viruses were harvested 24 h later and quantified by enzyme-linked immunosorbent assay for p24 CA . Viral infectivity was determined by challenging 1 ϫ 10 5 target TZM-␤-gal indicator cells (43) with viruses corresponding to 4 ng of p24 CA and measuring the accumulation of ␤-galactosidase activity at ϳ24 h using the Galacto-Star system (Applied Biosystems).
Antibodies and Immunoblot Analysis-Polyclonal antibodies specific for human APOBEC3F were raised in rabbits immunized with a synthetic peptide with an introduced Cys at the N terminus to allow coupling (CGLKYNFLFLDSKLQEILE), and purified by affinity chromatography according to standard procedures. The 14-3-3␥ protein was detected using a rabbit polyclonal antibody (C-16; Santa Cruz Biotechnology), and p24 CA was detected using the 24-2 monoclonal antibody (44). Virions were quantified by enzyme-linked immunosorbent assay, and material corresponding to 20 ng p24 CA was loaded onto a 20% (w/v) sucrose cushion in phosphate-buffered saline and concentrated by centrifugation at 20,000 ϫ g for 2 h at 4°C prior to standard immunoblot analysis. APOBEC3F proteins and 14-3-3␥ were detected in whole cell lysates of transfected 293T cells, and APOBEC3F proteins and p24 CA were detected in the corresponding virion lysates using relevant primary antibodies, horseradish peroxidase-conjugated secondary antibodies, and enhanced chemiluminescence.
HIV-1 DNA Sequence Analyses-35-mm subconfluent monolayers of 293T cells were infected with pseudotyped viral stocks containing 4 ng of p24 CA and maintained for 24 h. Total cellular DNA was purified using the DNeasy kit (Qiagen), treated with DpnI for 2 h at 37°C to digest any residual plasmid DNA, and subjected to high fidelity PCR (Advantage-HF 2 polymerase; BD Biosciences) using HIV-1-specific oligonucleotides: EcoRI.nef.s, 5Ј-CCGAATTCAGGCAGCTGTAGATCT-TAGCCACTT, and BamH1.U5.a, 5Ј-CAGGATCCGGTCTG-AGGGATCTCTAGTTAC. PCR products were gel-purified and subcloned into pBluescript using EcoRI and BamHI restriction sites. The resulting cloned fragments encoded a 650-bp nef-U5 region and were sequenced on an ABI 3730 sequencer according to the manufacturer's instructions.
Quantitative PCR-Viral stocks from transfected 293T cells were treated with 20 l/ml RQ1-DNase (Promega) for 2 h prior to use. Stocks corresponding to 150 ng of p24 CA were used to challenge 5 ϫ 10 6 SupT1 cells (a T-lymphoid line) on ice, followed by gentle rotation at 4°C for 2 h to allow virus binding. The cells were then washed in cold phosphate-buffered saline, resuspended in cold RPMI, and incubated at 37°C to initiate "synchronous" virus entry. One-fifth of each culture was removed at 0, 2, 4, 6, and 24 h, and total DNA was isolated using the DNeasy kit and resuspended in a final volume of 200 l. All of the DNA was treated with DpnI for 2 h, and 2 l was used for standard fluorescence-monitored real time PCR analysis. Primer/probe combinations specific for different reverse transcription intermediates were used: early, strong stop products (proviral coordinates 500 -635) were amplified using oHC64 (5Ј-TAACTAGGGAACCCACTGC) and oHC65 (5Ј-GCTA-GAGATTTTCCACACTG) and detected using oHC66 (5Ј-FAM-ACACAACAGACGGGCACACACTA-TAMRA); second strand transfer products (proviral coordinates 500 -695) were amplified using oHC64 and gagM661as (5Ј-CTGCGTCGAG-AGAGCTCCTCTGGTT) and detected using oHC66. The reactions were performed in triplicate, in TaqMan Universal PCR master mix (UNG-less) using 0.9 pmol/l of each primer and 0.25 pmol/l probe. After 10 min at 95°C, the reactions were cycled through 15 s at 95°C followed by 1 min at 60°C for 40 repeats, carried out on an ABI prism model 7900HT machine (Applied Biosystems). The ⌬vif HIV expression vector pIIIB/⌬vif was diluted into purified SupT1 cellular DNA to create standards that were used to calculate cDNA copy numbers and to confirm the linearity of the assays.

RESULTS
To begin to address whether APOBEC3F, like APOBEC3G, is able to exhibit antiviral activity in the absence of cytidine deamination, we generated a panel of APOBEC3F constructs expressing mutant proteins in which highly conserved amino acids had been replaced.
Expression of APOBEC3F Mutant Proteins-The APOBEC3F protein possesses two putative cytidine deaminase motifs with the conserved amino acid sequence His-Ala-Glu-Xaa 27 -Pro-Cys-Pro-Xaa-Cys. Four residues are crucial for the catalytic activity of the related APOBEC1 and APOBEC3G enzymes (31,45); a histidine and two cysteines co-ordinate a Zn 2ϩ ion, and a glutamic acid is involved in proton shuttling during the deamination process (46). We therefore used site-directed mutagenesis to introduce point mutations into the APOBEC3F cDNA and generated constructs coding for mutant proteins with amino acid changes in the N-terminal, C-terminal (single mutants), or both cytidine deaminase motifs (double mutants) (Fig. 1). The histidine was substituted with cysteine (predicted not to disrupt Zn 2ϩ coordination) or arginine, the glutamic acid was exchanged for glutamine, and the two cysteines were exchanged for serines.

APOBEC3F Proteins with Single Mutations in Either Cytidine
Deaminase Motif Retain Substantial Antiviral Activity-Wildtype and mutant APOBEC3F constructs were tested in a singleround infectivity assay to examine the effect of these point mutations on the antiviral activity of this protein (Fig. 2). Plasmids encoding wild-type or mutant APOBEC3F and HIV-1/ ⌬vif were co-transfected into 293T cells, virus supernatants were harvested and normalized, and TZM-␤-gal indicator cells were challenged (43). The wild-type APOBEC3F protein inhibited virus infectivity to ϳ1% compared with the vector-alone control. This particular APOBEC3F protein differs from the one we have used previously by four amino acids (3) and has ϳ3-fold greater antiviral activity (data not shown). All of the proteins carrying altered amino acids in one of the cytidine deaminase motifs retained substantial amounts of antiviral activity, with infectivity ranging from 1-2% (for H65C) to ϳ8% (for E67Q) and with no distinctive trend for N-terminal mutants (Fig. 2, white bars) versus C-terminal mutants (Fig. 2, gray bars). Aside from the H65C/H249C mutant, the doubly mutated proteins were essentially inactive (Fig. 2, hatched bars), presumably because these proteins are not packaged into virions (Fig. 3). The antiviral activity of APOBEC3G was also measured and was noted to be at least 10 times greater than for APOBEC3F, inhibiting infectivity to less than 0.05% compared with 0.5-1.7% for APOBEC3F (also refer to Fig. 6). All of these mutant proteins were tested for activity against wild-type virus and were found to have minimal effect on infectivity, indicating that resistance to inhibition by Vif had not been acquired (data not shown). APOBEC3F has two cytidine deaminase motifs based on sequence alignment with APOBEC-1, each containing conserved histidine, cysteine, and glutamic acid residues. Point mutations were introduced into the DNA at the sites coding for these residues in both the N-terminal and C-terminal domains of APOBEC3F. The letters in bold type indicate the altered residues as compared with the consensus sequence shown. The Role of Cytidine Deaminase Motifs in APOBEC3F Packaging-To assess the importance of the conserved residues in APOBEC3F for virion incorporation, ⌬vif viruses were produced in 293T cells in the presence of wild-type or mutant APOBEC3F proteins and purified through a 20% sucrose cushion. Virions and corresponding whole cell lysates were then subjected to immunoblot analysis (Fig. 3). The abundant cellular protein 14-3-3␥ was present in all cell lysates (loading control; Fig. 3, top panel) and, consistent with earlier work (47), was not detected in any viral lysates (data not shown), thus establishing the specificity of this packaging assay. The wild-type and mutant APOBEC3F proteins were expressed at similar levels in cells (Fig. 3, second panel); however, although the wild-type protein was efficiently packaged, some of the mutants were not (Fig. 3, third panel). First, none of the double mutant proteins were discernibly incorporated into virions aside from H65C/ H249C (Fig. 3, lanes 8, 11, 14, and 17 compared with lane 5). As previously noted, this histidine-to-cysteine change should not prevent Zn 2ϩ co-ordination, suggesting that this attribute may contribute to APOBEC3F packaging, possibly through helping to bind RNA. The inability of the double mutants to be packaged into virions corresponds well with the lack of antiviral activity for any of these proteins. Second, proteins with muta-tions in the N-terminal motif were packaged less efficiently than their C-terminally mutated counterparts (lanes 3, 6, 9, and  12 compared with lanes 4, 7, 10, and 13), with the exception of the C99S and C283S proteins, neither of which were packaged well (lanes 15 and 16). We therefore conclude that the N-terminal cytidine deaminase motif plays a more prominent role in APOBEC3F packaging than the C-terminal motif but that the latter does contribute to some extent. Finally, although the antiviral phenotypes of wild-type and mutated proteins are all titratable by themselves (supplemental Figs. S1 and S2), some poorly packaged proteins (e.g. E67Q) still exhibit significant antiviral effects. Although we do not yet fully understand such phenomena, one possibility is that these mutated proteins may have acquired attributes that enhance antiviral efficacy.
The C-terminal Cytidine Deaminase Motif of APOBEC3F Mediates Catalytic Activity-Single mutations in either the Cor N-terminal cytidine deaminase motifs of APOBEC3F did not abolish the antiviral activity of this protein (Fig. 2). We therefore wanted to investigate whether these mutations disrupted DNA editing function. HIV-1/⌬vif stocks virions were produced in the presence of wild-type or mutant APOBEC3F proteins and then used to infect 293T cells, and total cellular DNA was purified. High fidelity PCR was performed to amplify a 650-bp fragment running from within the nef gene into the U5 region of the 3Ј-LTR (3) These fragments were then inserted into pBluescript, and multiple clones were sequenced (Table 1). We have shown previously that, in addition to inducing G-to-A transitions in viral positive strands, APOBEC3F also induces a small, yet significant, number of C-to-T transitions (3), most likely because of the deamination of cytidines in virion RNA; mutation rates are therefore calculated for both G-to-A and C-to-T transitions. The mutation rates for wild-type APOBEC3F and the proteins with an altered N-terminal cytidine deaminase motif were very similar, whereas disrupting the C-terminal cytidine deaminase motif resulted in rates that were equivalent to the vector-alone control. These results were echoed when APOBEC3F proteins with mutations in the C-terminal motif were also found to be inactive in a previously described bacterial editing assay (data not shown) (16). Although it seems improbable, it should be recognized that APOBEC proteins that do not display measurable editing activity in these experiments could perhaps introduce mutations into regions of HIV-1 not sequenced here or into the tRNA Lys3 primer used for first strand DNA synthesis. Accordingly, we also sequenced a region of the env gene from reverse transcripts

. Expression and virion incorporation of APOBEC3F proteins.
Immunoblot analysis of APOBEC3F wild-type (WT) and mutant proteins was performed on both 293T cells transfected with APOBEC3F expression plasmids and concentrated virus produced from these cells. Equivalent amounts of protein were loaded, as judged by amounts of the cellular protein 14-3-3␥ (top panel) and the viral protein p24 CA (fourth panel). Band intensities of incorporated APOBEC3F proteins were measured first as relative to the p24 CA loading control, and then the numbers shown are expressed as a proportion of the strongest band present on the gel (set at 100).

Viral infectivity
No. clones sequenced

Total bp sequenced
Total no. mutations Average no. mutations (proviral coordinates 6881-7592) and found that the mutation rate for wild-type APOBEC3F was slightly higher (ϳ1 mutation/100 bp) but that the proteins with mutations in the C terminus were still unable to edit (data not shown). Therefore, and as we have previously illustrated with mutant APOBEC3G proteins (31), the antiviral action of APOBEC3F can be experimentally segregated from its ability to induce G-to-A hypermutation in nascent reverse transcripts. These data are also consistent with previous work that concluded that the C-terminal motif of APOBEC3F is responsible for editing activity (39,40). Indeed, in human APOBEC proteins with two deaminase motifs, it appears that it is the C-terminal motif that harbors catalytic function (31,39,40,48).

Wild-type and Mutant APOBEC3F Proteins Prevent the Accumulation of HIV-1/⌬vif Reverse Transcription Products-Previous
reports have shown that the accumulation of HIV-1/⌬vif reverse transcription products is lower in the presence of APOBEC3G (7, 39, 49 -52). It has been speculated that the uridine produced by deamination is recognized by cellular DNA repair enzymes, leading to degradation of the cDNAs (5, 7, 53). However, there has been conflicting experimental evidence addressing this issue (52,54), and it is alternatively possible that APOBEC proteins may act by causing aberrant degradation of cDNA or by interfering directly with cDNA synthesis. To help address these points, we used quantitative PCR to measure the accumulation of early reverse transcription products over a 24-h time course (Fig. 4). Viruses were produced in the presence of control vector, wild-type APOBEC3G, wild-type APOBEC3F, or mutant APOBEC3F proteins and then used to challenge SupT1 cells. DNA was harvested and purified after 0, 2, 4, 6, and 24 h and subjected to quantitative PCR analysis using primers that amplify strong stop (early) reverse transcripts. The quantity of reverse transcripts rose, peaked at ϳ6 h, and then declined for most infections, as is typical for retroviral infection (55,56). Accumulation of these products was inhibited almost entirely by wild-type APOBEC3G and to ϳ20% by wild-type APOBEC3F, a finding that reflects the less potent antiviral phenotype of this protein. Interestingly, the H65R, H249R, E67Q, and E251Q mutant APOBEC3F proteins all inhibited accumulation of the early reverse transcription products by roughly the same extent as the wild-type protein. A similar experiment was performed using lower amounts of wild-type or mutant APOBEC3F proteins, and again, antiviral activity corresponded well with reductions in reverse transcript accumulation (supplemental Fig. S1).
Wild-type APOBEC3G Inhibits Reverse Transcript Accumulation to a Greater Extent than Editing-deficient Mutants-Having found that APOBEC3F mutants deficient for cytidine deaminase activity inhibited HIV-1 cDNA accumulation to the same extent as the wild-type parental protein, we wanted to determine whether the same would hold true for analogous APOBEC3G mutant proteins (31). We therefore repeated the quantitative PCR analyses of reverse transcript levels using wild-type APOBEC3G and its nonediting derivatives H257R and E259Q (Fig. 5; wild-type and E251Q versions of APOBEC3F were included for comparison). Analysis of early (strong stop) cDNAs revealed a different pattern for APOBEC3G proteins than for APOBEC3F proteins (Fig. 5A). In particular, whereas wild-type APOBEC3G imposed a profound decrease in levels (dark blue), the presence of either of the editing deficient mutants (purple and light blue) resulted in modest decreases in levels to about half of those seen in the vectoralone infection (black line). Although the rank order of inhibitory phenotypes was maintained when a later replication intermediate (second strand transfer cDNA) was analyzed (Fig. 5, B  and C), the magnitudes of inhibition of viral DNA accumulation were substantially greater. For instance, at the 8-h time point, the E259Q mutant of APOBEC3G reduced the level of strong stop cDNA to ϳ30% and reduced second strand transfer to ϳ5% of the amount accumulated in the absence of any APO-BEC protein. It is therefore evident that although nonediting APOBEC3F mutant proteins are able to inhibit reverse transcript accumulation to the same extent as the wild-type protein FIGURE 4. Effect of APOBEC3F proteins on HIV-1 reverse transcript accumulation. A, quantitative PCR analysis was used to measure the amount of HIV-1/⌬vif early reverse transcription products (strong stop) made in the presence of wild-type (WT) or mutant APOBEC3G/F proteins. Equivalent amounts of ⌬vif virions produced from 293T cells expressing wildtype APOBEC3F (red), APOBEC3F H65R (pink), APOBEC3F H249R (light green), APOBEC3F E67Q (orange), APOBEC3F E251Q (dark green), wild-type APOBEC3G (dark blue), or no APOBEC (black) were added to SupT1 cells, and total DNA was harvested at the indicated times after infection. The relative levels of HIV-1 early reverse transcription products compared with standard samples are shown. The line representing APOBEC3F H65R (pink) falls below the line representing APOBEC3F E67Q (orange) and is therefore not visible. B, effects of APOBEC proteins on HIV-1/⌬vif infectivity. The same viral preparations described above were used in parallel to challenge TZM-␤-gal indicator cells, and productive infection was measured as the induction of ␤-galactosidase activity. Infectivity is measured in counts, and the y axis is plotted on a log scale. The colors are as for A. (Fig. 4), these nonediting APOBEC3G proteins show a clear reduction in this attribute. These differences in phenotype cannot be attributed to variations in packaging efficiency because the incorporation of the editing-deficient APOBEC proteins closely matches that of the wild-type proteins (Fig. 3 and data not shown) (31).

Inactivation of the C-terminal Cytidine Deaminase Motif Has a More Profound Effect on the Antiviral Function of APOBEC3G than It
Does on APOBEC3F-The aforementioned differences in the behavior of APOBEC3F and APOBEC3G mutant proteins prompted us to examine their antiviral phenotypes in further detail. Accordingly, stocks of HIV-1/⌬vif were produced in 293T cells with different doses of wild-type or mutant versions of APOBEC3G/F (viral DNA: APO-BEC DNA ratios of 1:1, 1:0.33, 1:0.067, and 1:0.0067; all with constant DNA levels) and used in single-cycle challenges of TZM-␤-gal cells ( Fig. 6; note the logarithmic scale of the y axes). Several points were evident when examining these titration data. First, as would be expected, the antiviral activity of all APOBEC3 proteins decreased at lower plasmid DNA concentrations, although wild-type APOBEC3G retained strong antiviral function even at the 1:0.067 ratio. Second, wild-type APOBEC3F (red line) has only slightly greater antiviral activity over its nonediting mutant counterparts (Fig. 6A). Third, in striking contrast, wild-type APOBEC3G (dark blue line) is a far more powerful antiviral factor than the mutant proteins, even though each mutant APOBEC3G protein inhibited infection by more than 80% at the highest dose (Fig. 6B). In fact, the antiviral activity of wild-type APOBEC3G was at least 10-fold greater than the mutant proteins at the three higher APOBEC cDNA concentrations tested.
In sum, the loss of editing function through mutation of the C-terminal deaminase motif disrupts the antiviral function of APOBEC3G to a much greater extent than for APOBEC3F. Furthermore, mutant APOBEC3G proteins that are unable to edit are also less able to suppress reverse transcript accumulation. This contrasts with APOBEC3F where the loss of cytidine deaminase activity has negligible effect on reverse transcript accumulation and considerably less impact upon antiviral activity, thus highlighting important differences between these two proteins. A, quantitative PCR analysis was used to measure the amount of HIV-1/⌬vif early reverse transcription products (strong stop) made in the presence of wild-type (WT) or mutant APOBEC3G/F proteins. Equivalent amounts of ⌬vif virions produced from 293T cells expressing wild-type APOBEC3F (red), APOBEC3F E251Q (green), APOBEC3G H257R (purple), APOBEC3G H259Q (light blue), wild-type APOBEC3G (dark blue), or no APO-BEC (black) were added to SupT1 cells, and total DNA was harvested at the indicated times after infection. The relative levels of HIV-1 early reverse transcription products compared with standard samples are shown. B, quantitative PCR analysis was performed as above but with primers designed to amplify later reverse transcription products (second strand transfer). The colors are as above. C, enlargement of the graph shown in B, displaying the reverse transcripts present at five copies and below. This is to allow a difference to be seen between samples when the copy numbers are very low. The same standards were used for all analyses, thereby allowing the efficiency of reverse transcription progression to be evaluated.

DISCUSSION
Previous work has demonstrated that mutant APOBEC3G proteins that cannot function as cytidine deaminases still exhibit antiviral activity (31). Because APOBEC3F is closely related to APOBEC3G and exhibits a similar antiviral phenotype (3, 6, 10, 13), we were eager to investigate whether this protein could also inhibit HIV-1 infection in the absence of hypermutation. Here, we show that APOBEC3F mutant proteins that are catalytically inert display antiviral phenotypes that are very close to that of the parental protein in terms of magnitude (Figs. 2 and 6). These results therefore provide further evidence that APOBEC proteins are able to function as antiviral factors by a mechanism that is distinct from the induction of hypermutation. These findings also further confirm that it is the C-terminal cytidine deaminase motif of the doublemotif human APOBEC proteins (APOBEC3G, F, and B) that mediates deamination (Table 1) (39,40,48). Why the N-terminal motifs are inactive remains unresolved; it is possible that these domains are intrinsically noncatalytic or that they adopt conformations that prevent appropriate access to nucleic acid substrates.
Additionally, we have shown that both cytidine deaminase motifs of APOBEC3F contribute to virion incorporation (Fig.  3). Specifically, our data with single domain mutants of APOBEC3F indicate that the N-terminal motif plays a principal role in packaging, perhaps because of RNA binding via the Zn 2ϩ coordinating elements, and that the incorporation of proteins with mutations in both motifs is negligible. Likewise, it has been demonstrated that the N-terminal motif of APOBEC3G, and in particular two phenylalanine residues, is important for the packaging of this protein (57).
Suppression of HIV-1 infectivity by APOBEC3F is reflected in a lack of accumulation of viral reverse transcripts (Figs. 4 and 5 and supplemental Fig. S1 for titration data). Other groups have reported lower accumulations of HIV-1 cDNAs in the presence of APOBEC3G (7, 39, 49 -52) or APOBEC3F (39), and it has been proposed that the recognition of edited sequences by the host DNA repair pathways may induce their degradation (5,7,53,54). However, we clearly observe lower accumulations of viral cDNA products in the presence of APOBEC3F proteins that are catalytically inactive (Figs. 4 and 5 and supplemental Fig. S1); in fact, the decreases in levels closely match those seen with wild-type APOBEC3F. Accordingly, we propose that the inability to produce sustained levels of these cDNAs constitutes a major mode of antiviral action for APOBEC proteins, irrespective of whether editing has or can occur. Importantly, this conclusion is consistent with recent data demonstrating that APOBEC-mediated antiviral effects are not dependent on the recognition of edited reverse transcripts by the major cellular uracil DNA glycosidase, UNG2 (52). Nevertheless, when the (C-terminal) cytidine deaminase active site is operational, then editing of viral cDNAs will occur and, through the mutational burden imposed, has the potential to amplify the inhibition of HIV-1 infection and replication. Interestingly, we also have experimental evidence that cytidine deamination is not sufficient to confer antiviral activity. Specifically, studies in our lab using APOBEC3G/F chimeric proteins have shown that it is possible to induce relatively high numbers of mutations in nascent HIV-1 cDNAs, with little or no accompanying antiviral activity as judged in single-cycle challenges of TZM-␤-gal cells (39). Rather, and as seen here, infectivity correlated closely with the levels of reverse transcripts measured in target cells.
We also investigated the accumulation of viral cDNAs for viruses produced in the presence of wild-type or editing-deficient APOBEC3G proteins (Fig. 5). Although the wild-type APOBEC3G protein exhibited a dramatic inhibition of reverse transcript accumulation, the mutant proteins showed a relatively modest inhibition of early (strong stop) cDNAs (Fig. 5A). One interpretation of this result is that the editing activity of APOBEC3G contributes to its ability to inhibit the accumulation of reverse transcripts. However, the observations that editing-deficient APOBEC3F mutants inhibit cDNA accumulation as well as the wild-type protein (Fig. 4) and also that UNG2 inhibition fails to reverse APOBEC-mediated viral inhibition (52) each tend to suggest otherwise (although potential contributions of UNG2-independent repair pathways must not be disregarded). Additionally, it cannot be ruled out that editing may participate in the inhibition of reverse transcript accumulation through a nondegradative mechanism via the deamina- . Titration experiments with wild-type and editing-deficient APOBEC3F and APOBEC3G proteins. 293T cells were transfected with decreasing amounts of wild-type or mutant APOBEC3G/F constructs. The amount of ⌬vif proviral DNA was kept constant, whereas the amount of APO-BEC DNA was added at the following ratios; 1:1, 1:0.33, 1:0.067, and 1:0.0067. Total DNA was made up to 3 g using empty vector. The viruses were produced, normalized, and used to challenge TZM-␤-gal indicator cells. Productive infection was measured as the induction of ␤-galactosidase activity. At the lowest DNA ratio (1:0.0067), infectivity (as measured by counts) was equivalent to viruses produced in the absence of any APOBEC (data not shown).
tion of viral or alternative substrates. Importantly, the rank order of inhibition of infection by these APOBEC proteins closely resembles their ability to inhibit the accumulation of reverse transcripts.
Finally, we were interested in comparing the overall antiviral activities of wild-type and editing-deficient versions of APOBEC3G/F in greater detail in titration experiments (Fig. 6). Whereas the nonediting APOBEC3F proteins show only moderately diminished antiviral activity relative to the wild-type protein (Fig. 6A), the analogous APOBEC3G mutants display a far greater loss in activity relative to the wild-type protein at decreasing DNA concentrations (Fig. 6B). Our favored interpretation of this dichotomy in behavior between the editingdeficient APOBEC proteins and their respective wild-type proteins is that cytidine deamination of HIV-1 reverse transcripts plays a more significant role in the overall antiviral effect of APOBEC3G than it does for APOBEC3F. However, as discussed earlier, it cannot be ruled out that the introduction of mutations into the C-terminal deaminase motif of APOBEC3G may disrupt other aspects of APOBEC protein function, potentially to a greater extent than do analogous mutations in APOBEC3F. The less severe effects of such APOBEC3G mutants on viral cDNA accumulation (relative to wild-type APOBEC3G) may be a consequence of such an effect (Fig. 5).
In sum, we have shown that APOBEC3F proteins that have lost the ability to hypermutate HIV-1 cDNAs retain strong antiviral phenotypes that are quite close in magnitude to that of the wild-type protein. Moreover, these mutant proteins are able to inhibit the accumulation of viral reverse transcripts as effectively as the parental protein. These results therefore reinforce the view that the inhibition of HIV-1 infection by APOBEC proteins is not mediated solely by cytidine deamination (31,58) and that the suppression of viral cDNA accumulation is an important aspect of the antiviral phenotype (39). Given the relatively similar behavior of mutated and wild-type APOBEC3F proteins in different assays, we suggest that this protein is a better system (than APOBEC3G) with which to study the nonediting activities and attributes of APOBEC proteins.