APOBEC3B and APOBEC3C Are Potent Inhibitors of Simian Immunodeficiency Virus Replication*

In the human genome the apolipoprotein B mRNA-editing enzyme catalytic polypeptide (APOBEC)3 gene has expanded into a tandem array of genes termed APOBEC3A-G. Two members of this family, APOBEC3G and APOBEC3F, have been found to have potent activity against virion infectivity factor deficient (Δvif) human immunodeficiency virus 1 (HIV-1). These enzymes become encapsidated in Δvif HIV-1 virions and in the next round of infection deaminate the newly synthesized reverse transcripts. The lentiviral Vif protein prevents the deamination by inducing the degradation of APOBEC3G and APOBEC3F. We report here that two additional APOBEC3 family members, APOBEC3B and APOBEC3C, have potent antiviral activity against simian immuno-deficiency virus (SIV), but not HIV-1. Both enzymes were encapsidated in HIV-1 and SIV virions and were active against Δvif SIVmac and SIVagm. SIV Vif neutralized the antiviral activity of APOBEC3C, but not that of APOBEC3B. APOBEC3B induced abundant G → A mutations in both wild-type and Δvif SIV reverse transcripts. APOBEC3C induced substantially fewer mutations. APOBEC3F was found to be active against SIV and sensitive to SIVmac Vif. These findings raise the possibility that the different APOBEC3 family members function to neutralize specific lentiviruses.

The interaction of Vif with APOBEC3G is species-specific (4,15,16). HIV-1 Vif binds to human APOBEC3G but not mouse, African green monkey (AGM), or rhesus macaque (MAC) APOBEC3G (4). As a result, noncognate APOBEC3G proteins tend to block the replication of both ⌬vif and wild-type HIV-1. Conversely, Vif of SIV mac , but not SIV agm , binds to human APOBEC3G; therefore, human APOBEC3G inhibits wild-type SIV agm but not wild-type SIV mac (4). The species specificity of the Vif-APOBEC3G interaction is determined by a single amino acid difference in APOBEC3G at position 128 that encodes a charged amino acid (16 -19). The encapsidation of APOBEC3G into retroviral virions is less virus-specific. APOBEC3G can be encapsidated by HIV, SIV, and murine leukemia virus (MLV) through a mechanism thought to involve binding to viral RNA, cellular RNA, or the viral nucleocapsid protein (20 -23). APOBEC3G can also be encapsidated in hepatitis B virus and is thought to inhibit replication without deamination of the viral DNA (24).
In the human genome, the APOBEC gene family consists of APOBEC1, APOBEC2, activation-induced deaminase (AID), and APOBEC3A-G (here referred to as APOBEC3 in general) (25). The mouse genome contains only a single APOBEC3 gene, suggesting that APOBEC3A-G is a fairly recent evolutionary expansion (26). Each APOBEC protein contains a catalytic domain characterized by a conserved amino acid motif consisting of a His and two Cys residues that coordinate a Zn 2ϩ and a Glu residue that acts as a proton shuttle during catalysis. In APOBEC3B, APOBEC3F, and APOBEC3G the structural unit has been duplicated such that the enzyme contains two potential catalytic domains (25). In addition, the APOBEC proteins are generally dimeric.
APOBEC3G was the first cytidine deaminase shown to have anti-lentiviral activity (2). Recently, its most closely related family member, APOBEC3F, was also shown to be active against HIV-1 and a target of Vif (8,9,27,28). APOBEC3F is expressed in lymphocytes where it is able to form heteromul-timers with APOBEC3G (9,27). The two enzymes differ in target sequence preference, in that APOBEC3F prefers TC, whereas APOBEC3G targets CC (9,27,28). Together such sites appear to be the major contributors to HIV-1 hypermutation in vivo.
Here, we characterized APOBEC3B and APOBEC3C for their activity against HIV-1, SIV mac , and SIV agm . Although they were hardly active against HIV-1, APOBEC3B and APOBEC3C were potent inhibitors of SIV. Both enzymes were efficiently encapsidated by HIV-1 and SIV. These findings raise the possibility that the various APOBEC3 family members protect against different lentiviruses and provide a rationale for the expansion of the gene family in primates.

EXPERIMENTAL PROCEDURES
APOBEC Expression Vectors-APOBEC3B expression vector was derived by cloning the full-length cDNAs into the pcDNA3.1(ϩ) (Invitrogen). The cDNAs were generated by reverse transcriptase PCR using the RNA isolated from phytohemagglutinin-activated human peripheral blood mononuclear cells with RNeasy kit (Qiagen). The cDNA was primed with oligo(dT) and extended with Superscript III reverse transcriptase (Invitrogen). The APOBEC3B sequence was amplified with primers that contained EcoR V and SalI restriction sites and encoded the 5Ј-influenza hemagglutinin (HA) tag, YPYDVPDYA. The amplicons were cleaved with EcoRV and SalI and ligated to similarly cleaved pcDNA3.1(ϩ). Expression vectors for human APOBEC3C, APOBEC2, and AID were constructed the same way as for APOBEC3B, except that BamHI and SalI sites were used for cloning of APOBEC3C and EcoRI and XhoI sites for APOBEC2 and AID. Expression vectors for HA-tagged human APOBEC3G and AGM APOBEC3G were constructed by similar methods as previously described (4). APOBEC3F expression vector was provided by Dr. M. Peterlin (8). All plasmids were confirmed by sequencing.
Cell Lines-The human embryonic kidney cell line 293T, human embryonic carcinoma cell line NCCIT, adherent human osteosarcoma cell line HOS, and HOS.CD4.X4 that expresses CD4 and CXCR4 were cultured in Dulbecco's modified Eagle's medium, 10% fetal bovine serum at 37°C with 5% CO 2 .
APOBEC Encapsidation-Virions were produced by transfection of 293T cells as for production of reporter viruses. Culture supernatant was harvested 2 days posttransfection, and the virions were pelleted by ultracentrifugation through 20% sucrose at 150,000 ϫ g for 1 h (Beckman). The pelleted virions were solubilized in 100 l of 1% Triton-containing buffer and normalized for p24/p27 concentration, which was measured by enzyme-linked immunosorbent assay. Cell lysates were prepared by lysing the cells in the same buffer and normalized for protein concentration. Lysates of virions (20 ng of p24/p27) and cells (20 g of protein) were analyzed on immunoblots probed with anti-HA monoclonal antibody (mAb) 16B12 (Covance) and developed with horseradish peroxidase-conjugated sheep anti-mouse immunoglobulin and ECL reagents (Amersham Biosciences). Equal loading of virions was confirmed on immunoblots probed with anti-p24/p27 mAb AG3.0 (National Institutes of Health AIDS Research and Reference Reagent Program).
Cytidine Deaminase Assay-This assay was modified from the deaminase activity assay for encapsidated APOBEC3G previously described (10). Briefly, virions were pelleted and solubilized in 100 l of virus lysis buffer containing 0.1% (v/v) Triton X-100 (10). Virus lysate containing 100 ng of p24 was mixed with 1 ϫ 10 5 cpm of 5Ј-end 32 P-labeled oligonucleotide containing the indicated target sequence in the deaminase buffer. After 5 h at 37°C, the reactions were heated to 90°C for 5 min and then incubated with uracil DNA glycosylase (UDG) (New England Biolab) in UDG buffer for another 30 min at 37°C. Subsequently, the reaction was brought to 0.15 M NaOH at 37°C for 30 min. The products were separated on a precast 15% TBE-urea PAGE (Invitrogen) and detected by autoradiography. A labeled marker oligonucleotide in which the target C was replaced with U was processed in parallel to indicate the position of the cleaved product.
Real-time PCR Quantitation of APOBEC3 mRNA in Primary Cell Populations-Peripheral blood mononuclear cells were separated from healthy donor blood by Ficoll (Amersham Pharmacia Biotech) centrifugation. To purify unactivated CD4 ϩ T cells, peripheral blood mononuclear cells were positively sorted on anti-CD4 Dynabeads and removed with Detachabead (Dynal Biotech) as previously described (32). The cells were further purified by negative selection on anti-CD8 and anti-CD14 (BD Biosciences) with goat anti-mouse IgG (Dynal Biotech)coated beads. The purified cells were 99.5% CD4 ϩ as determined by flow cytometry. Activated cells were prepared as previously described (33). Purified CD4 ϩ T cells were cultured in dishes coated with goat antimouse IgG (10 g/ml; CalTag Laboratories) followed by 3 g/ml anti-CD3 antibody OKT-3 (American Type Culture Collection) in medium supplemented with 1 g/ml soluble anti-CD28 antibody (BD Biosciences). The cells were cultured in RPMI containing 10% fetal calf serum and IL-2 (Chiron) prepared as previously described (33).
Monocytes were purified from peripheral blood mononuclear cells by positive selection on anti-CD14-coated magnetic beads (Miltenyi Biotech) using an AutoMacs (Miltenyi Biotech). The monocytes were Ͼ98% CD14 ϩ . Macrophages were generated by culturing the monocytes for 4 days in medium containing 50 ng/ml granulocyte-macrophage colonystimulating factor (R&D Systems).
Total RNA was extracted from 1 ϫ 10 6 cells using an RNeasy RNA isolation kit (Qiagen) and treated with DNase I. NCCIT cells, human heart, and small intestine total RNAs (Ambion) were used as controls. cDNA was generated from 100 ng of RNA using the Superscript III first-strand synthesis system (Invitrogen) primed with random hexamers. Specific cDNAs from 2 to 10 ng of total RNA were quantitated by Taqman quantitative PCR with specific primer/probe combinations on an ABI PRIS 7700 sequence detection system following the manufacturer's instruction. Primers and probes were designed using Primer Express software (Applied Biosystems) and were shown to be specific for a single APOBEC family member by testing against plasmid DNA. Standard curves were constructed using serially diluted plasmid. The data were normalized using 18 S rRNA primers and probe (Applied Biosystems), which was found to be reliable for comparison of resting and activated populations. Data are presented as the average of triplicates Ϯ S.D.

HIV-1 and SIV Encapsidate APOBEC3B and APOBEC3C-
To determine whether different APOBECs can be encapsidated by lentiviruses, virions were generated by cotransfection of 293T cells with wild-type or ⌬vif HIV-1 or SIV agm proviral clones and expression vectors for HA-tagged APOBECs. Viruscontaining culture supernatant was harvested 2 days posttransfection, the virions were pelleted by ultracentrifugation, normalized for p24/p27 by enzyme-linked immunosorbent assay, and then analyzed for APOBEC content by immunoblot analysis with anti-HA mAb. To determine the relative expression levels of the APOBEC proteins, lysates of the transfected cells were prepared and analyzed similarly on immunoblots (Fig. 1A).
The results showed that APOBEC3B and APOBEC3C were encapsidated into HIV-1 and SIV agm virions (Fig. 1B). Both enzymes were more efficiently encapsidated than APOBEC3G in HIV-1 and SIV. Controls with human and AGM APOBEC3G demonstrated the expected species specificity; HIV-1, but not SIV agm Vif, reduced the amount of human APOBEC3G that was encapsidated. Conversely, SIV agm , not HIV-1 Vif, reduced the AGM APOBEC3G encapsidation. This species specificity did not apply to APOBEC3B and APOBEC3C. HIV-1 and SIV agm Vif reduced the amount of encapsidated APOBEC3C, but not APOBEC3B (Fig. 1B). Two other members of the APO-BEC family, APOBEC2 and AID, were also efficiently incorporated into wild-type and ⌬vif HIV-1 virions (Fig. 1B). There was slightly more APOBEC3B and AID in the wild-type as compared with ⌬vif virus particles, a phenomenon previously noted for mouse and AGM APOBEC3G encapsidated into HIV-1 virions (4). HIV-1 Vif appeared to slightly reduce the amount of encapsidated APOBEC2.
APOBEC3B and APOBEC3C Inhibit SIV Infectivity-Although the different APOBECs could be encapsidated in HIV-1 and SIV, it was not clear whether they would have antiviral activity. To determine this, VSV-G pseudotyped wild-type and ⌬vif HIV-1, SIV agm , and SIV mac luciferase reporter viruses were prepared in 293T cells cotransfected with APOBEC expression vectors. The infectivity of the viruses was determined by infecting HOS cells with viruses normalized for p24/p27 and measuring luciferase activity 3 days postinfection (Fig. 2). The amount of APOBEC3G expression vector in the transfection was sufficient to reduce the infectivity of ⌬vif HIV-1 to close to background and to cause an ϳ2-fold reduction in wild-type virus infectivity. AGM APOBEC3G, which is resistant to HIV-1 Vif, potently inhibited both wild-type and ⌬vif HIV-1 ( Fig. 2A). APOBEC3B and APOBEC3C weakly inhibited HIV-1 (2-5-fold for APOBEC3B and Ͻ2-fold for APOBEC3C; Fig. 2A). The antiviral activity of APOBEC3C, but not APOBEC3B, was moderately relieved by HIV-1 Vif. In contrast, APOBEC3B and APOBEC3C were potent inhibitors of SIV agm and SIV mac (Fig.  2, B and C). APOBEC3B reduced SIV agm infectivity to near background. SIV Vif did not relieve APOBEC3B inhibition of the virus. APOBEC3C was potent against ⌬vif SIV agm , and in contrast to APOBEC3B, its inhibitory activity was relieved by SIV Vif. APOBEC3B and APOBEC3C activity on SIV mac was similar to that on SIV agm (Fig. 2, B and C). Human APOBEC2 and AID did not affect the infectivity of HIV-1, SIV agm , and SIV mac (Fig. 2, A-C).
Encapsidated APOBEC3B and APOBEC3C Deaminate the Reverse Transcripts-APOBEC3G and APOBEC3F mediate their antiviral activity by single-strand DNA deamination (9,10,27,28,34). To determine whether APOBEC3B and APOBEC3C also act through this mechanism, the sequence of viral reverse transcripts from newly infected cells was analyzed. VSV-G pseudotyped wild-type and ⌬vif HIV-1 and SIV agm with or without encapsidated APOBEC were prepared, normalized for p24, treated with DNase I, and then used to infect HOS cells. At 12 h postinfection, a fragment of the reverse transcript that contained the 3Ј-portion of env and 5Ј-portion of nef, a region that is highly susceptible to APOBEC3G deamination (10), was amplified and cloned. Nucleotide sequences were determined for at least 10 independent clones from each infection.
APOBEC3B and APOBEC3C generated a low frequency of G 3 A mutations on HIV-1 reverse transcripts (Fig. 3A). The number of mutations induced by APOBEC3C, but not APOBEC3B, was reduced by Vif. In contrast, APOBEC3B generated a high frequency of G 3 A mutations on SIV agm reverse transcripts ( Fig. 3A; quantitated in Fig. 3B). This mutational

FIG. 3. APOBEC3B and APOBEC3C induce G 3 A mutations in HIV-1 and SIV agm reverse transcripts.
A, a fragment in env-nef was amplified from HIV-1 and SIV agm reverse transcripts at 12 h postinfection. At least 10 independent HIV-1 and SIV agm nucleotide sequences were determined. The mutations in eight clones of each group are shown. Each mutation is denoted by a vertical line, color coded with respect to dinucleotide context: GG 3 AG (red), GA 3 AA (cyan), GC 3 AC (green), GT 3 AT (magenta), and non-G 3 A (black). B, the percentage of G nucleotides that were changed to A is plotted as the average of the sequenced clones from each SIV agm infection. C, the relative frequency of GG 3 AG, GA 3 AA, GC 3 AC, and GT 3 AT in SIV agm generated by APOBEC3B, APOBEC3C, and APOBEC3G is shown graphically as the percentage of G 3 A mutations in the indicated dinucleotide context. frequency was comparable with that induced by human APOBEC3G and was not affected by Vif. The pattern of bases changed by APOBEC3B and APOBEC3G was overlapping, with many shared hot spots, but APOBEC3B displayed a broader target sequence preference, targeting both GG and GA dinucleotides (Fig. 3C). APOBEC3C also induced G 3 A mutations on ⌬vif SIV agm reverse transcripts, but with a 9-fold lower frequency compared with APOBEC3B (Fig. 3B). These mutations were randomly distributed and lacked obvious hot spots or strong dinucleotide sequence preference (Fig. 3, A and  C). Vif reduced the number of mutations by APOBEC3C to near background (Fig. 3, A and B). Taken together, these findings suggest that APOBEC3B and APOBEC3C, as APOBEC3G, act by deaminating minus-strand reverse transcripts.
Deaminase Activity of APOBEC3B and APOBEC3C in Vitro-To investigate the basis for the selective inhibition of SIV, but not HIV-1 infectivity by APOBEC3B and APOBEC3C, the deaminase activity of the encapsidated enzymes was measured in an in vitro assay (10). APOBEC3-containing virions were produced in transfected 293T cells, pelleted, and solubilized. The virion lysate was then incubated with a 5Ј-endlabeled oligonucleotide containing the sequence CCCA, the favored target site for APOBEC3G (10,35). The deaminated oligonucleotide was subsequently cleaved at the deamination site by uracil DNA glycosylase and high pH treatment, and the cleaved product was visualized by autoradiography. Despite the much reduced antiviral activity of APOBEC3B against HIV-1 as compared with SIV agm , the cytidine deaminase activity of APOBEC3B released from HIV-1 and SIV agm virions was similar (Fig 4, A and B). Thus, the APOBEC3B encapsidated by HIV-1 remained catalytically active but the HIV-1 genome was largely resistant to it.
APOBEC3B was highly active on the target sequence CCCA (Fig. 4, A and B). In contrast, APOBEC3C was weakly active against CCCA. The target sequence preference of the APO-BECs was further analyzed by testing labeled oligonucleotide substrates that contained CCA, CCT, and CCG target sites (Fig. 4C). These are suboptimal targets for human and AGM APOBEC3G. As expected, human and AGM APOBEC3G were weakly active on these sequences in vitro (Fig. 4C). In contrast, APOBEC3B was highly active against each of these targets. APOBEC3C was poorly active on each substrate. These findings are consistent with the pattern of the mutations generated by APOBEC3B and APOBEC3C on the SIV reverse transcripts. The strand specificity of APOBEC3 was determined using single-stranded (ss) DNA, double-stranded (ds) DNA, and RNA-DNA hybrid as substrates in the assay. APOBEC3B was specific for ssDNA but not dsDNA or RNA-DNA hybrids, similar to APOBEC3G (Fig. 4D). APOBEC3C was also inactive against dsDNA and RNA-DNA hybrid.
MLV Was Relatively Resistant to APOBEC3B and APOBEC3C-We previously reported that APOBEC3G was efficiently encapsidated in MLV virions but had only a small effect on the infectivity of an MLV-EGFP reporter virus (4). To determine whether MLV might be sensitive to APOBEC3B or APOBEC3C, MLV-EGFP reporter viruses were prepared in 293T cells cotransfected with APOBEC3 expression vector. The virions were then harvested and tested for APOBEC3 encapsidation, infectivity, and deaminase activity. Human APOBEC3B, APOBEC3C, APOBEC3G, and AGM APOBEC3G were found to be encapsidated in MLV virions (Fig. 5A). Nonetheless, the APOBECs had only a small effect on infectivity of the virus (Fig. 5B). The encapsidated APOBEC3B was enzymatically active as shown in the in vitro deaminase assay (Fig.  5C). In particular, APOBEC3B-containing virions displayed more deaminase activity than APOBEC3G-containing MLV or HIV-1 particles. The activity of the MLV-encapsidated APOBEC3C and human/AGM APOBEC3G was relatively low in this assay.
APOBEC3F Is Active against SIV-In light of the virus specificity of the APOBEC3 family members, we tested APOBEC3F against SIV. As previously reported, APOBEC3F was active against HIV-1 but with a reduced sensitivity to Vif compared with APOBEC3G (Fig. 6A) (8,9,27,28). APOBEC3F was active against SIV agm , as was APOBEC3G, and both were resistant to SIV agm Vif (Fig. 6B). APOBEC3F was also active against ⌬vif SIV mac , but as with APOBEC3G it was sensitive to SIV mac Vif (Fig. 6C).
APOBEC3B and APOBEC3C Are Expressed in Lymphoid and Myeloid Cells-APOBEC3B and APOBEC3C would only be relevant to lentiviral replication if they were expressed in lymphoid or myeloid cells, the natural targets of the viruses. To measure expression of these genes in lymphocytes and monocytes, resting and activated human CD4 ϩ T cells and macrophages were prepared, and the expression of APOBEC3B, APOBEC3C, APOBEC3F, and APOBEC3G was analyzed by FIG. 4. Catalytic activity of encapsidated APOBEC3B and APOBEC3C. A, cytidine deaminase activity released from pelleted HIV-1 virions containing the indicated APOBEC3 was measured with 32 P-labeled deoxyoligonucleotide containing the target sequence CCCA. The deaminated product is indicated by an arrow. Virions prepared with no APOBEC were used as a control for background activity (lanes 1 and 2). B, cytidine deaminase activity in SIV agm virions. An oligonucleotide containing a dU in place of the target dC (CCUA) was used as a control to show the mobility of the cleaved product. C, the target site preference of the SIV agm -encapsidated APOBEC was tested on labeled deoxyoligonucleotides with the sequence CCA, CCT, or CCG. D, the substrate preference of the SIV agm -encapsidated APOBEC3B and APOBEC3C was tested on single-stranded DNA (ss), double-stranded DNA (ds), or RNA/DNA hybrid (R/D). Double-stranded substrates were formed by annealing the labeled CCCA deoxyoligonucleotide to unlabeled complementary DNA or RNA.
quantitative RT-PCR. The expression of these genes in human heart, small intestine, and the human embryonic carcinoma cell line NCCIT was also analyzed. APOBEC3B and APOBEC3C were expressed in heart and small intestine, whereas only APOBEC3B was expressed in NCCIT cells. APOBEC3F and APOBEC3G were expressed at low levels in the nonlymphoid cells tested. Each of these genes was expressed in activated CD4 ϩ T cells, and their expression was increased upon T cell activation (Fig. 7). These genes were also expressed in macrophages, whereas the level of APOBEC3F expression was relatively low. These results were consistent among all three donors. Thus, APOBEC3B and APOBEC3C are expressed in lymphoid cells where they could play a role in restricting retroviral replication. DISCUSSION We report here that two of the human APOBEC3 family members, APOBEC3B and APOBEC3C, have potent antiviral activity against two types of SIV, SIV mac and SIV agm . Interestingly, these deaminases were only marginally active against HIV-1 despite that they were efficiently encapsidated. APOBEC3F, which has been previously reported to inhibit HIV FIG. 6. APOBEC3F inhibits SIV infectivity. Wild-type and ⌬vif HIV-1, SIV agm , and SIV mac luciferase reporter viruses were produced in 293T cells in the presence or absence of APOBEC3G or APOBEC3F. Infectivity of the viruses was determined by quantitation of luciferase activity in HOS cells infected with equal amounts of viruses. (8,9,27,28), was also active against SIV. The specificity of the antiviral activity was not caused by differences in encapsidation, as APOBEC3B and APOBEC3C were encapsidated at relatively high levels in HIV-1, SIV, and MLV particles. Other APOBEC family members, such as AID and APOBEC2, were also encapsidated in HIV-1 and SIV but did not inhibit these viruses despite their single-strand DNA deaminase activity (8).
APOBEC3B was the most potent deaminase of the APOBEC family members that were tested. Like APOBEC3F and APOBEC3G, APOBEC3B has the duplicated, two-catalytic domain structure. In APOBEC3G it is the C-terminal catalytic domain that mediates deamination (36). The C-terminal catalytic domain of APOBEC3B is unique among APOBEC enzymes in having the two Zn 2ϩ coordinating cysteines separated by four amino acids instead of two. This spacing could result in a catalytic site that is less sequence-specific and more active. It is also interesting to note APOBEC3C is a single-unit-length enzyme and is efficiently encapsidated. Thus, it is possible for a single unit to contain sites for both encapsidation and enzymatic activity.
Whether APOBEC3B or APOBEC3C is active in vivo against HIV or SIV remains to be determined. However, it is clear that these enzymes were expressed in CD4ϩ T cells and macrophages, although their expression level appeared to be lower than that of APOBEC3G. Even though neither of the enzymes was strongly active against HIV-1, the resistance of APOBEC3B to Vif and its potent deaminase activity could allow it to contribute to the G 3 A hypermutation rate of the virus.
The virus specificity of the APOBEC3 antiviral activity is both surprising and difficult to explain. This effect was particularly pronounced for APOBEC3B, which is a potent deaminase that was efficiently encapsidated in HIV-1 yet caused only a few G 3 A changes in the reverse transcripts. In contrast, it generated a high frequency of mutations in SIV. During the preparation of this report, Bishop et al. (28) reported that APOBEC3B was active against HIV-1, but not MLV, unlike APOBEC3G that inhibited both viruses. We reported that both APOBEC3B and APOBEC3G were relatively ineffective against MLV regardless of their efficient encapsidation, reflecting again a specificity for particular retroviruses. Overall, SIV appeared to be most sensitive to APOBEC deamination among the viruses tested. SIV mac and SIV agm were susceptible to APOBEC3B, APOBEC3C, APOBEC3F, and APOBEC3G. HIV-1 was intermediate, being sensitive to APOBEC3F and APOBEC3G, whereas MLV was largely resistant to all of the deaminases tested. The APOBEC enzymes encapsidated in these virions were shown to be enzymatically active, arguing against a viral mechanism for silencing the enzyme. Furthermore, our findings suggest that the simple presence of an active deaminase in the virion is not sufficient to mediate viral reverse transcript deamination. AID, for example, is a highly FIG. 7. APOBEC3 expression in human primary CD4 ؉ T cells and macrophages. Relative mRNA levels of indicated APOBEC3 in human primary T cells, macrophages, heart, small intestine, and NCCIT cells were quantitated by real-time PCR and normalized by 18 S rRNA. The data for primary T cells and macrophages are representative of separate experiments using primary cells obtained from three healthy donors. active single-strand-specific cytidine deaminase and is efficiently encapsidated in HIV-1 but does not have detectable antiviral activity.
The molecular basis for the differences in virus susceptibility to the different APOBEC3 is a matter of speculation. It could reflect differences in the capsid structure of these viruses. It is possible that in MLV the nucleocapsid protein more tightly coats the viral genome, protecting it from the APOBEC enzyme, whereas in SIV and to some extent in HIV-1 the viral genome may be less protected. Alternatively, the viruses could differ in their mechanism of reverse transcription, such that encapsidated APOBEC3 is removed from the reverse transcription complex prior to the viral DNA being transiently exposed as single-stranded substrate. It is also possible that following virus entry the encapsidated deaminase dissociates more rapidly from the MLV reverse transcription complex during the process of virus uncoating.
Our findings raise the possibility that the various APOBEC3 family members function to restrict the replication of specific lentiviruses or of other viruses that go through a single-strand DNA intermediate. If this were the case, it would explain the fairly recent evolutionary expansion of the APOBEC3 family in primates, as the expanded family would more effectively protect against diverse retroviruses. In rodents, which are thought not to be subject to lentiviral infection, the selective pressure to expand APOBEC3 would have been absent. One possible scheme for the evolution of the gene family is that APOBEC3 originally served a physiological role in the cell, separate from its antiviral function. With the appearance of diverse lentiviruses, the gene may have expanded in primates to more effectively neutralize diverse lentiviruses. Diversification of the APOBEC3 family would make it more difficult for a single Vif to adapt to bind the different proteins. However, this antiviral mechanism was not entirely successful, judged by the wide distribution of primate and human lentiviruses.