A3G belongs to a family of cytidine deaminases that includes seven family members (A3A–H), all located in a gene cluster on chromosome 22 (Conticello et al., 2005; Jarmuz et al., 2002). Of note, mice contain only a single A3 gene located at a synthetic region on chromosome 15. Forty-five percent of the human genome consists of endogenous mobile genetic elements or retroelements, such as short interspersed nuclear elements or long interspersed nuclear elements, which pose a potential threat to genome integrity. It is thought that the human A3 family arose by gene duplication by unequal crossover due to strong evolutionary pressure to control these elements (Conticello et al., 2003; Sawyer et al., 2004). A3G and A3F are best known for their antiviral activity against HIV-1 (Chiu and Greene, 2008), although A3A, A3B, A3C, A3DE, and A3H can also defend against HIV-1 to some extent (discussed below).

The common characteristic of all these enzymes is the presence of one (A3A, A3C, A3H) or two cytidine deaminase domains (CDAs) (A3B, A3DE, A3F, A3G) (Conticello et al., 2003; Jarmuz et al., 2002). Although the two CDAs are highly similar, they have different functions (Jarmuz et al., 2002). In primates, the N-terminal domain mediates RNA binding, whereas the C-terminal domain mediates sequence specific cytidine deamination of single-stranded DNA (Friew et al., 2009; Gooch and Cullen, 2008; Hache et al., 2005; Iwatani et al., 2006; Li et al., 2004; Navarro et al., 2005).

CDAs are characterized by a conserved zinc-binding motif (C/H)-X-E-X_23~28-P-C-X_2~4-C (Conticello et al., 2005; Jarmuz et al., 2002). This enzymatically active site catalyzes the hydrolytic deamination at the C4 position of 2′-deoxycytidine, resulting in a 2′-deoxyuridine. During the deamination reaction, the cysteines coordinate a single zinc ion, while the key glutamate is involved in proton shuttling (Betts et al., 1994).

A3G and NC have a strong propensity to bind RNA, which may serve an adaptor for binding, as co-immunoprecipitation of NC and A3G is RNase sensitive (Svarovskaia et al., 2004). Indeed, RNA binding is likely indispensable for the packaging of A3G into virions (Luo et al., 2004; Navarro et al., 2005; Schafer et al., 2004; Svarovskaia et al., 2004; Zennou et al., 2004). Mutations in the N-terminus (amino acids 124–127) of A3G impair its RNA binding properties, thus greatly compromising its encapsidation into budding virions. Such mutants of A3G function as very poor inhibitors of HIV-1 infection (Burnett and Spearman, 2007; Huthoff et al., 2009; Huthoff and Malim, 2007; Navarro et al., 2005). Therefore, A3G takes advantage of the fundamental property of retroviral RNA packaging to be incorporated into the virus (Zennou et al., 2004).

G-to-A mutations occur throughout the viral genome but in an unequal distribution (Kijak et al., 2008; Koulinska et al., 2003; Suspene et al., 2006; Suspene et al., 2004; Yu et al., 2004a). Most hypermutations are present in the envelope (Env) and Nef regions and decrease toward the 5′ UTR. This gradient of hypermutation is explained in part by the nature of the reverse transcription reaction. For HIV, this reaction begins with the binding of the tRNA^Lys-3 to the primer-binding site of the viral RNA. The reverse transcriptase starts to synthesize minus-strand DNA, using the tRNA^Lys-3 as a primer. The minus-strand strong-stop DNA is then transferred to 3′ of the viral RNA. RNaseH degrades the plus-strand RNA and restores the enzymatic activity of A3G, enabling it to act on the single-strand minus DNA and deaminate dC to dU. Plus-strand synthesis is initiated from two RNaseH-resistant polypurine tracts, cPPT and 3′PPT, which remain associated with the minus-strand cDNA to serve as initiation sites.

The chance that a cytidine on the minus-strand will be deaminated correlates strongly with the time during which these DNA regions remain single stranded (Chelico et al., 2006; Suspene et al., 2006; Yu et al., 2004a). Therefore, the specificity of A3G for single-stranded DNA accounts for the highly polarized (5′→3′) mutational twin gradients within the viral genome, each containing the greatest number of mutations just 5′ to the cPPT and 3′PPT (Suspene et al., 2006). This deamination gradient could be further reinforced by a proposed model in which A3G, slides predominantly in the 3′→5′ direction on its minus strand DNA template (Chelico et al., 2006). This could also contribute to A3G’s preference for deaminating the 3′ C in the consensus sequence (5′-CCCA-3′) (Chelico et al., 2006). However, a more recent study suggests that A3G does not translocate on the DNA in a positionally correlated fashion by sliding or microscopic jumping but instead translocates in an uncorrelated fashion by macroscopic jumping or intersegmental transfer (Nowarski et al., 2008).

In rare cases, C→T mutations are detected in the U3 region of the 5′UTR, which reflects the deamination of a cytidine on the single plus strand. Of note, the primer binding site becomes single stranded through the action of RNaseH when it degrades the tRNA^Lys-3 bound at this site (Yu et al., 2004a).

A3G incorporated into virions resides in the viral core, bound in a ribonucleoprotein complex containing the viral genomic RNA as well as the viral NC, IN and Vpr proteins. A3G assembly with HIV RNA results in an inhibition of its intrinsic deaminase activity (Soros et al., 2007). How, then, is the virion-incorporated A3G ultimately activated so that it mediate mutation of the newly synthesized minus-strand DNA? It appears that the action of the viral RNase H during reverse transcription frees the enzyme from its RNA inhibitor, allowing full expression of its deaminase activity (Soros et al., 2007). These findings highlight a rather unexpected host-pathogen interaction in which the anti-HIV activity of A3G depends on its activation by an HIV enzyme.

Although A3G mediates hypermutation of the viral cDNA, it was also clear that wildtype A3G diminishes the accumulation of early and the late reverse transcription products in newly infected cells in a dose-dependent manner, with the greatest effect observed for late products (Anderson and Hope, 2008; Bishop et al., 2006; Guo et al., 2006; Holmes et al., 2007; Iwatani et al., 2007; Kaiser and Emerman, 2006; Luo et al., 2007; Mangeat et al., 2003; Mbisa et al., 2007). One possible explanation for this finding is that cellular host proteins degrade the hypermutated viral cDNA products. The resulting massive dU mutations could trigger the removal of the uracil in the nascent viral DNA by uracil-DNA-glycosylases (UNG2 and SMUG1), leading to abasic sites, and degradation by apurinic/apyrimidinic endonucleases (Schrofelbauer et al., 2005; Yang et al., 2007). However, more recent studies have shown that cells deficient in UNG2 (Kaiser and Emerman, 2006) or both UNG2 and SMUG1 (Langlois and Neuberger, 2008) do not rescue the accumulation of defective viral DNA.

Many laboratories have investigated how this deaminase-independent antiviral activity of A3 proteins reduces the amount of viral cDNA. Various, sometimes contradictory, reports have emerged, suggesting that A3G can interfere with primer tRNA annealing, minus- and plus-strand DNA transfer, primer tRNA progression and removal, and DNA elongation (Anderson and Hope, 2008; Bishop et al., 2006; Guo et al., 2006; Guo et al., 2007; Guo et al., 2009; Holmes et al., 2007; Iwatani et al., 2007; Li et al., 2007; Luo et al., 2007; Mangeat et al., 2003; Mariani et al., 2003; Mbisa et al., 2007).

Recently, studies employing a cell-free reverse transcriptase assay, lacking any nucleases, showed that the reduced HIV-1 cDNA levels are due to A3G-mediated inhibition of RT elongation. This finding suggests that A3G can physically impair reverse transcription, further excluding the possibility that increased cDNA degradation accounts for the lower levels of HIV-1 cDNA (Bishop et al., 2008) (Fig. 1, point 1).

Expression of A3G also inhibits viral DNA integration and provirus formation, probably by inducing defects in tRNA cleavage during plus-strand DNA transfer, leading to the formation of aberrant viral DNA ends (Fig. 1, point 3). These aberrant structures presumably interfere with the chromosomal integration of the double-stranded viral DNA required for provirus formation. Of note, cytidine deaminase activity was necessary to inhibit provirus formation, although the mechanism remains unclear (Mbisa et al., 2007). Interestingly A3G, A3F, and noncatalytic A3G mutants interact with components of the HIV-1 preintegration complex, namely integrase. This interaction might interfere with the structural integrity of this complex, resulting in diminished nuclear import and a lower integration rate (Luo et al., 2007).

In general, most of the cytidine deaminase–independent actions of A3 proteins are dose-dependent, with the greatest effects occurring at higher levels of expression, which may exceed physiological concentrations of enzyme (Holmes et al., 2007; Miyagi et al., 2007; Schumacher et al., 2008). However, in the aforementioned in vitro RT study where a hypermutation-independent decrease in viral cDNA synthesis was observed, wildtype A3G was used at levels analogous to that found endogenously (Bishop et al., 2008).

Comparison of noncatalytic mutants of A3G and A3F revealed that both proteins can function as antiviral proteins in the absence of cytidine deaminase activity, although an important portion of their antiviral activity depends on this enzymatic activity Overall, the loss of deaminase activity appears to have a more pronounced effect on the antiviral activity of A3G than A3F (Holmes et al., 2007).

A recent report identified four critical lysine residues (K297, K301, K303, and K334) in the C-terminal region of A3G that are required for Vif-mediated degradation (Iwatani et al., 2009). Mutation of these residues restored the antiviral function of A3G in the presence of Vif (Iwatani et al., 2009). However, another study showed evidence for the polyubiquitination and degradation of a lysine-free A3G in the presence of HIV-1 Vif (Shao et al, 2010). Other data suggest that the polyubiquitination of Vif is required for A3G degradation, rather than A3G polyubiquitination (Dang et al., 2008b). Therefore, it remains unresolved which residues in A3G, if any, are important for A3G polyubiquitination and proteasomal degradation.

HIV-1 Vif binds the E3 ubiquitin ligase complex through at least two interaction sites: it binds to the elongin C protein through its SOCS box, a S¹⁴⁴LQYLA¹⁴⁹ motif, (Mehle et al., 2004a; Yu et al., 2004b), and to cullin5 through a zinc-binding H¹⁰⁸x₅Cx_17-18Cx_3-5H¹³⁹ motif (Luo et al., 2005; Mehle et al., 2006) (Fig. 3B). Interestingly, the SOCS box of Vif is the most conserved region in different Vif proteins (Oberste and Gonda, 1992), indicating that this motif is crucial for the function of Vif. In addition, post-translational modifications may influence Vif’s function. Vif’s binding to elonginC can be negatively regulated by the phosphorylation of S144 in the SOCS box motif (Mehle et al., 2004a). An S144A mutation that prevents phosphorylation leads to the production of virions with poor infectivity, although this mutant effectively depletes A3G.

Vif also directly prevents the encapsidation of A3G, as shown by Vif’s ability to inhibit the packaging of the degradation-resistant A3G mutant C97A (Opi et al., 2007) and by the greater reduction of A3G levels in the virion than in the cell (Kao et al., 2007; Mariani et al., 2003; Schrofelbauer et al., 2004) (Fig. 2, point 3). These observations could indicate that Vif competes for the A3G binding site on viral genomic RNA or components of Gag, both of which are needed for A3G encapsidation (Kao et al., 2004; Mariani et al., 2003; Opi et al., 2007).

The interaction between A3G and Vif is critically dependent on amino acids 128–130 in A3G (Huthoff and Malim, 2007; Russell et al., 2009b) (Fig. 3B). Their importance is shown by the fact that the species specificity of HIV/SIV Vif is determined by D128 in A3G. More precisely, the African green monkey (agm) A3G protein contains a lysine (K) in lieu of aspartic acid (D) at position 128 and is not degraded by HIV-1 Vif. Substitution of D for K in agm A3G renders the enzyme sensitive to HIV-1 Vif, while mutation of D to K in human A3G makes it resistant to HIV-1 Vif but sensitive to SIV-1 Vif (Bogerd et al., 2004; Mangeat et al., 2004; Schrofelbauer et al., 2004; Xu et al., 2004). The corresponding sequence in Vif that likely interacts with the D128 region of A3G is D¹⁴RMR¹⁷. Substitution of Vif D¹⁴RMR¹⁷ with S¹⁴ERQ¹⁷ or S¹⁴EMQ¹⁷, the sequence found in agm Vif, allows the mutant HIV-1 Vif to degrade agm A3G but not human A3G. This species specificity was attributed to two positively charged amino acids (15 and 17) in Vif and their interaction with negatively charged D128 in human A3G (Schrofelbauer et al., 2006). Another region in Vif that is essential for the degradation of A3G is facilitated by amino acids 40–44 (Y⁴⁰RHHY⁴⁴) (Russell and Pathak, 2007; Yamashita et al., 2008). Of note, co-immunopreciptation experiments demonstrated that the Y⁴⁰RHHY⁴⁴ domain is actually the region necessary for the binding of Vif to A3G, while the D¹⁴RMR¹⁷ domain is more involved in a secondary step necessary for A3G degradation (Russell and Pathak, 2007). Indeed, Vif can bind the D128K A3G mutant but cannot facilitate its degradation (Russell and Pathak, 2007; Xu et al., 2004). Another region important for A3G binding to Vif is the amino acid stretch between 82 and 99 (Santa-Marta et al., 2005).

Other A3 family members besides A3G are also active against HIV-1, as will be discussed below. The most important is A3F, which is expressed in the natural target cells of HIV-1, but is produced at lower levels than A3G and is a less potent inhibitor of HIV-1 (Koning et al., 2009; Liddament et al., 2004; Simon et al., 2005; Xu et al., 2004; Zennou and Bieniasz, 2006). Vif can also degrade A3F, but with lower efficiency (Liddament et al., 2004; Zennou and Bieniasz, 2006).

Vif binds to A3F via the D¹⁴RMR¹⁷ domain, but not the Y⁴⁰RHHY⁴⁴ domain, which is essential for A3G degradation. Therefore Vif has two different binding sites—the Y⁴⁰RHHY⁴⁴ site, which is critical for A3G binding, and the D¹⁴RMR¹⁷ site, which is critical for A3F binding (Russell and Pathak, 2007). Interestingly, mutations in the A3G binding domain of Vif increase its sensitivity to A3F. Hence, it seems that Vif evolved to target the more potent A3G, even though Vif can reduce A3F with lower efficiency (Russell et al., 2009b) (Fig. 3A).

Other amino acids involved in the binding of Vif to A3F, but not to A3G, are W11, Q12, 74-76, and W79 (He et al., 2008; Russell and Pathak, 2007; Simon et al., 2005; Tian et al., 2006). In contrast, I9, K22, E45, Y40, and N48 are critical for Vif’s ability to degrade A3G, but not A3F (Russell and Pathak, 2007; Simon et al., 2005; Wichroski et al., 2005). The amino acids Y⁶⁹xxl⁷² (He et al., 2008; Pery et al., 2009; Yamashita et al., 2008) and the residues W5, W21, W38, N48, and W89 are needed to degrade A3G as well as A3F (Tian et al., 2006). Recently two groups identified W²¹KSLVK²⁶ as a novel motif in Vif critical for neutralizing A3G and A3F (Chen et al., 2009; Dang et al., 2009) with K22 and K26 playing especially important roles in this neutralization. However, these groups disagreed about the importance of S²³LV²⁵ residues for neutralization of A3F and A3G. One group identified SLV as important for both A3G and A3F neutralization (Dang et al., 2009), while the other found only L24 to be important for both A3G and A3F neutralization, and V25 important for the degradation of only A3F (Chen et al., 2009) (Fig. 3A). Recently two additional domains in Vif, L⁸¹GxGxSIEW⁸⁹ and E¹⁷¹DRW¹⁷⁴, were shown to be important for Vif neutralization of A3G and A3F. In the L⁸¹GxGxSIEW⁸⁹ domain, residues S⁸⁶IEW⁸⁹ are involved in Vif binding to A3G, A3F and Cul5; residue G84 is important for Vif binding to both A3G and A3F; and residues L81, G82 are involved in Vif binding to A3F. The E¹⁷¹DRW¹⁷⁴ domain is critical for Vif neutralization of A3F (Dang et al., 2010).

A3G and A3F also bind to different regions of Vif. A3G binds to the amino acid stretch 128–130 (Huthoff and Malim, 2007; Russell et al., 2009b), whereas A3F binds to amino acids 283–300 (Russell et al., 2009b).

A3B has a moderate activity against HIV-1. A3B is highly similar to A3G and A3F, and like A3G and A3F, it contains two CDAs. It can also bind to the nucleocapsid of HIV-1 Gag and therefore is packaged into the budding virus. Unlike A3G and A3F, A3B is resistant to Vif-induced degradation. However, A3B is expressed at extremely low levels in the natural target cells of HIV-1 and thus is unlikely to exert antiviral effects against HIV-1 under normal conditions (Bishop et al., 2004b; Bogerd et al., 2007; Doehle et al., 2005).

A3DE corresponds to another A3 protein with two CDA domains. It displays moderate activity against HIV-1, but Vif can facilitate its degradation through the 26S proteasome. In the absence of Vif, A3DE is encapsidated into virions and can deaminate CG on the minus-strand DNA, resulting in GC-to-AC mutations on the viral plus strand (Dang et al., 2006).

Vif also interacts with Pr55^gag (Simon et al., 1999). Vif was found in a complex together with Gag and the Gag-Pol precursor (Zimmerman et al., 2002). Therefore, Vif’s encapsidation into the virus might be mediated by an interaction with genomic RNA and partly by interaction with Gag and Gag-Pol (Bardy et al., 2001; Khan et al., 2001). In addition, a recent report suggests that virion encapsidation of Vif is dependent on Vif binding to A3G and A3F (Yamashita et al. 2010).

G-to-A hypermutations were first recognized in the env gene region during propagation of HIV-1 in vitro (Vartanian et al., 1991). Periodic G-to-A hypermutations are easily detectable in clinical samples from HIV-1 infected subjects and can account for >9% of the integrated viral DNA in cells (Gandhi et al., 2008; Kieffer et al., 2005; Land et al., 2008; Simon et al., 2005). Hypermutated virus could not be found in patient plasma, indicating that A3G/F function in vivo gives rise to hypermutated viral genomes that can be integrated but do not produce progeny viruses (Kieffer et al., 2005). Infection studies with ΔVif HIV-1 in CEM cells (which express A3G) and CEM-SS cells (which do not express A3G) show high levels of G-to-A mutations in viral DNA, as expected. However, significantly fewer mutations were detected in intracellular viral RNA, and even fewer in virion RNA (Russell et al., 2009a).

Genetic polymorphisms in A3G, A3F, or cullin5 and natural variations in their levels of expression could also affect HIV-1 pathogenesis. Indeed, in HIV-1-infected patients, A3G mRNA levels correlate inversely with HIV-1 viral loads and positively with CD4 cell counts (Jin et al., 2005), A3G mRNA levels are higher in long-term nonprogressors than in uninfected controls, and the lowest A3G mRNA levels are found in HIV progressors (Jin et al., 2005). Another study has described a significant increase in A3G expression in PBMCs from HIV-1-exposed seronegative individuals compared to HIV-1-seropositive patients or healthy controls (Biasin et al., 2007).

Specific single nucleotide polymorphisms (SNPs) in A3G and cullin5 can affect the rapidity of HIV-1 disease progression. An SNP in A3G (H186R), which is highly abundant in African Americans (37%) but rare in Europeans or European Americans (5%, <3%), is strongly associated with a more rapid decline in the number of CD4 T cells and accelerated progression to AIDS (An et al., 2004; Reddy, 2010). This can also be seen with a codon-changing variant in the fourth exon of A3G and a 3′ extragenetic mutation (Reddy et al., 2010). The cullin5 gene exhibits several different SNPs, which can be classified into two clusters or haplotypes. Interestingly, the two haplotypes produce opposing effects on CD4 T-cell counts during HIV-1 infection; the loss of these cells was delayed by cluster I and accelerated by cluster II. Cullin5 polymorphisms linked to accelerated CD4 T cell decline are associated with increased HIV-1 viral load (An et al., 2007). These findings raise the possibility that the Vif–cullin5 interface might represent a tractable target for small molecule development

Other studies indicate that A3G/F hypermutation of HIV-1 is associated with increased CD4 T-cells count (Land et al., 2008) and a 0.7 log¹⁰ reduction in viremia (Pace et al., 2006). However, some studies have not found a correlation between A3G/F mRNA levels or polymorphisms and viral load and/or CD4 T-cell counts (Cho et al., 2006; Do et al., 2005; Gandhi et al., 2008; Piantadosi et al., 2009; Reddy, 2010).

Interestingly, defective Vif alleles that cannot effectively neutralize A3G/F are readily detected in infected patients (Simon et al., 2005). Unopposed but nonlethal A3G/F action could promote viral sequence diversification in these patients. This raises the question of whether these G-to-A mutations might actually be beneficial for the virus, for example by promoting HIV-1 drug escape mutants or using A3G-induced mutations as a means to achieve rapid viral evolution.