APOBEC3G Antagonism by Vif or: When Structure Meets Biological and Evolutionary Studies

Abstract

Restriction factors serve as innate host defenses against viruses and act as critical barriers to cross-species transmission. In response, viruses have evolved accessory proteins to counteract restriction factors, enabling evasion of innate immune responses. The interplay between primate APOBEC3G (A3G) and lentiviral Vif exemplifies a molecular arms race between a restriction factor and its viral antagonist. This review integrates evolutionary and functional analyses of this system, showing how genetic signatures of molecular arms races map onto high-resolution cryo-EM structures. However, A3G’s interaction with Vif is not limited to the evolutionary dynamic interface, characterized by rapidly evolving residues under selective pressure, but also involves a conserved interface mediated by RNA binding which positions A3G for antagonism by Vif. These findings propose a model wherein Vif, and potentially other viral antagonists, target functional complexes using a dual strategy: leveraging both adaptive interfaces subject to evolutionary pressures and conserved interfaces that constrain host escape mechanisms.

Introduction

Lentiviruses, such as Human Immunodeficiency Virus (HIV) and Simian Immunodeficiency Virus (SIV), encode accessory proteins that counteract the antiviral defenses posed by host innate immune proteins known as restriction factors. Restriction factors function to block viral replication at various stages of the viral lifecycle. One example of a host encoded restriction factor is APOBEC3G (A3G) (1), as well as other members of the APOBEC3 (A3) family, APOBEC3A (A3A), APOBEC3B (A3B), APOBEC3C (A3C), APOBEC3D (A3D), APOBEC3F (A3F), and APOBEC3H (A3H). Of these, A3G appears to be the most potent anti-HIV protein, although A3C, A3D, A3F, and A3H also have varying levels of activity against HIV (2). All seven A3 proteins in the primate lineage are cytosine deaminases containing one or two conserved zinc-binding domains with a (Cys/His)-Xaa-Glu-Xaa_23–28-Pro-Cys-Xaa_2–4-Cys) motif (3). Although named for the mRNA-editing activity of the founding member, APOBEC1, the antiviral activity of the antiviral A3 proteins is primarily through the introduction of deleterious mutations in viral cDNA (4, 5). Lentiviruses have evolved an antagonist to APOBEC3G, called Vif. The retention of vif throughout lentiviral evolution, except in equine infectious anemia virus (6, 7), and its conserved role as an antagonist of host A3 proteins during infection, highlights its critical role in viral survival. The ability of the Vif protein to antagonize A3G and thereby facilitate successful viral replication depends on its capacity to bind A3G-containing complexes effectively. Therefore, understanding the molecular determinants of this interaction is crucial. Here, we highlight work elucidating the interface of A3G and Vif through both structural and evolution-guided approaches. These insights are critical for understanding the mechanisms by which HIV mediates immune evasion, as well as the determinants of viral spillover events that led to the origins of HIV-1 in humans.

The Vif: A3G “arms race” predicted sites of interaction

Viral antagonism of restriction factors often requires direct protein-protein interactions between the host and viral proteins. Therefore, host restriction factors and their viral antagonists are key players in the host-virus “molecular arms race”, an iterative cycle where hosts adapt to evade viral recognition, followed by viral counter-adaptation (8–11). These ongoing coevolutionary processes leave genetic signatures in restriction factors (9, 12, 13) in a process called “positive selection” where changes are favored to evade recognition or re-establish recognition. The iterative evasion/recognition between proteins drives adaptive changes in both the host and the pathogen genomes, fueling the coevolutionary dynamic of the molecular arms race (14, 15).

The ability of Vif to bind A3G is often species-specific (16–18). For example, HIV-1 Vif cannot counteract African Green monkey (AGM) and macaque A3G but can fully antagonize both human and chimpanzee A3G (17). Comparative analysis of A3G genes across primate species reveals that A3G has evolved to evade Vif antagonism, as evidenced by positive selection at residues that were functionally validated to be important determinants of species-specific Vif antagonism of A3G (13, 19, 20). These studies have shown that variations in A3 proteins across different species of primates have significantly influenced the evolution of the Vif protein, acting as barriers to cross-species transmission and selecting for Vif mutations that permit the antagonism of divergent A3 proteins (13, 21). For example, molecular differences between A3 proteins in hominids and Old World monkeys (OWM) have been pivotal barriers to cross-species transmission of SIVs and have shaped the evolution of HIV-1 (13, 21–23). It is now understood that while human and chimpanzee A3G are perfectly conserved at critical amino acids at the Vif-binding interface, AGM and macaque A3G exhibit differences in amino acid composition at these critical sites (Figure 1a, amino acids 128–130). Vif sequences are among the most divergent genes across all lineages of SIV (24), driven in part by evolutionary selection at A3G-interacting sites, as Vif adapts to the species-specific sequence identity of A3G.

Figure 1: Evolutionary and structural insights into A3G interaction with Vif E3 ligase complex.

(a) A3G acts as a barrier to cross-species transmission of SIV from Old World Monkeys (OWM) to hominid primates. Sequence alignments of residues 128–130 of primate A3G are indicated. Residues 128 and 130 undergo positive selection, with the most common amino acid at each position indicated in bold (left). Schematic showing the ancient origins of HIV-1 (right). SIV that infects chimpanzee (SIV_cpz) arose from a recombination of the SIV that infects red-capped mangabey monkey (SIV_rcm, red star) and the mustached guenon (SIV_mus). The amino acid of hominid primate A3G at position 128 is an aspartate, which limits infection of hominid primates by SIV from OWM where the Vif gene is adapted to antagonize a lysine or glutamate at this position. (b) Ribbon diagram of human A3G (blue) bound to substrate receptor of the Vif E3 ligase, referred to as VCBC, which is comprised of Vif (green) and the host proteins CBFβ (wheat), ELOC (purple) and ELOB (maroon). The copurified RNA molecule, shown in gold, includes four core nucleotides (NT1-NT4) that act as a molecular glue between A3G and Vif (PDB ID: 8CX0). Insets show zoom in views of the molecular arms race interface between Vif and A3G (top) and the RNA binding site (bottom). The same color code for the subunits is used throughout the review. Figure 1b adapted from reference (79) (CC BY 4.0).

The evolutionary pressure exerted on Vif by species-specific differences in A3G is well-illustrated in a study examining the four AGM subspecies: sabaeus, vervet, tantalus, and grivet monkeys (13). Each of these monkeys is infected with a distinct subtype of SIV, referred to as SIV_agm.Sab, SIV_agm.Ver, SIV_agm.Tan, and SIV_agm.Gri, respectively (24). Adaptive single-nucleotide polymorphisms in each species indicate that SIV infection in AGM selects for Vif-resistant A3G, which in turn drives a subspecies-specific counterrevolution in vif (13). Moreover, in vivo evolution experiments reveal that SIV_agm Vif evolves specificity to the A3G genotype within each host population (13). Specifically, codons 128 and 130 in AGM A3G were identified as polymorphic sites of adaptive change that confer resistance to Vif-mediated antagonism (13). These sites are located within the ¹²⁶FWKPDYQ¹³² motif, which had been well characterized as a critical region for Vif-A3G interactions (16, 18, 25–27). Previous studies analyzing A3G sequences across a broad range of primates identified the A3G locus as being under positive selection (19, 20). However, a more focused analysis of A3G sequences within the OWM clade revealed codons at positions 128 and 130 (Figure 1a, left), showing strong signals of diversifying selection (28). Thus, rescaling the species included in these analyses allowed for a more precise examination of evolution under a common selective pressure- in this case, the ongoing conflict with lentiviruses and the evasion of Vif recognition. Although there were additional codons under positive selection as well, importantly, positions 128 and 130 in A3G dictate species-specific recognition by Vif proteins (13, 16, 18, 26–31), with the D128 residue found in hominid A3G conferring resistance to most SIV Vifs, which evolved to antagonize OWM A3G that largely carry K128 (13, 16, 18, 26, 28, 32). Subsequently, the cryo-EM structure of human A3G and HIV-1 Vif validated the evolutionary predictions of the arms race interface between A3G and Vif (Figure 1b). Vif contains two domains: the α/β domain is comprised of an antiparallel beta sheet flanked by alpha helices that engage A3 proteins whereas the α domain contains three short alpha helices that bind components of the cellular E3 ligase. These interactions will be described in detail below.

The arms race between A3G and primate lentiviral Vif is further illustrated by a divergent evolution of A3G and Vif in the colobus monkey (Colobus guereza) A3G, which is broadly resistant to most SIV Vif proteins despite sharing the same amino acid identities at positions 128 and 130 as many Vif-susceptible species, such as sabaeus monkeys (28). However, the Vif protein encoded by the SIV that infects the olive colobus (SIV_olc) specifically antagonizes colobus A3G. This phenotype results from an amino acid insertion (⁶⁶SCK⁶⁸) within the A3G sequences of Colobinae monkeys, which, when removed, renders the protein susceptible to SIVagm.Sab Vif (28). These findings suggest that the insertion in colobus A3G causes a conformational change that conceals the canonical Vif-binding interface at positions 128 and 130, driving adaptive evolution in the vif genes of Colobinae-infecting SIVs.

Mechanism of A3G packaging and antiviral activity

The A3G antiviral restriction activity requires that the protein be packaged into virions and delivered to the target cell so that it is present at the site of viral reverse transcription, where it binds single-stranded viral DNA as the substrate for deamination (33). To be packaged into virions, A3 proteins must undergo RNA-dependent oligomerization (34, 35). Generally, A3 proteins do not exhibit a preference for the type of RNA bound (34, 35). Interaction with either host or viral RNA can result in A3G packaging into nascent virions (34, 35). However, A3G preferentially binds HIV-1 viral RNA in infected cells (36) and it has been suggested that this RNA-binding specificity mimics that of HIV nucleocapsid (NC), which also binds RNA to ensure incorporation into virions (36).

When packaged, A3G deaminates the second dC of 5′-CC dinucleotide sites in the newly synthesized viral minus-stranded ssDNA during reverse transcription (5, 30, 33, 37). Several factors contribute to the efficiency of viral ssDNA deamination, including the proximity of the target nucleotides to the 5’ end of the ssDNA, the orientation of the A3G C-terminal deaminase domain (CDA2) relative to the 5’ end of the ssDNA substrate, as well as the target ssDNA overall length (38). The dinucleotide preference exhibited by A3G is unique among A3 proteins (39, 40). For example, A3F prefers a 5′-TC dinucleotide substrate compared to the 5’-CC preference of A3G. This specificity influences the type and location of mutations introduced into the viral genome, thereby contributing differently to the viral mutational landscape. The A3 signature G-to-A hypermutations can be observed in HIV sequences derived from clinical samples (41–45). In fact, studies analyzing many independent sequences isolated from the same individual have found hypermutated sequences in 43% to 100% of HIV-infected individuals (42, 46–48). Such hypermutations can significantly impact viral fitness and replication. For example, extensive G-to-A hypermutation can lead to lethal mutagenesis of the virus, rendering it non-infectious. However, sub-lethal hypermutation levels may contribute to viral diversity and immune escape (49).

There is also evidence suggesting deamination-independent mechanisms of A3G-mediated HIV restriction. Introducing mutations into the active site of A3G, rendering it catalytically dead, still hinders replication of HIV-1, mouse mammary tumor virus, and murine leukemia virus to a certain degree (35, 50, 51). However, through experimental and mathematical investigation, one group has concluded that 99.3% of the antiviral effect of A3G is dependent on its deaminase activity (52). Importantly, the A3G deamination-independent mechanism also requires A3G to be packaged into daughter virions and may be mediated through an interaction between A3G and the viral reverse transcriptase, blocking the production of viral DNA (52).

Mechanism of A3G antagonism by Vif

Given that A3G requires encapsidation to exert its antiviral effects, Vif must prevent A3G packaging by facilitating its degradation in the producer cell. Specifically, HIV-1 Vif recruits various cellular proteins to assemble an E3 ubiquitin ligase complex, which ubiquitinates A3G, leading to its subsequent proteasomal degradation (53–55). The first crystal structure of HIV-1 Vif in complex with these host proteins revealed the precise binding interfaces involved in interactions with scaffold protein cullin 5 (CUL5), substrate adaptors elongin B (ELOB) and elongin C (ELOC), and core-binding factor β (CBFβ), which stabilizes Vif expression (56). Collectively, Vif, CBFβ, ELOB, and ELOC are referred to as the VCBC complex (Figure 2). CBFβ is best known as a binding partner of the RUNX family of transcription factors (57), and therefore it was surprising to find it as a component of the host CUL5 E3 complex that is necessary for Vif to antagonize A3G (58, 59). The binding interface between CBFβ and the α/β domain of Vif stabilizes the assembly of Vif with CUL5 by shielding a large hydrophobic patch on Vif from solvent exposure, thus protecting Vif from degradation (56). While this interaction is not required for Vif to bind A3G (58), it is essential for facilitating the ubiquitin-mediated degradation of A3G (58, 59). However, CBFβ is dispensable for Vif proteins encoded by non-primate infecting viruses. For example, Maedi-visna virus (MVV) Vif requires cyclophilin A (CypA) as a co-factor, while Bovine immunodeficiency virus (BIV) Vif assembles a host E3 complex without any cofactor (60). These results highlight different evolutionary paths taken by lentiviral Vif proteins to form complexes to degrade diverse species’ A3G proteins (61).

Figure 2: Structure of HIV-1 VCBC bound to an N-terminal fragment of cullin-5 (CUL5).

Vif, CBFβ, ELOC and ELOB co-fold into a structured substrate receptor for A3 family members. The proteins are depicted as transparent surfaces with underlying ribbons to highlight their secondary structures. Vif is divided into two domains: the α domain engages the CUL5 backbone and ELOBC heterodimer, while the α/β domain binds CBFβ, A3 family members, and regulatory subunits of the protein phosphatase PP2A. Additionally, Vif coordinates a zinc ion (Zn), which is essential for maintaining its structural integrity and interactions with other proteins in the complex.

Structure of Vif complex with ELOB, ELOC, CUL5, CBFβ, and A3G

The discovery of CBFβ enabled reconstitution of the Vif-E3 ligase activity on A3G using recombinant purified proteins and determination of the long sought-after structure of Vif by X-ray crystallography (56, 59). Co-expression of Vif with CBFβ along with ELOB and ELOC (ELOBC) in Escherichia coli (E.coli), yielded a soluble substrate receptor module that assembles CUL5 and a RING-box protein 2 (Rbx2) to form a Vif E3 ligase that can promote polyubiquitination of A3G but not Vif-resistant A3A (59). Detailed X-ray crystallographic studies of the VCBC complex identified specific Vif residues that are crucial for engaging with CBFβ, ELOB, ELOC and CUL5 (56, 62). HIV-1 Vif folds into two domains: the α domain contacts the ELOBC heterodimer and the cullin backbone, while the α/β domain interacts with CBFβ and A3 substrates (Figure 2). The two domains of Vif are rigidified by loops that engage a zinc ion that is buried between the domains and coordinated by a zinc finger motif with the sequence His-Xaa₅-Cys-Xaa_17–18-Cys-Xaa_3–5-His (HCCH motif). The α domain is comprised of two sequential helices that engage the cullin backbone and ELOBC respectively. The latter interaction exemplifies viral mimicry of host factors as the so called ‘BC box’, the binding site for the ELOBC complex, is found in cellular suppressor of cytokine signaling (SOCS) proteins and the Von Hippel-Lindau (VHL) protein. The α/β domain of Vif contains separate clusters of surface exposed residues implicated in binding of A3G, A3C/A3D/A3F and A3H, consistent with mutagenesis studies indicating a separation of function for these residues in antagonism of A3 family members (reviewed in (63)). Strikingly, the N-terminal β-strand of Vif extends the β-sheet of CBFβ, burying nearly 4800 Å of solvent accessible surface area between the two proteins, explaining why CBFβ is required to promote Vif folding and function in a variety of primate lentiviruses (21, 56). In addition to acting as a template for Vif folding, CBFβ may function to enable Vif antagonism of the extended A3 repertoire found in primates. In support, functional and structural studies indicate Vif and CBFβ form a composite interaction surface for A3F-CDA2 (64).

Adaptation of Vif to hominids at the origins of HIV-1

The immediate precursor to the HIV-1 Groups M/N vif gene is the vif gene from SIV_cpz, a virus that infects chimpanzees (65, 66) (Figure 1a, right). The SIV_cpz vif gene is derived from SIV_rcm, which infects the red-capped mangabey (Figure 1a, right) an Old World monkey species (32). The structure of the SIV_rcm Vif showed the conservation of VCBC interacting residues across broad SIV Vifs, suggesting that the formation of this complex to restrict A3G is evolutionarily conserved across primate lentiviral Vifs (21). While A3G proteins in chimpanzees and humans are highly conserved, A3G proteins in Old World monkeys, such as the red-capped mangabeys, show greater diversity. Despite the differences between Old World monkey and hominid A3G, the SIV_rcm Vif protein can still weakly antagonize hominid A3G, a result of evolutionary coincidence rather than direct adaptation (21, 32). Following the initial cross-species transmission, the SIV_rcm Vif adapted to fully antagonize hominid A3G, leading to the evolution of the SIV_cpz Vif protein. Given the high degree of similarity between human and chimpanzee A3G, it is unsurprising that the SIV_cpz Vif protein can fully antagonize human A3G, facilitating the emergence of HIV-1 Groups M/N (32). However, Vif from HIV-1 Groups O/P (as opposed to Groups M/N) were also influenced by adaptation in gorillas to a unique Q amino acid instead of a P at position 129 in gorilla A3G (31, 67) (Figure 1a, left).

While positive selection sequence analyses of A3G across a broad range of primates have been instrumental in identifying sequences in A3G that putatively interact with Vif, these approaches were less informative for identifying sequences in Vif that interact with A3G. The greater diversity, faster replication, and higher mutation rates of viruses lead to rapid adaptation to selective pressures, making it challenging to pinpoint specific drivers of adaptive change over an evolutionary time scale. Nonetheless, well before structures of Vif or A3G were available, insights into the amino acids critical for Vif-A3 interactions largely came from mutagenesis studies (29, 68–71). Moreover, Vif residues involved in the molecular arms race with A3G were implicated through prospective studies of Vif adaptations to A3G polymorphisms. For example, human and chimpanzee A3G contain the ¹²⁸DPD¹³⁰ motif and introducing a D128K mutation confers resistance to wildtype HIV-1 Vif (¹⁴DRMR¹⁷), but susceptibility to Vif proteins with ¹⁴SEMQ¹⁷ motif (29, 72). More broadly, in vitro evolution studies of Vif to antagonize the D128K or a D130R mutant suggested that Vif residues 14–17 are spatially close to A3G position 128 and Vif residue 82 in Loop 5 is close to A3G position 130 (29). Similarly, an in vivo evolution experiment in AGMs implicated Vif mutations at the equivalent position, Y83C, as a potential interfaces in Vif to target A3G (73).

The experimental evolution studies implicating changes in loop 5 of Vif as required for the adaptation of Vif to a new species were validated when the structure of the SIV_rcm Vif bound to ELOB, ELOC, and CBFβ was solved (21). Like many other OWMs, red-capped mangabeys have lysine (K) at position 128, which impairs SIV_rcm Vif’s ability to fully antagonize chimpanzee A3G (23, 32). The major structural changes in the VCBC complex between HIV-1 and SIV_rcm Vif were found in Loop 5 of Vif, and further functional viral infectivity assays revealed that a single amino-acid substitution in Loop 5 of Vif of SIV_rcm, Y86H, conferred specificity toward both chimpanzee and human A3G, and therefore the Vif adaptations at the origins of HIV-1 (21).

Cryo-EM structures unveil Vif-APOBEC3 evolutionary dynamics

While previous studies provided hints at the interface between Vif and A3G, the exact nature of their interactions would not be solved until a complete structure of the complex was determined. Moreover, the predicted interface from the mutagenesis and positive selection analyses appears smaller than expected for the protein size of Vif and A3G, suggesting additional interactions may be involved. Investigating these interactions required high-resolution structural studies; however, this endeavor had been a major challenge for the field because wild-type full-length A3 proteins are poorly behaved in solution (74, 75). We reasoned that co-expression of A3G, Vif and associated host factors using the insect cell expression system could facilitate co-assembly of the complex, obviating the need for solubility enhancing mutations or covalent fusion strategies as employed in prior studies (64, 75–78). This strategy enabled us to obtain a cryo-EM structure of the wild-type human A3G in complex with the substrate receptor complex (HIV-1 Vif, CBFβ, ELOB, and ELOC) at 2.7 Å resolution (79). Notably, as discussed later, this structure unveiled RNA as a molecular glue to hold A3G and Vif together.

Additional Vif-A3G structures were obtained using engineered A3G proteins that render A3G more soluble in E. coli. For instance, rhesus macaque (rhA3G), which is inherently more soluble than its human counterpart, was modified with a K128D mutation to enable recognition by HIV-1 Vif, facilitating its use in structural studies. (80). In parallel, guided by comparative amino acid sequence analysis with soluble A3 variants (A3A, A3B-CDA2, and A3G-CDA2), a soluble A3G (sA3G) construct that retained approximately 80% identity to the wild-type A3G was utilized to obtain the cryo-EM structure with HIV-1 Vif (81). Both the humanized rhA3G and sA3G were shown to be susceptible to Vif-mediated degradation, justifying their use for in vitro reconstitution of Vif-A3G complexes for structural analysis.

The cryo-EM structure of hA3G-VCBC showed that the molecular arms race interface is featured by hydrogen bonds between D128 and D130 of human A3G and R15 and Q83 of HIV-1 Vif, respectively (79, 81) (Figure 1b). Thus, the structure validated the predictions from positive selection (on the A3G side) (19, 20, 73), and the mutagenesis/forced evolution studies on the Vif side (29, 68–71). The cryo-EM structures of the A3G-Vif complex also facilitated further interpretation of how the Vif protein encoded by the SIV that infects the red-capped mangabey monkey adapted to enable antagonism of chimpanzee A3G, enabling cross-species transmission (23, 32). For example, the relevance of Vif R15 in the arms race interface with A3G is important since SIV_rcm Vif already possessed R15, explaining why SIV_rcm Vif, but not SIV_agm Vif is also to partially (but not fully) antagonize chimpanzee A3G (21, 23). Thus, SIV_rcm Vif was already poised for the cross-species adaptation to hominid A3G.

Further, comparative modeling of SIV_rcm Vif in complex with rcmA3G, informed by the HIV-1 Vif/hA3G cryo-EM structure, suggests a rewiring of the hydrogen bond network at the arms race interface, with Y86 of SIV_rcm Vif (corresponding to Q/H 83 in HIV-1 Vif) interacting with K128 of rcmA3G (Figure 3a) (79). This model indicates that adaptive changes at this interface are not one-to-one but instead result in a repacking of interactions mediated by SIV_rcm Vif Y86 and the corresponding residue of SIV_cpz/HIV-1 Vif (Q/H 83) in response to K-to-D change at position 128 of A3G as the virus adapted from red-capped mangabey to hominids (21). The electrostatic interaction between Vif and A3G that occur at the molecular arms race interface are specific but probably weak. In support, loss-of-function studies and comparative modelling studies have revealed additional key residue mediating Vif-A3G interactions. SIV_rcm Vif, like HIV-1 Vif, engages A3G through conserved hydrophobic interactions (Figure 3b) (21, 79).

Figure 3: Comparative modeling suggests adaptation of SIV_rcm Vif to hominid primate A3G results from a repacking of the molecular arms race interface.

Key residues lining the molecular arms race and RNA binding interface are shown in (a) and (b), respectively. (a) For HIV-1 Vif, residues R15 and Q83 interact with D128 and D130 of A3G through electrostatic and hydrogen bonding interactions whereas in SIV_rcm Vif, the equivalent positions (R16 and Y86) sample different sidechain rotamers to engage K128 and D130 of rcmA3G. (b) The interactions with RNA are predicted to be conserved between HIV-1 Vif and human A3G and complex formed between SIV_rcm Vif and rcmA3G. Hydrophobic contacts between RNA, Vif, and A3G complement the arms race interface, stabilizing the interaction and enhancing binding affinity. Figure adapted from reference (79) (CC BY 4.0).

RNA as a molecular glue: changing the Vif-A3G interaction paradigm

The Vif-A3G complex was previously believed to be exclusively mediated by protein-protein interaction. However, recent cryo-EM structures have challenged this perspective by showing RNA acts as a molecular glue that bridges the Vif-A3G interaction (79–81). In these structures Vif and A3G sandwich single-stranded RNA. A3G makes base specific interactions with RNA whereas Vif does not; moreover, structural studies of A3G in complex with RNA reveal the same binding mode with RNA as when in complex with Vif, leading us and others to propose Vif binds to preformed A3G/RNA complexes formed in cells.

In the absence of Vif, A3 proteins bind to HIV-1 genomic RNA and the nucleocapsid (NC) domain of HIV-1 Gag precursor for assembly into the virions (34, 36, 82–87). While the mechanism for A3G encapsidation remains unclear, biochemical and biophysical analyses suggested that A3G forms RNA-dependent oligomers recruited to the plasma membrane by the Gag protein for incorporation into budding virions (88). High-throughput sequencing of RNA isolated by crosslinking immunoprecipitation (CLIP-HTS) was used to investigate the RNA binding specificity of A3G and other A3 family proteins (A3F and A3H) known for their anti-HIV activity in infected cells and virions (36). Their analysis revealed a predilection for A3 binding to RNA with G-rich and A-rich sequences. This preference aligns with the composition of HIV-1 genomic RNA, which is rich in adenine, constituting 36.2% of its sequence, and low in cytosine, at 17.6% (89, 90). This adenine bias is even more pronounced in single-stranded regions of the viral RNA, reaching 47.5%, while AA is the most predominant dinucleotide sequence (90, 91). The preferential interaction of A3G with viral RNA, particularly in the single-stranded regions rich in adenine dinucleotides, suggests an evolutionary adaptation of A3G for efficient encapsidation into the HIV-1 and potentially other lentiviruses. It is also noteworthy that A3H exhibits a distinct binding pattern, favoring RNA duplexes of seven or more nucleotides, which is in contrast to A3G, whose binding sites on the HIV-1 genome are less likely to form such duplex structures (92). This divergence in RNA-binding preferences, with A3H favoring structured duplex or hairpin RNA and A3G recognizing di-adenine sequence motifs within single-stranded RNA, could ensure their concomitant packaging with HIV-1 RNA into the nascent virions (92, 93).

In the presence of Vif, A3G recognizes a di-purine motif, specifically the sequence 5’-AA-3’ and, to a lesser extent, 5’-GA-3’, in the single-stranded RNA. The preferential binding of AA sequences over GA sequences in A3G is supported by replacement studies, where the substitution of the AA with GA motif resulted in a significant reduction in binding affinity to humanized rhA3G (93). This sequence specificity is primarily attributed to the extensive interaction network involving aliphatic and aromatic stacking, as well as hydrogen bonding (Figure 4). For example, the 5’ base of AA is cradled by an amalgam composed of residues from the CDA1 (I26 and W127) and CDA2 (F268 and K270) of A3G, as well as the Vif protein (H42, H43, and Y44), illustrating a coordinated interaction crucial for Vif-mediated A3G degradation (Figure 4a). Additionally, the second adenine base of the 5’-AA-3’ is nestled within a hydrophobic pocket formed by residues of A3G-CDA1 (I26, W94, Y124, Y125, and W127), where it engages in hydrogen bonding with the main-chain amide of Y125 and the carbonyl groups of P25 and L123 (Figure 4b). This configuration favors the exclusive fitting of the adenine base due to the spatial clash between the O6 atom of guanine and the carbonyl groups of P25 and L123. Such an interaction network is critical for the preferential binding of AA dinucleotides over other sequences, thereby influencing the selective incorporation of hA3G in Vif-deficient HIV-1 (36). It is notable that the HIV genome is modified by N6-methylation (94). Based on the cryo-EM structure of A3G with HIV-1 Vif, modification of adenine by N6-methylation would be incompatible with A3G binding to RNA; therefore, it is tempting to speculate that N6-methylation could play a role in A3G packaging or antagonism by Vif.

Figure 4: Sequence-specific recognition of a purine dinucleotide by A3G and Vif at the RNA interface.

The RNA that bridges the Vif-A3G interface contains a purine dinucleotide that forms sequence-specific interaction with Vif and A3G. (a) The first purine base is stabilized by residues from A3G-CDA1 and CDA2, along with Vif. (b) The second adenine base is specifically accommodated in a hydrophobic pocket within A3G-CDA1, with key hydrogen bonds enhancing specificity. In both cases, the backbone amide and carbonyl groups of A3G form hydrogen bond donors and acceptors with the Watson-Crick faces of the purine dinucleotide. Dashed lines indicate hydrogen bonds. Blue, red and orange spheres represent nitrogen, oxygen and phosphate atoms, respectively. Figure adapted from reference (79) (CC BY 4.0).

Early studies had predominantly focused on the CDA1 domain when investigating RNA-binding residues within A3G (95). Indeed, each of the CDA1 residues now known to interact directly with the 5’-AA-3’ motif has been previously implicated in RNA-dependent A3G oligomerization, viral packaging, and Vif engagement (27, 96). However, recent cryo-EM structures have expanded this view by revealing that residues F268 and K270 in the CDA2 domain also play a role in RNA engagement (Figure 4a) (79–81). Alanine substitution of these CDA2 residues resulted in a significant decline in their affinity for the Vif E3 ligase complex, elucidating their indispensable contribution to the structural integrity of the A3G-RNA-Vif complex (80).

The cryo-EM structure of hA3G with VCBC also revealed Vif residues that make specific interactions with the bridging RNA, which are critical for antagonizing A3G (Figure 4) (79). H43 and Y44 are unique as they interact with both RNA and A3G, whereas K22, K26, Y40, and H42 engage exclusively with RNA. The selective recognition for RNA, as opposed to DNA, can be partially attributed to the interaction of Vif H42 with the 2′-hydroxyl group of 5’A ribose via hydrogen bonding (79). As thoroughly discussed below, the RNA-dependent interaction is crucial for Vif-mediated A3G ubiquitination and degradation. Unlike H42, H43, and Y44, which interact with the RNA base, residues K22, K26, and Y40 of Vif engage the RNA through interaction with backbone phosphate, relying on charge complementarities and hydrogen bonding (79).

Extensive mutagenesis studies demonstrated the importance of these residues for A3G degradation by Vif (71, 79). Particularly, mutations at positions 26 and 40 impaired Vif binding to RNA in vitro, which in turn compromised viral infectivity in non-permissive T cells (97, 98). Furthermore, residues of Vif and A3G that bind RNA are highly conserved in all primate lentiviruses and primate species, suggesting that the RNA molecular glue observed in the cryo-EM structure of HIV-1 Vif bound to hA3G is conserved in all primate lentiviral Vif-A3G complexes.

Formal proof that Vif requires RNA to promote ubiquitination of A3G was provided by in vitro reconstitution with recombinant purified proteins. Humanized rhA3G(K128D) requires RNA to copurify with and be ubiquitinated by the HIV-1 Vif E3 ligase (80). Likewise, RNA, but not DNA, promotes copurification and ubiquitination of sA3G by the HIV-1 Vif E3 ligase (81). The binding mode of a four-nucleotide core RNA molecular glue observed in the cryo-EM structures of these A3G/HIV-1 Vif complexes is the same as the wild-type hA3G/HIV-1 Vif complex. In particular, specificity for at least two nucleotides may be conferred through a composite binding pocket formed between A3G and Vif, and an adenine binding pocket on A3G (Figure 4). Though not tested, the structures suggest only single-stranded RNA could be accommodated by the Vif-A3G interface. Together, these data suggest the substrate of the Vif E3 ligase is not A3G but instead an A3G complex with single-stranded RNA.

In summary, A3G binding to Vif extends beyond the conventional arms race interface of diversifying selection to include conserved residues involved in RNA binding, underlining the indispensable roles of RNA in the antagonism between A3G and Vif.

Structural insights into A3G mono-ubiquitin priming by ARIH2

Cullin-RING E3 ligases promote ubiquitination by acting in the last step of a three enzyme (E1-E2-E3/E3) cascade (99). RING-between-RING E3 ligases work in unison with CRL family members to prime substrates by monoubiquitination, which then leads to polyubiquitin chain formation by a specific ubiquitin-conjugating enzyme or E2 (100). ARIH2 is essential for viral replication, Vif antagonism of A3G and accelerates the rate of polyubiquitin chain formation in vitro through this ‘priming-and-extending mechanism’ (101). The concordance between acceptor lysines on A3G targeted by Vif in cells and those targeted in vitro using recombinant purified ARIH2 and the Vif E3 ligase suggests ARIH2 is a bona fide coenzyme that functions with CRL5 to promote A3G degradation (101–103). Cryo-EM structural studies of ARIH1 bound to Cul1/Rbx1 and ARIH2 bound to Cul5/Rbx2 enabled us to construct a comparative model of how A3G is ubiquitinated by HIV-1 Vif E3 ligase (79, 104, 105). The model shows that ubiquitin acceptor lysines on A3G required for Vif mediated degradation are in close proximity to the ARIH2-ubiquitin conjugate posed for catalysis.

A model for A3G antiviral activity and Vif antagonism as a consequence of RNA binding

Several lines of evidence support the notion that Vif promotes ubiquitin-dependent degradation of A3G when it is in its most dangerous form for the virus: bound to RNA and in the process of being packaged. First, in the absence of Vif, CLIP-HTS studies suggest A3G binds purine-rich sequences present in mRNA, non-coding RNA, and lentiviral genomic RNA (36, 83). Second, functional analyses have shown that specific A3G residues, such as Y124 and W127, which are critical for RNA binding in vitro, also facilitate its packaging into capsids of vif-deficient viruses (27). Third, it is notable that lentiviral genomes are inherently enriched with purine-sequences (91). Finally, crystallographic studies of rhA3G bound to RNA reveal a binding mode that is conserved when in complex with Vif as determined by cryo-EM (80, 93). Collectively, these findings suggest that Vif targets A3G bound to various RNA molecules during infection, with a particular emphasis on degrading A3G associated with viral genomic RNA, a mechanism likely crucial for viral fitness.

The interaction between Vif and the A3G/RNA complex significantly impacts host strategies to escape viral antagonism. Mutations within the RNA-binding site of A3G would compromise its antiviral activity, thereby diminishing its contribution to host defense and, consequently, host fitness (27, 96). In contrast, positive selection pressures acting on A3G at the molecular arms race interface offer a mechanism for hosts to adapt and escape viral countermeasures. Despite its small size, this interface is a critical vulnerability for the virus, as it provides an avenue for the host to achieve evolutionary escape while simultaneously presenting a target that the virus can rapidly adapt to. This dynamic interplay at the molecular level has profoundly influenced the evolutionary trajectory of HIV-1.

Although A3G can bind various RNAs, during HIV-1 infection it preferentially binds viral genomic RNA in a manner that mimics the RNA-binding properties of the viral NC protein (36, 83, 84, 106). Interestingly, Vif has also been shown to possess viral RNA-binding capabilities (98, 107–109), suggesting a potential mechanism whereby the shared affinity for viral RNA may spatially co-localize Vif and A3G. This co-localization could be a strategic feature that enhances Vif’s ability to target A3G. The structural data for human A3G and HIV-1 Vif were obtained from protein complexes purified from RNase-treated cell lysates, restricting our capacity to fully characterize the RNA present at the shared Vif-A3G interface. Future investigations aimed at elucidating the RNA-binding specificity that underpins Vif and A3G interactions will be crucial to understanding the precise mechanisms by which Vif targets A3G.

Interactions of Vif with other host proteins

Although this review has primarily focused on the interaction between Vif and A3G, Vif also antagonizes A3C, A3D, A3F, and A3H (110–112). A critical and yet unresolved question is how Vif orchestrates the degradation of these multiple, diverse A3 proteins. Additionally, Vif exhibits a broader range of interactions by binding to and promoting the degradation of regulatory subunits of the protein phosphatase 2A (PP2A) family (113–118). Structures of the A3F CDA2, A3H, and PP2A regulatory subunits bound to Vif were recently determined by cryo-EM (64, 118, 119). While the binding sites of A3F, A3G, and A3H are genetically distinct to some extent, structural analysis predicts binding of A3G is mutually exclusively with A3F and PP2A regulatory subunits due to steric clashes (data not shown). This observation raises intriguing questions about potential trade-offs in Vif’s ability to antagonize different A3s. Specifically, it remains unclear whether Vif’s interaction with A3G must be suboptimal to accommodate binding requirements for other A3 proteins, or how these competing interactions shape the virus’s evolutionary strategies in response to A3 polymorphisms. Deep mutational scanning studies, conducted in cells expressing individual A3 proteins as well as the full A3 repertoire, could provide critical insights into these complex dynamics and inform our understanding of how HIV-1 balances these evolutionary constraints.