HIV-1 Vif is required for growth on certain cell types and in vivo

Lentiviruses are a genus within the broader family of retroviruses. In addition to the ubiquitous gag, pol and env genes, lentiviruses possess several accessory genes that have integral roles in viral pathogenesis (recently reviewed by Refs Reference Malim and Emerman3, Reference Gramberg, Sunseri and Landau4). The prototypical lentivirus, HIV-1, contains two accessory genes involved in the regulation of viral gene expression and four additional accessory genes more directly involved in pathogenesis. One of these encodes the virion infectivity factor (Vif).

The vif gene is located in the centre of the HIV-1 genome, overlapping the 3′ end of pol and the 5′ end of the accessory gene vpr. Expressed from a partially spliced, Rev-dependent subgenomic mRNA (Refs Reference Garrett, Tiley and Cullen5, Reference Schwartz, Felber and Pavlakis6), the protein product is 192 residues, highly basic and 23 kDa in size (Refs Reference Kan7, Reference Lee8, Reference Sodroski9). Vif is required for productive infection in vivo and for infection of primary CD4⁺ T cells, monocytes and macrophages ex vivo (Refs Reference Desrosiers10, Reference Strebel11, Reference Gabuzda12, Reference Gabuzda13). This requirement, however, is variable in some human T cell lines (Ref. Reference Fisher14), allowing these lines to be grouped as ‘permissive’ for the replication of Vif-deficient HIV-1 (e.g. CEM-SS, SupT1) or ‘nonpermissive’ (e.g. CEM, H9).

When produced in permissive cells, virions from Vif-deficient HIV-1 have no quantitative or qualitative defects relative to wild-type virus except for the absence of the Vif protein itself. (Refs Reference Strebel11, Reference Fisher14). When produced in nonpermissive cells, however, Vif-deficient HIV-1 virions rarely complete reverse transcription (Refs Reference Simon and Malim15, Reference Courcoul16, Reference Goncalves17, Reference Sova and Volsky18).

Given the above, two scenarios were considered likely to explain the phenotype of Vif-deficient viruses on nonpermissive cells: (1) permissive cells contain an endogenous factor that complements the function lacking in a Vif-deficient virus or (2) nonpermissive cells contain a Vif-suppressible factor with antiviral restriction activity. To determine which hypothesis was correct, two groups tested the infectivity of virions produced from heterokaryons formed by the fusion of permissive and nonpermissive cells (Refs Reference Madani and Kabat19, Reference Simon20). Under these conditions, the production of infectious virions would support the former hypothesis, while the production of noninfectious virions would support the latter. In fact, such heterokaryons formed from the fusion of permissive and nonpermissive cells produce virions with diminished infectivity, indicating that nonpermissive cells contain a Vif-sensitive dominant restriction factor.

APOBEC3G is a Vif-sensitive restrictor of HIV-1 infectivity with DNA cytidine deaminase activity

To identify the cellular factor suggested by cell fusion experiments, Sheehy et al. sought to isolate genes uniquely expressed in a nonpermissive cell line called CEM (Ref. Reference Sheehy21) by suppression subtractive hybridisation, which simultaneously suppresses the recovery of transcripts common to CEM and its permissive derivative CEM-SS and enriches the recovery of transcripts that are more abundant in CEM. Among many sequences, a cDNA encoding APOBEC3G (A3G) was isolated, and engineering naturally permissive CEM-SS cells to express A3G rendered these cells nonpermissive for the replication of Vif-deficient but not wild-type virus. A3G therefore accounts for the nonpermissive phenotype of CEM, although a number of studies since then suggest that other APOBEC3 proteins may also contribute to the nonpermissive phenotype (Table 1). Of particular note, passage of Vif-deficient HIV-1 on permissive T cells stably expressing A3G yields viral variants that grow efficiently in the presence of A3G but not in CEM cells (Ref. Reference Haché22), demonstrating that other factors beyond A3G are also capable of restricting HIV-1.

At approximately the same time, two other groups independently identified A3G, not for its antiviral activity but rather for its enzymatic activity, which provided a crucial clue about the mechanism by which A3G might restrict HIV-1. Jarmuz et al. identified A3G as part of a family of polynucleotide cytidine deaminases related to the mRNA-editing enzyme APOBEC1, which edits the apolipoprotein B mRNA (Ref. Reference Jarmuz23). (The acronym APOBEC derives from APOBEC1 and stands for ‘apolipoprotein B mRNA editing, catalytic polypeptide’; human APOBEC3 proteins are designated ‘apolipoprotein B mRNA editing, catalytic polypeptide-like 3′ followed by the letter, A–H, identifying the specific protein.) Meanwhile, Harris and colleagues hypothesised, based on homology between APOBEC1 and the DNA mutator activation-induced deaminase (AID/AICDA), which is involved in antibody diversification, that APOBEC1 and its homologues can act on DNA substrates. Indeed, APOBEC1, A3G, A3C and AID were all shown to mutate DNA (Refs Reference Petersen-Mahrt, Harris and Neuberger24, Reference Harris, Petersen-Mahrt and Neuberger25).

HIV-1 restriction by A3G is mediated by deamination of viral cDNA cytidine during reverse transcription

Soon after the fortuitous convergence of data showing A3G's capacity to restrict HIV-1 and deaminate DNA cytidines, several groups demonstrated that the major basis for retroviral restriction is the deamination of minus-strand viral cDNA during reverse transcription (Refs Reference Harris26, Reference Mangeat27, Reference Zhang28). The proviral DNA of a Vif-deficient virus grown in the presence of A3G becomes heavily mutated, with characteristic genomic plus-strand G-to-A transitions. This strand bias is only explained by deamination of the minus-strand reverse transcript, which after deamination from C-to-U templates A rather than G on the viral plus-strand (Refs Reference Harris26, Reference Mangeat27, Reference Zhang28, Reference Lecossier29) (Fig. 1). Furthermore, some catalytically defective A3G mutants are unable to as effectively reproduce the nonpermissive phenotype, which represents important additional evidence for the dependence of restriction on deaminase activity (Refs Reference Mangeat27, Reference Zhang28, Reference Browne, Allers and Landau30, Reference Holmes31, Reference Miyagi32, Reference Schumacher33).

Figure 1.

APOBEC3 and Vif function. (a) Vif-deficient HIV infection. In a ‘nonpermissive producer cell’ that expresses the DNA cytidine deaminase APOBEC3G (A3G), gene expression of Vif-deficient HIV-1 proceeds normally, but during assembly A3G (pink circles) is encapsidated along with normal virion components. Virions with encapsidated A3G bud from a producer cell and enter a target cell normally. Reverse transcription, however, is blocked by the presence of A3G, either by the direct inhibition of cDNA synthesis or the degradation of uridine-containing cDNA, arising from cytidine deamination. Uridine-containing cDNA templates A on the viral plus-strand, resulting in G-to-A hypermutation. A direct A3G-mediated block to integration has also been proposed (not shown). (b) Vif-proficient HIV infection. To prevent encapsidation, Vif expressed from a Vif-proficient virus links A3G to an E3 ligase complex, which results in the ubiquitination of A3G and its eventual degradation in the proteasome. Alternatively, Vif may directly inhibit encapsidation, lower A3G expression and/or inhibit A3G catalytic activity (not shown). Subsequent release, entry, reverse transcription and integration are shown proceeding normally. Abbreviations: dsDNA, double-stranded DNA; HIV, human immunodeficiency virus.

Although hypermutation is clearly a crucial part of the retrovirus-restriction mechanism, evidence also exists that the catalytic activity of A3G may not wholly explain its ability to restrict HIV-1 (Refs Reference Holmes31, Reference Shindo34, Reference Newman35, Reference Bishop, Holmes and Malim36, Reference Bishop37, Reference Iwatani38, Reference Guo39, Reference Yang40, Reference Li41, Reference Luo42). These studies suggest deaminase-independent effects on various specific steps in reverse transcription or integration (Fig. 1), although at least one report finds some of these same effects to be dependent on catalytic activity (Ref. Reference Mbisa43). While a substantial quantity of evidence suggests deaminase-independent restriction by A3G, the sheer diversity of the steps at which A3G is proposed to exert its deaminase-independent effects makes it highly improbable that all of these phenomena are physiological. In fact, several groups have suggested that the deaminase-independent effects of A3G may represent overexpression artifacts, as stable expression of physiological levels of catalytically inactive A3G has little if any effect on HIV-1 (Refs Reference Browne, Allers and Landau30, Reference Miyagi32, Reference Schumacher, Nissley and Harris44). Thus, it appears that deaminase activity is largely required for the inhibitory activity of A3G under physiological conditions. Whether this will also be the case for other APOBEC3 proteins, which may have more pronounced deaminase-independent effects (Ref. Reference Holmes31), remains to be seen. Moreover, it is important to note that deaminase-independent mechanisms may be more central to the restriction of many other retroelements including hepatitis B virus and LINE-1 (Refs Reference Turelli45, Reference Bogerd46, Reference Muckenfuss47, Reference Stenglein and Harris48, and reviewed recently by Refs Reference Holmes, Malim and Bishop49, Reference Bonvin and Greeve50).

It was proposed early on that the effect of A3G-mediated deamination might be to shunt mutated viral cDNA into a base-excision repair pathway that would ultimately result in the degradation of the viral genome (Refs Reference Harris26, Reference Harris51). Under this model, a uracil DNA glycosylase such as UNG would recognise and excise cDNA uracils, leaving an abasic site at which an endonuclease such as APEX (APEX1) would cleave the cDNA backbone. While one report supports such a scenario (Ref. Reference Yang52), different data sets focusing on uracil DNA glycosylase functions imply either reparative (Refs Reference Chen53, Reference Mansky54, Reference Priet55) or degradative (Ref. Reference Schröfelbauer56) effects on the viral genome. Alternatively, uracil DNA glycosylase activity may simply be irrelevant to the viral lifecycle (Refs Reference Schumacher33, Reference Kaiser and Emerman57). Further work will be required to elucidate the potential mechanisms by which viral cDNA is processed following APOBEC3-mediated deamination.

HIV-1 Vif preserves viral infectivity by preventing A3G encapsidation

Many mechanisms are plausible to explain how HIV-1 Vif counteracts the restrictive capacity of A3G. Most authors agree that this occurs by inhibiting A3G packaging, although the details of the mechanism(s) by which this inhibition occurs are debatable. At least part of the mechanism by which Vif overcomes restriction by A3G involves the Vif-mediated degradation of A3G through a proteasome-dependent pathway (Refs Reference Conticello, Harris and Neuberger58, Reference Marin59, Reference Sheehy, Gaddis and Malim60, Reference Stopak61, Reference Yu62, Reference Mehle63). Specifically, Vif binds both to A3G and to an E3 ubiquitin ligase complex consisting of cullin-5 (CUL5), ring-box protein 2 (RBX2/RNF7) and elongins B (TCEB2) and C (TCEB1). By bringing A3G into proximity with the ubiquitin ligase complex, Vif facilitates the ubiquitination and eventual degradation of A3G through the proteasome (Refs Reference Yu62, Reference Mehle64, Reference Yu65) (Fig. 1).

Evidence also exists suggesting that Vif can directly inhibit A3G encapsidation (Refs Reference Sheehy, Gaddis and Malim60, Reference Kao66, Reference Mariani67). Such a mechanism need not require degradation as an end point. Other reports indicate a Vif-dependent reduction in A3G expression (Refs Reference Stopak61, Reference Kao66), and it is also possible that Vif may function by directly inhibiting the catalytic activity of A3G (Ref. Reference Santa-Marta68).

The mechanisms proposed above are not necessarily mutually exclusive. While details may differ, an overarching message is readily apparent in the current Vif literature: preventing the encapsidation of A3G is crucial for the ability of Vif to preserve the infectivity of HIV-1. In fact, reducing the encapsidation of A3G by simply increasing particle production to titrate out the restriction factor is at least one mechanism by which HIV-1 can resist restriction in the absence of Vif function (Ref. Reference Haché22).

Multiple APOBEC3 family proteins have activity against HIV-1

While discussion to this point has focused on a basic description of the interactions between Vif and A3G, there are six additional proteins in the APOBEC3 family – APOBEC3A (A3A), APOBEC3B (A3B), APOBEC3C (A3C), APOBEC3DE (A3DE), APOBEC3F (A3F) and APOBEC3H (A3H) (Refs Reference Harris and Liddament2, Reference Jarmuz23, Reference Conticello69, Reference Wedekind70) – all of which have been implicated in the mutation and restriction of HIV-1 under certain conditions. A summary of references with data concerning the restriction activity and Vif sensitivity of each APOBEC3 protein is provided in Table 1; current in-depth reviews of the differential properties of these APOBEC3 proteins may be found elsewhere (e.g. Refs Reference Chiu and Greene1, Reference Malim71).

While the results apparent in Table 1 do reflect a lack of consensus on the ability of certain APOBEC3 proteins to function against HIV-1, there are areas of agreement. Most notably, it is well accepted that A3F and, more prominently, A3G have strong activity against HIV-1. Moreover, there is a second tier consisting of A3B and A3DE where antiviral activity is generally reproducible in overexpression single-cycle infectivity assays but may not be apparent in a more physiological spreading infection system. Further study of these proteins, particularly in spreading infections, will be required to determine definitively their ability to act as dominant restriction factors. It is generally accepted, however, that A3G, A3F, A3DE and perhaps A3H (see below) are Vif-sensitive, while A3B is not, which implies that these Vif-sensitive factors are likely the most relevant for the restriction of HIV-1 in vivo. That said, the fact that A3B expression increases significantly in lymphoid tissue from acutely infected patients along with A3F and A3G (Ref. Reference Li72) may also suggest a role for this Vif-insensitive protein.

A3H represents a special case since only one known allele appears to encode a protein that is stable in human cells, and stability is required for its ability to effectively restrict HIV-1 (Refs Reference OhAinle73, Reference Harari74, Reference Tan75). It is also presently unclear to what extent this APOBEC3 protein is Vif-sensitive (Table 1), although the fact that some have found it sensitive to Vif suggests that it may be important for the restriction of HIV-1. Perhaps even more muddled is the case of A3C. A3C may be sensitive to Vif, but its restriction activity against HIV-1 is relatively weak, arguing against a role for this APOBEC3 protein in vivo. It should be noted, however, that A3C is quite active against simian immunodeficiency virus (Ref. Reference Yu76). While some case can thus be made for the ability of six of seven human APOBEC3 proteins to dominantly restrict HIV-1, there are few findings to support such a role for A3A, as the references showing the strongest effects require the fusion of A3A to a helper protein that shuttles it into the viral core (see discussion below).

Beyond restriction and Vif-sensitivity phenotypes, it is important to emphasise one key characteristic that helps to distinguish among APOBEC3 proteins: the sequence context in which they typically mutate DNA cytidines to uridines (Table 1). A3G preferentially mutates cytidine residues that are preceded by another cytidine, with an even stronger preference for the context 5′-CCCA-3′ in minus-strand viral cDNA, where the underlined deoxycytidine is the nucleotide preferentially targeted (Refs Reference Harris26, Reference Mangeat27, Reference Zhang28, Reference Yu77). In plus-strand reverse-transcribed cDNA, this manifests as 5′-TGGG-3′ to 5′-TAGG-3′ mutations, or 5′-GG-3′ to 5′-AG-3′ in the less stringent dinucleotide context. Other APOBEC3 proteins, however, generally display a preference for the mutation of cytidines preceded by another deoxyribonucleotide, most often thymidine (e.g. (Refs Reference Bishop78, Reference Liddament79, Reference Wiegand80, Reference Zheng81). This results in 5′-TC-3′ to 5′-TU-3′ transitions, which manifest on plus-strand viral cDNA as 5′-GA-3′ to 5′-AA-3′ transitions.

The identification of these different sequence contexts is important for assessing which APOBEC3 proteins act on a given viral substrate in a context where multiple APOBEC3 proteins are expressed such as in vivo. Furthermore, assessing these contexts helps one to appreciate how hypermutation can be detrimental to a virus. Aside from deleterious amino acid or other mutations that may occur in unique sequence contexts, any UGG codon encoding tryptophan is susceptible to the creation of a UAG or perhaps even a UAA stop, particularly in the presence of A3G. Similarly, the AUG start codon may place the wobble base in an APOBEC3-susceptible hotspot depending on the identity of the next base in the sequence, thus making the ablation of initiating methionines another distinct possibility. Extensive mutation may also disrupt HIV-1 genomic RNA secondary structures, which are important for the regulation of HIV-1 gene expression (Ref. Reference Watts82).

APOBEC3 proteins mutate the HIV-1 genome

Even prior to the discovery of APOBEC3 proteins, the tendency for HIV-1 to acquire G-to-A mutations was evident (Refs Reference Janini83, Reference Vartanian84, Reference Vartanian85) – a phenomenon analogous to the original observation of hypermutation in spleen necrosis virus (Ref. Reference Pathak and Temin86). These early reports generally noted a preferential sequence context for hypermutation of 5′-GA-3′ and, to a lesser extent, 5′-GG-3′, suggesting that, given what is now known, multiple APOBEC3 proteins account for these early descriptions of hypermutation (Ref. Reference Liddament79).

While more-recent studies at times note stronger preferences for one context or another under a given condition (Refs Reference Kieffer87, Reference Pace88, Reference Piantadosi89), a theme uniting recent literature on the in vivo hypermutation of HIV-1 is that it frequently occurs in both major dinucleotide contexts. One study indicates, for example, that hypermutated sequences separate roughly into thirds: sequences with hypermutation predominantly in a 5′-GA-3′ context, sequences with hypermutation predominantly in a 5′-GG-3′ context, and sequences with hypermutation occurring in roughly equal proportions in each context (Ref. Reference Land90). Another reports the occurrence of 5′-GG-3′ context mutations in approximately 95% of hypermutated sequences versus the occurrence of 5′-GA-3′ context mutations in approximately 60% of hypermutated sequences (Ref. Reference Gandhi91). Furthermore, mutations in both contexts frequently occur in the same cloned sequence. For example, 5′-GG-3′ hypermutations occur within 95% of predominantly 5′-GA-3′ context clones (Ref. Reference Gandhi91), while another data set shows a very slight preference for hypermutation in a 5′-GA-3′ context with a strong 5′-GG-3′ component (Ref. Reference Ulenga92).

It is reasonable to conclude, therefore, that multiple APOBEC3 proteins are active in vivo and that at least one of these is A3G with its strong 5′-GG-3′ context preference. The identities of the APOBEC3 proteins responsible for the reported 5′-GA-3′ hypermutations, however, remain unknown. While A3F is certainly a likely candidate based on its expression, Vif sensitivity and restriction profiles (see below; Table 1), it is entirely possible that other APOBEC3 proteins account for at least a portion of this hypermutation. A more careful elucidation of the expression and restriction capacities of APOBEC3 proteins in vivo as well as the interactions of these proteins with HIV-1 in vitro will be required to define the true nature of the in vivo APOBEC3 repertoire, which may itself vary from person to person (e.g. Ref. Reference Kidd93).

Regardless of which APOBEC3 proteins are actively hypermutating HIV-1, it is clear that their overall effect is significant. Estimates of the frequency of hypermutated sequences in vivo range from approximately 7% to 20% with wide variation in the proportions seen in any one individual (Refs Reference Kieffer87, Reference Pace88, Reference Piantadosi89, Reference Land90, Reference Gandhi91). In fact, these figures almost certainly underestimate the total proportion of proviral sequences mutated by APOBEC3 proteins. Since a sequence must typically reach a given threshold of G-to-A mutations to be defined as ‘hypermutated’, methods of quantifying hypermutants fail to account for sequences that have been mutated by APOBEC3 proteins at a lower level. Furthermore, extremely hypermutated sequences may fail to replicate and amplify by PCR while still potentially contributing to the overall mutational load of a viral population through recombination (Ref. Reference Mulder, Harari and Simon94). As briefly discussed above, the deamination of cDNA cytidines may also create a substrate for the downstream degradation of the viral genome, which would further diminish the apparent proportion of viral genomes affected by APOBEC3 proteins.

APOBEC3 proteins may contribute to viral adaptation

Several authors have previously discussed the theoretical possibility that drug-resistance mutations could be attributable to APOBEC3-mediated hypermutation, citing the frequent occurrence of drug-resistance mutations in sequence contexts consistent with the action of APOBEC3 proteins (Refs Reference Berkhout and de Ronde95, Reference Haché, Mansky and Harris96, Reference Pillai, Wong and Barbour97). Combined with the common occurrence of hypermutation in vivo, it would seem that a role for APOBEC3 proteins in aiding the evolution of HIV-1 drug resistance is likely. Nevertheless, the actual data in the literature on this topic are still limited. For example, one study demonstrates that, in the presence of a vif allele defective for the neutralisation of A3G, HIV-1 can readily acquire a particular drug-resistance mutation by recombination with a lethally hypermutated genome residing in the same cell (Ref. Reference Mulder, Harari and Simon94). Computer modelling to assess the likelihood of drug-resistance mutation acquisition from A3G action further suggests that HIV-1 does acquire drug-resistance mutations from A3G activity at a low rate (Ref. Reference Jern98). Also of interest, a recent paper on HIV-1 sequence evolution proposes that APOBEC3 proteins may be involved in the acquisition of immune-escape mutations early in HIV-1 infection (Ref. Reference Wood99).

While the consensus of the data to date is that APOBEC3 proteins are likely to contribute to HIV-1 evolution, many questions remain unanswered. For example, how often does this actually occur in vivo, and what is the relative contribution of APOBEC3 proteins versus other sources of mutation such as reverse transcriptase? Which APOBEC3 proteins are primarily responsible? To what extent does selective pressure contribute to the amplification of mutations caused by APOBEC3 proteins? Much work remains to be done in determining the impact of APOBEC3 proteins on the adaptability of HIV-1. The field would particularly benefit from studies that expand beyond A3G to consider the APOBEC3 proteins likely to cause 5′-GA-3′ context mutations.

Correlation of hypermutation levels, APOBEC3 expression and clinical indicators

While hypermutation clearly occurs in vivo, the data concerning its correlation with clinical parameters of infection are less clear; a summary of pertinent research on this topic is provided in Table 2. For example, one study shows a significant correlation between hypermutation in both 5′-GA-3′ and 5′-GG-3′ contexts and reduced viral load (Ref. Reference Pace88), consistent with the higher levels of hypermutation found in another group of patients with low viral loads (Ref. Reference Vázquez-Pérez100). Strikingly, the authors of the first study conclude that the drop in viral load attributable to hypermutation is notably greater than that attributable to other known protective alleles such as CCR5Δ32. Corroborating this effect, 17 Kenyan women with significant hypermutation levels have a statistically significant increase in CD4⁺ cell counts (Ref. Reference Land90). Furthermore, a positive correlation is found between adenosine content and CD4⁺ cell counts throughout this entire cohort of over 200 Kenyan women when the total adenosine content of patient proviruses is taken as a surrogate for hypermutation.

Table 2.

Summary of linkage studies between clinical indicators and hypermutation, APOBEC3 expression or APOBEC3 polymorphism

Abbreviations: A3B, APOBEC3B; A3F, APOBEC3F; A3G, APOBEC3G; AIDS, acquired immune deficiency syndrome; HIV, human immunodeficiency virus.

By contrast, hypermutation does not significantly correlate with viral load in a cohort of Senegalese female sex workers, although levels of hypermutation do generally correlate with A3G mRNA expression (Ref. Reference Ulenga92). Additional evidence suggesting no correlation between hypermutation and improved clinical status has been found by comparing patients on highly active antiretroviral therapy (HAART) with elite suppressors (patients who maintain high CD4⁺ cell counts and undetectable viral loads) (Ref. Reference Gandhi91). In contrast with the Senegalese cohort, however, no correlation between A3G expression and the fraction of hypermutated sequences existed in these groups. A third cohort, consisting of 28 Kenyan women, also shows no correlation between hypermutation and viral load or CD4⁺ cell counts (Ref. Reference Piantadosi89).

Others have more directly assessed the impact of specific APOBEC3 proteins by correlating their mRNA expression levels with clinical indicators; results of these studies are also summarised in Table 2. Using this approach, strong negative correlations are found between A3G mRNA levels and viraemia, concordant with positive correlations between A3G mRNA levels and CD4⁺ cell counts (Refs Reference Vázquez-Pérez100, Reference Jin101), although one study does not reproduce these associations (Ref. Reference Cho102). Nevertheless, other data indicate that patients with a low viral set point have significantly higher levels of A3G mRNA both preinfection and postinfection than patients with a high viral set point; analysis for A3F yields the same correlation with postinfection A3F mRNA levels but not preinfection mRNA levels (Ref. Reference Ulenga103). Furthermore, the peripheral blood mononuclear cells of HIV-exposed seronegative patients show higher A3G expression and decreased susceptibility to ex vivo infection (Ref. Reference Biasin104), and HIV-exposed seronegative individuals have higher A3G expression levels than healthy controls (Ref. Reference Vázquez-Pérez100).

Thus, the preponderance of evidence indicates a correlation between higher A3G expression and improved clinical status. This correlation may be less apparent or even absent, however, when hypermutation is taken as a surrogate for APOBEC3 action. Failing to find a correlation when substituting hypermutation for APOBEC3 function may not be that surprising, though, given that reducing the probability of successful reverse transcription and integration appears to be an important part of the mechanism by which APOBEC3 proteins restrict HIV-1. Interpretations of data correlating hypermutation with clinical indicators are somewhat confounded because it is not clear what has occurred when no correlation is found. If APOBEC3 proteins are highly effective, the viral genome may simply not persist long enough to integrate. Furthermore, it is possible that some APOBEC3 repertoires and/or individual APOBEC3 alleles may be protective, leading some patients to be selectively excluded from studies due to a lack of apparent HIV-1 infection. Regardless, the positive data in these studies, particularly those focusing on APOBEC3 expression, do indicate promise for therapies aimed at the APOBEC3–Vif axis. The field would benefit from further research correlating clinical indicators with both hypermutation and the expression of more APOBEC3 family members in larger patient cohorts. Individuals known to be relatively resistant to HIV-1 should also be closely examined for APOBEC3 sequence and copy number variations.

APOBEC3 polymorphisms may be important for restriction activity against HIV-1

Relatively little is known about the functional relevance of human APOBEC3 polymorphisms, but several attempts have been made to correlate clinical status with APOBEC3 variations, as summarised in Table 2. One survey of the APOBEC3G alleles in a French cohort divided into rapid and slow progressors reports 29 polymorphisms, two of which affect the amino acid sequence, but none of these correlates with the rate at which these groups progress to disease (Ref. Reference Do105). Others find that the H186R allele identified in this cohort does, however, correlate with CD4⁺ cell depletion and rapid progression to conditions defining AIDS (acquired immune deficiency syndrome) in African American patients (Ref. Reference An106), although there is no apparent association with risk of transmission. Another allelic variant, C40693T in intron 4 of APOBEC3G, is associated with an increased risk of infection in a cohort of Caucasian patients (Ref. Reference Valcke107).

While these associations are suggestive, their functional basis is far from clear. For example, there is no apparent difference between the wild-type and H186R alleles of APOBEC3G in an in vitro single-cycle infectivity assay (Ref. Reference Do105), and no experimental data are available for the C40693T variant. As the authors of the latter study rightly point out, however, the close association of APOBEC3G with other APOBEC3 genes on chromosome 22 raises the distinct possibility that a given association may actually reflect linkage with an undetected variation elsewhere in the locus rather than a direct effect of the identified polymorphism (Ref. Reference Valcke107).

In the case of APOBEC3H, activity against HIV-1 is strongly influenced by polymorphisms. Four major haplotypes of human APOBEC3H exist, but only one of these appears to be expressed stably and able to inhibit HIV-1 (Refs Reference OhAinle73, Reference Harari74, Reference Tan75). This active variant, APOBEC3H haplotype II, is present in roughly 50% of individuals of African descent, 10–18% of individuals of European descent and 3–4% of individuals of Asian descent (Ref. Reference OhAinle73). Another common APOBEC3 polymorphism is an APOBEC3B deletion, which occurs in approximately 20% of the world's population (Ref. Reference Kidd93). In contrast with APOBEC3H haplotype II, however, deletion of APOBEC3B is relatively rare in African and European populations but common in East Asian (36.9%), Amerindian (57.7%) and Oceanic (92.9%) populations.

Whether these or other unidentified allelic variants are of significance for susceptibility to HIV-1 infection or disease progression is unknown, although at least one recent report has found a statistically significant increased risk of HIV-1 infection and progression in individuals with homozygous APOBEC3B deletions (Ref. Reference An108). It is important to point out, however, that most APOBEC3 proteins have evolved under positive selection, implying that the challenges posed by ancient and presumably divergent pathogens have resulted in the rapid evolution of these restriction factors (Refs Reference OhAinle109, Reference Sawyer, Emerman and Malik110). While this selective pressure may not have been applied by HIV-1 given its relatively recent arrival in the human population, selection for APOBEC3 proteins better able to restrict other pathogens may have inadvertently selected cross-reactivity for the restriction of HIV-1. Thus, it is reasonable to predict that, despite the relative dearth of data presently available, variation at the genetic level will account for some effect on HIV-1 in vivo. Furthermore, it is possible that ongoing selection by the HIV-1 epidemic itself may result in the expansion of protective APOBEC3 alleles not yet identified.

Studies of APOBEC3 effects in primary cell types

CD4⁺ T cells are the primary reservoir of HIV-1 infection in vivo, but relatively little is known about the APOBEC3 repertoire found in these cells. One report suggests that A3G forms the basis of a largely uncharacterised, deaminase-independent target-cell restriction of incoming HIV-1 virions in resting CD4⁺ T cells (Ref. Reference Chiu111), an effect enhanced by treatment with interferon α (IFN-α/IFNA1), which upregulates A3G expression (Ref. Reference Chen112). The reproducibility of this phenomenon, however, has been questioned (Refs Reference Santoni de Sio and Trono113, Reference Kamata, Nagaoka and Chen114). Regardless, evidence indicates that among activated CD4⁺ T cells, the Th1 subset of helper T cells expresses greater levels of A3G and A3F, which are further enhanced by the autocrine action of IFN-γ (IFNG) (Ref. Reference Vetter115). Greater expression, in turn, results in greater encapsidation of APOBEC3 proteins in HIV-1 produced from these cells and consequently lower viral infectivity than in the Th2 subset of helper T cells, even for Vif-proficient viruses. In general, most APOBEC3 family proteins with the exception of A3A and, to a lesser extent, A3B appear to be expressed in primary CD4⁺ T cells; among T cell subsets, however, A3G is expressed at somewhat higher levels than A3F (Ref. Reference Koning116).

Studies with primary myeloid lineage cells have also noted a prominent effect of IFN treatment on the ability of APOBEC3 proteins to restrict HIV-1, linking higher expression of A3G and A3A to the relative resistance of monocytes versus differentiated macrophages to infection (Ref. Reference Peng117). Similarly, A3G and A3F are responsible for preventing HIV-1 from infecting immature dendritic cells, an effect that augments as dendritic cells mature and increase their A3G expression (Ref. Reference Pion118). Expression of IFN-α and, to a lesser extent, IFN-γ have also been linked to increased APOBEC3 levels and increased resistance to HIV-1 infection; this effect is so dramatic that the upregulation of APOBEC3 proteins may effectively overwhelm normal Vif function and enable restriction of wild-type HIV-1 in macrophages (Refs Reference Peng117, Reference Peng119). Of note, IFN-α clearly upregulates all APOBEC3 proteins in macrophages and dendritic cells with the possible exceptions of A3B and A3C, indicating that these are part of the normal IFN response (Ref. Reference Koning116). This upregulation is particularly astounding for A3A (Refs Reference Koning116, Reference Peng117). Given the common expression of many APOBEC3 proteins in primary cell types and their IFN inducibility, it is tempting to speculate that the combination of different APOBEC3 proteins may yield emergent properties not readily apparent in the majority of in vitro studies to date, which almost exclusively focus on the effects of single APOBEC3 proteins.

Indirectly increasing encapsidation by preventing the degradation of APOBEC3 proteins

A pharmaceutical approach

Prevention of the degradation of APOBEC3 proteins would be most simply accomplished by blocking the interaction of APOBEC3 proteins with Vif (Fig. 2a and 2b). To that end, a great deal of effort has been expended attempting to elucidate the important interacting regions on Vif and on APOBEC3 proteins (Refs Reference Russell and Pathak120, Reference Simon121, Reference Zhang122, Reference He123, Reference Tian124, Reference Yamashita125, Reference Huthoff126, Reference Huthoff and Malim127, Reference Russell128, Reference Zhang129, Reference Pery130, Reference Chen131, Reference Dang132 and others cited elsewhere in this article). While a thorough discussion of the many details involved in these investigations is beyond the scope of this review, a schematic summary of important regions on Vif and on A3G with particular attention to their putative interaction domains is provided in Figure 3, and biologically relevant regions of A3G are highlighted on a recently published full-length structural model of this restriction factor (Ref. Reference Harjes133) in Figure 4.

Figure 3.

Important domains in Vif and APOBEC3G. While many papers have identified residues critical to Vif function, only putative continuous interaction domains in Vif (a) and APOBEC3G (A3G) (b) are depicted here for the sake of clarity. Exceptions are made for two important lysines in Vif (K22 and K26) as well as the C-terminal zinc-binding domain of Vif. Vif and A3G are internally to scale, but Vif is depicted at twice its actual size relative to A3G. HIV-1 Vif residues shown are those found in the HXB2 reference sequence (accession number K03455). A3G corresponds to reference sequence NP_068594. Note that K26 and ²³SLV²⁵ of Vif are required for neutralisation of A3G and A3F, respectively, but these residues may not mediate direct interaction with either APOBEC3 protein. Similarly, one group finds that while A3G D128 is important in permitting the degradation of A3G, it does not mediate direct binding to Vif. See main text for references and discussion. HIV, human immunodeficiency virus.

Figure 4.

Important domains of APOBEC3G on a full-length model structure. Important domains of APOBEC3G (A3G) identified by genetic analysis are highlighted on a full-length model structure of A3G (Ref. Reference Harjes133) with colouration carried out in Pymol. See main text for references and discussion. Side chains are shown for key regions, and the right panel represents a rotation of 180º about the y-axis. Red spheres, zinc; green side chains (¹²⁴YYFW¹²⁷), encapsidation determinant (overlaps 126–127 start of Vif-binding region ¹²⁶FWDPDYQ¹³²); red side chain (D128), a key residue for Vif sensitivity; blue side chains (¹²⁹PDYQ¹³²), the remainder of the putative Vif-binding site; yellow side chains (H65 & E67 and H257 & E259), the HxE of the pseudocatalytic and catalytic N- and C-terminal domains, respectively; orange side chains (⁹⁶PC⁹⁷ & C100 and ²⁸⁷PC²⁸⁸ & C291), the PCxxC of the active and pseudoactive N- and C-terminal domains, respectively.

In surveying this literature, it is worthwhile to note that several discontinuous surfaces on Vif have been implicated in the binding of specific APOBEC3 proteins (Fig. 3a). This indicates that the APOBEC3-interaction surfaces on HIV-1 Vif are structurally complex and that three-dimensional structures of Vif bound to APOBEC3 proteins will likely be required both to understand and to effectively target this interaction. In contrast, what is known of the Vif-interaction surfaces on APOBEC3 proteins indicates that, while such surfaces may be distinct on different APOBEC3 proteins (Refs Reference Zhang122, Reference Russell128), they are relatively continuous (Fig. 3b). It may be possible, therefore, to protect APOBEC3 proteins by simply designing a molecule that binds the host protein and shields it from interaction with Vif (Fig. 2b), a strategy that would likely hamper the ability of HIV-1 to develop resistance to this putative therapy (Ref. Reference Harris and Liddament2). Conversely, it may be possible to inhibit Vif interaction with either APOBEC3 proteins or with the proteasomal machinery using a molecule that binds Vif itself (Fig. 2a).

Although a small-molecule inhibitor of the APOBEC3–Vif interaction is not yet clinically available, a significant recent development is the identification of RN-18, a small-molecule antagonist of HIV-1 Vif (Ref. Reference Nathans134). Interestingly, RN-18 decreases Vif levels and thereby increases A3G (and A3F and A3C but not A3B) levels when these viral and cellular proteins are present together. This small molecule is also able to inhibit the growth of Vif-proficient HIV-1 in cultured nonpermissive cells. Thus, while further research is necessary to understand the mechanism by which this and other putative inhibitors function, RN-18 has provided proof of principle that the APOBEC3–Vif axis can be targeted by small molecules.

Another promising development in the APOBEC3 field is the growing body of structural information on the deaminase domain of A3G (Refs Reference Harjes133, Reference Chen135, Reference Holden136, Reference Furukawa137) and related proteins such as APOBEC2 (Ref. Reference Prochnow138). As progress continues toward the expansion of our structural knowledge of A3G and other APOBEC3 proteins (reviewed recently by Ref. Reference Bransteitter, Prochnow and Chen139), it may one day be possible to visualise the APOBEC3–Vif interaction structurally and thereby rationally design interaction inhibitors. Half of this putative interaction surface is depicted in the A3G full-length model presented in Figure 4. In the meantime, the field would benefit from further genetic study of the APOBEC3–Vif interaction, both for the sake of reproducing current results and for identifying mutants that may further our mechanistic understanding of the APOBEC3–Vif interaction in anticipation of complementary structural data. For more detail on such efforts, see ‘Further reading, resources and contacts’ below.

Directly increasing APOBEC3 encapsidation