Controlling lentiviruses: Single amino acid changes can determine specificity

The human immunodeficiency virus (HIV) is a member of the lentivirus family of retroviruses that we share with other primates. The known strains of primate lentiviruses fall into six major lineages identified from >25 species of African primates (1), and it is clear from phylogenetic analysis of lentiviral sequences that their evolutionary history involves a number of cross-species transmissions. Indeed, HIV-1 is a close relative of a virus found in chimpanzees (simian immunodeficiency virus from chimpanzees, or SIVcpz), and HIV-2 is a relative of a virus found in sooty mangabeys (SIVsmm). However, it is not clear whether other members of the primate lentivirus family could spread to humans under the right transmission conditions. Thus, it is important to understand the potential barriers that govern cross-species transmissions from animals to human. Two articles in this issue of PNAS (2, 3) and also an article in press at PNAS (4) indicate that a single amino acid position in a host antiviral enzyme called Apobec3G (5) may be one such barrier that protects humans from primate lentiviruses found in the widely distributed family of AGMs (Chlorocebus aethiops) (Fig. 1).

Fig. 1.

Host-specific interactions of Apobec3G with viral Vifs. Human Apobec3G encodes aspartate (D) at amino acid 128 and functionally interacts with HIV-1 Vif but not with SIVagm Vif, whereas AGM Apobec3G encodes a lysine (K) at amino acid 128 and functionally interacts with SIVagm Vif but not with HIV-1 Vif. The functional interactions can be reversed by mutating human Apobec3G to K at amino acid 128 and by mutating AGM Apobec3G to D at amino acid 128. Interactions are denoted by + or -. A positive interaction indicates that Apobec3G will be sensitive to Vif, and a negative interaction indicates that it will be resistant.

Innate cellular antiviral responses limit the damage that viruses inflict on their hosts. However, such defensive measures create strong selective pressures leading to the evolution of viral countermeasures. The study of the viral strategies to evade host antiviral responses has often led to better understanding of underlying cellular processes, for example, control of IFN responses, antigen processing, antiapoptotic responses, and gene silencing. A more recently described example of an innate antiviral response overcome by the acquisition of a specific viral countermeasure is the Apobec3G protein. Apobec3G is a cellular protein that is incorporated into the virion of retroviruses and acts during the process of reverse transcription (copying viral RNA into DNA) to direct the deamination of cytidine to uridine on the minus strand of viral DNA. This results in massive mutation of guanidine to adenines on the coding strand (6-9) and/or degradation of the viral genome (10). Consequently, lentiviruses (with one known exception) have acquired a gene called vif, whose gene product prevents the incorporation and subsequent antiviral activity of Apobec3G by binding and targeting it down a proteasome-dependent degradation pathway (11-13).

Apobec3G belongs to a superfamily of nucleic acid deaminases that includes Apobec1, activation induced deaminase (AID), and Escherichia coli cytidine deaminase (ECCDA). Although the substrates for these deaminases vary, the amino acid sequence and catalytic mechanism for cytidine deamination is largely conserved across the family. The apobec3 locus shows evidence of relatively recent expansion and tandem duplication resulting in five genes and two probable pseudogenes (Apobec3A-Apobec3G) (14) with distinct patterns of expression, although their normal functions are unknown (15).

Previous studies suggested that the Vif protein of various primate lentiviruses typically functioned only in the host species from which the virus was derived (16), and the species-specific exclusion of Apobec3G in virions by Vif (17) accounted for this host specificity. That is, human Apobec3G was inhibited by Vif from HIV-1 but not by Vif from the SIV of AGMs (SIVagm), whereas AGM apobec3G was inhibited by Vif from SIVagm but not by Vif of HIV-1 (Fig. 1). This finding raised the possibility that each lentivirus has a Vif protein suitably evolved for evading the innate antiretroviral defense strategy by the given Apobec3G of its usual host.

Humans encode an aspartate at amino acid 128 (D128) in Apobec3G, whereas AGM encodes a lysine at that position (K128) (Fig. 1). Remarkably, swapping the two amino acids reverses the species specificity: human Apobec3G (D128K) is no longer inhibited by HIV Vif but is inhibited by SIVagm Vif, and vice versa (Fig. 1). Therefore, WT HIV has the same phenotype as HIV lacking the vif gene if it infects a human cell harboring the D128K mutation in Apobec3G. These results suggest that a single amino acid is the sole determinant of the differential interaction of the two Vif proteins with the Apobec3G proteins of their respective hosts.

An obvious difference between the amino acids positioned at 128 in AGM and humans is a net charge of two. By making several different mutations at this position, two groups concluded that human Apobec3G must be neutral or negatively charged at amino acid 128 to functionally interact with HIV Vif, whereas AGM Apobec3G must be neutral or positively charged at amino acid 128 to functionally interact with SIVagm Vif (2, 3). Based on homology modeling to E. coli cytidine deaminase, amino acid 128 lies in a loop between the catalytic domain and the pseudocatalytic domain (2). This predicted solvent-exposed loop would be free to participate in direct interactions with a binding partner. Additionally, the charge restrictions at position 128 suggest a putative electrostatic interaction with the binding surface of Vif. On the other hand, amino acid 128 of Apobec3G may not directly be involved in the physical binding to Vif because another group showed that a fragment of Apobec3G from amino acids 54 to 124 was sufficient for binding to Vif (18). Rather, the charge at amino acid 128 may prevent binding of some Vifs at a nearby interaction site; for example, changing amino acid 128 in human Apobec3G to a small, neutrally charged amino acid allows the protein to bind to both HIV-1 and SIVagm Vif. Alternatively, amino acid 128 may be involved in an effector function of Vif leading to Apobec3G degradation (4). The crystal structure of Apobec3G (or even better, cocrystals with Vif), together with the high sequence homology between the primates, will facilitate an understanding of the species-specific interaction between Apobec3G and Vif proteins and may allow for the design of inhibitors of the binding.

Because of the high mutation rate of RNA viruses, the large number of rounds of replication, and the immune selection on viral proteins, viral genes evolve much more rapidly than the antiviral host genes. Among the different Vif proteins from primate lentiviruses, the amount of amino acid conservation is <40% between members of different classes of primate lentiviruses. It will be interesting to determine the minimum number of steps it takes to change the specificity of Vif from one virus into the Vif of another. However, the ORF that encodes Vif is constrained by partial overlap with two other ORFs, so its evolutionary course is unlikely to be simple. Although on a relatively short evolutionary time frame the Vif proteins of the primate lentiviruses may have evolved to counteract Apobec3G of the host species, on longer evolutionary scales, Apobec3G may be evolving away from the interaction with Vif. This would be unlikely in humans, whose infection by HIV is a relatively recent event, and possibly not in chimpanzees, where the infection rates are low. But it could well be the case in the genus Cercopithecus, which may harbor the greatest diversity of primate lentiviruses. It is furthermore possible that Apobec3G protects us against other, more ancient viral infections, and the signatures of positive evolution will be seen in the human apobec3G gene once more primate sequences are available.

Given the large reservoir of primate lentiviruses for which humans have potential contact (19), there remains the question of whether additional retroviruses could be transmitted to humans. Schröfelbauer et al. (2) argue that a SIV had to adapt to use amino acid D128 of chimpanzee Apocbec3G as a prerequisite for subsequent transmission to humans; however, this may not be the case. Rather, it is more likely that among the diverse strains of SIV there are Vif proteins that are already able to inactivate human Apobec3G. The origin of SIVcpz is illuminating in this regard. It is has been hypothesized that chimpanzees originally acquired SIVcpz through smaller monkeys on which they prey (20). Indeed, it appears that SIVcpz is a hybrid virus whose ancestors may be similar at the one end to a virus that is found in red-capped mangabeys (Cercocebus torquatus) (SIVrcm) and at the other end to a virus that is found in the greater spot-nosed monkey (Cercopithecus nicitans) (SIVgsn) (21). Specifically, the Vif protein of SIVcpz is more likely to have been contributed from SIVrcm (Paul M. Sharp, personal communication). Because SIVrcm is able to replicate in primary human cells, its Vif protein it likely to already have the capability to inactivate human Apobec3G. Additionally, the Vif protein of SIV from macaques (SIVmac, or SIVsmm in its original form) interacts with both macaque and human apobec3G (16, 17), and there are even some strains of SIV isolated from a subspecies of African green monkeys that already replicate in primary human cells to some extent (22) and thus likely already encode a Vif protein that can inactivate human Apobec3G. Thus, we argue that the ability of an SIV to overcome the human antiviral defense, Apobec3G, is already present in the wild. To identify which strains of SIV are of particular concern, it would be useful to screen many different strains by using a functional screen for Vif action on human Apobec3G. This could be done with DNA samples collected even in the absence of the isolation of replication-competent viruses.

That said, Vif is not the only barrier to cross-species transmission. Although the viral receptors and coreceptors are fairly well conserved among the different primates, there are interesting cellular restrictions between viruses and different primate hosts early after virus entry that are as yet still largely uncharacterized (reviewed in ref. 23), and there is an additional cellular restriction at the late stage of viral budding that is the target of another viral protein, Vpu (24). Thus, viruses must simultaneously evolve around multiple host defenses, and this complex interplay between antiviral genes and viral countermeasures creates selective pressures that leave an imprint on the evolutionary pattern of primates and their parasites.