Ancient adaptive evolution of the primate antiviral DNA-editing enzyme APOBEC3G

Abstract

Host genomes have adopted several strategies to curb the proliferation of transposable elements and viruses. A recently discovered novel primate defense against retroviral infection involves a single-stranded DNA-editing enzyme, APOBEC3G, that causes hypermutation of HIV. The HIV-encoded virion infectivity factor (Vif) protein targets APOBEC3G for destruction, setting up a genetic conflict between the APOBEC3G and Vif genes. This kind of conflict leads to rapid fixation of mutations that alter amino acids at the protein-protein interface, referred to as positive selection. We show that the APOBEC3G gene has been subject to strong positive selection throughout the history of primate evolution. Unexpectedly, this selection appears more ancient than, and is likely only partially caused by, modern lentiviruses. Furthermore, five additional APOBEC genes in the human genome appear to be engaged in similar genetic conflicts, displaying some of the highest signals for positive selection in the human genome. Despite being only recently discovered, editing of RNA and DNA may thus represent an ancient form of host defense in primate genomes.

Figure 1. The Primate APOBEC Family

(A) The human genome contains nine known members of the APOBEC family. AID and APOBEC1 are located approximately 900 kb apart on human Chromosome 12. The primate-specific APOBEC3 cluster of six genes resides on human Chromosome 22, and likely arose through a series of gene duplication events (Jarmuz et al. 2002; Wedekind et al. 2003). The single APOBEC3-like gene found in mouse resides on Chromosome 15 (not shown), which is syntenic to human Chromosome 22 (Sheehy et al. 2002). There is EST evidence for both APOBEC3D and APOBEC3DE (see Materials and Methods), and we treat these as three separate transcripts in our analysis because currently there is no evidence for the relevant protein products. (B) All members of the APOBEC family contain an active site that encodes a zinc-dependent cytidine deaminase domain with the HXE, PCXXC signature (Mian et al. 1998), a linker peptide, and a pseudoactive domain (Navaratnam et al. 1998; Jarmuz et al. 2002). The active and pseudoactive domains are related by structure only, and likely originated from a gene duplication event followed by degeneration of the catalytic activity of the pseudoactive domain. Several members of the human APOBEC3 gene cluster (APOBEC3B, 3DE, 3F, and3G) have undergone an additional duplication/recombination event and now contain two each of the active and pseudoactive sites (Jarmuz et al. 2002; Wedekind et al. 2003), as does the single APOBEC3-like gene found in mouse. DOI:10.1371/journal.pbio.0020275.g001

Figure 2. APOBEC3G Has Been Under Positive Selection for at Least 33 Million Years

The ω values and actual numbers of non-synonymous and synonymous changes (R:S, included in parentheses) in APOBEC3G are indicated on the accepted primate phylogeny (Purvis 1995) that includes five hominids, five OWMs, and two NWMs. OWMs diverged from hominids about 23 million years ago, whereas NWMs diverged around 33 million years ago (Nei and Glazko 2002). ω values were calculated using the PAML package of programs using the free-ratio model that allows ω to vary along each branch. In some instances, zero synonymous substitutions lead to an apparent ω of infinity. HIV/SIV-infected species are indicated by asterisks. DOI:10.1371/journal.pbio.0020275.g002

Figure 3. Episodic Positive Selection on Different Regions of the APOBEC3G Gene

(A–C) Sliding window (300-bp window; 50-bp slide) analysis of Ka and Ks was performed on three representative pairs of primate APOBEC3G sequences, between two hominids (human–orangutan) (A), between two OWMs (crested macaque–baboon) (B), and between two NWMs (tamarin–woolly monkey) (C). Ka/Ks, Ka, and Ks are plotted against the length of the gene (with a schematic of protein domains along the x-axis) to illustrate that different domains of APOBEC3G have undergone positive selection, depending on which lineage is examined. The value for ω, indicated by Ka/Ks, is not shown for part of the crested macaque–baboon comparison (B), because Ks is zero in this region (see plot below). (D) A schematic of the domains of human APOBEC3G illustrates the N-terminal domain (aa 1–29), the two active sites (aa 30–120 and 215–311), and the pseudoactive sites (aa 162–214 and 348–384). Also illustrated is the Vif-interaction domain of APOBEC3G (aa 54–124) (Conticello et al. 2003) as well as the single amino acid residue responsible for species-specific sensitivity to Vif (aspartic acid 128; cross shape in linker 1) (Bogerd et al. 2004; Schrofelbauer et al. 2004). PAML (Yang 1997) was used to identify individual residues (codons) that have significant posterior probabilities of ω greater than 1.0 (see Materials and Methods). Those codons with posterior probabilities greater than 0.95 and greater than 0.99 are indicated by open and closed inverted triangles, respectively (listed in Figures S2 and S3). This represents only a subset of the residues that are likely to be under positive selection, highlighting those residues that have repeatedly undergone non-synonymous substitutions. For instance, residue 128 is not highlighted, as it has a posterior probability of only 0.55 because it has undergone only one fixed non-synonymous change (along the OWM lineage). Domains have been defined by protein sequence alignment to APOBEC1 (Jarmuz et al. 2002). The first pseudoactive domain is likely to include in its C-terminus a second duplication of the N-terminal domain, although this boundary cannot be resolved because of sequence divergence. DOI:10.1371/journal.pbio.0020275.g003

Figure 4. Selective Pressures on APOBEC1, APOBEC2, and APOBEC3E

Sliding window analysis (250-bp window; 50-bp slide) was performed on three APOBEC genes. Although APOBEC1 demonstrates purifying selection when the whole gene is analyzed (Table 1), the sliding window analysis of a human–orangutan comparison reveals a window (aa 1–100) in the first active site (dark gray bar), which shows evidence of positive selection (p < 0.01). Sliding window analysis of APOBEC2, which is also evolving under purifying selection (Table 1), does not show any windows where ω is greater than one. APOBEC3E, which gives the strongest signal for positive selection (Table 1), has ω greater than one for almost all windows. (Note that ω is not plotted where Ks = 0). DOI:10.1371/journal.pbio.0020275.g004