Editing at the Crossroad of Innate and Adaptive Immunity

See allHide authors and affiliations

Science  18 Feb 2005:
Vol. 307, Issue 5712, pp. 1061-1065
DOI: 10.1126/science.1105964

This article has a correction. Please see:


Genetic information can be altered through the enzymatic modification of nucleotide sequences. This process, known as editing, was originally identified in the mitochondrial RNA of trypanosomes and later found to condition events as diverse as neurotransmission and lipid metabolism in mammals. Recent evidence reveals that editing enzymes may fulfill one of their most essential roles in the defense against infectious agents: first, as the mediators of antibody diversification, a step crucial for building adaptive immunity, and second, as potent intracellular poisons for the replication of viruses. Exciting questions are raised, which take us to the depth of the intimate relations between vertebrates and the microbial underworld.

Compared with alternative splicing, a process in which fragments of RNA transcripts are differentially cut and pasted, editing seems a rather subtle fine-tuning of the coding capacity of genomes. Yet the power of this process for genetic diversification is overwhelmingly illustrated by its crucial role in allowing B lymphocytes to generate high-affinity antibodies against an almost infinite variety of antigens. In mammals, there are two main classes of editing enzymes, both of which deaminate encoded nucleotides: one generates inosine (I) from adenine (A), the other uridine (U) from cytidine (C). Both classes of proteins are characterized by the presence of closely homologous zinc-coordinating catalytic domains, and phylogenetic analyses indicate that this now diversified family of proteins evolved from an ancestral cytosine deaminase involved in pyrimidine metabolism (1). Activation-induced deaminase (AID) likely was the first editing deaminase to appear in vertebrates, as this subphylum emerged from the chordata and started to develop an adaptive immunity (2). The pattern of chromosomal distribution and the degree of homology of the 11 AID-related cytidine deaminase genes found so far in humans and other primates indicate that the other members of the family arose through a series of gene transpositions and duplications (36) (Fig. 1). It is interesting that this process appears to have undergone a sharp acceleration in primates, because the genome of rodents, for instance, contains only four AID-related cytidine deaminase genes.

Fig. 1.

The primate cytidine deaminases family. Schematic representation of the 12 known cytidine deaminases, with chromosomal location of corresponding gene indicated on left. All family members share at least one zinc-dependent cytidine deaminase active site (AS). APOBEC3B, -3G, and -3F are the result of a further duplication; APOBEC3DE is encoded by a predicted splicing variant. Also indicated is the domain implicated in HIV-1 Vif binding (red stars), including the critical Asp128 residue (red diamond) responsible for its species specificity.

APOBEC1 [apolipoprotein B (apoB)–editing catalytic subunit 1] was the first identified mammalian cytidine deaminase (7). Like its paralogs, APOBEC1 exhibits a narrowly tissue-specific distribution. Expressed in enterocytes, it is responsible for introducing a premature nonsense codon in the apoB mRNA by changing cytidine 6666 to uridine. The resulting truncated form of apoB and its liver-specific full-length counterpart play distinct roles in lipid metabolism. RNA editing by APOBEC1 critically depends on its homodimerization and on its recruitment by ACF (APOBEC1 complementation factor), an RNA-binding protein that recognizes a mooring sequence in the apoB mRNA. An “editosome” is thereby formed, the activity of which is modulated by auxiliary components (8).

Cytidine Deamination and the Emergence of Adaptive Immunity

When mature B lymphocytes migrate to secondary lymphoid organs, such as spleen and lymph nodes, and encounter antigens, they become activated and rapidly proliferate. At the same time, their genome undergoes two kinds of alterations at the immunoglobulin gene locus: class switch recombination (CSR) and somatic hypermutation (SHM). CSR replaces the constant region of the immunoglobulin heavy chain gene (CH), which triggers a switch of immunoglobulin isotype from the original IgM to IgG, IgE, or IgA. SHM introduces point mutations in the variable (V), antigen-binding region of both heavy and light chain genes and thus creates a repertoire of B cells from which producers of high-affinity antibodies are then selected (9). Although apparently dissimilar, these two genetic events both depend on activation-induced deaminase. In birds, diversification of the antigen-binding site of antibodies is accomplished by another mechanism, gene conversion (GC), but it too requires AID. Even though it was initially suggested that AID might act by editing the RNA encoding for some DNA-modifying enzyme, several lines of recent evidence support a model whereby CSR, SHM, and GC all result from the initial conversion by AID of cytidine to uridine in the immunoglobulin locus DNA in a transcription-dependent and strand-biased fashion (10, 11). It is noteworthy that the enzyme uracil DNA glycosylase (UNG), a DNA repair enzyme that removes uracil from single- and double-stranded DNA, is also required for all three AID-dependent processes. This has been interpreted by most as supporting the DNA-editing model of antibody diversification. However, the recent demonstration that catalytically inactive ung2 mutants are functional in tissue culture–based systems of CSR adds a level of complexity that precludes definitive conclusions regarding the mechanism of AID action (12).

A Viral Shield Reveals a Cellular Weapon

After the human immunodeficiency virus (HIV) was first isolated in the early 1980s, the sequencing of its genome rapidly delineated nine genes, the products of which were soon assigned a variety of structural, enzymatic, or regulatory functions. All but one, vif (virion infectivity factor, initially called sor), would largely elude the perspicacity of researchers for close to another 20 years. It did not take long to demonstrate that the ∼200–amino acid–long Vif protein, which accumulates in the cytoplasm of infected cells late in the viral life cycle, is important for virion infectivity (13), but more tedious efforts were required to reveal first that vif-mutated virions can enter cells normally but yield markedly reduced levels of proviral DNA (the integrated form of the reverse-transcribed RNA genome) and, second, that this defective phenotype is entirely conditioned by the cell releasing the virus, not by its next target cell (14, 15). ΔVif-permissive and ΔVif-restrictive cells were thus distinguished from each other, the latter, not surprisingly, was composed of the cells normally infected by HIV in vivo, namely, primary T lymphocytes and macrophages. An important breakthrough came with the demonstration that ΔVif restrictiveness is a dominant phenotype and, hence, reflects an intracellular antiviral activity specifically countered by Vif (16, 17). Finally, through DNA subtraction, an approach requiring much patience, a member of the human cytidine deaminase family, APOBEC3G, was identified as this intracellular restriction factor (18). This discovery immediately set up intense efforts to elucidate the mechanisms of APOBEC3G action against HIV and of its blockade by Vif. It ended up unveiling a far broader defense system than was initially suspected.

Cytidine Deamination as a Broad Line of Defense Against Exogenous Retroelements

APOBEC3G is a cytoplasmic protein, the product of one of the eight APOBEC3 genes found on human chromosome 22 (3) (Fig. 1). In the absence of Vif, it is packaged into HIV particles during assembly, apparently through formation of a complex with the RNA-recruiting nucleocapsid (NC) region of the Gag viral protein, an interaction possibly strengthened by the nonspecific binding of RNA to NC (1921). APOBEC3G subsequently associates with the viral reverse-transcription complex, where it deaminates cytidine residues to uridine in the nascent minus-strand viral DNA (Fig. 2, upper left). These dU-rich transcripts are then either degraded or serve as templates for the synthesis of plus-strand DNA, thus yielding proviruses that are largely nonfunctional due to G-to-A hypermutation (2225). Vif counters this antiviral defense by connecting APOBEC3G with a poly-ubiquitinating complex comprising elongin B/C, Cul5, and Rbx1 (26). This prevents the virion incorporation of the deaminase and triggers its proteasomal degradation, which allows the production of HIV particles that can fully express their infectious potential.

Fig. 2.

One family of proteins, three antiviral mechanisms. (Top left) When expressed in HIV producer cells in the absence of Vif, huAPOBEC3F, -3G, or -3B is packaged in outgoing virions. In target cells, APOBEC3F, -3G, and, to a lesser extent, -3B deaminate cytidines to uridines in the growing minus-strand DNA reverse-transcribed from the viral genomic RNA. Many of these dU-rich reverse transcripts are degraded (52). The others serve as templates for the synthesis of G-to-A hypermutated plus-strand DNA. The resulting proviruses are nonfunctional. (Top right) Rat APOBEC1, possibly packaged in the outgoing virions, can also deaminate cytidines in the HIV viral genomic RNA. This leads to G-to-A changes in the minus-strand DNA and, hence, to C-to-T mutations in the plus-strand DNA. The resulting hypermutated proviral DNAs can integrate but are defective. (Bottom) After infection of APOBEC3G- or APOBEC3F-expressing cells, the partially double-stranded DNA HBV genome is repaired by cellular enzymes, and the various viral RNAs are transcribed and translated normally. However, there is a defect either in the assembly of capsids containing pregenomic RNA or in the stability of the viral reverse transcription complex. The end result is a sharp drop in the production of HBV DNA-containing virions. [Illustration: Katharine Sutliff/Science]

The antiretroviral spectrum of human APOBEC3G extends well beyond HIV, because it can also block other lentiviruses, such as the simian immunodeficiency virus (SIV) and equine infectious anemia virus (EIAV), as well as the gammaretrovirus murine leukemia virus (MLV) (23, 25) (Table 1). APOBEC-mediated editing thus constitutes a barrier to the cross-species transmission of several retroviral pathogens.

Table 1.

Genetic and functional characteristics of human and rodent cytidine deaminases. Extensive comparisons between primate and human genomes reveal that AID, APOBEC2, and APOBEC3A have been submitted to purifying rather than positive selection. Reported patterns of expression; antiviral activity against HIV, HBV, and MLV; and sensitivity to HIV-1 Vif action are indicated for each member of the family. ND, not determined.

Cytidine deaminasePositive selectionReported expressionAntiviral activityVif sensitivity
    APOBEC1 + Small intestine, spleen - - ND ND
    APOBEC2 - Heart and muscles ND ND ND ND
    AID - Activated B cells - - ND ND
    APOBEC3A - Keratinocytes - - ND ND
    APOBEC3B + PBLs, many tumor cell lines ± - ND -
    APOBEC3C + PBLs, hear, spleen, testes, ovary, prostate, thymus... many tumor cell lines - - ND ND
    APOBEC3D + None found so far ND ND ND ND
    APOBEC3E + Pseudogene? ND ND ND ND
    APOBEC3DE + None found so far ND ND ND ND
    APOBEC3F + PBLs, macrophages, spleen, testes, ovary + - + ±
    APOBEC3G + PBLs, macrophages, spleen, lung, testes, ovary + + + +
    APOBEC3H ND Lymphoid tissue, placenta ND ND ND ND
    muAPOBEC1 ND Spleen> liver, brain, kidney, lungs, testes, oocytes, ovary, embryonic stem cells - ND ND ND
    muAPOBEC2 ND Heart - ND ND ND
    muAID ND Lymph nodes, oocytes, ovary > thymus, spleen - ND ND ND
    muAPOBEC3 ND Spleen, lungs > testes, PBLs? + - ND -

Additional members of the human APOBEC3 family are endowed with activity against HIV, for instance, APOBEC3F, which is largely coexpressed and perhaps coregulated, with APOBEC3G. The two enzymes, however, can be distinguished by their target sequence consensus: APOBEC3G favors the 5′-CC dinucleotide (underline marks the target), whereas APOBEC3F prefers 5′-TC (2730). Human APOBEC3B also exhibits moderate levels of activity against HIV, with an APOBEC3F-like consensus, but is resistant to the action of Vif (27). However, the immediate relevance of this finding is unclear, because this protein is at best weakly antiviral and is expressed in small amounts in the natural targets of HIV. Future drugs aimed at blocking the interaction between Vif and APOBEC3G and APOBEC3F may thus have the potential to constitute valid additions to current AIDS therapies.

The editing of retroviral reverse transcripts by APOBEC family members suggested that these enzymes might also act on other types of retroelements. Like retroviruses, hepatitis B virus (HBV) replicates through reverse transcription. However, although the retroviral RNA genome is copied into DNA, for the most part, once the virus enters target cells, HBV and other hepadnavirus particles contain a partially double-stranded DNA genome, synthesized from a pregenomic RNA by the viral reverse transcriptase within subviral core particles in the cytoplasm of virus producer cells. Recent experiments demonstrate that this process can be efficiently inhibited by human APOBEC3G and APOBEC3F (31, 32) (Fig. 2, bottom). However, the editing function of the cellular enzymes is dispensable for this effect, which rather reflects either an inhibition of HBV pregenomic RNA packaging or a destabilization of the viral reverse transcription complex, precluding HBV DNA accumulation. In the HepG2 hepatoma cell line, the same phenomenon prevails, but some APOBEC3G-mediated HBV DNA editing can also be detected (33).

These data raise at least three questions. First, do antiviral cytidine deaminases participate in the noncytopathic clearance of HBV that takes place in most patients acutely infected with this virus? APOBEC3 family members are expressed at only low levels in hepatocytes, but they could be induced by HBV infection, for instance, under the influence of cytokines. HBV replication is rapidly abolished in the liver of HBV transgenic mice treated with an interferon-inducing agent; this inhibition reflects primarily loss of capsids containing pregenomic RNA, reminiscent of the observed effect of APOBEC3G in tissue culture (31, 34). Also, recent data indicate that stimulation of the protein kinase C (PKC)/mitogen-activated protein kinase kinase (MEK)/extracellular signal–regulated kinase (ERK) pathway enhances APOBEC3G expression (35). A second question is: Would it be possible to induce the expression of antiviral cytidine deaminases in the liver of individuals chronically infected with HBV, either pharmacologically or by gene transfer, in order to eradicate the virus? Whereas the constitutive expression of wild-type forms of these proteins might carry the risk of introducing mutations in the cellular DNA (36), catalytically inactive variants fully conserving their antiviral potential would be devoid of such a side effect. Finally, does cytidine deamination explain the occasional emergence of G-to-A mutated HBV isolates, including HbeAg-negative strains and vaccine escape variants (37)?

Human antiviral cytidine deaminases can thus inhibit retroelements by at least two distinct mechanisms: single-stranded DNA editing (for retroviruses) and noncatalytic blockage of DNA accumulation (for hepadnaviruses). A third mode of antiviral action for this class of proteins was discovered by examining how rat APOBEC1 interferes with HIV replication: In that case, the editing process targeted not only the virus minus-strand DNA but also its genomic RNA (38) (Fig. 2, top right). This raises the possibility that APOBEC-mediated C-to-U editing might affect some of the innumerable viruses that replicate entirely through RNA. Accordingly, the spectrum of innate antiviral resistance conferred by cytidine deaminases might be of considerable breadth.

Mechanisms of Viral Escape

Pathogenic retroviruses seem a relatively new entry in the realm of factors that can cause human diseases. Nevertheless, the brutality of the AIDS epidemic cruelly illustrates the limits of our resistance to this class of pathogens. Evolution allows viruses to overcome host defenses, and retroviral inhibition by cytidine deamination has been no exception to this rule. The acquisition of a vif gene is the best-documented viral countermeasure against APOBEC proteins, but, interestingly, it is effective in an almost completely species-specific manner. Vif-defective HIV-1, for instance, is blocked by APOBEC3G from human, chimpanzee, rhesus macaque, and African green monkey (AGM) and by mouse APOBEC3. However, the Vif protein of HIV-1 can only counter human and chimpanzee APOBEC3G and is ineffective against the rhesus macaque, AGM, and mouse orthologs of the enzyme. Conversely, Vif from SIVAGM is active against AGM but not against human APOBEC3G (39). Remarkably, a single amino acid difference at position 128 of these two highly homologous proteins governs their virus-specific binding to Vif, hence, their inactivation by the viral protein (4042) (Fig. 1).

The vif-devoid MLV is efficiently blocked by human APOBEC3G but resists APOBEC3F and APOBEC3B through unknown mechanisms (Table 1). Unsurprisingly, MLV also escapes the action of murine APOBEC3, the antiviral cytidine deaminase found in its normal host (27), seemingly by avoiding the packaging of the enzyme during virion assembly (43). This raises questions about how other viruses might protect themselves against cytidine deaminases. For instance, does EIAV, which again does not contain a vif gene, proceed similarly in order to spread in its equine host, if one assumes that there is a horse antiviral cytidine deaminase? Alternatively, does this lentivirus hide a vif-like function in some other of its genes or does it avoid replicating in cell types that express the cellular enzyme?

What about human T cell leukemia virus (HTLV), another vif-less retrovirus that happens to infect the same CD4-positive T lymphocytes as HIV, that is, it targets cells expressing APOBEC3G and APOBEC3F? Has this virus evolved alternative strategies to counter the antiviral cytidine deaminases, for instance, by modulating their expression at the transcriptional level, or does it tolerate T cells, which it ultimately transforms, as dead ends for its replication? HTLV exhibits notoriously low levels of infectivity in tissue culture, but the basis of this phenotype likely depends on multiple factors.

Discovering the full antiviral spectrum of APOBEC family members may be a complicated task. Indeed, viruses that replicate in a given species will generally have evolved ways to escape inhibition by the editing cytidine deaminases expressed in this host. So far, the antiviral potential of this class of innate resistance factors was revealed by testing viruses that were either invalidated by specific mutations (as for ΔVif HIV) or confronted with noncognate APOBEC proteins (as for MLV and human APOBEC3G).

Endogenous Targets for APOBEC-Mediated Editing? Evolutionary Considerations

The nature of the physiological targets of APOBEC1 and AID suggests that at least some other members of the family edit cellular genes, whether at the RNA or DNA level. However, evolutionary data support a model whereby most APOBEC proteins play a major role at the interface between hosts and pathogens. The mouse genome encodes orthologs of AID, APOBEC1 and APOBEC2, but contains only one APOBEC3 gene, compared with eight in primates. This recent expansion of the APOBEC gene cluster was further modulated by repeated episodes of positive selection during primate evolution, as indicated by the accumulation of nonsynonymous nucleotide changes in several of these genes, particularly at positions coding for charged residues likely to participate in protein-protein interactions (5, 6). Although only human APOBEC3B, -3F, and -3G have so far been demonstrated to carry antiviral activity, a sequence comparison between the genomes of Homo sapiens and other primates indicates that APOBEC1, -3C, -3D, and -3E have also been subjected to this type of positive selection, hence, have likely participated, at least occasionally, in host defense (5). Only the heart- and muscle-restricted APOBEC2 and the keratinocyte-specific APOBEC3A seem to have escaped this process and thus may have cellular sequences as their exclusive targets.

Identifying the genetic conflicts that have shaped the APOBEC family is a task of great interest. The apparent antagonistic coevolution of these proteins with invading genetic elements started long before the appearance of modern lentiviruses (5, 6), which points to additional pressures and is consistent with the demonstrated effect of these antivirals on other types of retroelements. The presence of a single APOBEC3 protein in the mouse may reflect the development of an alternative, yet unidentified, form of innate antiviral immunity in this species. However, there is a striking evolutionary coincidence between the expansion of the APOBEC gene cluster and the abrupt drop in retrotransposon activity that took place in primates, compared with rodents (44).

Retrotransposons cumulatively account for about one-third of the human genome and are commonly classified as those that bear long terminal repeats (LTRs), also known as endogenous retroviruses, and those that do not, of which only L1 elements (LINE-1, long interspersed nucleotide element–1) are autonomous (by contrast with SINE elements and Alu repeats) (45). Retrotransposons actively participate in genome remodeling and, as such, may facilitate adaptation. However, they also represent genetic threats, because they can disrupt genes by insertional mutagenesis, promote translocations and other rearrangements by recombination, generate aberrant cellular transcripts from their promoter, and exert antisense effects. The mouse genome contains up to 3000 active L1 elements and shows evidence for ongoing LTR retrotransposition. By comparison, only 40 to 60 L1 elements are active in humans, and all human endogenous retroviruses (HERVs) appear lethally incapacitated. The presence of human-specific HERV proviruses and their absence in corresponding loci of other primates, nevertheless, indicate that a subset of these LTR retrotransposons has been active during the recent evolution of hominoids.

Although transcriptional silencing via cytosine methylation of CpG dinucleotides is a major mechanism of retrotransposon control, one can speculate that destabilization or hypermutation of their reverse transcripts by cytidine deamination also participates in this process. Particularly intriguing are the high levels of human APOBEC3G and APOBEC3F detected in testes and ovaries (3) and the extensive level of genome demethylation that prevails during the preimplantation phase of embryonic development, two settings in which expression of retroelements is thought to occur. APOBEC3G has been shown not to inhibit human L1 retrotransposition in a tissue culture–based assay (46), which is perhaps not too surprising considering that L1 reverse transcription occurs in the nucleus, whereas APOBEC3G is a cytoplasmic protein. In contrast, HERVs carry strong functional homologies with exogenous retroviruses and, as such, could be sensitive to APOBEC-mediated cytidine deamination. Short of the identification of active HERVs and of a tissue culture–based assay for measuring their retrotransposition ability, a careful comparison of the A/G content of endogenous retroviruses with that of related exogenous retroelements might give hints regarding their past exposure to cytidine deaminases. In strong support of a protective role for cytidine deaminases against genome destabilization by endogenous retroelements, recent experiments with murine intracisternal A-particles and MusD elements indicate that at least some murine endogenous retroviruses are sensitive to murine APOBEC3 and human APOBEC3G (47, 48).

When Viruses Utilize the Host Deaminases

Even though Vif efficiently counters the antiviral action of APOBEC3G and APOBEC3F, some level of G-to-A mutation can be detected when wild-type HIV is produced in the presence of high levels of the deaminases. This apparent flaw may actually serve the virus by increasing its genetic diversity, thus enabling escape from adaptive immune responses or antiviral drugs. Even though the error-prone reverse transcriptase is likely to be the major factor driving HIV diversification, G-to-A hypermutation has been documented in proviral DNA sequences isolated from peripheral blood lymphocytes of HIV-infected individuals. In many cases, these nucleotide changes were extensive and clearly resulted in complete virus inactivation. However, specific G-to-A changes, often in sequences fitting the APOBEC3F and, more rarely, the APOBEC3G consensus, have been shown to govern resistance to protease inhibitors used in AIDS therapies (49). Because it is slightly less sensitive than APOBEC3G to Vif action, APOBEC3F may be a more important factor for HIV hypermutation in vivo (28).

An even more convincing case for the hijacking of deaminating enzymes by viral pathogens is provided by adenosine deaminases. These enzymes seem to act exclusively on free adenosine or on RNA. ADARs (adenosine deaminases that act on RNA) can induce codon changes (inosine is read as guanosine, G, during translation); alter splicing; or modify the subcellular localization of their substrate (50). Such editing was first shown to account for isoforms of the glutamate and serotonin receptors in the central nervous system. However, many inosine-containing RNAs have been detected, in particular in the brain, where ADARs most likely exert broad regulatory influences. ADARs are also interferon-inducible proteins that can target a wide array of viral RNAs with a variety of consequences, often beneficial to the virus. For minus-strand RNA viruses such as measles, A-to-I hypermutation has been proposed to promote persistence. For the DNA polyomavirus, ADAR-mediated editing facilitates the sequential expression of early and late viral transcripts. Finally, for the agent of hepatitis D, an A-to-I change at a specific position of the antigenomic RNA is essential to remove an amber codon and to allow synthesis of the viral capsid.

Many More Questions for the Curious Mind

Although defining the range of antiviral activities of editing enzymes certainly represents a most exciting challenge, many other questions need to be addressed. The biochemistry of APOBEC-mediated retroviral editing is still, in many regards, a black box. Is an editosome at play, as for APOBEC1? If so, are its additional components of cellular or viral origin? A unique opportunity to characterize the determinants of this reaction may be provided by the possibility of its completion within purified HIV-1 virions (24). What is the noncatalytic mechanism that underlies the blockade of HBV by several APOBEC family members? What regulates the expression of these antiviral proteins, in particular, in organs targeted by APOBEC-sensitive viruses? Do polymorphisms in APOBEC genes shape interindividual susceptibilities to human pathogens (51)? What is the tridimensional structure of the APOBEC-Vif interface, and does it open a window for the design of pharmacological inhibitors that could complement current AIDS therapies? Answers to these and other questions will take us to new frontiers in understanding the fine balance between vertebrates and their genetic invaders.

References and Notes

View Abstract

Navigate This Article