Activation-induced deaminase catalyses two processes that diversify antibodies. But this enzyme need not work alone: a partner links it to its substrate — single-stranded DNA — and to DNA-repair molecules.
One of the cardinal features of immune systems is that they show enough diversity to mount a defence against virtually any pathogen. The B cells of the immune system, for instance, produce a staggering repertoire of protective antibody proteins. In human and mouse B cells, three DNA-modifying processes generate such diversity: V(D)J recombination, somatic hypermutation and class switch recombination. The discovery that a single enzyme — activation-induced deaminase (AID) — is both necessary1,2 and sufficient for the latter two processes was a breakthrough in understanding the underlying molecular mechanisms. On page 992 of this issue, Chaudhuri and colleagues3 advance our knowledge further, revealing a protein that can assist in AID-induced antibody diversification.
Each of us inherits a limited set of antibody, or immunoglobulin, genes. But the number of pathogens that we might encounter is potentially limitless — hence the requirement for DNA-modifying processes to generate antibody diversity. An antibody contains one of several specific constant regions (which spark off different immune responses), joined to a variable region (which recognizes a particular foreign molecule, or antigen). V(D)J recombination, which occurs in immature B cells, generates the variable region, by randomly shuffling the V, D and J segments of the immunoglobulin genes (Fig. 1). This produces a spread of B cells whose antibodies can recognize different antigens.
Two other processes further diversify antibody genes in mature B cells that have encountered their cognate antigen. As the name suggests, class switch recombination changes the class of antibody that is produced so as to elicit the most appropriate immune response. This occurs by combining the variable region of the gene with appropriate constant regions. Meanwhile, somatic hypermutation introduces single mutations to the variable region, potentially increasing the antibody's affinity for its antigen.
Several years ago, it was discovered that both class switch recombination and somatic hypermutation require the AID protein1,2. But it has been less easy to work out exactly what AID does. The protein shows a close relationship to APOBEC-1, an enzyme that converts cytosine nucleotides to uracils in RNA. So it was initially suggested1 that AID also ‘edits’ RNA — specifically, a messenger RNA that encodes an endonuclease. Endonucleases are enzymes that generate nicks in DNA; the thinking was that the endonuclease in question, appropriately edited, would thereby initiate class switch recombination and somatic hypermutation.
An alternative idea was that AID might deaminate cytosine nucleotides in DNA, not RNA1,4. In this model, the precise details of subsequent events would depend on how cells attempt to repair the cytosine-to-uracil mutations (uracils in DNA are not usually tolerated). Although there is no definitive proof for this theory, it is supported by several observations. First, AID mutates cytosines to uracils in Escherichia coli3. These bacteria should not express the putative target mRNA, arguing against the RNA-editing model. Second, AID can bind to and deaminate DNA (in single-stranded form)5,6,7,8,9. Third, AID associates with the immunoglobulin-gene region in B cells undergoing class switch recombination10. And finally, uracil-N-glycosylase, an enzyme that removes uracils from DNA, is required for normal somatic hypermutation and class switch recombination11,12. Chaudhuri and colleagues' latest findings3 lend additional strong support to the DNA-deamination idea.
DNA is generally double-stranded. But, as mentioned above, AID targets single-stranded DNA. How could such a single-stranded substrate arise in vivo? The answer seems to lie in the fact that single-stranded regions of DNA are transiently exposed on the surface of the enzyme RNA polymerase during transcription — the process by which genes are copied into mRNA. These exposed regions are generally on the opposing strand to that bearing the gene being expressed. So, in E. coli, AID targets the single-stranded region that is complementary to a transcribed gene9. Transcription of appropriate regions of the immunoglobulin genes also seems to be a prerequisite for somatic hypermutation and class switch recombination.
But that's not the end of the story. In B cells undergoing somatic hypermutation, mutations can be made in either DNA strand. And there are certain immunoglobulin-gene hypermutation ‘hotspots’, defined by the sequence RGYW (where R is adenine or guanine, G is guanine, Y is cytosine or thymine, and W is adenine or thymine)13. This sequence preference is due at least in part to AID itself, as the deamination of single-stranded DNA by AID in vitro is skewed to RGYW motifs8.
But how is AID targeted to these single-stranded regions during somatic hypermutation? Chaudhuri et al. previously5 made the key observation that AID purified from B cells deaminated DNA that was provided in single-strand conformation, but was inactive on DNA substrates — akin to the variable regions of immunoglobulin genes — that had been transcribed by the T7 RNA polymerase enzyme. In contrast, crude AID preparations were active in both reactions. The implication was that AID does not function alone to target and deaminate these substrates.
Chaudhuri et al.3 have now used a classical biochemical approach to search for factors that might work with AID. They found replication protein A (RPA), a protein that binds single-stranded DNA and is important in various aspects of DNA metabolism. The authors discovered that transcription-dependent deamination by AID in vitro depends on RPA, and that it correlates with the RGYW content of the DNA substrate. Furthermore, a physical interaction between these two proteins is required for AID to bind to DNA, in an RGYW-dependent manner, during T7-polymerase-mediated transcription. These findings suggest that RPA stabilizes the single-stranded DNA produced during transcription, to enhance AID activity. It is not difficult to imagine that RPA could also serve as an entry point for additional mutational factors and for DNA-repair factors — such as uracil-N-glycosylase and mismatch-repair proteins — that are known to associate with RPA.
Chaudhuri et al. also show that only phosphorylated AID — which is produced specifically by B cells — interacts with RPA. Because the overexpression of AID in non-B cells is sufficient to support RGYW-focused somatic hypermutation and class switch recombination, it is generally accepted that AID is the only B-cell-specific factor that is required for these processes. However, in non-B cells mutations are strongly biased towards guanines and cytosines, whereas in B cells, mutations are found on all four nucleotides. Chaudhuri and colleagues' findings suggest that a B-cell-specific interaction between RPA and phosphorylated AID is required to support bona fide somatic hypermutation, and may explain the guanine:cytosine bias in cells in which such an interaction cannot occur.
Chaudhuri and colleagues' findings3 enrich our understanding of somatic hypermutation, by defining AID's first partner and by providing a conceptual link between AID, single-stranded DNA exposed during transcription, and DNA repair. Their exciting results also raise several questions. How does RPA contribute in vivo? How is the phosphorylation of AID regulated? If the RPA–AID complex specifically targets hypermutation hotspots, how does hotspot targeting occur in non-B cells? Are uracil-N-glycosylase or mismatch-repair proteins recruited to the deamination site through RPA, through AID, or through the AID–RPA complex? Finding this complex is an essential first step towards answering such questions. It also, coincidentally, lends weight to the idea that AID works on DNA, not RNA.
About this article
Nature Communications (2015)