Promoting Antibody Diversity

BIOCHEMISTRY OF AID

"These are enzymes that are not easy to work with," says Michael Neuberger of the Laboratory of Molecular Biology, Cambridge, UK. Only present in relatively small amounts in B cells, AID is both inactive and insoluble when made in a test tube, he adds. "There's a lot going on that keeps its activity in check," which makes sense for a protein that causes gene mutation. "And so the business of showing catalytic activity on DNA was nontrivial," says Neuberger.

Because early reports had failed to show activity for the enzyme on RNA, singled-stranded (ss) DNA, or double-stranded (ds) DNA, University of Southern California biologist Myron Goodman thought he'd try something different: a RNA-DNA hybrid. His group's Hot Paper reports activity only when the RNA strand was digested with an RNAse.³ So AID was not working on the hybrid directly, but rather while another enzyme destroyed the RNA strand, the RNAse was "allowing AID to do what it does," says Goodman.

The study also confirmed earlier predictions that AID changes cytosine to uracil, says Patricia Gearhart, an immunologist at the National Institute on Aging in Bethesda, Md. But Goodman's demonstration was not definitive; AID purifications were incomplete, leaving the possibility that other factors were responsible for the observed activity. Also, it was expressed as a glutathione-S-transferase fusion protein in insect cells. "People were saying it could still be RNA editing if you had the real AID protein, instead of this artificial thing," says Gearhart.

Two weeks later, Fred Alt's paper appeared using the real thing: AID highly purified from mammalian B cells. He also reported activity on ssDNA, not dsDNA or RNA, generating uracil.⁴ Alt, a molecular immunologist at Harvard's Children's Hospital, credits his coauthor, Jayanta Chaudhuri, now at Memorial Sloan Kettering Cancer Center in New York City, with the "intuition and insight" that AID might require modification to be active. "So the idea was to get AID from a source where we knew it should be fully active in the processes for which it was supposed to work," says Alt. Alt's experiments also induced transcription to see AID activity. "That got everyone tuned into transcription," says Gearhart.

A third biochemical study, by Nina Papavasiliou at Rockefeller University, used AID purified from Escherichia coli.⁵ Commenting on the three biochemical studies that came out that year, Nussenzweig says, "The assays were different in each case, but the answer was the same."

GENETICS OF AID

Neuberger's work in 2002 first turned everybody's attention to DNA.⁶ "What Michael did was to express AID in E. coli and show that it could mutate DNA in E. coli," says Nussenzweig, "meaning that it had all the activity required for making somatic hypermutation."

In this issue's fourth Hot Paper, Nussenzweig showed that DNA actually was the substrate for the enzyme in this system. He used an inducible promotor on a gene, and "when we induced the gene, we got more mutation," he says.⁷ Thus, ssDNA created during transcription was the substrate for AID. "Transcription is important because it liberates the substrate."

© 2004 Nature Publishing Group

Activation-induced cytidine deaminase (AID) deaminates cytidine residues in DNA, converting them into uridine. The U-G mismatch can then be processed either by base excision repair (BER) or mismatch repair (MMR). In somatic hypermutation, the mismatch can be replicated to produce a C-to-T mutation, nick processing by UNG can produce an abasic site, or a gap can be filled in by error prone polymerases. During ' class-switch recombination, stag' gered nicks are thought to be processed by unknown exonucleases or error-prone end-filling leading to blunt double-strand breaks that can be ligated to similarly created breaks on the downstream. (Adapted from M. Nussenzweig, F.W. Alt, Nature Medicine, 10:1304–5, 2004.)

Both Neuberger and Nussenzweig exploited a key feature of the E. coli system: the step subsequent to cytosine deamination by AID. Uracil does not belong in DNA and needs repair; this is the work of uracil DNA glycosylase (UNG). If AID turns C to U in DNA, "then a mutation in an UNG or that kind of enzyme should interfere with its action," explains Nussenzweig. "And it does. I think that's the clearest evidence, and it's genetic."

Researchers allow that not all doubt has been put to rest. Tasuku Honjo, whose group at Kyoto University originally described AID, still uses an RNA-editing mechanism to explain AID action. Enzymes related to AID (known as APOBEC family proteins) use this mechanism, and when put into E. coli, they can mutate DNA (and thus are a critical positive control in studying AID). Honjo was unavailable for comment. "This field still has some very basic disputes at the biochemical level," says Nussenzweig. The argument goes: "Yes, you have shown that it works in E. coli and you have shown that it works in vitro – on ss DNA substrates," explains Nussenzweig, "but just because it can do that doesn't mean it does do that."

AID IN TARGETING

Meanwhile, researchers are building the DNA case. The fact that enzyme activity was related to transcription "suggests that the targeting to immunoglobulin genes may be related to the higher amounts of transcription," says Gearhart. "So now all of a sudden everyone's thinking targeting."

Regulation of a process as potentially damaging as DNA hypermutation is obviously an important issue. Alt has identified one potential regulatory factor, replication protein A, a ssDNA binding protein.⁸

AID can clearly work on both strands of DNA, despite early evidence that the enzyme targets only the nontemplate strand. Biochemically, AID has specificity for certain DNA base sequences in immunoglobulin genes known to be mutational hot spots. Whether AID induces point mutation or double-strand breaks may depend on the frequency of a three-base sequence known as WRC. In the variable region of the gene, processing of an abasic site (after the U is excised) by mistmatch repair or error-prone polymerase results in mutation. A higher frequency of WRC motifs in the switch region increase the likelihood that abasic sites result in stagger breaks and the recombination that underlies class switch.

"What's really nice about all this work is that it ties together in vivo phenotypic observations of hypermutation, which we've known for about 20 years now, [with] these beautiful genetics experiments and biochemistry," says Gearhart. Perhaps it is a case where a field has been looking for answers for so long, that once there was a useful clue, everything seemed to fall into place.