Class (Switch) Wars

A "C" CHANGE

Michael Neuberger at the Medical Research Council Laboratory of Molecular Biology, Cambridge, UK, has posited a different hypothesis: that AID works directly on DNA. Such a mechanism explains several experimental observations that could not be explained by the RNA-editing model, such as an apparent G-to-C mutation bias in the first phase of SHM. "Everything fell into place once one conceived of DNA deamination," Neuberger says. Further, deficiencies in the mismatch recognition protein MSH2 refocused CSR breakpoints to sequences that looked like SHM "hotspots," which could be explained if these common DNA sequences were deamination targets for AID.

In this issue's Hot Papers, Neuberger and colleagues presented evidence for this alternate mechanism.⁴⁵⁶ Their proposal that AID directly deaminates DNA "broke open the whole mechanism of hypermutation," says National Institutes of Health immunologist Patricia Gearhart.

The researchers tested two predictions: first, that AID works on DNA, and second, that alteration of the uracil repair pathway would perturb diversification in a predictable way. Evidence supporting the first prediction came from overexpression of AID in Escherichia coli, where they found deaminated cytosine residues in its DNA.⁴

"He got this E. coli to mutate to beat the band," says University of Southern California biochemist Myron Goodman. That the C-to-U mutations were greatly enhanced in cells lacking uracil-DNA glycosylase (UNG), an enzyme that removes uracil from DNA, further suggested that AID works as a DNA-processing enzyme. In two subsequent Hot Papers, Neuberger's group climbed the evolutionary ladder, performing similar experiments in chicken cells and in mice.⁵⁶

Yet Honjo vociferously denies these claims and has recently published evidence that, he says, forces a reexamination of the DNA deamination model.⁷ Honjo reasons that if uracil removal were required to initiate cleavage by base-excision repair, then UNG inhibition would block the DNA cleavage step of CSR. To test this, his group blocked UNG's DNA-binding activity with a specific inhibitor protein (Ugi) and found that DNA cleavage was not inhibited. Further, they showed that UNG mutants that lacked the ability to remove uracil "rescued" CSR in UNG-deficient B cells.

The DNA deamination model proposes that AID generates uracil within antibody gene DNA. The UNG enzyme normally removes uracil. Low-fidelity polymerases that replicate across this site mainly generate transversion mutations (which exchange a purine for a pyrimidine, and vice versa). If UNG is inhibited, high-fidelity DNA polymerases will instead replicate across the uracil to generate transition mutations (which exchange bases of similar structure), reading uracil as thymine. Neuberger and colleague Javier Di Noia posited that if uracil in antibody genes is indeed an intermediate in diversification, then UNG inhibition would shift the pattern of mutation to favor transitions.

To test this, Neuberger and Di Noia inhibited UNG in chicken DT40 B cells, which were engineered to perform somatic hypermutation, with uracil glycosylase inhibitor protein (Ugi).⁵ More than 80% of the resulting mutations represented transitions. "That again followed exactly the prediction that AID was acting on DNA," notes Janet Stavnezer of the University of Massachusetts School of Medicine, Worcester. A subsequent experiment that investigated antibody SHM in mice deficient in UNG generated similar results.⁶ It showed CSR inhibition, providing additional support for the DNA deamination model of antibody diversification.

MORE EVIDENCE

Neuberger's evidence was indirect, however; still needed was biochemical evidence that AID worked directly on DNA. Michel Nussenzweig and F. Nina Papavasiliou, both at Rockefeller University, Fred Alt at Harvard Medical School, and Goodman showed that AID catalyzes the deamination of single-stranded but not double-stranded DNA in vitro and in vivo; however, they found no evidence that the enzyme deaminates RNA.⁸⁹ Both Honjo's group and Nussenzweig and colleagues demonstrated that AID localizes to the nucleus.¹⁰¹¹ And, Akira Shimizu and colleagues at Kyoto University found that AID targets antibody immunoglobulin switch-region DNA during CSR when AID is overexpressed in splenic B cells;¹² Alt later confirmed this finding with endogenous AID.

Alt's group also showed that AID cooperates with a single-stranded DNA binding protein, called replication protein A (RPA), to bind to "hotspots" in the immunoglobulin variable region during transcription, and lead directly to DNA deamination.¹³ "That further supports the DNA model because it means that whatever AID does, it does with a DNA-binding protein, and a single-strand DNA-binding protein at that," says Nussenzweig.

NOT THE LAST WORD

Honjo has not been swayed by Neuberger's results, however, and interprets his own in a manner consistent with an RNA-editing model. "Many people think that DNA deamination is correct, but you can't decide by majority rule," says Honjo. Chief among his arguments is AID's similarity to APOBEC-1. "What was attractive about his hypothesis was and is that AID has homology and is linked to RNA-editing enzymes," says Matthew Scharff of the Albert Einstein College of Medicine, Bronx, NY.

But Neuberger argues that other members of the APOBEC family (including APOBEC-1) can mutate DNA in vitro and that APOBEC-1's RNA editing function may be the exception rather than the rule. "APOBEC-1 is a very late evolutionary arrival. It's only found in mammals. And I would see APOBEC-1's role in RNA editing as a paradigm [as] peculiar," he says.

Honjo has continued to chip away at the DNA-editing model, citing experiments demonstrating that AID can't induce CSR without de novo protein synthesis.¹⁴ "He argued that that fit perfectly with the model," says Stavnezer, because after AID mutates mRNA, the mRNA would need to be translated to produce a protein, the endonuclease that triggers CSR. Some argue that other de novo proteins could be involved. "You could get protein synthesis for a million different reasons," says Goodman.

Honjo claims that his evidence that CSR can still take place in the presence of mutated, inactive UNG contradicts Neuberger's model. "Many people will be convinced by this," Honjo says. "I don't think the [DNA editing] people have had the last word."

Aileen Constans aconstans@the-scientist.com