The Molecular Retrovirology Unit started out back in the mid 1980s working on the molecular biology of HIV, hence the name. We were the first to publish the sequence of HIV. One of our findings back in 1991 was the phenomenon of retroviral hypermutation (Vartanian et al., 1991). This ensues when human APOBEC3G encounters the HIV genome in the absence of the virally encoded protein Vif.

APOBEC3G is part of a seven gene locus at 22q13.1 of which six encode catalytically functional cytidine deaminases that have, unusually, single stranded DNA as their substrate. The result of cytidine deamination is DNA peppered with uridine (dU). As dU base pairs as dT, hypermutants encode monotonously substituted C residues. OK. The degree of editing is, however, phenomenal; from a few % up to 90% and more. This results in a meltdown of genetic information, essentially death by deamination.

This fitted with the initial take that they were part of innate immunity, and the observation that some APOBEC3 (A3) genes could be up-regulated by interferons. It transpires that the situation is not so simple – déjà entendu? Recently we showed that human mitochondrial and nuclear DNA could be hyperedited in non-inflammatory contexts (Suspène et al., 2011). While several A3 deaminases can machine gun (hyperedit) mtDNA, only one, APOBEC3A, could reach nuDNA. Cytidine deamination initiates DNA catabolism with UNG2 and APE1 being downstream components.

As ever new methodologies allow one to see farther. In our case, the breach was opened up by the development of a variant of PCR that allows selective amplification of hyperedited genomes. Referred to as 3DPCR (Suspène et al., 2005) it exploits the fact that AT rich DNA melts at a lower temperature - an AT base pair is held together by two hydrogen bonds, while there are three for a GC pair. As old as the Double Helix itself, Watson and Crick didn’t quite see it this way back in 1953 (Figure 1).


The third hydrogen bond in a guanine–cytosine base pair (bottom) was missed in the 1953 descriptions of DNA (top).

Getting back to edited human DNA, and our recent publication (Suspène et al., 2011), the Unit has shifted from HIV into cancer research. However, as the field is teaming with lots of excellent groups and will be thoroughly transformed by cancer genomics, we are concentrating on cellular hyperplasia and dysplasia that precede cancer. An emerging trait from cancer genomics is the thousands of CG->TA transitions compared to a healthy genome. APOBEC3 enzymes fit this bill precisely: not only do they uniquely make C->T transitions (after all G->A is C->T on the other strand), but also they are capable of making large numbers of mutations in one go. Actually logs more, which suggests strong negative selection. And strong negative selection will generate bottlenecks and allow fixation large numbers of neutral or nearly neutral mutations, which could explain the larger number of so-called “passenger” mutations in cancer genomes.

Our working hypothesis is that massive APOBEC3 editing is apoptotic while a little editing wounds, rather than kills, the host cell genome so generating a mutant population upon which natural selection acts. Rounds of low level A3 editing, perhaps in the context of persistent inflammation would lead to waves of purifying selection resulting in the emergence of a cancer genome.

APOBEC3A expression in transfected HeLa cells

The Unit is financed by the Institut Pasteur and France’s three major cancer agencies.