Response: Cautious interpretation of assessment of AID variant activities using cells with endogenous AID expression

Response

In their correspondence, although van Maldegem and colleagues confirmed much of what we reported,¹  they concluded instead that activation-induced cytidine deaminase (AID) splicing variants were nonfunctional. They first used a green fluorescence protein (GFP) reversion assay and demonstrated that AID variants can mutate this construct permitting GFP expression. However, they also found that various anticipated negative controls exhibited activity, prompting them to conclude that AID variants are nonfunctional. Regarding the interpretation of these results, it was not clear to us if they were challenging our results by concluding that AID variants were functionally inactive, or if they were questioning the accuracy of the assay itself. We believe it is difficult to interpret results emerging from experimentation in which the negative controls behave unexpectedly. Indeed, if their assertion is correct, it is reasonable that one should also be skeptical about the functionality of the full-length AID protein (AID-FL). We agree there is merit in use of alternative negative controls provided they are shown to indeed qualify as true negative controls.

The authors also provide evidence that GFP reversion frequency correlates with the expression level of yellow fluorescence protein (YFP), and that YFP also acquired mutations which would be an artifact, possibly contributing to the GFP⁺ population. The authors argue this may be further evidence that the GFP assay is not reliable in their hands and further suggest endogenous AID may be active in the NIH 3T3 cells used in their assay. Based on the presence of endogenous AID expression alone, we think the assay used under their conditions appears to be problematic. However, whether this is specific to an NIH 3T3–based assay in general or is unique in this case due to some unexpected contamination remains to be seen. However, our assay uses the mouse pre–B-cell line 70Z/3, and these cells do not express any detectable endogenous AID^2,3  thereby lessening the possibility this was an issue in our system.

The authors also used an in vitro deamination assay that showed bacterially expressed AID variants did not display any detectable activity, whereas AID-FL did. It remains possible that the lack of activity displayed by the variants reflects instead a lack of proper protein folding or alterations in the specificity for the synthetic DNA substrate. Most importantly, although this assay has been used in several publications, its relevance to the in vivo function of AID proteins remains largely unknown. Indeed, other investigators recently set out to specifically address this question and concluded that “DNA deamination activity does not represent the physiological function of AID.”⁴ 

Finally, because intracellular localization of AID may spatially regulate the function of AID protein in cells the authors used localization studies to argue that because AID-ivs3 and AID-dE4 did not undergo nucleocytoplasmic shuttling like AID-FL, this is additional evidence that AID variants are functionally inactive. We believe it is problematic to view AID localization as a surrogate measure of activity because localization and deaminase activity are each dictated by their specific sequences. It is also worth noting that we only minimally tagged FL-AID and its variants with a 10-amino-acid Flag epitope, which differs considerably from the 238-amino-acid GFP protein used by the authors. Actually, the dissonant localization of N-terminally and C-terminally tagged AID or its variants observed by the authors suggests that those proteins are altered by the GFP tag.