Targeting of the Activation-Induced Cytosine Deaminase Is Strongly Influenced by the Sequence and Structure of the Targeted DNA

The variable regions of immunoglobulin (Ig) genes encode the antigen binding sites of antibodies for estimated billions of different antigenic determinants. Thousands to millions of different antibody binding sites are created when the hundred or so variable, diversity, and joining genes for Ig heavy and light chains are recombined and diversified by nucleotide deletions and insertions at the V(D)J joints in developing B lymphocytes. In mature B lymphocytes, the rearranged V(D)J sequences are extensively further diversified by somatic hypermutation (SHM) during exposure to antigens that react with the specific Igs comprising the B-cell receptors. SHM is initiated by the activation-induced cytosine deaminase (AID) (11, 18; reviewed in reference 25). It is likely that in vivo AID deaminates cytosines in DNA since inactivation of the uracil glycosylase Ung greatly increases the proportion of transitions from C to T (16). This is probably due to copying the uracil (product of cytosine deamination) with adenine during replication. Furthermore, expression of AID in uracil glycosylase-deficient Escherichia coli results in mutations (14, 17, 23), most of which are also due to replication of uracil (14). While AID clearly can act on DNA, a role as an editor for the mRNA of a specific endonuclease (1) cannot be ruled out. In any case, whether it acts in vivo as a DNA cytosine deaminase or whether it creates an endonuclease or other factor that promotes SHM, three major questions about targeting of the somatic hypermutation process remain to be explained. All three must be understood in view of the finding that Ig gene transcription is required for SHM (2, 13). First, how does SHM target Ig genes and a few other genes expressed in mutating B lymphocytes without affecting all transcribed genes? There must be SHM-specific cis elements in all SHM target genes. One potential targeting element is a binding site for certain helix-loop-helix proteins present in all Ig enhancers (10). However, since this 6-bp element is abundant in the genome, it remains to be determined whether it is essential for SHM and how it would act as a targeting signal. Second, within the genes altered by SHM, only about 1 to 2 kb downstream of the promoter is targeted. This mutation distribution has been shown to relate to initiation of transcription (13), but how the mutation process is terminated in the middle of a transcribed gene is not understood. This report addresses a third issue of SHM targeting, the DNA strand specificity of AID. In vitro experiments by several laboratories have shown that AID targets single-stranded DNA but not double-stranded DNA (4, 7, 15). Double-stranded DNA was targeted in vitro when it was transcribed, but then mainly on the nontranscribed strand (3, 5, 6, 23). Also, when AID is expressed in E. coli, it can target endogenous or introduced genes, but again, mainly on the nontranscribed strand (14, 17, 23). In vivo, however, clearly both DNA strands are targeted (24). This poses the question of whether the in vitro experiments can give reliable clues about the in vivo situation. Our laboratory has shown recently that AID targets double-stranded DNA in vitro on both strands when the DNA is supercoiled (21). In the plasmid substrates we used, the cytosine deaminations were seen mainly in two reportedly negatively supercoiled regions. Based on this finding, we proposed that during SHM in vivo, negative supercoils upstream of the elongating RNA polymerase during transcription may provide access to AID. Here we investigate the role of the DNA sequence and transcription in AID targeting using two plasmid constructs in which either the T7 or the T3 promoter drives transcription of different target sequences. Each plasmid carries two antibiotic resistance markers: one constitutively active and one that requires AID deamination to create a functional ATG translation initiation codon. This independence of the analysis from antibiotic selection allowed us to show that both the transcribed and the nontranscribed DNA strands can be efficiently targeted by AID during transcription. The additional unexpected finding was that the primary sequence greatly influences which DNA strand is targeted by AID and that this difference is independent of the distribution of AID hot spots.