AID protein is up-regulated 48 h after stimulation of cells to undergo CSR. Western blots of 40 μg cytoplasmic and 40 μg nuclear extracts prepared from WT or aid^−/− splenic B cells cultured as indicated. Extracts from freshly isolated resting B cells are shown in the left-most lanes. The four blots were incubated with anti-AID antiserum simultaneously, and exposed to film simultaneously. A 42-kD protein in the nuclear extracts that cross-reacts with the AID antibody is shown to demonstrate nearly equivalent loading of protein in the different lanes. The 26-kD protein in cytoplasmic extracts is likely to be a cross-reacting protein, because a band this size also is observed sometimes in aid^−/− cells (unpublished data). (A) Time course showing induction of AID in WT cells treated with LPS plus IL-4. (B) Blots of cytoplasmic and nuclear extracts from cells treated under the indicated conditions for 48 h. All conditions included BLyS.

Mouse Sμ sequence with DSB sites from WT cells indicated. These breaks were obtained by LM-PCR using the indicated primer (5′Smu) located at the 5′ end of Sμ sequence. The arrows point to the base that was deleted, where the linker is ligated. Breaks occurring at G:C bp at the underlined base in WRC/GYW are indicated by bold arrows. Other breaks are indicated with gray arrows. Early experiments used the Smu probe (nt 453–480) to detect the cloned breaks, and thus, missed some of the breaks occurring upstream of the tandem repeats, which start at ∼820 in this sequence. Therefore, the distribution of break sites is not entirely representative. Furthermore, although additional long break fragments due to breaks occurring at the 3′ end of this segment were cloned, often the break site could not be determined because of difficulty in alignment with the tandem Sμ repeats. We did not include any break sites cloned more than once, because of concerns about amplification of previously cloned segments.

DSBs detected by LM-PCR are induced in Sμ and Sγ3 segments in wild type (WT) but not in aid^−/− splenic B cells 48 h after induction of CSR. (A) Sμ LM-PCR products (amplified with the 5′Sμ primer) from WT and aid^−/− cells treated as indicated for the indicated hours were blotted and hybridized with the Sμ probe. The PCR amplification of the GAPDH gene are shown below the blots as an internal control for template input: threefold dilutions of 1630 cell equivalents. (B) Cμ LM-PCR products from WT cells and aid^−/− B cells detected with the Cμ probe. These blots are from the same experiment shown in (A). (C) Sγ3 LM-PCR products from WT and aid^−/− B cells treated for 48 h with LPS plus anti-δ dextran detected with the Sγ3 probe. To detect Sγ3 breaks, the template input was increased by threefold (threefold dilutions of 4890 cell equivalents) over that used to detect Sμ and Cμ breaks. (D) Most DSBs in Sμ and Sγ3 are UNG-dependent. LM-PCR products from splenic B cells activated for 48 h with LPS plus IL-4 detected with the Sμ5′ probe or Sγ3 probe; threefold dilutions of 2770 cell equivalents. The finding of AID-dependent DSBs is reproducible and was obtained in all six of the six independent experiments performed, always tested in two or more ligations.

Most DSBs are staggered, and a model for conversion of staggered DSBs to blunt DSBs is shown. (A) Most DSB breaks are staggered, as demonstrated by an increase in Sμ LM-PCR products detectable after treatment of activated B cell DNA with 0.5 U T4 DNA polymerase and 200 μM dNTP at 11°C for 20 min before linker-ligation. Shown are threefold dilutions of 1,085 cell equivalents. The same DNA samples were analyzed plus and minus T4 DNA polymerase. (B) A model for the types of end-processing that might result in the pattern of DSBs observed after AID-UNG instigated ss break formation. The Cs on the flaps represent nt attacked by AID, and the G:C bp in the boxes indicates the first nucleotide of the deleted segment. This diagram is for DSBs detected using a primer at the 5′ end of the S region, as used in the experiments reported in this study, except in Fig. S1. Because of the finding that the blunt DSBs occur preferentially at G on the nontranscribed strand, it is likely that 5′ overhangs mostly are converted to blunt DSBs by fill-in DNA synthesis (left-side pathway), rather than by exonuclease or endonuclease activities (middle pathway). 3′ overhangs cannot be filled in and might be removed by exonuclease 1 and ERCC1-XPF (right-side pathway), both of which are known to participate in CSR (33, 34). This model also predicts that break sites detected by a primer located at the 3′ end of Sμ would occur preferentially at C on the top strand relative to the sequence itself. This is what we observed, because the G:C ratio of break sites detected with the 3′ primer (1.0) is much less than that for the 5′ primer (15.7). One of the many puzzles that remain is what removes the deoxyribose phosphate group left after AP endonuclease acts. To obtain blunt-end ligation at G:C bp, this group must be removed from the 5′ ends of the breaks.