Addendum: RAG-1 and ATM coordinate monoallelic recombination and nuclear positioning of immunoglobulin loci

Nat. Immunol. 10, 655–664 (2009); published online 17 May 2009; corrected after print 18 May 2009, 19 August 2009 and 12 March 2010; addendum published online 12 March 2010.

In our published article, we found that RAG-dependent γ-H2AX foci showed monoallelic colocalization with Igh and Igk alleles in developing B lymphocytes. We sought to determine whether this was due to a stochastic cleavage mechanism or a regulated cleavage process. In Figure 6a–c and Supplementary Tables 9 and 10 initially published, we made a calculation we thought would yield the predicted number of cells in which we would expect to find biallelic colocalization of γ-H2AX foci if recombinase targeting were stochastic and compared that with the actual number obtained. The percentage of cells with biallelic colocalization of γ-H2AX foci in wild-type cells undergoing recombination was significantly lower than the predicted frequency. We believed this calculation supported the idea that monoallelic V(D)J recombination occurs as a result of the restriction of RAG-mediated cleavage to one allele and is not simply a result of stochastic recombination.

It was called to our attention, however, that our calculation of the predicted frequencies of biallelic colocalization of γ-H2AX foci was incorrect and should have been done on an allele basis, not on a per-cell basis. After consulting with biostatisticians and after further reflection, we realized that it was not possible to accurately calculate a predicted frequency on either basis, because such a calculation of predicted frequencies rests on the assumption that pre-pro-B cells, pro-B cells and pre-B cells are homogeneous populations in which all alleles are equally available for cleavage. This assumption is clearly not true for the following reasons. Some cells have two alleles available for rearrangement, whereas others have already undergone a nonfunctional rearrangement on one allele and therefore only have one allele left to rearrange. Also, the calculation does not take account of cells that have functionally rearranged one allele, as these cells move to the next developmental stage. Consequently, the data presented in Figure 6a–c and in Supplementary Tables 9 and 10 as initially published do not allow us to make any statements regarding whether monoallelic recombination in wild-type cells is due to a stochastic process or a regulated process. Thus, the discussion of Figure 6 on page 659 is now obsolete. However, we feel that the remainder of the text is still supported.

We feel it is important to correct the article and to alert the community to the caveats raised above regarding attempts to make this sort of calculation. We therefore reexamined the data through a more straightforward lens, the χ² test, and present here the results. We assessed the number of wild-type and Atm^−/− cells at various stages of development showing monoallelic or biallelic association of γ-H2AX with immunoglobulin loci (Addendum Fig. 1). These data represent the percentage of cells with monoallelic or biallelic association of γ-H2AX with Igh or Igk; the actual number of cells analyzed (n) and statistical P values comparing wild-type and Atm^−/− cells are in Addendum Tables 1 and 2. The numbers of cells analyzed differ from those originally published, as we have added new experiments to make this statistical calculation more robust. The frequency with which γ-H2AX is associated with immunoglobulin alleles varies between experiments because the vagaries of both FISH and mice. Because of this, we always assessed wild-type controls alongside the mutants, and the results consistently showed differences between wild-type and Atm^−/− mice. ATM deficiency caused a statistically significant increase in biallelic cleavage, as shown by biallelic γ-H2AX foci in both pre-pro-B cells and pre-B cells, with the greatest difference occurring in pre-B cells. Our analysis of monoallelic and biallelic cleavage in pro-B cells in the absence of ATM is complicated because D_H-J_H recombination has already occurred, creating many breaks in chromosomes that are not repaired rapidly in the absence of ATM, which decreases the number of alleles available for possible V_H-DJ_H rearrangements in the next developmental stage. Pro-B cells thus contain many broken and missing alleles, and as we assigned scores only to the association of γ-H2AX with immunoglobulin alleles in cells with two alleles, the numbers we could analyze were lower. We believe this leads to a misleadingly small set of pro-B cells with biallelic cleavage from ATM-deficient mice. These results, together with the molecular data presented in Figure 7 of the original article, provide support for the conclusion that RAG cleavage activates ATM-mediated signals that inhibit further V(D)J recombination to suppress biallelic recombination and diminish the risk of chromosome translocations.

Figure 1: Wild-type and Atm^−/− cells at various stages of development with monoallelic or biallelic association of γ-H2AX with Ig loci.

Results are presented as the percentage of pre-pro-B, pro-B and pre-B cells assigned scored as described in Addendum Tables 1 and 2. Data are a composite of three independent experiments.

Table 1 Statistical analysis of γ-H2AX foci co-localization with lg alleles

Table 2 Addendum Table 2 Statistical analysis of data in Addendum Table 1 using Fisher's exact test and 3 × 2 contingency: