The DNA breakage-and-repair mechanism that generates antibodies of different classes has, in theory, a 50% chance of occurring correctly. But this recombination turns out to be heavily biased towards productive events. See Letter p.134
Antibodies are proteins that recognize and neutralize invading pathogens. To accomplish this, antibodies (also called immunoglobulins) must access different tissues and recruit killer molecules and immune cells. These effector functions depend on the antibody's 'constant' region, which varies in protein sequence between the different classes of immunoglobulin — IgM, IgG, IgA and IgE. To produce antibodies appropriate to a particular infection, the class can be changed through an inducible genomic rearrangement called class-switch recombination (CSR)1. This process can occur in two orientations, one of which results in a 'productive' rearrangement whereas the other prevents antibody production. Theoretically, these two events have an equal probability of occurring, which would give CSR a 50% failure rate that would limit the efficiency of antibody responses. But in this issue, Dong et al.2 (page 134) demonstrate that CSR is highly non-random, with 90% of events resulting in a functional rearrangement.
Antibodies of the IgM class are the first to be produced when B cells of the immune system are stimulated by encounter with a pathogen. As the immune response progresses, the IgM constant region is replaced by another one, depending on the infection: IgG antibodies are effective against viruses and bacteria and are the antibodies induced by vaccination; IgA antibodies protect mucosal surfaces; and IgE antibodies attack certain parasites.
The constant regions defining the antibody classes are all encoded by the antibody heavy-chain gene (Igh), and the sequence for each is preceded by a distinct repetitive 'switch' (S) sequence — Sμ, Sγ, Sα and Sɛ (Fig. 1). During CSR, the enzyme activation-induced deaminase (AID) causes simultaneous DNA double-strand breaks (DSBs) at the Sμ and another S region1. The DSBs, which can be up to 200 kilobases apart, are then joined by non-homologous end joining (NHEJ), a ubiquitous pathway for repairing broken chromosomes3. For productive CSR, the broken ends of the two separate DSBs must be joined in the orientation that places the new constant region in place of that of IgM and circularizes and deletes the intervening sequence (Fig. 1). Joining in the other orientation inverts the region between the DSBs, inactivating the antibody gene.
By adapting a high-throughput technique4 that enables analysis of the sequence and relative orientation of massive numbers of junctions between two DSBs, Dong et al. determined the relative frequency of deletion compared with inversion after inducing CSR. Their results show unambiguously that CSR is heavily biased towards the productive orientation, which they found in more than 90% of the joins between two switch regions broken by AID.
The authors further explored this finding by introducing into the genome of antibody-producing B cells sequences that are recognized by the enzyme I-SceI (a yeast enzyme that induces DSBs in DNA). They then looked at the orientation outcomes when one or both DSBs were made by I-SceI at a non-switch-region sequence, compared with breaks created by AID at an S region. It had been shown previously that replacing both S regions with I-SceI recognition sites allows CSR following expression of I-SceI5. Dong et al. find that I-SceI-initiated recombinations lack orientation bias, even when multiple I-SceI sites in tandem are used to mimic the repetitive nature of S regions. This lack of bias probably contributes to the relative inefficiency of CSR following I-SceI-induced DSBs compared with normal CSR5.
By contrast, the authors find that CSR between an I-SceI-induced non-S-region break and an AID-induced S-region break within the Igh gene is more than 75% biased towards the productive orientation. This result, together with the greater than 90% orientation bias of CSR between two AID-induced S-region breaks, suggests that the S regions and/or AID contribute to enforcing the productive orientation. However, although seemingly necessary, the S-region and AID are not sufficient for this bias — Dong et al. found no orientation preference when the AID-induced and the I-SceI-induced breaks were in two different chromosomes, a set-up that mimics AID-dependent chromosomal translocations6.
Previous work had shown that inter-chromosomal fusions between an I-SceI-induced and a spontaneous DSB are unbiased4. Future analysis of junctions between AID-induced or I-SceI-induced breaks in various genomic locations will make it possible to test the authors' hypothesis that the spatial organization of the Igh, in which the S regions are brought into contact with each other within a physically restrained topological domain (Fig. 1), is a key determinant of the orientation preference of CSR. A recent paper shows that two S regions inserted into another gene targeted by AID (the immunoglobulin light-chain gene) are not joined to one another even if efficiently broken by AID7. Instead, the individual DSBs in each S region are just rejoined. This observation is consistent with a role for the specific topology of Igh in promoting the productive joining of breaks between S regions.
Dong et al. obtained further insight into the mechanism of CSR by studying cells lacking DNA-repair factors that are involved in the process. They found that orientation bias was reduced in the absence of the enzyme ATM kinase, which coordinates the response to AID-induced damage8, and in cells lacking the DNA-binding proteins H2AX, Rif-1 or 53BP1, which act to protect broken ends from being resected, thereby promoting NHEJ8,9. The authors propose that inhibiting end resection accentuates an intrinsic predisposition of CSR to proceed in a specific orientation, dictated by the topology of Igh, by allowing NHEJ to repair breaks that are not correctly paired and could join in either orientation. However, 53BP1 prevents end resection by recruiting Rif-1 (ref. 9), and Dong et al. find that decreasing resection in 53BP1-deficient cells did not restore the orientation preference, thus revealing a resection-independent role for 53BP1 in determining orientation. Accordingly, 53BP1 is required for normal CSR but dispensable for I-SceI-mediated CSR3,8. Putative 53BP1 functions include pairing of the S regions and influencing the topology of Igh9. Dissecting the role of 53BP1 in orientation bias will be one of the most interesting challenges arising from this work.
The only other example of orientation-biased DNA rearrangement is VDJ recombination, the process that assembles the antibody genes during B-cell development10. The mechanisms inducing this bias and that of CSR, as revealed by Dong and colleagues, are poorly understood and probably different. However, it is unlikely to be a coincidence that both have evolved to function in the most effective way to ensure the production of antibodies and thereby an efficient immune response.