Ataxia Telangiectasia Mutated and MSH2 Control Blunt DNA End Joining in Ig Class Switch Recombination

Bcells recognize and eliminate Ags by producing Igs, also known as Abs. Each Ig is composed of two H chain (IgH) and two L chain (IgL) polypeptides, which are held together by disulfide bonds to form the characteristic “Y” shape structure of the Ig (1). The N termini of IgH and IgL comprise the variable region of each polypeptide, and together they form the Ag binding site of the Ig, whereas the C region of IgH imparts the effector function of the Ig.

In a process known as class-switch recombination (CSR), mature B cells generate dsDNA breaks (DSBs) within the IgH gene locus (Igh) to diversify the Ig repertoire through a DNA recombination event and produce new Ig isotypes (2–4). CSR is mediated by the enzymatic activity of activation-induced cytidine deaminase (AID) (5–7), which converts deoxycytidines into deoxyuridines in the repetitive switch (S) region that precedes each Igh constant coding exon (5–7). Noncanonical repair of these U:G mutations in DNA by either mismatch repair (MMR) or base excision repair (BER) creates staggered DNA breaks that promote recombination between distal S regions (8–15) and expression of alternative Igh constant coding exons.

During BER, uracil DNA glycosylase (UNG) generates an abasic site by removing the AID-generated uracil base, allowing for subsequent cleavage of the phosphodiester backbone by apurinic/apyrimidinic endonuclease 1 (APE1) to generate an ssDNA break (SSB) (13, 16). In contrast, MMR utilizes a heterodimer of mutS homologs mutS homolog 2 (MSH2) and MSH6 to recognize DNA base pair mismatches and recruit the mutL homology 1 (MLH1) and PMS1 homolog 2 (PMS2) heterodimer to cleave the phosphodiester backbone 5′ of the mismatch (6). Exonuclease-1 (EXO1) degrades the DNA to remove the U:G mismatch (17). Staggered SSBs in close proximity constitute DSBs, which are converted into recombination events. Mice deficient in MSH2 or UNG show reduced levels of switched isotypes as compared with wild-type (WT) mice (8). Furthermore, simultaneous loss of both MSH2 and UNG blocks CSR completely, demonstrating the complementary function of MMR and BER in CSR (8).

Activation of BER and MMR during CSR is mediated by the phosphorylation of AID at serine 38 (pS38-AID), which promotes the interaction of AID with APE1 (18). Mice with a homozygous knock-in mutation of the phosphorylation site (Aicda^S38A/S38A), which changes S38 to an alanine, have significantly reduced CSR (19, 20). Aicda^S38A/S38A mice with homozygous null mutations in Ung or Msh2 show a CSR phenotype comparable to Ung^−/−Msh2^−/− mice, indicating a critical role of AID phosphorylation in activating BER and MMR during CSR (21). Furthermore, AID phosphorylation at S38 requires ataxia telangiectasia mutated (ATM), a PI3K-related kinase (PIKK) that responds to DSBs and is required for WT CSR (22, 23). Atm^−/−mice show reduced levels of pS38-AID and impaired AID/APE1 interaction, suggesting that ATM promotes phosphorylation of AID to induce its interaction with APE1 and drive CSR through BER (18). However, the mechanism by which ATM signals noncanonical, recombinogenic BER during CSR remains unclear.

After BER- and MMR-dependent DSBs are formed in the recombining S regions, nonhomologous end joining (NHEJ) factors ligate the broken ends and promote DNA recombination rather than canonical DSB repair (24, 25). Notably, the absence of NHEJ factors, such as Ku70, Ku80, X-ray repair cross complementing 4 (XRCC4), or DNA ligase IV, reduces CSR efficiency and increases Igh chromosomal translocations (26–28). Without NHEJ, CSR can be completed through a microhomology-mediated end joining pathway, also known as alternative end joining (A-EJ), which is more error-prone and less efficient than NHEJ (29, 30). Recombined S–S junctions from WT B cells show ligations resulting primarily from NHEJ (26, 27). Most of these NHEJ-repaired S–S junctions show short stretches of homology, or microhomology, between donor and acceptor S regions of <4 nt, and the remaining ∼30% of S–S junctions contain no homology between the recombining S regions and are characterized as a blunt end join (31). Recombined S–S regions that have been repaired by A-EJ contain longer stretches of homology, which skew toward ≥5 nt (26, 31), yet the conditions under which A-EJ is favored over NHEJ during CSR remain unclear. Atm^−/− B cells exhibit longer microhomologies in recombined S–S junctions, indicating that ATM is required for NHEJ, and A-EJ mediates recombinational repair in the absence of ATM (23, 32). Interestingly, MSH2 has been implicated in regulating A-EJ during CSR because Sμ–Sγ3 junctions from B cells lacking MSH2 exhibited no microhomology above 5 nt (10). In this study, we demonstrate an epistatic role for Atm to Msh2 in CSR. Loss of MSH2 and ATM in B cells leads to a reduction in CSR concurrent with a block in blunt-end S–S junctions, suggesting a role for MSH2 in enforcing NHEJ in the absence of ATM. Additionally, to our knowledge, we report a novel synthetic lethality between Atm and Msh2, which implicates ATM or MSH2 as a potential therapeutic target for cancers driven by mutations in either gene.

C57BL/6 and Atm^−/− (stock no. 008536) mice were purchased from The Jackson Laboratory (33). Msh2^−/− mice were a gift from H. te Riele (34), Aicda^−/− mice were a gift from T. Honjo (35), Atm^fl/f mice were a gift of Shan Zha (36), and CD23-Cre mice were a gift of Meinrad Busslinger (37). Husbandry of and experiments with mice were conducted according to protocols approved by The City College of New York Institutional Animal Care and Use Committee.

To obtain embryonic day 13.5 (E13.5) embryos, females 3–6 mo of age were placed into a cage previously occupied by male mice 1 d prior to introducing the male mouse. Timed matings were set up by limiting the mating period to 1 d. Pregnant female mice were euthanized 14 d after mating (E13.5), and embryos were collected from the uterine horn immediately after euthanasia. To genotype the embryos, one hindlimb and tail were incubated in lysis buffer (10 mM Tris, 0.1 M EDTA, 0.5% SDS, and 200 μg/ml proteinase K) overnight at 56°C. Genomic DNA was isolated by isopropanol precipitation and resuspended in TE buffer. To genotype newborn mice (postnatal day 0 [P0]), pregnant females were monitored daily for neonates ∼3 wk after mating. Tail biopsies were collected from neonates for genotyping. When a dead carcass or stillborn neonate was found in the litter, tissue samples or tail biopsies were collected and genotyped when possible. Adult mice were identified by tail biopsy ∼4 wk after birth (postnatal day 28 [P28]), when the mice were weaned from the nursing female. The primers for each genotype are as follows: Atm (ATMB, 5′-CGAATTTGCAGGAGTTGCTGAG-3′; ATMF, 5′-GACTTCTGTCAGATGTTGCTGCC-3′; and ATMN, 5′-GGGTGGGATTAGATAAATGCCTG-3′), Msh2 (MshF, 5′-GCTCACTTAGACGCCATTGT-3′; MshWR, 5′-AAAGTGCACGTCATTTGGA-3′; and MshKR, 5′-GCCTTCTTGACGAGTTCTTC-3′), CD23-Cre (CD23cre_1, 5′-GCGACTAGTAACTTGTTTATTGCA GCTTAT-3′; and CD23cre_2, 5′-GCGTCCGGAAGACACCACATCCCAATTCTT-3′), and Atm^f (ATMgF86723, 5′-ATCAAATGTAAAGGCGGCTTC-3′; ATM_BAC13, 5′-CATCCTTTAATGTGCCTCCCTTCGCC-3′; and ATM_BAC7, 5′-GCCCATCCCGTCCACAAT ATCTCTGC-3′). PCR with DreamTaq (Thermo Fisher Scientific, no. K1082) was performed for 30 cycles using the following conditions: 95°C, 30 s; melting temperature, 30 s; 72°C, 30 s. The melting temperature (annealing temperature) for Atm and Msh2 was 55°C, CD23-Cre 58°C, and Atm^f 63°C.

B cells were purified from the spleens of mice (aged 2–7 mo of age) through negative selection with anti-CD43 magnetic beads (Miltenyi Biotec) and subsequently cultured in RPMI 1640 media (Life Technologies) supplemented with 10% FBS (Atlanta Biologicals), 2 mM l-glutamine, 1× penicillin/streptomycin (Corning), and 47.3 μM 2-ME. For switching to IgG1, B cells cultured at 1 × 10⁶ cells/ml with 25 μg/ml LPS (Sigma-Aldrich, catalog no. L7261) and 12.5 ng/ml IL-4 (R&D Systems). B cells were split 1:2 (v/v) 48 and 72 h after isolation. To stimulate CSR to IgG3, 25 μg/ml LPS was used. After 96 h poststimulation, B cells were harvested for analysis by flow cytometry. Live cells were kept on ice and stained in Life Technologies PBS (pH 7.4) (catalog no. 10-010-049) supplemented with 2.5% FBS. To identify switched B cells, samples were stained with anti-IgG1 (BD Biosciences; allophycocyanin conjugate, catalog no. BDB550874 or PE conjugate, catalog no. BDB550083) or anti-IgG3 (BD Biosciences; FITC conjugate catalog no. BDB553403). DAPI was used for live/dead exclusion. Samples were analyzed on a BD LSR II cytometer. For proliferation assays, B cells were labeled with carboxylic acid, acetate, succinimidyl ester (SNARF) (Thermo Fisher Scientific, catalog no. S22801) and analyzed by flow cytometry at time of labeling (D0), and every 24 h thereafter.

Protein extracts were prepared from IgG1-stimulated B cells at 72 h poststimulation. Cells were lysed in Nonidet P-40 lysis buffer supplemented with proteinase inhibitor tablet (MilliporeSigma). Nonidet P-40 cell lysates were resolved on 8% SDS-PAGE gels, transferred overnight to Immobilon (MilliporeSigma), and immunoblotted with anti-ATM (GeneTex, catalog no. GTX70103), anti-MSH2 (Abcam, catalog no. ab70270), anti–α-tubulin (Sigma-Aldrich, catalog no. T9026-100UL) or anti-AID (38) Abs. Immunoblots were developed using anti-mouse HRP or anti-rabbit HRP using the Pierce ECL Western blotting substrate (Thermo Fisher Scientific, catalog no. PI32106) and imaged on an Azure A300 chemiluminescent imaging system. RNA was extracted from IgG1-stimulated B cells at 72 h poststimulation using TRIzol reagent (Thermo Fisher Scientific), and cDNA was prepared using the ProtoScript II kit (NEB). Germline transcription was analyzed using quantitative PCR with SYBR Green (Roche) and primers for Sμ (IμF, 5′-CTCTGGCCCTGCTTATTGTTG-3′; and CμR, 5′-GAAGACATTTGGGAAGGACTGACT-3′) or Sγ1 (Iγ1, 5′-GGCCCTTCCAGATCTTTGAG-3′; and Cγ1, 5′-GGATCCAGAGTTCCAGGTCACT-3′). Reactions were analyzed on a LightCycler 96 (Roche).

Mice (aged 2–6 mo) were injected i.p. with 50 μg of 4-hydroxy-3-nitrophenylacetyl conjugated to chicken γ globulin (NP-CGG) (Biosearch Technologies) in Imject Alum (Thermo Fisher Scientific) (39). Mice were given a boost of NP-CGG in Imject Alum 10 d after primary immunization. Sera were obtained from blood that was collected on day 0 (preimmunization) and day 7 and day 21 (postimmunization) through cheek bleeds (days 0 and 7) or cardiac puncture (day 21). To analyze Ig serum titers, ELISA assays were performed as previously described (21).

DNA from IgG1-stimulated B cells was isolated from ∼1.5 × 10⁶ cells 72 h after stimulation. Sμ–Sγ1 junctions were amplified using nested PCR primers: Sμ1b (5′-AACTCTCCAGCCACAGTAATGACC-3′) and Sγ1.1 (5′-CTGTAACCTACCCAGGAGACC-3′) were used for the first cycle of PCR; SμF2 (5′-GAGAAGGCCAGACTCATAAAGCT-3′) and Sγ1.1R2 (5′-GTCGAATTCCCCCATCCTGTCACCTATA-3′) were used in the second-round PCR, as previously described (40, 41). Both rounds of PCR were performed for an initial 10 cycles at 94°C for 15 s, 52°C for 30 s, and 72°C 2 min, followed by 23 more cycles with 5 s added per cycle. All PCRs were performed using Q5 high-fidelity DNA polymerase (NEB) according to the manufacturer’s instructions. Five identical PCR reactions were setup for each B cell stimulation. The reactions were pooled, concentrated in a speed-vacuum, and resolved on a 1% agarose gel. PCR products (0.5–1 kb) were excised from the gel, cloned into cloning vector pJET1.2 (Thermo Fisher Scientific, catalog no. K1231), and Sanger sequenced using the T7 forward primer. Sequenced junctions were aligned against reference Sμ (MUSIGCD07) and Sγ1 (MUSIGHANB) in National Center for Biotechnology Information BLAST with the low complexity filter disabled.

Data figures were generated using GraphPad Prism 7.01. Unless otherwise noted, statistical analyses were completed by one-way ANOVA in Prism.

BER and MMR function redundantly to direct CSR in B cells (8). ATM regulates the phosphorylation of AID and its interaction with APE1, a BER protein (18). To determine the role of ATM in BER-dependent CSR, we bred Atm^−/− and Msh2^−/− mice to generate Atm^−/−Msh2^−/− mice. Atm^−/− mice carry a PGKneo modified exon at position 5790 in the Atm locus that recapitulates mutations observed in ataxia telangiectasia patients (33). Msh2^−/− mice contain a targeted disruption 5′ of the exon encoding the MSH2 ATP binding domain, resulting in a frameshift and mRNA decay (34). Atm^−/− and Msh2^−/− single knockout mice survive into adulthood, albeit with mild immune defects and a predisposition to lymphomagenesis (42, 43). Surprisingly, of the 247 mice genotyped at P28, no Atm^−/−Msh2^−/− mice were found (Table I, p = 3.3 × 10⁻⁴), unveiling a novel synthetic lethal relationship. To determine the developmental stage at which the lethality occurs, embryos were collected from timed mating 2 wk after conception (E13.5) and genotyped. Mendelian ratios of E13.5 embryos from a dihybrid cross were obtained (Table II). Newborn pups (P0) from Atm^+/−Msh2^−/− × Atm^+/−Msh2^+/−crosses yielded less than half of the expected number of P0 Atm^−/−Msh2^−/− neonates; however, these findings are not statistically significant (Table III; p = 0.292). Collectively, these data indicate that the Atm^−/−Msh2^−/− lethality occurs between E13.5 and P0 with an incomplete penetrance of the synthetic lethal phenotype.

To examine the function of ATM and MSH2 in CSR, we bred Msh2^−/− mice to mice that carry a conditional knockout Atm allele (Atm^f/f) and transgenic mice expressing Cre recombinase under the control of a B cell–specific promoter (CD23-Cre⁺) (36, 44). The CD23-Cre⁺Atm^f/fMsh2^−/− mice, which carry a conditional deletion of Atm in B cells and a germline deletion of Msh2, are hereafter referred to as Atm^D/DMsh2^−/− mice. Unlike Atm^−/−Msh2^−/− mice, Atm^D/DMsh2^−/− mice were obtained at the expected Mendelian frequencies (Supplemental Table I), suggesting that the loss of ATM and MSH2 in B cells does not contribute to the Atm^−/−Msh2^−/− lethality. To investigate the loss of ATM and MSH2 on Ig production, Atm^D/DMsh2^−/− and control mice were immunized with NP-CGG, a T cell–dependent Ag (45, 46). Serum Igs were analyzed by ELISA prior to and 21 d after a primary and booster immunization of NP-CGG. Atm^−/−, Aicda^−/−, Msh2^−/−, and WT mice showed IgM and IgG1 titers that were consistent with previously published reports (Fig. 1) (21, 23, 35, 42). Atm^D/DMsh2^−/− IgM titers pre- and postimmunization did not differ significantly from WT, Atm^−/−, Msh2^−/−, Atm^D/D, Msh2^−/−, and CD23-Cre⁺ mice (Fig. 1). Given that mature B cells produce IgM without undergoing CSR, these data suggest that Atm^D/DMsh2^−/− B cells do not have any defects in plasma cell development, Ig production, or Ig secretion.

To determine whether ATM and MSH2 are required for switched Ig isotypes, IgG1 serum concentrations were quantified before immunization and 21 d after a primary and booster immunization with NP-CGG (Fig. 1). Consistent with previous reports, IgG1 is absent in Aicda^−/− mice and significantly reduced in Msh2^−/− mice (average of 1041 μg/ml) (35, 42). Atm^−/− mice exhibit a more pronounced reduction in IgG1 (average of 554 μg/ml) as compared with Msh2^−/− mice (13, 14). The average IgG1 titer of Atm^D/D mice (354 μg/ml) did not differ significantly from Atm^−/− mice pre- or postimmunization, suggesting that the conditional deletion of Atm effectively recapitulates the IgG1 phenotype observed in the germline deletion of Atm. To determine whether the reduced IgG1 resulted from defective germinal center B cell development, Peyer’s patches from NP-CGG–immunized mice were stained with anti-B220 Ab and peanut agglutinin and analyzed by flow cytometry (Supplemental Fig. 1A). Atm^D/DMsh2^−/− mice showed comparable germinal center B cell numbers as compared with WT, Atm^D/D, and Msh2^−/− mice, suggesting that the reduced IgG1 in Atm^D/DMsh2^−/− mice was not a result of defective germinal center formation but rather a B cell dysfunction in CSR.

To investigate whether Atm^D/DMsh2^−/− B cells contain an intrinsic defect in CSR, Atm^D/DMsh2^−/− and control splenic B cells were purified and stimulated in vitro with LPS + IL-4 to induce CSR to IgG1. Consistent with previous reports (4, 22, 42), CSR to IgG1 is significantly reduced in Atm^−/− and Msh2^−/− B cells (Fig. 2). However, trace amounts of ATM protein are detectable by immunoblot in Atm^D/D B cells (Supplemental Fig. 1B). Additionally, Atm^D/D and Atm^−/− B cells show no significant difference in CSR to IgG1 (p = 0.990). Furthermore, CD23-Cre⁺ B cells have no significant difference in IgG1 CSR as compared with WT cells (p = 0.999), demonstrating that the defect in CSR is not due to aberrant B cell–specific expression of Cre recombinase. As observed in the serum IgG1 of Atm^D/DMsh2^−/− mice postimmunization (Fig. 1), Atm^D/DMsh2^−/− B cells exhibit significantly reduced, but not ablated, IgG1 (Fig. 2, p < 0.0001). Atm^D/DMsh2^−/− mouse IgG3 serum titers and Atm^D/DMsh2^−/− B cells that were stimulated to switch to IgG3 were similarly reduced (Supplemental Fig. 2), indicating that the defects in CSR in Atm^D/DMsh2^−/− mice and B cells are not isotype specific. To determine whether the reduction in IgG1 was due to reduced proliferation, Atm^D/DMsh2^−/− or control B cells were labeled with the far-red dye SNARF 24 h after purification and monitored every 24 h thereafter for dilution of the dye via flow cytometry. No significant differences in cell proliferation were observed between WT, Atm^D/D, Msh2^−/−, and Atm^D/DMsh2^−/− B cells (Supplemental Fig. 3A). CD23-Cre⁺ B cells diluted the SNARF dye quicker than WT B cells, indicating faster proliferation of CD23-Cre⁺ B cells (Supplemental Fig. 3A). To examine whether the reduction in IgG1 in Atm^D/DMsh2^−/− B cells was due to defects in germline transcription at the recombining S regions (47), germline transcripts for μ and γ1 were quantified from LPS + IL-4–stimulated Atm^D/DMsh2^−/− or control B cells. No significant difference in μ and γ1 germline transcripts was observed across genotypes (Supplemental Fig. 3B, 3C), suggesting that the reduction in CSR observed in Atm^D/DMsh2^−/− B cells is not due to transcriptional defects. Furthermore, immunoblots for AID showed that Atm^D/DMsh2^−/− B cells expressed AID comparably to WT, Atm^D/D, and Msh2^−/− B cells (Supplemental Fig. 1B, 1C), indicating that the CSR defect in Atm^D/DMsh2^−/− B cells is not due to the absence of AID. In sum, the reduced CSR in Atm^D/DMsh2^−/− B cells likely results from defects in the molecular pathways regulating the formation or repair of DSBs in S regions rather than in plasma cell development, B cell proliferation, transcriptional activation of recombining S regions, or AID expression.

CSR occurs through an intrachromosomal recombination between repetitive, intronic S regions that precede each constant coding exon, resulting in the expression of a new Ig isotype and deletion of the intervening constant exons (10). Ligation of the DSBs in the recombining S regions requires an end-joining pathway to repair the DSB and complete the recombination. In WT B cells, end joining during CSR requires primarily NHEJ wherein recombined S–S junctions are ligated with no or a few nucleotides of homology (<4 nt) (48). However, in the absence of NHEJ, B cells completing CSR can use an A-EJ pathway that requires resection of the broken DNA to generate short regions of ≥5 nt of homology at the recombined S–S junction (40, 48, 49). Therefore, sequencing the recombined S–S junctions analyzes and quantifies the type of DNA repair and end-joining pathway that is used in recombining S regions. During end joining, the activity of ATM coupled with the amount of DNA resection at the site of recombination determines whether NHEJ will be activated (50, 51). Atm^−/− B cells show a reduction in blunt end joins and an increase in microhomologies >4 nt in S–S junctions, suggesting that ATM drives NHEJ during CSR (22). In contrast, S–S junctions from Msh2^−/− B cells are joined using blunt or limited microhomology, suggesting that MSH2 promotes A-EJ (10).

To determine whether NHEJ or A-EJ functions in the absence of ATM and MSH2, we analyzed Sμ–Sγ1 junctions from WT, Atm^D/D, Msh2^−/−, and Atm^D/DMsh2^−/− B cells that were stimulated with LPS + IL-4 to induce CSR to IgG1 in vitro (Fig. 3). Sμ–Sγ1 junctions from Msh2^−/− B cells contained primarily blunt end joins or short-stretch microhomology, as previously described for Sμ–Sγ3 junctions in Msh2^−/− B cells (10, 11, 52). Consistent with Sμ–Sγ1 junctions from Atm^−/− B cells (22), Sμ–Sγ1 junctions from Atm^D/D B cells trend toward longer stretches of microhomology and reduced blunt joins as compared with WT B cells (Fig. 3A), although these trends have not reached statistical significance. Almost 20% of the Sμ–Sγ1 junctions in Atm^D/DMsh2^−/− B cells have microhomology lengths of ≥5 nt (Fig. 3A). Most strikingly, Sμ–Sγ1 junctions from Atm^D/D Msh2^−/− B cells show a significant loss in blunt end joins (Fig. 3A,p = 0.007), suggesting a severe block in NHEJ in Atm^D/DMsh2^−/− B cells. Additionally, coupled with the reduction in blunt end joins, Atm^D/DMsh2^−/− B cells display a stark increase in insertions at Sμ–Sγ1 junctions as compared with WT B cells (Fig. 3A, p = 0.021). Significant differences in blunt end joins and insertions are also observed between Atm^D/DMsh2^−/− and Msh2^−/− B cells (p = 0.001 and p = 0.003, respectively). Although we observed a smaller total percentage of insertions at Sμ–Sγ1 junctions in Msh2^−/− B cells as compared with previous Sμ–Sγ3 analysis (10), these insertions were more than triple the length of that observed in WT Sμ–Sγ1 junctions (Fig. 3B). The average inserted sequence at Sμ–Sγ1 junctions in Msh2^−/− B cells was 80 nt as compared with 12 nt in WT B cells. Similarly, Atm^D/D insertion lengths were longer (average 55 nt) (Fig. 3B). Interestingly, the Sμ–Sγ1 junctions in the Atm^D/DMsh2^−/− B cells skew toward insertion lengths of ≥100 nt, with the largest insertion size of almost 400 nt; however, the average insertion lengths in Sμ–Sγ1 junctions of Atm^D/DMsh2^−/− B cells was not statistically different from the average insertion lengths in Atm^D/D and Msh2^−/− Sμ–Sγ1 junctions. The significant loss of blunt end joins coupled with an increase in insertions observed in Sμ–Sγ1 junctions of Atm^D/DMsh2^−/− B cells indicates that MSH2 enforces blunt end joining in the absence of ATM and identifies a distinct recombinogenic repair in Atm^D/DMsh2^−/− B cells as compared with Atm^D/D or Msh2^−/− B cells. Collectively, these data suggest that the DSBs in Atm^D/DMsh2^−/− S regions are joined in a process even more inefficient than either NHEJ or A-EJ that result in an inefficient and sloppy end joining (Fig. 4).

Phosphorylation of AID and the subsequent interaction of pS38-AID with APE1 depends on ATM (18), suggesting that ATM may regulate CSR through BER. Based on genetic crosses that inactivated BER and MMR (8), genetic inactivation of Atm and Msh2 was predicted to block both BER and MMR pathways, respectively, and abrogate CSR. Surprisingly, dihybrid crosses of Atm^+/−Msh2^+/− yielded no adult Atm^−/−Msh2^−/− mice, suggesting a synthetic lethal interaction between mutations in Atm and Msh2 (Table I). However, Msh2^−/− mice with a conditional deletion of Atm in B cells (Atm^D/DMsh2^−/−) are viable in adulthood, indicating that the Atm^−/−Msh2^−/− lethality is not B cell–dependent. Although the loss of Atm and Msh2 induces a late embryonic to early postnatal lethal phenotype in mice (Tables II, III), the underlying cellular mechanism behind this lethality remains unknown. Synthetic lethality results when the simultaneous loss of two genes leads to cell death, whereas the loss of each gene alone is tolerated and the cells are viable (53–55). In mice, synthetic lethalities between genes encoding DNA damage response proteins usually occur during gestation and early postnatally (56–60). For example, mutations in the genes for the DNA-dependent protein kinase catalytic subunit (DNA-PKcs) and ATM lead to a synthetic lethal phenotype and death of mouse embryos at E7 (61). Interestingly, loss of XRCC4 or DNA ligase IV results in a late embryonic lethality similar to the Atm^−/−Msh2^−/− embryos (62, 63). The significant loss of blunt end joins in Atm^D/DMsh2^−/− B cells suggests a block in NHEJ in the absence of ATM and MSH2 (Fig. 3A). However, whether ATM- and MSH2-deficient mouse embryos exhibit any defects in genomic instability and consequently cell proliferation, cell cycle arrest, or cell death remains to be determined. Alternatively, the reduction in observed Atm^−/−Msh2^−/− mice could result from metabolic defects. ATM and MSH2 respond directly or indirectly to reactive oxidative species (ROS), which are produced as byproducts of respiration (64, 65). Atm^−/− cells contain higher levels of intracellular ROS as compared with WT cells, and the mechanism by which ATM responds to ROS is independent of its role in DSBs (66, 67). Additionally, MSH2-deficient cells have increased 8-oxo-G–containing DNA, a byproduct of oxidative damage (68). Thus, the synthetic lethality between Atm and Msh2 may result from cellular respiratory failure or mitochondrial damage.

Contrary to the original hypothesis, Atm^D/DMsh2^−/− B cells complete CSR, albeit at severely impaired levels (Fig. 2, Supplemental Fig. 2A). These data suggest that ATM functions epistatically to BER and MMR during CSR. Alternatively, ATM may be dispensable for BER-driven CSR. Interestingly, Atm^−/−Ung^−/− mice develop to adulthood and are obtained at Mendelian frequencies (data not shown). Atm^−/−Ung^−/− mouse IgG1 titers are comparable to Atm^−/− mouse IgG1 titers (data not shown), further supporting an epistatic role for ATM. Contrary to the hypothesized role of ATM modulating BER through pS38-AID and APE1, IgG1 titers in Atm^D/DMsh2^−/− mice were significantly reduced pre- and postimmunization but not ablated (Fig. 1, Supplemental Fig. 2B). To examine the DNA repair pathway used in the absence of ATM and MSH2, Sμ–Sγ1 junctions were sequenced in Atm^D/DMsh2^−/− B cells that were stimulated to switch to IgG1. Despite the smaller sample size from Sanger sequencing (Fig. 3), the Sμ–Sγ1 junction data are comparable to a high-throughput genome-wide sequencing of stimulated B cells (69) and confirms that ATM promotes NHEJ of S–S junctions in IgG1-switched B cells. Previous work in MSH2-deficient B cells stimulated with LPS indicated an absence of long microhomologies, suggesting a role for MMR in driving A-EJ (10). In contrast to the Sμ–Sγ3 junctions (10), Msh2^−/− B cells show Sμ–Sγ1 junctions that are comparable to WT B cells (Fig. 3A). A possible explanation for the discrepancy in the phenotypes reported between Sμ–Sγ3 and Sμ–Sγ1 is that Sγ3 and Sγ1 have different recombination phenotypes in the absence of MSH2, possibly due to different repeat sequences (2, 52, 70). However, in the absence of the NHEJ factor XRCC4, ablating MMR further decreases CSR, supporting the hypothesis that MMR can drive A-EJ (24). Interestingly, the reduced CSR in B cells doubly deficient for XRCC4 and MSH2 is comparable to CSR in the Atm^D/DMsh2^−/− B cells (Fig. 2, Supplemental Fig. 2A) (24). Furthermore, blunt end joins in Sμ–Sγ1 junctions of Atm^D/DMsh2^−/− B cells are almost negligible, which indicates a block in NHEJ (Fig. 3A) and is consistent with the phenotype observed in the NHEJ-deficient Xrcc4^−/− B cells (24). Collectively, this would suggest that MSH2 may enforce blunt end joining in the absence of NHEJ (24). Previously, Stavnezer and Schrader (71) proposed a model wherein MSH2 functions to convert staggered SSBs into DSBs through end processing of distal C residues on opposite strands, either by 3′ loading of EXO1 to the MMR complex (72, 73) or through the recruitment of DNA polymerase η by MSH2/MSH6 to synthesize nucleotides at resected SSBs until a blunt DSB has been reached (74). Following this model of MSH2-dependent DSB conversion, perhaps the presence or absence of ATM determines the density of DSBs within a recombining S region, possibly through the recruitment of APE1 by ATM-dependent phosphorylation of AID (18). Thus, in the absence of ATM fewer DSBs are generated, constituting the need for MSH2 blunt end joining. Speculatively, NHEJ and A-EJ may function suboptimally in ATM- and MSH2-deficient B cells to create a heterogeneous phenotype in end joining where the absence of ATM and MSH2 exerts an epistatic suppression of NHEJ, which may account for the reduction in blunt joins.

Alternatively, other NHEJ proteins may compensate for the absence of ATM (61, 75). During V(D)J recombination, which utilizes NHEJ at conserved recombination signal sequences cleaved by RAG1/2, NHEJ is driven mostly by DNA-PKcs (76–78). ATM participates in the repair of RAG-induced DSBs; however, it is insufficient to rescue NHEJ during V(D)J recombination in the absence of DNA-PKcs (79, 80). Interestingly, MSH2 is dispensable for NHEJ in V(D)J recombination despite its known involvement in CSR (81), suggesting that although both CSR and V(D)J use NHEJ, there remain uncharacterized mechanistic differences between the processes. In CSR, DNA-PKcs can partially function in NHEJ in the absence of ATM (82), which could account for the lack of significant difference in the percentage of short stretches of microhomologies (1–4 nt) in Atm^D/DMsh2^−/− B cells as compared with the control genotypes (Fig. 3). However, the extent to which DNA-PKcs can compensate for the absence of ATM may depend on the amount of resection at the broken ends. When limited resection of DSB ends has occurred, the Ku70/80 complex efficiently binds the DSB (83), which allows for recruitment of DNA-PKcs, Artemis, XRCC4, and LigIV to efficiently join the broken ends through NHEJ (84, 85). Because the Ku complex has low affinity for resected DNA (86), the initiation of resection is a critical determinant for repair pathway choice. Interestingly, there is evidence to suggest that A-EJ driven by resection occurs in CSR. RPA, a ssDNA-binding protein that regulates DNA resection (87, 88), interacts with pS38-AID at S regions (19). Furthermore, B cells defective in 53BP1 exhibit an increase in ATM-mediated end resection that is repaired through A-EJ during CSR (89). Interestingly, Atm^D/DMsh2^−/− B cells show an increased percentage of Sμ–Sγ1 junctions with insertions, many of which are some of the longest in length (up to 380 bp) across all of the assayed genotypes (Fig. 3B). This phenotype is comparable to that observed in the absence of 53BP1, suggesting that Atm^D/DMsh2^−/− insertions may have an increase in end resections that have not been repaired through A-EJ. An alternative explanation for the increased insertions could be that the absence of ATM and MSH2 leads to an increase in resection with no mechanism to repair through NHEJ or A-EJ and thus requires the need for a salvage pathway. Single strand annealing (SSA) has been observed in yeast in the case of resected DNA homology-based repair but in the absence of functional HR, leading to high rates of mutagenesis (90–92). In this way, the increase in insertions observed in Atm^D/DMsh2^−/− B cells may indicate an attempt at salvage repair through SSA. Additionally, members of the SAGA (Spt-Ada-Gcn5-acetyltransferase) deubiquitylation complex such as USP22, ENY2, and ATXN7 may modulate NHEJ, A-EJ, and SSA during CSR (93); however, the role of ATM and MSH2 in regulating SAGA remains unknown. Collectively, this may indicate that ATM and MSH2 are required to repair resected DNA to drive A-EJ (Fig. 4).

We hypothesize that pS38-AID–mediated interaction with RPA-coated resected ssDNA may signal through ATM in a positive feedback loop to increase the density of resected DNA breaks in recombining S regions (18, 94). We propose a model whereby the phosphorylation of AID through an uncharacterized ATM–protein kinase A signaling cascade activates A-EJ in CSR (Fig. 4). Hypothetically, in the absence of resection (i.e., blunt end joins), the Ku complex efficiently binds the broken ends and recruits DNA-PKcs–mediated NHEJ. However, when the DSB ends have been resected, NHEJ is driven by ATM. In the absence of ATM, MSH2 promotes the conversion of staggered SSBs into DSBs through the recruitment of either EXO1 or polymerase η to generate blunt ends that can be joined through NHEJ. We speculate that activation of ATM indirectly induces the phosphorylation of S38 on AID, which in turn may cointeract with resected SSB coated by RPA (95), facilitating microhomology-mediated strand invasion into the acceptor S region to complete repair through A-EJ. In the absence of ATM and MSH2, resected DNA may also be repaired through SSA due to the extensive resection.

The authors have no financial conflicts of interest.

Mouse husbandry and colony maintenance was supported by Harry Acosta, Alfredo Diaz, Anthony Pacheco, and Iris Falquez from the City College of New York Marshak Animal Care Facility. Thank you to Drs. Alicia Meléndez, Christine Li, Scott Keeney, and Linda Spatz for providing intellectual support, and Dr. Allysia Matthews, Dr. Simin Zheng, and Sarah (Jee Eun) Choi for critical reading of the manuscript.