The ataxia telangiectasia mutated kinase controls Igκ allelic exclusion by inhibiting secondary Vκ-to-Jκ rearrangements.

Allelic exclusion is enforced through the ability of antigen receptor chains expressed from one allele to signal feedback inhibition of V-to-(D)J recombination on the other allele. To achieve allelic exclusion by such means, only one allele can initiate V-to-(D)J recombination within the time required to signal feedback inhibition. DNA double-strand breaks (DSBs) induced by the RAG endonuclease during V(D)J recombination activate the Ataxia Telangiectasia mutated (ATM) and DNA-dependent protein kinase (DNA-PK) kinases. We demonstrate that ATM enforces Igκ allelic exclusion, and that RAG DSBs induced during Igκ recombination in primary pre-B cells signal through ATM, but not DNA-PK, to suppress initiation of additional Igκ rearrangements. ATM promotes high-density histone H2AX phosphorylation to create binding sites for MDC1, which functions with H2AX to amplify a subset of ATM-dependent signals. However, neither H2AX nor MDC1 is required for ATM to enforce Igκ allelic exclusion and suppress Igκ rearrangements. Upon activation in response to RAG Igκ cleavage, ATM signals down-regulation of Gadd45α with concomitant repression of the Gadd45α targets Rag1 and Rag2. Our data indicate that ATM kinases activated by RAG DSBs during Igκ recombination transduce transient H2AX/MDC1-independent signals that suppress initiation of further Igκ rearrangements to control Igκ allelic exclusion.

Assembly of Ig and TCR genes from variable (V), diversity (D), and joining (J) segments is the pervasive means by which antigen receptor (AgR) diversity is generated (Brady et al., 2010). V(D)J recombination is initiated by the RAG1/RAG2 (RAG) endonuclease that induces DNA double-strand breaks (DSBs) adjacent to participating segments (Schatz and Ji, 2011) and completed by DSB repair factors that process V(D)J coding ends (CEs) into coding joins (CJs; Helmink and Sleckman, 2012). AgR assembly occurs during and is required for lymphocyte differentiation. IgH genes are assembled through DH-to-JH recombination, followed by VH-to-DJH rearrangements on one allele at a time in pro–B cells (Rajewsky, 1996). IgH chains expressed from in-frame VHDJH joins can bind λ5/Vpre–B chains to form pre-BCRs that signal inhibition of VH rearrangements, proliferation, and differentiation into pre–B cells (Rajewsky, 1996). The two-thirds of cells that assemble out-of-frame VHDJH joins can attempt to assemble in-frame VHDJH joins on the second allele (Rajewsky, 1996). Igκ genes are assembled from Vκ and Jκ segments on one allele at a time in G1 phase pre–B cells (Rajewsky, 1996). Igκ chains expressed from VκJκ joins can bind IgH chains to form κ+ BCRs that are subject to selection (Rajewsky, 1996; Nemazee, 2006). Non-autoreactive BCRs signal inhibition of Igκ recombination and differentiation into B cells (Nemazee, 2006). Autoreactive BCRs induce additional Igκ rearrangements that replace VκJκ complexes, a process known as Igκ editing (Nemazee, 2006). Pre–B cells that assemble out-of-frame VκJκ joins can attempt to assemble in-frame VκJκ joins on the other allele (Rajewsky, 1996).

Most lymphocytes express surface AgR chains from a single allele. Allelic exclusion is enforced through the ability of Ig and TCR chains expressed from one allele to signal feedback inhibition of V-to-(D)J rearrangements on the second allele (Brady et al., 2010; Vettermann and Schlissel, 2010). To achieve allelic exclusion, only one allele can initiate V-to-(D)J recombination in the time required for feedback inhibition. V-to-(D)J recombination requires CTCF-mediated looping between RAG accessible V segments and RAG-bound D/J segments (Guo et al., 2011; Schatz and Swanson, 2011). In pre–B cells, Igκ loci replicate asynchronously and the early replicating allele is preferentially rendered accessible and selected for Igκ recombination (Mostoslavsky et al., 2001). The time between replication of Igκ loci might be sufficient to enable Igκ chains from the first allele to prevent Igκ rearrangements on the second allele. Yet experiments that show Igκ allelic exclusion is enforced by asynchronous replication between Igκ alleles have not been reported.

The feedback model for allelic exclusion hypothesized that V(D)J recombination could activate transient intracellular signals that inhibit recombination on the second allele (Alt et al., 1980). RAG DSBs activate DNA-dependent protein kinase (DNA-PK), which forms an endonuclease with Artemis that processes CEs (Ma et al., 2002). RAG DSBs also activate Ataxia Telangiectasia mutated (ATM), which phosphorylates proteins to coordinate the cellular DSB response (Bredemeyer et al., 2006, 2008). In pre–B cells, RAG DSBs signal through ATM to initiate a genetic program that controls differentiation (Bredemeyer et al., 2008). ATM promotes high-density histone H2AX phosphorylation along RAG-cleaved loci (Savic et al., 2009). H2AX phosphorylation creates binding sites for MDC1, which retains activated ATM kinases around DSBs (Lou et al., 2006). The pools of activated ATM bound and not bound to H2AX/MDC1 exhibit different signaling capabilities (Celeste et al., 2002; Lou et al., 2006). In G1 phase cells, ATM promotes CJ formation independent of H2AX and MDC1 (Bredemeyer et al., 2006; Yin et al., 2009; Helmink et al., 2011). H2AX phosphorylation is detectable over only one Igκ locus in most pre–B cells including those with paired Igκ alleles (Hewitt et al., 2009). The fraction of cells with H2AX phosphorylation over both Igκ loci is fivefold higher in Atm mice relative to wild-type mice (Hewitt et al., 2009), suggesting that Igκ recombination initiates on a single paired allele and ATM bound to this allele acts on the other allele to prevent bi-allelic rearrangements (Hewitt et al., 2009). Detection of bi-allelic Igκ chromosome breaks in Atm-deficient pre–B cell lines provided support for this model (Hewitt et al., 2009).

Here, we show in mice that inactivation of ATM causes a higher frequency of B cells expressing surface Igκ chains from both alleles. We show in primary pre–B cells that DSBs induced during Igκ recombination signal through ATM, but not DNA-PK, to suppress further Igκ rearrangements. Neither H2AX nor MDC1 is required for the ability of ATM to enforce Igκ allelic exclusion or inhibit Igκ recombination. Upon activation in response to RAG DSBs, ATM signals down-regulation of Gadd45α with concomitant repression of the Gadd45α targets Rag1 and Rag2. Our data indicate that ATM kinases activated by Igκ cleavage transduce transient H2AX/MDC1-independent signals that suppress further Igκ rearrangements and thereby enforce Igκ allelic exclusion.

RESULTS AND DISCUSSION

To determine whether ATM enforces Igκ allelic exclusion, we used Igκ mice that express an allotypic marker that enables analysis of surface Igκ expression from each allele using anti-Igκm and anti-Igκh antibodies (Casellas et al., 2001). Because ∼1/3 of CJs are assembled in-frame and Igκ chains are required for differentiation, 20% is the maximal frequency of B cells that can exhibit Igκ allelic inclusion, assuming simultaneous Igκ rearrangements on both alleles and no Igκ feedback inhibition and selection for/against dual-Igκ+ cells (Alt et al., 1984). However, as a result of the CJ formation defect in Atm cells (Bredemeyer et al., 2006), <20% of Atm-deficient Igκ B cells should express surface Igκ chains from both alleles if ATM controls Igκ allelic exclusion. Because a small fraction of Igκ B cells exhibits Igκ allelic inclusion (Casellas et al., 2001; Velez et al., 2007; Fournier et al., 2012), we bred Igκ and Mb1-creAtm mice to more readily generate Mb1-creAtm mice with B lineage–specific Atm inactivation and littermate control Atm mice (frequency of 0.5) compared with Atm and Atm littermate mice (frequency of 0.25). We observed equal high-level expression of Igκm and Igκh on 1.98 ± 0.14% of bone marrow and 3.68 ± 0.10% of splenic B cells from Mb1-creAtm mice (Fig. 1). The frequencies of cells expressing high levels of surface Igκm and Igκh were ∼2.5-fold greater in Mb1-creAtm mice as compared with littermate Igκ mice (bone marrow: 0.92 ± 0.04 vs. 1.98 ± 0.14%, P < 0.0001; spleen: 1.48 ± 0.06 vs. 3.68 ± 0.10%, P < 0.0001) and age-matched Mb1-creAtm mice (Fig. 1). These data demonstrate that Igκ allelic exclusion is not enforced as strictly in Atm-deficient B cells.

Figure 1.

ATM enforces (A) Analysis of Igκ and Igκ expression on bone marrow and splenic B cells isolated from the indicated mice. Shown are the gates used to quantify B cells expressing Igκ and Igκ at equal high levels and the percentages of cells in these gates. (B) Graph showing the mean frequencies of Igκ allelic inclusion in bone marrow and splenic B cells of the indicated mice. Data are from four independent experiments conducted on five Igκ, six Mb1-creAtm littermates, and three age-matched Mb1-creAtm mice. Error bars are SEM. ***, P < 0.0001.

Less strict enforcement of Igκ allelic exclusion could be a result of loss of signals that coordinate initiation of Igκ rearrangements between alleles and/or slower CJ formation on the first allele that leads to initiation of Igκ recombination on the second allele before feedback inhibition. To assess if ATM enforces Igκ allelic exclusion by transducing signals from RAG DSBs on one allele to suppress recombination on the other allele, we needed to quantify initiation of Igκ rearrangements independent of ATM functions in CJ formation. For this purpose, we used an approach that we developed to elucidate responses to RAG DSBs in primary pre–B cells (Bednarski et al., 2012). We generated Rag1 or Artemis mice expressing IgH and BCL2 transgenes (Rag1 and Artemis mice); these mice each exhibit a block in B cell development at the pre–B cell stage caused by inability to assemble Igκ genes (Bednarski et al., 2012).

Southern analysis of Jκ cleavage can be used to quantify initiation of Igκ rearrangements (Fig. 2 A). Because IL7 promotes pre–B cell proliferation and represses Igκ accessibility and RAG expression (Billips et al., 1995; Amin and Schlissel, 2008; Johnson et al., 2008; Malin et al., 2010), culturing Rag1 or Artemis bone marrow in IL7 leads to outgrowth of pre–B cells with germline (GL) Igκ loci (Fig. 2 B). Yet a fraction of IL7-cultured Artemis cells accumulates Igκ CEs because IL7 does not block Igκ accessibility and RAG expression (Fig. 2 B). Removal of Rag1 or Artemis cells from IL7 leads to G1 arrest and stimulation of Igκ accessibility and RAG expression (Bednarski et al., 2012); Rag1 cells retain GL Igκ loci, whereas Artemis cells lose GL Igκ loci and accumulate Jκ1 and Jκ2 CEs (Fig. 2, B and C). These Igκ DSBs activate ATM-mediated signals, as indicated by phosphorylation of p53 after removal of Artemis but not Rag1 cells from IL7 and by inhibition of these responses by the KU55933 ATM kinase inhibitor (Bednarski et al., 2012).

Figure 2.

ATM inhibits (A) Schematic of the GL Jκ locus and EcoR1/SacI fragments of Jκ loci in GL configuration or with CEs for the functional Jκs. Indicated are the sizes of the restriction fragments and the probe used to visualize these. (B–I) Southern blot analyses (B, D, G, and H) and quantification of Jκ cleavage (C, E, G, and I) in the indicated pre–B cells before and at indicated times after IL7 removal or IL7 removal and KU55933 addition. GL Jκ loci and Jκ CE bands are labeled. Southern blots with a TCRβ probe for DNA content are also shown. Data are from three independent experiments. Error bars represent SEM. Asterisks (*, P = 0.05–0.01; **, P = 0.001–0.01; ***, P < 0.0001) indicate Jκ cleavage that is significant from the comparable time in untreated Artemis (C, E, and G) or Scid:IgH:BCL2 (I) cells.

Because CEs are not processed into CJs in Artemis cells (Bednarski et al., 2012), these cells enable us to assess if ATM signals inhibition of Igκ recombination independent of ATM function in CJ formation. We grew Artemis pre–B cells in IL7, removed IL7, placed cells back into culture in the absence or presence of KU55933, isolated DNA at 0, 48, and 72 h after IL7 removal, and then conducted Southern blots. We detected GL Jκ loci and a low level of Jκ1 CEs in IL7-cultured cells (Fig. 2, D and E). At 48 and 72 h after IL7 removal, we observed decreased GL Jκ loci, increased levels of Jκ1 CEs, and appearance of Jκ2 CEs (Fig. 2, D and E). In contrast, at 48 and 72 h after IL7 removal and KU55933 addition, we detected near loss of GL Jκ loci, higher levels of Jκ1 CEs, and appearance of Jκ2, Jκ4, and Jκ5 CEs (Fig. 2, D and E). To rule out off-target KU55933 effects, we analyzed Jκ cleavage in Atm cells. As compared with IL7-cultured Artemis cells, we found fewer GL Jκ loci and higher levels of Jκ CEs in Atm cells cultured in IL7 (Fig. 2, B and C). We detected near complete loss of GL Jκ loci and increased levels of Jκ4 and Jκ5 CEs after IL7 removal from Atm cells (Fig. 2, B and C). The relative levels of GL Jκ loci and Jκ CEs indicate that RAG DSBs were induced on roughly half the Igκ alleles in Artemis cells, and on almost all Jκ alleles in the absence of ATM. These data show that RAG DSBs induced during Igκ recombination signal through ATM to suppress initiation of further Igκ rearrangements.

Because DNA-PK is activated by DSBs, we also assessed if DNA-PK inhibits Igκ recombination. We observed no changes in levels of GL Jκ loci or Jκ CEs after IL7 withdrawal of Artemis cells in the presence or absence of the NU7026 DNA-PK kinase inhibitor (Fig. 2, F and G). Because Artemis phosphorylation by DNA-PK is required for CJs (Ma et al., 2002), Scid cells that lack DNA-PK are equal to Artemis cells with respect to CE accumulation (Savic et al., 2009). We observed similar loss of GL Jκ loci and elevated levels of Jκ1 and Jκ2 CEs after IL7 removal from Scid:IgH:BCL2 cells (Fig. 2, H and I), relative to Artemis cells (Fig. 2, B and C). Yet we detected near loss of GL Jκ loci, increased levels of Jκ1 and Jκ2 CEs, and appearance of Jκ4 CEs after IL7 removal and KU55933 addition to Scid:IgH:BCL2 cells (Fig. 2 H, I). These data demonstrate that ATM, but not DNA-PK, transduces signals that inhibit Igκ rearrangements.

H2ax and Mdc1 cells exhibit impaired ATM-mediated activation of some checkpoints (Lou et al., 2006) but normal ATM-mediated activation of other checkpoints (Celeste et al., 2002), indicating that the pools of ATM bound and not bound to H2AX/MDC1 exhibit different signaling capabilities. To determine if the ATM kinases bound to H2AX/MDC1 along RAG-cleaved Igκ loci enforce Igκ allelic exclusion, we analyzed Igκ, Atm, H2ax, and Mdc1 mice. We found equal high levels of Igκm and Igκh expression on bone marrow and splenic B cells from Igκ, H2ax, and Mdc1 mice (Fig. 3 A). The frequencies of cells exhibiting Igκ allelic inclusion in H2ax and Mdc1 mice were lower than in Atm mice (Fig. 3 A). To assess whether the soluble pool of activated ATM kinases controls initiation of Igκ rearrangements between alleles, we analyzed Jκ cleavage in Artemis, H2ax, and Mdc1 pre–B cells. Because H2AX and MDC1 protect CEs from degradation (Helmink et al., 2011), we quantified Igκ cleavage by calculating loss of GL Jκ band intensity relative to DNA content. We found that cleavage occurred on roughly half of the Igκ alleles in Artemis, H2ax, and Mdc1 cells (Fig. 3, B and C) rather than on almost all Igκ alleles as observed in Atm cells (Fig. 2 C). These data demonstrate that ATM kinases activated by Igκ cleavage transduce H2AX/MDC1-independent signals that suppress further Igκ rearrangements and thereby enforce Igκ allelic exclusion.

Figure 3.

ATM controls Igκ recombination and allelic exclusion independent of H2AX and MDC1. (A) Graphs showing the mean frequencies of Igκ allelic inclusion in bone marrow and splenic B cells isolated from the indicated mice. Data are from at least three independent experiments conducted on five Igκ mice, three Atm, six H2ax, and seven Mdc1 age-matched mice. Error bars represent SEM. *, P = 0.05–0.01; **, P = 0.001–0.01; ***, P < 0.001. (B and C) Southern analyses (B) and graphs showing quantification (C) of Jκ cleavage in the indicated pre–B cells before and at indicated times after IL7 withdrawal. GL Jκ loci and Jκ CE bands are labeled. Southern blots with a TCRβ probe for DNA content are also shown. Data are from three independent experiments conducted on four Artemis, three H2ax, and two Mdc1 cultures. Artemis and H2ax error bars represent SEM and Mdc1 error bars represent SD.

We previously showed that RAG DSBs signal through ATM to increase levels of many mRNAs, including several encoding lymphocyte-specific proteins (Bredemeyer et al., 2008). To test if RAG DSBs signal through ATM to decrease expression of proteins necessary for Igκ recombination, we quantified Rag1 and Rag2 mRNA and protein in Artemis cells before and after IL7 removal, both in the absence or presence of KU55933. We found that ATM inhibition led to greater levels of Rag1 and Rag2 mRNA and protein after IL7 removal (Fig. 4, A, B, and E). In pre–B cells, the Gadd45α stress-regulated protein transduces signals that promote Rag1/Rag2 transcription and maintain basal RAG expression (Amin and Schlissel, 2008). We found that ATM inhibition also led to higher levels of Gadd54α mRNA and protein after IL7 removal and accumulation of Jκ CEs in Artemis cells (Fig. 4, C, D, and E). These data demonstrate that RAG DSBs induced in pre–B cells activate and signal through ATM kinases to transcriptionally down-regulate RAG expression.

Figure 4.

ATM represses RAG expression. (A and B) Graphs showing the mean levels of Rag1 or Rag2 mRNA (A) or protein (B) relative to β-actin mRNA or protein in Artemis cells before and at indicated times after IL7 withdrawal or IL7 withdrawal and addition of ATM inhibitor KU55933. (C and D) Graphs showing the mean levels of Gadd45α mRNA (C) or protein (D) relative to β-actin mRNA or protein in Artemis cells before and at indicated times after IL7 withdrawal or IL7 withdrawal and KU55933 addition. (A–D) Data were generated from three independent experiments. Error bars represent SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001. (E) Western blot data from an experiment in B and D. (F) Model showing how the ATM-dependent RAG negative feedback loop suppresses further Igκ rearrangements on either allele to control Igκ allelic exclusion and ensure broad utilization of Jκ segments in selected Igκ chains.

We conclude that ATM controls Igκ recombination and allelic exclusion through a negative feedback loop. Initial RAG cleavage on the early replicating Igκ allele induces ATM-dependent signals that repress RAG activity to provide time for selection of Igκ genes before further Igκ recombination on either allele. This mechanism would inhibit additional Igκ recombination on both alleles regardless of Igκ locus positioning. Non-autoreactive BCRs would signal RAG repression and feedback inhibition to enforce Igκ allelic exclusion (Nemazee, 2006), whereas autoreactive BCRs would induce RAG expression and additional Igκ rearrangements on either allele to promote Igκ editing and cause Igκ allelic inclusion (Casellas et al., 2001). ATM inactivation after assembly of out-of-frame CJs would lead to RAG reexpression and reinitiation of Igκ recombination on either allele. In addition to controlling allelic exclusion, the ability of ATM to repress RAG activity in response to RAG cleavage likely slows initiation of secondary Igκ rearrangements to ensure broad utilization of Jκ segments in selected Igκ chains. This ATM-dependent negative feedback loop also likely enforces IgH and TCR-β allelic exclusion, ensures broad utilization of D/J segments in other AgR chains, and controls recombination among loci in pro–T cells. ATM also functions in a negative feedback loop to repress the SPO11 endonuclease and limit the number of DSBs during meiosis (Lange et al., 2011), suggesting that pressure to regulate DSB induction during programmed recombination events selected for the evolution of ATM-dependent mechanisms that feedback inhibit tissue-specific nuclease activity.

MATERIALS AND METHODS

Mice.

All mice were bred and housed under specific pathogen-free conditions at the Children’s Hospital of Philadelphia or the Washington University School of Medicine. All mice were on a 129/B6 mixed background. All analyses were conducted on littermate or age-matched mice. All experiments were performed in accordance with national guidelines and regulations and approved by the Institutional Animal Care and Use Committees of both institutes.

Primary pre–B cell culture.

Bone marrow cells were harvested from 4–6-wk-old mice and cultured in media containing IL7 at 5 ng/ml for 7–10 d at 2 × 106 cells/ml. For IL7 withdrawal experiments, cells were harvested, washed in PBS, resuspended, and maintained at 2 × 106 cells/ml in IL7-free media in the presence or absence of 15 µM ATM (KU-55933; Sigma-Aldrich) or 20 µM DNA-PKs (NU7026; Sigma-Aldrich) inhibitors.

Southern blots.

Southern blot analyses of RAG Igκ cleavage were performed as previously described (Bednarski et al., 2012). Quantification of loss of Igκ GL bands was conducted using ImageJ software (National Institutes of Health) and referenced to TCRβ loading control bands.

Flow cytometric analysis.

Cells isolated from the bone marrow and spleens of 6–8-wk-old mice were stained with the following antibodies: APC/Cy7 anti–mouse B220 (RA3-6B2; BD), APC anti–mouse TCR-β (H57-597; BD), PE anti–mouse κ (187.1; BD), and biotin anti–human κ (SouthernBiotech). Cell suspensions were depleted of red blood cells with NH4Cl lysis buffer and stained in PBS containing 3% FCS and 0.25 mM EDTA. Before staining, FC receptors were blocked using anti-CD16/CD32 (2.4G2; BD). Data were collected on an LSR II and analyzed with FlowJo software (Tree Star). Surface Igκ expression was assayed on single, DAPI−, TCR-β−, and B220+ cells. Single, live cells were gated on the basis of forward and side scatter and DAPI exclusion (D1306; Invitrogen).

qRT-PCR.

RNA was isolated as previously described (Bednarski et al., 2012). Each reaction was run in triplicate. Rag1 and Rag2 mRNA levels were normalized to β-actin mRNA levels. The primers used were: Rag1 forward, 5′-TGGGAATCGTTTCAAGAGTGAC-3′, and Rag1 reverse, 5′-CATCTGCCTTCACGTCGATCC-3′; Rag2 forward, 5′-ACACCAAACAATGAGCTTTCCG-3′, and Rag2 reverse, 5′-CCGTATCTGGGTTCAGGGAC-3′; Gadd45α forward, 5′-CCGAAAGGATGGACACGGTG-3′, and Gadd45α reverse, 5′-TTATCGGGGTCTACGTTGAGC-3′; or β-actin forward, 5′-TCATCACTATTGGCAACGAGCGGTTC-3′, and β-actin reverse, 5′-TACCACCAGACAGCACTGTGTTGGCA-3′.

Western blots.

Whole cell lysates isolated from 30 × 106 IL7 withdrawn pre–B cells were run on 10% Tris-Glycine gels (Novex), transferred to PVDF membrane (Immobilon; Millipore), blocked in Odyssey Blocking Buffer (LI-COR Biosciences) and probed with the following monoclonal antibodies: rabbit anti–mouse Rag1 #23 and rabbit anti–mouse Rag2 #39 (Coster et al., 2012), rabbit anti–mouse Gadd45α (D17E8; Cell Signaling Technology), rabbit anti–mouse β-actin (Sigma-Aldrich), and IRDye 800CW goat anti–rabbit IgG (LI-COR Biosciences). Western blots were visualized and quantified using the Odyssey Infrared Imaging System (LI-COR Biosciences). Rag1, Rag2, and Gadd45α protein levels were normalized to β-actin protein levels.

Statistical analyses.

All p-values were generated by Student’s t test using Prism (GraphPad Software).