Generating and repairing genetically programmed DNA breaks during immunoglobulin class switch recombination

Role of germline transcription in generating AID substrates

Ig heavy chain constant (C_H) exons are organized as independent transcriptional units composed of a cytokine-inducible promoter upstream of a non-coding “I-exon”, the intronic S region, and the corresponding C_H exons¹⁸. T cell–dependent (for example, cytokines and CD40L) or T cell–independent (for example, lipopolysaccharide) stimuli (or both) activate transcription of recombining S regions (Figure 1), which is absolutely required for CSR. The primary germline transcript is spliced into a mature, polyadenylated transcript with no known protein product and is frequently referred to as a “sterile” germline transcript¹⁹. Genetic deletion of specific I-exons abolishes germline transcription and CSR to the corresponding isotype^20,21. Germline transcription initiating from the I-exons and proceeding through the S regions to the C_H exons creates the ssDNA substrates for AID within the transcribed S regions. Each S region varies in length (1–10 kb) and consists of tandem repetitive units that contain a G-rich non-template strand. Deleting the repetitive units within the S regions or replacing the S regions with small core S-region sequences significantly impairs CSR and demonstrates an essential role for these sequences during CSR^22–25. Recent data suggest that the repetitive, G-rich non-template strand forms G-quadruplex (G4) structures that facilitate cooperative AID oligomerization at S regions²⁶. In addition, the tandem repeats of 5′-AGCT-3′ within the core S regions recruit AID and its kinase, protein kinase A (PKA), to the S regions via the 14-3-3 adaptor proteins, which specifically recognize the 5′-AGCT-3′ repeats (Xu et al., 2010²⁷; reviewed in Xu et al., 2012²⁸).

Germline transcription of S regions creates R-loops, wherein the newly transcribed RNA hybridizes to the template DNA to form a stable RNA:DNA hybrid that exposes the non-template DNA as ssDNA, which is the substrate for AID^24,29–32. Inversion of the mouse Sγ1 sequence, which converts the G-rich non-template strand to a G-rich template strand, impairs R-loop formation and CSR without affecting germline transcription²⁴. These data demonstrate the inherent ability of G-rich S regions to form R-loops that likely contain G4 structures, which facilitate AID recruitment^26,33.

Although R-loop formation at non-Ig loci may redirect AID activity to other regions of the B cell genome³⁴, AID is significantly enriched at the Igh locus, suggesting that factors beyond R-loop formation also restrict AID to the Igh locus during CSR^14,35. AID interacts with RNA polymerase II and its associated proteins, such as Spt5³⁶, PAF1³⁷, and the FACT histone chaperone complex³⁸. In addition, RNA polymerase II, which has stalled at the repetitive G/C-rich S regions, can recruit the RNA exosome to degrade the nascent RNA transcript and facilitate AID deamination of the non-template and template DNA strands^39,40.

Role for germline transcripts in targeting AID to S regions

Germline transcription is necessary but not sufficient for CSR. In mice that lack the Iγ1 exon splice donor site, CSR to IgG1 was abolished despite active S-region transcription^41–43, suggesting that either the RNA processing machinery (for example, spliceosome) or the processed transcripts are required for CSR. CTNNBL1, a component of the spliceosome, interacts with AID and is required for CSR and SHM⁴⁴. Knockdown of the splicing regulator PTBP2 reduces AID at S regions and impairs CSR^45,46. These data demonstrate that the spliceosome plays an essential role in localizing AID to S regions.

Sα RNA expression from a plasmid in trans enhanced CSR to IgA in the B-cell line Bcl₁B₁⁴⁷, suggesting that spliced, intronic S-region RNA derived from germline transcripts have a functional role during CSR. More recently, these S-region RNAs were shown to recruit AID to S-region DNA sequences^33,48. Intronic switch RNAs were known to be spliced from primary transcripts to form lariats that undergo hydrolytic degradation, which is catalyzed by the debranching enzyme DBR1⁴⁹. Knockdown of DBR1 in CH12F3 cells reduces CSR; however, expression of switch RNAs in trans bypasses the lariat debranching step in DBR1 knockdown cells to rescue both CSR and AID recruitment to S regions in a sequence-specific manner^33,50. In addition, AID bound directly and selectively to sense S-region transcripts, which were shown to form highly stable four-stranded G4 structures^33,50. A putative G4 RNA binding motif in AID was identified and mutations in this domain abrogated AID interactions with G4 switch RNA and consequently the localization of AID to S regions and wild-type levels of CSR^33,50. Interestingly, a mutation in the RNA binding motif of AID (G133V) has been identified in patients with Hyper-IgM Syndrome who show severe CSR defects⁵¹. From these studies, a new regulatory model for AID localization to S regions was proposed whereby the non-coding, intronic S-region RNA, which is produced following germline transcription and splicing, binds to AID to target AID to sites of DNA recombination (S regions) and promote AID-mediated DNA deamination and CSR in a sequence-specific manner. This model connects the data demonstrating the role of germline transcription and splicing in CSR with the binding of AID to S-region RNA and DNA and identifies a critical function for S-region RNA in CSR beyond germline transcription and splicing (Figure 2).

Figure 2. Proposed model for RNA-dependent targeting of AID during class switch recombination.

Upon B-cell activation, germline transcription is initiated from a cytokine-inducible promoter (P) and primary germline transcripts are generated from the I-S-C_H sequences, which encode the I-exon, switch (S) region, and constant coding exons (C_H). These transcripts are spliced to form a mature non-coding, germline transcript and an intronic S lariat. The latter is further processed by the debranching enzyme DBR1 to form a linear S-region transcript. Linear S transcripts fold into G-quadruplex RNA, which is bound by AID. The complex of S-RNA and AID is guided to transcribed S-region DNA as a result of the complimentary between the S-RNA and the transcribed S region. AID, activation-induced cytidine daminase; DBR1, debranching RNA lariats 1.

Epigenetic regulation of AID localization to S regions

Epigenetic modifications of the Igh locus during CSR have been proposed to control the recruitment of AID to S regions (comprehensively reviewed in 52). Changes in histones H4 and H3 methylation patterns have been associated with altered levels of CSR, although the functional significance of these modifications during CSR remains unclear. Conditional deletion of two methyltransferases (Suv4-20h1 and Suv4-20h2), which are responsible for histone H4 lysine 20 di- and tri-methylation (H4K20me2 and H4K20me3), in B cells leads to a 50% reduction in CSR⁵³. AID interacts with SUV4-20H1 and SUV4-20H2 in 293F cells and localizes these methyltransferases to S regions to promote SUV4-20-mediated histone trimethylation in B cells undergoing CSR⁵⁴, suggesting cooperative targeting of methyltransferases and AID to recombining S regions. H3K9me3 tethers AID to Sμ through its interaction with KRAB domain-associated protein 1 (KAP1) and heterochromatin protein 1 (HP1)⁵⁵, and combinatorial phospho-Ser10 and acetyl-Lys9 modification of H3 (H3K9acS10ph) mediates AID recruitment to S regions by stabilizing S-region DNA binding of 14-3-3, which in turn interacts with AID⁵⁶. Enrichment of H3K9me3, histone H3 lysine 9 acetylation (H3K9ac), and histone H3 lysine 4 trimethylation (H3K4me3) at recombining, transcribed S regions^57–59 and a reduction in CSR in B cells deficient in Pax interaction with transcription-activation domain protein-1 (PTIP), which is responsible for H3K4 methylation, suggest additional epigenetic mechanisms of regulating AID localization to S regions⁵⁷. However, the functional relevance of H4K20 and H3K9 methylation in the recruitment of AID to S regions during CSR remains unclear, as some data demonstrate that H3K9 tri-methylation and H4K20 methylation (mono-, di-, and tri-methylation) are reduced at recombining S regions^55,60. Additional work is required to decipher the epigenetic code at S regions, which will further elucidate the role of post-translational modification of histones in the localization and stabilization of AID at S regions during CSR.

Converting deaminated DNA into DSBs by BER and MMR

CSR requires BER and MMR pathways to generate DNA breaks in recombining S regions. Defects in either BER or MMR alone significantly impair CSR, whereas combined BER and MMR deficiency (for example, UNG^−/−MSH2^−/−) completely blocks CSR in vitro and in vivo^9,63. In the BER pathway for CSR, AID-generated dU in S-region DNA is removed by uracil DNA glycosylase (UNG) to generate an abasic site, which is cleaved by the apurinic/apyrimidinic endonuclease (APE1) to create a single-strand break (SSB) in the DNA^9,64–66 (Figure 1). Adjacent SSBs on complementary DNA strands constitute a DSB, which is an obligate intermediate in CSR. Human and mouse B cells with inactivating mutations in UNG exhibit impaired DSB formation at S regions and a severe block in CSR^66–69. Impaired recruitment of UNG to recombining S regions in Rev1-deficient B cells reduces CSR in vitro and in vivo⁷⁰. Likewise, mice heterozygous for an APE1 null mutation and CH12F3 cells with a homozygous deletion of APE1 have significantly diminished CSR^71–73.

The AID-generated dU:dG mismatch can also be processed into SSBs through MMR⁹. In this pathway, an MSH2-MSH6 heterodimer recognizes the dU:dG mismatch and recruits a complex of MLH1/PMS2/EXO1 to repair the mismatch (Figure 1). PMS2 (PMS1 homolog 2) generates a SSB distal to the mismatch and subsequently exonuclease 1 (EXO1) converts the DNA breaks into ssDNA gaps by excising the segment of the DNA containing the dU in a 5′-to-3′ direction³. EXO1 excision of dU-containing sequences on opposite DNA strands thus would generate DSBs that are required for CSR^18,74. Consistent with this proposed role for PMS2 and EXO1 in converting deaminated S regions into DSBs, humans or mice with inactivating mutations in PMS2 or EXO1 have significant impairments in CSR because of defects in DSB formation in S regions^75,76. PMS2- and EXO1-mediated excision of dU:dG mismatched DNA creates a DSB with a 5′ overhang that can be resolved into a blunt (or nearly blunt) DSB by DNA polymerases (η and θ), which subsequently is used by proteins of the C-NHEJ and A-EJ pathways to complete CSR^74,77,78.

Positive feedback loop to amplify DNA breaks through AID phosphorylation

Despite the overwhelming genetic and biochemical data demonstrating the role of BER and MMR in CSR, the mechanism by which BER and MMR are subverted (or coopted) to promote recombinogenic repair of S regions rather than canonical repair remains uncharacterized. Hypothetically, AID may generate a high density of dU:dG mismatches within the S regions that cannot be repaired by canonical BER and MMR pathways. To maintain genomic integrity, BER and MMR are shunted toward recombinogenic repair and thus CSR. AID phosphorylation regulates the balance of canonical and recombinogenic repair that is mediated by BER and MMR downstream of AID-dependent deamination of S regions⁶².

Phosphorylation of AID at Ser38 (pS38-AID) is critical for CSR as mice harboring a homozygous S38A knock-in mutation (AID^S38A/S38A) have a significant reduction in CSR^79,80. S38 lies within a consensus cAMP-dependent PKA phosphorylation site^80–82. A hypomorphic PKA-RIα knock-in mutant (RIαB) substantially impairs CSR and blocks phosphorylation of AID at S regions⁸³, indicating that PKA is required for AID phosphorylation at S38. Multiple isoforms of protein kinase C (PKC) can phosphorylate AID at S38 in vitro⁸⁰; however, the regulation of PKC-mediated AID phosphorylation in vivo remains unknown. Although the mutant AID (S38A) protein retains wild-type levels of deaminase activity in vitro and binding to S-region DNA in vivo, AID^S38A/S38A B cells cannot efficiently generate DSBs at recombining S regions⁶². These data in conjunction with biochemical data demonstrating the indirect interaction of pS38-AID with APE1 strongly suggest that pS38-AID is required for DSB formation⁶². Endogenous wild-type AID in UNG^−/−MSH2^−/− B cells or catalytically inactive AID cannot be phosphorylated at S38 and consequently cannot bind to APE1; however, treating these cells with ionizing irradiation to induce DSBs restores both AID phosphorylation and APE1 binding, suggesting that the conversion of AID-dependent S-region DNA deamination into single-strand breaks by BER (APE1) or MMR (PMS2/EXO1) is required for AID phosphorylation. Thus, AID phosphorylation at S38 is required for, and dependent on, DNA breaks⁶². These findings suggest the existence of a positive feedback loop wherein a low density of DNA breaks leads to AID phosphorylation, APE1 binding, and additional DNA breaks, which in turn activate more AID phosphorylation (Figure 3). Consistent with this model, ATM, a serine/threonine protein kinase that activates DNA repair pathways in response to DSBs, is required for wild-type levels of AID phosphorylation and APE1 interaction⁶². This model uncovers a previously undescribed role for ATM as a molecular rheostat that couples targeted DNA double-strand break formation with non-canonical, recombinogenic DNA repair to promote Ig gene diversification (Figure 3).

Figure 3. A hypothetical positive feedback loop generates a high density of DNA double-strand breaks to promote wild-type CSR.

AID-mediated deamination of S regions generates DNA breaks that induce PKA-dependent AID phosphorylation at serine-38 (pS38-AID) and subsequent binding of APE1 and RPA to pS38-AID. Recruitment of APE1 to S regions generates additional DNA breaks, inducing additional AID phosphorylation through an unidentified ATM-dependent mechanism of activating PKA. AID, activation-induced cytidine daminase; APE1, apurinic/apyrimidinic endonuclease 1; ATM, ataxia telangiectasia mutated; CSR, class switch recombination; PKA, protein kinase A; RPA, replication protein A.

ATM-dependent and -independent DNA damage responses during CSR

ATM plays a role not only in the stabilization of S-region DSBs through the proposed synapsis and joining of S regions but also in the generation of DSBs through the phosphorylation of AID and the subsequent interaction of AID with APE1^62,90–92. Activation of ATM kinase activity requires binding of the Mre11/RAD50/Nbs1 (MRN) complex to DSBs, which induces ATM-dependent phosphorylation of proteins mediating cell cycle checkpoints (for example, p53) and DNA repair, such as H2AX, MDC1, Nbs1, and 53BP1^60,93,94. The DNA damage response initiated by ATM promotes the assembly of macromolecular foci flanking DSBs and provides docking sites for DNA repair proteins to bind and stabilize DNA ends to promote recombinogenic repair during CSR. Null mutations in ATM substrates impair CSR and increase chromosomal abnormalities and translocations^90–92. 53BP1 deficiency leads to the most robust defect in CSR^95–97. Mutation or deletion of 53BP1 results in a 90% defect in CSR with a significant proportion of chromosomal aberrations involving the Igh locus^92,98 as well as a high frequency of Sμ internal deletions in cells stimulated for CSR⁹⁹. 53BP1 promotes the synapsis and long-range joining of S regions⁶⁰ and protects DNA ends from end resection to direct DNA repair toward NHEJ^100,101. Consistent with these roles for 53BP1, ATM-mediated phosphorylation of 53BP1 recruits Rap-1 interacting factor (Rif1) to sites of DNA damage to protect DNA ends from resection and to promote DNA repair¹⁰². Accordingly, Rif1-deficient B cells are significantly impaired in CSR¹⁰² (Figure 4). Additionally, 53BP1 can be recruited to S-region DSBs through ATM-independent pathways. 53BP1 interacts with H4K20me2 at sites of DNA damage. Depletion of the histone methyltransferase MMSET in the CH12F3 B cell line decreases both H4K20me2 levels and 53BP1 accumulation at S regions, thereby impairing CSR^103,104. Furthermore, the recruitment of 53BP1 to DSBs has been shown to require the RNF8- and RNF168-dependent histone ubiquitination pathway. RNF8^−/− and RNF168^−/− B cells have decreased 53BP1 at S regions and a concomitant reduction in CSR^105–108. Because CSR is more dramatically reduced in 53BP1-deficient B cells as compared to ATM-, H2AX-, MDC1-, or RNF8-deficient B cells, 53BP1 also has a function during CSR that is independent of the ATM/γH2AX/MDC1/RNF8 DNA damage response^{90,91,96,97,105–110}. Data showing reduced Eα interactions with Eμ and the γ1 promoter in LPS- or LPS+IL4-stimulated 53BP1^−/− B cells demonstrate a role for 53BP1 in S-S synapsis during CSR⁶⁰.

Figure 4. Resolution of DSBs generated by MMR or BER following AID-dependent deamination of S regions is accomplished by multiple pathways.

ATM directly or indirectly phosphorylates proteins (for example, H2AX, MDC1, 53BP1, and AID) and stabilizes protein complexes that aid in the formation and resolution of DSBs during class switch recombination. A-EJ, alternative end-joining; AID, activation-induced cytidine daminase; BER, base excision repair; cNHEJ, classical non-homologous end-joining; DSB, double-strand break; HR, homologous recombination; MMR, mismatch repair.

More recently, BRCT-repeat inhibitor of hTert expression (BRIT1) has been implicated as a novel effector of the DNA repair phase of CSR¹¹¹. BRIT1 is a ubiquitously expressed protein that is rapidly recruited to DSBs after ionizing radiation through its C-terminal BRCT repeat domain, which is necessary for its interaction with phosphorylated H2AX (γH2AX)¹¹². As predicted, successful CSR requires BRIT1 interaction with γH2AX at recombining S regions¹¹¹. In addition, the BRIT1-γH2AX pathway is further modulated by the interaction of γH2AX with MDC1 in CSR. Although BRIT1 or MDC1 deficiency alone leads to a moderate reduction in CSR, loss of both BRIT1 and MDC1 together markedly impairs CSR¹¹¹. Thus, BRIT1 likely serves as a scaffold to recruit factors that resolve DSBs at S regions downstream of ATM (Figure 4).

End-joining

Homologous recombination (HR) and NHEJ are the two major pathways for DSB repair in mammalian cells¹¹³. HR is restricted to the S/G₂ phase of the cell cycle and requires large stretches of homology, whereas NHEJ is active throughout the cell cycle and requires little or no homology. Since CSR-associated DSBs are observed primarily during the G₁ phase of the cell cycle and do not have extended stretches of homology, NHEJ is generally considered the major pathway in the joining of DSBs during CSR^95,113. Consistent with this, mutations in the canonical NHEJ components Ku70/Ku80 heterodimer (Ku), XRCC4, and DNA ligase IV (Lig4) severely compromise CSR, while mutations in non-canonical NHEJ proteins such as DNA-dependent protein kinase catalytic subunit (DNA-PKcs), Artemis, and XRCC4-like factor (XLF or Cernunnos) increase chromosomal translocations even though CSR is not severely impaired⁹⁵.

B cells lacking core NHEJ components are capable of residual CSR, mediated by microhomology-biased A-EJ pathway (Figure 4). Although A-EJ is a poorly defined DNA repair mechanism guided by microhomology between two DSBs, factors from other DNA repair pathways, including XRCC1, Ligase III, Mre11, Parp1, and CtIP, have been shown to be necessary for A-EJ during CSR^114–117. More recently, Rad52, an HR repair factor, was shown to facilitate microhomology-mediated A-EJ that favors intra–S region recombination and competes with Ku to mediate inter–S region DSB recombination¹¹⁸. Whether A-EJ is physiologically necessary in order to complete a productive class switch reaction is extensively discussed in 95.

While C-NHEJ and A-EJ are the primary end-joining pathways ligating DSBs within IgH during CSR, HR has been proposed to repair AID-induced off-target DSBs. Deficiency of the Rad51 paralog XRCC2, a key component of HR-mediated repair, significantly enhances AID-dependent genome-wide DNA damage^119,120. Notably, AID-expressing human chronic lymphocytic leukemia cells are hypersensitive to HR inhibitors and this is possibly due to AID-dependent synthetic cytotoxicity from unrepaired DSBs at non-Ig loci¹²¹. Thus, HR is essential for repairing AID-generated DSBs and dysregulated AID activity may provide a novel therapeutic approach to treat B cell malignancies.