RAG endonuclease is a Y-shaped heterotetramer, which contains two units of RAG1 catalytic enzymes and two units of RAG2 regulatory co-factors (Kim et al., 2015; Ru et al., 2015; Kim et al., 2018; Ru et al., 2018). Both RAG1 and RAG2 are required for the physiological V(D)J recombination (Schatz et al., 1989; Oettinger et al., 1990). RAG1 has the DNA binding and cleaving activity to cut the heptamer of RSSs to generate blunt RSS ends and hairpin-associated coding ends (Figures 2B–D) (McBlane et al., 1995; van Gent et al., 1995; Alt et al., 2013). RAG1 interacts with numerous nucleolar proteins to modulate recombination activity in the nucleus (Brecht et al., 2020), and the N-terminal region of RAG1 regulates the efficiency and pathways of synapsis for V(D)J recombination (Beilinson et al., 2021). RAG2 has no DNA cleavage activity, but it is required to enhance RAG1 catalytic activity. RAG2 binds to DNA by recognizing trimethylation of lysine 4 on histone H3 (H3K4me3), which is a histone marker of active chromatin including promoters and enhancers (Matthews et al., 2007; Teng et al., 2015). The abundance of RAG2 protein is cell cycle-dependent which undergoes ubiquitin-dependent degradation when lymphocytes transit from G1 to the S phase (Li et al., 1996; Teng and Schatz, 2015). Also, RAG2 interacts with RAG1 to abolish RAG1 aggregation to initiate V(D)J recombination during the G1 phase (Brecht et al., 2020; Gan et al., 2021). The regulation of RAG2 promotes RAG-mediated V(D)J recombination in B cells during the G1 phase; meanwhile, it suppresses the generation of undesired DSBs and translocations to ensure the genome stability.

RAG not only binds the bona fide RSSs flanked by the V, D, and J gene segments for physiological V(D)J recombination but also can capture and cut cryptic targets besides RSSs at a low frequency (Hu et al., 2015), which might lead to translocations related to B- and T-cell lymphoma (Mahowald et al., 2008). RAG can generate robust recombination between Dβ1 and Jβ1-1 when the Dβ1 and Jβ1-1 segments of T-cell receptor β (TCRβ) are inserted into the c-Myc locus (c-Myc-DJβ cassette). Meanwhile, the c-Myc-DJβ cassette insertion activates RAG activity to capture and cut the cryptic targets (convergent-orientated “CAC” motifs) linearly. Interestingly, RAG cryptic targets are restricted to the 1.8 Mb c-Myc domain anchored by CTCF binding elements (CBEs). Also, RAG cryptic activity within a domain also applies to other domains across the genome (Hu et al., 2015). Moreover, RAG extends its activity to the cryptic targets outside of a domain by deleting the CBE-mediated boundaries (Hu et al., 2015; Zhang Y. et al., 2019). This evidence suggests that RAG scans linearly to capture and cut the convergent-orientated CAC motifs within a domain.

The RAG scanning process can also explain the physiological D to J_H recombination and V_H to DJ_H recombination. The plasmid-based studies indicate that the RSS sequence, not RAG scanning, determines the utilization of D-RSSs (Gauss and Lieber, 1992), while the high-throughput HTGTS-V(D)J-seq analysis of large amounts of D-RSS-inverted v-Abl progenitor (pro)-B-cell lines supports that RSS orientation, not the RSS sequence, plays a key role in deletional D to J_H recombination, indicating that RAG scanning promotes the utilization of the downstream D-RSSs during physiological D to J_H recombination (Figure 2C) (Zhang Y. et al., 2019). J_H-RSS-bound RAG initiates scanning from RC to the upstream D segments until aligning and cutting one downstream D-RSS with J_H-RSS, leading to the generation of DJ_H recombination products (Figures 2C–H) (Zhang Y. et al., 2019). After DJ_H recombination, DJ_H-RSS-bound RAG initiates scanning to the upstream V_H segments and cuts a convergent-orientated V_H-RSS with DJ_H-RSS to complete the V_H to DJ_H recombination, which is supported by the V_H inversion experiments in mice (Hill et al., 2020; Dai et al., 2021). The V_H region inversion eliminates V_H utilization and increases the utilization of newly formed CAC motifs within the inverted region, which strongly supports that RAG scanning promotes the capture of convergent-orientated V_H-RSS in the physiological V_H to DJ_H recombination (Dai et al., 2021).

D to J_H joining occurs within the loop domain anchored upstream by the two divergent CBEs-formed IGCR1 between D and V_H and downstream by the ten tandem CBE-formed super anchor (3′CBEs) (Guo et al., 2011; Alt et al., 2013). V_H to DJ_H recombination needs the neutralization of IGCR1 anchor and V_H-associated CBEs, which allows RAG scanning to the upstream V_Hs (Guo et al., 2011; Alt et al., 2013; Jain et al., 2018). The depletion of CTCF in v-Abl pro-B cells increases the utilization of distal V_Hs, indicating that RAG scans through the CBEs after removing CTCF-mediated anchors in v-Abl pro-B cells (Ba et al., 2020). Moreover, the depletion of Wapl, a cohesin unloader, in v-Abl pro-B cells also increases the utilization of distal V_Hs (Dai et al., 2021), which is consistent with the downregulation of Wapl in normal pro-B cells. It is likely that downregulated Wapl might neutralize CBE-mediated blocks to enhance RAG scanning to the upstream V_Hs, leading to the generation of more diverse antibody repertoires during physiological V(D)J recombination. RAG activity mainly focuses on the targets within the dynamic chromatin impediments including the CTCF-bound chromatin, highly transcribed chromatin, RAG-bound chromatin, and even catalytic-dead Cas9-bound chromatin (Zhang Y. et al., 2019; Zhang et al., 2022). The aforementioned evidence strongly supports that cohesin-mediated loop extrusion is the underlying mechanism of RAG scanning-mediated V(D)J recombination.

Intrinsic and extrinsic stress-induced DSBs are the most harmful DNA lesions to genome integrity, which trigger DNA damage response (DDR) by recruiting DDR factors to the DSBs for repairing. ATM and its downstream phosphorylated targets (H2AX, 53BP1, and MDC1) are the key DDR factors, which play crucial roles in repairing general DSBs and maintaining genome stability (Weitering et al., 2021).

RAG-initiated DSBs also recruit DDR factors during V(D)J recombination. ATM and ATM-phosphorylated p53 are recruited to the RAG-initiated DSBs to surveil the intermediates in V(D)J recombination, protecting against the potentially aberrant oncogenic translocations (Perkins et al., 2002). Also, coding joining is decreased with more un-joined coding ends in ATM-deficient pre-B cells, indicating that ATM stabilizes RAG-initiated DSBs during V(D)J recombination (Bredemeyer et al., 2006). 53BP1-deficient mice have relatively normal B-cell compartments and no substantial block in V(D)J recombination (Manis et al., 2004), while 53BP1-deficiency is also found to impair the distal V to DJ joining at the TCRα locus, suggesting a specific role of 53BP1 in maintaining genomic stability during long-range joining of DSBs (Difilippantonio et al., 2008). H2AX is recruited to the RAG-initiated DSBs at the TCRα locus (Chen et al., 2000), while it is not required for coding join formation or lymphocyte development (Bassing et al., 2002), suggesting that it only functions as a general surveillance machinery to prevent translocations during V(D)J recombination (Yin et al., 2009). MDC1-deficiency has no major block for V(D)J recombination or lymphocyte development (Lou et al., 2006). The recently identified shieldin complex, composed of MAD2L2/REV7, SHLD1, SHLD2, and SHLD3, is also dispensable for V(D)J recombination and lymphocyte development (Ghezraoui et al., 2018; Ling et al., 2020). Altogether, DDR factors have relatively modest or no effect on V(D)J recombination, suggesting the redundant roles of these DDR factors with others during V(D)J recombination (more discussion in the next section).

Intrinsic and extrinsic stress-induced DSBs are mainly repaired by homologous recombination (HR) and C-NHEJ. HR mainly functions in the late S and G2 phases, which uses sister chromatids as templates for error-free DNA repair. C-NHEJ repairs almost all DSBs outside of S and G2 phases and is the major DSB repair pathway in both dividing and non-dividing cells (Zhao et al., 2020).

RAG-initiated DSBs are exclusively repaired by C-NHEJ, resulting from the synapsis of breaks held by the RAG post-cleavage complex (PCC) (Figure 2D) (Teng and Schatz, 2015; Libri et al., 2021). RAG2 truncations or charge-neutralizing mutations switch the DSB repair pathway from C-NHEJ to alternative end-joining (A-EJ) and HR (Corneo et al., 2007; Coussens et al., 2013; Gigi et al., 2014). RAG interacts with the core NHEJ factors Ku70/Ku80 (Figure 2E) (Raval et al., 2008), and Ku70 suppresses A-EJ in G1-arrested pro-B cells (Liang et al., 2021). The deficiency of Ku70 has a severe combined immunodeficiency (SCID) phenotype and severely impairs the formation of coding joins and RSS joins (Gu et al., 1997; Ouyang et al., 1997). The deficiency of Ku80 arrests lymphocyte development at early progenitor stages and induces a profound impairment in V(D)J recombination (Nussenzweig et al., 1996; Zhu et al., 1996). The Ku70/80 complex recruits another two core C-NHEJ factors, namely, XRCC4 and ligase 4, to the DSBs for end joining (Figure 2G). XRCC4 is a scaffolding protein to stabilize ligase 4 to form the ligation complex for ligating the DSB ends. XRCC4- or ligase 4-deficient mice die during the late embryonic development, resulting from the p53-dependent apoptosis (Barnes et al., 1998; Frank et al., 1998; Gao et al., 1998). Deleting p53 in XRCC4-deficient or ligase 4-deficient mice rescues the lethality, while has no rescues for the impaired V(D)J recombination and lymphocyte development (Frank et al., 1998; Gao et al., 2000). So the four core C-NHEJ factors are absolutely required for V(D)J recombination.

In addition to the conserved core C-NHEJ factors, there are several other C-NHEJ factors including DNA-PKcs, Artemis, XLF, and PAXX. DNA-PKcs is recruited to the RAG-initiated coding ends (Lieber, 2010) and phosphorylates Artemis to activate its endonuclease activity, leading to the removal of the coding end-associated hairpins (Figure 2E) (Ma et al., 2002). Before the DNA-PKcs-Artemis-processed coding ends get joined, DNA polymerases (Polμ, Polλ) and terminal deoxynucleotidyl transferase (TdT)-mediated nucleotide additions can further increase the junction diversity (Figure 2F) (Zhao et al., 2020). DNA-PKcs not only play roles in processing coding ends for coding joins, but also functions in RSS joins. The deficiency of DNA-PKcs and DDR factors severely impairs RSS joins, suggesting DNA-PKcs has redundant roles with DDR factors in RSS joins (Gapud et al., 2011; Zha et al., 2011b). In contrast to other C-NHEJ factors, XLF seems to be dispensable for V(D)J recombination as the deficiency of XLF has no measurable impact on V(D)J recombination (Li et al., 2008), while V(D)J recombination is almost abrogated by the deficiency of both XLF and ATM or one of its downstream DDR factors, suggesting functional redundancy of XLF with DDR factors during V(D)J recombination (Zha et al., 2011a; Liu et al., 2012; Oksenych et al., 2012). PAXX, a paralog of XLF, is also dispensable for V(D)J recombination, but the deficiency of both PAXX and XLF almost abrogates V(D)J recombination (Kumar et al., 2016). The new identified ERCC6L2 interacts with other C-NHEJ factors and plays functionally redundant roles with XLF during V(D)J recombination (Figure 2G) (Liu et al., 2020). These aforementioned C-NHEJ factors have relatively less influence on V(D)J recombination than the core C-NHEJ factor, resulting from the functional redundancy with DDR factors or other unknown factors.

AID is essential for both CSR and SHM (Muramatsu et al., 2000). As a paralog of the RNA-cytosine deaminase APOBEC family, AID is originally proposed to be an RNA editing enzyme (Muramatsu et al., 1999; Muramatsu et al., 2000), while large amount of evidence supports that AID functions as a DNA deaminase to deaminate deoxycytidine (dC) to deoxyuridine (dU) (Feng et al., 2020). AID preferentially targets the dC in short DGYW (D = A/G/T, Y=C/T, W = A/T) motifs within the V(D)J exons (Figure 3A) and S regions (Figure 4A) for SHM and CSR, respectively (Rogozin and Diaz, 2004). AID-initiated dU causes the mismatch with deoxyguanine (dG), which can be converted into the point mutation or DSB by base excision repair (BER) and mismatch repair (MMR) during SHM and CSR (Figures 3C, 4A) (Hwang et al., 2015; Methot and Di Noia, 2017).

BER and MMR are two complex DNA repair processes which can function as error-free repair and error-prone repair (Figures 3B,C) (Hwang et al., 2015; Methot and Di Noia, 2017). BER repairs the AID-initiated dU from the recognition and excision of dU by UNG. APE cleaves the DNA to generate a nick at the UNG-initiated abasic site. The nick is further processed to generate a gap, which is filled by DNA polymerase β and sealed by DNA ligase 1/3. MMR repairs the AID-initiated dU from the recognition of the mismatch by MSH2/6, which further recruits MLH1 and PMS2. Exo1 excises the DNA sequences adjacent to the mismatch to generate a gap, which is filled by DNA polymerase δ and sealed by DNA ligase 1. Instead of accurate repair by BER and MMR (Figure 3B), mutagenic repair frequently occurs after AID-initiated dU during CSR and SHM (Figures 3C, 4A). Recent studies indicate that FAM72a influences the usage of error-prone vs. error-free DNA repair by regulating UNG2 abundance during CSR and SHM (Figures 3B,C, 4A) (Feng et al., 2021; Rogier et al., 2021).

DDR factors can also be recruited to the AID-initiated DNA lesions, and these DDR factors are required for CSR as the deficiency of the individual ATM, H2AX, or 53BP1 decreases the CSR frequency (Reina-San-Martin et al., 2003; Lumsden et al., 2004; Manis et al., 2004; Reina-San-Martin et al., 2004; Franco et al., 2006; Reina-San-Martin et al., 2007; Bothmer et al., 2010). 53BP1 and H2AX are the downstream targets of ATM, but 53BP1-deficiency has a much more dramatic effect than that of ATM- or H2AX-deficiency (Dong et al., 2015; Panchakshari et al., 2018). RIF1 is a downstream factor of 53BP1 to inhibit DSB end resection and RIF1-deficiency significantly decreases CSR (Chapman et al., 2013; Di Virgilio et al., 2013). The shieldin complex functions downstream of 53BP1-RIF1 pathway and the deficiency of shieldin components have similar phenotype as that of 53BP1-deficiency (Xu et al., 2015; Dev et al., 2018; Ghezraoui et al., 2018; Gupta et al., 2018; Noordermeer et al., 2018). This 53BP1 pathway can compete with MRN/CtIP activity to protect DNA ends during CSR (Mirman and de Lange, 2020). The deficiency of these DDR factors variably increases the resection of AID-initiated DSBs and increases the utilization of longer microhomology for end joining (Dong et al., 2015; Panchakshari et al., 2018). This evidence indicates that DDR factors inhibit resection to maintain the integrity of AID-initiated DSBs for the efficient C-NHEJ pathway (Figure 4B), while the deficiency of DDR factors switches the end joining from C-NHEJ to the less efficient A-EJ which is prone to use longer microhomology (Figure 4C).

The core C-NHEJ factor deficiency completely abolishes V(D)J recombination and blocks lymphocyte development, while core C-NHEJ factor-deficiency only decreases but not abrogates CSR, suggesting other less efficient end-joining pathways can join the AID-initiated breaks when C-NHEJ is absent during CSR (Yan et al., 2007; Boboila et al., 2012). This less efficient end-joining pathway is identified as A-EJ (Boboila et al., 2012) (Figure 4C).

C-NHEJ is the major DSB end-joining pathway during CSR (Figure 4B). The deficiency of the individual core C-NHEJ factor, Ku70, Ku80, XRCC4, or ligase 4, impairs CSR (Casellas et al., 1998; Manis et al., 1998; Pan-Hammarstrom et al., 2005; Yan et al., 2007; Han and Yu, 2008; Panchakshari et al., 2018). In addition to the core C-NHEJ factors, DNA-PKcs and Artemis are also necessary for joining AID-initiated DSBs during CSR (Franco et al., 2008). The deficiency of XLF impairs CSR (Zha et al., 2011a), while deficiency of PAXX, a paralog of XLF, has no influence on CSR (Kumar et al., 2016). ERCC6L2 is identified as a new NHEJ factor and ERCC6L2-deficiency impairs CSR. Surprisingly, ERCC6L2 deficiency does not increase the resection of AID-initiated break ends, but it significantly increases the inversional end joining during CSR (Liu et al., 2020). ERCC6L2 regulates the orientation-biased end joining without affecting the DSB end resection via an unprecedented mechanism during CSR.

A-EJ is activated when C-NHEJ or DDR factors are absent during CSR (Figure 4C). The deficiency of ligase 4 shares some similar features as that of DDR factor deficiency, including the increase of DSB resection, utilization of longer microhomology, and decrease of CSR frequency (Panchakshari et al., 2018). A-EJ is relatively less intelligible than C-NHEJ. PARP1 can respond to DNA damage and bind to the break sites during A-EJ (Wei and Yu, 2016). Then ligase 1 and ligase 3, the key joining factors of the A-EJ pathway, play redundant roles in joining AID-initiated DSBs during CSR (Lu et al., 2016; Masani et al., 2016). Several exonucleases and endonucleases can also enhance DSB resection and promote A-EJ during CSR (Bai et al., 2021; Sun et al., 2021). Further studies are required to figure out the whole picture of the A-EJ pathway in CSR and other physio-pathological processes.

AID-initiated CSR occurs within the ∼200 kb constant region of the IgH locus in mature B cells. Chromatin loop extrusion is proposed to be the underlying mechanism of CSR, which promotes the formation of the CSR center, transcriptional activation of acceptor S regions, synapsis of donor Sμ and an activated acceptor S region, and deletional joining of AID-initiated DSBs during CSR (Figure 4D) (Zhang X. et al., 2019; Zhang et al., 2021; Zhang et al., 2022). In addition to the physiological CSR process, the loop extrusion-mediated CSR model can also explain some abnormal switching events within the CSR center, including the IgH locus suicide recombination between Sμ and 3′RR (Peron et al., 2012), the ectopic S region switching after CBE insertion in the IgH constant region (Zhang X. et al., 2019) or 3′CBEs deletion (Zhang et al., 2021) and the Sμ-Sγ3 switching after inserting Sμ, Sγ3, and core 3′RR in the Igκ locus (Le Noir et al., 2021).

In resting B cells, cohesin is loaded onto either the active iEμ-Sμ region or the downstream 3′RR enhancer region to initiate loop extrusion. Cohesin-mediated loop extrusion brings these two active regions, namely, iEμ-Sμ and 3′RR into proximity to form a basal loop, in which the iEμ-Sμ and 3′RR serve as dynamic loop anchors. This basal loop is termed as a dynamic CSR center (Zhang X. et al., 2019). When B cells get activated, loop extrusion brings the primed acceptor S region into the CSR center, where it gets transcriptionally activated by 3′RR. Then, the activated acceptor S region loads cohesin to initiate loop extrusion to bring the donor Sμ into close proximity with the activated acceptor S region, leading to the synapsis of two S regions in the CSR center (Zhang X. et al., 2019).

AID can target different locations of the synapsed donor Sμ and acceptor S region at different times within the CSR center. Once AID initiates a DSB within an S region, the DSB ends will be pulled toward the opposite direction by loop extrusion and stalled by the associated cohesin rings. The two pairs of ends held by cohesin rings will be joined deletionally to generate the productive CSR products (Zhang X. et al., 2019). The disruption of the synapsis structure by inserting CBEs that have a convergent orientation to 3′CBEs between donor Sμ and acceptor Sα significantly increases the inversional joining without influencing DSB end resection, which means that the loop extrusion-mediated perfect synapsis of the donor Sμ and acceptor S region is required for the deletional end-joining during CSR (Zhang X. et al., 2019). Loop extrusion-mediated deletional end-joining is consistent with the cohesin accumulation at DSBs (Kim et al., 2002; Strom et al., 2004). Loop extrusion is also proposed to be the underlying mechanism of DNA damage repair. Loop extrusion-mediated ATM scanning along the chromatin adjacent to the DSB site phosphorylates H2AX until reaching the loop anchor to form DNA damage repair foci (Arnould et al., 2021), which shares some similar features to the loop extrusion-mediated deletional end-joining during CSR (Zhang X. et al., 2019). Loop extrusion might have more general roles in DNA damage repair, DSB end joining, and recombination processes.

Upon activation by antigens, mature B cells can undergo CSR and SHM. CSR occurs prior to the mature B cells entering GC, where the V(D)J exons get mutated (Roco et al., 2019). Unlike the critical roles of DDR factors in CSR, ATM, 53BP1, and H2AX are dispensable for the V(D)J exon SHM. The deficiency of the individual ATM, 53BP1, or H2AX has no effect on the SHM frequency of the V(D)J exon (Reina-San-Martin et al., 2003; Manis et al., 2004; Reina-San-Martin et al., 2004). On the other hand, the checkpoint signaling via the ATR/Chk1 axis is downregulated by the transcription factor Bcl-6 in GC B cells, suggesting that negative regulation of the ATR/Chk1 axis is required for efficient SHM in vivo (Ranuncolo et al., 2007; Polo et al., 2008; Frankenberger et al., 2014; Bello and Jungnickel, 2021). However, Chk2 has opposite effects to Chk1 in the regulation of SHM. The deficiency of Chk2 decreases the SHM frequency, resulting from the defects of C-NHEJ and increase of the Chk1 activity (Davari et al., 2014). So, the ATR/Chk1/Chk2-mediated checkpoint signaling of the DNA damage response is crucial for the physiological SHM.