Developing B lymphocytes remodel the variable regions of their immunoglobulin (Ig) loci through V(D)J recombination, generating a vast array of antigenic specificities (1–3). Upon antigen encounter, mature B lymphocytes undergo class switch recombination (CSR) which targets the constant (C_H ) genes of the Ig heavy chain (IgH) locus, ultimately leading to a change of the constant domain of Ig molecules. CSR thus enables activated B cells to switch from the expression of the initial IgM to the expression of downstream isotypes (IgG, IgE or IgA) with novel effector functions (4–7). CSR occurs between highly repetitive, GC-rich, switch (S) sequences, located upstream of the C_H gene exons, except Cδ. CSR to a particular S region is induced by specific external stimuli including antigens, mitogens, cytokines, and inter-cellular interactions, and requires transcription across S regions, directed by the so-called I promoters (4–7).

Switch transcription is regulated by various long-range cis-acting elements, including enhancers and insulators (8, 9). The major control element is a super-enhancer called 3’ Regulatory Region (3’RR), composed of four enhancers that act in synergy to activate upstream I promoters (8, 9). Regulation of I promoters involves dynamic conformational changes that are controlled in a developmental stage-, and stimulus-dependent manner (9). In resting B cells, the 3’RR engages in stable interactions with Eµ enhancer, located upstream of C_H genes (10), forming a CSR centre (CSRC) (11). Upon activation, the primed I promoter is brought to the CSRC where synapsis between Sµ and the partner switch sequence is promoted by Cohesin-mediated loop extrusion that is dynamically impeded by Eµ enhancer and the 3’RR ( (11). Reviewed in Ref (9)).

Switch transcription generates long non-coding RNAs that promote accessibility of S sequences, through secondary structures such as R loops and G quadruplexes (12–15), to the enzyme Activation-Induced cytidine Deaminase (AID), which is absolutely required for CSR (16, 17). AID initiates the process by deaminating cytosines to uracils (5, 18). Processing of uracils by base excision and mismatch repair pathways leads to double-strand break (DSB) intermediates, at the Sµ donor region and one of the acceptor S regions (Sγ, Sε, Sα). The DSBs are monitored by components of the DNA damage response pathway (such as ATM, 53BP1 and H2AX) and repaired by the classical and alternative non-homologous end joining pathways (hereafter C-NHEJ and A-NHEJ, respectively) (19–25), ultimately fusing Sµ and the acceptor S region, with a strong bias towards deletional joining (26). C-NHEJ and A-NHEJ use different components and have different signatures at switch junctions. The C-NHEJ pathway (whose core components include Ku70, Ku80, XRCC4, and ligase 4) favors blunt ends or ends with limited MH (≤3 bp), and is the major repair pathway in CSR (27, 28). The A-NHEJ pathway uses components such as CtIP, MRN, and PARP-1, favors ends with larger MHs (≥4 bp) and involves extensive end resection (27–30).

The S regions differ in size, ranging from _~2 kb (Sε) to _~12 kb (Sγ1), and there is evidence that the number of tandem repeats determines, at least in part, the efficiency of CSR (31, 32). They also differ in the nature of their tandem repeats (4, 33). Sµ, Sε and Sα core repeats are short (5 bp), consisting of units such as GAGCT and GGGG/CT, whereas the Sγ (Sγ3, Sγ1, Sγ2b and Sγ2a in the mouse) core repeats, which also contain GAGCT and GGGG/CT units, are longer (48-49 bp) and more complex (4–6, 34). Sµ has a higher sequence homology with Sα and Sε than with Sγ. Likewise, Sγ3 sequence for instance displays higher sequence homology with the other Sγ than with Sµ, Sε and Sα (4, 28, 34). In this regard, it was suggested that A-NHEJ could play an important role in CSR involving S partners with substantial sequence homology (22, 25, 35). However, the extent to which S core repeats’ peculiarities influence their CSR efficiency and the choice of the NHEJ pathway is still unclear.

Various studies revealed that Sµ region displays specific features that distinguish it from downstream S regions. For instance, Sµ is transcribed along B cell development, whereas other S regions are mainly transcribed in activated mature B cells (9). Additionally, Sµ is the most repetitive and displays the highest density of AID target motifs (4), in particular of the evolutionary conserved AGCT motif (36) (see Supplementary Table 1). Following activation for CSR, internal switch deletions (ISDs) are detected at Sµ region at a higher frequency than at downstream S regions [e.g (35, 37–41)] Moreover, mice deficient for components of the DNA damage response or C-NHEJ feature defects in CSR but not in Sµ ISDs (35, 39–42), suggesting that DNA repair mechanisms involved in ISDs differ, at least in part, from those involved in genuine CSR [discussed in (27)]. Several non-mutually exclusive hypotheses could be put forward to account for Sµ specificity including Eµ enhancer proximity, continuous transcription, chromatin structure, preferential recruitment of AID, and differential recruitment of repair pathways. Thus, the molecular basis of Sµ specificity remains elusive.

We reasoned that by putting a downstream S sequence under the control of the known elements that regulate Sµ, we could investigate if that S region can acquire Sµ properties. To this end, we used a mouse line in which Iγ3 promoter was replaced by a pre-rearranged VDJ-Eµ cassette (43), leaving intact the endogenous Sµ and Sγ3 regions. In this setting, the two S regions have roughly the same size, are almost equally distant from Eµ enhancer, but differ in the nature of their core repeats and the density of AID target motifs. Here, we focused on the recombinational activity of the two S regions. We show that Sγ3 acquired most of Sµ properties but displayed a distinctive propensity for longer MH in both ISDs and CSR.

We have previously shown that replacement of Iγ3 switch promoter by a PV_H-VDJ-Eµ cassette (hereafter A150 mutation or mouse line) (Supplementary Figure 1A) leads to an accumulation of partially rearranged DJ_H alleles and a drastic reduction of V_H-DJ_H recombination (43). Consequently, IgM expression is severely impaired in A150 homozygous mice and B cell development is driven by IgG3 (43) (and Supplementary Figure 1B). In this study, we used this mouse line to investigate if Sγ3 region, in its new setting, has acquired the recombinational properties of Sµ.

By putting Sγ3 region under the control of the known elements that regulate Sµ and by comparing the recombinational activities of Sµ and Sγ3 in the same conditions (same allele, same stimulation conditions), we provided evidence that Sγ3 acquired most of the features of Sµ. In addition to its continuous and constitutive transcription along B cell development (43) (and the present study), Sγ3 efficiently recruited AID, underwent high frequency of ISDs, and acted as a powerful donor site. Remarkably, Sγ3 acted this way despite the fact that it has a lower density of AID target motifs generally, and of the hot AGCT motif specifically (Supplementary Table 1). On the other hand, Sγ3 displayed a distinguishing feature, i.e. an increased usage of MH both in ISDs and in CSR regardless of the switch acceptor region or the stimulation condition.

It should be stressed that the A150 genetic setting has its own limitations. The fact that B cell development in A150 mice is driven by IgG3 (43) instead of IgM has already been noted. Mechanistically, it is presently unclear to what extent insertion of Eµ enhancer upstream of Sγ3 has perturbed various parameters that are important for CSR including the global architecture of the IgH constant region, CSRC interactions (see below), transcription of a subset of S regions and its correlation with CSR efficiency, as well as the (co-)transcriptional events and the epigenetic landscape at Sγ3 itself, which are crucial for AID recruitment and activity (7, 9). These topics clearly require further investigations.

With these caveats in mind, we found that surface expression of IgG1 and IgA was reduced in activated A150 B cells. This cannot be readily explained by defective switch transcription as Sγ1 transcript levels were only moderately reduced, while Sα (and Sε) transcript levels were normal. At the quantitative level, the precise threshold of switch transcript levels required for efficient CSR has not been determined yet. Nonetheless, the high frequency of CSR to Sγ1, as seen at the genomic level with CSR-HTGTS, suggests that the modest decrease of A150 Sγ1 transcript levels is not the critical issue.

A likely explanation stems from the lingering efficiency of Sµ as a donor site and the combinations of alleles with different rearrangement status. Indeed, switched sIg positive A150 B cells (other than IgG3) can originate from CSR events involving either Sµ or Sγ3. However, the majority of A150 alleles are in a DJ_H configuration and only a small fraction undergoes proximal V_H-DJ_H recombination (43), of which 2/3 are in principle out-of-frame. Consequently, despite efficient recombining activity of Sµ at the genomic level, most of its recombination products lead to dead-ends at the Ig level. This is not the case when Sγ3 (located downstream of a pre-rearranged, in-frame VDJ sequence) acts as a donor site. Therefore, most of sIg positive cells likely derive from recombination events involving Sγ3 on alleles that did not undergo Sµ recombination (i.e. that did not delete Sγ3). Thus, the recombining activity of Sµ makes it difficult to establish a strong correlation between sIg expression and CSR events at the genomic level in A150 line. However, this issue was circumvented by using CSR-HTGTS, which provided a powerful tool to track, at the nucleotide resolution level, in allele- and orientation-independent manner, the recombination events involving both Sµ and Sγ3.

Although we cannot formally exclude a potential contribution of the PV_H promoter in A150 setting, acquisition of Sµ properties by Sγ3 is likely due to the proximity of Eµ enhancer. In normal B cells, Sµ is known to undergo CSR on both partially rearranged DJ_H alleles and fully rearranged V_HDJ_H alleles (9). In A150 context, V_H-DJ_H recombination is severely impaired in developing B cells (43), and only a small fraction of activated mature B cells express sIg (data not shown), but this does not prevent Sµ from acting as a powerful switch donor site as clearly shown by CSR-HTGTS. Our data therefore strongly suggest that provision of Eµ enhancer is sufficient to induce a high frequency of ISDs and to confer a robust donor function to A150 Sγ3 despite its different core repeats and the lower density of AID motifs. One possible explanation is that the ectopic Eµ enhancer ensures high levels of Sγ3 transcription, enabling efficient recruitment of AID. However, A150 Sγ3 region recruited AID as efficiently as Sµ despite comparatively lower levels of Sγ3 transcripts. On the other hand, we found comparable levels of Sγ3 transcripts in LPS-activated WT and A150 B cells. Nonetheless, AID was significantly enriched at Eµ-driven A150 Sγ3, but was only at the background level in WT Sγ3 (within the sensitivity limits of our ChIP assay). Taken together, these observations suggest that the apparent preferential targeting of Sµ by AID in activated normal B cells is not the consequence of specific properties of Sµ primary sequence such as repetitiveness or density of AID motifs, or of a higher transcriptional activity, but results, at least in part, from specific properties conferred by Eµ enhancer proximity. In this context, the proximity of Eµ enhancer can explain, at least in part, the relatively high frequency of sequential switching to Sε in normal B cells. Indeed, CSR to IgE is known to occur directly (Sμ/Sε) or sequentially (Sμ/Sγ1/Sε) [e.g.( (49–53)]. The presence of Eµ upstream of the hybrid Sµ/Sγ1 likely promotes the subsequent recombination to Sε. We do not infer from the above discussion that primary sequence peculiarities of S sequences have no importance. They have, in particular with respect to the mechanistic aspects of DSBs repair (see below).

CSR takes place in CSRCs and involves Cohesin-mediated loop extrusion that is impeded by Eµ enhancer and the 3’RR (11) [Reviewed in (9)], as well as the super-anchor located downstream of the IgH locus which focuses loop extrusion on the upstream constant region (54). Our findings could be explained by a model whereby the two Eµ enhancers and the 3’RR co-exist in a « ménage à trois » within the CSRC. Alternatively, there may be competition between the two Eµ elements such that only one lies close to the 3’RR (Supplementary Figure 8). The high frequency of Sµ/Sγ1 and Sγ3/Sγ1 recombination on one hand, and the low frequency of Sµ/Sγ3 recombination on the other hand, favor the view that only one Eµ enhancer lies within the CSRC at a time. Nonetheless, both enhancers may co-exist in a small fraction of the CSRCs allowing the low levels Sµ/Sγ3 synapsis. Further analyses are needed to unravel the dynamics of Sµ and Sγ3 sequences within the CSRC.

In agreement with the notion that C-NHEJ is the major repair pathway during CSR (27, 28), the vast majority of A150 B cells displayed either direct or low MH joins regardless of the switch donor site. The same holds true for ISDs. Nonetheless, there was a relatively higher MH usage by Sγ3 in the context of both ISDs and CSR irrespective of the stimulation condition. Overall, the increase was moderate (_~2-fold) but highly reproducible and statistically significant. Based on the criterion of MH extent, this suggests that Sγ3 tends to favor the recruitment of A-NHEJ regardless of the outcome of the DNA DSBs (i.e. ISDs or CSR) and of the acceptor site. This bias appears therefore to be directed by Sγ3, not by sequence homology with the partner S regions. We propose the following speculative model to account for this finding. Upon B cell activation, AID initially targets the switch donor region ultimately leading to multiple and heterogeneous DNA ends that recruit C-NHEJ and A-NHEJ pathways. When the partner S region is targeted by AID, the DNA ends that did not undergo short-range repair (ISDs) at the donor site will engage in bona fide CSR while tethering the components of the pathway initially recruited.

The increased usage of MH by Sγ3 likely reflects the complexity of its repeats and therefore the complexity of the DNA ends generated, and potentially the kinetics of repair. For instance, by focusing on the most abundant motif, the core Sµ is virtually a multimer of AGCT(G)GGGT motifs whereas the AGCT units are relatively more distant within Sγ3 repeats. It is plausible that if AID-initiates nicks on both strands of the palindromic, overlapping (55) AGCT motifs (be it at Sµ or Sγ3), or on very close AGCT motifs (more frequently at Sµ than Sγ3), the resulting ends would require very limited resection (and/or filling), ultimately leading to C-NHEJ-mediated repair. In contrast, if AID initiates nicks at distant motifs on opposite strands (more frequent at Sγ3 than Sµ), the long overhangs would require more extensive resection (and/or filling), favoring MH search and usage and A-NHEJ-mediated repair. This is in agreement with the notion that the structure of staggered DSBs influences the mode of end processing and recruitment of A-NHEJ during CSR (56–58). Recent work strongly suggests that accumulation of DNA : RNA hybrids at S regions due to deficiency of the RNA exosome catalytic subunit, DIS3, yields longer overhangs and increased MH (59). In this regard, it is possible that A150 Sγ3 DNA : RNA hybrids are relatively more stable and processed by the RNA exosome with a slower kinetics than their Sµ counterparts.

It is arguable if increased usage of MH is promoted by DNA damage response deficiencies (39–41, 60–63) and/or by other factors. The most recent evidence shows that 53BP1- and, to a lesser extent, ATM-, H2AX- and Rif1-deficiencies significantly increase MH-mediated CSR junctions (64, 65). Our data are in line with the notion that A-NHEJ can operate in the presence of intact DNA damage response. This does not exclude the possibility of enhanced MH usage in the context of defective DNA damage response. On the other hand, RAD52 has been shown to play an important role in MH-mediated A-NHEJ during CSR, notably by facilitating a KU-independent DNA DSB repair (66). To what extent the DNA damage response and RAD52 are involved in Sγ3 related A-NHEJ in CSR are questions for future investigations. Finally, the A-NHEJ was initially thought to prevail in CSR upon C-NHEJ deficiency, and whether it is efficient in C-NHEJ-proficient cells was much debated (27, 29, 30). Our findings support the notion that A-NHEJ can operate in C-NHEJ-proficient cells (65, 67) undergoing ISDs and CSR.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/geo/, GSE174296.For FACS data / FlowRepositery ID : FR-FCM-Z3SX Access with the following link : https://flowrepository.org/id/RvFrFuSmwmwDXnmHOsLKZvnlUyMgrKRdBybKuJo4HfMcahREDfK4mNLE3OHJvSYG.

The animal study was reviewed and approved by The Regional Ethical Committee (Accreditation N° E31555005). All experiments on mice have been carried out according to the CNRS ethical guidelines.

CO, XZ, NP, FA, and AK actively participated to the experimental design of the study. CO, XZ, NK, AK, and FA designed CSR-HTGTS and interpretation of the data. CL and F-ZB contributed to experiments. AD handled the mouse lines. All authors participated in the scientific discussion for manuscript writing, read and approved the manuscript. AK designed the project and obtained financial grants and agreement of the relevant ethic committees to perform the study. All authors contributed to the article and approved the submitted version.

This work was supported by the Agence Nationale de la Recherche [ANR-21-CE15-0019], the Institut National du Cancer [INCA_9363, PLBIO15-134], the Fondation ARC pour la Recherche sur le Cancer [PJA 20191209515], the Ligue Contre le Cancer (Ligue Régionale : comités de l’Ex Région Midi-Pyrénées). FA is an investigator of the Howard Hughes Medical Institute. CO was a fellow of the Ministry of Higher Education & Research and recipient of a fellowship from the “Fondation pour la Recherche Médicale”. TRI- IPBS has the financial support of ITMO Cancer Aviesan (National Alliance for Life Science and Health) within the framework of Cancer Plan.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.