All of CH12F3 cell lines in this study were cultured with in RPMI 1640 (10-040-CV, Corning) supplemented with 15% FBS (FSP500, ExCell Bio), 100 mΜ β-mercaptoethanol (0482-250 ml, Amresco), 20 μM HEPS (25-060-CI, Corning), 2 mM L-Glutamine (25-005-CI, Corning), 1× MEM non-essential amino acid (25-025-CI, Corning), 1 mM sodium pyruvate (25-000-CI, Corning), 1× penicillin streptomycin (30-002-CI, Corning). 293T and Phoenix Ampho were maintained in DMEM (10-013-CV, Corning) supplemented with 10% FBS (ExCell Bio) and 1× penicillin streptomycin.

PSpCas9(BB)-2A-GFP (pX458) plasmid was obtained from Addgene (#48138). All the gRNA oligonucleotides were cloned into pX458. All the oligonucleotides sequences were listed in Appendix information, Supplementary Table S1. pMSCV-IRES-GFP II (pMIG II) plasmid was obtained from Addgene (#52107). pLKO.1 puro plasmid was obtained from Addgene (#8453). pMD2.G and psPAX2 plasmids were kindly gifted by the F.W.A. laboratory.

The gene deletion strategies were performed according to the essential domain of genes reported (Taccioli et al., 1998; Babbe et al., 2009; Schaetzlein et al., 2013; Broderick et al., 2016; Panchakshari et al., 2018; van Wietmarschen et al., 2018). WT CH12F3 cell line or its mutants were nucleofected with a pair of pX458 vector with two gRNAs flanking one or two exons using the 4D Nucleofector Kit (solution SF, protocol CA-137; Lonza). At 24–48 h post-nucleofection, the GFP positive cells were sorted with Beckman Coulter MoFlo Astrios EQs and plated into 96-well plates. Single cell clones were marked and screened by PCR. Positive clones were further confirmed by western blot analysis or T-A cloning and sequencing.

The primary antibodies used in this study were as follows: anti-ATM Rabbit antibody (D2E2, #2873, Cell Signaling Technology), anti-γ-Tubulin antibody (#5886, Cell Signaling Technology), anti-Mre11 Antibody (#4895, Cell Signaling Technology), anti-DNA-PKcs (G-12, SC-390849, Santa Cruz), anti-CtIP (D-4, SC-271339, Santa Cruz), anti-EXD2 antibody (20138-1-AP, Proteintech), anti-phospho KAP1 (S824) antibody (A304-146A-M, Bethyl Laboratories), anti-β-Actin antibody (66009-1-Ig, Proteintech), anti-AID monoclonal antibody (mAID-2, 14-5959-82, eBioscience), and anti-Flag M2 antibody (F1804-50UG, Sigma-Aldrich). The antibodies for flow cytometry analysis were anti-Mouse lgM-APC (17-5790-82, eBioscience), anti-Mouse lgA-PE (12-4204-83, eBioscience), and anti-Mouse lgG1-PE (406608, Biolegend).

Mirin (M9948-5 MG, Sigma-Aldrich), PFM01 (SML1735-5mg, Sigma-Aldrich), Ku55933 (SML 1109-5 mg, Sigma-Aldrich), and AZD1390(S8680-5 mg, Selleck) were dissolved in DMSO and stored at –20°C. Cells were exposed to X-rays generated by a Rad Source RS2000 Irradiator (160 kv, 25 mA) to induce DNA damage.

ShRNAs specific to Mre11, CTIP, and DNA2 were cloned into pLKO.1 puro vector. All the shRNA sequences were listed in Appendix information, Supplementary Table S1. The plko.1 vector cloned with specific shRNA sequence and the packaging plasmids pMD2. G and psPAX2 were co-transfected into HEK293T cell to produce lentiviruses with polyethylenimine (PEI) transfection reagent. Cell supernatants were collected after 48 h post-transfection and filtered with a sterile 0.45-µm syringe filter to remove cell debris. The lentiviruses were concentrated by virus precipitation Solution (ExCell Bio) and resuspended in complete medium. CH12F3 or its mutants were infected with lentivirus by centrifuging at 32°C 1,000×g 60 min. After transduction for 48 h, the cells were selected with 0.5 μg/ml puromycin for 5–7 days.

The EXO1 cDNA sequence was obtained using reverse transcription from total RNA extracted from WT CH12F3 cell line. EXO1 mutants EXO1^EK was obtained with site-directed mutagenesis. The C terminus of cDNA was added with 3× flag tag by two sequential PCR rounds. The EXO1/EXO1^EK-3× flag were cloned into pMIG II. Retrovirus vector pMIG II-EXO1^WT/EXO1^EK-3× flag were transfected into Phoenix Ampho cell to produce retrovirus with PEI. Retrovirus was concentrated as lentivirus did. EXO1-deficient CH12F3 were transduced with retrovirus by centrifuging at 32°C 1,000×g for 60 min. After transduction for 3–4 days, the GFP positive cells were sorted with Beckman Coulter MoFlo Astrios EQs. To confirm the expression of EXO1^WT/EXO1^EK in EXO1-deficient CH12F3 was rescued, the infected cells were lysed for western blot analysis with anti-flag primary antibody.

WT CH12F3 cell line or its mutants at a density of 5 × 10⁴ cells/mL or 1 × 10⁵ cells/mL were stimulated with 1 μg/ml anti-CD40 (16-0401-86, eBioscience), 20 ng/ml IL4 (214-14, PeproTech), and 1 ng/ml TGF-β (96-100-21-10, PeproTech) for 72 h. Cells were collected and analyzed by flow cytometry. Data were presented as mean ± SD from independent experiments (Student’s t-test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, n. s indicates non-significant differences).

For CRISPR/Cas9-initiated CSR (Cas-CSR) in CH12F3 cells, sgRNAs targeting up- and down-stream S regions (Sμ and Sγ1) were transfected into CH12F3 cells via electroporation. SgRNAs were cloned into px458 plasmids. After transfection, CSR to IgG will increase gradually. CSR level to other Ig in KO cells was normalized to the GFP + ratio of 24 h after transfection.

Cells were collected and lysed in RIPA buffer with fresh proteinase inhibitors. The cell lysate was centrifuged and quantified by the BCA assay (23225, Thermo). The collected cell lysate was denatured by boiling in loading buffer at 100°C for 10 min, loaded into the wells of SDS/PAGE to separate, and transferred to PVDF membranes (IPVH00010, Merck). The membranes were blocked by 5% skim milk in PBST for 1 h at room temperature, probed with indicated primary antibodies overnight at 4°C, washed 3 times with PBST, incubated with recommended HRP-conjugated second antibody (7074s, Cell Signaling Technology) for 1 h at room temperature, washed 3 times with PBST, and visualized with HRP substrate peroxide solution.

Total RNA was extracted using TRIzol reagent (15596026, Invitrogen). RNA was reverse transcribed into cDNA by the reverse transcription system (RR037A, Takara). SYBR Premix Ex Taq kit (RR820A, Takara) was used to perform qRT-PCR on LightCycler480 Real-Time PCR System (Roche). Relative gene expression levels were obtained based on the 2^−∆∆Ct method with Hprt as internal reference control. Primers for qRT-PCR are listed in Supplementary Table S1.

HTGTS libraries were constructed as described (Dong et al., 2015). Briefly, genomic DNA of CH12F3 or its mutants were extracted after stimulation for 3 days. The genomic DNA was sonicated and amplified by LAM-PCR with 5′ Sμ biotin primer (5′-CAG​ACC​TGG​GAA​TGT​ATG​GT-3′). The Biotinylated products of PCR were captured by Dynabeads MyOne streptavidin C1 beads (Invitrogen), ligated with bridge adapters on-bead. The ligated products were amplified by second-PCR to add adaptor. Then, the products of PCR were blocking with endonuclease Afill to remove germline genomic DNA fragment. The third round PCR was performed to add Illumina Miseq-compatible adapters to conduct MiSeq sequencing. The HTGTS data were analyzed as described (Dong et al., 2015; Panchakshari et al., 2018). Data were presented as mean ± SEM (Student’s t-test, *p < 0.05, **p < 0.01, ***p < 0.001).

Statistical analysis was performed in GraphPad Prism 7.01. Data was reported as mean and SD except that the HTGTS analysis was reported as mean and SEM. Unpaired two-tailed Student t test or two-way ANOVA was used to examine the significant difference between samples. The asterisks stand for significant differences (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, n. s indicates non-significant differences).

HTGTS sequencing data have been deposited at the Sequence Read Archive (SRA) with a project #PRJNA728565, with an access URL:https://dataview.ncbi.nlm.nih.gov/object/PRJNA728565.

Previous reports indicated that germline deletion of Mre11 or CtIP confers early embryonic lethality in mice, and mutant MEF cells showed altered proliferation and genome instability (Buis et al., 2008; Chen et al., 2005). To first examine the role of Mre11 in class switching, we utilized two different short hairpin RNAs (shRNA) expressed from lentiviral vectors to silence its expression in mouse mature B cell lymphoma cell line CH12F3 that can be stimulated to specifically undergo isotype switching to IgA (Figure 1A). ShRNA-mediated knock down of Mre11 expression in CH12F3 appears not affecting the overall proliferation of cells (Supplementary Figure S1A). When stimulated by the combination of αCD40/IL-4/TGF-β, shMre11 cells showed similar level of mature Iμ and Iα germline transcription, and the protein level of AID was not perturbed by Mre11 silencing (Supplementary Figures S1B,S1C). IgA expression in shMre11 cells showed a mild defect by surface staining (Figures 1B,C, Supplementary Figure S2A). To distinguish whether the endonuclease or exonuclease of Mre11 is involved in CSR by c-NHEJ, we treated CH12F3 cells with small chemical inhibitor Mirin or PFM01 that specifically inhibit Mre11’s exo- or endonuclease activity, respectively (Shibata et al., 2014), and discovered that only Mirin, but not PFM01 treatment conferred a mild but significant defect in IgA levels (Figure 1D, Supplementary Figure S2B). Next, we used three different shRNA to silence expression of CtIP in CH12F3 cell (Figure 1E). While CtIP knockdown did not affect the Iμ and Iα germline transcription, AID protein level and cell proliferation rate did exhibit small decline by shCtIP-3# (Supplementary Figures S1D-F). However, all three shCtIP-infected cells showed similar IgA levels at around 70–80% of values of WT, implying that CtIP contributes to class switching in WT cells largely independent of AID protein regulation. A recent study reported EXD2 as an exonuclease that functions with Mre11 for DSB resection and HR (Broderick et al., 2016). We generated EXD2 knockout CH12F3 cells by CRISPR/Cas9 (Supplementary Figures S3A–S3C), and surface staining indicated that EXD2 was not required for IgA switching by c-NHEJ (Supplementary Figure S3D).

To further investigate whether these resection initiation proteins are required for CSR by A-EJ, we knocked-down Mre11 in Lig4 ^−/− CH12F3 cells (Figure 1A). Again, Mre11 silencing in Lig4 ^−/− cells did not change Iμ/Iα germline transcription and AID protein level, or overall proliferation (Supplementary Figures S1A–S1C). As previously reported, Lig4 ^−/− cells switched to IgA at an efficiency about ∼30% of that of WT CH12F3 cells. Mre11 silencing by two different hairpin RNAs significantly further reduced IgA levels by about one third to half (Figures 1B,C). Treating Lig4 ^−/− cells with either Mirin or PFM01 also reduced IgA switching levels to close to 50% of DMSO-treated control cells (Figure 1D, Supplementary Figure S2B). Similarly, shRNA-mediated knockdown of CtIP in Lig4 ^−/− cells further impairs IgA switching by more than 50% (Figures 1F,G, Supplementary Figure S2C). A rather mild effect on A-EJ was observed by Exd2 deletion in Lig4 ^−/− cells (Supplementary Figures S3C,S3D), suggesting that the stimulation of Mre11’s exonuclease activity by Exd2 is negligible during B cell class switching. Taken together, these data suggest that while efficient CSR in wild type cells requires Mre11/CtIP to various extent, Mre11 and CtIP play more important roles in A-EJ-mediated CSR.

To further explore the molecular signature of end joining in cells deficient for Mre11 or CtIP, we utilized High Throughput Genome-wide Translocation Sequencing (HTGTS) to characterize Sμ-Sα junctions and MH usage pattern in stimulated CH12F3 wild type and mutant cells (Figure 2A). HTGTS with a 5’ Sμ anchor primer fine-maps joining from AID-initiated DSBs occurring in upstream Sμ to those in Sα region and genome wide. Consistent with no or moderate defect in c-NHEJ CSR by FACS in WT CH12F3 cells, Mre11 or CtIP knockdown cells showed nearly no difference in the percentage of Sμ-Sα joining compared with scramble controls (Supplementary Figure S4). When examining junctions mapped to the Sα region for details, however, we indeed observed a small but significant decrease in the ratio of direct versus inversional Sμ-Sα joining in shMre11 cells (Figures 2B,C). The percentages of junctions falling into Cα represents joining of Sμ to Sα DSBs resected into distal region (Figure 2A). In shMre11 cells, we observed a small but significant increase in the Sα DSBs long resection (Figure 2D), and MH usage of Sμ-Sα joining showed a slight decrease in “blunt” (MH = 0) and increase in MH = 1 joins (Figure 2E), consistent with a role for Mre11 in activating DDR kinase ATM that is critical for suppressing resection and MH usage. In contrast, CtIP-silenced CH12F3 cells showed identical Sα DSB resection and MH pattern in Sμ-Sα junctions compared with control cells (Figures 2D,E), indicating that CtIP does not play a critical role in the joining step of c-NHEJ-mediated CSR.

We then analyzed S-S joining pattern of Mre11 or CtIP-silenced Lig4 ^−/− cells with HTGTS. Scramble control virus transduced Lig4 ^−/− cells had significantly decreased in direct Sα joining and concomitant decrease in the ratio of direct versus inversional Sα junctions (Figures 2B–D). Mre11 or CtIP knockdown further decreased direct Sα joining percentage, consistent with IgA surface staining data (Figure 2B). Interestingly, we found no significant difference in Cα distal junctions and MH usage between Lig4 ^−/− cells infected with scramble or shMre11/shCtIP virus (Figures 2D,E), indicating that while Mre11/CtIP is partly required for A-EJ events in Lig4-deficient cells, they are not required for the long resection activity into Cα region in these cells. In addition, we indeed discovered an obviously decreased ratio of direct versus inversional Sα junctions in shCtIP infected Lig4 ^−/− cells (Figures 2B,C), implicating a unique role for CtIP in A-EJ in this context.

To study the role of long range DSB resection factors in CSR, we first deleted Exo1 by CRISPR/Cas9 in CH12F3 cells (Supplementary Figures S5A,B). Consistent with previous reports, IgA switching in Exo1 knockout cells in both WT and Lig4 ^−/− backgrounds was extremely low (Supplementary Figure S5C) due to severe defect in mismatch repair that is critical to convert AID-initiated lesions into DSBs (Bardwell et al., 2004). Except for its exonuclease activity, Exo1 also has a structural function to facilitate the assembly of high-order protein complex. The Exo1^E109K mutation that is exonuclease-dead has been shown to retain mismatch repair activity but is defective in DSB resection and HR-related functions (Schaetzlein et al., 2013). We thus reintroduced the Exo1^E109K (referred to as Exo1^EK hereafter) mutation by retrovirus back to Exo1-deleted WT and Lig4 ^−/− cells (Supplementary Figure S5D), and both Exo1^WT and Exo1^EK fully rescued the near-null IgA switching in not only Exo1 ^−/− cells, but also Lig4 ^−/− Exo1 ^−/− cells (Figure 3A, Supplementary Figure S5E), indicating that exonuclease-embedded DSB resection function of Exo1 is not required for either c-NHEJ or A-EJ-mediated CSR. To further test whether Exo1 plays any role in joining of non-AID initiated DSBs, we introduced simultaneous blunt end breaks at Sμ and Sγ1 by CRISPR/Cas9 (Supplementary Figure S5F) and tested switching to IgG1 in WT, Lig4 ^−/− , Exo1 ^−/− , or double-mutant cells at 24, 48, and 72 h post-transfection. After normalization with transfection efficiency, efficient joining of Sμ-Cas9 DSBs to Sγ1-Cas9 breaks generated 40–60% of IgG1+ cells; as expected, Lig4 ablation reduced IgG1 switching efficiency by more than half to only 10–20% (Figure 3B, Supplementary Figures S5G,5H). In addition, we found that Exo1 deletion in either in WT or Lig4 ^−/− cells did not reduce Cas9-mediated IgG1 switching than the corresponding controls (Figure 3B). Taken together, we conclude that Exo1 is not required for joining AID or Cas9-generated DSBs by either c-NHEJ or A-EJ pathways.

We then asked whether DNA2/BLM-mediated long range resection is required for efficient A-EJ. Two specific shRNA efficiently knocked down the mRNA expression of DNA2 by 50–70% in both wild type and Lig4 ^−/− CH12F3 cells (Supplementary Figure S6A); accordingly, the IgA CSR efficiency in WT cells was slightly decreased by about 20%, and Lig4 ^−/− cells with shDNA2 exhibited an IgA CSR decline by 30% (Figure 3C, Supplementary Figure S6H). This deficiency can be at least partly attributed to impaired proliferation caused by DNA2 knockdown in WT and Lig4 ^−/− cells (Supplementary Figure S6B). To test the role of BLM in A-EJ, we first deleted with CRISPR/Cas9 the exon 8 of Blm gene upstream of the helicase domain; this mutation rendered ablation of BLM by premature termination of translation (Babbe et al., 2009) (Supplementary Figures S6C,S6D). The resultant BLM^Δhelicase cells were indistinguishable in IgA switching compared with WT cells, whilst BLM^Δhelicase in the Lig4 ^−/− background slightly but significantly reduced IgA CSR compared to control cells (Supplementary Figure S6E). Due to severe slow proliferation caused by helicase domain disruption, we generated another Blm mutation by Cas9 to delete the exon 19 that encodes Helicase-and-RNaseD-like-C-terminal (HRDC) domain of BLM (Supplementary Figures S6F,G). The HRDC domain interacts with the ATPase domain of BLM that may affect its helicase activity (Newman et al., 2015) and has been shown to be required for annealing of complementary single-strand DNA and Holliday junction resolution (Wu et al., 2005; Newman et al., 2015; van Wietmarschen et al., 2018). Although BLM^ΔHRDC CH12F3 cells proliferate and switch to IgA normally, BLM^ΔHRDC in Lig4 ^−/− background displayed a substantial decrease in CSR efficiency (Figure 3D, Supplementary Figures S6l), indicating that the HRDC domain of BLM is required for A-EJ, but not c-NHEJ-mediated class switching.

To gain more insights on the mechanism of how BLM participates in A-EJ, we performed HTGTS assay in WT and Lig4 ^−/− cells with BLM^ΔHRDC mutation and analyzed the pattern of S-S joining and MH in these cells (Supplementary Figure S7). Compared with WT cells, the BLM^ΔHRDC mutant exhibited a slightly decrease in the proportion of Sα junctions and a mild increase in junctions involving long Sα resection (Supplementary Figures S7B,C). When examining MH pattern of Sμ-Sα junctions, we recovered no significant difference between WT and BLM^ΔHRDC mutant cells (Supplementary Figure S7D). In contrast, a significant decrease in the proportion of long Sα resection junctions in Lig4 ^−/− cells with BLM^ΔHRDC mutation was observed (Figures 3E,F), indicating the HRDC activity is required for the joining of long-resected Sα breaks in Lig4 ^−/− cells. Similar to aforementioned cells in WT background, no significant difference in the MH profile in Sμ-Sα junctions was observed in BLM^ΔHRDC cells compared with the corresponding Lig4 ^−/− control (Figure 3G).

AID-initiated DSB recruits and activates ATM. While ATM positively regulates c-NHEJ-mediated CSR, its role in A-EJ has not been carefully examined. To this end, we first applied a highly selective ATM inhibitor AZD1390 that suppressed c-NHEJ mediated CSR at low concentrations with no effect on germline transcription and AID expression (Figure 4A, Supplementary Figures S8C,E). Treating Lig4^−/− cells with AZD1390 further reduced IgA switching by about half (Figure 4A). The CSR reduction by AZD1390 treatment in Lig4 ^−/− cells was phenocopied by Atm knockout (Figure 4B, Supplementary Figure S8A). In addition, we observed more phosphorylation of KAP1 that was ATM-dependent in IR-irradiated Lig4-deficient cells compared with WT control, indicating persistent ATM activation in A-EJ cells (Supplementary Figure S8B). Together, we concluded that ATM kinase activity is required for both c-NHEJ and A-EJ-mediated CSR.

We then examined the Sμ-Sα joining pattern with ATM deletion and ATM kinase inhibition by the HTGTS assay. Atm knockout or kinase activity inhibition in CH12F3 cells similarly resulted in increased Sα DSBs resection, increased MH usage, and decreased ratio of deletional versus inversional Sμ-Sα joining (Figures 4C–E). Surprisingly, in Lig4 ^−/− cells with Atm ablation, although the percentage of Sα deletional joining over total IgH junctions decreased (Supplementary Figures S8F-H) that was consistent with further reduced IgA surface expression than Lig4 ^−/− cells by flow cytometry, other parameters including Sα DSBs long resection, MH usage, and ratio of deletional versus inversional Sμ-Sα junctions remained unchanged compared with controls (Figures 4C–E). We concluded that while ATM plays important roles in both c-NHEJ and A-EJ-mediated CSR, it did not appear to control Sα long resection in Lig4-deficient cells. We also examine the potential role of another PIKK, DNA-PKcs in CSR. Consistent with the reported role of DNA-PKcs in promoting c-NHEJ to IgG in primary mouse B cells (Franco et al., 2008; Callén et al., 2009), while deleting DNA-PKcs by CRISPR/Cas9 (Supplementary Figures S9A–D) significantly diminished IgA in WT CH12F3 cells, knocking out DNA-PKcs in Lig4 ^−/− cells did not further reduce IgA CSR (Supplementary Figure S9E), indicating that DNA-PKcs does not play a role in A-EJ.

Previous reports suggested ATM may promote DSBs resection through phosphorylating CtIP as the CtIP-T859A mutant compromised resection and HR repair (Peterson et al., 2013; Wang et al., 2013). To investigate the relationship between ATM and Mre11/CtIP in end joining during CSR, we first silenced Mre11 or CtIP expression by shRNA in ATM ^−/− cells (Supplementary Figure S10A). While shMre11 in ATM ^−/- cells showed only a mild effect on IgA + cells by surface staining, silencing CtIP slightly decreased (∼20%) IgA switching compared with scramble control (Figure 5A). However, in Lig4 ^−/− ATM ^−/− cells, silencing either Mre11 or CtIP (Supplementary Figure S10B) conferred significantly more defect in CSR up to 50% lower than scramble control, a phenotype much more severe than that in Atm deletion alone (Figure 5B), suggesting that Mre11 and CtIP function in part via different pathways than ATM in promoting A-EJ in Lig4-deficient cells.

To further confirm the above observation, we treated Mre11 or CtIP-silenced WT or Lig4 ^−/− cells with DMSO or AZD1390 before stimulating them for CSR. Consistent with aforementioned findings, silencing only CtIP but not Mre11 in DMSO-treated WT cells resulted in a mild IgA switching defect (Figure 5C). As expected, ATMi treatment decreased IgA switching of scramble control cells compared with DMSO treatment; no significant difference in switch efficiency was observed between ATMi-treated control and shMre11 cells (Figure 5C). In contrast, ATMi-treated shCtIP cells showed further IgA defect than scramble control cells under the same treatment (Figure 5C). These data confirmed an epistatic effect between ATM and Mre11, and a non-overlapping function of ATM and CtIP in c-NHEJ-mediated cells. In stark contrast, ATMi treatment indeed caused more severe switching defect in Lig4 ^−/− cells infected with either shMre11 or shCtIP compared with that of scramble control (Figure 5D). Taken together, these data implicated that while ATM and Mre11 are epistatic in c-NHEJ-mediated CSR, it functions at least in part independent of Mre11 and CtIP in promoting A-EJ-mediated CSR in Lig4 ^−/− cells.

Ours and others’ previous reports had indicated that 53BP1-deficient B cells underwent greatly impaired CSR characterized by significantly increased S region DSB resection and MH usage in S-S junctions (Bothmer et al., 2013; Dong et al., 2015), representing a scenario similar to A-EJ comparing to c-NHEJ factor ablation. To investigate whether Mre11/CtIP-mediated resection plays any role in A-EJ in 53BP1-deficient cells, we first silenced these two factors with lentiviral expressed shRNA (Supplementary Figure S11A) and stimulated them for switching. However, in contrast to what we observed in Lig4 ^−/− cells, silencing Mre11 or CtIP in 53bp1 ^−/− cells did not lead to further defect in IgA switching compared with that of scramble control (Figure 6A). ATM kinase inhibition by AZD1390 in 53bp1 ^−/− cells resulted in very mild, if there was any, additive decline in switching than DMSO controls (Supplementary Figure S11B). Similarly, Atm gene knockout by CRISPR/Cas9 in 53bp1 ^−/− cells (Supplementary Figures S11C,D) did not change its switching efficiency (Figure 6B), suggesting that ATM and its kinase activity are not required for A-EJ in this setting.

We also deleted resection factor EXD2 in 53bp1 ^−/− cells and results showed EXD2 depletion did not further reduce IgA switching (Supplementary Figures S11E,F). Similar to the observations with WT CH12F3 cells, deleting Exo1 in 53bp1 ^−/− nearly completely abolished switching, and re-introducing either EXO1^WT or Exo1^E109K mutants back to the 53bp1 ^−/− Exo1 ^−/− double-mutant cells restored IgA switching to the 53bp1 ^−/− level (Supplementary Figures S11G–I), suggesting that Exo1-mediated DSB resection was not required for A-EJ in 53BP1-deficient cells. Taken together, it appears that neither short-range nor long-range resections factors were critical for A-EJ in 53BP1-deficient cells, and the requirement for resection activity in A-EJ-mediated CSR is context dependent.

To gain more insight into the effect of Mre11/CtIP and ATM kinase on class switch in 53bp1 ^−/− cells, we applied HTGTS assay with Mre11/CtIP-silenced or Atm-deleted 53bp1 ^−/− cells. Consistent with the IgA FACS staining data, HTGTS revealed similar levels of Sμ-Sα joining in all cells with shMre11 or shCtIP or ATM deletion compared with 53bp1 ^−/− control cells (Supplementary Figures S12A,B). Surprisingly, we found that Mre11 or CtIP silencing did not change the overall pattern of Sα joining in 53bp1 ^−/− cells in that deletional versus inversional Sα joins and levels of SαDSB resection were similar in shMre11 and shCtIP-infected 53bp1 ^−/− cells compared with those in scramble control cells (Figures 6C,D, Supplementary Figure S12C). However, we indeed observed a significant increase in the percentage of direct junctions and junctions with 1 base pair of microhomology sequences in shMre11 cells (Supplementary Figure S12D), implicating that Mre11 and CtIP may have different roles in mediating MH-mediated joining in this context. Atm gene deletion, on the other hand, resulted in significantly decreased Sα DSB long resection without changing MH usage in Sα junctions (Figures 6C,D, Supplementary Figure S12D). Taken together, we concluded that while only ATM activity is required for Sα DSB resection in 53BP1-deficient cells, neither Mre11/CtIP nor ATM is required for the residual switching in this setting.