Orientation Regulation of Class-switch Recombination in Human B Cells

During early B lymphocyte development, RAG initiates V(D)J recombination to assemble the V region of Ig genes, resulting in a vast repertoire of Ag receptors (1, 2). In mature B cells, class-switch recombination (CSR) allows previously rearranged Ig H chain V domains to be expressed in association with a downstream C region, leading to the production of different Ig isotypes with improved effector functions (1, 3). The CSR process depends on activation-induced cytidine deaminase (AID) (4), which converts deoxycytidines to deoxyuridines in repetitive intronic sequences upstream of each C region, referred to as switch (S) regions. IgD expression is usually the result of alternative splicing of one long RNA encoding both IgM and IgD or, very rarely, recombination between Sµ and the cryptic S region–like sequence upstream of Cδ (5–8). Switching to IgG, IgA, and IgE is controlled by the IgH 3′-regulatory region (3′RR), whereas enigmatic IgD CSR occurs in the absence of this regulation (8).

Both V(D)J recombination and CSR require DNA double-strand breaks (DSBs) as intermediates, and the recognition, processing, and joining of DSBs generated during these processes require the classical nonhomologous end joining (c-NHEJ) pathway. c-NHEJ is initiated by the recruitment of the core DSB sensors Ku70/Ku86, which facilitate the recruitment of several accessory factors, including DNA-PKcs, Artemis, and XLF (9–12). DNA ligase 4 and its cofactor XRCC4 subsequently repair DSBs by end joining. The alternative end joining (A-EJ) pathway functions in parallel with c-NHEJ factors, although at lower rates and with a greater dependence on junctional microhomologies (MHs) (1). Although A-EJ is excluded from the V(D)J recombination process, it is sufficient to support some levels of CSR, as shown in various c-NHEJ–deficient mouse models or DNA repair factor–deficient patients (13–22).

Linear amplification–mediated high-throughput genome-wide translocation sequencing (LAM-HTGTS) is a method that allows investigation of the patterns of DSB-induced alterations, such as deletions, insertions, inversions, and inter- and intrachromosomal translocations (23–26). Studies based on LAM-HTGTS in murine CSR-activated lymphocytes and lymphoma CH12F3 cell lines have demonstrated that AID-dependent DSBs generated during CSR are programmed to join primarily in a productive orientation, resulting in deletional recombination and excision circles (24). Inversional joining events are rare during CSR-activated repair in wild-type murine cells. In contrast, ISceI-mediated or Cas9 gRNA-generated DSBs are prone to both deletional and inversional joining without significant orientation bias, even when they occur outside or in place of S regions within the Igh locus (1, 24). Similar to the mechanism underlying RAG chromatin scanning during V(D)J recombination (4, 27, 28), the preference of deletional end joining of CSR in murine B cells has recently been explained by a chromatin loop extrusion model (29). Cohesion-mediated loop extrusion supports the formation of CSR centers in naive B cells. In Ag-stimulated mature B cells, this mechanism further promotes the alignment and deletional joining of the acceptor and donor S regions in the CSR centers (29).

However, little is known about the regulation of CSR orientation in human B cells. In this study, we applied LAM-HTGTS to develop a high-throughput assay for systematically analyzing CSR patterns involving all Ig isotypes in human B cells. This enables us to study the orientation of various Ig class/subclass switching events in an unbiased and quantitative manner. We showed that deletional recombination is dominant in human B cells, with some variations depending on the Ig isotypes (IgG versus IgA) and the age of individuals (children versus adults). Furthermore, we compared the characteristic features of CSR in healthy donors with those in patients deficient in DNA-PKcs or Artemis, two c-NHEJ factors. On the basis of these comparisons, we identified two novel, to our knowledge, CSR signatures and demonstrated that features of S regions and the efficiency of the c-NHEJ machinery are important for the regulation of CSR orientation in human B cells.

Two patients with c-NHEJ defects were included in the study. The DNA-PKcs–deficient patient has previously been described (30, 31). She was diagnosed with T⁻B⁻NK⁺ radiosensitive SCID at 5 mo of age and carried disease-causing homozygous mutations (c.9185 T>G; L3062R) in the PRKDC gene. B and T cells were almost absent from her peripheral blood, but NK cells were detected (30). The case of an Artemis-deficient patient has not been described before. He was 2 y old, and the second child was born to Iranian consanguineous parents. His birth history was unremarkable, and he was well up until the age of 5 mo, when he was referred to the immunodeficiency clinic due to chronic fever, oral thrush, recurrent respiratory infections, and failure to thrive. At admission, he also presented with lymphadenopathy and splenomegaly due to BCGosis and was treated with antituberculosis (rifampin, isoniazid, ethambutol), antiviral (acyclovir), and antifungal (fluconazole) medicines. On the basis of the severity of manifestation and the presence of lymphopenia (lymphocytes 540 cells/µl, leukocytes 4400 cells/µl) in the initial evaluation, an advanced immunologic work-up (Table I) was conducted, a diagnosis of T⁻B⁻NK⁺ SCID was made, and i.v. Ig replacement therapy was initiated. The absence of a mitogen-induced proliferative response was evident in the patient, and he was a candidate for hematopoietic stem cell transplant for 2 y. Unfortunately, before the identification of an appropriate donor, he developed multiple complications, including pulmonary aspergillosis, and died of respiratory failure. Genetic analysis revealed that the patient carried a homozygous deletion in exons 1–3 of the DCLRE1C gene. The study was approved by the institutional review boards at the Karolinska Institutet.

PBMCs were isolated from patient blood samples, buffy coats from healthy adult blood donors (n = 7), or young healthy donors (1–8 y old; n = 9). Genomic DNA was subsequently extracted using DNeasy blood and tissue kits (Qiagen, Hilden, Germany).

B cells were enriched from PBMCs from adult blood donors (n = 3) by an EasySep human B cell enrichment kit (STEMCELL Technologies, Vancouver, BC, Canada), subsequently cultured in RPMI 1640 medium supplemented with 10% FBS plus 1% penicillin-streptomycin (Thermo Fisher Scientific, Waltham, MA), and stimulated with 10 ng/ml IL-10 (Immunotools, Friesoythe, Germany) together with 300 ng/ml CD40L (AdipoGen Life Sciences, San Diego, CA). After 3 d of stimulation, the cells were collected, and genomic DNA was extracted using DNeasy blood and tissue kits (Qiagen).

LAM-HTGTS was performed as described with modifications and optimization for the specific amplification of human CSR junctions (26) (Fig. 1). Briefly, 10–20 µg of genomic DNA was collected and sonicated using the Bioruptor system for three cycles, with 25 s of working time and 60 s of resting time in each cycle under “low energy” output. The DNA smear ranged from 0.2 kb to 2 kb, with a peak at ∼750 bp, and was validated using agarose gel electrophoresis. Linear amplification PCR was subsequently performed with a human Sµ biotin primer located 150 bp upstream of the core repetitive Sµ region (5′-/5BiosG/GGGTATCAAGTAGAGGGAGACAA-3′). The bridge adapter was ligated to the 3′ end of the fragments (adapter-upper 5′-GCGACTATAGGGCACGCGTGGNNNNNN-NH2, adapter-lower 5′-phosphorylation/CCACGCGTGCCCTATAGTCGC-NH2), and the second PCR was performed with I5/I7 primers (I5 5′-CCCTACACGACGCTCTTCCGATCTtcagcgAGCCTGGCTGTGCAGGAAC-3′, I7 5′-CAGACGTGTGCTCTTCCGATCTCGATCTGACTATAGGGCACGCGTGG-3′), followed by the third PCR to add the Illumina sequencing primer and a barcode to the fragments. Finally, the amplified DNA products were run on 1% agarose gels in TAE buffer, and fragments between 500 bp and 1000 bp were excised, purified using a QIAquick gel extraction kit (Qiagen), and sequenced on the MiSeq platform (2 × 250 bp or 2 × 300 bp). Various “prey” sequences were identified (“caught”) using the Sµ primer “bait” (Fig. 1); these sequences included S region sequences involved in deletional and inversional recombination events and “off-target” sequences located outside the S regions.

The sequencing data were analyzed following a previously described pipeline (2) with modifications to detect human CSR junctions, including those resulting from sequential switching events. Briefly, the raw sequence reads were first aligned to the hg19 human reference genome, and the junctions were detected and filtered by default parameters. After running the previously described pipeline (2), we recovered two types of junctions on the basis of S region features. First, we recovered all junctions with prey sequences localized in IGH loci that were originally removed due to alignment with Sµ region sequences. Second, the sequence reads that included more than two S region fragments were recovered. After removing the duplicated reads, a dataset of unique junctions was generated and used in the subsequent analyses.

Alignments between read sequences and the reference genome were determined using a score-based greedy extension approach. Two points were awarded for each matched base pair, whereas each mismatched base pair resulted in a deduction of 10 points. A successful alignment was defined as having a score that continued to increase to at least 50 points. The alignment end or breakpoint is determined next to the mismatch if fewer than five matches are found after that mismatch. The remaining read sequence was then aligned using the same scoring approach. The bait primer was designed to locate in the Sμ region, so the bait alignment should also be in the Sμ region (the donor S region). Prey is defined as the next alignment on the read immediately following the “bait” and not aligned to adapter sequences. Prey start breakpoints located in S regions were included in the analysis of CSR junctions, and the breakpoints outside the S regions were analyzed as off-target events. The coordinates of each S region in hg19 are as follows: Sµ, chr14:106322322–106328217; “Sδ”, chr14:106312010–106318298; Sγ3, chr14:106237742–106304737; Sγ1, chr14:106209407–106233251; Sα1, chr14:106175001–106188131; Sγ2, chr14:106111126–106134586; Sγ4, chr14:106092402–106109540; Sε, chr14:106068064–106090813; Sα2, chr14:106054731–106066403; and “ψSε": chr14: 106188864–106207810. The coordinates for the two 3′RRs are as follows: 3′RR1, chr14:106136377–106173505; 3′RR2, chr14:106023274–106053274.

MH is characterized by the overlap between the Sμ alignment (bait) and a downstream S region alignment (prey) on the reads. There may be at most one or two mismatches within the MH region, determined by the scoring approach.

The patient samples were tested in parallel with age-matched control subject samples. In total, 9 and 10 experiments were performed using DNA from PBMCs from seven healthy adult donors and nine healthy pediatric controls, respectively, and seven experiments were performed using DNA samples from IL-10+CD40L-stimulated B cells from three healthy adult donors. Two experiments were performed on samples from DNA-PKcs–deficient and Artemis-deficient patients. The data from the various experiments were reproducible (with largely similar patterns of CSR junctions), both when considering the same donor in repeated experiments and when comparing several healthy donors. When comparing the frequency of each type of CSR junction, the data from different experiments were first normalized to 1000 total junctions per sample per experiment and subsequently merged to generate the respective control and patient groups.

The analysis of CSR junctional signatures was performed using a nonnegative matrix factorization (NMF)-based unsupervised learning algorithm (32, 33). Junctions with only the Sμ region involved (Sµ-Sµ) or with nontemplate insertions exceeding 3 bp were excluded from this analysis. The remaining junctions were categorized into 64 types per sample. This categorization considered eight major acceptor S regions (Sδ, Sγ3, Sγ1, Sα1, Sγ2, Sγ4, Sε, and Sα2), each with eight types of features at the junctions (1–3-bp insertion, blunt-end joining, with MH of 1–5 bp, 6–10 bp, 11–20 bp, 21–40 bp, >40 bp, or inversion). Subsequently, the prevalence of each junctional type in every sample was calculated. Next, NMF was employed on the prevalence matrix (junctional types × samples) interactively 100 times for different numbers of signatures (1–10). The optimal number of signatures was determined on the basis of a high reproducibility (>90%) and a relatively small number of errors (34). Finally, the optimal signature features (junctional types × signatures) and signature exposure of each sample (signatures × samples) were used to visualize the results.

One million PBMCs per donor were stained for viability with live/dead Aqua (Thermo Fisher Scientific) at room temperature for 20 min in PBS in the dark. The cells were subsequently washed and eventually stained with IGS-1 panel Abs (BD Biosciences, Franklin Lakes, NJ) at a final volume of 100 µl, along with the following Abs: BUV737 CD27, BUV395 IgM, BV786 CD38, BV421 IgD, and PE-Cy7 CD19, and PE-Cy5 CD20. The samples were fixed with 0.4% paraformaldehyde for 10 min at 4°C, washed, and resuspended in acquisition buffer. All the data were collected with a Sony ID7000TM spectral cell analyzer (Sony Biotechnology Inc., CA) and analyzed with ID7000 software (Sony, version 1.1.0.11041) and FlowJo version 10.8.2 software (BD Biosciences, Ashland, OR).

CSR junctions with insertions ≥4 bp were further analyzed by realigning the inserted fragments to S region sequences using BLAST 2.9.0+. To obtain more precise BLAST results, we included additional bases before and after the breakpoint; that is, 6 bp were added from each side to the 4–7-bp insertions, resulting in 16–19 total bp, and 5 bp were added from each side to the 8–10-bp insertions, resulting in 18–20 total bp. The following criteria were used for alignment: (1) The alignment length should be more than half the inserted fragment length; (2) the inserted fragment sequences had greater than 90% identity with S region sequences; and (3) the alignment S region sequences with the highest identity were identified as the alignment results. All the alignments that met the above criteria were included in the analysis of junctions derived from sequential switching. When one insertion had identical alignment to more than one S region sequence, the count was divided by the number of alignments.

The joining between Sµ and regions outside the IGH loci was referred to as off-target events. To characterize hotspots of off-target events, we examined breakpoints across the entire genome. We first partitioned the genome into 1-Mb bins and calculated the number of junctions in each bin. We then defined a bin as an off-target hotspot if the number of junctions in that region was more than 5-fold greater than the SD of the number of junctions observed genome-wide. The breakpoints identified in off-target hotspots were then annotated using ANNOVAR and compared with potential fragile sites from the HumCFS database, previously reported recurrent DNA break cluster genes (12), and AID off-targets in human or mouse studies (35–37).

The nonparametric Mann–Whitney U test was used for the statistical analyses.

To determine the overall profile of CSR in human B cells, we adapted the LAM-HTGTS–based assay previously used to study CSR in mouse B cells (24) to amplify CSR junctions from human cells (Fig. 1). By applying this method, we first analyzed the distribution of Ig classes/subclasses in PBMCs from healthy donors. Using a human Sµ primer as bait, we obtained and analyzed 28,679 unique junctions from healthy adult donors (PBMC.adu) and 22,839 unique junctions from healthy children (PBMC.ch). Close to half of the detected junctions have only Sµ sequences (Sµ-Sµ), which could be either small deletions or internal recombination within the donor Sµ region, representing attempted/failed CSR events (Fig. 2A). The remaining junctions represent authentic or successful CSR involving both donor and acceptor S regions. In healthy adults, recombination junctions resulting from switching to IgA1 (Sµ-Sα1) and IgG1 (Sµ-Sγ1) were the most frequently detected, comprising 30.39% and 23.73% of the total authentic CSR junctions, respectively. These were followed by other Ig subclasses in the following order: Sα1>Sγ1>Sα2>Sγ2>Sγ3>Sγ4>Sε (Fig. 2B). The CSR profile in healthy adult donors detected by the LAM-HTGTS method largely correlated with the serum levels of Ig classes/subclasses in adults, considering the shorter half-life of IgA compared with IgG. Furthermore, in selected donors tested, the CSR profile based on the CSR junctions generally matched the Ig class/subclass distribution measured by spectral flow cytometry (Supplemental Fig. 1). In healthy children, switching to IgA1 was the most frequent event (35.46% of total authentic CSR junctions), with the order of Sα1>Sγ1>Sα2>Sγ3>Sγ2>Sγ4>Sε (Fig. 2B). Compared with adults, pediatric donors showed significantly less switching to IgG2 (3.92% versus 10.22% in adults; Mann–Whitney U test; p = 0.0004) (Fig. 2B, 2C), consistent with previously described serology data (38). Notably, we also detected junctions between Sµ and the region flanked by Cµ and Cδ, which consists of the cryptic “Sδ” (Supplemental Fig. 2). These junctions accounted for 0.39% and 0.86%, respectively, of the total authentic CSR junctions in the PBMC.adu and PBMC.ch samples, suggesting that in addition to alternative splicing, bona fide IgD CSR can occasionally occur in human B cells (Fig. 2B). Moreover, we detected rare recombination (0.21–0.28%) between Sµ and the repetitive sequences upstream of ψCε, which is a truncated pseudogene unable to encode the ε chain but retains a C_H4 exon. These repetitive sequences showed high similarity with Sε sequences and can thus be assigned as a “ψSε” region. Finally, we detected exceedingly rare (0.02–0.12%) recombination events between Sµ and the repetitive sequences within the two 3′RRs (Fig. 1E, 1F; Fig. 2B), which may result in the expression of ψCγ (3′RR1) or potentially locus suicide recombination (3′RR2), deleting the entire IgH C region gene cluster (39).

We further analyzed the distribution of Ig classes/subclasses in purified B cells from healthy adult donors. These B cells (referred to as “Sti-B”) were cultured in vitro for 3 d with a cytokine combination that is known to activate CSR in human B cells (IL-10+CD40L) (3, 40, 41). Altogether, 173,334 unique junctions were characterized from these activated B cells, with a significantly greater proportion of authentic CSR junctions than in PBMCs (58.66% compared with 44.34% in PBMC.adu; Mann–Whitney U test; p = 0.0057) (Fig. 2A). This combination of activators also drove switching to all IgG subclasses, resulting in proportionally increased Sµ-Sγ junctions compared with Sµ-Sa junctions (Sγ1>Sα1>Sγ2>Sα2>Sγ3>Sγ4>Sε; Fig. 2B–2D).

Using the same LAM-HTGTS technique, we further analyzed the overall CSR profiles of PBMCs from two patients with mutations in genes encoding the c-NHEJ factors (DNA-PKcs and Artemis). Both patients had T^-B^-NK⁺ SCID clinical presentations, but mutations in the respective genes may still cause a “leaky” phenotype, because a small number of B cells can be detected in Artemis-deficient patients, and a few CSR junctions can be amplified from DNA-PKcs–deficient patients, using a nested PCR approach (31). Using the more sensitive and higher-throughput LAM-HTGTS method, we could indeed obtain more CSR junctions from these patients, and the results were compared with those from healthy children due to their young age (Fig. 2). In DNA-PKcs– and Artemis-deficient patients, a lower proportion of authentic junctions was observed (Fig. 2A).

The vast majority of recombination events occurred between the Sµ and Sa regions in DNA-PKcs– and Artemis-deficient cells, whereas Sµ-Sγ junctions were barely detectable (Fig. 2B–2D). Notably, the proportion of Sµ-Sε recombination appeared to be greater in Artemis-deficient patients (5.09% versus 1.54%) than in healthy children; however, this difference was not significant (Mann–Whitney U test; p = 0.2857). In conclusion, our unbiased analysis revealed markedly lower overall CSR efficiency in patients with c-NHEJ deficiency. This is reflected by the lower proportion of authentic junctions detected in these patients. Additionally, the distribution of Ig class/subclass switching was altered, with an almost complete shift to IgA switching in the DNA-PKcs– and Artemis-deficient cells.

We next analyzed the proportion of deletional (productive) and inversional (nonproductive) junctions in cells from patients and control subjects. In PBMCs from healthy adults and children, deletional recombination events were clearly dominant (92.5–95.7% of the total authentic CSR junctions; Fig. 3A). Notably, inversions were less frequently observed at Sµ-Sα junctions than at Sµ-Sγ junctions in control subjects (5.6% versus 12.2% in adults, p = 0.0006; 2.9% versus 12.0% in children, p < 0.0001; Mann–Whitney U test). A comparison of the two healthy donor groups revealed that the proportion of inversions was similar at Sµ-Sγ junctions (12.2% versus 12.0%) but slightly lower at Sµ-Sα junctions (2.9% versus 5.6%; Mann–Whitney U test; p = 0.0512) in children than in adults. Deletional recombination events were also dominant in DNA-PKcs– and Artemis-deficient patients (99.3% and 94.7%, respectively). Too few Sµ-Sγ junctions were detected in these patients to allow an accurate comparison with those identified for Sµ-Sα junctions.

Because nonproductive recombination of B cells due to inversions may theoretically be selected against, we further analyzed short-term cultured B cells from healthy donors. Deletional recombination events remained dominant in the Sti-B samples (92.3%); however, the proportion of inversions was significantly greater at Sµ-Sα junctions than in the PBMCs (13.2% versus 5.6% in PBMC.adu; Fig. 3A). Overall, as described in mouse models (29), CSR is programmed for deletional recombination in human B cells, with some variations depending on the isotype, cell activation, and age of the individuals.

CSR junctions occurred broadly across the S regions, with a vast majority of the breakpoints located within the repetitive sequences (core S region sequences) in both c-NHEJ–deficient patients and control subjects (Fig. 3B). We next used an NMF-based method, which has been applied to explore mutagenesis processes in the cancer genome (32, 33), to identify potential signatures of CSR junctions (i.e., unique combinations of features of CSR junctions). We cataloged 64 main recombination types between donor Sµ and acceptor S regions (eight main acceptor S regions × eight features of the junctions [1–3-bp short insertions, blunt-end joining, various lengths of MHs or inversions]) and extracted two distinct CSR junctional patterns/signatures from our dataset (Fig. 4A). Signature 1 consists of both deletional (87%) and inversional (13%) junctions and both Sµ-Sγ (72%) and Sµ-Sα (28%) junctions. Signature 2 consists almost entirely of deletional junctions (99%), with a dominance of Sµ-Sα junctions (96%) and low frequencies of Sµ-Sε (3%) and Sµ-Sγ (1%) junctions. Importantly, the vast majority of signature 1 CSR junctions had MHs less than 6 bp, regardless of the Ig isotype, whereas most of the signature 2 CSR junctions had at least 6 bp MHs (Fig. 4A). When focusing only on Sµ-Sα junctions, they can be further divided into two groups on the basis of overall signatures: One group consists of junctions with shorter MHs that can result from either deletional recombination or inversions, and the other group consists of junctions exclusively due to deletional recombination and is characterized by longer MHs (Supplemental Fig. 3). Notably, more than 97% of the junctions in DNA-PKcs–deficient cells and all (100%) junctions in Artemis-deficient cells were assigned to signature 2 (Fig. 4B). Thus, signature 2 likely results from MH-mediated A-EJ activity in the absence of c-NHEJ factors such as DNA-PKcs or Artemis. Conversely, signature 1 might represent c-NHEJ activity. In PBMCs from healthy adults and children, signature 1 was dominant (66% and 52%, respectively), although signature 2 was also present. In cytokine-stimulated, short-term cultured B cells, signature 1 became even more dominant (97%), suggesting that c-NHEJ remains the major pathway contributing to CSR, particularly in IgG switching, in normal B cells.

To explore the reason underlying the different requirements for c-NHEJ during IgA/IgG switching and deletional/inversional recombination during CSR, we further analyzed MH use at the recombination junctions and the degree of sequence homology between different S regions (Fig. 4C–4E). For this analysis, CSR junctions with short nontemplate insertions were excluded. In healthy donors, CSR junctions resulting from deletional recombination between Sµ and Sγ were characterized by the absence, or only short stretches, of MHs (Fig. 4C). The average lengths of MHs for all Sµ-Sγ junctions were 2.7 bp and 2.3 bp in the PBMC.adu and PBMC.ch groups, respectively. In the same donors, Sµ-Sα junctions due to deletional recombination were characterized by overall longer MHs, with average lengths of 9.1 bp in the PBMC.adu samples and 12.2 bp in the PBMC.ch samples (Fig. 4C). In c-NHEJ–deficient patients, a further and significant shift to MH-based joining was observed at the Sµ-Sα junctions (16.4 bp in DNA-PKcs–deficient cells, 16.3 bp in Artemis-deficient cells; Fig. 4C). A similar shift to that of longer MH-based joining was observed at the Sµ-Sµ internal deletional junctions (Fig. 4C). These findings can potentially be explained by the substantially greater sequence homology between Sµ and Sα than between Sµ and Sγ (Fig. 4E, Supplemental Fig. 2). Thus, in normal B cells, MH-mediated end joining is more frequently used during deletional recombinational switching to IgA versus IgG, possibly due to the difference in the degree of homology between the donor and acceptor S regions. When the function of c-NHEJ is reduced or impaired, IgA switching is relatively less affected, because the high homology between the Sµ and Sα regions can still support some level of MH-mediated A-EJ. In contrast, IgG switching is significantly compromised under these conditions, because the recombination between the Sµ and Sγ regions is more dependent on c-NHEJ due to the low homology between the two S regions.

We proceeded with an analysis of MH use at CSR junctions resulting from inversional recombination. Among healthy donors, these inversional junctions consistently demonstrated either a lack of or short MHs, regardless of the Ig class involved (Fig. 4D). Although no significant changes were detected between the Sµ-Sγ (20) and Sµ-Sγ junction groups, the average MH length was notably shorter at the Sµ-Sα (20) junctions (2.3 bp in Sti-B, 1.8 bp in PBMC.adu, and 2.9 bp in PBMC.ch) than at the deletional Sµ-Sα junctions (6.2 bp in Sti-B, 9.1 bp in PBMC.adu, and 12.2 bp in PBMC.ch) in normal B cells (Fig. 4D). Similar results were observed for the Sti-B samples. Only a very limited number of inversional Sµ-Sα (20) junctions from Artemis-deficient patients were available for analysis, but these junctions also exhibited notably shorter average lengths of MHs than did the deletional Sµ-Sα junctions (3.3 bp versus 16.3 bp; Fig. 4C, 4D). Similar results were obtained at the Sµ-Sµ junctions in comparison with the Sµ-Sµ internal deletion junctions (Fig. 4C, 4D). These findings can be explained by the significantly reduced homology between the Sµ locus and the reverse orientation of the Sα1-2 and Sγ1 − 4 loci as well as the Sµ region itself compared with the forward orientation of these S regions (Fig. 4E, Supplemental Fig. 2). Thus, because of the nature of S region sequences, inversions more strictly rely on c-NHEJ, because MH-based A-EJ cannot be efficiently used when the acceptor S region is in a reverse orientation. This specifically affects IgA switching, because the low homology between the Sµ and Sγ regions to begin with (in the cis direction) means that deletional recombination between the two regions already heavily relies on c-NHEJ. When the function of c-NHEJ is impaired, as in DNA-PKcs– and Artemis-deficient patients, in addition to significantly compromised IgG switching, the orientation of IgA switching will also be affected, where almost only the in cis orientation is possible.

The chromatin loop extrusion-mediated CSR model may also explain sequential switching events (28, 29). After the recombination of the Sµ region and the first acceptor S region, chromatin can undergo further remodeling. This allows a subsequent acceptor S region to come into proximity with the already recombined S regions, setting the stage for another round of CSR. An alternative pathway for sequential CSR, referred to as “reversed sequential CSR,” which involves recombination of the downstream S regions followed by switching to the donor Sµ region, has also been proposed (42). Using the LAM-HTGTS–based method, we identified a small proportion of switching events involving three S regions in activated B cells and PBMCs from both control groups and c-NHEJ–deficient patients. Some of these events were described as small deletions or internal recombination, with two distinct sequences of the same S region (e.g., Sµ-Sµ-Sα2) or three distinct Sµ pieces (Sµ-Sµ-Sµ). Other events represented bona fide sequential switching events, where CSR occurred between Sµ and the first acceptor S region (S2) and then between S2 and an additional acceptor S region (S3) (Fig. 5A).

Multiple types of sequential switching were identified in the healthy donors (Fig. 5B, 5C). Significantly more sequential switching events were observed in PBMCs from healthy adults than in those from children (1.02% versus 0.69% of total authentic CSR events; Mann–Whitney U test; p = 0.0268). The most common sequential switching events in both healthy donor groups involved the two Sα regions (Sµ-Sα1-Sα2), and sequential switching events involving Sγ3 regions (or reversed Sγ3) were also frequent in these groups. In IL-10+CD40L-activated B cells from adult control subjects, junctions consisting of more than two S fragments were identified in 0.54% of the total authentic CSR junctions, which was significantly lower than that in the adult PBMC control subjects (Mann–Whitney U test; p = 0.02). The most common sequential switching events in these activated B cells are Sµ-Sγ3-Sγ1 and Sµ-Sα1-Sγ1, which is in line with the overall increase in IgG CSR events in these cells. Some CSR junctions identified in various control groups (e.g., Sµ-Sα2-Sα1 and Sµ-Sα1-Sγ1) may have resulted from transswitching involving both chromosomes. Similarly, sequential switching involving inversions such as Sµ-Sγ3(rev )-Sα2 might also have resulted from transswitching activity. Overall, the junctions detected due to transswitching were exceedingly rare, accounting for less than 0.42% of the total authentic CSR events.

Sequential switching events were also identified in c-NHEJ–deficient patients, with overall frequencies of 0.65% and 0.68% for DNA-PKcs– and Artemis-deficient patients, respectively (Fig. 5B, 5C). No significant difference in the frequency of various types of sequential switching was observed between the patient and PBMC control groups, probably because only a very small number of junctions were detected in the patients.

In addition to junctions between Sµ and downstream S regions and other repetitive sequences located in the IGH loci, we detected numerous joining events between Sµ and regions outside the IGH loci. These recombination events were referred to as off-target events (Fig. 1I). Thirty-five off-target hotspots were identified in patients and control subjects (Supplemental Fig. 4). Notably, 4 of these hotspot regions (11.4%) were associated with previously described AID off-target genes, 12 (34.3%) were located at chromosomal fragile sites, and 4 (11.4%) were related to previously described recurrent DNA break cluster genes (12), suggesting that our method can detect both AID-dependent and AID-independent rare recombination events in human B cells, as in mouse B cells (26). AID-independent events could be related to errors in DNA replication or rare insertions in S regions, as previously described in patients with chronic infections (43, 44) (Table I).

Using LAM-HTGTS technology, we developed a method that can simultaneously detect all types of CSR junctions in human B cells in an unbiased, high-throughput, and quantitative manner. We showed that in PBMCs from healthy adults and children, Sµ-Sα junctions, followed by Sµ-Sγ1 junctions, were the most frequent types of CSR junction detected. Compared with those in healthy adult donors, a slight increased proportion of Sµ-Sα junctions and a significantly reduced proportion of Sµ-Sγ2 junctions were observed in healthy children. Furthermore, we demonstrated that productive, deletional recombination is strongly favored in both PBMCs and cytokine-stimulated B cells. This orientation specificity supports a major role for the chromatin loop extrusion-mediated CSR mechanism in human B cells, as elucidated in mouse B cell studies, where in resting B cells, loop extrusion juxtaposes the intronic enhancer and Sµ region and the 3′RR enhancer to function as a CSR center. In activated B cells, transcription subsequently activates cohesin loading, leading to secondary loop extrusion that directionally aligns a downstream acceptor S region with the donor Sµ region for deletional CSR (28, 29).

To further explore the potential mechanisms that differentially regulate the orientation of IgG and IgA CSR in human B cells, inspired by cancer genome mutational signature analysis, we identified two major signatures of CSR junctions by applying a similar NMF-based method. The first signature is IgG dominant (72%) and deletional recombination dominant (87%), with no or short MHs at the CSR junctions, likely reflecting c-NHEJ activity in normal human B cells; the second signature is characterized by an almost complete shift to IgA switching (96%) and deletional recombination (99%) and significantly longer MH use at the recombination junctions, which may represent MH-based A-EJ activity in the absence of c-NHEJ factors such as DNA-PKcs or Artemis. By analyzing the junctional MH and the degree of sequence homology between different S regions, we further suggested that in human B cells, the orientation and pattern of CSR are also influenced by distinct features of the S regions involved. Because of the relatively low homology between the Sµ and Sγ regions, even when the two S regions are aligned in the right direction, deletional recombination relies on the efficiency of the c-NHEJ machinery. As a result of the high homology between the Sµ and Sα regions, in cis, but not in the reverse direction, deletional but not inversional recombination during IgA switching can be supported by A-EJ. Additional patients with c-NHEJ deficiency, including ligase IV– and XRCC4-deficient patients (14, 45), should be tested in the future to validate our results.

There are additional differences in the gene organization of the Ig H chain C region (IGHC) locus between humans and mice, which might contribute to the differential regulation of IgG and IgA CSR (3). For example, the human IGHC gene locus contains nine known functional genes and two pseudogenes organized into two γ-γ-ε-α blocks, with two 3′RRs located downstream of the Cα1 and Cα2 regions, respectively. Furthermore, the human IGHC locus is still evolving, with 5–20% of populations exhibiting duplication of single or multiple genes (46). The mouse IGHC locus, on the other hand, is composed of eight functional genes, including four Cγ genes but only one Cα gene and one 3′RR. CTCF-binding elements downstream of the 3′RR function as critical insulators for loop extrusion-mediated CSR activities in mouse B cells (47). Whether the two blocks of human IG genes are regulated by the two 3′RRs and associated CTCF-binding elements individually or cooperatively remains to be proved by further studies.

We also characterized several rare events in healthy donors, such as switching to IgD (0.14–0.86%), ψCε (0.03 − 0. 8%), or ψCγ (0.0 −0.05%). The sequence of the ψCγ gene suggested that there was no major structural defect in the gene itself, with only loss of the amino acid at position 15 in the CH3 exon and no change in the open reading frame (48). Thus, the lack of the S region is most likely the reason for the absence of corresponding protein expression. Our data suggest that “switchlike” repetitive sequences in 3′RR1 can be used to mediate recombination, potentially leading to the expression of ψCγ, which encodes a new IgG subclass, IgG5, albeit at a very low level. We also observed rare recombination events between the Sµ region and the 3′RR2 region (0.02–0.12%), which could be referred to as locus suicide recombination, resulting in the deletion of the entire IGHC gene cluster (39, 49).

In addition to studying the mechanistic details of the orientation control of CSR, we can also apply the LAM-HTGTS method in studies where Ig isotype specificity is relevant, such as following the immune response to infections. Following an acute infection, the CSR profile is expected to have some alterations. Indeed, we observed a slightly greater proportion of CSR junctions (46.5% versus 44.3% in PBMC adult control subjects), representing authentic CSR events in PBMCs from seven COVID-19 convalescent donors (17–122 d postinfection; unpublished data). Furthermore, we observed significantly greater IgA1 and IgA2 switching in COVID-19 convalescent donors than in noninfected, nonvaccinated healthy individuals (L. Du, H. Wan, H. Marcotte, and Q. Pan-Hammarström, unpublished observations), suggesting that the underlying mechanism could be of interest for follow-up. This method can also be used to study patients with mutations in transcription factors that are expected to influence the Ig subclass switching profile, such as APRIL deficiency (50) or T-bet deficiency (51). Furthermore, this method will be useful for the mechanistic study of extremely rare recombination events in normal human B cells, because these events might be more prevalent in selected human diseases, such as IgE switching in allergies and immunodeficiencies, IgD switching in hyper-IgD syndrome, and locus suicide switching in autoinflammatory/autoimmune disorders. Furthermore, LAM-HTGTS technology can be further optimized for profiling V(D)J recombination (28) and detecting translocation/insertion events in human cells.

The authors have no financial conflicts of interest.