Rad52 mediates class-switch DNA recombination to IgD

In B cells, IgD is expressed together with IgM through alternative splicing of primary VHDJH-Cμ-s-m-Cδ-s-m RNAs, and also through IgD class switch DNA recombination (CSR) via double-strand DNA breaks (DSB) and synapse of Sμ with σδ. How such DSBs are resolved is still unknown, despite our previous report showing that Rad52 effects the ‘short-range’ microhomology-mediated synapsis of intra-Sμ region DSBs. Here we find that induction of IgD CSR downregulates Zfp318, and promotes Rad52 phosphorylation and recruitment to Sμ and σδ, thereby leading to alternative end-joining (A-EJ)-mediated Sμ-σδ recombination with extensive microhomologies, VHDJH-Cδs transcription and sustained IgD secretion. Rad52 ablation in mouse Rad52−/− B cells aborts IgD CSR in vitro and in vivo and dampens the specific IgD antibody response to OVA. Rad52 knockdown in human B cells also abrogates IgD CSR. Finally, Rad52 phosphorylation is associated with high levels of IgD CSR and anti-nuclear IgD autoantibodies in patients with systemic lupus erythematosus and in lupus-prone mice. Our findings thus show that Rad52 mediates IgD CSR through microhomology-mediated A-EJ in concert with Zfp318 downregulation. IgD is expressed, predominantly together with IgM, via mRNA alternative splicing, but IgD class switch recombination (IgD CSR) has also been reported. Here the authors show, using Rad52-deficient mouse and human B cells, that IgD CSR is mediated by Rad52 through an alternative, microhomology-based end-joining pathway of DNA repair.

I gD has been an enigmatic antibody class for many years, despite being evolutionarily ancient and highly conserved across species [1][2][3][4][5][6] . As primordial as IgM, IgD appeared in cartilaginous fishes, amphibians and occurs in fishes, rodents, cattle, and humans 2,7 . As an example, in Xenopus, the Igδ exon cluster is in the same position, immediately 3′ of the Igμ locus, as it exists in mammals 7 . In mice and humans, IgD is expressed primarily as a transmembrane IgD receptor together with IgM with identical antigen specificity on naïve mature B cells in the form of BCR. IgD also exists as a secreted antibody. In humans, circulating IgD occurs at concentrations up to more than twothousand folds greater than IgE (10-250 μg/ml vs.~0.1 μg/ml), the rarest peripheral blood Ig class. IgD is secreted by IgM − IgD + plasmablasts and plasma cells differentiated from B cells in lymphoepithelial organs of aerodigestive mucosae, including palatine and pharyngeal tonsils. IgM − IgD + B cells and plasma cells can also be found in the lachrymal, salivary and mammary glands 3 . In addition to existing as a free molecule, IgD can occur on the surface of innate effector cells, including basophils, mast cells, and monocytes 1,8,9 . IgD bound to these cells would enhance immune surveillance and exert proinflammatory and antimicrobial effects 1,8,9 . These include triggering basophils to secret IL-4, IL-5, and IL-13 upon antigen engagement or attenuating basophil or mast cell allergic degranulation induced by IgE coengagement 1 . Thus, IgD would contribute to mucosal homeostasis by modulating effector cells reactivity to microbial commensals and pathogens 5,6 .
Identifying the stimuli and molecular mechanisms that underpin IgD expression is important to understand the regulation of IgD secretion throughout the body. The immediately proximal location of Cμ and Cδ gene loci as an integrated transcriptional unit allows these two Ig isotypes to be coordinately regulated 10,11 . In naive IgM + IgD + B cells, (membrane) mIgM and mIgD are co-expressed by alternative splicing of long primary transcripts consisting of rearranged V H DJ H exons and downstream Cμ and Cδ exons (V H DJ H -Cμ-s-m-Cδ-s-m). Alternative splicing of the same long primary V H DJ H -Cμ-s-m-Cδ-s-m transcripts also leads to expression of (secreted) sIgM and sIgD 2,8 . Transcription of long primary V H DJ H -Cμ-s-m-Cδ-s-m RNA requires the zinc finger ZFP318 repressor of transcriptional termination, which, as shown with genetically modified mouse models, obliterates the effect of the transcriptional termination sites (TTS) intercalated between the Cμ and Cδ exon clusters 10,11 (Fig. 1a). IgD can also be expressed through class-switch DNA recombination (CSR), by which IgM + IgD + B cells juxtapose V H DJ H DNA from the Cμ (IgM) to the Cδ (IgD) exons cluster, giving rise to V H DJ H -Cδm RNA transcripts and IgM − IgD + B cells 1,5,8,9,12 (Fig. 1b). In human and mouse nasopharyngeal and intestinal lymphoid tissues, a significant proportion of mucosal B cells class-switch to IgM − IgD + B cells, which subsequentially differentiate to plasmablasts and plasma cells 1,3,5,6 . Generally, CSR to IgD (Cδ) is a less frequent event than CSR to IgG (Cγ), IgA (Cα) or IgE (Cε), perhaps a reflection among other factors of the peculiar structure of the pseudo-switch σδ region lying immediately upstream of Cδ exons. Compared to the canonical Sμ, Sγ, Sα and Sε regions lying 5′ of the respective Igμ, Igγ, Igα, and Igε loci, σδ is shorter and contain differing motifs of nucleotide (nt) repeats 2,5,8,13,14 . These would provide an unconventional substrate for AID-mediated insertion of DSBs, possibly more prone to end-resection and generation of single-strand overhangs for Sμ-σδ recombination and expression of postrecombination V H DJ H -Cδ RNA transcripts 2,8,[13][14][15] .
Unlike CSR to IgG, IgA, and IgE, the mechanism of CSR to IgD remains unknown. Recombination involving Sμ DSB ends with DSB ends in downstream Sμ, Sγ, Sα, or Sε region is effected by non-homologous end-joining (NHEJ), one of the two major DNA DSB repair pathways, the other being homologous recombination (HR) 16,17 . HR accurately repairs resected (staggered) DSB ends using a sister chromatid as a homologous single-strand template during cell cycle S-G2. It critically effects error-free DSB repair in somatic cells and helps orchestrate chromosome segregation in meiosis. In contrast to HR, NHEJ is a homology-independent error-prone DSB repair process. It synapses blunt or virtually blunt DSB ends that lack substantial joining complementarity to form "direct" junctions, predominantly in G1 but also throughout the whole cell cycle 16 . NHEJ requires Ku70/Ku86 and in CSR mediates efficient long-range synapses of Sμ DSB ends with Sγ, Sα, and Sε DSB ends, leading to IgG, IgA, and IgE 15 . The finding, however, that reduction or deletion of Ku70/Ku86 led to reduced but still substantial CSR to IgG1 and IgG3 supported the existence of an alternative CSR end-joining (A-EJ) pathway [18][19][20] . This, like HR, would join resected DSB ends, thereby giving rise to S-S junctions involving microhomologies. Unlike HR, however, the A-EJ pathway juxtaposes DSB overhangs to be joined without using a homologous template as a guide. Rather, it utilizes differing extents of sequence complementarity (homology) between the upstream and downstream resected DSB overhangs to align the to-be DNA junctions 21 . As we have shown, HR factor Rad52 competes with NHEJ factors Ku70/Ku86 for binding to S region DSB ends and synapses DSB ends by A-EJ through microhomology-mediated end-joining (MMEJ) 20 , as inferred from increased NHEJ-mediated IgG, IgA, and IgE CSR events with even fewer S-S junction microhomologies in Rad52 −/− B cells in vivo and in vitro 20 . This together with the increased CSR to IgD in B cells lacking 53BP1, which protects S regions DSB ends from resection and facilitates long-range NHEJ to IgG, IgA, and IgE [22][23][24] , as well as other findings of ours showing reduced intra-Sμ region DSB short-range rejoining in Rad52 −/− B cells 20 led us to hypothesize that Rad52 mediates CSR to IgD through short-range Sμ-σδ DSB recombination by annealing to Sμ and σδ DSB single-strand resected ends.
By activating Rad52 −/− mouse B cells and RAD52 siRNA knockdown human B cells with different stimuli together with molecular genetic methods, here we show that Rad52 critically effects Sμ-σδ DNA recombination and CSR to IgD, through a microhomology-mediated (MM) A-EJ process which is associated with junctional Sμ-σδ somatic point-mutations. For this synaptic process, Rad52 is phosphorylated and recruited to Sμ and σδ regions, concomitant with downregulation of the TTS repressor Zfp318. Rad52-mediated Sμ-σδ CSR is required to mount an IgD-specific antibody response and is upregulated in systemic lupus erythematosus (SLE) patients and lupus mice, leading to high levels of total IgD and antinuclear antigen IgD autoantibodies. Thus, our experiments unveil a previously unrecognized and essential role of the Rad52 HR factor in a mammalian MM A-EJ DSB repair process which underpins a unique modality of CSR in health and disease.

Results
Definition of stimuli that induce Sμ-σδ CSR in mouse and human B cells. Toward testing our hypothesis that Rad52 mediates CSR to IgD, we first identified the stimuli that induce naïve IgM + IgD + B cells to undergo Sμ-σδ recombination. In naïve IgM + IgD + B cells, mIgM and mIgD are expressed by alternative splicing of long primary V H DJ H -Cμ-s-m-Cδ-s-m mRNAs-the Cδ locus is located immediately downstream of the Cμ locus in the same transcriptional unit, allowing these two loci to be coordinately regulated at the transcriptional level 1,2,4,6 ( Fig. 1a). As CSR can be induced in a T-dependent or T-independent fashion 15,25 , we used CD40 ligand CD154 (for mouse and human B cells), TLR4 ligand LPS (for mouse B cells), Fig. 1 Expression of cell surface and secreted IgD and IgM, as well as Iμ-Cδ transcripts by alternative splicing, alternative transcription termination and CSR. a Alternative splicing and alternative transcription termination underpin the expression of germline Iμ-Cμ and Iμ-Cδ transcripts, as well as membrane and secreted IgM and IgD in B cells. Expression of IgD stems from either Zfp318-dependent alternative mRNA splicing or Sμ-σδ CSR. In the presence of Zfp318, which represses the transcription termination sites (TTS) of the Cμ gene, mature B cells constitutively transcribe long primary V H DJ H -Cμ-Cδs-m transcripts initiated by the V H promoter. These long primary transcripts undergo alternative splicing which removes intronic regions, leading to dual expression of mature V H DJ H -Cμs and V H DJ H -Cδm transcripts encoding IgM and IgD. In the absence of Zfp318, transcription stops at Cμ TTS, resulting in a shorter primary transcript, which does not contain Cδ exons, and leads to expression of a mature V H DJ H -Cμ−s-m transcript only. Mature B cells also transcribe Iμ, Cμ, and Cδ regions under control of the Iμ promoter. When Zfp318 is expressed, unswitched mature B cells constitutively transcribe long primary Iμ-Cμ−s−m-Cδ-s−m transcripts, which undergo alternative splicing to removes intronic regions, leading to dual expression of germline Iμ-Cμ and Iμ-Cδ transcripts. In the absence of Zfp318, Iμ promoter-initiated transcription stops at Cμ TTS, and only germline Iμ-Cμ transcripts are expressed. b Expression of Iμ-Cδ transcripts, and membrane and secreted IgD by CSR. Schematic representation of CSR from IgM to IgD. The Sμ region recombines with the σδ region and loops out the intervening DNA, which forms a switch circle. The recombined DNA is transcribed leading to expression of V H DJ H -Cδ−s-m and Iμ-Cδ transcripts, initiated by the V H and Iμ promoters, respectively-in this case, Iμ-Cδ transcripts are generated as post-recombination transcripts. Graphics depict portion of the IgH locus and the resulting primary and mature transcripts. Inset depicts the detection of Sμ-σδ junctional DNA (CSR to IgD) by nested PCR amplification followed by Southern-blotting using specific Sμ and σδ probes (Southern-blotting of amplified recombined Sμ-σδ DNA from human naïve and germinal center B cells). The amplified Sμ-σδ DNA is sequenced for further analysis of the junctional sequence as well as identification and census of mutations. iEμ, IgH intronic enhancer; Iμ, intervening μ exon; μm, exon encoding the transmembrane region of IgM; δm, exon encoding the secretory piece of IgM; σδ, noncanonical switch-like region 5′ to Cδ; δs, exon encoding the secretory region of IgD; Cδm, exon encoding the transmembrane region of IgD. Dotted gray lines show splicing of primary transcripts to yield secreted and transmembrane forms of IgM and IgD. and TLR9 ligand CpG (for human B cells) in conjunction with different cytokines and/or BCR-cross-linking to induced CSR to IgD. Unlike IgG, IgA, and IgE, IgD can be expressed in the absence of CSR on surface of mature naive IgM + IgD + B cells, at high levels in a Zfp318-dependent fashion 10,11 . Accordingly, Aicda knockout B cells, in which CSR to IgD is abolished, can display robust surface IgD, at a level comparable to wild-type B cells 5 . Therefore, expression of IgD, as detected by surface/intracellular staining and/or flow cytometry, does not necessarily reflect Sμ-σδ recombination events, thereby making detection of recombined Sμ-σδ DNA the direct and positive proof of CSR to IgD.
Sμ-σδ junctions are enriched in microhomologies and abetted by somatic mutations in mouse and human B cells. The mechanisms effecting CSR can leave an S-S synaptic signature 20,26 . As we previously showed, Rad52 mediates A-EJ of resected DSB ends by juxtaposing overhangs with nucleotide complementarities, thereby giving rise to Sμ-Sx DNA junctions with microhomologies 20 . Next-generation sequencing of >100,000 recombined Sμ-Sx DNA junctions from mouse and human B cells in vitro and in vivo showed that Sμ-σδ junctions contained significantly more microhomologies (p < 0.01) than Sμ-Sγ1 or Sμ-Sα DNA junctions (representative frequencies and lengths of microhomologies in human and mouse B cells are depicted in Fig. 3a; representative human and mouse intra-σδ and junctional Sμ-Sx sequences are depicted in Supplementary Figs. 1-3), indicating that an MMEJ 21 process underpinned Sμ-σδ synapses. In both human and mouse B cells, the microhomologies in Sμ-σδ junctions were significantly more extensive than those in Sμ-Sγ1 and, to a lesser extent, Sμ-Sα junctions (Supplementary Figs. 1-4). As one example, in human tonsil B cells, 100% of analyzed Sμ-σδ junctions contained microhomologies, consisting of 2-13 nucleotides (mean = 6.30), while only 21% of Sμ-Sγ1 junctions contained microhomologies, consisting of 1-6 nucleotides (mean = 0.72) (Fig. 3a). Interestingly, there were a few common S-S sequences shared by recombined Sμ-σδ DNA junctions in human tonsil B cells and blood naïve B cells stimulated in vitro by CpG plus IL-2 and IL-21, suggesting that select Sμ and σδ DSB hotspots underpin Sμ-σδ DNA recombinations. A high frequency of microhomologies was also evident in the synaptic repair process of intra-σδ DSBs, evocative of what we showed in intra-Sμ DSBs 20 . Consistent with the greatest occurrence of microhomologies in Sμ-σδ junctions, Sμ is better suited for complementary DNA single-strand annealing with σδ than Sγ1 or, to a lesser extent, Sα (mouse) or Sα1 (human), based on various b Human and mouse Sμ and σδ regions consist of repetitive motifs, which are better-suited substrates for Rad52-mediated MMEJ than those in Sμ and Sγ1 or Sμ and Sα. As such, they can facilitate the formation of microhomologies. Repetitive sequence elements in mouse and human Sμ, σδ, Sγ1 and Sα that can potentially form microhomologies were identified by Pustell Matrix dot plot using MacVector software and are depicted by small dots. Intensity of dots depicts frequency and degree of complementarity of respective sequences. c Somatic point-mutations in Sμ and σδ regions abetting recombined Sμ−σδ DNA junctions in IgD class-switched human and mouse B cells in vivo and in vitro. Mutations were identified in a 48-506 nt stretch of Sμ or σδ regions in unique Sμ-σδ DNA recombination sequences. Each dot represents an individual sequence. Sequence data were pooled from three individuals in each group. Box and whiskers plots show median, quartiles, maximum and minimum of mutation frequencies in Sμ and σδ regions. In pie charts, the size of slices denotes the proportion of transcripts with the same number of mutations and the gray hue denotes the number of point mutations per transcript. Center of pie shows the total number of independent sequences analyzed. Below the pie charts is the overall mutation frequency (change/base). **p < 0.01, ***p < 0.001, ns: not significant (unpaired two-tailed t-test). Source data are provided as a Source Data file. numbers and contexts of these DNA regions discrete motifs, such as [G n ]AGCT repeats (Sμ, Sγ, and Sα) or AGCTGAGCTG repeats (Sμ and σδ), as revealed by Pustell Matrix dot-plot analysis (MacVector software) (Fig. 3b). During CSR, somatic mutations are introduced into the (upstream) donor Sμ and the (downstream) acceptor Sx regions, as effected by AID-targeting of these DNA regions. Indeed, Sμ-σδ DNA junctions were associated with somatic point-mutations in human and mouse B cells. These were more frequent in the σδ region than the Sμ area abetting the Sμ-σδ junction (e.g., 0.559 × 10 −2 vs. 0.973 × 10 −2 change/base in mouse spleen B cells in vivo and 1.251 × 10 −2 vs. 1.985 × 10 −2 change/base in mouse B cells stimulated by LPS plus IL-4 in vitro) (Fig. 3c). Thus, the high frequency of microhomologies in Sμ-σδ junctions supports a role of Rad52 in mediating CSR to IgD, while the somatic point-mutations in σδ reflects the targeting by AID.
Rad52 is critically required for Sμ-σδ synapsing. Having established that LPS or CD154 pus IL-4 induced CSR to IgD in mouse B cells, we used these stimuli and Rad52 −/− B cells together with appropriate controls (LPS alone, LPS plus TGF-β and RA, CD154 alone, or CD154 plus TGF-β and RA) and the same approach used in the experiments of Fig. 2 to investigate whether Rad52 was required for CSR to IgD. LPS plus IL-4 and CD154 plus IL-4 failed to induce Sμ-σδ recombination in Rad52 −/− B cells but not in Rad52 +/+ B cells, while either treatment efficiently induced Sμ-Sγ1 and Sμ-Sε recombinations in the same Rad52 −/− B cells (Fig. 4a)-in Rad52 −/− B cells, LPS, and LPS or CD154 plus TGF-β and RA induced CSR to IgG3 and IgA, respectively. As expected, CSR to IgD as well as IgG, IgA, and IgE was ablated in Aicda −/− B cells. Finally, the failure of Rad52 −/− B cells and Aicda −/− B cells to undergo CSR to IgD was associated with significantly decreased secretion of IgD (Fig. 4b). Thus, Rad52 is critical for Sμ-σδ DNA recombination and seemingly important for IgD secretion.
Rad52 is required to mount an antigen-specific IgD antibody response. To determine the role of Rad52 in mediating a specific IgD antibody response, we immunized Rad52 −/− and Rad52 +/+ mice with ovalbumin (OVA, 20 μg i.p., three times). Rad52 −/− mice showed no Sμ-σδ recombination in spleen, mesenteric lymph nodes (MLNs), or Peyer's patch B cells (Fig. 5a). The lack of CSR to IgD was specific, as B cells in such mice showed Sμ-Sγ1 and Sμ-Sα DNA recombinations like mouse Rad52 +/+ B cells, which also underwent Sμ-σδ recombination. In Rad52 −/− mice, B cells showed Sμ-Sγ1 and Sμ-Sα DNA junctions with fewer and shorter microhomologies than in Rad52 +/+ mice ( Fig. 5b and Supplementary Figs. 5,6), a reflection of the lack of Rad52 20 . Rad52 −/− mice also showed significantly decreased total and OVA-specific IgD in circulating blood, bronchoalveolar lavage (BALF), feces (free or bound to fecal bacteria), and IgDproducing cells in MLNs and lamina propria, as compared to their Rad52 +/+ mouse counterparts ( Fig. 5c-h and Supplementary Fig. 7a). This contrasted with the normal or elevated total and OVA-specific IgM, IgG1, and IgA levels in the same Rad52 −/− mice, as predicted based on our previous findings 27 . Thus, Rad52 is critical for mounting an antigen-specific classswitched IgD response.
Rad52 is modulated and phosphorylated by IgD CSR-inducing stimuli, and it is recruited to Sμ and σδ. We analyzed Rad52, Ku70, Ku86, and Aicda transcripts as well as the respective Rad52, Ku70, Ku86, and AID proteins, including phosphorylated Rad52 (p-Rad52 has been shown to display enhanced single-strand DNA annealing activity 28 ), in B cells induced to undergo CSR to IgD. Mouse naïve IgM + IgD + B cells stimulated by LPS plus IL-4 and human naïve IgM + IgD + B cells stimulated by CD154 plus IL-4 and IL-21 increased Ku70/Ku86 and Ku70/Ku86 expression at 24-48 h concomitant with significantly greater expression of Aicda and AID, which was virtually undetectable at time 0, while somewhat downregulating Rad52 and Rad52. Rad52 protein, however, was increasingly phosphorylated within the same time range (Fig. 6a-c). Further supporting its role in CSR to IgD, Rad52 was recruited to Sμ, σδ (and Sγ1) in B cells stimulated by LPS plus IL-4, which induced Sμ-σδ (and Sμ-Sγ1) DNA recombination, but not by stimuli that did not induce Sμ-σδ recombination, i.e., LPS alone or LPS plus TGF-β and RA, as shown by chromatin immunoprecipitation (ChIP) using an anti-Rad52 Ab-the specificity of the ChIP Rad52 recruitment assay being emphasized by the lack of chromatin immunoprecipitation in Rad52 −/− B cells (Fig. 6d, e). Recruitment of Rad52 but not Ku70/Ku86 to σδ in CSR to IgD, as induced by LPS plus IL-4, contrasted with that of Ku70/Ku86 to Sγ3 and Sα regions as induced in CSR to IgG3 and IgA (Fig. 6f), possibly a reflection of the competition of these HR and NHEJ elements for binding to S region DSB ends 20 . Notably, LPS plus IL-4 induced recruitment of Rad52 but not Ku70/Ku86 to σδ, while inducing mostly Ku70/ Ku86 recruitment to Cγ1, consistent with the efficient LPS pus IL-4 induction of CSR to IgG1, mediated mainly by NHEJ 20 . Thus, Rad52 modulation and, importantly, Rad52 phosphorylation are induced by IgD CSR-inducing stimuli to recruit Rad52 to σδ region DNA.
Stimuli that induce Sμ-σδ DNA recombination downregulate ZFP318/Zfp318 and lead to IgD secretion. Next, we addressed the expression of mIgD and sIgD and its regulation by stimuli inducing CSR to IgD. Resting naïve IgM + IgD + B cells expressed mIgD and mIgM, but virtually no sIgD or sIgM, reflecting high  Fig. 8b), an epigenetic activation mark, at the Zfp318 gene locus. Zfp318 downregulation was specific to IgD CSR, as it did not occur in response to IgA CSRinducing stimuli (LPS plus TGF-β and RA). ZFP318 downregulation concomitant with decreased mIgD expression and increased IgD secretion was reproduced in human B cells submitted to IgD CSR-inducing stimuli (CpG plus IL-2 and IL-21) but not IgD CSR non-inducing stimuli (CpG plus IL-4 and IL-21) (Fig. 7f). Similarly, ZFP318 transcripts and ZFP318 protein were downregulated in human B cells undergoing IgD CSR in vivo, as in peripheral blood and tonsils (Fig. 7g, h). Zfp318 downregulation was independent of and likely preceded expression of AID or Rad52, as revealed by virtual absence of Zfp318 transcripts in LPS plus IL-4-induced Aicda −/− B cells, Rad52 −/− B cells, and Rad52 +/+ B cells, all of which lost mIgD expression as compared to similar B cells stimulated by IgA CSR-inducing stimuli (LPS plus TGF-β and RA) ( Fig. 7i-k, Supplementary Fig. 7d). Thus, the stimuli that specifically induce CSR to IgD downregulate ZFP318/ Zfp318 independently of AID or Rad52 expression and prior to Sμ-σδ DNA recombination.
RAD52 knockdown reduces Sμ-σδ DNA recombination and IgD secretion in human B cells. The high frequency of microhomologies in Sμ-σδ junctions of human tonsil B cells in vivo and human B cells induced to undergo CSR to IgD in vitro ( Fig. 3a and Supplementary Figs. 1, 4) suggested to us that RAD52 also mediates Sμ-σδ DNA recombination in human B cells. We purified naïve IgM + IgD + B cells from peripheral blood of three healthy subjects and knocked down using RAD52-specific siRNAs RAD52 transcripts and RAD52 protein by up to 75% and 95%, respectively. In these B cells, Sμ-σδ recombination, as induced by CpG plus IL-2 and IL-21, was virtually abolished, while AICDA or AID expression and Sμ-Sγ1 recombination were not altered (Fig. 8a-c). The reduced Sμ-σδ DNA recombination in RAD52 knockdown human B cells was associated with decreased expression of V H DJ H −Cδs transcripts, without significant alteration of V H DJ H −Cδm transcripts (Fig. 8d). The critical role of RAD52 in human CSR to IgD was emphasized by RAD52 recruitment to Sμ and σδ regions in human naïve IgM + IgD + B cells induced to undergo CSR to IgD (by CpG plus IL-2 and IL-21) in vitro, human tonsil (IgD + ) B cells undergoing CSR to IgD in vivo, but not in unstimulated naïve IgD + IgM + B cells (Fig. 8e). Thus, Rad52 critically mediates CSR to IgD through Sμ-σδ recombination in human B cells.
Sμ-σδ DNA recombination leads to IgD plasma cell differentiation. To determine whether the substantial IgD secretion we observed upon induction of CSR to IgD (Figs. 4b, 5c-h, 7b, f) reflected increased plasma cell differentiation, we analyzed human naïve IgM + IgD + B cells induced to undergo CSR to IgD by CpG plus IL-2 and IL-21. More than 13% of these B cells became mIgM − and intracellular IgD + compared to 6.8% of their counterparts stimulated by CpG plus IL-4 and IL-21, not undergoing to CSR to IgD (Fig. 9a, Supplementary Fig. 9a). More than 90% of the IgM − IgD + B cells emerging from CpG plus IL-2 and IL-21 simulation expressed BLIMP-1 and almost 60% were CD27 + CD38 + versus about 10% of IgM − IgD + B cells from CpG plus IL-4 and IL-21 stimulation expressing BLIMP-1 and <12% being CD27 + CD38 + . Among mouse naïve IgM + IgD + B cells induced to undergo CSR to IgD by LPS plus IL-4, about 25% expressed intracellular IgD. All these B cells also expressed Blimp-1 and 70% or more acquired CD138 ( Fig. 9b and Supplementary  Fig. 9b). By contrast, among naïve IgM + IgD + B cells induced to undergo CSR to IgA (by LPS plus TGF-β and RA), about 50% expressed intracellular IgD and mIgD. Virtually none of these IgD + B cells, however, expressed Blimp-1 or acquired surface CD138 (Figs. 7a, 9b). The relevance of IgD CSR to plasma cell differentiation and sustained IgD secretion was furthered by analysis of three human myelomas, two IgD and one IgA. Both IgD myelomas displayed Sμ-σδ DNA, but not Sμ-Sα DNA recombination (Fig. 9c). Conversely, the IgA myeloma showed Sμ-Sα, but not Sμ-σδ DNA recombination. Thus, IgD + B cells emerging by CSR would be prone to differentiate into IgD-secreting plasmablasts/plasma cells for sustained IgD secretion. And such IgD + B cells may function as precursors of neoplastic IgD + transformants.
B cell Rad52 phosphorylation, increased CSR to IgD and IgD autoantibodies in systemic autoimmunity. Serum IgD have been suggested to increase in patients with inflammatory autoimmune diseases, such as SLE 29 and rheumatoid arthritis 30 , and in hereditary autoinflammatory syndromes, most notably the hyper-IgD syndrome (HIDS) [31][32][33][34] . In healthy humans, many circulating B cells make IgD that react with self-antigens including nuclear components, such as DNA 35 . We found SLE patients to display significantly higher levels of circulating IgD, including IgD specific for nuclear antigens, than their healthy subject controls (Fig. 10a, b). Similarly, we found lupus-prone MRL/Fas lpr/lpr mice to display higher levels of total IgD and/or IgD antinuclear autoantibodies than their wildtype C57BL/6 mouse counterparts in serum, feces and BALF as well as increased IgD-coated bacteria in feces (Fig. 10b-e). Overall, this was a Rad52 deletion ablates in vivo Sμ-σδ DNA recombination and reduces IgD production. Rad52 +/+ and Rad52 −/− mice were immunized with OVA in alum i.p. a Recombined Sμ-σδ, Sμ−Sγ1, and Sμ−Sα DNAs in spleen, mesenteric lymph nodes (MLNs), and Peyer's patches (PPs) B cells, as analyzed by nested PCR using forward Iμ and reverse Cδ, Sγ1, or Sα primers, respectively, followed by Southern-blotting using specific Sμ, σδ, Sγ1, or Sα probe, as indicated. Data are one representative of three independent experiments yielding comparable results. b Sμ-σδ, Sμ-Sγ1, and Sμ-Sα junctional DNAs were amplified by nested PCR and sequenced by MiSeq. The length and numbers of nucleotide overlaps (microhomologies) in Sμ-σδ, Sμ−Sγ1, and Sμ−Sα junctional DNAs are shown by violin plots. Each symbol represents a unique sequence (n = 45 per group). c-f Titers of total IgD in serum, BALF, and feces, as analyzed by dot-blotting using rat anti-mouse IgD mAb-titers of total IgM, IgD, IgG1, and IgA as well as OVA-binding IgM, IgD, IgG1, and IgA as analyzed by specific ELISAs. Each dot represents datum from one individual mouse (n = 5-8 per group, as indicated). Data are mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ns: not significant (unpaired two-tailed t-test). No adjustments were made for multiple comparisons.  (Fig. 10f, g and Supplementary Fig. 9c). In MRL/Fas lpr/lpr mice, the elevated IgD levels reflected the increased numbers of IgD-producing cells and increased B cell Sμ-σδ DNA recombination in bone marrow, spleen, MLNs and Peyer's patches (Fig. 10h). Increased CSR to IgD in MRL/Fas lpr/lpr mice was characterized by greater frequency and length of microhomologies in Sμ-σδ as compared to Sμ-Sγ1 and Sμ-Sα junctional sequences, as well as a high frequency of somatic pointmutations in areas abetting Sμ-Sδ DNA junctions (Fig. 10i, j and  Supplementary Fig. 10). In both SLE patients and lupus MRL/ Fas lpr/lpr mice, the high level of CSR to IgD, total IgD and IgD nuclear autoantibodies likely stemmed from the higher level of B cell Rad52 and/or p-Rad52 (Fig. 10k). Further, consistent with the high levels of Rad52 and/or p-Rad52 expression and CSR to IgD, expression of ZFP318/Zfp318 was decreased in B cells of such SLE patients and lupus MRL/Fas lpr/lpr mice (Fig. 10l). Thus, B cells expressing high levels of phosphorylated Rad52 would underpin the high levels of IgD and IgD autoantibodies to nuclear antigens in SLE patients and in lupus MRL/Fas lpr/lpr mice. As shown in these mice, Sμ-σδ DNA recombination events involving high frequency of junctional microhomologies occur in circulating B cells and B cells in different body districts, including the respiratory tract, gut lymphoid formations, spleen and splanchnic draining lymph nodes, giving rise to high levels of IgD autoantibodies locally and systemically.

Discussion
The mechanism of CSR to IgG, IgA, and IgE is quite well understood, as mediated by Ku70/Ku86-dependent NHEJ, although occurrence of a "residual" IgM to IgG CSR in B cells lacking Ku70/Ku86 expression has suggested the existence of a Ku70/Ku86-independent A-EJ synaptic mechanism [18][19][20] . Mice lacking 53BP1, in which NHEJ-dependent CSR to IgG, IgA and IgE was significantly decreased-53BP1 protects resection of DSB ends, thereby skewing the synaptic process toward NHEJshowed increased CSR to IgD and increased circulating IgD levels 22,23 , suggesting that the short-range Sμ-σδ CSR is mediated by a 53BP1-independent synaptic process involving resected DSB ends and entailing a high frequency of Sμ-σδ junctional microhomologies. This together with our previous demonstration that Rad52 plays a central role in synapsing intra-Sμ region resected DSB ends as well as c-Myc/IgH locus translocations also involving resected DSB ends-both processes entailing significant junctional microhomologies-prompted us to hypothesize that Rad52 synapses Sμ and σδ DSB involving complementary overhangs for CSR to IgD 20 . Indeed, here we showed that Rad52 effects CSR to IgD through a MM A-EJ mechanism in concert with Zfp318 modulation of primary V H DJ H −Cμ-s-m-Cδ-s-m transcription in mouse and human B cells ( Supplementary  Fig. 11), thereby unveiling a previously unknown and critical role of the HR factor Rad52 in a specific mammalian DNA repair process.
We have provided here unequivocal evidence that Rad52 is critical for CSR to IgD in vitro and in vivo, in mouse and human B cells. Sμ-σδ DNA recombination was ablated in mouse Ku70/KU70, Ku86/KU86, and Rad52/RAD52 transcripts while significantly upregulating Aicda/AICDA in both mouse and human B cells. This was concomitant with induction of AID and moderate decrease in Rad52 protein, which, in fact, was increasingly phosphorylated at Tyr104. Rad52 Tyr104 phosphorylation boosts Rad52-mediated DNA single-strand annealing and is possibly effected by c-ABL kinase 28 . In human and mouse B cells, Rad52 phosphorylation was promoted by IgD CSRinducing stimuli. Rad52 involvement in CSR to IgD was further emphasized by recruitment of this protein to the σδ region (in addition and necessarily to Sμ) in vivo in human tonsil IgD + B cells, as well as in vitro, in mouse and human naïve B cells induced to undergo CSR to IgD, but no or only marginally in similar B cells undergoing CSR to IgG3 or IgA. Instead, these B cells recruited Ku70/Ku86 to Sγ3 and Sα, consistent with a dominant contribution of NHEJ to CSR to IgG3 and IgA.
Rad52 is a member of the eponymous epistasis group for DSB repair that shows strong evolutionary conservation 17,36 . In Saccharomyces cerevisiae, Rad52 is a key element of the HR pathway, and its deletion or mutation impairs DNA DSB repair 37,38 . Indeed, yeast Rad52 is a DSB recombination mediator and a facilitator of annealing of complementary DNA singlestrands 39,40 . As a cofactor of Rad51, it can form nucleoprotein filaments with single-strand DNA and promotes strand pairing by overcoming the inhibitory effect of replication protein A (RPA) 20,41 . Rad52 mutation or even deletion results in no obvious abnormalities in viability or functions in mammalian cells. As we have shown, Rad52 −/− mice displayed no obvious alteration of immune system elements, including B cells 20 , possibly owing to the presence of mammalian gene paralogues, such as BRCA2 and RAD51, which by encoding functions related to Rad52, can compensate for the absence of this HR factor 42 .
Human BRCA2 functions as a recombination mediator by facilitating RAD51 nucleoprotein filament formation 40,43-45 . Nevertheless, human BRCA2 cannot facilitate annealing of RPA-coated DNA, a function that Rad52 carries out efficiently in the absence of BRCA2 46 . This together with Rad52 involvement in DSB repair at stalled or collapsed replication forks points at a unique role of this HR factor in catalyzing single-strand annealing in homologydirected DNA repair in human cells [47][48][49] . Indeed, in these cells, Rad52 would mediate synapses of intra-V gene segment resected DSBs during SHM 50 .
Our identification of Rad52 as essential in the IgD CSR Sμ-σδ synaptic process provides, to the best of our knowledge, the first demonstration of a critical and dedicated role of this factor in mammalian DNA DSB repair. The short-range Rad52-mediated Sμ-σδ recombination of resected DSB ends adds to the other Rad52-mediated short-range DSB recombination we recently uncovered: intra-Sμ region DSB recombination 20 . This, like Sμ −σδ synapsing, entails insertion of significant junctional nucleotide microhomologies 20 . In this function, as in CSR to IgD, Rad52 is not fungible in mouse or human B cells. The identification of Rad52 as the critical element in Sμ-σδ synapsis sheds light on the mechanistic nature of the CSR microhomologymediated A-EJ DSB repair pathway, originally referred to as A-NHEJ 19 . As per our current findings, the CSR A-EJ pathway uses HR Rad52 to synapse upstream and downstream DSB overhangs through MMEJ, hence our definition of this process as MM A-EJ. This would not require a homologous template as a guide, as the HR pathway does. Although DNA polymerase θ has been suggested to contribute to A-EJ 21 , our previous findings do not support a role of this polymerase in Rad52-mediated intra-Sμ DSB recombination or Sμ−Sγ1, Sμ−Sγ3, Sμ−Sγ2a/Sγ2c, and Sμ −Sα recombination (CSR to IgG1, IgG3, IgG2a/IgG2c, and  Recombined Sμ-σδ and Sμ-Sγ1 DNA in the transfected B cells 120 h after RAD52 siRNA transfection, as well as Sμ-σδ DNA in tonsil IgM -IgD + and blood naive IgM + IgD + B cells were analyzed by nested PCR using forward Iμ and reverse Cδ or Sγ1 primers followed by Southern-blotting using indicated specific probes. Data are from three independent experiments. b Expression of RAD52 and AICDA transcripts was analyzed 48 h after RAD52 siRNA or Scra siRNA transfection by qRT-PCR and normalized to HPRT expression. Data are mean ± SEM of three independent experiments. **p < 0.01, ns: not significant (unpaired two-tailed t-test). c Expression of RAD52 and AID proteins were analyzed 72 h after RAD52 siRNA or Scra siRNA transfection by specific Western blotting. Data are representative of three independent experiments. d Expression of V H DJ H −Cδm and V H DJ H −Cδs transcripts as analyzed by semi-quantitative RT-PCR with of serial two-fold dilution of cDNA templates. Data are representative of three independent experiments. e RAD52 is recruited to σδ region DNA in human B cells undergoing CSR to IgD. Recruitment of RAD52 to Sμ and σδ region DNA in human blood naive IgM + IgD + B cells stimulated for 120 h with CpG plus IL-2 and IL-21, as analyzed by specific ChIP-qPCR. Data are mean ± SEM of three or four independent experiments. *p < 0.05, **p < 0.01 (unpaired two-tailed t-test). Source data are provided as a Source Data file.   -tailed t-test). b Human and mouse antinuclear autoantibodies (ANAs), as visualized by indirect immunofluorescence microscopy on HEp-2 cells that were incubated with serum from a healthy human subject, an SLE patient, a C57BL/6 mouse, or a MRL/Fas lp/lpr mouse, as revealed by FITC-labeled rat mAb to human or mouse IgD. Scale bar = 50 μm. c Total IgD in serum, feces and BALF as analyzed by dot-blotting, and concentrations of IgD autoantibodies to dsDNA or histone in serum of C57BL/6 and MRL/Fas lpr/lpr mice, as analyzed by specific ELISAs. Each dot represents datum from one individual mouse. Data are mean ± SEM of 3-9 mice, as indicated. **p < 0.01, ***p < 0.001 (unpaired two-tailed t-test). d IgD concentrations in serum, feces and BALF from C57BL/6 and MRL/Fas lpr/lpr mice, as analyzed by dot-blots. Shown are dot-blots from one C57BL/6 and one MRL/Fas lpr/lpr mouse, representative of 3-9 C57BL/6 and MRL/Fas lpr/lpr mice. e Bacteria-bound IgD and IgA in feces from C57BL/6 and MRL/Fas lpr/lpr mice, as analyzed by flow cytometry. f Expression of V H DJ H −Cδm, V H DJ H −Cδs, V H DJ H −Cμm, and V H DJ H −Cμs transcripts in bone marrow (BM), spleen and mesenteric lymph nodes (MLNs), as analyzed semi-quantitative RT-PCR by serial two-fold dilutions of cDNA templates. Shown are RT-PCR data from one C57BL/6 mouse and one MRL/Fas lpr/lpr mouse, representative of three C57BL/6 and three MRL/Fas lpr/lpr mice. g IgD + B cells in lamina propria, MLNs and Peyer's patches (PPs) of C57BL/6 and MRL/Fas lpr/lpr mice, as visualized by fluorescent microscopy. Scale bar = 100 μm. h Recombined junctional Sμ-σδ, Sμ-Sγ1, and Sμ-Sα DNAs in bone marrow, spleen, MLNs, and PPs B cells from C57BL/6 and MRL/Fas lpr/lpr mice as analyzed by specific nested PCR using forward Iμ and reverse Cδ, Sγ1 or Sα primers, followed by Southern-blotting using indicated probes. Data are representative of three independent experiments. i Sμ-σδ, Sμ-Sγ1 and Sμ-Sα junctional DNAs in non-immunized MRL/Fas lpr/lpr mice, as amplified by nested PCR and sequenced by MiSeq. The length and numbers of nucleotide overlaps (microhomologies) in Sμ-σδ, Sμ-Sγ1, and Sμ-Sα junctional DNAs are rendered by violin plots. Each dot represents a unique sequence (n = 45 per group). j Somatic point-mutations in Sμ and σδ regions abetting recombined Sμ−σδ DNA junctions in IgD class-switched spleen B cells from three MRL/ Fas lpr/lpr mice. Mutations were identified in a 48-506 nt stretch of Sμ or σδ regions in unique Sμ-σδ DNA recombination sequences. Each dot represents an individual sequence. ns: not significant (unpaired two-tailed t-test). Box and whiskers plots show the median, quartiles, maximum, and minimum of mutation frequencies in Sμ and σδ regions. In pie charts, the size of slices denotes the proportion of sequences with the same number of mutations and the gray hue denotes the number of point-mutations per sequence. Center of pie shows the total number of independent sequences analyzed. Below the pie charts is the overall mutation frequency (change/base). k Expression of phosphorylated Rad52 (p-Rad52), Rad52 and β-Actin proteins in peripheral blood B cells from healthy human subjects and SLE patients as well as B cells from C57BL/6 mice and MRL/Fas lpr/lpr mice, as analyzed by specific Western blotting using rabbit anti-p-Rad52 Ab or anti-β-Actin mAb-p-Rd52 (Y104) Ab detected endogenous levels of Rad52 protein only when phosphorylated at tyrosine 104. l Expression of ZFP318 and AICDA transcripts in B cells from healthy human subjects and SLE patients (left panel), as well as Zfp318 and Aicda transcripts in MLNs from C57BL/6 and MRL/Fas lpr/lpr mice (right panel), as analyzed by specific qRT-PCR. Data are mean ± SEM of three healthy human subjects, three SLE patients, three C57BL/6 mice, and three MRL/Fas lpr/lpr mice. *p < 0.05 (unpaired two-tailed t-test). No adjustments were made for multiple comparisons. Source data are provided as a Source Data file. IgA) 20 . While Rad52 MM A-EJ functions as a backup pathway for CSR in B cells defective in NHEJ, it also synapses DSB ends in presence of a full complement of NHEJ elements 51 , as exemplified by microhomologies in S-S junctions in a minor proportion of B cells that class-switched to IgG, IgA and IgE, and the disappearance of such microhomologies upon Rad52 ablation 20 .
As we showed here, Rad52 works in concert with Zfp318 to modulate IgD expression through an interplay of alternative RNA splicing and DNA recombination, the latter after AID intervention. Zfp318 represses the TTS intercalated between Cμ and Cδ exons within the Ighμ/Ighδ loci transcriptional complex unit 10,11 , thereby allowing for expression of long primary V H DJ H -Cμ-s-m-Cδ-s-m RNA. Zfp318, however, also simultaneously allows for the continuous expression of primary Iμ-Cμ-s-m-Cδ-s-m RNA transcripts. In fact, albeit possibly more abundant, hence their predominant detection in our specific PCR assays, Iμ-Cδ transcripts (in secretory or membrane form) are identical in sequence to their postrecombination Iμ-Cδ counterparts (Fig. 1a, b). During B cell development, Zfp318 expression closely parallels mIgD expression 10,11 . And, consistent with its repression of the TTS intercalated between the Cμ and Cδ exons complex, the Zfp318 protein is expressed during the transition from immature IgM + IgD − to mature IgM + IgD hi B cell 10,11 . As we showed here, naïve mature B cells which express high levels of mIgD also express high levels of Zfp318 transcripts and Zfp318 protein. In these B cells, stimuli that induced Sμ-σδ DNA recombination, yielding primary V H DJ H -Cδ-s-m RNA transcripts, induced profound downregulation of Zfp318 transcripts and Zfp318 protein, suggesting that relieving Zfp318-mediated TTS repression is a prerequisite for Sμ-σδ DNA recombination to unfold. Conversely, as we also showed here, naïve mature IgM + IgD + B cells submitted to stimuli that induced CSR to isotypes other than IgD, such as IgA (by LPS plus IL-4 and RA), further upregulated Zfp318 transcripts and Zfp318 protein, concomitant with no Sμ-σδ recombination, but rather allowing for massive expression of mIgD rather than sIgD.
The role of Zfp318 as gene transcription regulator is highly specific for IgD, as genome-wide transcriptome analysis of B cell Zfp318-deficient (Vav-Cre dependent deletion) mice identified Sva as the only other gene altered in expression 11interestingly, Sva is also involved in alternative splicing, albeit outside the IgH locus 52 . Zfp318 would be under the control of 5'AMP-activated protein kinase (Ampk). This is phosphorylated by Lbk1 53 , whose signaling triggers the B cell GC reaction. Indeed, Lbk1's failure to activate Ampk or Ampk loss specifically muted Zfp318 expression and IgD transcription 53 . In contrast, activation of Ampk by phenformin impaired GC formation 53 , likely by heightening Zfp318 expression and possibly in addition to other mechanisms. This would result in increased expression of primary V H DJ H −Cμs-m-Cδ-s-m RNA transcripts but not Sμ-σδ DNA recombination, suggesting that CSR to IgD is one of the multiple and complex events inherent to GC formation. This is generally triggered by naturally occurring, generally, microbial stimuli, as in tonsil GCs and GCs or other secondary lymphoid formations in aerodigestive mucosae 1,5,6 . Consistent with the contention that Ampk mediates the regulation of Zfp318 as well as the contrasting impact of IgD CSR-inducing (LPS plus IL-4) and non-inducing stimuli (LPS plus TGF-β and RA) on expression of Zfp318, stimulation of both human and mouse cells by LPS has been shown to result in dephosphorylation/inactivation of Ampk, while similar cell stimulation by TGF-β would result in rapid phosphorylation/activation of this protein kinase 54 .
In our experiments, stimuli that induced CSR to IgD (e.g., LPS plus IL-4 in mouse B cells, and CD154 plus IL-2 and IL-21 in human B cells) also downregulated Zfp318 expression which, in turn, reduced V H DJ H −Cδm transcript level and mIgD, while greatly increasing V H DJ H −Cδs transcripts and sIgD. This argues for CSR to IgD to be critical for significant IgD secretion. Indeed, stimuli that induced Sμ-σδ recombination and IgD secretion also induced plasmablast/plasma cell differentiation, as shown by Blimp-1, CD38, and CD27 expression in human B cells, and Blimp-1 and CD138 expression in mouse B cells. A similar outcome was not produced by stimuli that did not induce Sμ-σδ recombination and IgD secretion in both mouse and human B cells. Thus, while alternative splicing of long primary V H DJ H −Cμs-m-Cδ-s-m RNA transcripts in B cells that have not undergo CSR would make some contribution to the overall level of IgD production in vivo, CSR to IgD would be required for substantial and sustained IgD production, as secreted by plasmablasts/plasma cells or by neoplastic transformants, such as IgD myeloma cells. The limited IgD amounts detected in supernatants of mouse or human B cells primed by stimuli that induced high levels of mIgD but not Sμ-σδ synaptic recombination would result from translation of alternative spliced long primary V H DJ H −Cμ-s-m-Cδ-s-m RNA transcripts as well as some "shedding" of mIgD.
Bacteria and viruses have been suggested to play an important role in driving CSR to IgD, generally through stimulation of TLRs in gut and respiratory lymphoid tissues, and mesenteric lymph nodes, possibly leading to emergence of plasmablasts and plasma cells secreting IgD 1,3-5,52,55-57 . Circulating IgD are increased in HIV patients with frequent respiratory infections, and in mice undergoing acute viral infections, suggesting a role for IgD in airways mucosae antimicrobial protection 1,3,5,55,57,58 . In the in vivo T-dependent antibody response to OVA, ablation of Rad52 (Rad52 −/− mice) resulted in reduced levels of total and specific IgD not only in circulating blood but also in total and/or bacteria-bound IgD in feces as well as decreased numbers of IgD + B cells in lamina propria and MLNs, a privileged site of IgD CSR 12 . In these mice, the marginal level of IgD in BALF suggests that in the lack of CSR, the compensatory expression of IgD by alternative splicing contributes to overall IgD levels systemically and in the gut but only minimally in respiratory mucosae, where the level of IgD is would result almost exclusively from CSR. As predicted by our previous findings 20 , the decreased IgD levels in Rad52 −/− mice were associated with increased IgG1 and IgA as well as greatly reduced microhomologies in Sμ-Sγ1 and Sμ-Sα junctions. This reflected the lack of Rad52 contribution to the synaptic process underpinning such junctions as well as the lack of Rad52 competition with Ku70/Ku86 for binding to DSB free ends 20 , resulting solely in Ku70/Ku86-mediated NHEJ, a process that limits microhomologies to 0-3 nt 16 .
Information on the contribution of IgD to autoimmune responses is scant and contradictory. Self-antigen-binding and mostly polyreactive IgD occur in healthy subjects, much like IgM or even IgG and IgA do 35,[59][60][61][62] . Interestingly, in healthy humans, circulating mature B cells class-switched to IgD have been found to make IgD reactive with self-antigens, including nuclear antigens, such as single-strand DNA and double-strand DNA 35 . High levels of IgD have been reported in rheumatoid arthritis patients and thought to possibly be implicated in the pathogenesis of the disease 30 . IgD expression, however, has been speculated to exert an inhibitory effect on B cell autoreactivity, as suggested by elevated autoantibody production, increased deposition of immune complexes in kidneys, and severe nephritis in lupus-prone C56BL/6lpr mice with deletion of the Igδ locus 63,64 . Our findings showed total and self-reactive IgD (dsDNA, histone, RNP/ Sm or RNA and ANAs) to be elevated in the circulation of SLE patients and lupus-prone MRL/Fas lpr/lpr mice. The latter displayed higher levels of IgD in serum, BALF, and feces, than their wildtype C57BL/6 counterparts. Such high IgD levels reflected CSR recombinations that included Sμ-σδ junctions with extensive microhomologies and high frequency of somatic mutations in the DNA areas abetting Sμ-σδ junctions. IgD CSR occurred in different districts, such as bone marrow, spleen, MLNs, and Peyer's patches, and was reflected in IgD + B cells in those districts. This together with high B cell levels of phosphorylated Rad52 and low levels Zfp318 indicated that in murine and human lupus, IgD autoantibodies stem from extensive B cell Sμ-σδ recombination, as also supported by high levels of AID, rather than Zfp318-dependent alternative splicing of primary V H DJ H −μ-s-m-Cδ-s-m RNA transcripts. Our findings do not suggest a "protective" role of IgD in autoimmunity 63,64 , while supporting a role of CSR to IgD in systemic lupus autoantibody responses.
Collectively, our data outline a critical and dedicated role of Rad52 in mediating the synapsis of Sμ with σδ DSB resected ends. They also provide the first demonstration of Rad52 as essential element of the poorly understood A-EJ process underpinning resolution of resected end DSBs in B cell IgH locus. In malignant B cells, Rad52 is involved in DNA recombination events that give rise to DNA deletions and translocations. As we previously showed, Rad52 ablation reduced the frequency of c-Myc/IgH translocations in mouse p53 −/− B cells by >70%, with the residual translocations containing limited microhomologies 20 . Whether Rad52 intervention extends to other modalities of A-EJ in neoplastic and non-neoplastic lymphoid mammalian cells remains to be determined. The importance of the Rad52 function newly unveiled here is further emphasized by our demonstration that this highly conserved HR element is critical for CSR to IgD in not only mouse but also human B cells. This together with the further reduction of the physiologically moderate microhomologies in Sμ-Sγ1, Sμ-Sγ3, Sμ-Sα, and Sμ-Sε junctions in Rad52 −/− B cells (current data and refs. [18][19][20] solidifies the role of Rad52 as critical mediator of the A-EJ backup pathway underpinning the residual CSR to IgG, IgA and IgE in the absence of Ku70/Ku86 proteins 20 . Our findings also showed how stimuli that induce Sμ-σδ recombination coordinate Rad52 function, as enabled by phosphorylation, with downregulation of Zfp318, unique repressor of the TTS intercalated between the Cμ and Cδ loci, whose activity allows transcription throughout V H DJ H −Sμ-Cμ-s-m-σδ-Cδ-s-m and Iμ-Cμ-s-m-σδ-Cδ-s-m. Further, they indicate that CSR to IgD is required for sustained IgD secretion and, possibly, a prerequisite for IgD plasma cell differentiation. They also add new and significant information to a potential role of CSR to IgD, as promoted by Rad52 phosphorylation, in systemic autoimmunity. Finally, they would provide important new molecular information to approach the virtually unexplored mechanistic underpinning of hyper-IgD syndrome, a relatively rare but a severe autoinflammatory disease associated with mevalonate kinase deficiency (due to MVK recessive mutations) and exorbitant levels of IgD 31,32,65 .

Methods
Mice. Rad52 −/− mice were generated by replacing exon 3 of the Rad52 gene with positive selection marker neomycin, as driven by the phosphoglycerate kinase (PGK) promoter, and an upstream mouse sequence functioning as a transcription terminator (Dr. Albert Pastink, Leiden University, Leiden, The Netherlands) 36 . Rad52 −/− mice were backcrossed to C57/BL6 mice for more than six generations. No full length or truncated Rad52 protein was produced from the disrupted allele 36 . Rad52 −/− mice were viable and fertile, and showed no gross abnormalities. Aicda −/− mice (C57BL/6 background) 66 were obtained from Dr. Tasuku Honjo (Kyoto University, Kyoto, Japan). C57BL/6 mice (Stock # 000664) and MRL/Fas lpr/lpr mice (Stock # 000485) were purchased from Jackson Laboratory (Bar Harbor, Maine). For OVA immunization, Rad52 +/+ and Rad52 −/− mice were injected intraperitoneally with 20 μg of OVA in 100 μl alum (Imject Alum, Pierce) i.p., three times, at day 0, 7 and 14. One week after the last injection, the mice were euthanized using a euthanasia CO 2 chamber (filled at a rate of 20-30% CO 2 chamber volume per minute, which has been shown to cause the least amount of distress to rodents) for ex vivo analysis. All mice were housed in the Specific Pathogen-Free/SPF facility of the University of Texas Health Science Center at San Antonio. Both male and female mice aged 8-12 weeks were used for the experiments. The Institutional Animal Care and Use Committees (IACUC) of the University of Texas Health Science Center at San Antonio approved all animal protocols.
Fluorescence microscopy. To analyze IgM and IgD-producing cells in the lamina propria and PPs, the intestine was folded into a "Swiss-roll", fixed with PFA (4%) and embedded in paraffin. Ten-micrometer sections were cut and heated at 80°C to adhere to the slide, washed four times in xylene for 2 min, dehydrated two times with 100% ethanol for 1 min, two times with 95% ethanol for 1 min, and washed two times in water for 1 min. Antigens were unmasked using 2 mM EDTA in 100°C for 40 mins followed by a cooling step at 25°C on the bench top, three times washing with 1x TBS and blocking using 10% BSA for 15 min. Slides were again washed three times with 1x TBS and stained with FITC-anti-IgD mAb (clone 11-26 c.2a, 405713, BioLegend), PE anti-mouse IgM mAb (clone II/41, 12-5790-82, Invitrogen), and FITC-anti-mouse IgA mAb (clone mA-6E1, 11-4204, Invitrogen) for 2 h in a dark moist chamber. After washing three times with Triton X-100 (0.1%) in TBS, slides were air-dried, and coverslips were mounted with ProLong ® Gold Antifade Reagent with DAPI (Invitrogen). Fluorescence images were captured using a 10x objective lens with a Zeiss Axio Imager Z1 fluorescence microscope. To analyze IgD-producing cells in MLNs, 10 μm MLN sections were prepared by cryostat and loaded onto positively charged slides, fixed in cold acetone, and stained with FITC-anti-IgD mAb (clone 11-26 c.2a, 405704, BioLegend), or PE anti-mouse IgM mAb (clone RMM-1, 406507, BioLegend), respectively, for 1 h at 25°C in a moist chamber. Coverslips were then mounted using ProLong ® Gold Antifade Reagent using DAPI (Thermo Fisher), before examination with a fluorescence microscope.
Mouse and human antibodies and autoantibodies. Titers of mouse serum, BALF or fecal total IgM, IgG1 and IgA and OVA-binding IgD, IgM, IgG1 and IgA, mouse serum dsDNA-or histone-binding IgD and total IgD in in vitro culture supernatants of stimulated human B cells, as well as human serum total and dsDNA-, RNA-, histone-or RNP/Sm-binding IgD were measured using specific ELISAs, as we described 20,25,68,69 . Total IgD in in vitro culture supernatants of stimulated mouse B cells or in mouse serum, BALF or feces were measured by dot-blotting with serially two-fold diluted samples. Mouse and human antinuclear autoantibodies (ANAs) were detected using Hep-2 cells and specific FITC-anti-mouse IgD (clone 11-26 c.2a, 405704, BioLegend) and FITC-anti-human IgD (clone IA6-2, 348205, BioLegend).
Bacteria-bound IgD and IgA were detected in feces by flow cytometry, as we described 27 . Briefly, feces (10 mg) were suspended in 100 μl 1x PBS (filtered through 0.2 μm filter), homogenized and centrifuged at 400 × g for 5 min to remove large particles. The supernatant was then centrifuged at 8000 g for 10 min to remove non-bound antibodies. The bacterial pellet was suspended in 1 ml of PBS with 1% (w/v) BSA. After fixation with 7.2% formaldehyde for 10 min at room temperature, bacteria were washed with PBS, and stained with PE-anti-IgD mAb (clone 11-26 c.2a; 405705, BioLegend) or PE-anti-mouse IgA mAb (clone mA-6E1, 12-4204-82, Invitrogen) on ice for 30 min, washed with PBS, and further resuspended in 1 × PBS containing 0.2 μg ml −1 DAPI for flow cytometry analysis. All events that stained with DAPI were considered as bacteria.
S-S region DNA recombinations and S region somatic mutations. Genomic DNA was prepared from human or mouse B cells using QIAmp DNA Mini Kit (Qiagen) or from paraffin-embedded human IgD or IgA myeloma tissue sections (obtained from the University of Arkansas for Medical Science) using Quick-DNA™ FFPE Kit (Zymo Research). Recombined Sμ-σδ, Sμ-Sγ1, Sμ-Sα, and Sμ-Sε DNA were amplified by specific nested PCR involving two sequential rounds of PCR using Phusion™ high-fidelity DNA polymerase (Thermo Scientific™) and two sets of nested specific primers (forward Iμ and reverse Cδ or Sx primers) (Custom synthesized by Eurofins Genomics, Supplementary Table 1). The first and second rounds of PCR were performed at 98°C for 30 sec, 58°C for 45 sec, 72°C for 4 min (30 cycles). Amplified genomic DNA was fractionated through 1.0% agarose, blotted onto Hybond-N + membranes (GE Healthcare) and hybridized to biotinlabeled internal Sμ and σδ, Sγ1, Sγ3, Sα, or Sε specific probes 70 . Detection was performed using the Chemiluminescent Nucleic Acid Detection Module (Thermo Fisher Scientific) according to the manufacturer's instructions. For sequence analysis of the recombined DNA, PCR products were purified using a QIAquick PCR purification kit (Qiagen). The amplified library was tagged with barcodes for sample multiplexing, and PCR was enriched and annealed to the required Illumina clustering adapters. High-throughput 300-base pair (bp) paired-end sequencing was performed by the University of Texas Health Science Center San Antonio Genome Sequencing Facility using the Illumina MiSeq platform. S-S junctions and somatic mutations in the S regions were analyzed by sequence alignment as performed by comparing PCR products sequences with germline Sμ and σσδ, Sγ1, or Sα sequences using National Center for Biotechnology Information BLAST (www.ncbi.nih.gov/BLAST).
RAD52 knockdown in human B cells. The human RAD52-specific siRNA oligo duplex (TT320001, Locus ID 5893) and non-effective Trilencer-27 Flurescencelabeled transfection control siRNA duplex (SR30002) were obtained from Origene Technologies. The siRNA duplexes were used to transfect purified human naïve B cells using the Human B Cell Nucleofector TM Kit (VPA-1001, LONZA) and Amaxa Nucleofector TM Device (LONZA). Transfected B cells were then stimulated with CpG ODN 2395 (Custom synthesized by Eurofins Genomics) plus IL-2 and IL-21 for 96 h before genomic DNA extraction for analysis of Sμ-σδ and Sμ-Sγ1 DNA recombination. Expression of RAD52 and AICDA transcripts were analyzed by qRT-PCR using specific primers 24 h after transfection. Expression of RAD52, AID, and β-ACTIN proteins were analyzed by immune-blotting 24 h after transfection.
High-throughput mRNA-Seq. RNA was isolated from cells using the Directzol RNA Microprep Kit (Zymogen Research), according to manufacturer's instructions and as previously described 69 . RNA integrity was verified using an Agilent Bioanalyzer 2100 (Agilent). Next generation RNA-Seq for mRNA and non-coding RNA was performed by the Genome Sequencing Facility at University of Texas Health Science Center San Antonio. High-quality RNA was processed using an Illumina TruSeq RNA sample prep kit v2 or TruSeq Small RNA Sample Prep kit following the manufacturer's instructions (Illumina). Clusters were generated using TruSeq Single-Read Cluster Gen. Kit v3-cBot-HS on an Illumina cBot Cluster Generation Station. After quality control procedures, individual mRNA-Seq or small RNA-Seq libraries were then pooled based on their respective 6-bp index portion of the TruSeq adapters and sequenced at 50 bp/sequence using an Illumina HiSeq 3000 sequencer. Resulting reads were checked by assurance (QA) pipeline and initial genome alignment (Alignment). After the sequencing run, demultiplexing with CASAVA was employed to generate the Fastq file for each sample. All sequencing reads were aligned with their reference genome (UCSC mouse genome build mm9) using TopHat2 default settings, and the Bam files from alignment were processed using HTSeq-count to obtain the counts per gene in all samples 67 . Quality control statistical analysis of outliers, intergroup variability and distribution levels, were performed for statistical validation of the experimental data.
Statistical analysis. Statistical analysis was performed using Excel (Microsoft) or GraphPad Prism ® software. P-values were determined by paired and unpaired Student's t-tests; P-values < 0.05 were considered significant.
IRB for use of human tissues and peripheral blood as well as IACUC for use of mice. For the use of DNA procured from formalin-fixed paraffin-embedded tissues obtained from the University of Arkansas for Medical Science, the study was reviewed by the University of Arkansas for Medical Sciences Institutional Review Board (IRB) which determined that this project is not human subject research as defined in 45 CFR 46.102. Human B cells were purified from PBMCs of healthy subject buffy coats obtained from South Texas Blood and Tissue Center, San Antonio, Texas, under the "Healthy Volunteer Blood Donor Program". SLE patient B cells were purified from PBMCs obtained under the University of Texas Long School of Medicine IRB HSC 20140234H (including informed consent) "Class switching, somatic hypermutation and plasma cell differentiation in B cells". Informed consents were obtained from the study participants. Studies involving mice and mouse-derived materials were performed under the University of Texas Long School of Medicine IACUC 20200019AR "Somatic hypermutation, classswitch DNA recombination and plasma cell differentiation in antibody and autoantibody responses".
Antibodies. All polyclonal Abs and mAbs used in these studies are listed in Supplementary Table 2, including Cat. number, clone number (for mAbs), vendor and working dilution.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The mRNA-sequencing data are available in NCBI Gene Expression Omnibus (GEO) through GEO Series accession number GSE156904. S-S junction sequencing data has been deposited in NCBI Sequence Read Archive (SRA) under BioProject PRJNA789895. The raw numbers for charts and graphs are available in the Source Data file whenever possible. The source data underpinning Figs. 2a-c, 3a, c, 4a, 5a-f, 6a-f, 7a, c-h, j, k, 8a-e, 9c, and 10a-c, g-k are available in the Source Data file. All other data supporting the findings of this study are available within the article and its supplementary information files and from the corresponding author upon reasonable request. Source data are provided with this paper.