Published online 5 May 2021 Nucleic Acids Research, 2021, Vol. 49, No. 9 5057–5073 doi: 10.1093/nar/gkab322 Activation-induced deaminase is critical for the establishment of DNA methylation patterns prior to the germinal center reaction Francesc Català-Moll1,2 , Anna G. Ferreté-Bonastre1,2,† , Tianlu Li1,2,† , Dieter Weichenhan3 , Pavlo Lutsik3 , Laura Ciudad1,2 , Ángel F. Álvarez-Prado4 , Javier Rodrı́guez-Ubreva1,2 , Christian Klemann5 , Carsten Speckmann5,6 , Amaya Vilas-Zornoza7 , Hassan Abolhassani8 , Mónica Martı́nez-Gallo9 , Romina Dieli-Crimi9 , Jacques G. Rivière10,11,12 , Andrea Martı́n-Nalda10,11,12 , Roger Colobran9,12,13 , Pere Soler-Palacı́n10,11,12 , Sven Kracker14 , Lennart Hammarström8 , Felipe Prosper7 , Anne Durandy14 , Bodo Grimbacher5,15,16,17,18 , Christoph Plass3 and Esteban Ballestar 1,2,* 1 Epigenetics and Immune Disease Group, Josep Carreras Research Institute (IJC), 08916 Badalona, Barcelona, Spain, 2 Chromatin and Disease Group, Cancer Epigenetics and Biology Programme (PEBC), Bellvitge Biomedical Research Institute (IDIBELL), 08908 L’Hospitalet de Llobregat, Barcelona, Spain, 3 Division of Cancer Epigenomics, German Cancer Research Center (DKFZ), Heidelberg 69120, Germany, 4 B Cell Biology Laboratory, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), 28029 Madrid, Spain, 5 Institute for Immunodeficiency, Center for Chronic Immunodeficiency (CCI), Medical Center, Faculty of Medicine, University of Freiburg, 79110 Freiburg, Germany, 6 Faculty of Medicine, Center for Pediatrics and Adolescent Medicine, Medical Center, University of Freiburg, Germany, 7 Center for Applied Medical Research (CIMA), University of Navarra, 31008 Pamplona, Spain, 8 Division of Clinical Immunology, Department of Laboratory Medicine, Karolinska Institutet, Karolinska University Hospital Huddinge, SE 14186 Stockholm , Sweden, 9 Immunology Division, Hospital Universitari Vall d’Hebron and Diagnostic Immunology Research Group, Vall d’Hebron Research Institute (VHIR), Barcelona, Spain, 10 Pediatric Infectious Diseases & Immunodeficiencies Unit, Hospital Universitari Vall d’Hebron, Vall d’Hebron Research Institute (VHIR), Autonomous University of Barcelona, Barcelona, Spain, 11 Infection in Immunocompromised Pediatric Patients Research Group, Vall d’Hebron Institut de Recerca (VHIR), Hospital Universitari Vall d’Hebron, 08035 Barcelona, Spain, 12 Jeffrey Modell Diagnostic and Research Center for Primary Immunodeficiencies, Barcelona, Spain, 13 Department of Cell Biology, Physiology and Immunology, Universitat Autònoma de Barcelona (UAB), Bellaterra, Catalonia, Spain, 14 Laboratory of Human Lymphohematopoiesis, Imagine Institute, INSERM UMR 1163, Université de Paris, Paris, France, 15 German Center for Infection Research (DZIF), Satellite Center Freiburg, Germany, 16 Centre for Integrative Biological Signalling Studies (CIBSS), Albert-Ludwigs University, Freiburg, Germany, 17 RESIST, Cluster of Excellence 2155 to Hanover Medical School, Satellite Center Freiburg, Germany and 18 Institute of Immunity & Transplantation, Royal Free Hospital, University College London, UK Received February 07, 2020; Revised April 15, 2021; Editorial Decision April 16, 2021; Accepted April 19, 2021 ABSTRACT body deficiency. AID-mediated deamination has been proposed as leading to active demethylation of 5- Activation-induced deaminase (AID) initiates anti- methycytosines in the DNA, although evidence both body diversification in germinal center B cells by supports and casts doubt on such a role. In this deaminating cytosines, leading to somatic hyper- study, using whole-genome bisulfite sequencing of mutation and class-switch recombination. Loss-of- HIGM2 B cells, we investigated direct AID involve- function mutations in AID lead to hyper-IgM syn- ment in active DNA demethylation. HIGM2 naı̈ve and drome type 2 (HIGM2), a rare human primary anti- memory B cells both display widespread DNA methy- * To whom correspondence should be addressed. Tel: +34 935572800; Fax: +34 934651472; Email: eballestar@carrerasresearch.org † The authors wish it to be known that, in their opinion, the second and third authors should be regarded as Joint Second Authors. Present address: Christian Klemann, Department of Paediatric Pulmonology, Allergy and Neonatology, Hannover Medical School, Hannover, Germany. C The Author(s) 2021. Published by Oxford University Press on behalf of Nucleic Acids Research. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com 5058 Nucleic Acids Research, 2021, Vol. 49, No. 9 lation alterations, of which ∼25% are attributable to analyses, or the study of different cells (in vitro activated or active DNA demethylation. For genes that undergo isolated GC B cells). active demethylation that is impaired in HIGM2 in- On the other hand, different in vitro studies suggest dividuals, our analysis indicates that AID is not di- that 5mC is a poorer substrate than C, although early rectly involved. We demonstrate that the widespread studies showed that human AID can deaminate 5mC (5). For instance, Abdouni and colleagues showed that the ef- alterations in the DNA methylation and expression ficient deamination of 5mC by zebrafish AID is due to profiles of HIGM2 naı̈ve B cells result from prema- the flexibility of its structure, in comparison with that of ture overstimulation of the B-cell receptor prior to other AID orthologs, including human AID (14). In ad- the germinal center reaction. Our data support a role dition, the comparison of AID with other members of for AID in B cell central tolerance in preventing the the AID/APOBEC family have shown that human AID expansion of autoreactive cell clones, affecting the deaminates 5mC only weakly because the 5-methyl group correct establishment of DNA methylation patterns. fits poorly in its DNA-binding pocket (15). Larijani et al. have shown that methylated-CpG motifs, but not their un- methylated counterparts, are in fact protected from AID- INTRODUCTION mediated deamination (16). Nabel and colleagues have also Hyper-IgM syndrome type 2 (HIGM2) is a rare primary shown that steric requirements for cytosine deamination are antibody deficiency, with autosomal recessive inheritance, one intrinsic barrier to the proposed function of deaminases characterized by loss-of-function mutations in activation- in DNA demethylation (17). induced deaminase (AID) (1), an enzyme required for sev- In parallel, the discovery of alternative enzymatic path- eral crucial steps of B cell terminal differentiation. AID con- ways that lead to bona fide active DNA demethylation verts deoxycytosines (dCs) into deoxyuracils (dUs), produc- through ten-eleven translocation methylcytosine dioxyge- ing dU:dG mismatches that are removed by mismatch re- nase (TET)-mediated oxidation of methylcytosines (18,19) pair and base-excision repair (2). Deaminase activity is re- raised more doubts about the possibility that AID redun- quired for somatic hypermutation (SHM) and class-switch dantly plays such a role. There is currently no consensus recombination (CSR) of immunoglobulin (Ig) genes, which about whether AID is involved in mediating DNA demethy- are necessary processes for affinity maturation and antibody lation in specific cell contexts. diversification within the germinal centers (GC) (3,4). AID Whole-genome analysis has shown the occurrence of a deficiency results in the absence of CSR and SHM, and vast amount of demethylation associated with B cell differ- leads to lymphoid hyperplasia (1). HIGM2 patients have entiation. Changes occur mostly during naı̈ve B cell activa- normal or elevated serum IgM levels with severe reduction tion, yielding memory B cells (20,21) that coincide with the of IgG, IgA and IgE, resulting in considerable susceptibility highest peak of AID expression (3). Naı̈ve B cells start to to bacterial infections (1). proliferate upon activation by antigen encounter. Then they In addition to its role in CSR and SHM, AID has express AID which triggers the secondary diversification of been proposed to participate in active DNA demethylation antibodies by SHM and CSR. This is followed by affinity through deamination of 5-methylcytosine (5mC), leading maturation which finally leads to (a) a new cycle of SHM or to a mismatch that is converted to G:C by thymine DNA (b) terminal differentiation into memory or plasma B cells glycosylase (TDG), followed by base-excision repair. The depending on the affinity of the B cell receptor (BCR) for potential role of AID in active DNA demethylation was the cognate antigen (22). It has been originally reported that first proposed by Petersen-Mahrt and colleagues (5). Initial AID is targeted to Ig genes with the involvement of RNA studies in this topic mainly focusing on non-lymphoid cells, polymerase II (23). Nevertheless, it has since been described such as including zebrafish embryos and heterokaryon- that AID is also recruited to non-Ig genes to mediate recur- based reprogramming, supported that model (6–8). These rent mutations and these genes are collectively termed ‘AID studies subsequently led to its study in B cells, given that off-targets’ (24,25). activated B cells display the highest levels of AID expres- In this study, we took advantage of the exceptional possi- sion (9). During the past decade, conflicting reports have bility to investigate the direct role of AID in active demethy- both supported and discounted a role in active demethyla- lation by comparing the complete DNA methylomes of tion for AID in that context [reviewed in (10)]. For instance, naı̈ve and memory B cells of HIGM2 patients with those of Fritz et al. (11) performed reduced-representation bisulfite healthy individuals. By studying two sibling patients with sequencing (RRBS) of mouse splenic naı̈ve B cells from wild a homozygous mutation for AID that results in a severely type and AID-deficient mice, activated ex vivo for 72 h and truncated enzyme, we were able to determine its direct link found no significant differences in their DNA methylation with DNA methylation defects and infer its catalytic activ- profiles. Similar conclusions were obtained looking at GC ity in relation to active DNA demethylation. B cells from AID-deficient mice and using MethylCap-Seq Our results show that the absence of AID catalytic ac- (12). However, more recently, Dominguez and colleagues tivity affects DNA methylation in naı̈ve and memory B showed that the transit of B cells through the GC is asso- cells. The majority of the changes occurring in the transition ciated with marked locus-specific loss of methylation and from naı̈ve to memory B cells arise from passive demethy- increased methylation diversity, both of which are lost in lation and are linked to late-replicating domains. How- Aicda−/− animals (13). Methodological aspects could ex- ever, for those potentially associated with active demethy- plain the discrepancies between these studies, including the lation, we found no evidence of direct involvement of AID, limited coverage and resolution of the three aforementioned and our analysis suggests that TET enzymes are responsi- Nucleic Acids Research, 2021, Vol. 49, No. 9 5059 ble for DNA methylation changes in this cell context. The extracted with a Maxwell RSC Cultured Cells DNA kit increased DNA demethylation noted in naı̈ve B cells of (Promega), following the manufacturer’s protocol. HIGM2 patients is associated with premature demethyla- tion of BCR downstream genes prior to the GC reaction. Tagmentation-based whole-genome bisulfite sequencing. Indeed, we found that these changes are related to the ex- For whole-genome bisulfite sequencing, 30 ng of genomic pansion of autoreactive clones, which suggests a major role DNA was used to produce four independent barcoded for AID in preventing the expansion of such clones under sequencing libraries per DNA sample using the tagmen- normal conditions. tation method (27). Sequencing of the TWGBS libraries was done on a HiSeq 2000, PE 125 bp mode. Bisulfite sequencing reads were processed by the DKFZ bisul- MATERIALS AND METHODS fite analysis workflow. In brief, the reads were trimmed Human samples using Trimmomatic, pre-processed and aligned using MethylCTools, with default parameters (V. Hovestadt, S. The patients included in the study fulfilled the diagnostic Picelli, B. Radlwimmer, M.Z. and P.L., unpublished data), criteria for hyper-IgM syndrome type 2, based on ESID which uses the Burrows-Wheeler alignment algorithm (28). clinical diagnostic criteria (26) and genetic confirmation of Following quality control of bisulfite conversion (>99.5% AICDA mutation and exclusion of other primary and sec- in all samples) and of read-mapping (80–90% could be ondary causes of immunodeficiencies. Samples come from mapped on average), we performed methylation calling the Medical Center of the University Hospital, University using methylCtools. A summary of the sequencing data for of Freiburg, Freiburg, Germany and Hospital Universitari each sample is provided in Supplementary Table S1. Vall d’Hebron, Barcelona, Spain. The Committees for Hu- man Subjects of the local hospitals approved the study, which was conducted in accordance with the ethical guide- Differentially methylated region (DMR) calling. DMRs lines of the 1975 Declaration of Helsinki. All samples were were detected with the DeNovoDMR algorithm included in compliance with the guidelines approved by the local in the Specific Methylation Analysis and Report Tool ethics committee and all donors (and/or their parents) re- (SMART2) (29) using all the default parameters except for ceived oral and written information about the possibility the number of CpGs per segment, which was set to 4, the that their blood would be used for research purposes. absolute mean methylation difference, which was set to 0.2, and a threshold value of P of 0.01. Only those CpGs with a coverage of ≥5 in all samples were considered in the Isolation of B cell populations. Peripheral blood mononu- construct of the SMART input matrix. DMR calling was clear cells (PBMCs) were obtained from blood. After Ficoll- performed on all possible comparisons between naı̈ve and Isopaque density centrifugation (Rafer, Zaragoza, Spain), memory B cells for both control and HIGM2 patients. collected cells were washed twice with ice-cold PBS, fol- lowed by centrifugation at 2000 rpm for 5 min. Next, cells were labeled with antibodies to CD19 – FITC (Miltenyi Bisulfite pyrosequencing. 500 ng of genomic DNA was Biotec, clone LT19), CD27 – APC (Miltenyi Biotec, clone converted with an EZ DNA Methylation-Gold kit (Zymo M-T271), IgD – PE (SouthernBiotech, Cat. No. 2032-09) Research), following the manufacturer’s instructions. and IgM – PerCP/Cy5.5 (BioLegend, clone MHM-88) for Bisulfite-treated DNA was PCR amplified using primers 20 min on ice in a staining buffer (PBS with 4% FBS and (see Supplementary Table S2) designed with PyroMark As- 2 mM EDTA). Naı̈ve B cells (CD19+ CD27– IgD+ ) and say Design 2.0 software (Qiagen). Finally, PCR amplicons unswitched memory B cells (CD19+ CD27+ IgD+ ) were were pyrosequenced with the PyroMark Q24 system and obtained by FACS sorting on a MoFlo Astrios (Beck- analyzed with PyroMark CpG software (Qiagen). man Coulter). Purified samples were pelleted and stored at −80◦ C. For isolation of naı̈ve autoreactive B cells. Total B cells Processing of chromatin immunoprecipitation (ChIP) assays were isolated from PBMCs using positive selection with with sequencing (ChIP-Seq) datasets MACS CD19 microbeads (Miltenyi Biotec). Next, cells Sequencing reads from ChIP-seq experiments from the were stained with CD27-APC (Miltenyi Biotec, clone M- BLUEPRINT consortium (21) were mapped to the hg19 as- T271), IgD – PE (SouthernBiotech, Cat. No. 2032-09), sembly of human reference genome using Burrows-Wheeler HLA-DR – PE-Cy7 (eBioscience, clone LN3), 9g4 primary Aligner (BWA) v0.7.13 (with parameters -q 5, -l 32, -k ab (igm Bioscience) and donkey anti-rat IgG (H + L) – 2). After removing reads with MAPQ <30 with Sequence Alexa Fluor 488 (invitrogen). 9g4+ naı̈ve B cells (CD27– Alignment/Map (SAMtools) v1.2, PCR duplicates were IgD+ 9g4+ ) and 9g4- naı̈ve B cells (CD27– IgD+ 9g4– ) were eliminated using the Picard function available in MarkDu- obtained by FACS sorting on a BD FACSAria II (BD plicates software v1.126. Peak calling was performed us- Biosciences). Purified samples were pelleted and stored at ing macs2 (with parameters -p 1e-2 –nomodel –shift 0 - −80◦ C. B –SPMR). Only peaks with an overlap of ≥0.5 between replicates were considered. Histone mark signals around Genomic DNA extraction. For whole-genome bisulfite se- DMR sets were extracted with the annotatePeaks.pl algo- quencing, DNA was extracted with a QIAamp DNA mi- rithm available in Hypergeometric Optimization of Motif cro kit (Qiagen) according to the manufacturer’s protocol. EnRichment (HOMER) software v4.10.3 (with parameters: For pyrosequencing analysis (detailed below), DNA was size = 10 000, hist = 10). 5060 Nucleic Acids Research, 2021, Vol. 49, No. 9 Super-enhancer identification. H3K27ac ChIP-seq data statistical test. Levels of significance are indicated as: from BLUEPRINT database were used to identify the *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Non- super-enhancer regions, as described previously (30) using significance (P ≥ 0.05) was indicated as ‘ns’. Rank-Ordering of Super-Enhancers (ROSE) software. An enhancer stitching distance of 15 kb was used along with a Public RRBS of B cell activation. Data of EBV and 2.5 kb transcriptional start site (TSS)-exclusion window. CD40/IL4 B cell activation were downloaded from the NCBI Gene Expression Omnibus (GSE49629) (36). Methy- DNA methylation data analysis. Hierarchical clustering lation calls from RRBS data were filtered so that only those was carried out based on Pearson correlation distance met- CpGs with a minimum of five reads per position in all sam- rics and average linkage criteria. For low-dimensional anal- ples were retained. Since RRBS genomic coverage is signif- ysis, we used the t-distributed stochastic neighbor embed- icantly lower than T-WGBS we only tested the methylation ding (t-SNE) method implemented in the Rtsne v0.15 pack- status of positions common to two datasets. age. Transcription factor motifs were enriched for each set of BCR activation and EBV infection. For naı̈ve B cell acti- DMRs using HOMER software v4.10.3. Specifically, we vation and infection experiments, we obtained buffy coats used findMotifsGenome.pl algorithm (with parameters - from anonymous donors through the Catalan Blood and size given -cpg) to search for significant enrichment against Tissue Bank (CBTB). The CBTB follows the principles a background sequence adjusted to have similar CpG and of the World Medical Association (WMA) Declaration GC contents. of Helsinki. Before providing the first blood sample, all Transcription factor binding analysis was performed in- donors received detailed oral and written information and terrogating the overlap between the different sets of DMRs signed a consent form at the CBTB. PBMCs were isolated with ChIP-seq data for transcription factors available for using Ficoll-Paque gradient centrifugation. Total B cells GM12878 lymphoblastoid cell line from the ENCODE were isolated from PBMCs using positive selection with Project (31). The enrichment factor was calculated against MACS CD19 microbeads (Miltenyi Biotec). Next, cells random regions as a background, and P values were calcu- were stained with CD27-APC (Miltenyi Biotec, clone M- lated using Fisher’s exact test. Finally, the transcription fac- T271) and IgD – PE (SouthernBiotech, Cat. No. 2032–09) tors downstream of the BCR signaling pathway were man- and naı̈ve B cells were sorted as CD27– IgD+ . For EBV in- ually annotated from a curated database (32). fection, pure naı̈ve B cells were incubated with B95–8 cell Chromatin states and histone mark enrichments analysis supernatant for 3 h at 37ºC in order to infect them with for NBC, GC B cells and ncsMBC were assessed using a EBV, they were harvested after 30 days. For BCR activa- custom adaptation of the EpiAnnotator R package (33) us- tion, pure naı̈ve B cells were cultured in RPMI medium ing BLUEPRINT data (21). DMRs were converted to hg38 1640 GlutaMAXTM-1 (Gibco, Life Technologies) contain- assembly with the liftOver function in the rtracklayer v1.42 ing 20% fetal bovine serum (Gibco, Life Technologies), 1% R package. sodium pyruvate and 1% Penicillin Streptomycin (Gibco Replication timing data in the GM12878 lymphoblastoid Thermo Fisher Scientific, MA, USA). Cells were activated cell line were obtained from the UW Repli-seq track of the with 10 ␮g/ml anti-IgM (SouthernBiotech), 0.1 ␮g/ml UCSC Genome Browser. Genomic replication timing val- MEGACD40L® Protein (Enzo) and 50 ng/ml IL21 (Tebu- ues were binned in deciles where the first decile contained bio) for 24 h (ChIP) or 7 days (bisulfite pyrosequencing). the regions of latest replication and the last decile the re- RNA-sequencing. RNA-sequencing libraries from total gions of earliest replication. Identified DMRs were then mRNA were prepared and purified using Illumina Stranded overlapped to these regions by genomic location. mRNA Prep, Ligation kit (Illumina). Sequencing was DMR annotation for genetic context location was performed on a NextSeq500 (Illumina) with a NextSeq performed using the annotatePeaks.pl algorithm in the 500/550 High Output Kit v2.5 (75 Cycles, Illumina, HOMER software v4.10.3. To determine the location rela- 20024906) to generate 38-nucleotide paired-end reads at a tive to a CpG island (CGI), we used ‘hg19 cpgs’ annotation read depth of at least 20 million reads per sample. in the annotatr v1.8 R package. GREAT software (34) was used to enrich downstream RNA-seq data analysis. For bioinformatics analyses, raw pathways and gene ontologies. We used the single nearest sequence reads were initially processed using FastQC gene option for the association between genomic regions (Babraham Institute, Cambridge, UK) for quality control, with genes. and then adapter sequences and poor quality reads were Genes obtained with HOMER annotation software were removed using TrimGalore that uses Cutadapt for trim- used to study correlations between their associated DMRs ming. Quality-filtered reads were then mapped to the hu- and gene expression data from healthy naı̈ve, memory and man genome (hg19) using STAR, and only the uniquely GC B cells from BLUEPRINT database, as well as with the mapped reads were kept. Read counts were calculated using AID off-target genes dataset obtained from Álvarez-Prado HTSeq-count. Differentially expressed genes were identi- et al. (35). fied using R package DESeq2 (fold change ≥ 2 and adjusted All statistical analysis (excluding T-WGBS and ChIP- P-value < 0.05). Inference of TF activities from expression seq analyses) were done in R v3.5.1. Data distribu- values were calculated using DoRothEA (37). Kyoto En- tions were tested for normality. Normal data were tested cyclopedia of Genes and Genomes (KEGG) database was using two-tailed unpaired Student’s t-tests; non-normal used to perform functional enrichment analysis of differen- data were analyzed with the appropriate non-parametric tially expressed genes. Nucleic Acids Research, 2021, Vol. 49, No. 9 5061 ChIP-PCR. Cell-sorter isolated naı̈ve B cells from healthy Global inspection of DNA methylation confirmed pre- donors were activated as described above for 24 h. 1 × 106 vious studies (20,21) that the transition from naı̈ve to unstimulated and stimulated B cells were crosslinked with memory B cells is accompanied by global demethylation 1% methanol-free formaldehyde (Thermo Fisher) for 10 of the genome in healthy controls (Figure 1D). However, min subjected to 15 min of sonication utilizing Covaris HIGM2 patients showed a partial impairment of demethy- M220 Focused-ultrasonicator (Covaris, MA, USA). ChIP lation during naı̈ve-to-memory B cell differentiation (Fig- experiments were performed using the LowCell# ChIP kit™ ure 1D, E), which is compatible with a potential role of protein A, according to manufacturer’s instructions (Di- AID as a demethylating enzyme. The transition from naı̈ve agenode, Liege, Belgium). Anti-BATF antibody (#8638) to memory B cells in healthy controls is characterized by and corresponding rabbit IgG (#2729) were both obtained the demethylation of 30175 DMRs. In contrast, compar- from Cell Signaling. Protein binding was analyzed by real- ing naı̈ve to memory B cells in HIGM2 patients only iden- time quantitative PCR, and data are represented as a ratio tified 4803 hypomethylated DMRs (Supplementary Table of the enriched fraction with respect to input. ChIP primers S3). Furthermore, we also observed that naı̈ve B cells were are shown in Supplementary Table S2. more demethylated in HIGM2 patients than in healthy con- trols (a total of 2936 hypomethylated DMRs) (Figure 1F). RESULTS Taken together, these global observations suggest that AID loss not only affects the DNA methylation patterns in the Study strategy transition from naı̈ve to memory B cells but also has a po- We obtained peripheral blood from two sibling HIGM2 tential role in establishing the B cell DNA methylome in ear- patients, both with the same homozygous mutation lier stages of development. for the AICDA gene, and two sex-matched healthy controls. Specifically, the patients carried a deletion A high proportion of DNA demethylation events identified in (Exon 2 c.22 40del19) that generates a frameshift vari- HIGM2 are due to passive demethylation of late-replicating ant (p.Arg8Asnfs*19) that affects the DNA binding re- domains gion of AID located within its nuclear localization sig- nal domain (Figure 1A). This deletion impairs both CSR Recent studies have shown that a high proportion of the and SHM (38). We confirmed the impact on CSR by in- demethylation events occurring in cancer and in differenti- specting the peripheral B cell compartment by flow cy- ation processes are associated with high proliferation rates tometry. The two HIGM2 patients were characterized by (40). Such demethylation events take place in regions known the absence of class-switched memory B cells (csMBC; as ‘partially methylated domains’ (PMDs), which are char- CD19+ CD27+ IgM– IgD– ), as described (1) (Figure 1B). In acterized by repressive chromatin and low-GC density (40), contrast, classic non-class switched memory B cells (nc- and do not appear to occur in regions of high methylation sMBC; CD19+ CD27+ IgM+ IgD+ ) and naı̈ve B cells (NBC; (HMD: highly methylated domains). PMDs encompass re- CD19+ CD27– IgM+IgD+) were present in patients (Figure gions that undergo late replication and their demethyla- 1B). It has been shown that, under physiological condi- tion is a passive event, which is a result of inefficient DNA tions, ncsMBC cells display certain levels of SHM at the remethylation during DNA replication (36,40,41). Recent immunoglobulin locus (20), which supports the expression reanalysis of the B cell lineage DNA methylation profiles of AID during their maturation in GCs. Hence, the com- published by the BLUEPRINT consortium (21) has shown parison between DNA methylation profiles of NBC and nc- the occurrence of demethylation of PMDs in the transi- sMBC isolated from healthy and HIGM2 individuals is an tion towards memory B cells and antibody-secreting plasma adequate model to investigate the potential role of AID in cells (42). This highlights the importance of separating DNA demethylation. DNA methylation analysis into PMD and non-PMD re- gions when interrogating the occurrence of active demethy- lation processes in order to exclude those changes due to HIGM2 patients display an aberrant methylation profile in DNA replication-dependent or passive demethylation. naı̈ve and unswitched memory B cells To address this matter, we first identified total dif- We performed tagmentation-based whole-genome bisulfite ferentially methylated regions (DMRs) by performing sequencing (T-WGBS), a version of the WGBS method that pairwise comparisons of DNA methylation between all allows analysis of limited DNA amounts (27), in NBC and samples. Specifically, we compared naı̈ve to memory ncsMBC (henceforth referred to as ‘naı̈ve’ and ‘memory’ B cells in both control and HIGM2 patients (con- B cells) isolated from two HIGM2 siblings and two sex- trol NBC versus control ncsMBC, HIGM2 NBC versus matched healthy controls (Figure 1C). Pearson correlation HIGM2 ncsMBC), as well as comparing HIGM2 patients and t-distribution stochastic neighbor embedding (t-SNE) to controls for both naı̈ve and memory B cells (con- between samples were highly reproducible between repli- trol NBC versus HIGM2 NBC, control ncsMBC versus cates with correlation coefficient to be >0.9 (Supplementary HIGM2 ncsMBC). We then divided the DMRs into PMD Figure S1A, B). We also compared our DNA methylation and HMD regions, which were previously identified by data from healthy controls with public data from the In- Zhou et al. (40) to be present either in all cell types (com- ternational Cancer Genome Consortium (ICGC) (39) and mon) or cell-type specific (partial). We found that the major- Oakes et al. (20), in which we observed high correlation ity of DMRs overlapped with PMDs that were either com- (Pearson correlation >0.85), thereby confirming the robust- mon or partially shared by cell types (72.5%, Figure 2A). ness of our data (Supplementary Figure S1C, D). We then classified DMRs into two groups: the first com- 5062 Nucleic Acids Research, 2021, Vol. 49, No. 9 A 26 190 1 8 22 40 55 94 108 181 198 AIDWT NH2 NLS Catalytic domain APOBEC-like domain NES COOH 13 26 113 123 173 198 DNA binding Hotspot recognition loop CSR specific 190 1 8 22 40 55 94 108 181 198 AID ∆22-40 NH2 NLS Catalytic domain APOBEC-like domain NES COOH B C Control HIGM2 0.3% 38.6% 5.2% 88.0% ncsMBC ncsMBC NBC ncsMBC NBC ncsMBC IgD IgD csMBC csMBC 46.1% 9.9% 1.9% 3.9% IgM IgM Naïve Memory Naïve Memory 20.2% 0.9% 34.5% 6.3% + + + + CD19 CD19 CD19 CD19 CD27-IgD+ CD27+IgD+ CD27-IgD+ CD27+IgD+ CD27 CD27 Control Control HIGM2 HIGM2 78.0% 0.9% 38.7% 20.5% CD19 1.7% 85.5% CD19 2.0% 92.1% Exon 2 Exon 2 NBC NBC - - c.22_40del19 c.22_40del19 Homozygous Homozygous IgD IgD 3.6% 9.8% 1.48% 4.38% IgM IgM D E 1 1 Naïve Memory 2 Control HIGM2 2 3 % DNA methylation memory 100 r = 0.88999 100 r = 0.94683 3 4 4 75 75 5 5 50 50 6 6 25 25 DNA methylation level 7 0 0 7 0 25 50 75 100 0 25 50 75 100 8 % DNA methylation naïve % DNA methylation naïve 8 > 0.90 < 0.55 9 9 DNA methylation difference F Naïve Memory 10 < 0.03 > 0.25 100 100 % DNA methylation Control 10 11 11 75 75 12 12 13 50 50 13 14 25 25 14 15 r = 0.95507 r = 0.94643 15 16 0 0 16 17 0 25 50 75 100 0 25 50 75 100 17 18 18 HIGM2 Control 19 19 % DNA methylation HIGM2 % DNA methylation HIGM2 20 20 21 21 22 22 Figure 1. DNA methylomes of HIGM2 B cell subpopulations determined by T-WGBS. (A) Graphical representation of wild type (WT) AICDA gene and truncated AID of HIGM2 patients used in this study. The mutation in HIGM2 AID harbors a deletion from 22–40 which encompasses the DNA binding domain. (B) Representative examples of cell sorting strategy of B cell populations (naı̈ve B cells, NBC; classic non-class-switched memory B cells, ncsMBC; class-switched memory B cells, csMBC). (C) Description of the B cell subpopulations analyzed, including surface markers and details of mutation. (D) Circos plot of DNA methylation of naı̈ve B cells (inner circle) and unswitched memory B cells (outer circle) for controls (right) and HIGM2 patients (left). Histogram tracks represent the average methylation levels of 10 Mb windows. Heatmap shows the DNA methylation differences between naı̈ve and memory B cells (E) Density scatter plots of average DNA methylation in windows of 500 bp in naı̈ve and memory B cells in controls (left) and HIGM2 patients (right). (F) Density scatter plots of average DNA methylation in windows of 500 bp comparing HIGM2 and control for naı̈ve (left) and memory (right) B cells. Nucleic Acids Research, 2021, Vol. 49, No. 9 5063 A B C 100 PMD-DMRs non-PMD-DMRs 80 60 1.00 outside CGI % DMRs 60 Shelf Shore DNA methylation level 40 Island 0.75 % DMRs 40 20 0 0.50 nd Rs s r o u -DM MR -D 20 ckg MD MD Ba P n -P no 0.25 100 0 non-coding 80 D D TTS D D 0.00 % DMRs HM HM PM PM 60 3'UTR on al on al ïve ry ïve ry ïve ry ïve ry intron m rti m rti Na mo Na mo Na mo Na mo m pa co m pa Me Me Me Me exon co 40 5'UTR Control HIGM2 Control HIGM2 promoter-TSS 20 Intergenic 0 D un d MR s MR s gro D-D -D ck MD Ba PM - P PMD-DMRs non-PMD-DMRs no n active promoter weak promoter F 10 1 poised promoter Log 2 (FC) strong enhancer 1 1.5 strong enhancer 2 9 2 non-PMD-DMRs weak enhancer 0,0 transcription PMD-DMRs transcription elongation weak transcription % DMR -1.5 8 3 Polycomb repressed 10 20 30 40 50 heterochromatin/low signal heterochromatin/ epressed ïve er ed ïve er ed Na ent itch Na ent itch 4 lC sw lC sw 7 rm ina Un rm ina Un Ge Ge Significant (qv < 0.01) Percentage of DMRs 6 5 no yes 25 50 75 Early Late Replication E H3K27ac H3K36me3 H3K4me3 G H3K27me3 H3K4me1 H3K9me3 **** PMD-DMRs non-PMD-DMRs 4 1.5 Mean coverage (10bp bin) Log10 (FPKM + 1) 3 1.0 2 0.5 1 0 -5 -2.5 0 2.5 5 -5 -2.5 0 2.5 5 non-PMD-DMRs PMD-DMRs Distance from DMR (kbp) Figure 2. Detection and characterization of partially methylated regions. (A) Bar plot showing the percentage of DMRs in PMD and HMD (partially and highly methylated domains, respectively) regions annotated by Zhou and colleagues (40). ‘Common’ refers to PMD and HMD regions common to all cell types analyzed by Zhou and colleagues. ‘Partial’ refers to those regions detected only in a fraction of the cell types analyzed. (B) Box and violin plots summarizing the distribution of DNA methylation levels for controls and HIGM2 patients, in which the DMRs were characterized as inside (PMD-DMRs) and outside (non-PMD-DMRs) of PMDs. (C) Location proportions of PMD-DMRs and non-PMD-DMRs in the context of CpG islands (CGIs) and gene-related regions. (D) Bubble chart depicting the enrichment (red) or depletion (blue) of chromatin states defined from ChromHMM ChIP-seq data of healthy naı̈ve, GC and unswitched memory B cells. Color scale represents the logarithmic fold change, circle size indicates the percentage of DMRs in the chromatin state, and the edge color indicates the statistical significance of the enrichment (black: significant, none: not significant; q < 0.01). (E) ChIP-seq data of the represented histone marks in healthy GC B cells were downloaded from the BLUEPRINT database and mean coverage was calculated for 10bp bins ±5 kb centered around the midpoints of PMD-DMRs (left) and non-PMD-DMRs (right). (F) Circular representation of the proportion of PMD- (outer circle) and non-PMD-DMRs (inner circle) in deciles of replication timing data of GM12878 cell line. Color scale represents the proportion of DMRs in each decile. The first and last deciles correspond to the regions of latest and earliest replication, respectively. (G) Box and dot plots showing the distribution of expression values of associated genes of PMD- and non-PMD-DMRs. FPKM refers to fragments per kb of transcript per million mapped reads. Expression data from healthy GC B cells downloaded from BLUEPRINT database. Statistical tests: two-tailed Fisher’s exact (D) and unpaired Wilcoxon’s (G) tests (**** P-value < 0.0001). 5064 Nucleic Acids Research, 2021, Vol. 49, No. 9 prising DMRs that coincided with common PMDs (PMD- 522 DMRs (containing 450 different genes) fulfilled these DMRs), and the second containing DMRs that did not conditions (Supplementary Table S4). overlap with common PMDs (non-PMD-DMRs). To val- We then compared P-AID DMRs associated genes with idate the identity of these PMD-DMRs, inspection of the data obtained from a mouse model that allows the identifi- methylation values showed that the two groups of DMRs cation of targets of AID deamination events (35), in which had intermediate methylation values in memory B cells, 271 non-Ig genes were identified to be dependent on AID although non-PMD-DMRs had slightly lower methyla- catalytic activity and referred to as AID off-target genes. We tion levels (Figure 2B). Utilizing functional genomic fea- found low correspondence between these two sets of genes ture analysis, we found that most PMD-DMRs were lo- in which only 6 genes (corresponding to 6 DMRs) overlap cated outside of CpG islands and in intergenic regions (Fig- with the identified AID off-target genes (Figure 3B). We ure 2C). In relation to chromatin states, we observed that also tested for the presence of described AID hotspot se- DMRs occurring in PMDs were mainly associated with het- quences (RGYW and WRCY) (44) in our DMRs and found erochromatic regions, while non-PMD-DMRs were highly a significant increase for the WRCY hotspot with respect to enriched at enhancers and active promoters, in concor- the background, although it appeared too low (<1 hotspot dance to what was previously described as generally associ- per 100 bp) to be of biological relevance (Figure 3C). On the ated with replication-independent demethylation (43) (Fig- other hand, we found that DMRs associated with AID off- ure 2D). This was also confirmed by overlapping DMRs target genes underwent demethylation in HIGM2 patients with ChIP-seq data of healthy GC B cells obtained from (Supplementary Figure S2B). the BLUEPRINT database, in which non-PMD-DMRs Two recent studies have characterized the genomic and were enriched in enhancer marks, including H3K4me1, epigenomic features of AID off-target regions in mice. Inde- H3K4me3 and H3K27ac (Figure 2E). It has been previ- pendently, they found that AID targets regions with conver- ously described that PMDs are characterized by their as- gent transcription from intragenic super-enhancers (45,46). sociation with late-replication domains (40), hence we ana- In this sense, there were no significant differences regard- lyzed DMRs based on their association with late or early ing gene localization between the DMRs associated with phase replication. First, genomic density data obtained AID off-targets and P-AID DMRs (Supplementary Fig- from GM12878 Repli-seq (UW Repli-seq, UCSC Genome ure S2C and Supplementary Table S4). However, AID off- Browser) were divided into deciles by their replication tim- target DMRs exhibited greater enrichment of enhancer re- ing and then overlapped by genomic location with identi- gions, as observed by both ChromHMM (Figure 3D) and fied DMRs. We confirmed that PMD-DMRs were mainly histone mark (Supplementary Figure S2D) analyses, and found in late-replicating regions (Figure 2F) and were ac- were associated with more transcriptional activity of as- companied by lower expression of associated genes in GC sociated genes in GC B cells (Figure 3E). We also found B cells (Figure 2G). Taking all these observations into ac- that, although the two DMR groups had a similar percent- count, our results suggest that most of the DNA methyla- age of overlap with super-enhancers (P-AID 21%, AID off- tion changes occur in bona fide PMDs. However, the exis- target 33.3%; Figure 3F), the super-enhancers of the AID tence of a set of non-PMD-DMRs (∼27%) located in highly off-targets had a stronger signal for H3K27ac (Figure 3G, active regions suggests the potential participation of DNA H; Supplementary Figure S2E) and greater transcriptional replication-independent or, in other words, active demethy- activity of their associated genes than P-AID (Supplemen- lation events, and AID could directly drive these events. tary Figure S2F). Finally, we hypothesized that if AID had a role mediating active DNA demethylation, we would expect to see differences in CpG sites containing WRCY hotspots HIGM2-associated defects in DNA methylation in the tran- at the super-enhancers of AID off-target genes. However, sition from naı̈ve to memory B cells do not have the features such differences were not observed (Supplementary Figure of AID targets S2G). Given all the previous considerations, including the removal Taken together, our results suggest that, despite the differ- of DNA methylation changes related to DNA replication ences in DNA methylation associated with B cell activation (PMD-DMRs), our model allows us to examine whether between wild type and AID-deficient B cells, if we assume AID has a direct role in mediating demethylation in the that mechanisms and targets are conserved between mouse B cell lineage. In this context, GC B cells, in the transi- and human, such demethylation may not be directly associ- tion from naı̈ve to memory, displayed the highest AICDA ated with AID catalytic activity. mRNA levels (Supplementary Figure S2A) (as obtained While the proposed mechanism of DNA demethyla- from the BLUEPRINT database). Hence, it is plausible that tion by AID implies that the 5mC conversion in thymine defective AID in HIGM2 patients could play a role in aber- could be repaired and replaced with an unmethylated cy- rant DNA demethylation occurring during the transition tosine (5), the removal of methyl groups of cytosines me- from naı̈ve to memory B cells within the GC. diated by TET enzymes involves the generation of oxida- To explore this possibility, we first identified DMRs tion intermediates, including 5hmC (18) (Supplementary whose DNA demethylation could potentially be driven by Figure S3A). It is therefore plausible that AID-dependent AID (P-AID DMRs). These DMRs are demethylated in the demethylation would associate with lower levels of 5hmC transition from naı̈ve to memory B cells in healthy controls compared to TET-dependent demethylation. Hence, we an- but not in HIGM2 patients. Furthermore, DMRs that were alyzed publicly-available DNA methylation data obtained already aberrantly demethylated in naı̈ve HIGM2 B cells from GC and naı̈ve B cells of Aicda–/– mice (13). We selected compared to controls were excluded (Figure 3A). A total of three sets of CpGs: (i) CpGs demethylated in the transition Nucleic Acids Research, 2021, Vol. 49, No. 9 5065 A B C 1.00 RGYW WRCY 20 ns ** 265 ns ns DNA methylation level Hotspots per 100 base pair ns * AID off-target genes 15 0.50 6 10 444 5 P-AID 0.00 0 ïve ry ïve ry Na mo Na mo Me Me d d un ID ets un AID rge ts gro P-A arg gro P- -ta Control HIGM2 Ba ck off-t Ba ck ff AID AIDo D E 10 1 AID off-targets P-AID active promoter weak promoter 9 2 P-AID poised promoter Log2 (FC) strong enhancer 1 AID off-targets strong enhancer 2 1.5 weak enhancer % DMR 0.0 8 3 transcription transcription elongation -1.5 20 40 60 80 weak transcription Polycomb repressed heterochromatin/low signal 4 7 heterochromatin/repressed ïve er ed ïve er ed Na ent itch Na ent itch 6 5 in al C Un sw in al C Un sw G erm G erm Significant (qv < 0.01) Percentage of DMRs Expression no yes 10 20 30 40 20Kb CD83 F G H DMR Super-enhancers H3K27ac signal at super-enhancers * H3K27ac Super-enhancer 1e+05 Enhancer 1086 20Kb BCL6 108 5e+04 DMR 85 2 408 4 164 H3K27ac AID off-targets AID off-targets P-AID Super-enhancer P-AID Enhancer Figure 3. Indirect involvement of AID in DNA demethylation dynamics during B cell activation. (A) Box and violin plot representation of DNA methy- lation of DMRs that are potentially demethylated directly by AID (P-AID). These DMRs were filtered by the presence of demethylation between control naı̈ve and memory B cells which were not present in HIGM2 B cells. Furthermore, DMRs that displayed demethylation in HIGM2 naı̈ve B cells compared to controls were excluded. (B) Venn diagram showing the overlap between associated genes of P-AID DMRs and the human orthologues of mouse AID off-target genes as defined by Álvarez-Prado et al. (35). (C) Box and violin plots displaying the frequency of WRCY/RGYW (W = A/T; R = G/A; Y = C/T) hotspots per 100 bp in DMRs of each subset. Background was generated as 100bp windows of the whole genome. (D) Bubble plot depicting the enrichment (red) or depletion (blue) of chromatin states. Color represents the logarithmic-fold change, size shows the percentage of DMRs in the chromatin state, and the edge indicates the statistical significance of the enrichment (black: significant, none: not significant, q-value < 0.01). Chromatin state data was obtained from ChromHMM, generated in GC B cells. (E) Circular representation of proportion of DMRs in deciles of genes whose expression is ranked from low to high expression in GC B cells. The first and last deciles correspond to the least and most highly expressed genes, respectively. Color scale represents the proportion of DMRs in each decile. Gene expression data was obtained from the BLUEPRINT database. (F) Venn diagram of the overlap between P-AID and AID off-target DMRs with super-enhancers in GC B cells. Super-enhancer identification is described in the Materials section utilizing H3K27ac ChIP-seq data from the BLUEPRINT database. (G) Box plots showing H3K27ac ChIP-seq peaks of super-enhancers associated with AID off-target (red) and P-AID (green) DMRs. One-sided unpaired Wilcoxon’s test was used to examine signal intensity differences. (H) DNA methylation and H3K27ac profiles in the vicinity of two representative genes. DMR color indicates the type of DMR, either P-AID (green) or AID off-target DMRs (red). H3K27ac ChIP-seq signals from GC B cells are shown in dark orange. Super-enhancers and enhancers are depicted. Statistical tests: two-tailed unpaired Wilcoxon’s (C) and Fisher’s exact (D) tests (* P-value < 0.05, ** P-value < 0.01; ns is not significant). 5066 Nucleic Acids Research, 2021, Vol. 49, No. 9 from naı̈ve to GC B cells in WT but not in Aicda–/– mice, Some genes that have altered DMRs, such as BATF (Figure termed mouse potential AID targets (mouse P-AID); (ii) 4D) and MEF2A (Supplementary Figure S4E), are crucial CpGs significantly demethylated in naı̈ve-to-GC transition to B cell development (47,51). in both WT and knockout mice and, therefore, more likely To explore the possibility that aberrant DNA methyla- to be TET-dependent, termed positive control; (iii) CpGs tion in HIGM2 naı̈ve B cells may be due to pre-activation that did not present demethylation in the transition from outside the GC, we first analyzed the enrichment for TF naı̈ve to GC B cells neither in WT nor in Aicda–/– mice, binding motifs with HOMER software in HIGM2 naı̈ve termed negative control (differential DNA methylation < DMRs. Some of the most enriched TFs are downstream 0.05) (Supplementary Figure S3B). Next, we merged this of the BCR pathway, including BATF, IRF8 and RELA data with public hydroxy-meDIP-seq data of mouse B cell (32) (Supplementary Figure S4F). We then validated these activation (47) to determine the hydroxymethylation status results through enrichment analysis of the ChIP-seq data of these three groups of CpGs. We found that mouse P- available for GM12878 cells from the ENCODE consor- AID CpGs presented similar levels of hydroxymethylation tium (31) and we again observed enrichment of TFs down- to those that are TET-dependent (positive control) com- stream of BCR activation in DMRs found to be altered in pared to negative control (Supplementary Figure S3C). On naı̈ve B cells of HIGM2 patients (Figure 4E). the other hand, we only found 21 genes associated with The type III latency state of the Epstein-Barr virus mouse P-AID CpGs that overlapped with the aforemen- (EBV) is characterized by the constitutive activation of tioned AID off-target genes (18), which represents <8% the BCR and CD40 pathways (52,53), both of which of the AID-off target genes (Supplementary Figure S3D). are major signaling pathways essential for B cell activa- Taken together, these results suggest that, at least in a mouse tion (54). In that sense, the B lymphoblastoid cell line model, the majority of demethylation associated with B cell GM12878 presents a type III latency state (55) and is differentiation may not be directly driven by AID activity. therefore a good model to investigate the involvement of BCR/CD40 in aberrant DNA methylation of naı̈ve B cells in HIGM2 patients. Firstly, we represented DNA methy- AID deficiency results in premature demethylation of the lation levels of publicly-available data of EBV-transformed BCR pathway of naı̈ve B cells and CD40L/IL-4-activated B cells (36) in regions that over- Our initial analysis suggested that alterations in DNA lapped with DMRs identified in naı̈ve B cells of HIGM2 methylation not only occurred in the transition from naı̈ve patients and observed that EBV transformation effectively to memory B cells, but were already present in naı̈ve B cells reproduced the aberrant DNA demethylation presented in of HIGM2 individuals in comparison with healthy con- HIGM2 patients (Figure 4F). Conversely, such changes trols. Specifically, HIGM2 naı̈ve B cells appeared to be more did not take place when activation was performed with demethylated than those of healthy controls (Figures 1F CD40/IL-4 alone, suggesting that the BCR pathway has a and 4A). AID expression has customarily been associated significant role in driving aberrant DNA methylation (Fig- with the GC reaction (3), however, more recent evidence ure 4F). Furthermore, these DNA methylation effects were suggests that AID may have a role in earlier stages of B cell not reproduced in hypermethylated DMRs (Supplementary development (48,49). Figure S5A). To confirm these results, we transformed naı̈ve Comparing naı̈ve B cells from HIGM2 to control, we de- B cells isolated from healthy donors with EBV and, after 30 tected 2152 hypomethylated DMRs (Figure 4A) and 127 days, we interrogated DNA methylation of the set of CpGs hypermethylated DMRs (Supplementary Figure S4A), ex- that we had previously identified to be aberrantly demethy- cluding PMDs. Both hypo- and hypermethylated DMRs lated in HIGM2. We observed that these CpGs underwent were mostly found outside CpG islands in intergenic regions demethylation following EBV-mediated transformation of and introns (Supplementary Figure S4B). Genetic annota- naı̈ve B cells (Supplementary Figure S5B). Additionally, we tion of DMRs in CpG islands reveals an association with directly stimulated BCR in isolated healthy naı̈ve B cells and promoters (Supplementary Figure S4C and Supplementary observed similar demethylation of the same CpGs, which Table S5). However, while hypomethylation was associated supported the direct role of BCR in their demethylation with enhancer regions, hypermethylation was mainly en- (Figure 4G). Taken together, our findings suggest that al- riched in promoters (Supplementary Figure S4D), in agree- terations in DNA methylation of HIGM2 naı̈ve B cells may ment with its reported regulatory role in the ‘spurious’ ini- at least be partly due to the aberrant pre-activation of the tiation of transcription (50). BCR at some point of B cell development prior to the GC We observed that the DMRs that were hypomethylated reaction. in naı̈ve B cells from HIGM2 patients underwent normal Since ChIP-seq analysis in GM12878 cells identified demethylation during the transition from naı̈ve to memory BATF to bind to almost 50% of regions that correspond cells in healthy controls (Figure 4A). Our results suggest to DMRs in HIGM2 naı̈ve B cells (Figure 4E), we there- that the demethylation of these DMRs may be essential for fore hypothesize that BATF might be the main media- correct naı̈ve-to-memory B cell differentiation, and prema- tor of demethylation in naı̈ve-to-memory transition, and ture demethylation observed in HIGM2 naı̈ve B cells may its aberrant recruitment may drive the aberrant demethy- indicate aberrant pre-activation outside of the GC. This is lation observed in HIGM2 naı̈ve B cells. BATF, together consistent with the finding that genes associated with these with IRF4 (56,57), is a regulator of B and T cell acti- DMRs became upregulated during the activation of B cells vation via the BCR and TCR pathways respectively. We in the GC (Figure 4B) and were associated with functional observed that 87% of the ChIP-seq peaks of IRF4 over- categories related to B cell activation via BCR (Figure 4C). lapped with BATF peaks in GM12878 (Figure 4H). As ex- Nucleic Acids Research, 2021, Vol. 49, No. 9 5067 A B C Abnormal lymphopoiesis -07 Abnormal mononuclear cell differentiation 1.0 p = 5e Abnormal lymphocyte morphology Abnormal adaptive immunity Abnormal leukopoiesis DNA methylation level Immune system process Log10 (FPKM +1) 4 Immune response Regulation of immune system process Leukocyte activation Lymphoid cancer 0.5 Lymphoma Genes involved in immune system 2 Positive regulation of immune system process dsDNA virus infectious disease Lymphoblastic leukemia Genes involved in cytokine signaling in immune system DNA virus infectious disease 0 Genes involved in adaptive immune system 0.0 IL2 signaling events mediated by PI3K ïve ry ïve ry ïve ter BCR signaling pathway Na mo Na mo Na en Me Me na lC B cell activation r mi Oxidative stress response Control HIGM2 Ge T cell activation D 0 Ras pathway 30 60 90 - Log10 (Binominal FDR Q-Value) 1.00 DNA methylation level Disease Ontology Mouse Phenotype 0.75 PANTHER Pathway Biological Process MSigDB Pathway 0.50 0.25 0.00 E C2H2 IRF RHD bZIP 8 ETS POU,Homeobox SRF Log 2 (Fold enrichment of motif) 10kb JUND 6 RELA BATF DMR MEF2A BCL11A MEF2C 4 IRF4 ATF2 BATF NFATC1 POU2F2 H3K27ac EGR1 SPI1 2 Super-enhancer ELK1 Enhancer FOS Control Control HIGM2 HIGM2 0 Naïve Memory Naïve Memory 0 20 40 60 Percent of DMRs containing motif F G 1.00 BATF CpG1 CD58 CpG1 CD80 CpG1 CD58 CpG2 CD80 CpG2 HIF1A CpG2 DNA methylation level 90 * * ns ** ns ns 0.75 80 70 DNA methylation level 60 50 0.50 40 naïve activated HIF1A CpG1 IRF4 CpG1 MEF2A CpG1 IRF4 CpG2 MEF2A CpG2 90 ns ns ** ns ** 0.25 80 70 60 0.00 50 40 ïve ry ïve ry ed L4 V Na mo Na mo lat L/I EB naïve activated naïve activated naïve activated naïve activated naïve activated Me Me mu 40 sti CD Un Control HIGM2 I BATF CD58 CD80 H 125 100 ns * ns * 40 ns ns 10 30 75 20 50 5 25 10 Relative to input BATF 0 0 0 naïve activated naïve activated naïve activated 16959 IgG 15460 BATF HIF1A IRF4 MEF2A 40 IRF4 60 ns ns ns ns ns * 20 30 2262 40 20 10 20 10 0 0 0 naïve activated naïve activated naïve activated Figure 4. Role of AID in early B cell development. (A) Box and violin plot representation of DNA methylation of hypomethylated DMRs identified comparing control and patient naı̈ve B cells. (B) Box plot showing the distribution of expression values of DMR-associated genes in naı̈ve (green) and GC 5068 Nucleic Acids Research, 2021, Vol. 49, No. 9 pected, we confirmed that regions with BATF and IRF4 tioned categories (Supplementary Figure S6B). Utilizing binding had lower DNA methylation levels in naı̈ve B cells DoRothEA TF analysis, we observed that DEGs identified of HIGM2 patients compared to controls (Supplementary in naı̈ve B cells of HIGM2 patients also displayed increased Figure S5C). This is also confirmed for regions that con- TF activity compared to controls (Figure 5C), where sim- tained JUND binding, another TF downstream of BCR ilar results were observed for memory B cells (Supplemen- (Supplementary Figure S5D). This did not occur in re- tary Figure S6C). Integration of the DNA methylation and gions enriched for other B cell-intrinsic TF binding motifs, expression datasets in naı̈ve B cells did not show a clear such as RAD21, ZNF274 and MYC (Supplementary Fig- correlation when comparing HIGM2 versus controls (Fig- ure S5D). Using mRNA transcription data from IRF4 and ure 5D). Interestingly, hypomethylated DMRs in HIGM2 BATF knockouts from GM12878 cells (55), we determined naı̈ve B cells showed slightly increased expression levels in that genes with binding motifs for both TFs displayed ex- HIGM2 naı̈ve B cells in respect to control naı̈ve B cells. pression changes for many of these genes associated with The naı̈ve B cell expression levels of hypomethylated DMRs DMRs (Supplementary Figure S5E). Furthermore, we ob- of HIGM2 patients were more similar to those of memory served that BATF was specifically recruited to genes that B cells, reinforcing the notion of a premature activation of displayed defective DNA demethylation in HIGM2 patients these cells in HIGM2 (Figure 5E). when naı̈ve B cells were activated via BCR stimulation (Fig- ure 4I). Altogether, these results suggest that a significant fraction of DMRs identified in HIGM2 naı̈ve B cells may AID deficiency causes blockade of central B cell tolerance be associated with the recruitment of the BATF/IRF4 com- with an expansion of pre-activated autoreactive B cells plex to these genomic sites. Our results suggest that naı̈ve B cells are pre-activated in AID deficient patients. However, the stage of B cell differ- entiation at which this alteration is produced remains to be The transcriptome of AID-deficient naı̈ve B cells also reflects established. Two independent studies reported a potential alterations in the BCR pathway role for AID in removing autoreactive B cells during the We then inspected whether loss of function of AID had central B cell tolerance process in the bone marrow (48,49). an impact on the transcriptomic profiles of both naive Specifically, immature B cells with auto-reactive BCR were and memory B cells. To this end, we performed RNA- activated and went into a secondary receptor editing pro- seq of both naı̈ve and memory B cells from the same two cess with an increase in AID and recombination-activating aforementioned HIGM2 patients. In parallel, we performed gene 2 (RAG2). If the autoreactive BCR did not lose self- RNAseq with two healthy controls. antigen affinity, the genomic instability induced by the over- Comparison of the naı̈ve B cells transcriptomes revealed exposure to high levels of AID led to apoptosis. However, statistically significant expression changes of a number of AID deficiency reduces the genomic damage that causes the genes in HIGM2 compared to controls (DEGs: differen- expansion of autoreactive B cells (48,49). With that in mind, tially expressed genes). DEGs were considered to be statisti- we hypothesized that the presence of naı̈ve B cells with a cally significant when fold change ≥2 and adjusted P-value pre-activation methylation signature in HIGM2 patients is <0.05. 3223 genes became upregulated in HIGM2 naı̈ve B a consequence of the impairment of the central B cell toler- cells in relation to healthy controls, whereas 3146 genes were ance mechanism that causes autoreactive naı̈ve B cells to ac- downregulated. (Figure 5A and Supplementary Table S6). cumulate. In fact, it has been reported that 21% of HIGM2 We then performed Gene Ontology analysis of DEGs using patients suffer some type of autoimmune disease (58). GREAT software which revealed enrichment of categories To assess this hypothesis, we first checked whether there related to EBV activation and BCR signaling pathway (Fig- is an expansion of naı̈ve autoreactive B cells in HIGM2 ure 5B), consistent with our previous findings related to patients with respect to controls. To this end, we used a DNA methylation alterations in naı̈ve B cells of HIGM2. commercial antibody against 9G4+ IgG used to detect au- Interestingly, we also observed differences in gene expres- toreactive clones in autoimmune diseases like systemic lu- sion in memory B cells (2199 upregulated genes, 2384 down- pus erythematosus and rheumatoid arthritis (59,60). We ob- regulated) (Supplementary Figure S6A and Supplemen- served an expansion of the naı̈ve B cell compartment in tary Table S6) which were also enriched in the aforemen- HIGM2 patients in comparison with healthy controls (Fig- ←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− (red) B cells. Gene expression data obtained from BLUEPRINT database. FPKM refers to fragments per kilobase of transcript per million mapped reads. (C) Results of gene set enrichment analysis using GREAT software. The plot depicts the top five enriched terms for five annotation databases ranked by P-values from the binominal distribution. All the depicted enriched terms passed a significance threshold of adjusted P-value < 0.05. (D) Smoothed DNA methylation data of BATF-associated DMRs in HIGM2 naı̈ve B cells and H3K27ac ChIP-seq signal, super-enhancer and enhancer location in GC B cells. H3K27ac ChIP-seq data obtained from BLUEPRINT database. (E) Bubble scatter plot of transcription factor ChIP-seq peaks determined in GM12878 lymphoblastoid cell line in regions that correspond to DMRs identified in HIGM2 naı̈ve B cells. Transcription factors downstream of BCR signaling are colored according to the transcription factor family. Bubble size corresponds to the logarithm of adjusted P-values. ChIP-seq data were obtained from ENCODE consortium (31). FDR refers to false discovery rate. (F) Box and violin plots summarizing the distribution of DNA methylation of regions that correspond to HIGM2-identified DMRs in resting B cells (Unstimulated), B cells activated with CD40L/IL4 and B cells infected with Epstein- Barr virus (EBV). Data of resting and activated B cells obtained from Hansen et al. (36) (G) Dot plot showing the DNA methylation values determined by pyrosequencing of naı̈ve B cells (naı̈ve) and B cells activated with anti-BCR/CD40L/IL-21 (activated) from healthy donors. (H) Venn diagram representing the overlap between BATF and IRF4 ChIP-seq peaks in the GM12878 cell line. (I) Dot and bar plot representation of BATF ChIP-qPCR results of naı̈ve B cells (naı̈ve) and naı̈ve B cells activated with anti-BCR/CD40L/IL-21 (activated) from healthy donors. Statistical tests: Student’s t-test (G, I) (* P-value < 0.05, ** P-value < 0.01, ns is not significant). Nucleic Acids Research, 2021, Vol. 49, No. 9 5069 Figure 5. RNA sequencing analysis of naı̈ve B cells transcriptomes from HIGM2 patients and healthy donors. (A) Volcano plot representation of the transcriptomic comparison of HIGM2 with control naı̈ve B cells. In blue, genes that are differentially expressed (absolute of fold change ≥2 and adjusted P-value < 0.05). (B) Gene Ontology analysis representation of differentially expressed genes. All represented categories are statistically significantly enriched (adjusted P-value < 0.05). Gene ratio refers to the ratio between the unique genes present in the DMRs that are associated with a Gene Ontology term and all the genes potentially included in that ontology gene set. Dot size represents the number of times each term is represented in the DMRs. Dot color represents adjusted P-value. (C) Heatmap representation showing TF activity predicted using DoRothEA from mRNA expression of differentially expressed genes in control and HIGM2 naı̈ve B cells. Color represents normalized enrichment scores (NES) where blue and red represent a decreased and increased activity, respectively, in comparison to background. (D) Correlation between DNA methylation differences and gene expression scaled counts ratio in naı̈ve B cells from HIGM2 and controls. Colors represent genetic context location. (E) Box and violin plot representation of expression values per sample group from the genes associated with hypomethylated DMRs in HIGM2 naı̈ve B cells. Statistical tests: Spearman’s rank correlation test (D) and two-tailed unpaired Wilcoxon’s (E) (ns is non significant). ure 6A, B). We did not find an expansion of 9G4+ cells in an expansion of pre-activated autoreactive naı̈ve B cells as HIGM2 patients with respect to controls (Figure 6C). How- a consequence of central B cell tolerance impairment medi- ever, we observed a significant increase of mean fluorescence ated by AID deficiency. intensity for 9g4 staining (Figure 6A, D), as well as, an ex- pansion of high 9g4+ naı̈ve B cells (Figure 6E). Next, we DISCUSSION determined the methylation status of a selection of genes by pyrosequencing of 9G4– (non-autoreactive) and 9G4+ Our results show that AID deficiency in HIGM2 syndrome (autoreactive) naı̈ve B cells and found that autoreactive B results in the acquisition of aberrant DNA methylation cells had lower levels of DNA methylation than their non- profiles in naı̈ve and memory B cells. First, we observed autoreactive counterparts (Figure 6F). Overall, our results the presence of HIGM2-associated DNA methylation al- suggest that the demethylation in naı̈ve B cells of HIGM2 terations occurring in the transition from naı̈ve to memory patients compared with control donors is associated with B cell in which the majority are due to DNA replication- 5070 Nucleic Acids Research, 2021, Vol. 49, No. 9 A F BATF CpG 1 CD58 CpG 1 CD58 CpG 2 NBC * * * * PBMCs 100 75 5.05% IgD CD27 9g4 50 Control HIGM2 CD80 CpG 1 CD80 CpG 2 HIF1A CpG 1 * * * * 100 75 NBC IgD DNA methylation level 6.24% 50 CD27 9g4 HIF1A CpG 2 IRF4 CpG 1 IRF4 CpG 2 * * B C * * 100 % of naïve compartment % of B compartment * 10 ns 80 75 70 8 50 60 6 50 4- 4+ MEF2A CpG 1 MEF2A CpG 2 ve 4 9G 9G aï * * * * N Control HIGM2 Control HIGM2 100 D E % of naïve compartment 75 * * Mean fluorescence 400 7 50 intensity 5 300 4- 4- 4+ 4+ ve ve 9G 9G 9G 9G aï aï 3 N 200 N Control HIGM2 Control HIGM2 Figure 6. Central B cell tolerance impairment by AID deficiency. (A) Representative example of strategy for sorting naı̈ve autoreactive B cells. (B) Dot plot showing the percentage of naı̈ve B cells in the B cell compartment of HIGM2 patients and controls. (C) Dot plot showing the percentage of autoreactive naı̈ve B cells in the naı̈ve B cell compartment of HIGM2 patients and controls. (D) Dot plot showing the mean fluorescence of 9g4+ gate. (E) Dot plot showing the percentage of high 9g4+ naı̈ve B cells in the naı̈ve B cell compartment of HIGM2 patients and controls. (F) Dot plot showing the DNA methylation values determined by pyrosequencing of control naı̈ve B cells (naı̈ve), HIGM2 patient 9g4– naı̈ve B cells (9g4– ) and HIGM2 patient 9g4+ naı̈ve B cells (9g4+ ). Statistical tests: student t-test (* P-value < 0.05, ns is not significant). mediated mechanisms. Second, compared to AID mouse addressing the potential participation of AID in demethy- models, we observed that AID may not be directly involved lation had not considered late-replicating domains or the in the small proportion of those DNA methylation alter- relationship between DNA methylation changes and the ations attributable to active demethylation. Third, the com- genomic features of AID targets. In our study, we deter- parison of naı̈ve B cells in HIGM2 and healthy controls mined that most of the changes taking place during the showed premature demethylation of genes downstream of transition from naı̈ve to memory B cells occur through pas- the BCR in AID-deficient individuals, which was associated sive demethylation. It is possible that memory B cells of with the expansion of autoreactive B cell clones prior to the healthy controls undergo more divisions than HIGM2 pa- GC reaction. This reinforces a novel role for AID in pre- tients, similarly to what has been described for Aicda–/– venting the expansion of autoreactive B cell clones, affect- mice (61). This could explain the partial impairment of pas- ing the DNA methylation profiles of naı̈ve B cells. sive demethylation in HIGM2 patients. For active changes, Our study indicates that AID does not play a direct role in no associations with AID targets were found. These find- mediating active demethylation in the transition from naı̈ve ings are also in line with those by Álvarez-Prado et al. (35), to memory B cells. This transition is associated with the which indicated that AID-mediated mutation frequencies highest proportion of DNA methylation changes of the en- are too low. Such low frequency would be unlikely to pro- tire B cell differentiation process (20,21) and also coincides duce a perceptible effect at the level of DNA methylation. with the highest peak of AID expression. Previous studies In relation to the results obtained by Dominguez and col- Nucleic Acids Research, 2021, Vol. 49, No. 9 5071 leagues (13), it is remarkable that even without the removal cannot assume that the pre-activation of autoreactive naı̈ve of PMDs in their analysis, the overlap between the DNA B cells in HIGM2 accounts for all the observed alterations methylation changes and AID-associated dsDNA breaks in genome-wide DNA methylation. It is possible that other was very small, which again reinforces the notion that AID mechanisms may actively contribute to this phenomenon. may not be directly involved in DNA demethylation events Our results, therefore, indicate that the enhanced demethy- of B cell maturation. A limitation of our analysis was the lation of BCR downstream targets in HIGM2 naı̈ve B cells lack of effective human models to study the direct contribu- may be the result of the expansion of autoreactive B cell tion of AID activity to active demethylation. Therefore, we clones as a consequence of the absence of AID. Further- cannot discard the possibility that the use of mouse models more, in order to generalize our findings to all variants of may not be directly attributable to the functions of human HIGM2 it would be important to extend these analyses to AID in the context of HIGM2. However, comparing Aicda- a bigger cohort of patients with different genetic mutations /- mouse model with a mouse model deficient for BER and of the AICDA gene (38), as they can have a different impact MMR pathways (Ung-/- Msh2-/-), we were able to deduce on B cell function. that DNA methylation during B cell maturation, at least in mice, was not dependent on AID enzymatic activity. Fur- thermore, there is a 92% identity between mouse and human DATA AVAILABILITY AID protein sequences, in which the catalytic and hotspot DNA methylation and expression data for this publica- recognition domains are conserved. tion have been deposited in the NCBI Gene Expression Another limitation of our analysis is that we work under Omnibus and are accessible through GEO Series accession the assumption that all deaminase activity is attributed to number GSE172012 (SuperSeries). AID and that no other targets exist apart from AID off- targets and Ig genes. Without further evidence, we cannot effectively conclude that AID enzymatic activity is not es- SUPPLEMENTARY DATA sential to mediate DNA demethylation during B cell mat- Supplementary Data are available at NAR Online. uration. Moreover, AID hotspots were originally identified as genomic locations of SHM (29), but further experiments would be required to verify that they are direct targets of ACKNOWLEDGEMENTS AID deaminase activity in the context of human B cells. We thank CERCA Programme/Generalitat de Catalunya A second major conclusion of our study concerns the and the Josep Carreras Foundation for institutional sup- identification of DNA methylation defects in naı̈ve B cells port. This study makes use of data generated by the from HIGM2 patients in relation to healthy controls. Cus- BLUEPRINT Consortium (Funding for the BLUEPRINT tomarily, AID expression has been regarded as being re- project provided by the European Union’s Seventh Frame- stricted to GC B cells, but some evidence suggests that AID work Programme (FP7/2007–2013) under grant agreement may also have a role in central B cell tolerance (48,49,62). no 282510––BLUEPRINT). A full list of the investigators Along this line, during B cell development, these cells not who contributed to the generation of the data is available only become activated in the GC but also in previous stages from www.blueprint-epigenome.eu. We are also very grate- of differentiation in the bone marrow, where self-reactive ful to Dr Javier di Noia, Dr Almudena Ramiro, Dr Javier immature B cells are activated in a process character- Carmona and Dr Biola M. Javierre for useful feedback. ized by the upregulation of both AID and recombination- activating gene 2 (RAG2) and the downregulation of the anti-apoptotic MCL-1 (63). In this context, AID activity FUNDING increases the probability of genomic damage with the sub- Spanish Ministry of Science, Innovation and Universi- sequent activation of apoptosis through p53, which is also ties [SAF2017-88086-R to E.B.]; cofunded by FEDER enhanced by the inhibition of the anti-apoptotic proteins funds/European Regional Development Fund (ERDF)––a BCL2 and MCL-1 (49). At this point, self-reactive imma- way to build Europe. E.B is supported by Instituto ture B cells that are unable to correct their affinity for de Salud Carlos III (ISCIII), Ref. AC18/00057, asso- self-antigens by receptor editing are eliminated. In patients ciated with i-PAD project (ERARE European Union with AID deficiency, this mechanism of cell removal is im- program); P.L. and C.P. are supported by the German paired and autoreactive cells accumulate (49). Indeed, we Cancer Aid project CO-CLL [70113869]; B.G. is funded noted that HIGM2 patients accumulated more autoreac- by the Deutsche Forschungsgemeinschaft [GR1617/14- tive B cells than healthy donors, a finding that is compati- 1/iPAD, SFB1160/2 B5, RESIST–EXC 2155–Project ID ble with the previously described high frequency of autoim- 390874280, CIBSS–EXC-2189–Project ID 390939984]; mune disorders in this type of patients (58). The failure in BMBF [GAIN 01GM1910A]. Funding for open access AID function in these patients could be responsible for the charge: Spanish Ministry of Science, Innovation and smaller degree of genomic damage that promotes the ex- Universities [SAF2017-88086-R]. pansion of autoreactive naı̈ve B cells. These self-reactive B Conflict of interest statement. None declared. cells, owing to the persistent activation of their BCR dur- ing negative selection in the bone marrow, display a more demethylated profile in genes downstream of the BCR com- REFERENCES pared with non-autoreactive naı̈ve B cells. However, we have 1. Revy,P., Muto,T., Levy,Y., Geissmann,F., Plebani,A., Sanal,O., only replicated this data in a selected number of CpGs and Catalan,N., Forveille,M., Dufourcq-Lagelouse,R., Gennery,A. et al. 5072 Nucleic Acids Research, 2021, Vol. 49, No. 9 (2000) Activation-induced cytidine deaminase (AID) deficiency a continuum of disease phenotypes in chronic lymphocytic leukemia. causes the autosomal recessive form of the hyper-IgM syndrome Nat. Genet., 48, 253–264. (HIGM2). Cell, 102, 565–575. 21. Kulis,M., Merkel,A., Heath,S., Queirós,A.C., Schuyler,R.P., 2. Honjo,T., Kinoshita,K. and Muramatsu,M. (2002) Molecular Castellano,G., Beekman,R., Raineri,E., Esteve,A., Clot,G. et al. mechanism of class switch recombination: linkage with somatic (2015) Whole-genome fingerprint of the DNA methylome during hypermutation. Annu. Rev. Immunol., 20, 165–196. human B cell differentiation. Nat. Genet., 47, 746–756. 3. Muramatsu,M., Kinoshita,K., Fagarasan,S., Yamada,S., Shinkai,Y. 22. Victora,G.D. and Nussenzweig,M.C. (2012) Germinal Centers. Annu. and Honjo,T. (2000) Class switch recombination and hypermutation Rev. Immunol., 12, 117–139. require activation-induced cytidine deaminase (AID), a potential 23. Storb,U. (2014) Why does somatic hypermutation by AID require RNA editing enzyme. Cell, 102, 553–563. transcription of its target genes? In: Advances in Immunology. 4. Arakawa,H., HauschiLd,J. and Buerstedde,J.M. (2002) Requirement Academic Press Inc., Vol. 122, pp. 253–277. of the activation-induced deaminase (AID) gene for immunoglobulin 24. Chiarle,R., Zhang,Y., Frock,R.L., Lewis,S.M., Molinie,B., Ho,Y.J., gene conversion. Science, 295, 1301–1306. Myers,D.R., Choi,V.W., Compagno,M., Malkin,D.J. et al. (2011) 5. Morgan,H.D., Dean,W., Coker,H.A., Reik,W. and Genome-wide translocation sequencing reveals mechanisms of Petersen-Mahrt,S.K. (2004) Activation-induced cytidine deaminase chromosome breaks and rearrangements in B cells. Cell, 147, deaminates 5-methylcytosine in DNA and is expressed in pluripotent 107–119. tissues: Implications for epigenetic reprogramming. J. Biol. Chem., 25. Klein,I.A., Resch,W., Jankovic,M., Oliveira,T., Yamane,A., 279, 52353–52360. Nakahashi,H., Di Virgilio,M., Bothmer,A., Nussenzweig,A., 6. Rai,K., Huggins,I.J., James,S.R., Karpf,A.R., Jones,D.A. and Robbiani,D.F. et al. (2011) Translocation-capture sequencing reveals Cairns,B.R. (2008) DNA demethylation in zebrafish involves the the extent and nature of chromosomal rearrangements in B coupling of a deaminase, a glycosylase, and Gadd45. Cell, 135, lymphocytes. Cell, 147, 95–106. 1201–1212. 26. Seidel,M.G., Kindle,G., Gathmann,B., Quinti,I., Buckland,M., van 7. Bhutani,N., Brady,J.J., Damian,M., Sacco,A., Corbel,S.Y. and Montfrans,J., Scheible,R., Rusch,S., Gasteiger,L.M., Grimbacher,B. Blau,H.M. (2010) Reprogramming towards pluripotency requires et al. (2019) The European Society for Immunodeficiencies (ESID) AID-dependent DNA demethylation. Nature, 463, 1042–1047. registry working definitions for the clinical diagnosis of inborn errors 8. Sabag,O., Zamir,A., Keshet,I., Hecht,M., Ludwig,G., Tabib,A., of immunity. J. Allergy Clin. Immunol. Pract., 7, 1763–1770. Moss,J. and Cedar,H. (2014) Establishment of methylation patterns 27. Wang,Q., Gu,L., Adey,A., Radlwimmer,B., Wang,W., Hovestadt,V., in ES cells. Nat. Struct. Mol. Biol., 21, 110–112. Bähr,M., Wolf,S., Shendure,J., Eils,R. et al. (2013) 9. Muramatsu,M., Sankaranand,V.S., Anant,S., Sugai,M., Tagmentation-based whole-genome bisulfite sequencing. Nat. Kinoshita,K., Davidson,N.O. and Honjo,T. (1999) Specific Protoc., 8, 2022–2032. expression of activation-induced cytidine deaminase (AID), a novel 28. Li,H. and Durbin,R. (2010) Fast and accurate long-read alignment member of the RNA-editing deaminase family in germinal center B with Burrows-Wheeler transform. Bioinformatics, 26, 589–595. cells. J. Biol. Chem., 274, 18470–18476. 29. Liu,H., Liu,X., Zhang,S., Lv,J., Li,S., Shang,S., Jia,S., Wei,Y., 10. Ramiro,A.R. and Barreto,V.M. (2015) Activation-induced cytidine Wang,F., Su,J. et al. (2016) Systematic identification and annotation deaminase and active cytidine demethylation. Trends Biochem. Sci., of human methylation marks based on bisulfite sequencing 40, 172–181. methylomes reveals distinct roles of cell type-specific hypomethylation 11. Fritz,E.L., Rosenberg,B.R., Lay,K., Mihailović,A., Tuschl,T. and in the regulation of cell identity genes. Nucleic Acids Res., 44, 75–94. Papavasiliou,F.N. (2013) A comprehensive analysis of the effects of 30. Whyte,W.A., Orlando,D.A., Hnisz,D., Abraham,B.J., Lin,C.Y., the deaminase AID on the transcriptome and methylome of activated Kagey,M.H., Rahl,P.B., Lee,T.I. and Young,R.A. (2013) Master B cells. Nat. Immunol., 14, 749–755. transcription factors and mediator establish super-enhancers at key 12. Hogenbirk,M.A., Heideman,M.R., Velds,A., van den Berk,P.C.M., cell identity genes. Cell, 153, 307–319. Kerkhoven,R.M., van Steensel,B. and Jacobs,H. (2013) Differential 31. Feingold,E.A., Good,P.J., Guyer,M.S., Kamholz,S., Liefer,L., programming of B cells in AID deficient mice. PLoS One, 8, e69815. Wetterstrand,K., Collins,F.S., Gingeras,T.R., Kampa,D., 13. Dominguez,P.M., Teater,M., Chambwe,N., Kormaksson,M., Sekinger,E.A. et al. (2004) The ENCODE (ENCyclopedia of DNA Redmond,D., Ishii,J., Vuong,B., Chaudhuri,J., Melnick,A., Elements) Project. Science, 306, 636–640. Vasanthakumar,A. et al. (2015) DNA Methylation dynamics of 32. Kandasamy,K., Sujatha Mohan,S., Raju,R., Keerthikumar,S., germinal center B cells are mediated by AID. Cell Rep., 12, Sameer Kumar,G.S., Venugopal,A.K., Telikicherla,D., Navarro,D.J., 2086–2098. Mathivanan,S., Pecquet,C. et al. (2010) NetPath: a public resource of 14. Abdouni,H., King,J.J., Suliman,M., Quinlan,M., Fifield,H. and curated signal transduction pathways. Genome Biol., 11, R3. Larijani,M. (2013) Zebrafish AID is capable of deaminating 33. Pageaud,Y., Plass,C. and Assenov,Y. (2018) Enrichment analysis with methylated deoxycytidines. Nucleic Acids Res., 41, 5457–5468. EpiAnnotator. Bioinformatics, 34, 1781–1783. 15. Wijesinghe,P. and Bhagwat,A.S. (2012) Efficient deamination of 34. McLean,C.Y., Bristor,D., Hiller,M., Clarke,S.L., Schaar,B.T., 5-methylcytosines in DNA by human APOBEC3A, but not by AID Lowe,C.B., Wenger,A.M. and Bejerano,G. (2010) GREAT improves or APOBEC3G. Nucleic Acids Res., 40, 9206–9217. functional interpretation of cis-regulatory regions. Nat. Biotechnol., 16. Larijani,M., Frieder,D., Sonbuchner,T.M., Bransteitter,R., 28, 495–501. Goodman,M.F., Bouhassira,E.E., Scharff,M.D. and Martin,A. 35. Álvarez-Prado,Á.F., Pérez-Durán,P., Pérez-Garcı́a,A., Benguria,A., (2005) Methylation protects cytidines from AID-mediated Torroja,C., de Yébenes,V.G. and Ramiro,A.R. (2018) A broad atlas deamination. Mol. Immunol., 42, 599–604. of somatic hypermutation allows prediction of activation-induced 17. Nabel,C.S., Jia,H., Ye,Y., Shen,L., Goldschmidt,H.L., Stivers,J.T., deaminase targets. J. Exp. Med., 215, 761–771. Zhang,Y. and Kohli,R.M. (2012) AID/APOBEC deaminases 36. Hansen,K.D., Sabunciyan,S., Langmead,B., Nagy,N., Curley,R., disfavor modified cytosines implicated in DNA demethylation. Nat. Klein,G., Klein,E., Salamon,D. and Feinberg,A.P. (2014) Large-scale Chem. Biol., 8, 751–758. hypomethylated blocks associated with Epstein-Barr virus-induced 18. Tahiliani,M., Koh,K.P., Shen,Y., Pastor,W.A., Bandukwala,H., B-cell immortalization. Genome Res., 24, 177–184. Brudno,Y., Agarwal,S., Iyer,L.M., Liu,D.R., Aravind,L. et al. (2009) 37. Garcia-Alonso,L., Iorio,F., Matchan,A., Fonseca,N., Jaaks,P., Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in Peat,G., Pignatelli,M., Falcone,F., Benes,C.H., Dunham,I. et al. mammalian DNA by MLL partner TET1. Science (80-.)., 324, (2018) Transcription factor activities enhance markers of drug 930–935. sensitivity in cancer. Cancer Res., 78, 769–780. 19. Ito,S., Dalessio,A.C., Taranova,O. V., Hong,K., Sowers,L.C. and 38. Durandy,A., Peron,S., Taubenheim,N. and Fischer,A. (2006) Zhang,Y. (2010) Role of tet proteins in 5mC to 5hmC conversion, Activation-induced cytidine deaminase: structure-function ES-cell self-renewal and inner cell mass specification. Nature, 466, relationship as based on the study of mutants. Hum. Mutat., 27, 1129–1133. 1185–1191. 20. Oakes,C.C., Seifert,M., Assenov,Y., Gu,L., Przekopowitz,M., 39. Hudson,T.J., Anderson,W., Aretz,A., Barker,A.D., Bell,C., Ruppert,A.S., Wang,Q., Imbusch,C.D., Serva,A., Koser,S.D. et al. Bernabé,R.R., Bhan,M.K., Calvo,F., Eerola,I., Gerhard,D.S. et al. (2016) DNA methylation dynamics during B cell maturation underlie Nucleic Acids Research, 2021, Vol. 49, No. 9 5073 (2010) International network of cancer genome projects. Nature, 464, 52. Cahir-McFarland,E.D., Carter,K., Rosenwald,A., Giltnane,J.M., 993–998. Henrickson,S.E., Staudt,L.M. and Kieff,E. (2004) Role of NF-kappa 40. Zhou,W., Dinh,H.Q., Ramjan,Z., Weisenberger,D.J., Nicolet,C.M., B in cell survival and transcription of latent membrane protein Shen,H., Laird,P.W. and Berman,B.P. (2018) DNA methylation loss 1-expressing or Epstein-Barr virus latency III-infected cells. J. Virol., in late-replicating domains is linked to mitotic cell division. Nat. 78, 4108–4119. Genet., 50, 591–602. 53. Mancao,C. and Hammerschmidt,W. (2007) Epstein-Barr virus latent 41. Lister,R., Pelizzola,M., Dowen,R.H., Hawkins,R.D., Hon,G., membrane protein 2A is a B-cell receptor mimic and essential for Tonti-Filippini,J., Nery,J.R., Lee,L., Ye,Z., Ngo,Q.-M.M. et al. B-cell survival. Blood, 110, 3715–3721. (2009) Human DNA methylomes at base resolution show widespread 54. Luo,W., Weisel,F. and Shlomchik,M.J. (2018) B cell receptor and epigenomic differences. Nature, 462, 315–322. CD40 signaling are rewired for synergistic induction of the c-Myc 42. Durek,P., Nordström,K., Gasparoni,G., Salhab,A., Kressler,C., de transcription factor in germinal center B cells. Immunity, 48, 313–326. Almeida,M., Bassler,K., Ulas,T., Schmidt,F., Xiong,J. et al. (2016) 55. Ma,Y., Walsh,M.J., Bernhardt,K., Ashbaugh,C.W., Trudeau,S.J., Epigenomic profiling of human CD4+ T cells supports a linear Ashbaugh,I.Y., Jiang,S., Jiang,C., Zhao,B., Root,D.E. et al. (2017) differentiation model and highlights molecular regulators of memory CRISPR/Cas9 screens reveal Epstein-Barr virus-transformed B cell development. Immunity, 45, 1148–1161. host dependency factors. Cell Host Microbe, 21, 580–591. 43. Wu,X. and Zhang,Y. (2017) TET-mediated active DNA 56. Ise,W., Kohyama,M., Schraml,B.U., Zhang,T., Schwer,B., Basu,U., demethylation: mechanism, function and beyond. Nat. Rev. Genet., Alt,F.W., Tang,J., Oltz,E.M., Murphy,T.L. et al. (2011) The 18, 517–534. transcription factor BATF controls the global regulators of 44. Rogozin,I.B. and Kolchanov,N.A. (1992) Somatic hypermutagenesis class-switch recombination in both B cells and T cells. Nat. Immunol., in immunoglobulin genes. II. Influence of neighbouring base 12, 536–543. sequences on mutagenesis. BBA - Gene Struct. Expr., 171, 11–18. 57. Ochiai,K., Maienschein-Cline,M., Simonetti,G., Chen,J., 45. Qian,J., Wang,Q., Dose,M., Pruett,N., Kieffer-Kwon,K.R., Resch,W., Rosenthal,R., Brink,R., Chong,A.S., Klein,U., Dinner,A.R., Liang,G., Tang,Z., Mathé,E., Benner,C. et al. (2014) B cell Singh,H. et al. (2013) Transcriptional regulation of germinal center B super-enhancers and regulatory clusters recruit AID tumorigenic and plasma cell fates by dynamical control of IRF4. Immunity, 38, activity. Cell, 159, 1524–1537. 918–929. 46. Meng,F.L., Du,Z., Federation,A., Hu,J., Wang,Q., 58. Quartier,P., Bustamante,J., Sanal,O., Plebani,A., Debré,M., Kieffer-Kwon,K.R., Meyers,R.M., Amor,C., Wasserman,C.R., Deville,A., Litzman,J., Levy,J., Fermand,J.P., Lane,P. et al. (2004) Neuberg,D. et al. (2014) Convergent transcription at intragenic Clinical, immunologic and genetic analysis of 29 patients with super-enhancers targets AID-initiated genomic instability. Cell, 159, autosomal recessive hyper-IgM syndrome due to activation-induced 1538–1548. cytidine deaminase deficiency. Clin. Immunol., 110, 22–29. 47. Lio,C.W.J., Shukla,V., Samaniego-Castruita,D., González-Avalos,E., 59. Richardson,C., Chida,A.S., Adlowitz,D., Silver,L., Fox,E., Chakraborty,A., Yue,X., Schatz,D.G., Ay,F. and Rao,A. (2019) TET Jenks,S.A., Palmer,E., Wang,Y., Heimburg-Molinaro,J., Li,Q.-Z. enzymes augment activation-induced deaminase (AID) expression via et al. (2013) Molecular basis of 9G4 B cell autoreactivity in human 5-hydroxymethylcytosine modifications at the Aicda superenhancer. systemic Lupus erythematosus. J. Immunol., 191, 4926–4939. Sci. Immunol., 4, eaau7523. 60. Moura,R.A., de la Torre,I., Leandro,M.J., Edwards,J.C.W., 48. Kuraoka,M., Holl,T.M., Liao,D., Womble,M., Cain,D.W., Fonseca,J.E. and Cambridge,G. (2012) 9G4 expression on Reynolds,A.E., Kelsoe,G., Kuraoka,M., Cain,D.W., Womble,M. autoantibodies to citrullinated peptides in patients with early et al. (2011) Activation-induced cytidine deaminase mediates central inflammatory arthritis and established rheumatoid arthritis. Ann. tolerance in B cells. Proc. Natl. Acad. Sci. U.S.A., 108, 11560–11565. Rheum. Dis., 71, A1–A93. 49. Cantaert,T., Schickel,J.-N.N., Bannock,J.M.M., Ng,Y.-S.S., 61. Zaheen,A., Boulianne,B., Parsa,J.Y., Ramachandran,S., Massad,C., Oe,T., Wu,R., Lavoie,A., Walter,J.E.E., Gommerman,J.L. and Martin,A. (2009) AID constrains germinal Notarangelo,L.D.D. et al. (2015) Activation-induced cytidine center size by rendering B cells susceptible to apoptosis. Blood, 114, deaminase expression in human B cell precursors is essential for 547–554. central B cell tolerance. Immunity, 43, 884–895. 62. Meyers,G., Ng,Y.-S., Bannock,J.M., Lavoie,A., Walter,J.E., 50. Neri,F., Rapelli,S., Krepelova,A., Incarnato,D., Parlato,C., Basile,G., Notarangelo,L.D., Kilic,S.S., Aksu,G., Debre,M., Rieux-Laucat,F. Maldotti,M., Anselmi,F. and Oliviero,S. (2017) Intragenic DNA et al. (2011) Activation-induced cytidine deaminase (AID) is required methylation prevents spurious transcription initiation. Nature, 543, for B-cell tolerance in humans. Proc. Natl. Acad. Sci. U.S.A., 108, 72–77. 11554–11559. 51. Herglotz,J., Unrau,L., Hauschildt,F., Fischer,M., Kriebitzsch,N., 63. Nemazee,D. (2006) Receptor editing in lymphocyte development and Alawi,M., Indenbirken,D., Spohn,M., Müller,U., Ziegler,M. et al. central tolerance. Nat. Rev. Immunol., 6, 728–740. (2016) Essential control of early B-cell development by Mef2 transcription factors. Blood, 127, 572–581.