ABSTRACT
Mutations in activation induced deaminase (AID) lead to hyper-IgM syndrome type 2 (HIGM2), a rare human primary antibody deficiency. AID-mediated cytosine deamination has been proposed as mediating active demethylation, although evidences both support and cast doubt on such a role. We here made use of HIGM2 B cells to investigate direct AID involvement in active DNA demethylation. HIGM2 naïve and memory B cells both display widespread DNA methylation defects, of which approximately 25% of these defects correspond to active events. For genes that undergo active demethylation that is impaired in HIGM2 individuals, we did not observe AID involvement but a participation of TET enzymes. DNA methylation alterations in HIGM2 naïve B cells are related to premature overstimulation of the B-cell receptor prior to the germinal center reaction. Our data supports a role for AID in B cell central tolerance in preventing the expansion of autoreactive cell clones, affecting the correct establishment of DNA methylation patterns.
INTRODUCTION
Hyper-IgM syndrome type 2 (HIGM2) is a rare primary antibody deficiency characterized by loss-of-function mutations in activation-induced deaminase (AID)1, an enzyme required for several crucial steps of B cell terminal differentiation. AID converts deoxycytosines (dCs) into deoxyuracils (dUs), producing dU:dG mismatches that are removed by mismatch repair and base-excision repair2–6. Deaminase activity is required for somatic hypermutation (SHM) and class-switch recombination (CSR) of immunoglobulin genes, which are necessary processes for affinity maturation and antibody diversification within the germinal centers1,7,8. AID deficiency results in the absence of CSR and SHM, and leads to lymphoid hyperplasia1. HIGM2 patients have normal or elevated serum IgM levels with severe reduction of IgG, IgA, and IgE, resulting in considerable susceptibility to bacterial infections1.
In addition to its role in CSR and SHM, AID has been proposed as participating in active DNA demethylation through deamination of 5-methylcytosine, leading to a mismatch that is converted to G:C by thymine DNA glycosylase (TDG), followed by base-excision repair. During the past decade, conflicting reports have both supported and discounted such a role for AID (reviewed in9). For instance, in three independent studies using B cells from AID-deficient mice, two reported the absence of DNA methylation changes10,11, whereas the third study found DNA methylation differences relative to wild type mice12. However, methodological aspects could explain the discrepancies between these studies. In parallel, the discovery of alternative enzymatic pathways that lead to bona fide active DNA demethylation through ten-eleven translocation methylcytosine dioxygenase (TET)-mediated oxidation of methylcytosines13–15 raised more doubts about the possibility that AID redundantly plays such a role. There is currently no consensus about whether AID is involved in mediating DNA demethylation in specific cell contexts.
Whole-genome analysis has shown the occurrence of a vast amount of demethylation associated with B cell differentiation. Changes occur mostly during naïve B cell activation, yielding memory B cells16,17 that coincide with the highest peak of AID expression1,7,18. Naïve B cells start to proliferate upon activation by antigen encounter. Then they express AID which triggers the secondary diversification of antibodies by SHM and CSR. This is followed by affinity maturation which finally leads to a) a new cycle of SHM or b) terminal differentiation into memory or plasma B cells depending on the affinity of the B cell receptor (BCR) for the cognate antigen19.
In this study, we took advantage of the exceptional possibility to address the direct role of AID in active demethylation by comparing the complete DNA methylomes of naïve and memory B cells of HIGM2 patients with those of healthy individuals. By studying two sibling patients with a homozygous mutation for AID that results in a severely truncated enzyme we were able to determine its direct link with DNA methylation defects and infer its catalytic activity in relation to active DNA demethylation.
Our results show that the absence of AID catalytic activity affects DNA methylation in naïve and memory B cells. The majority of the changes observed in the transition from naïve to memory B cells arise from passive demethylation and are linked to late-replicating domains. However, for those potentially associated with active demethylation, we found no evidence of a direct involvement of AID, and our analysis indicates that TET enzymes are responsible for DNA methylation changes in this cell context. The increased DNA demethylation noted in naïve B cells of HIGM2 patients is associated with a premature demethylation of BCR downstream genes prior to the germinal center reaction. Indeed, we found that these changes are related to the expansion of autoreactive clones, which suggests a major role for AID in preventing the expansion of such clones under normal conditions.
RESULTS
Study strategy
We obtained peripheral blood from two HIGM2 sibling patients, both with the same homozygous mutation for AID, and two healthy controls. Specifically, the patients carried a deletion (Exon 2 c.22_40del19) that generated a frameshift variant (p.Arg8Asnfs*19) that affects the majority of AID, including its catalytic domain. We inspected the peripheral B cell compartment by flow cytometry. As previously described, HIGM2 patients are characterized by the absence of class-switched memory B cells (CD19+CD27+IgM−IgD−, csMBC)1. Nevertheless, classic non-class-switched memory B cells (CD19+CD27+IgM+IgD+, ncsMBC) and naïve B cells (CD19+CD27−IgM+IgD+, NBC) are present in patients (Fig. 1a). Under physiological conditions, ncsMBC cells display certain levels of SHM at the immunoglobulin locus16, which supports the expression of AID during their maturation in germinal centers1,7,18. Therefore, the comparison between the DNA methylation profiles from NBC and ncsMBC of healthy and HIGM2 individuals are an adequate model for testing the potential role of AID in demethylation.
HIGM2 patients display an aberrant methylation profile in naïve and unswitched memory B cells
We performed tagmentation-based whole-genome bisulfite sequencing (T-WGBS), a version of the WGBS method that allows analysis of limited DNA amounts20, for two biological replicates of each of the two aforementioned B cell subsets of the HIGM2 and of healthy controls (from now on referred to as “naïve” and “memory” cells) (Fig. 1b). Pearson correlation and t-distribution stochastic neighbor embedding (t-SNE) between samples was highly reproducible between replicates (correlation coefficient > 0.9, Supplementary Fig. 1a,b). We also compared our methylation data from healthy controls with public data from the International Cancer Genome Consortium (ICGC)21 and Oakes et al.16, thereby confirming the robustness of our data (Supplementary Fig. 1c,d).
Global inspection of DNA methylation confirmed that, as reported16,17, transition from naïve to memory B cells is accompanied by global demethylation of the genome (Fig. 1c). However, the same comparison in HIGM2 patients showed a partial impairment of demethylation during B cell differentiation (Fig 1c,d), compatible with a potential role of AID as a demethylating enzyme. Unexpectedly, we also observed that naïve B cells were more demethylated in HIGM2 patients than in healthy controls (Fig. 1e). Taken together, these global observations suggest that AID loss not only affects the DNA methylation patterns in the transition from naïve to memory B cells, but also has a significant role in establishing the B cell methylome in earlier stages of development.
A high proportion of DNA demethylation events identified in HIGM2 are due to passive demethylation of late-replicating domains
Recent studies have shown that a high proportion of the demethylation occurring in cancer and in differentiation processes, are associated with high proliferation rates. Such demethylation takes place in regions known as ‘partially methylated domains’ (PMDs), rather than ‘highly methylated domains (HMDs). PMDs are characterized by late replication, and their demethylation is a passive event, as a result of inefficient DNA remethylation during DNA replication22–27. Recent reanalysis of the B cell lineage DNA methylation profiles published by the BLUEPRINT consortium17 has shown the occurrence of demethylation of PMDs in the transition towards memory cell and antibody-secreting plasma cells28. This highlights how critical it is to separate the analysis of DNA methylation changes produced inside and outside PMDs when examining the occurrence of active demethylation processes to exclude those changes due to passive demethylation.
To address this matter, we examined the overlap of the differentially methylated regions (DMRs) corresponding to all comparisons with the PMDs and HMDs obtained by Zhou et al.23. We found that the majority of DMRs overlap with PMDs (72.5%, Fig. 2a). DMRs were therefore classified into two sets: the first comprised those DMRs that coincided with common PMDs (PMD-DMRs), and the second contained all the other DMRs (non-PMD-DMRs). Global inspection of the methylation values showed that the two groups of DMRs had intermediate methylation values in memory B cells, although non-PMD-DMRs had lower methylation levels (Fig. 2b), suggesting the presence of active demethylation events.
Next, we analyzed the functional genomic features of the two groups of DMRs to confirm whether it is appropriate to use the PMD/HMD annotation with our data. We found that the PMD-DMRs were enriched in intergenic regions (Fig. 2c). Considering the connection between the two groups of DMRs and the chromatin states, we found that DMRs occurring in PMDs were mainly associated with heterochromatic regions, while non-PMD-DMRs were highly enriched at enhancers and active promoters, which were mainly associated with active demethylation29–36 (Fig. 2d,e).
PMDs are characterized by their association with late-replication domains22,23. To test this property in our data, we divided the GM12878 Repli-seq data into deciles and measured the percentage of DMRs in each category. We confirmed that, indeed, the DMRs annotated as PMDs were mainly found in late-replication regions (Fig. 2f) and were accompanied by lower expression levels of associated genes in germinal center B cells (Fig. 2g). Taking all these observations into account, our results indicated that most of the DNA methylation changes in all the comparisons occur in PMDs. However, the existence of a set of non-PMD-DMRs (~27%) located in highly active regions suggests the potential participation of active demethylation events, which could now be interrogated for the potential direct participation of AID.
HIGM2-associated defects in DNA methylation in the transition from naïve to memory B cells do not have the features of AID targets
Given all the previous considerations, including the removal of DNA methylation changes related to DNA replication (PMD-DMRs), our model allows us to examine whether AID has a direct role in mediating demethylation in the B cell lineage. In this context, germinal center B cells in the transition from naïve to memory, displayed the highest levels of AID mRNA expression (Supplementary Fig. 2a)1,7,37. This makes the comparison between naïve and memory B cells the most suitable to explore the direct role of AID in active demethylation.
To this end, we selected those DMRs with methylation dynamics consistent with a potential demethylation mediated by AID (P-AID DMRs). These are defined as DMRs that are demethylated in the transition from naïve to memory B cells, without such changes in the HIGM2 patients and having similar methylation levels in naïve cells of controls and HIGM2 patients (Fig. 3a). A total of 522 DMRs (containing 450 different genes) fulfilled these conditions.
We first investigated the overlap between the genes contained in the P-AID DMRs and 271 described off-target AID genes (AID targets outside the locus of immunoglobulins)38 (add PMID: 17485517; PMID: 9751748; PMID: 11460166; PMID: 18273020) and found a low correspondence (Fig. 3b). We also tested the presence of described AID hot spots39 and found a significant increase for the WRCY hot spot with respect to the background, although the increase seemed too slight to be of biological relevance (Fig. 3c). On the other hand, we found that DMRs associated with off-target AID genes were normally demethylated in HIGM2 patients (Supplementary Fig. 2b).
Two recent studies have characterized the genomic and epigenomic features of AID off-target regions. Independently, they found that AID targets regions with convergent transcription from intragenic super-enhancers40,41. In this sense, although there were no significant differences regarding gene localization between the DMRs associated with AID off-targets and P-AID DMRs (Supplementary Fig. 2c), the former exhibited greater enrichment of enhancer regions (Fig. 3d and Supplementary Fig. 2d) that was associated with more transcriptional activity of associated genes in germinal center B cells (Fig. 3e). We also found that, although the two DMR groups had a similar percentage of overlap with super-enhancers (P-AID 21%, AID off-target 33.3%; Fig. 3f), the super-enhancers of the AID off-targets had a stronger signal for H3K27ac (Fig. 3g,h and Supplementary Fig. 2e) and greater transcriptional activity of their associated genes than P-AID (Supplementary Fig. S2f). Finally, we hypothesized that if AID had a role mediating active DNA demethylation, we would expect to see differences in CpG sites in a WRCY context at the super-enhancers of AID off-target genes. However, no such differences were observed (Supplementary Fig. 2g). Taken together, our results suggest that, despite the differences in DNA methylation associated with B cell activation between wild type and AID-deficient B cells, such demethylation is not directly associated with AID catalytic activity.
Demethylation during activation of B cells involves TET family proteins
Our findings ruled out a role for AID in directly mediating DNA demethylation in human B cells and contradicted those of a previous study addressing this hypothesis in mouse B cells. However, the conclusions of that study were based on there having been a significant enrichment with SHM target genes and AID-associated dsDNA breaks (as was also the case in the present study), but with an overlap between the two of less than 10%12.
While the removal of methyl groups of cytosines mediated by TET enzymes involves the generation of oxidation intermediates14,31, the proposed mechanism of DNA demethylation by AID implies that the 5mC conversion in a thymine could be repaired and replaced with an unmethylated cytosine9,42 (Fig. 4a). Given these considerations, we hypothesized that if AID deaminates 5mC and yields thymine in mouse B cells at the population level, AID-dependent demethylation events would be associated with lower levels 5mhC than those that are TET-dependent. To address this possibility, we merged the methylation data of Dominguez and colleagues12 with public hydroxy-meDIP-seq data of mouse B cell activation43. We selected a set of CpGs demethylated in wt but not in AID−/− mice (mouse potential AID targets, P-AID), a second CpG set that is significantly demethylated under both conditions and more likely to be TET-dependent (positive control), and a third group of CpGs without demethylation (negative control; delta DNA methylation < 0.05) (Fig. 4b). We found that mouse P-AID CpGs overlapped little with the described AID off-target genes (< 8%, Fig. 4c). Next, we determined the hydroxymethylation status of the three groups of CpGs, and found that mouse P-AID CpGs presented similar levels of hydroxymethylation to those that are TET-dependent (positive control), and significantly higher levels than CpGs selected as negative controls (Fig. 4d). Taken together, these results indicate that demethylation changes occurring during B cell maturation are mediated by TET proteins, and if AID has a role it should be only in an indirect manner.
AID deficiency results in premature demethylation of the BCR pathway of naïve B cells
Our initial analysis suggested that alterations occur in DNA methylation in naïve cells of HIGM2 individuals in comparison with healthy controls. Specifically, HIGM2 naïve B cells appeared to be more demethylated than those of healthy controls. AID expression has customarily been associated with the germinal center reaction1,7,18. However, more recent evidence suggests that AID might have a role in earlier stages of B cell development44–46.
Comparing HIGM2 and control naïve B cells, we detected 2152 hypomethylated DMRs (Fig. 5a) and 127 hypermethylated DMRs (Supplementary Fig. 3a), PMDs having been excluded. Both groups of DMRs were mostly found in intergenic regions and introns (Supplementary Fig. 3b). However, while hypomethylation was associated with enhancer regions, hypermethylation was mainly enriched in promoters (Supplementary Fig. 3c), in agreement with its reported regulatory role in the ‘spurious’ initiation of transcription47.
We observed that DMRs in this comparison, i.e. differentially methylated between naïve cells of HIGM2 patients and healthy controls, underwent a similar change in DNA methylation during the transition from naïve to memory cells in controls and HIGM2 patients, suggesting that the DNA methylation alterations in HIGM2 naïve cells correspond to changes that occur later in differentiation, as if these cells were pre-activated outside the germinal center (Fig. 5a). This is consistent with the finding that genes associated with these DMRs became upregulated during the activation of naïve to memory cells in the germinal center (Fig. 5b) and were associated with functional categories related to B cell activation via BCR (Fig. 5c). Some genes that have altered DMRs, such as BATF (Fig. 5d)43,48,49 or MEF2A (Supplementary Fig. 3d)50–52, are crucial to B cell development.
To explore the possibility that the changes between naïve B cells of HIGM2 patients and controls are due to pre-activation outside the germinal center, we tested the enrichment for transcription factor motifs in the DMRs. Some of the most enriched TFs are downstream of the BCR pathway53 (Supplementary Fig. 3e). We then validated these results through enrichment analysis of the ChIP-seq data available for GM12878 cells from the ENCODE consortium54 and found them to be in accordance with those of the motif enrichment analysis and showed enrichment of downstream TFs of the BCR (Fig. 5e). In addition, some of these TFs were associated with altered DMRs such as BATF, MEF2A, NFATC1, BCL11A and IRF4.
The type III latency state of the Epstein-Barr virus (EBV) is characterized by the constitutive activation of the BCR and CD40 pathways55–59, both of which are major signaling pathways that function during B cell activation60. In this sense, the B lymphoblastoid cell line GM12878 presents a type III latency state61 and is therefore a good model for testing if the changes at the altered DMRs are indeed produced via BCR/CD40. To test this hypothesis, we checked the methylation status of altered DMRs in public DNA methylation data of EBV-transformed B cells and CD40L/IL-4-activated B cells24. We observed that EBV transformation effectively reproduced methylation changes of the altered DMRs. Conversely, such changes did not take place with CD40/IL-4 activation, suggesting that the BCR pathway has a significant role (Fig. 5f and Supplementary Fig. 3f). We confirmed these results by transforming naïve B cells with EBV. After 30 days, we used pyrosequencing to test a selection of genes that were aberrantly demethylated in HIGM2 naïve B cells. We found that these genes underwent demethylation following EBV-mediated transformation of naïve B cells (Fig. 5g and Supplementary Fig. 3g). Taken together, our findings showed that the changes between HIGM2 naïve B cells and those of the controls are due to the aberrant pre-activation of the BCR at some point of B cell development prior to the germinal center reaction.
The analysis of motif enrichment and ChIP-seq data in GM12878 indicated that BATF is one of the main potential mediator of demethylation by TET proteins via their recruitment in the generation of altered DMRs in HIGM2 naïve B cells (~50% overlap with the DMRs). BATF is a regulator of B and T cell activation (BCR and TCR pathways, respectively), in cooperation with IRF448,62,63. We observed that 87% of the ChIP-seq peaks of IRF4 overlapped with BATF peaks in GM12878 (Fig. 6a). As expected, we were able to confirm that regions with BATF and IRF4 binding had lower DNA methylation levels in naïve cells of HIGM2 patients than in controls (Fig. 6b), as well as JUND, another TF downstream of BCR (Supplementary Fig 4a). This did not occur in regions enriched for other B cell-intrinsic TF binding motifs (Supplementary Fig. 4a). Using mRNA transcription data from IRF4 and BATF knockouts from GM12878 cells61, we determined that genes with binding motifs for both TFs displayed expression changes for many of these genes associated with altered DMRs (Fig. 6c). All these results suggest that a significant fraction of the altered DMRs in HIGM2 naïve B cells may be associated with the recruitment of the BATF/IRF4 complex to these genomic sites. Such recruitment might facilitate TET protein-mediated hydroxymethylation and the consequent demethylation of those regions, as recently reported43.
AID deficiency causes blockade of central B cell tolerance with an expansion of pre-activated autoreactive B cells
Our results suggest that naïve B cells are pre-activated in AID deficient patients. However, the stage of B cell differentiation at which this alteration is produced remains to be established. Two independent studies reported a potential role for AID in removing autoreactive B cells during the central B cell tolerance process in bone marrow. Specifically, the immature B cells with auto-reactive BCR were activated and went into a secondary receptor editing process with an increase in AID and recombination-activating gene 2 (RAG2). However, if the autoreactive BCR did not lose self-antigen affinity the genomic instability induced by the overexposure to high levels of AID led to apoptosis. However, AID deficiency reduces the genomic damage that causes expansion of autoreactive B cells44,45. With that in mind, we hypothesized that the presence of naïve B cells with a pre-activation methylation signature in HIGM2 patients is a consequence of the impairment of central B cell tolerance that causes autoreactive naïve B cells to accumulate. In fact, it has been reported that 21% of HIGM2 patients suffer some kind of autoimmune disease64.
To assess this hypothesis, we first checked whether there is an expansion of naïve autoreactive B cells in HIGM2 patients with respect to controls. To this end, we used a commercial antibody against 9G4+ IgG used to detect autoreactive clones in autoimmune diseases like systemic lupus erythematosus and rheumatoid arthritis65–68. We observed an expansion the naïve B cell compartment in HIGM2 patients in comparison with healthy controls (Fig. 7a,b). We did not found an expansion of 9G4+ in HIGM2 patients with respect to controls (Fig. 7c). However, we observed a significant increase of mean fluorescence intensity for 9g4 staining (Fig. 7a,d), as well as, an expansion of high 9g4+ naïve B cells (Fig. 7e). Next, we determined the methylation status of a selection of genes by pyrosequencing of 9G4− (non-autoreactive) and 9G4+ (autoreactive) naïve B cells, and found that autoreactive B cells had lower levels of DNA methylation than their non-autoreactive counterparts (Fig. 7f). Overall, our results suggest that the demethylation in naïve B cells of HIGM2 patients compared with control donors is associated with an expansion of pre-activated autoreactive naïve B cells as a consequence of central B cell tolerance impairment mediated by AID deficiency.
DISCUSSION
Our results show that AID deficiency in HIGM2 syndrome results in the acquisition of aberrant DNA methylation profiles in naïve and memory B cells. Two major conclusions emerge from the study of this phenomenon. First, the analysis of the HIGM2-associated alterations occurring in the transition from naïve to memory B cell rules out the direct involvement of AID in active demethylation. Second, the comparison of naïve B cells in HIGM2 and healthy controls shows premature demethylation of genes downstream of the BCR in AID-deficient individuals, whichis associated with the expansion of autoreactive B cell clones, prior to the germinal center reaction. This reinforces a novel role for AID in preventing the expansion of autoreactive B cell clones, affecting the DNA methylation profiles of naïve B cells.
Our study unequivocally demonstrates that AID does not play a direct role through its catalytic activity in mediating active demethylation in the transition from naïve to memory B cells. This transition is that associated with the highest proportion of DNA methylation changes of the entire B cell differentiation process16,17 and also coincides with the highest peak of AID expression. Previous studies addressing the potential participation of AID in demethylation had considered late-replicating domains or the relationship between DNA methylation changes and the genomic features of AID targets. In our study, we determined that most of the changes taking place during the transition from naïve to memory B cells occur through passive demethylation. No associations with AID targets were found for the other changes. These findings are also in line with those of Álvarez-Prado and colleagues38, who have indicated that AID-mediated mutation frequencies are too low, In this sense, such low frequency but we unlikely to produce a perceptible effect at the level of DNA methylation.
A second major conclusion of our study concerns the identification of DNA methylation defects in naïve B cells from HIGM2 patients in relation to healthy controls. Customarily, AID expression has been regarded as being restricted to germinal center B cells, but some evidence suggests that AID may also have a role in central B cell tolerance44,45,69. In keeping with this, during B cell development, these cells not only become activated in the germinal center but also in previous stages of differentiation in the bone marrow. In that location, self-reactive immature B cells are activated in a process characterized by the upregulation of both AID and recombination-activating gene 2 (RAG2) and the downregulation of the anti-apoptotic MCL-170–72. In this context, AID activity increases the probability of genomic damage with the subsequent activation of apoptosis through p53, which is also enhanced by the inhibition of the anti-apoptotic proteins BCL2 and MCL-173. At that point, self-reactive immature B cells that are unable to correct their affinity for self-antigens by receptor editing are eliminated. In patients with AID deficiency, this mechanism of cell removal is impaired and autoreactive cells accumulate44. Indeed, we noted an accumulation of autoreactive B cells in HIGM2 patients than in healthy donors, a finding that is compatible with the previously described high frequency of autoimmune disorders in this type of patient74. The failure in AID function in these patients could be responsible for the smaller degree of genomic damage that promotes the expansion of autoreactive naïve B cells. These self-reactive B cells, owing to the persistent activation of their BCR during negative selection in the bone marrow, display a more demethylated profile in genes downstream of the BCR compared with non-autoreactive naïve B cells. Our results therefore indicate that the enhanced demethylation of BCR downstream targets in HIGM2 naïve B cells may be the result of the expansion of autoreactive B cell clones as a consequence of the absence of AID. Our results are also consistent with the recent observation that class-switch recombination occurs infrequently in germinal centers 75.
METHODS
Human samples
Patients who fulfilled the diagnostic criteria for hyper-IgM syndrome type 2 were included in the study based on ESID clinical diagnostic criteria76 and genetic confirmation of AICDA mutation and exclusion of other primary and secondary causes of immunodeficiencies. Samples come from the Medical Center of the University Hospital, University of Freiburg, Freiburg, Germany and Hospital Universitari Vall d’Hebron, Barcelona, Spain. The Committees for Human Subjects of the local hospitals approved the study, which was conducted in accordance with the ethical guidelines of the 1975 Declaration of Helsinki. All samples were in compliance with the guidelines approved by the local ethics committee and all donors (and/or their parents) received oral and written information about the possibility that their blood would be used for research purposes.
Isolation of B cell populations
Peripheral blood mononuclear cells (PBMCs) were obtained from blood. After Ficoll-Isopaque density centrifugation (Rafer, Zaragoza, Spain), collected cells were washed twice with ice-cold PBS, followed by centrifugation at 2000 rpm for 5 min. Next, cells were labeled with antibodies to CD19 – FITC (Miltenyi Biotec, clone LT19), CD27 – APC (Miltenyi Biotec, clone M-T271), IgD – PE (SouthernBiotech, Cat. No. 2032-09) and IgM – PerCP/Cy5.5 (BioLegend, clone MHM-88) for 20 min on ice in staining buffer (PBS with 4% FBS and 2 mM EDTA). Naïve B cells (CD19+ CD27− IgD+) and unswitched memory B cells (CD19+ CD27+ IgD+) were obtained by FACS sorting on a MoFlo Astrios (Beckman Coulter). Purified samples were pelleted and stored at −80°C.
For isolation of naïve autoreactive B cells. Total B cells were isolated from PBMCs using positive selection with MACS CD19 microbeads (Miltenyi Biotec). Next, cells were stained with CD27-APC (Miltenyi Biotec, clone M-T271), IgD – PE (SouthernBiotech, Cat. No. 2032-09), HLA-DR – PE-Cy7 (eBioscience, clone LN3), 9g4 primary ab (igm Bioscience) and donkey ant-rat IgG (H + L) – Alexa Fluor 488 (invitrogen). 9g4+ naïve B cells (CD27− IgD+ 9g4+) and 9g4-naïve B cells (CD27− IgD+ 9g4−) were obtained by FACS sorting on a BD FACSAria II (BD Biosciences). Purified samples were pelleted and stored at −80°C.
Genomic DNA extraction
For whole-genome bisulfite sequencing, DNA was extracted with a QIAamp DNA micro kit (Qiagen) according to the manufacturer’s protocol. For pyrosequencing experiments, DNA was extracted with a Maxwell RSC Cultured Cells DNA kit (Promega).
Tagmentation-based whole-genome bisulfite sequencing
For whole-genome bisulfite sequencing, 30 ng of genomic DNA was used to produce four independent barcoded sequencing libraries per DNA sample using the tagmentation method20. Sequencing of the TWGBS libraries was done on a HiSeq 2000, PE 125 bp mode. Bisulfite sequencing reads were processed by the DKFZ bisulfite analysis workflow. In brief, the reads were trimmed using Trimmomatic, pre-processed and aligned using MethylCTools, with default parameters (V. Hovestadt, S. Picelli, B. Radlwimmer, M.Z. and P.L., unpublished data), which uses the Burrows-Wheeler alignment algorithm77. Following quality control of bisulfite conversion (>99.5% in all samples) and of read-mapping (80-90% could be mapped on average), we performed methylation calling using methylCtools. A summary of the sequencing data for each sample is provided in Supplementary Table 1
DMR calling
Differentially methylated regions were detected with the DeNovoDMR algorithm included in the Specific Methylation Analysis and Report Tool (SMART2)78 using all the default parameters except for the segment CpG number threshold, which was set to 4, the 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 construct of the SMART input matrix. DMR calling was performed or all possible comparisons between naïve and memory B cells for both control and HIGM2 patients.
Bisulfite pyrosequencing
500 ng of genomic DNA was converted with an EZ DNA Methylation-Gold kit (Zymo Research), following the manufacturer’s instructions. Bisulfite-treated DNA was PCR amplified using primers (see Supplementary Table 2) designed with PyroMark Assay Design 2.0 software (Qiagen). Finally, PCR amplicons were pyrosequenced with the PyroMark Q24 system and analyzed with PyroMark CpG software (Qiagen).
ChIP-seq data processing
Sequencing reads from ChIP-seq experiments from the BLUEPRINT consortium79 were mapped to the hg19 assembly of human reference genome using Burrows-Wheeler Aligner (BWA) v0.7.1377 (with parameters -q 5, -l 32, -k 2). After removing reads with MAPQ < 30 with Sequence Alignment/Map (SAMtools) v1.280, PCR duplicates were eliminated using the Picard function available in MarkDuplicates software v1.12681. Peak calling was performed using macs282 (with parameters -p 1e-2 --nomodel --shift 0 -B --SPMR). Only peaks with an overlap of ≥ 0.5 between replicates were considered. Histone mark signals around DMR sets were extracted with the annotatePeaks.pl algorithm available in Hypergeometric Optimization of Motif EnRichment (HOMER) software v4.10.3 (with parameters: size = 10000, hist = 10).
Super-enhancer identification
H3K27ac ChIP-seq data were used to identify the super-enhancer regions, as described previously83 using Rank-Ordering of Super-Enhancers (ROSE) software. An enhancer stitching distance of 15 kb was used along with a 2.5 kb transcriptional start site (TSS)-exclusion window.
Data analysis
Hierarchical clustering was carried out based on Pearson correlation distance metrics and average linkage criteria. For low-dimensional analysis we used the t-distributed stochastic neighbor embedding (t-SNE) method implemented in the Rtsne v0.15 package.
Transcription factor motifs were enriched for each set of DMRs using HOMER software v4.10.3. Specifically, we used the findMotifsGenome.pl algorithm (with parameters -size given -cpg) to search for significant enrichment against a background sequence adjusted to have similar CpG and GC contents.
Transcription factor binding analysis was performed interrogating the overlap between the different sets of DMRs with ChIP-seq data for transcription factors available for GM12878 cell line from the ENCODE Project54. The enrichment factor was calculated against random regions as a background, and P values were calculated using Fisher’s exact test. Finally, the transcription factors downstream of the BCR signaling pathway were manually annotated from a curated database53.
Chromatin states and histone mark enrichments analysis for NBC, germinal center B cells and ncsMBC were assessed using a custom adaptation of the EpiAnnotator R package84 using BLUEPRINT data79. DMRs were converted to hg38 assembly with the liftOver function in the rtracklayer v1.42 R package.
Replication timing data in the GM12878 lymphoblastoid cell line were obtained from the UW Repli-seq track of the UCSC Genome Browser. Replication timing values were binned in deciles to perform the overlap with the DMR groups.
DMR annotation for genetic context location was performed using the annotatePeaks.pl algorithm in the HOMER software v4.10.3. For determine the location relative to a CpG island (CGI), we used ‘hg19_cogs’ annotation in the annotatr v1.8 R package.
GREAT software85 was used to enrich downstream pathways and gene ontologies. We used the single nearest gene option for the association between genomic regions with genes.
All statistical analysis (excluding T-WGBS and ChIP-seq analyses) were done in R v3.5.1. Data distributions were tested for normality. Normal data were tested using two-tailed unpaired Student’s t-tests; non-normal data were analyzed with the appropriate non-parametric statistical test. Levels of significance are indicated as: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Non-significance (P ≥ 0.05) was indicated as ‘ns’.
Public RRBS of B cell activation
Data of EBV and CD40/IL4 B cell activation were downloaded from the NCBI Gene Expression Omnibus (GSE49629)24. Methylation calls from RRBS data were filtered, so that only those CpGs with a minimum of five reads per position in all samples were retained. Since RRBS genomic coverage is significantly lower than T-WGBS we only tested the methylation status of positions common to two datasets.
EBV infection
For naïve B cell EBV infection experiments, we obtained buffy coats from anonymous donors through the Catalan Blood and Tissue Bank (CBTB). The CBTB follows the principles of the World Medical Association (WMA) Declaration of Helsinki. Before providing the first blood sample, all donors received detailed oral and written information and signed a consent form at the CBTB. PBMCs were isolated using Ficoll-Paque gradient centrifugation. Total B cells were isolated from PBMCs using positive selection with MACS CD19 microbeads (Miltenyi Biotec). Next, cells were stained with CD27-APC (Miltenyi Biotec, clone M-T271) and IgD – PE (SouthernBiotech, Cat. No. 2032-09) and naïve B cells were sorted as CD27−IgD+. Pure naïve B cells were incubated with B95-8 cell supernatant for 3 h at 37ºC in order to infect them with EBV. Finally, cells were collected after 30 days.
DATA AVAILABILITY
The data that support the findings of this study are available from the NCBI Gene Expression Omnibus (GEO).
CODE AVAILABILITY
Code and data processing scripts are available from the corresponding author upon request
FUNDING
We thank CERCA Programme/Generalitat de Catalunya for institutional support. E.B. was funded by the Spanish Ministry of Economy and Competitiveness (MINECO; grant numbers SAF2014-55942-R and SAF2017-88086-R) and was cofunded by FEDER funds/ European Regional Development Fund (ERDF) – a way to build Europe. E.B is supported by RETICS network grant from ISCIII (RIER, RD16/0012/0013).
CONTRIBUTIONS
F.C.-M. designed and performed experimental experiments and bioinformatics analysis; F.C.-M., D.W. and C.P. generated TWGBS data; P.L. supervised some bioinformatics analysis; F.C.-M and A.F.Á.-P performed AID hot spot analysis; F.C.-M and J.R.-U. design sorting strategies; C.K., C.S, H.A., M.M.-G, R.D., P.S.-P., S.K., L.H., A.D., and B.D. contributed with clinical material and clinical interpretation of the results; E.B. conceived and supervised the study; F.C.-M. and E.B. wrote the manuscript; All authors participated in discussions and interpretation of the data and results.
ACKNOWLEDGEMENTS
This study makes use of data generated by the BLUEPRINT Consortium. A full list of the investigators who contributed to the generation of the data is available from www.blueprint-epigenome.eu. Funding for the project was provided by the European Union’s Seventh Framework Programme (FP7/2007–2013) under grant agreement no 282510—BLUEPRINT. We are also very grateful to Dr Javier di Noia, Dr Almudena Ramiro, Dr Javier Carmona and Dr Biola M. Javierre for useful feedback.
REFERENCES
- 1.↵
- 2.↵
- 3.
- 4.
- 5.
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.
- 26.
- 27.↵
- 28.↵
- 29.↵
- 30.
- 31.↵
- 32.
- 33.
- 34.
- 35.
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.
- 57.
- 58.
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.
- 67.
- 68.↵
- 69.↵
- 70.↵
- 71.
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵