TET enzymes augment activation-induced deaminase (AID) expression via 5-hydroxymethylcytosine modifications at the Aicda superenhancer

See allHide authors and affiliations

Science Immunology  26 Apr 2019:
Vol. 4, Issue 34, eaau7523
DOI: 10.1126/sciimmunol.aau7523

Epigenetic enablers of B cell activation

Initiation of immunoglobulin class switching and somatic hypermutation during B cell activation requires tightly regulated expression of the activation-induced cytidine deaminase (Aicda) gene. Lio et al. investigated how the TET2 and TET3 enzymes, responsible for depositing 5-hydroxymethylcytosine epigenetic marks on genomic DNA, contribute to regulation of Aicda expression by studying B cells from mice in which the two enzymes can be inducibly deleted. This analysis identified two key elements (TetE1 and TetE2) within the Aicda superenhancer that are marked by TET2 and TET3 for progressive hydroxymethylation followed by demethylation, thereby facilitating chromatin accessibility and Aicda expression. The results demonstrate that kinetic mapping of 5-hydroxymethylcytosine on genomic DNA is a powerful tool for identification of key sites of active gene transcription during lymphocyte activation and differentiation.


TET enzymes are dioxygenases that promote DNA demethylation by oxidizing the methyl group of 5-methylcytosine to 5-hydroxymethylcytosine (5hmC). Here, we report a close correspondence between 5hmC-marked regions, chromatin accessibility and enhancer activity in B cells, and a strong enrichment for consensus binding motifs for basic region-leucine zipper (bZIP) transcription factors at TET-responsive genomic regions. Functionally, Tet2 and Tet3 regulate class switch recombination (CSR) in murine B cells by enhancing expression of Aicda, which encodes the activation-induced cytidine deaminase (AID) enzyme essential for CSR. TET enzymes deposit 5hmC, facilitate DNA demethylation, and maintain chromatin accessibility at two TET-responsive enhancer elements, TetE1 and TetE2, located within a superenhancer in the Aicda locus. Our data identify the bZIP transcription factor, ATF-like (BATF) as a key transcription factor involved in TET-dependent Aicda expression. 5hmC is not deposited at TetE1 in activated Batf-deficient B cells, indicating that BATF facilitates TET recruitment to this Aicda enhancer. Our study emphasizes the importance of TET enzymes for bolstering AID expression and highlights 5hmC as an epigenetic mark that captures enhancer dynamics during cell activation.


The three TET proteins (ten-eleven-translocation; TET1, TET2, and TET3) are Fe(II)- and α-ketoglutarate–dependent dioxygenases that catalyze the stepwise oxidation of 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC), and 5-carboxylcytosine (5caC) (1, 2). Together, these oxidized methylcytosine (oxi-mC) bases are intermediates in DNA demethylation and may also function as stable epigenetic marks. 5hmC, the most stable and abundant product of TET enzymatic activity, is highly enriched at the most active enhancers and in the gene bodies of the most highly expressed genes, and its presence at enhancers correlates with chromatin accessibility (2, 3). TET proteins regulate several fundamental biological processes, including lineage commitment, and play important roles in embryonic, neuronal, and hematopoietic development (3).

TET proteins, particularly TET2 and TET3, have critical roles in B cell differentiation and malignancy (2). We and others have previously shown that deletion of Tet2 and Tet3 at early stages of mouse B cell development with an Mb1-Cre transgene resulted in impaired light-chain rearrangement and developmental blockade, leading to acute precursor B cell–derived leukemia with 100% penetrance (4, 5). Inducible deletion of Tet1 and Tet2 using Mx1-Cre promoted the development of acute lymphoblastic leukemia derived from precursor B cells, and global loss of Tet1 caused B cell lymphomas with an extended latency (6). In humans, TET2 mutations are frequently observed in diffuse large B cell lymphoma, a malignancy derived from germinal center (GC) B cells (7, 8), suggesting that TET proteins may regulate mature B cell function. However, because of the pleiotropic functions of TET proteins, studies of TET-mediated gene regulation are best performed in systems where TET genes are deleted acutely rather than during development.

After their development in the bone marrow, mature B cells migrate to peripheral lymphoid tissues where they encounter antigen and follicular T helper cells in GCs and participate in the generation of functional immune responses (9). In GCs, B cells diversify the variable regions of immunoglobulin (Ig) chains in a process known as somatic hypermutation (SHM) and also undergo class switch recombination (CSR) to replace the immunoglobulin M (IgM) constant region with those of other isotypes (IgG1, IgA, etc.). Both CSR and SHM are orchestrated by the enzyme AID (activation-induced cytidine deaminase, encoded by Aicda) (1012). AID promotes CSR and SHM by generating DNA double-strand breaks at Ig switch regions and point mutations at Ig variable regions, respectively (13). Because of its high mutagenic potential (14, 15), AID expression is normally restricted to activated B cells and is tightly regulated.

Here, we investigated the role of TET proteins during mouse B cell activation by mapping 5hmC distribution genome-wide and integrating the data with previous studies of transcriptional and epigenetic changes during B cell activation (16, 17). We deleted the Tet2 and Tet3 genes acutely using CreERT2 to avoid secondary effects caused by prolonged TET deficiency during differentiation. We show that TET2 and TET3 regulate CSR by controlling the activation-induced up-regulation of AID mRNA and protein and that they act downstream of the basic region/leucine zipper (bZIP) transcription factor BATF (basic leucine zipper transcription factor, ATF-like), which is induced during B cell activation with more rapid kinetics than Aicda and binds concomitantly with TET proteins to two TET-responsive elements in the Aicda locus, TetE1 and TetE2. Our study demonstrates the role of TET proteins in CSR in activated B cells and provides a detailed description of the general mechanism whereby TET proteins influence cell activation and differentiation.


B cell activation promotes genome-wide deposition of 5hmC

We immunoprecipitated genomic DNA with antibodies to cytosine-5-methylenesulfonate (CMS-IP) (18, 19) to analyze the kinetics of 5hmC deposition in murine B cells activated with lipopolysaccharide (LPS) and interleukin-4 (IL-4), a well-characterized in vitro system for studying CSR (Fig. 1A). The vast majority of 5hmC-marked regions (~160,000) were shared between pre- and postactivated B cells (Fig. 1B and fig. S1A). Of the ~9500 differentially hydroxymethylated regions (DhmRs) in 72-hour-activated B cells versus naïve B cells, the majority (8454) had increased 5hmC (DhmR72h-up), whereas a much smaller number (928) had decreased 5hmC (DhmRdown) (Fig. 1, B and C). DhmRs were typically located more than 10 kb from the closest transcription start site (TSS) (fig. S1B), and their 5hmC levels progressively changed with time after activation [DhmR72h-up (Fig. 1D) and DhmR72h-down (fig. S1C)].

Fig. 1 Dynamic changes in 5hmC during B cell activation.

(A) Flowchart of experiments. (B) Comparison between 5hmC-enriched regions in unstimulated and activated B cells (72 hours). One hundred ninety-three regions represented only in naïve B cells were not shown. (C) Number of DhmRs. (D) Kinetics of 5hmC modification in the DhmR72h-up (left). There is no increase in the same number of randomly chosen common 5hmC-marked regions (middle) and random regions (right). 5hmC enrichment is shown as normalized reads per 100-bp bin. (E) The 85 and 1953 regions with increased 5hmC in 24-hour- and 48-hour-activated B cells relative to naïve B cells show decreased “methylation” (bisulfite-resistant cytosine, 5mC + 5hmC) at their centers 48 hours after activation. Average methylation was calculated for each 200-bp bin across 6 kb. (F) Motif enrichment analysis of DhmR72h-up. Common 5hmC-enriched regions were used as background for analysis. The y axis indicates the fold enrichment versus background, circle size indicates the percentage of regions containing the respective motif, and the color indicates the significance (log10 P value). (G) 5hmC is enriched at active (H3K4me1+ H3K27Achi) relative to poised (H3K4me1+ H3K27Aclo) enhancers in both activated and naïve B cells. The y and x axes indicate the levels (log2) of H3K4me1 and H3K27Ac relative to input, respectively. (H) A substantial fraction of superenhancers (352 of 459, 76.7%) identified by high H3K27Ac enrichment overlap with DhmR72h-up. Fisher’s exact test was used to analyze the significance. ***P < 0.01 (P = 8.9 × 10−266). n.s., not significant. (I) Ccr4 locus (mm10; chr9:114,484,000 to 114,501,000) as an example of a region with increased 5hmC, increased H3K27Ac, and decreased CpG methylation after activation. The red track indicates CpGs that were included for analysis based on coverage. (J) Kinetics Ccr4 mRNA expression (by RNA-seq) in activated B cells. See also fig. S1.

The oxi-mC produced by TET proteins are known intermediates in DNA demethylation (2, 20). To relate 5hmC to changes in DNA methylation, we compared 5hmC distribution in naïve and 72-hour-activated B cells with published whole-genome bisulfite sequencing (WGBS) data on B cells activated for 48 hours under similar conditions (17). Although WGBS cannot distinguish 5mC from 5hmC (21), 5hmC is typically a small fraction (1 to 10%) of 5mC (22); thus, we refer to the WGBS signal as “DNA methylation” here. Most (1097 of 1168, 94%) differentially methylated regions (DMRs) were demethylated at 48 hours and marked by 5hmC (fig. S1, D and E). Regions that showed increased 5hmC at 24, 48, and 72 hours (DhmR24h-up, DhmR48h-up, and DhmR72h-up) also showed decreased DNA methylation (Fig. 1Eand fig. S1F). Overall, regions of 5hmC deposition correspond to regions of DNA hypomethylation during B cell activation, as expected from the well-established role of 5hmC as an intermediate in DNA demethylation. Motif enrichment analysis of the 8454 DhmR72h-up and 1097 DMR48h-down regions showed that both were enriched in consensus binding sequences for transcription factors of the nuclear factor κB (NF-κB) (Rel homology domain) and bZIP families, as well as for “composite” IRF:bZIP motifs (Fig. 1F and fig. S1G) (2325).

To discern the relationship between 5hmC and enhancers, we stratified naïve and activated B cell enhancers, defined by H3K4 monomethylation (H3K4me1), on the basis of the level of H3K27 acetylation (H3K27Ac), a modification that tracks with enhancer activity (26). 5hmC was most highly enriched at active (H3K4me1+ H3K27Ac+) relative to poised (H3K4me1+ H3K27Ac) enhancers (Fig. 1G). Moreover, more than 75% of previously identified superenhancers in activated B cells, defined by H3K27Ac, overlapped with at least one DhmR72h-up region (Fig. 1H) (27): For instance, a 3′ distal element at the Ccr4 locus showed activation-dependent gain of 5hmC and H3K27Ac, associated with concomitant loss of methylation at specific CpGs and increased mRNA expression (Fig. 1, I and J). 5hmC was also associated with accessible chromatin defined by ATAC-seq (assay for transposase-accessible chromatin using sequencing) (4, 5, 28) (see below), and kinetic analysis of active enhancers showed that 5hmC levels correlated with enhancer activity (fig. S1H). Together, our data show that 5hmC modification and DNA demethylation correlate with enhancer activity during B cell activation.

Comparison of wild type and Tet2/3 double-deficient B cells identifies TET-responsive regulatory elements

TET2 and TET3 are the two major TET proteins expressed in B cells (Fig. 2A). To evaluate the role of TET proteins in regulating B cell function, we treated CreERT2 Tet2fl/fl Tet3fl/fl Rosa26-YFPLSL for 5 days with tamoxifen to delete Tet2 and Tet3 (double knockout, DKO). Tet2fl/fl Tet3fl/fl Rosa26-YFPLSL wild-type (WT) mice were used as controls; YFP reports Cre expression. B cells were isolated and activated with LPS and IL-4 (Fig. 2B). Tet2 and Tet3 were both efficiently deleted in B cells (Fig. 2C), and the YFP+ cells showed similar frequencies of mature splenic follicular B cells (fig. S2, A and B).

Fig. 2 Comparison of 5hmC modification in WT and Tet2/3 DKO B cells.

(A) Mean mRNA expression for TET family members (from RNA-seq). TPM, transcripts per million. (B) Description of mice and flowchart of experiment. (C) Tet2 and Tet3 are efficiently deleted. Tet2 and Tet3 expression in B cells from tamoxifen-treated WT control and Tet2/3 DKO mice [as described in (B)] were analyzed by qRT-PCR. Data were normalized to Gapdh within sample and subsequently to the value from WT. A representative experiment is shown (two independent experiments with three technical replicates). ***P < 0.01. (D) Number of differential 5hmC-enriched regions (DhmR) between WT and Tet2/3 DKO B cells. (E) Kinetics of 5hmC enrichment from WT (left) and Tet2/3 DKO (right) at DhmRs72h between WT and DKO [72 hours, (D)]. Regions with decreased 5hmC in DKO are shown above (WT > DKO) and those with increased 5hmC are shown below (DKO > WT). 5hmC enrichment is shown in normalized reads per 100-bp bin. (F) Motif enrichment analysis of DhmR72h-WT > DKO (analyzed as in Fig. 1F). See also fig. S2.

Global 5hmC levels were similar in WT and Tet2/3 DKO B cells before activation but showed a perceptible decrease by 48 hours after activation, consistent with passive loss of 5hmC as a function of cell division (fig. S2C) (2, 20). Starting at 48 hours, 5hmC levels were substantially lower in Tet2/3 DKO compared with WT B cells, indicating that TET2 and TET3 actively oxidize 5mC to 5hmC during B cell activation (fig. S2C). Around 2300 5hmC–enriched regions were significantly different between WT and DKO at 72 hours of activation, with substantially more regions gaining 5hmC in control compared with Tet2/3 DKO B cells (Fig. 2, D and E); most were located >10 kb from the nearest TSS (fig. S2D). Of the 2139 “TET-regulated” DhmRs with higher 5hmC in WT compared with Tet2/3 DKO B cells (“WT > DKO DhmR”), 2020 (94.4%) significantly overlapped with DhmR72-up regions; most showed decreased DNA methylation at their centers (fig. S2E) and were enriched for RHD, bZIP, and composite IRF:bZIP motifs (Fig. 2F).

TET2 and TET3 regulate Ig CSR

To assess the effect of Tet2/3 deletion on the antibody response in vivo, we treated Tet2 fl/fl Tet3 fl/fl Rosa26-LSL-YFP CreERT2 and control mice with tamoxifen before immunization with 4-hydroxy-3-nitrophenylacetyl–conjugated ovalbumin (NP-OVA) (Fig. 3A). Acute deletion of Tet2/3 resulted in increased numbers of total cells and B cells in draining popliteal lymph nodes by day 7 after immunization (fig. S3A), consistent with our previous observations that Tet2/3 deficiency results in increased cell survival and/or proliferation (28). The overall percentage of GC B cells (CD19+ GL7+ Fas+) was similar between WT and DKO (gated on YFP+) (Fig. 3, B and C), but there was a significant increase in the frequency of NP-specific B cells (Fig. 3D). There was also a moderate (~25%) decrease in GL7 MFI in Tet2/3 DKO GC B cells compared with WT control GC B cells (fig. S3B), indicating that although TET proteins are, in general, important for cell differentiation, acute deletion of Tet2/3 in B cells had only a limited effect on GC B cell differentiation. Because acute Tet2/3 deletion using the CreERT2 system affects multiple cell types, we did not investigate the in vivo B cell phenotype further in Tet2fl/fl Tet3fl/fl CreERT2 Rosa26 YFPLSL mice. The most notable phenotype in these mice, however, was the consistent decrease in CSR from IgM to IgG1 (Fig. 3, E and F), demonstrating a role for TET proteins in regulating antibody responses in vivo, particularly the CSR.

Fig. 3 TET proteins facilitate CSR.

(A) Flowchart of in vivo experiment. f.p., footpad. (B) Top: Flow cytometric analysis of CD19+GL7+Fas+ GC B cells at the draining popliteal lymph nodes from WT and Tet2/3 DKO mice after NP-OVA immunization as in (A). Bottom: Decreased IgG1-switched cells among GC B cells in Tet2/3 DKO (YFP+ GC B–gated) compared with WT mice (GC B–gated). (C to F) Quantification of experiments shown in (B). Data shown are aggregated results from two independent experiments. Means and SEs are shown. WT, n = 11; DKO, n = 12. (G) Flowchart of in vitro IgG1 switching. Cells were labeled with Cell trace violet and activated for 4 days with LPS and IL-4. (H and I) Flow cytometry plots (H) and quantification (I) of IgG1-switched B cells in Tet2/3 DKO (n = 4) and WT (n = 4) mice. Representative of at least three independent experiments. (J) Circular γ1 transcript was quantified by qRT-PCR and normalized to Gapdh and then to the level of WT. Representative experiment of two independent experiments with three technical replicates. (K) Flowchart of in vitro IgA switching. Cells were activated for 5 days with anti-CD40, rmIL-4, rmIL-5, and rhTGFβ. (L to N) Flow cytometry plots (L) and quantification (M and N) of IgG1-switched (M) and IgA-switched cells (N). Representative of three independent experiments with three technical replicates. Statistical significance was calculated using unpaired two-tailed t test. *P < 0.05, ***P < 0.01. See also fig. S3.

To determine whether the CSR phenotype was B cell intrinsic, B cells from tamoxifen-treated mice were labeled with a proliferation-tracking dye (CellTrace Violet) and activated with LPS and IL-4 for 4 days (Fig. 3G). Consistent with the CSR defect in vivo, IgG1 switching in Tet2/3 DKO B cells in vitro was consistently lower relative to WT B cells (Fig. 3, H and I). The impaired CSR in Tet2/3 DKO was not due to Cre activity (fig. S3C); the defect was cell intrinsic based on mixing experiments (fig. S3D); and the difference was not due to altered proliferation (Fig. 3H and fig. S3E). Correlating with the decrease in CSR, the expression of circular γ1 transcript was decreased in Tet2/3 DKO B cells (Fig. 3J). Furthermore, CSR to IgA was also decreased in Tet2/3 DKO relative to WT B cells (Fig. 3, K to N). The loss of Tet2/3 also resulted in a decrease in the differentiation of CD138+ plasma blasts/cells after in vitro activation (fig. S3F). Reconstitution of Tet2/3 DKO B cells with the enzymatically active catalytic domain of mouse TET2 (Tet2CD) (29) restored CSR almost to control levels (fig. S3, H and I). An enzymatically inactive mutant of Tet2CD (Tet2CDHxD) also partially restored CSR, suggesting that both TET enzymatic activity and scaffolding function have a role in CSR (fig. S3, G to I) (30). Together, these results indicate that Tet2 and Tet3 are required for optimal CSR both in vitro and in vivo.

Because the CSR defect in Tet2/3 DKO B cells was ~50% of control, we asked whether deletion of all three TET proteins might have a more notable effect. Consistent with the very low expression of Tet1 in mature B cells (Fig. 2A), the CSR defect in Tet1/2/3 TKO mice was comparable to that observed in Tet2/3 DKO mice (fig. S3, J and K). These results indicate that TET2 and TET3 are the major TET proteins that regulate CSR in B cells.

TET2 and TET3 regulate expression of the cytidine deaminase AID

CSR is a highly regulated process and involves multiple pathways (31). RNA sequencing (RNA-seq) analysis identified a relatively small number of genes differentially expressed between WT and Tet2/3 DKO B cells (fig. S4, A and B); among these was Aicda, which encodes AID, a protein essential for CSR. Quantitative reverse transcription polymerase chain reaction (qRT-PCR) and Western blot analyses confirmed a >50% decrease in Aicda mRNA in Tet2/3 DKO relative to WT B cells after activation (Fig. 4, A and B, and fig. S4, C and D), a phenotype that resembled the dampened CSR in AID haploinsufficient mice (fig. S4, E to J) (32, 33). Although Tet2 mRNA expression showed only minor changes in unstimulated versus stimulated B cells (Fig. 2A), Tet2 protein expression was low in unstimulated and 24-hour-stimulated B cells, with increased expression observed at 48 hours after stimulation (fig. S4D). The late TET2 induction parallels the late kinetics of increase in 5hmC (Fig. 1C).

Fig. 4 TET2/3 facilitate CSR by regulating expression of the cytidine deaminase AID.

(A) qRT-PCR analysis of Aicda mRNA expression in WT and Tet2/3 DKO B cells activated 4 days with LPS and IL-4. Aicda expression was normalized to Gapdh and then to the level in WT. Data shown are representative of two independent experiments with three technical replicates. *P < 0.05. (B) Immunoblotting of whole-cell lysates from day 4 activated WT and Tet2/3 DKO with indicated antibodies. Left lane contains lysate from the AID-KO CH12 B cells as a control for the specificity of anti-AID antibody. β-Actin was used as loading control. Data shown are representative of two independent experiments. See also fig. S4D. (C and D) WT and Tet2/3 DKO B cells were transduced with empty vector (Thy1.1, left), WT AID (AIDWT, middle), or catalytically inactive AID (AIDH56R/E58Q, right). Gated on live-transduced B cells (CD19+ Thy1.1+). Representative flow cytometry plots (C) and quantification (D) are shown. Data shown are representative of three independent experiments. ***P < 0.01. See also fig. S4.

To determine whether the decrease in AID expression was fully responsible for the CSR defect, we expressed WT and catalytically inactive AID (AIDH56R/E58Q) (34) in WT and Tet2/3 DKO B cells via retroviral transduction. Retroviral expression of catalytically active AID (AIDWT), but not inactive AIDH56R/E58Q, largely rescued the CSR defect in Tet2/3 DKO B cells (Fig. 4, C and D, bottom panels). Similar to previous observations (12), expression of AID in WT cells also increased the frequency of IgG1+ cells (Fig. 4, C and D, top left and middle panels). Because Tet2/3 were not required for expression of the μ and γ1 germline transcripts essential for CSR (fig. S4K), the bulk of the CSR defect in Tet2/3 DKO B cells can be attributed to the decrease in expression of Aicda mRNA and AID protein, leading us to test the hypothesis that TET proteins control Aicda expression through distal regulatory element(s) of the Aicda gene.

Genome-wide analyses identify TET-responsive regulatory elements in the Aicda locus

Multiple conserved regulatory elements influence Aicda expression (fig. S5A), and their deletion markedly decreased Aicda expression in activated B cells (17, 3537). Of these, the Aicda 5′ enhancer at −26 kb in the Mfap5 gene, the intergenic 5′ enhancers, and the intron 1 enhancer noticeably gain H3K27Ac and lose 5mC upon activation and have been collectively termed the Aicda “superenhancer” (Fig. 5A, middle and bottom tracks) (17, 27).

Fig. 5 TET2 and TET3 control Aicda expression via TET-responsive elements TetE1 and TetE2.

Diagram shows two conserved TET-responsive elements TetE1 and TetE2 located 5′ of the Aicda gene (labeled with green rectangles and gray shades). (A) Top two tracks: ChIP-seq analysis showed that TET2 (blue track) specifically bound to multiple elements in the Aicda locus (mm10; chr6:122,523,500 to 122,576,500) after activation when compared with Ig control (gray track). Middle track (green): Increased H3K27Ac after activation. Bottom track: Activation induced DNA demethylation at TetE1 and TetE2. WGBS showing DNA methylation (5mC + 5hmC) in naïve and 48-hour-activated B cells (mCG; black track). CpGs included in the analysis are indicated by red lines (red track). Bottom track indicates the conserved DNA elements among placental animals (“Conserve”). A previously identified superenhancer is indicated. For TET2 and Ig, scales indicate per 10 million reads; H3K27Ac, quantile-normalized reads; BS, percentage of bisulfite-resistant cytosine. (B) Activation induced TET2/3-dependent 5hmC deposition at Aicda distal elements. 5hmC modification was analyzed in WT and Tet2/3 DKO B cells activated as in Fig. 3G. Differentially 5hmC-enriched regions between WT and DKO after 72 hours activation are shown (WT > DKO DhmR72h). Scales indicate quantile-normalized reads. (C) TET2/3 deposit 5hmC and demethylate TetE1. Top: CpG modifications (5hmC, 5mC, and C) at TetE1 were analyzed by oxBS-seq using DNA isolated from WT and Tet2/3 DKO B cells before (0 hours) and after (72 hours) activation. Bottom: Quantification of the probability of 5mC modification on all CpGs in TetE1. The difference between samples was compared by using the Wilcoxon rank-sum test. *P < 0.05, ***P < 0.01. See also fig. S5H.

Chromatin immunoprecipitation sequencing (ChIP-seq) for TET2 showed that each of these regulatory elements was occupied by TET2 in 72-hour-activated B cells (Fig. 5A, top two tracks). Among these, the −26-kb Mfap5 intronic region and the −8-kb 5′ intergenic region (here termed TetE2 and TetE1, respectively) were TET-regulated: B cell activation induced a TET-dependent increase in 5hmC (Fig. 5B), placing them in the category of WT > DKO DhmRs (Fig. 2, D and E). TetE1 appears to be the prime target for TET2/3 due to its larger gain of 5hmC after activation (Fig. 5B). In contrast, regions such as the Aicda promoter were marked by 5hmC even before activation, indicating that the 5hmC at these regions was likely generated during a previous stage of B cell differentiation and then maintained until the emergence of naïve peripheral B cells.

To confirm the importance of the TetE1 enhancer in Aicda regulation, we deleted the enhancer using CRISPR in CH12F3 cells, a B cell line that can class-switch from IgM to IgA upon activation with anti-CD40/IL-4/TGFβ (fig. S5, B and C). We tested four clones with homozygous deletions; all showed decreased expression of Aicda mRNA, and in three of these, there was almost no detectable CSR (fig. S5, D and E), confirming a previous report in the context of a bacterial artificial chromosome (BAC) transgene that TetE1 was essential for Aicda expression (36). B cell activation also induced hydroxymethylation at the IgH locus, most notably upstream of the IgG1 promoter (fig. S5F). Given that Tet2/3 DKO B cells expressed similar levels of IgG1 germline transcripts (fig. S4K) and that ectopic expression of AID could rescue the impairment of IgG1 CSR, the significance of TET-mediated DNA modification/demethylation at the IgH locus remains to be determined.

B cell activation induces strong H3K27Ac and DNA demethylation at TetE1 (Fig. 5A). Because bisulfite sequencing does not distinguish 5mC from 5hmC, we used oxidative bisulfite sequencing (oxBS-seq) to assess the levels of 5mC, 5hmC, and unmodified C at TetE1 in WT and Tet2/3 DKO cells [neither BS-seq nor oxBS-seq distinguish unmodified C from 5fC and 5caC, but these modified bases are ~10-fold and ~100-fold less abundant than 5hmC (2)]. CpGs in both TetE1 and the Aicda promoter displayed similar levels of 5mC and 5hmC before activation (Fig. 5C and fig. S5H; compare 0-hour panels). At 72 hours after activation, there was a substantial decrease in 5mC in WT B cells; in contrast, both TetE1 and the Aicda promoter were methylated in Tet2/3 DKO B cells (Fig. 5C, bottom panel; compare 72-hour panels). These results indicate that TET2 and TET3 regulate Aicda expression by binding to and depositing 5hmC at TetE1 and TetE2.

TET2 and TET3 maintain chromatin accessibility at two Aicda TET-responsive elements, TetE1 and TetE2

Active regulatory regions are typically found in accessible regions of chromatin (38) and are marked by 5hmC (4, 28). To assess the dynamics of chromatin accessibility, we performed ATAC-seq in B cells stimulated with LPS and IL-4. Activated B cells displayed progressive chromatin remodeling (fig. S6A). Regions with increased 5hmC after activation (DhmR72h-up) also showed increased chromatin accessibility after activation and vice versa (fig. S7A; blue box-and-whisker plots).

To understand the relationship between TET function and chromatin accessibility, we performed ATAC-seq on WT and Tet2/3 DKO B cells activated as in Fig. 3G. Of a total of ~28,000 accessible regions (fig. S6, B and C), only a minor fraction (~1.5%; 421 of 28,137) showed significant changes in accessibility between WT and Tet2/3 DKO B cells, and the differences were observed late, at 72 hours after activation (fig. S6, B to D). Of the 292 potentially TET-regulated differentially accessible regions (DARs), defined as showing decreased accessibility in Tet2/3 DKO compared with WT B cells (WT > DKO DARs), the majority were located distal to the TSS (fig. S7B) and a significant proportion of these (110 of 292, 37.7%) showed a TET2/3-dependent increase in 5hmC at 72 hours compared with unstimulated cells (DhmR72up) (figs. S6C and S7C, top). In contrast, the 129 DKO > WT DARs that were less accessible in WT compared with Tet2/3 DKO B cells and the 27,716 commonly accessible DARs were present in both TSS-proximal and TSS-distal regions and did not show significant changes in 5hmC (fig. S6C, middle and bottom panels, and fig. S7, B and C). Analysis of DNA methylation at 48 hours after activation showed that WT > DKO DARs were further demethylated after activation, whereas DKO > WT DARs were already substantially demethylated in naïve B cells and showed no further changes after activation (fig. S7D). Moreover, the WT > DKO DARs were enriched in bZIP and BATF:IRF motifs (fig. S7E), similarly to those in DhmR72h-up (Fig. 1F) and WT > DKO DhmRs (Fig. 2F). Together, these data support our previously observed correlation of 5hmC modification with changes in chromatin accessibility.

Focusing on the Aicda locus, we found that activation was associated with increased accessibility at the Aicda enhancers TetE1 and TetE2 (fig. S6D). The 5hmC modification continuously increased at these two elements until 72 hours, with a higher level of deposition at TetE1 (see Fig. 5B). In contrast, the time course of increase in chromatin accessibility was quite different at the two enhancers (fig. S6D): TetE2 showed a rapid increase in accessibility at 24 hours after activation, whereas the time course of increase in TetE1 accessibility was slower, matching that of 5hmC deposition (compare Fig. 5B and fig. S6D). Consistent with the increased accessibility, several chromatin remodelers and histone acetyltransferases, including BRG1, CHD4, p300, and, to a lesser extent, GCN5, were recruited to TetE1 and TetE2 in 24-hour-activated B cells (fig. S7F). We noticed a slight decrease in chromatin accessibility at TetE1 and TetE2 in Tet2/3 DKO B cells compared with WT cells at 72 hours, suggesting that TET proteins are important for maintaining chromatin accessibility at these enhancers (fig. S6E). Loss of TET proteins had no significant effect on chromatin accessibility at the IgH locus (fig. S5G). Together, these data point to a consistent link between bZIP-family transcription factors, TET catalytic activity, and chromatin accessibility.

BATF acts upstream of TET at Aicda enhancers

Before enhancers are established during development, cell lineage specification, or activation, certain key transcription factors bind to nucleosome-associated regions and recruit chromatin remodeling complexes and histone-modifying enzymes to create active enhancers (26). To identify potential pioneer transcription factors for the Aicda locus, we took advantage of our previous motif enrichment analyses (Figs. 1F and 2F and fig. S1G). We had observed strong enrichment for consensus binding motifs for bZIP transcription factors, at regions that progressively gained 5hmC as a function of activation (Fig. 1F, DhmRup), regions that lost DNA methylation upon activation (fig. S1G, DMR48h-down), regions with higher 5hmC in WT compared with Tet2/3 DKO B cells (Fig. 2F, WT > DKO DhmRs), and regions with higher accessibility in WT versus Tet2/3 DKO B cells (fig. S7E, DAR72h WT > DKO).

On the basis of these data, we focused on bZIP transcription factors expressed in activated B cells. Consistent with previous observations (39, 40), Batf mRNA and protein were induced after activation (Fig. 6A and fig. S8B) and their expression preceded that of Aicda, as expected if BATF regulated Aicda mRNA induction (figs. S8, A and B, and S4D). In contrast, expression of Bach1 and AP-1 (Fos and Jun) family members was either low throughout (Fosl1, Fosl2, JunD, and Bach1) or moderate to high in unstimulated B cells, potentially because these cells contained a minor population of memory B cells that express high levels of Fos and Jun (fig. S8, C to E; and (41). Given the kinetics, we examined the importance of BATF in subsequent experiments.

Fig. 6 BATF facilitates TET protein–mediated hydroxymethylation at TetE1.

(A) Mean mRNA expression (RNA-seq) of Batf family (Batf1 to Batf3). (B and C) BATF binding correlates with 5hmC enrichment. WT BATF peaks were divided into two clusters on the basis of the pattern of 5hmC distribution. (B) Cluster 1 showed a broad 5hmC distribution, with the 5hmC level remaining unchanged after activation and in the absence of TET2/3 (top, compare “5hmC from WT” with “5hmC from DKO”). In contrast, a substantial proportion of regions in cluster 2 showed progressive TET-dependent 5hmC modification after activation (bottom), as further illustrated in (C) as line plots. Data shown are mean enrichment per 100-bp bin. (D) Recruitment of BATF and other transcription factors to Aicda enhancers. Top three tracks: Genome browser view of BATF-binding in unstimulated and 72-hour-activated WT (blue) and Tet2/3 DKO B cells (red) at the Aicda locus. Note that loss of Tet2/3 has no significant effect on BATF recruitment (compare WT and DKO, tracks 2 and 3; see also fig. S8G). Activation also induced E2A and PU.1 binding to Aicda enhancers (orange and purple tracks). Green tracks indicate H3K27Ac. Coordinates are chr6:122,523,500 to 122,576,500 (mm10). See also fig. S8. (E) BATF is required for 5hmC modification at TetE1. Batf-WT and Batf-KO B cells were activated with LPS and IL-4 for 4 days. 5hmC modification at TetE1 was quantified using AbaSI-qPCR.

BATF is essential for T and B cells during humoral responses (39, 40), and Batf-KO B cells are defective in CSR (fig. S8F) (39, 40). Genome-wide analysis of Batf binding by ChIP-seq in 72-hour-activated WT and Tet2/3 DKO B cells showed very few overall differences (fig. S8G), indicating that BATF functioned upstream or independently of TET enzymes. Nevertheless, one of two distinguishable sets of BATF ChIP-seq peaks (cluster 2 in Fig. 6B) was TET-regulated, because the peaks in this cluster showed a progressive TET2/3-dependent increase in 5hmC after activation (Fig. 6, B and C). In contrast, BATF peaks in cluster 1 showed no significant activation-dependent increase in 5hmC (Fig. 6B, top panel). Overall, about one-third of regions with activation-induced 5hmC (DhmR72-up) overlapped with BATF peaks (fig. S8H), indicating that in addition to BATF, other transcription factors also have a role in facilitating TET-mediated 5hmC generation at the Aicda locus. Despite this strong functional interaction, we did not observe a substantial direct protein-protein interaction between BATF and TET2 in coimmunoprecipitation experiments in which the effects of nucleic acids were excluded (fig. S8I), suggesting that the interaction is dependent on additional factors.

BATF bound strongly at the TetE1 and TetE2 enhancers in the Aicda locus and, to a lesser extent, to the −21-kb intergenic enhancer located between TetE1 and TetE2 (Fig. 6D, 72-hour WT and DKO; second and third tracks). Consistent with the lack of BATF expression in unstimulated cells, there was no enrichment of BATF occupancy at the Aicda enhancers at 0 hours (Fig. 6D, 0-hour WT; top track). This binding pattern resembles that of TET2 (Fig. 5A), as well as that of E2A and PU.1 (Fig. 6D) (4244). Moreover, BATF and JUNB associated with TETE1 and TETE2 in a human B cell lymphoblast (fig. S8L), suggesting that the binding of BATF to these enhancers is evolutionarily conserved. To determine whether BATF acted upstream of TET, we analyzed 5hmC deposition at the TET-responsive element TetE1 in WT and Batf-deficient B cells. We found unambiguously that the absence of BATF abolished activation-induced hydroxymethylation at TetE1 (Fig. 6E). Our results are consistent with the hypothesis that BATF facilitates the recruitment of TET2 and/or TET3 to TetE1 and TetE2 and increases Aicda expression by promoting 5hmC modification and DNA demethylation at these upstream Aicda enhancers.

As mentioned above, BATF is essential for Aicda regulation. However, we cannot rule out the involvement of additional transcription factors, including other bZIP family members, in this process. Although IRF4 is required for Aicda expression and binds to TetE1 (45, 46), 5hmC levels at TetE1 were unaffected in Irf4-deficient B cells after LPS/IL-4 stimulation (fig. S8M). Thus, depending on cell type and conditions of stimulation, certain transcription factors preferentially function together with TET proteins, whereas others could be responsible for additional aspects of locus remodeling and gene expression.


TET proteins oxidize 5mC to 5hmC, a stable epigenetic mark and an intermediate in DNA demethylation. Because of the pleiotropic effects of TET proteins in cells, it has been challenging to address the specific roles of TET proteins by using TET-deficient mice. To circumvent this problem, we used the inducible tamoxifen-CreERT2 system to delete Tet2 and Tet3 in mature B cells. Our data show that TET2 and TET3 are required for efficient CSR. A primary mechanism involves TET-mediated regulation of the expression of Aicda, the essential DNA cytosine deaminase for CSR. BATF, potentially with other transcription factors, helps recruit TET proteins to two major TET-responsive regulatory elements that we have newly defined in the Aicda locus, TetE1 and TetE2. TET2 and TET3 convert 5mC to 5hmC at these regulatory elements, leading to DNA demethylation, sustaining enhancer accessibility and augmenting Aicda expression.

The biological consequences of TET loss of function are determined by several factors: the time course of Tet2 and Tet3 gene deletion, the stability of TET2 and TET3 mRNA and protein, and the rate of cell division that determines the rate of passive (i.e., replication dependent) dilution of 5hmC. At each cell division, hemi-methylated CpGs are recognized by the DNMT1/UHRF1 maintenance DNA methyltransferase complex and converted back to symmetrically methylated CpGs, whereas hemi-hydroxymethylated CpGs are ignored and so are diluted by half (1, 2). Consequently, 5hmC is present at comparable levels in quiescent (nondividing) WT and Tet2/3 DKO B cells, thus enabling us to study the effects of acute TET deletion in activated, proliferating B cells. The progressive replication-dependent loss of 5hmC and consequent dilution of both 5mC and 5hmC is likely to be required for optimal gene expression, explaining the long-standing observation that the induction of Aicda expression during B cell activation and the induction of cytokine genes during T helper 2 (TH2) cell differentiation are both tightly coupled to cell division (47, 48).

An optimal level of AID is crucial to maintain the necessary balance between effective antibody immune responses and unintentional C > T mutations caused by AID-mediated DNA cytidine deamination. Although Aicda haploinsufficiency results in dampened antibody responses (fig. S4, E to J) (32, 33), uncontrolled AICDA expression is associated with B cell malignancies (49). Thus, the level and activity of AID are meticulously controlled by diverse mechanisms including a tight transcriptional regulatory program (50). At the Aicda locus, at least six regulatory elements have been identified (fig. S5A); five of them, located at distances ranging from −29 to +5 kb relative to the TSS, are collectively termed the Aicda superenhancer (17, 27). The enhancers at −26, −21, −8, and +13 kb are all necessary for inducing Aicda expression in activated B cells, on the basis of deletion of individual enhancers in mice and the CH12 B cell line (36, 37). Even in naïve B cells where Aicda is not expressed, the Aicda promoter is already highly enriched in 5hmC and the −21-, intronic, and +13-kb Aicda enhancers display 5hmC and H3K27Ac (fig. S5A). The 5hmC modification at the −26-, −21-, and +13-kb Aicda enhancers is apparent as early as the pro–B cell stage of B cell development (4), suggesting that TET-mediated 5hmC modification acts to “bookmark” regulatory elements necessary for proper gene expression in progeny cells after activation.

The vast majority of 5hmC-marked regions are present in common between naïve and activated mature B cells (this study) and between WT and TET-deficient invariant natural killer T cells (iNKT) cells (28), supporting the hypothesis that most 5hmC-marked regions in any given cell type were laid down during previous developmental stages and thus are constitutively modified. In contrast, activation-induced 5hmC modification occurs at only a few distal elements in B cells (Fig. 1C), and 5hmC levels at these elements correlate strongly with activation-induced increases in enhancer activity defined by H3K27Ac (Fig. 1G) (51). Moreover, the majority of previously described B cell superenhancers (17, 27) harbor at least one activation-induced 5hmC-modified regulatory element (Fig. 1H). In the particular case of Aicda, we identified activation-induced 5hmC modification at two major TET-responsive elements, TetE1 and TetE2, both part of a superenhancer cluster located 5′ of the Aicda gene (Fig. 5A) (17, 27). 5hmC modification at these elements was apparent by 48 hours (Fig. 5B), preceding the marked up-regulation of Aicda mRNA at 72 hours (fig. S4C). Tet2/3 deficiency almost eliminated the activation-induced 5hmC modification at both enhancers and resulted in diminished expression of both Aicda mRNA and AID protein (Figs. 4, A and B, and 5B and fig. S4C). Thus, our in vitro data indicate that TET proteins and 5hmC are important for Aicda expression by enabling the TetE1 and TetE2 enhancers to function at full capacity. However, we cannot rule out that in addition to decreasing AID expression, TET deficiency affects CSR indirectly by impeding B cell differentiation in vivo.

Our data strongly suggest that the bZIP transcription factor BATF is a major bZIP transcription factor responsible for TET recruitment to the Aicda locus, consistent with the identification that the bZIP motif is the singularly most enriched motif that correlated with DNA demethylation after B cell activation in human (52). BATF is induced at the mRNA level before Aicda induction in activated B cells (Fig. 6A), and Batf deficiency in B cells is associated with a marked impairment of AID expression and CSR (39, 40). Although loss of TET2 and TET3 had no significant effect on global BATF binding (Fig. 6B and fig. S8G), BATF was required for 5hmC modification at TetE1 (Fig. 6E). Composite bZIP:IRF motifs and AP-1 motifs that support BATF:JUN:IRF and BATF:JUN cooperation, respectively, were enriched in our genome-wide 5hmC, ATAC, and DNA methylation datasets (Figs. 1F and 2F and figs. S1G and S7E), consistent with previous findings that B cells lacking BATF or IRF4/IRF8 show impaired Aicda induction and CSR (39, 40, 45, 46, 53). We propose that together with additional transcription factors, BATF:JunB and BATF:IRF facilitate the recruitment of TET proteins as well as chromatin remodeling complexes to diverse enhancers including the Aicda enhancers TetE1 and TetE2 in activated B cells, thereby promoting enhancer accessibility, 5hmC deposition, and DNA demethylation. We note, however, that the phenotype of Batf-KO mice and B cells is considerably more severe than that of Tet2/3 DKO mice and B cells. Thus, TET2/3 proteins likely function as one of several modulators of BATF function in CSR.

Our data emphasize the utility of 5hmC mapping for easy, one-step analysis of transcriptional and epigenetic landscapes in any cell type of interest. 5hmC is a quintessential epigenetic modification that is most highly enriched at active enhancers and in the gene bodies of highly transcribed genes; thus, the relative levels of 5hmC at enhancers and gene bodies provide good estimates of enhancer function and the magnitude of transcription, respectively (51). 5hmC mapping by CMS-IP sufficed to identify all known enhancers in the Aicda locus, in a manner that was superior to both H3K27Ac and TET2 ChIP-seq and changes in 5hmC identified enhancers relevant to any particular process of cell activation or differentiation separately from all enhancers in the genome. Given its high chemical stability, the fact that its measurement requires only purified DNA and the availability of methods for its sensitive and specific detection, 5hmC is an appealing epigenetic mark for studying gene regulation. Overall, 5hmC distribution provides information analogous to that from ATAC-seq and ChIP-seq for enhancer histone modifications, and moreover requires only DNA and not live cells, effectively providing a transcriptional history of any given cell type written in DNA. If the genome is akin to an encyclopedia, 5hmC highlights those entries most relevant to a particular biological process.


Study design

The aim of this study was to characterize the function of TET enzymes and TET-mediated 5hmC modification in regulating B cell function. The genome-wide modification of 5hmC was analyzed as a function of time after B cell activation in vitro. To analyze the role of TET proteins in B cells in vitro and in vivo, Tet2 and Tet3 were acutely deleted using a tamoxifen-inducible system. The impact of TET loss of function on B cells was analyzed by flow cytometry, RNA-seq, and ATAC-seq. Potential TET-regulated regulatory elements and the transcription factors involved were identified on the basis of integrative genome-wide analyses.


Tet2fl/fl and Tet3fl/fl mice were generated as previously described (54, 55). C57BL/6J (000664), Ubc-CreERT2 (008085; located at Ndor1 locus; described as CreERT2 herein), Rosa26-LSL-EYFP (006148), and AID-Cre (007770) were obtained from the Jackson Laboratory. The LSL (LoxP-STOP-LoxP) cassette in Rosa26-LSL-EYFP mice contains a strong transcriptional stop flanked by two LoxP sites and was used as an indicator of the Cre activity. All mice used were 8 to 16 weeks in the C57BL/6 background and kept in a specific pathogen-free animal facility at La Jolla Institute and were used according to protocols approved by the Institutional Animal Care and Use Committee. To induce CreERT2-mediated deletion, we intraperitoneally injected Cre-expressing and control mice with 2 mg of tamoxifen (Sigma) dissolved in 100 μl of corn oil (Sigma) daily for 5 days. H. Singh provided splenocytes from Cd19-Cre Irf4-flox mice.

B cell isolation and CSR

B cells were isolated with the EasySep Mouse B cell isolation kit (STEMCELL Technologies, Canada) from splenocytes. To induce CSR from IgM to IgG1, B cells (5 × 105 to 1 × 106 cells/ml) were activated with LPS (25 μg/ml) from Escherichia coli O55:B5 (Sigma, St. Louis, MO) and rmIL-4 (10 ng/ml) at 37°C (5% CO2). For IgA switching, cells were activated with anti-CD40 (1 μg/ml, clone 1C10, BioLegend), rmIL-4 (10 ng/ml, Peprotech), rmIL-5 (10 ng/ml, Peprotech), and rhTGFβ1 (transforming growth factor β 1, 1 ng/ml). Media were composed of RPMI 1640 (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% FBS, 1× minimum essential medium (MEM) non-essential amino acids, 10 mM Hepes, 2 mM Glutamax, 1 mM sodium pyruvate, 55 μM 2-mercaptoethanol, and gentamicin (50 μg/ml) (all from Thermo Fisher Scientific, Waltham, MA, USA). To enhance CreERT2-mediated deletion, we cultured cells from CreERT2 mice in the presence of 1 μM 4-hydroxytamoxifen (Tocris). All cytokines used above were from PeproTech (Rocky Hill, NJ, USA).


For NP-OVA (Biosearch) immunization, the hapten-conjugated protein was diluted to 1 mg/ml in PBS, mixed with 1 volume of Alhydrogel (Invivogen), and injected into hind footpads (10 μg in 20 μl per injection). GC response was analyzed 7 days later, and the two draining popliteal lymph nodes were pooled for analysis. Hapten-specific B cells were identified by positive staining with NP-phycoerythrin (Biosearch Technologies).

Retroviral transduction and two-step CSR

Retrovirus was produced by transfecting PlatE cells with murine stem cell virus–based retroviral vectors and pCL-Eco. Naïve B cells were stimulated with F(ab′)2 goat antimouse IgM (5 μg/ml; Jackson ImmunoResearch) and LPS (10 μg/ml) at 1 × 106 cells/ml for 24 to 48 hours. Note that this stimulation condition inhibits CSR, and AID expression was delayed (56). Retrovirus was added to the cells in the presence of 20 mM Hepes and Polybrene (0.8 μg/ml; Millipore) and centrifuged at 3000 rpm at 32°C for 90 min. Cells were transferred back to a 37°C, 5% CO2 incubator for another 24 hours. To induce CSR, cells were washed once with warm media and activated with LPS and IL-4 as above for 48 hours. Under this condition, CSR was inhibited and started to class-switch only after LPS/IL-4 activation.

Flow cytometry

Primary cells and in vitro cultured cells were stained in fluorescence-activated cell sorting (FACS) buffer (1% bovine serum albumin, 1 mM EDTA, and 0.05% sodium azide in PBS) with indicated antibodies for 30 min on ice. Cells were washed and then fixed with 1% paraformaldehyde (diluted from 4% with PBS; Affymetrix) before flow cytometric analysis using FACS Canto II and FACS LSR II (BD Biosciences). Antibodies and dye were from BioLegend, eBioscience, and BD Pharmingen. Data were analyzed with FlowJo (FlowJo LLC, Ashland, OR, USA).


Proteins isolated from B cells with radioimmunoprecipitation assay (RIPA) buffer were resolved using NuPAGE 4 to 12% bis-tris gel (Thermo Fisher Scientific) and transferred from gel to polyvinylidene difluoride membrane using Wet/Tank Blotting Systems (Bio-Rad). Membrane was blocked with 5% nonfat milk (Bob’s Red Mill) in TBSTE buffer [50 mM tris-HCl (pH 7.4), 150 mM NaCl, 0.05% Tween 20, and 1 mM EDTA] and incubated with indicated primary antibodies, followed by secondary antibodies conjugated with horseradish peroxidase (HRP), and signal was detected with enhanced chemiluminescence reagents and x-ray film.


Coimmunoprecipitation was performed similar to previously described (4). Briefly, in vitro activated B cells (48 hours) were washed twice with cold PBS and then resuspended in swelling buffer [5 mM tris (pH 7.5), 2 mM MgCl2, and 3 mM CaCl2] at 10 × 106 cells/ml. After 10 min on-ice incubation, cells were pelleted (400g, 5 min) and resuspended in swelling buffer with 10% glycerol. An equal volume of lysis buffer (1% NP-40 in swelling buffer with 10% glycerol) was added to the cells with constant mixing. Cells were incubated on ice for 5 min, pelleted (400g, 5 min) and resuspended in buffer C [10 mM Hepes (pH 7.9), 400 mM NaCl, and 1 mM EDTA] supplemented with benzonase (500 U/ml; Sigma) at 10 × 106 cells/ml, and incubated at 4°C for 30 min with constant mixing. Debris was removed by centrifuge at 13,000 rpm for 10 min, and supernatant (nuclear fraction) was recovered. An equal volume of 2× conversion buffer [10 mM tris-HCl (pH 7.5), 280 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.2% sodium deoxycholate, and 0.2% Triton X-100] was added to the nuclear proteins. For immunoprecipitation, 10 μg of rabbit anti-TET2 (Abcam) or control rabbit Ig (Santa Cruz Biotechnology) was added to the nuclear extract with 30 μl of protein A Dynabeads (Thermo Fisher Scientific) in the presence of benzonase (500 U/ml) and ethidium bromide (10 μg/μl), both of which inhibit the indirect binding between nuclear proteins via DNA. Reaction was carried out overnight at 4°C and washed three times with RIPA buffer without SDS [50 mM tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, 0.5% sodium deoxycholate, and 1% NP-40]. Proteins were eluted from beads by heating at 70°C for 10 min with 1× LDS (lithium dodecyl sulfate) sample buffers (Thermo Fisher Scientific) with 10% 2-mercaptoethanol (Sigma). Immunoprecipitated proteins were analyzed as described above using immunoblotting with Rabbit TrueBlot Anti-Rabbit IgG HRP (Rockland) as secondary antibody.

RNA extraction, cDNA synthesis, and qRT-PCR

Total RNA was isolated with RNeasy Plus Kit (Qiagen, Germany) or with Trizol (Thermo Fisher Scientific, Waltham, MA, USA) following the manufacturers’ instructions. cDNA was synthesized using SuperScript III reverse transcriptase (Thermo Fisher Scientific), and qRT-PCR was performed using FastStart Universal SYBR Green Master Mix (Roche, Germany) on a StepOnePlus Real-time PCR system (Thermo Fisher Scientific). Gene expression was normalized to Gapdh. Primers are listed in table S3.

BS and oxBS sequencing

The BS and oxBS procedures were performed as previously described (22). Briefly, three PCR products containing C, mC, or hmC pertaining to different regions of ʎ phage genome were used as spike-ins at a ratio of 1:200 of the genomic DNA. A total of 1.5 μg of genomic DNA mixed with spike-ins was ethanol-precipitated, of which 1 μg of the DNA was oxidized with potassium perruthenate (KRuO4; Sigma) before BS treatment (for oxBS) using the MethylCode Bisulfite Conversion Kit (Thermo Fisher Scientific), and 0.5 μg of DNA was directly used for BS treatment. The BS- and oxBS-treated DNA were amplified using respective PCR primers as well as primers specific to the spike-in PCR products with KAPA Uracil+ PCR mix (Roche). The amplified products were pooled and libraries were prepared using the NEB Ultra II library preparation kit (NEB) according to the manufacturer. The libraries were sequenced paired-end 250 base pairs (bp) by 250 bp using MiSeq with the MiSeq reagent kit v2 (500 cycles; Illumina). Primers are listed in table S3.

Genome-wide 5hmC mapping by CMS-IP

Techniques for immunoprecipitation of DNA with antibodies are plagued by the fact that Igs nonspecifically immunoprecipitate DNA sequences containing CA and other DNA repeats (57). However, CMS is a derivative formed only after the reaction of 5hmC with sodium bisulfite, and anti-CMS antibodies recognize CMS with strong sensitivity and selectivity (19). Lentini et al. showed that the CMS-IP method is free of the nonspecific background of immunoprecipitation of DNA repeats, most likely because bisulfite treatment results in C > T conversions and/or single-stranded, as supposed to double-stranded, DNA was used for IP (57).

CMS-IP was performed essentially as previously described (4, 18, 19). Briefly, genomic DNA isolated from naïve and activated B cells was spiked with unmethylated lambda phage cI857 Sam7 DNA (Promega, Madison, WI, USA) and a PCR amplicon from a puromycin-resistant gene at a ratio of 200:1 and 100,000:1, respectively. DNA (5 to 10 μg in 130 μl tris-EDTA buffer) was sheared with a Covaris E220 using microTUBE for 4 min. DNA was cleaned up with Ampure XP beads, processed with NEBNext End Repair and A-tail Modules (NEB, Ipswich, MA, USA), and ligated to methylated Illumina adaptors (NEB). DNA was then bisulfite-treated (MethylCode, Thermo Fisher Scientific), denatured, and immunoprecipitated with anti-CMS serum (in-house) and a mixture of protein A and G Dynabeads (Thermo Fisher Scientific). Because our goal for this study was to identify regions that undergo significant activation-induced 5hmC modification after LPS + IL-4 stimulation, we normalized the data within each sample using total sequencing read counts for each individual time point, without using 5hmC-containing oligonucleotides as spike-ins to take into account the progressive dilution of 5hmC that occurs as a function of cell division (fig. S2C). Libraries for immunoprecipitated DNA were generated by PCR with barcoded primers (NEBNext Multiplex Oligos for Illumina; NEB) for 15 cycles using KAPA HiFi HotStart Uracil+ ReadyMix (Roche), followed by a cleanup with Ampure XP beads (Beckman Coulter), and sequenced with a HiSeq 2500 (Illumina, San Diego, CA, USA) with paired-end 50-bp reads.

Locus-specific analysis of 5hmC with Aba SI–qPCR

Genomic DNA (200 ng) was treated with T4 β-glucosyltransferase (Thermo Fisher Scientific) in the presence of UDP-glucose to glycosylate 5hmC at 37°C overnight. Half of the reaction was digested with Aba SI (NEB), which is specifically active for glycosylated 5hmC for 4 hours at 25°C followed by 15 min at 65°C to inactivate enzymes. The uncut sample was processed as above without the addition of Aba SI. Equal amount of DNA from the above reactions was used as template for RT-PCR as described for RNA qRT-PCR using primers TetE1-CMS-qF and TetE1-CMS-qR. Data shown were normalized to the signal obtained from WT naïve B cells. To monitor the degree of digestion, samples were spiked in 1 pg of control DNA with a single 5hmC-modified CpG (EpiMark 5hmC and 5mC Analysis Kit; NEB). The relative amount of 5hmC was calculated by the percentage of decrease in qPCR signals in the digested portion relative to the undigested portion. As a control to monitor nonspecific digestion, a genomic region containing CpG motifs but without 5hmC modification in B cells (Foxp3 CNS2) was amplified with Foxp3-CNS2-qF and Foxp3-CNS2-qR. Primers are listed in table S3.

DNA dot blot

DNA dot blot was performed as previously described (4, 22). To analyze 5hmC abundance, genomic DNA was treated with sodium bisulfite as above. DNA was diluted twofold serially with TE buffer, denatured in 0.4 M sodium hydroxide and 10 mM EDTA at 95°C for 10 min, and then immediately chilled on ice. An equal volume of ice-cold 2 M ammonium acetate (pH 7.0) was added and incubated on ice for 10 min. Denatured DNA were spotted on a nitrocellulose membrane using a Bio-Dot apparatus (Bio-Rad), washed with 2× SSC buffer (300 mM NaCl and 30 mM sodium citrate), and baked in a vacuum oven at 80°C for 2 hours. To detect CMS, we rehydrated membrane with TBSTE buffer and blocked it with 5% nonfat milk (Bob’s Red Mill) in TBSTE buffer. CMS was detected with primary rabbit anti-CMS antisera (in-house) following the procedures above for immunoblotting.

Chromatin immunoprecipitation sequencing

ChIP was performed as described before (4). Briefly, cells were fixed with 1% formaldehyde (Thermo Fisher Scientific) at room temperature for 10 min at 1 × 106 cell/ml in media, quenched with 125 mM glycine, and washed twice with ice-cold PBS. Cells were pelleted, snap-frozen with liquid nitrogen, and stored at −80°C until use. For TET2-ChIP, activated cells were centrifuged at 250g for 5 min, and cell pellets were resuspended in 37°C PBS with 2 mM disuccinimidyl glutarate to cross-link proteins for 30 min at room temperature. Formaldehyde was added to a final concentration of 1%, and the cells were incubated at room temperature for 10 min with nutation. Quenching and cell storage were performed as above. To isolate nuclei for sonication, cell pellets were thawed on ice, lysed with lysis buffer [50 mM Hepes (pH 7.5), 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40, and 0.25% Triton X-100] for 10 min at 4°C with rotation, and washed once with washing buffer [10 mM tris-HCl (pH 8.0), 200 mM NaCl, 1 mM EDTA, and 0.5 mM EGTA] and twice with shearing buffer [10 mM tris-HCl (pH 8.0), 1 mM EDTA, and 0.1% SDS]. Nuclei were resuspended in 1 ml of shearing buffer and sonicated with Covaris E220 using 1-ml milliTUBE (Covaris, Woburn, MA) for 18 to 20 min (duty cycle, 5; intensity, 140 W; cycles per burst, 200). After sonication, insoluble debris was removed by centrifugation at 20,000g. Buffer for chromatin was adjusted with 1 volume of 2× conversion buffer [10 mM tris-HCl (pH 7.5), 280 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.2% sodium deoxycholate, 0.2% Triton X-100, and 1% Halt protease inhibitors with (for H3K27Ac) or without (for BATF, TET2) 0.1% SDS]. Chromatin was precleared with washed protein A (Thermo Fisher Scientific) for 2 hours and incubated with antibodies and protein A Dynabeads overnight (all procedures were at 4°C with rotation). For H3K27Ac ChIP, bead-bound chromatin was washed twice with RIPA buffer [50 mM tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, 0.5% sodium deoxycholate, 1% NP-40, and 0.1% SDS], once with high-salt wash buffer [50 mM tris-HCl (pH 8.0), 500 mM NaCl, 1 mM EDTA, 1% NP-40, and 0.1% SDS], and once with TE [10 mM tris-HCl (pH 8.0) and 1 mM EDTA]. For BATF ChIP, all wash buffers were as above but without SDS. For TET2 ChIP, beads were washed three times with RIPA buffer without SDS and once with TE. Chromatin was eluted from beads with elution buffer (100 mM NaHCO3, 1% SDS, and 1 mg/ml RNase A; Qiagen) twice for 30 min each at 37°C with constant shaking. NaCl and proteinase K (Ambion) were added to the eluted chromatin at concentrations of 250 mM and 0.5 mg/ml, respectively, and de-cross-linked at 65°C overnight with constant shaking. DNA was purified with Zymo ChIP DNA Clean & Concentrator Capped Column (Zymo Research, Irvine, CA, USA). Library was prepared with NEB Ultra II Library Prep Kit (NEB) following the manufacturer’s instruction and was sequenced on an Illumina HiSeq 2500 (single-end 50-bp reads).


Procedures were as previously described (4). Briefly, 50,000 cells were collected by centrifugation and washed once with 50 μl of ice-cold PBS and centrifuged at 600g for 5 min at 4°C. Cell pellets were resuspended in 50 μl of cold lysis buffer [10 mM tris-HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl2, and 0.1% IGEPAL CA-630] and spun down immediately at 500g for 10 min at 4°C. Supernatant was discarded and nuclei were resuspended in 50 μl of transposition reaction mix (25 μl of 1× TD buffer from Illumina, 2.5 μl of Tn5 transposase, and 22.5 μl of H2O), incubated at 37°C for 30 min, and DNA was purified with a Qiagen MinElute kit (Qiagen). Library was amplified with KAPA HiFi HotStart Real-time PCR Master Mix (Roche) using indexed primers and sequenced on an Illumina HiSeq 2500 (paired-end 50-bp reads).

RNA sequencing with Smart-seq

Smart-seq was performed as described previously (28, 58). Briefly, total RNA was isolated from naïve and activated B cells with Trizol (Thermo Fisher Scientific), and the integrity of the RNA was accessed with TapeStation RNA analysis ScreenTape or BioAnalyzer RNA Pico Kit (Agilent). Ten nanograms of RNA was reverse-transcribed using oligo-dT30 VN primer in the presence of Template Switching Oligo (TSO) with SuperScript II reverse transcriptase. cDNA was preamplified with IS PCR primers, and PCR products were cleaned up with Ampure XP beads. One nanogram of the PCR product was used to generate library using Nextera XT Library Prep Kit (Illumina), and tagmentated DNA was amplified for a 12-cycle PCR and purified with Ampure XP beads. Libraries were sequenced on an Illumina HiSeq 2500 with single-end 50-bp reads.

Statistical analyses

Statistical analyses and bar plots were performed and plotted with Prism 7 or R (v3.3.3). The bar graph and dot plots shown indicate mean and SE. Most experiments were analyzed using two-tailed unpaired t test or Wilcoxon rank-sum test, as indicated in the figure legends unless otherwise stated.


Materials and Methods

Fig. S1. TET-mediated DNA hydroxymethylation correlates with demethylation and enhancer activity.

Fig. S2. Phenotypic features of WT and Tet2/3 DKO B cells.

Fig. S3. TET family proteins are important for B cell–intrinsic CSR.

Fig. S4. Decreased Aicda expression in Tet2/3 DKO B cells.

Fig. S5. TET-responsive element TetE1 regulates CSR and Aicda mRNA expression in the CH12 B cell line.

Fig. S6. TET proteins sustain enhancer accessibility.

Fig. S7. Analysis of TET-dependent accessible regions.

Fig. S8. AP-1 proteins in activated B cells.

Table S1. Differentially expressed genes between WT and Tet2/3 DKO B cells.

Table S2. Time-series analysis of RNA expression in WT B cells.

Table S3. Primer sequences.

Table S4. Reagents and resources.

Table S5. Raw data.

References (5973)


Acknowledgments: We would like to thank U. Basu for providing the AID antibody; K. Yu for Aicda-KO CH12F3 cells; H. Singh for splenocytes from Cd19-cre Irf4-flox mice; P. Casali and H. Zan for discussions; L. Hempleman for assisting with animal experiments; C. Kim, L. Nosworthy, D. Hinz, and R. Simmons (LJI Flow Cytometry Core) for cell sorting; J. Day and N. Wlodychak (LJI Functional Genomics Center) for assistance with next-generation sequencing; and S. Bélanger for advice on immunization. Funding: C.-W.J.L. was supported by a Cancer Research Institute Irvington Postdoctoral Fellowship and the Independent Investigator Fund (Kyowa Hakko Kirin/LJI). V.S. was supported by a Leukemia and Lymphoma Society Postdoctoral Fellowship (grant ID: 5463-18). D.S.-C. and E.G.A. were supported by the CONACYT/UCMEXUS fellowship from Mexico-US. This work was supported by the National Institutes of Health (NIH) grants R35 CA210043, R01 AI109842, and AI128589 (to A.R.) and R01 AI127642 (to D.G.S.). F.A. and A.C. were partially supported by Institute Leadership Funds from the LJI and by NIH grants R01 MH111267 and R35 GM128938. Purchase of the Illumina HiSeq 2500 and the BD FACSAria II was supported by equipment grants NIH S10OD016262 and NIH S10RR027366, respectively. Author contributions: C.-W.J.L. and V.S. conceptualized experiments, acquired and analyzed the data, and performed statistical analyses for in vitro and in vivo experiments. D.S.-C. and E.G.-A. performed the majority of the bioinformatics and related statistical analyses (ATAC-seq, CMS-IP, enhancer analysis, WGBS, oxBS-seq, and ChIP-seq) and proofread the manuscript. A.C. performed the expression clustering analysis (TC-seq) and advised on RNA-seq analysis. C.-W.J.L. performed initial bioinformatics analyses, motif analysis, and data visualization. X.Y. provided key reagents and assistance for oxBS-seq experiments. F.A. supervised the bioinformatics analysis and reviewed the manuscript. D.G.S. provided advice and key reagents, interpreted data, and reviewed the manuscript. A.R. supervised and interpreted the data. C.-W.J.L., V.S., and A.R. wrote the manuscript. Competing interests: A.R. is a member of the Scientific Advisory Board of Cambridge Epigenetix. The other authors declare no competing financial interests. Data and materials availability: Data for CMS-IP, ChIP-seq, ATAC-seq, RNA-seq, and oxBS-seq have been deposited to the National Center for Biotechnology Information Gene Expression Omnibus (accession number GSE116208). All (other) data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Immunology

Navigate This Article