Role of Dot1L and H3K79 methylation in regulating somatic hypermutation of immunoglobulin genes

Significance

The somatic hypermutation of immunoglobulin (Ig) genes is usually required to make antibodies that protect us from foreign and toxic substances. This mutational process is mediated by activation-induced deaminase (AID) and is targeted to the variable (V) part of the Ig genes that encode the antigen-binding site of antibody molecules, leading to higher affinity and more effective antibodies. AID acts on single-stranded DNA that is created during the transcription of the V region. Here, we show that multiple factors that are involved in the release, from pausing to elongation of transcription, contribute to the hypermutation process, supporting the likelihood that transcriptional pausing and elongation make V region single-stranded DNA available to AID-induced mutations.

Abstract

Somatic hypermutation (SHM) and class-switch recombination (CSR) of the immunoglobulin (Ig) genes allow B cells to make antibodies that protect us against a wide variety of pathogens. SHM is mediated by activation-induced deaminase (AID), occurs at a million times higher frequency than other mutations in the mammalian genome, and is largely restricted to the variable (V) and switch (S) regions of Ig genes. Using the Ramos human Burkitt’s lymphoma cell line, we find that H3K79me2/3 and its methyltransferase Dot1L are more abundant on the V region than on the constant (C) region, which does not undergo mutation. In primary naïve mouse B cells examined ex vivo, the H3K79me2/3 modification appears constitutively in the donor Sμ and is inducible in the recipient Sγ1 upon CSR stimulation. Knockout and inhibition of Dot1L in Ramos cells significantly reduces V region mutation and the abundance of H3K79me2/3 on the V region and is associated with a decrease of polymerase II (Pol II) and its S2 phosphorylated form at the IgH locus. Knockout of Dot1L also decreases the abundance of BRD4 and CDK9 (a subunit of the P-TEFb complex) on the V region, and this is accompanied by decreased nascent transcripts throughout the IgH gene. Treatment with JQ1 (inhibitor of BRD4) or DRB (inhibitor of CDK9) decreases SHM and the abundance of Pol II S2P at the IgH locus. Since all these factors play a role in transcription elongation, our studies reinforce the idea that the chromatin context and dynamics of transcription are critical for SHM.

Somatic hypermutation (SHM) and class-switch recombination (CSR) of immunoglobulin (Ig) genes occur in germinal centers allowing B cells to express high-affinity antibodies that protect us from various foreign antigens. Mutations caused by activation-induced deaminase (AID) are largely restricted to antibody/Ig variable (V) regions and switch (S) regions to independently initiate V region SHM or constant (C) region isotype switching. The AID-induced mutation rate of human and mouse Ig genes is a million times higher than the rate of other mutations throughout the genome (1). In B cells, AID off-target mutation occurs in non-Ig genes but at a lower rate than in Ig genes, and leads to translocations in loci such as Bcl-6 and c-Myc causing B cell malignancies (2).

AID converts cytosine to uracil preferentially at WRC (W = A/T, R = A/G, mutated bases underlined) hotspots and specially at overlapping hotspots, such as AGCT, which contain AID hotspots opposite each other on both strands (3). AID-induced G:U mismatches or abasic sites are subsequently repaired by replication, base-excision repair, or mismatch repair pathways (4, 5). In the case of error-prone mismatch repair, polymerase η (Pol η) generates the mutations that largely occur at A:T sites (1). RNA polymerase II (Pol II)–mediated transcription of Ig genes is required for efficient SHM, presumably by exposing single-stranded DNA (ssDNA) that is the substrate for AID (4, 6). Members of many transcription-related complexes, such as Spt5 (subunit of DSIF complex) and SSRP1 (subunit of FACT), play important roles in SHM (7–9). A recent study revealed that many of the distinctive characteristics of SHM, including the relative distribution and frequency of mutation of hotspots in a particular human V region, can be duplicated biochemically using transcription by core members of the human RNA Pol II complex along with purified AID (10). However, in the chromatin in intact cells, these processes are more complicated and are regulated by the Ig superenhancers that undergo both sense and antisense transcription and interact with the V and S regions (11–13). These interactions can lead to double-stranded DNA breaks and chromosomal rearrangements that seldom occur in V regions. There is also considerable evidence that stalling and elongation of Pol II contributes to AID-induced mutations (14–16). In addition, a number of factors involved in RNA processing—including the RNA exosome (17, 18), long-noncoding RNAs (16, 19), splicing regulator SRSF1-3 (20), and the RNA binding factor ROD1 (21)—contribute to the regulation of AID-induced mutations. However, all of these factors are associated with highly transcribed genes throughout the genome and it is still unclear how AID-induced mutations are largely restricted to the Ig V and S regions.

While it is clear that local DNA context, such as hotspots and various AID-associated factors, are very important in mediating AID mutations, there must be some additional factors that restrict AID targeting to the V region. Changes in chromatin modifications have been reported to be associated with and play a role in AID-induced hypermutation (22). We screened for chromatin modifications that are enriched in and distinguish the variable region. Among those, H3K79me was particularly interesting because genome-wide studies reveal that H3K79me2/3 is enriched in highly transcribed genes and is often found downstream of the transcription start site (TSS), where it has an important role in transcription elongation (23). In addition, H3K79me has a number of distinct characteristics. It is located in the core globular domain of H3 rather than in the histone tail, where most of the modifications are found. Dot1L is a distributive enzyme that sequentially adds mono-, di-, and trimethylation to H3K79. Most importantly, it is the only known methyltransferase (HMT) for H3K79. Dot1L is not only unusual in its specificity for H3K79 but it differs from other methyltransferases in that it does not contain the SET catalytic domain found in all of the other methyltransferases (23). It also differs from other HMTs in that it modifies H3K79 in the context of intact chromatin but not as free histone or histone peptides, suggesting that it requires cross-talk with other histone modifications (24). Most importantly, genome-wide studies in B cells have found an increased abundance of H3K79me2 in AID off-target sites (25, 26).

While Dot1L has very recently been shown to have a role in B cell differentiation and in CSR, and it is highly expressed in germinal center B cells (27), its role and that of H3K79me in V region mutation has not been examined in detail. To study this, we used the human Ramos Burkitt’s lymphoma germinal center-like B cell line that expresses AID and undergoes V region mutations (28). Although this cell line mutates at a lower rate than primary B cells in vivo, it mimics the pattern of AID-induced V region mutations seen in normal human B cells (29). We found that H3K79me2/3 was enriched in the V region and Sμ region compared with the unmutated C region. Inhibition and knockout (KO) of Dot1L in the Ramos cell line demonstrated that Dot1L and H3K79me are associated with V region SHM. Using primary mouse B cells stimulated ex vivo, we also found H3K79me2/3 was related to downstream constant region switching. The KO of Dot1L resulted in an even greater decrease in SHM than the inhibitors and the KO reduced the levels of nascent RNA transcription and the abundance of transcription elongation complexes in the IgH locus. These findings reveal that H3K79 methylation plays a role in the regulation of SHM and extends our knowledge of the changes in chromatin context and dynamics of transcription associated with this process.

Results

The Ig V Region Is Enriched for H3K79me2/3.

In seeking markers that might contribute to the targeting of AID to the V region, we looked for chromatin modifications that were more abundant on the mutating V region than the nonmutating downstream IgM constant (Cμ) region in the Ramos cell line. Ramos is a human Burkitt’s lymphoma germinal center-like cell line that constitutively expresses AID and mutates its endogenous IGHV4-34 region (28). Using chromatin immunoprecipitation (ChIP)-qPCR analysis, we found that H3K79me2/3 was more abundant on the V than the Cμ region (Fig. 1A, sites E and F compared with sites I and J). In this study, we used anti-H3K79me3 antibody, but since it slightly cross-reacts with H3K79me2, we called it H3K79me2/3 even though it is more specific to H3K79me3. In contrast, H3K79me1 was more abundant in the Cμ region than in the V region (Fig. 1B). In addition, H3K79me2/3, but not H3K79me1, was enriched at the 5′ end of the Sμ switch region (Fig. 1 A and B, site H), which undergoes high rates of “sterile” transcription and mutation in Ramos (12). H3K79me was not abundant 5′ to the V region exon (Fig. 1 A and B, sites A and B) that is close to the TSS, which does not undergo AID-induced mutations (30). The differences in H3K79 methylation were not due to variation in H3 itself since it had a relatively constant abundance between the V and C regions (Fig. 1C). Moreover, the abundance of Dot1L (Fig. 1D), which is the specific methyltransferase for H3K79, spatially correlated with the abundance of H3K79me2/3 in the IgH gene, including not being abundant near the TSS. Furthermore, H2BK120ub, which stabilizes Dot1L (24), is also associated with the V region in Ramos cells, although it is even slightly higher in the C region (Fig. 1E). The gene body of BTG1, which has been reported to be an off-target site of AID in mouse B cells (26, 31), accumulated both H3K79me1 and H3K79me2/3, while nontranscribed CD4 showed no detectable H3K79 methylation (Fig. 1 A and B). We also examined the abundance of other histone modifications that have been associated with off-target sites that are mutated by AID (25, 26). The active transcription marks H3K4me3 and H3K27ac were enriched in the V region exon and in the Sμ region (SI Appendix, Fig. S1 A and B). These marks are also enriched in the AID off-target site BTG1. H3K36me3, which is associated with transcription elongation and mismatch repair, was found to increase toward 3′ portion of the Ig gene body (SI Appendix, Fig. S1C), as has been reported for other highly transcribed genes genome-wide (32).

Fig. 1.

Dot1L and H3K79me2/3 are enriched in immunoglobulin variable and switch regions. (A–E) Distributions of H3K79me2/3, H3K79me1, H3, Dot1L and H2BK120ub in the Ig heavy-chain locus in Ramos cells were determined by ChIP-qPCR assays. H3K79me2/3 was detected by an anti-H3K79me3 antibody but labeled 2/3 due to its putative cross-reactivity to H3K79me2. For all the ChIP experiments throughout the paper, the results were normalized to percentage of input DNA and the nonspecific IgG control was subtracted. Error bars in figures represent the SD among three independent experiments. Amplification regions are indicated as letters in the map on the left at the bottom. (F) The abundance of H3K79me2/3 across the IgH gene in splenic mouse B cells before or after stimulation ex vivo with 50 μg/mL LPS plus 50 ng/mL IL-4 was measured by ChIP assays. Error bars represent the SD between two independent experiments each using two mice. *P < 0.05 and **P < 0.01. Amplification regions are indicated in the right-hand map at the bottom.

A number of factors that associate with AID and are involved in mutation, such as Pol II, ROD1 (21), and the RNA exosome (17) have been shown to be present on the V region prior to the recruitment of AID. We therefore asked whether H3K79me2/3 abundance is independent of AID recruitment by comparing the abundance and location of H3K79me2/3 in AID-overexpressing Ramos cells with cells that express a very low level of endogenous AID and do not undergo detectable V region mutation (33) (SI Appendix, Fig. S2A). We also measured the abundance and location of H3K79me2/3 in a reporter Ramos cell line in which both an mCherry and the endogenous VH4-34 are fused, providing a larger target (34) (more information in the next section). In this cell line the AID is restricted to the cytoplasm until it is induced to transfer to the nucleus with 4-Hydroxytamoxifen (4-OHT) (SI Appendix, Fig. S2B). In both situations the distribution and abundance of H3K79me2/3 across the IgH gene were not significantly different, suggesting that the generation of this modification on the V regions did not require either AID or the mutational process and that H3K79me would, like the other AID-associated factors, preexist on the chromatin prior to AID recruitment.

We further investigated H3K79 modification in the primary naïve mouse splenic B cells undergoing CSR from IgM to IgG1 through lipopolysaccharide (LPS) plus interleukin (IL)-4 stimulation ex vivo. The naïve splenic B cells highly transcribe their Sμ region to express IgM but do not transcribe the downstream γ1 switch region. H3K79me2/3 was more abundant on the transcribed activated Sμ region (Fig. 1F, sites L and M) than the unexpressed downstream receipt Sγ1 region (Fig. 1F, sites R and S), even though there were relatively few mutations in Sμ before stimulation (35). After 48 or 72 h of stimulation with LPS+IL-4, the cells underwent CSR (SI Appendix, Fig. S3 A and B) and actively expressed AID (SI Appendix, Fig. S3C). H3K79me2/3 is detected and increases at Sγ1 and Cγ1 (Fig. 1F, sites R to T), concurrent with switching to IgG1. In contrast, the Sγ3 and Cγ3 regions that are not activated by LPS+IL-4 stimulation maintain low levels of H3K79me2/3 (Fig. 1F, sites O to Q). These results not only suggest that H3K79me2/3 may be associated with CSR of Ig genes, but that this chromatin modification is induced in parallel with the process of AID mutations within the S regions of these primary B cells. We also observed a relatively high occupancy of H3K79me2/3 associated with the JH4 intron (Fig. 1F, site K), which undergoes mutations in Peyer’s patch germinal center B cells (35). This provides further evidence that H3K79me2/3 is associated with V region SHM.

Inhibition of Dot1L by EPZ004777 Reduces H3K79me2/3 and SHM in the V Region.

To explore the role of Dot1L and H3K79me in V region SHM, we used a reporter Ramos cell line that did not have easily detectable endogenous AID activity. The endogenous V region was replaced with an mCherry sequence containing several AID hotspots that was linked in frame to the 5′ end of the endogenous IGHV4-34 with an autodegradable T2A peptide (34). This reporter Ramos cell line was stably transfected with an ectopic AID-estrogen receptor (ER) so that AID was retained in the cytoplasm but could be induced with 4-OHT to enter the nucleus and mutate both mCherry and the IGHV4-34 region. Since AID mutates both the mCherry and the V region at approximately the same rate, the loss of mCherry could be used as a surrogate for V region mutation after only 7 d of 4-OHT induction (34). There is a background of ∼2.6% of mCherry loss population (SI Appendix, Fig. S4A), perhaps due to the leakage of some AID-ER from the cytoplasm into the nucleus and the low level of endogenous AID, but the loss of mCherry fluorescence is readily detectable with 4-OHT induction and not appreciably affected by the addition of DMSO (Fig. 2A), which is required to dissolve the inhibitors and add to the cells. There are several chemical compounds designed to bind Dot1L and compete with the methylation substrate SAM, and block Dot1L’s enzymatic activity. In this study, we used both the EPZ004777 and EPZ-5676 inhibitors of Dot1L, which have been used in preclinical or clinical trials to treat mixed-lineage leukemia (MLL)–rearranged leukemia (36, 37). The Ramos reporter cells were pretreated with the inhibitor for 24 h and then the transfer of AID into the nucleus was induced with 4-OHT for an additional 7 d in the presence of the Dot1L inhibitors. Treatment with either Dot1L inhibitor reduced the frequency of mutation in a dose-dependent manner compared with 4-OHT+DMSO as a control (Fig. 2A and SI Appendix, Fig. S4A). After subtracting the background, the intermediate dose of EPZ004777 (10 μM) reduced the frequency of mutation by ∼35%, while the highest dose of EPZ004777 (20 μM) decreased the mutation frequency by ∼47%.

Fig. 2.

Inhibition of Dot1L reduces SHM and decreases H3K79me2/3 in IgH gene. (A) Mutation frequency represented by the loss of mCherry fluorescent cells in the reporter Ramos cell line with 7 d of induction of AID-ER by 0.25 μM 4-OHT combined with DMSO or different doses of the Dot1L inhibitors (dissolved in DMSO). Error bars represent the variation among the three independent experiments. The frequency of mCherry⁻ cells of ∼2.6% in an aliquot that was not treated with 4-OHT served as a background control and was subtracted. (B) The overall frequency of mutation and frequency of mutations in different types of DNA motifs were analyzed by UMI-based next-generation sequencing, as described in Materials and Methods. DMSO-treated control cells were compared with cells treated with 20 μM EPZ004777 together with 7 d of 4-OHT induction. Except for the “Total sequence count,” the background was subtracted in all the calculations. For the calculation of “Mutation frequency,” “Mutated V regions,” and “Mutations/V region,” the corresponding absolute number was divided by the total numbers of nucleotides in each group. For the bottom part of the table, the respective absolute number of the mutated sites was divided by the total number of motifs or sites in each group. (C) Percentages of different cell cycle phases were detected by fluorescence-activated cell sorting (FACS) using propidium iodide staining for reporter Ramos cells treated with 0.25 μM 4-OHT plus DMSO alone or 20 μM EPZ004777 or 20 μM EPZ-5675. SD from three independent experiments is shown by error bars. (D) Different histone modifications including H3K79me were detected by Western blots using the histone fraction from reporter Ramos cells with 4-OHT induction combined with increasing doses of EPZ004777 treatment or DMSO alone. (E) The protein level of AID, IgM, and transcription-related factors was detected by Western blots using the whole-cell lysate from reporter Ramos cells with 4-OHT induction combined with increasing doses of EPZ004777 treatment or DMSO alone. (F and G) The abundance of H3K79me2/3 (F) and H3K79me1 (G) were measured by ChIP assays from cells treated with 4-OHT plus DMSO or 20 μM EPZ004777 in three independent experiments. PCR amplification sites are indicated in the map on the bottom. *P < 0.05, **P < 0.01, and ***P < 0.001.

To further confirm the decrease in frequency of mutation resulting from of Dot1L inhibition, we performed next-generation sequencing. Unique molecular identifiers (UMI) were added, which allows us to remove duplicate sequences due to PCR amplification and reduce the error rate of the sequencing process (Materials and Methods) (38). The deep sequencing revealed that treatment with 20 μM EPZ004777 resulted in a 27% decrease in the frequency of mutation compared with the DMSO control (Fig. 2B). This decrease is composed of a decrease in the percent of V regions that are mutated and in the frequency of mutation in the V regions with mutations (Fig. 2B). These can be considered as independent measures of mutation and confirm the decreases in mutation based on mCherry loss through FACS. In addition, the mutation rate in AID regular (WRC and GYW) and overlapping WGCW hotspots was also reduced upon Dot1L inhibitor treatment (Fig. 2B). Since AID-induced V region mutations are restricted to a brief period in G1 (39) and inhibitors of Dot1L do have some toxicity, we examined the cell cycles of the treated and untreated cells and did not see any differences, even at the highest dose (Fig. 2C and SI Appendix, Fig. S4B). The site-by-site mutation profile provides base pair resolution and reveals that the mutation frequency was decreased across the whole V region without dramatic changes of the mutation distribution after EPZ004777 treatment (SI Appendix, Fig. S5).

Consistent with the dose-dependent inhibition of SHM after Dot1L inhibitor treatment and the mono- to di- to trimethylation action of Dot1L, Western blots of the histone fraction from reporter Ramos cells showed a global dose-dependent decrease in H3K79me1, -me2, and -me3 (Fig. 2D). Different antibodies were used for H3K79me1, -me2, and -me3, so we cannot quantify the different effect on me1, me2, and me3. In contrast, EPZ004777 treatment did not affect the genome-wide methylation in the chromatin of other lysines, such as H3K4me3 and H3K36me3, confirming the specificity of Dot1L in H3K79 methylation in Ramos cells (Fig. 2D). Moreover, the inhibitor did not decrease the cellular levels of other transcription-related factors, such as Pol II, nor the expression of IgM and AID (Fig. 2E and SI Appendix, Fig. S4C), which are crucial for SHM. Importantly, we found the inhibition of Dot1L by EPZ004777 reduced H3K79me2/3 (Fig. 2F) abundance throughout the IgH locus, especially in the V region, while H3K79me1 was increased in the V region but decreased in the C region (Fig. 2G), confirming that EPZ004777 blocks Dot1L’s sequential methylation of H3K79 without completely inhibiting Dot1L allowing H3K79me1 to accumulate on the V region. In the off-target gene BTG1, both H3K79me1 and H3K79me2/3 were decreased by EPZ004777, but it is important to note that we only looked at one site on this gene. The reduction of H3K79me2/3 was not the result of reduced Dot1L levels since neither the abundance of the Dot1L protein across the Ig heavy-chain locus nor global expression of Dot1L was affected by EPZ004777 treatment (SI Appendix, Fig. S4 C and D). Additionally, the inhibitor treatment did not significantly affect the RNA level of mCherry and VH4-34 but it did decrease the RNA level of the C region (SI Appendix, Fig. S4C). This may be a result of the decrease of both for H3K79me1 and H3K79me2/3 on the C region, supporting the important role of H3K79me in Ig heavy-chain transcription. Taken together, these findings suggest that Dot1L and its target H3K79me are important for V region SHM. The continued presence of the Dot1L protein and of H3K79me1 could explain why the inhibition of Dot1L only partially blocked SHM and suggests that the H3K79me1 could also play a role in SHM.

KO of Dot1L Decreases SHM by Reducing both H3K79me1 and H3K79me2/3.

In order to independently confirm the role of Dot1L in SHM, we generated CRISPR/Cas9 KO of Dot1L in the same reporter Ramos cell line as described above. We designed guide RNA (gRNA) to target the first exon, which is a part of the catalytic domain of the Dot1L gene, hoping to completely disrupt Dot1L translation. We did not identify detectable Dot1L protein or methylated H3K79 in Dot1L-KO cells, and Dot1L-KO did not affect the expression of transcription-related factors such as Pol II, BRD4, CDK9, and Spt5 (Fig. 3A). While there was a slight reduction of IgM and AID expression (Fig. 3A and SI Appendix, Fig. S6A), the KO of Dot1L resulted in a much bigger ∼70% reduction of SHM as measured by the loss of mCherry after 7 d of 4-OHT treatment (Fig. 3B). This is a greater loss of SHM than with even the highest dose of EPZ004777 treatment, and provided an independent confirmation that Dot1L and H3K79me facilitate SHM in the Ramos cell line. We also observed a decrease of mCherry loss in the background of Dot1L-KO in which the cells were not treated with 4-OHT (Fig. 3B). This was predictable since, as we described above, the background comes from the mutations by leaked AID-ER and endogenous AID activity. As would be expected, the KO of Dot1L resulted in a virtual loss of Dot1L and H3K79me2/3 from the V region (Fig. 3 C and D). Although there appeared to be some H3K79me1 associated with the V region (Fig. 3E), we do not know if this is just background or some other factors can contribute to a small amount of H3K79me1. The important finding is that the apparently reciprocal increase in H3K79me1 seen with the inhibitor (Fig. 2G) did not occur in the KO. In summary, both the inhibitor treatment and KO of Dot1L experiments suggest H3K79me contributes to AID-targeted SHM of the V region.

Fig. 3.

KO of Dot1L reduces SHM and decreases H3K79me2/3 on the IgH gene. (A) Dot1L and H3K79me levels and the expression of other factors were detected by Western blots using the whole-cell lysate from WT or Dot1L-KO reporter Ramos cells with 7-d induction of 4-OHT. (B) Mutation frequency represented by the loss of mCherry in the reporter Ramos WT and Dot1L-KO cells with 7 d of 4-OHT induction. The data from three independent induction experiments are summarized. (C–E) The distribution of Dot1L, H3K79me1, and H3K79me2/3 were detected by three independent ChIP assays from WT or Dot1L-KO cells with error bars representing the SD. PCR amplification sites are indicated in the map at the bottom. *P < 0.05, **P < 0.01, and ***P < 0.001.

KO of Dot1L Decreases Nascent Transcription and Elongation of the IgH Locus.

Genome-wide studies established that H3K79me is usually localized within the gene body and positively correlates with active transcription (32). To address whether Dot1L and H3K79me promote SHM by affecting the rate of transcription, we first performed nascent RNA reverse-transcription quantitative polymerase chain reaction (RT-qPCR) to detect the transcription rate of the IgH locus. Using a recently published protocol (40), we first extracted chromatin-bound total RNA, then by depleting mRNA using Oligo dT magnetic beads, we purified nascent RNA and did RT-qPCR (Fig. 4A). The abundance of nascent transcripts decreases from the 5′ to the 3′ end of IgH locus and the KO of Dot1L reduced the nascent transcription compared with WT throughout the IgH gene (Fig. 4B), suggesting that Dot1L and presumably H3K79me contribute to maintaining a high rate of transcription at the IgH locus.

Fig. 4.

Dot1L-KO reduced nascent transcription of Ig heavy chain and decreased the abundance of transcription elongation related factors. (A) Methodology for how nascent RNA q-PCR was performed (also see Materials and Methods). (B) Relative expression of nascent RNA transcripts across the IgH gene was measured by real-time PCR using WT or Dot1L-KO cells with 4-OHT treatment. Three independent experiments were performed and the SD is shown by error bars. (C–F) The abundance of transcription related factors was measured by ChIP assays in three independent experiments. PCR amplicon sites are indicated in the map at the bottom. *P < 0.05, **P < 0.01, and ***P < 0.001.

Dot1L plays a critical role in transcription elongation presumably by its involvement with factors (41) from other elongation complexes, such as P-TEFb. The activity of P-TEFb is highly regulated by BRD4 since it can both recruit P-TEFb and stimulate the kinase activity of P-TEFb to phosphorylate the C-terminal domain of Pol II (32). This is especially interesting, since it has been proposed that pausing and elongation of Pol II plays a role in making ssDNA accessible to AID (15, 42). We examined the abundance of transcription elongation factors at the IgH locus in relation to Dot1L protein levels. In the Dot1L-KO cell line, the abundance of Pol II and Pol II S2P, which is the elongating form of Pol II, were significantly reduced throughout the IgH locus (Fig. 4 C and D), suggesting that Dot1L and H3K79me are required to maintain the rate of elongation. Furthermore, BRD4 and CDK9 (a subunit of P-TEFb) elongation complexes were both present at the V region exon (Fig. 4 E and F) and Sμ but not the C region, and their abundance was significantly decreased in the Dot1L-KO cell line. Consistent with this, EPZ004777 treatment also slightly reduced BRD4 and CDK9 abundance at the IgH locus (SI Appendix, Fig. S4 E and F). Taken together, these findings again support the idea that the transcription elongation rate was reduced by decreasing H3K79me in the Dot1L-KO cells and indicate that H3K79me may affect SHM by regulating transcription in the IgH locus. This finding is consistent with previous studies showing that AID-induced mutations are associated with the released elongating Pol II from stalling/pausing sites, and this could also be related to premature termination (14–16, 34).

One way to increase AID-induced mutations might be to increase the accessibility of the newly transcribed V region DNA. By ATAC-seq, we found that only 2.8% of the total identified peaks (310,860) were significantly affected by Dot1L-KO. Thus, Dot1L did not greatly affect the chromatin accessibility genome-wide, as shown in SI Appendix, Fig. S6B. This is not surprising because genome-wide studies showed that only a small number of genes are enriched for H3K79me (23). Furthermore, the chromatin accessibility of the IgH locus in the reporter Ramos cells was not changed by Dot1L-KO (SI Appendix, Fig. S6C), suggesting the effect of Dot1L-KO on SHM of the V region is not due to a change of chromatin accessibility in the IgH gene. Indeed, a recent study using H3K79me2/3 as a mark to identify putative enhancers found the majority of accessible chromatin regions are not affected by the loss of H3K79me2/3 when treated with Dot1L’s inhibitor (43). We conclude that the loss of Dot1L and H3K79me does not change global chromatin accessibility nor the accessibility of the IgH gene at least as measured by ATAC-seq in Ramos cells.

Inhibition of BRD4 or P-TEFb Decreases Mutation Rate of the V Region.

We further examined the role of transcription elongation and of BRD4 and P-TEFb in SHM in the Ramos system. We treated the reporter Ramos cells separately with inhibitors of BRD4 and CDK9. Although an earlier study using a different cell line and a much shorter treatment time did not find an effect of JQ1 (inhibitor of BRD4) on SHM (44), in the Ramos cell line JQ1 and DRB, which is an inhibitor of CDK9, individually reduced the mutation rate in a dose-dependent manner (Fig. 5A). Additionally, the combination of JQ1 or DRB with EPZ004777 further reduced V region mutation (Fig. 5B), confirming that BRD4 and P-TEFb complexes are associated with V region SHM. The treatments with these inhibitors did not affect the cell cycle (SI Appendix, Fig. S7), which is important for AID-induced mutation. JQ1 works to block the recruitment of BRD4 and DRB blocks the enzymatic activity of CDK9. Treatment of 50 nM JQ1 or 10 μM DRB did not affect the protein level of BRD4, CDK9, AID, or IgM (Fig. 5C). Although it has been shown that high-dose DRB treatment blocks the release of pausing to elongation, resulting in an increased Pol II density near the TSS (45), our results using a much lower dose but longer time found that the inhibition of BRD4 or CDK9 resulted in some reduction of Pol II at the mCherry/VH3-34 region and also reduction of its S2 phosphorylated form, especially in the V region but also in the C region (Fig. 5 D and E). Moreover, the abundance of BRD4 and CDK9 was decreased in the IgH gene when treated with JQ1, while there was a dramatic increase in BRD4 and CDK9 abundance when treated with DRB, even though DRB is an inhibitor of CDK9 (Fig. 5 F and G). This increased BRD4 and CDK9 may be recruited to the IgH locus to compensate for the inhibition of CDK9’s activity since CDK9 is a critical component of the transcription elongation complexes. In summary, these studies further support the role of Dot1L and its target H3K79me in facilitating SHM by affecting the transcription dynamics of the IgH gene.

Fig. 5.

Inhibition of BRD4 and P-TEFb reduces SHM. (A) The mutation frequency was measured by the loss of mCherry in cells treated with 4-OHT plus different concentrations of JQ1 or DRB compared with a DMSO control. Three independent experiments were performed to obtain SDs shown by error bars. (B) Mutation frequency was measured by FACS for mCherry loss with combinations of 10 μM EPZ004777 and 20 nM JQ1 or 5 μM DRB plus 4-OHT treatments using the reporter Ramos cell line. Data represent three independent experiments. (C) The expression of BRD4, CDK9, and other related factors was detected by Western blot using the whole-cell lysate from the reporter Ramos cell line treated with 4-OHT plus 50 nM JQ1 or 10 μM DRB compared with DMSO. (D–G) The abundance of transcription related factors was measured by ChIP assays in three independent experiments. Reporter Ramos cells were stimulated with 0.25 μM 4-OHT for 7 d combined with 50 nM JQ1 or 10 μM DRB. PCR amplification regions are indicated in the map at the bottom. *P < 0.05, **P < 0.01, and ***P < 0.001.

Discussion

Over the years many studies have suggested that aspects of the transcriptional process play a central role in the selective introduction of AID-induced mutations to the Ig V and S regions in order to generate the mutation rate of ∼10⁻³/bp per generation that is observed in these regions in germinal center B cells (1). There is a rough correlation with rate of transcription and frequency of AID mutation and AID physically interacts with Pol II and directly or indirectly interacts with more than 20 other factors that are involved in Pol II pausing and elongation, splicing, and other aspects of the processing of nascent RNA (4), and with chromosomal regions that include superenhancers (2, 12). In addition, transcription-dependent ssDNA that is the substrate for AID has been identified in the cross-linked chromatin of Ig V regions in chicken, mouse, and human B cells (7, 31, 46, 47). Biochemical studies did reveal that the substrate for AID is ssDNA (48) and transcription studies using T7 bacteriophage, bacterial RNA polymerases, and purified core human Pol II can duplicate the hotspot specificity of the ability of AID (10, 15, 49–51). Although transcription-dependent AID-induced mutations can be mimicked in vitro by biochemical assays, AID targeting of chromatin-bound and highly structured DNA (52) in the nucleus is likely to require additional factors and chromatin modifications, including Dot1L and H3K79me and many other factors that are described above to generate the very high rate of AID mutation, especially in the complementary determining regions of the V region. All of these and other studies have supported the idea that ssDNA substrates for AID are created during pausing, backtracking, and perhaps premature termination of transcription at sites within the coding exon of the V region (15, 34, 42).

The fact that AID mutations do not occur until a few hundred base pairs downstream from the TSS (53) further distinguishes this process from the pausing seen in many genes 20 to 40 nucleotides downstream from the TSS. This is consistent with the increased abundance of the various elongation factors in the coding exon for the V region observed in this study and may explain the many local restrictions on the AID mutational process that are superimposed on the larger-scale chromosomal changes that occur. Nevertheless, since there are many highly transcribed genes in B cells that do not undergo AID-mediated hypermutation and the factors associated with AID hypermutation, except AID, are ubiquitous (5), it has been difficult to discover exactly how the ssDNA substrates for AID become selectively available and mutated in the Ig genes. Furthermore, it is generally believed that the transcription bubble itself is fully occupied with the many protein components of Pol II and its regulatory factors. However, there is flexibility in the interactions of the elongation complex and nucleosomes (54), perhaps providing access for AID. An alternative, though not mutually exclusive, suggestion is that supercoiled DNA at the trailing edge of the transcription bubble or sites of premature terminations are the source of the ssDNA that is mutated by AID (15, 17, 34). It is also possible that overlapping AID hot spots or some neighboring non-B DNA structure, such as a G-quadruplex, could specifically recruit and activate AID, as seems to occur in the S regions, but that has not been as extensively demonstrated in V regions (55, 56). An understanding of the exact process that allows the targeting of AID to its substrate in the Ig genes must also explain how the processivity of AID seen in purely biochemical systems is restricted in B cells to allow only one mutational event per cell cycle (57, 58). While this restricted processivity of AID mutation can be explained to some degree by the low enzymatic efficiency of AID (59), it has been suggested that this restriction on multiple mutations may also help protect off-target sites from deleterious mutations (4).

H3K79 methylation is found in less than 30% of the genes and H3K79me2/3 in only a very small percentage of those genes genome-wide, suggesting that this modification has a very special role (23). It is associated with active transcription and enriched in the gene body rather than in the promoter (32). A recent study found that H3K79me2/3 can also serve as a mark for putative intragenic enhancers and cross-talk with H3K27ac in those enhancer regions (43), indicating H3K79me2/3 may play a role in enhancer-related topologically associated domains (TADs) (52). Importantly, it has been demonstrated recently that SHM susceptibility is correlated with TADs, which are enriched with H3K27ac (52), suggesting cis-regulating elements and TADs may contribute to AID-specific targeting. Studies of Dot1L complexes reveal that they may interact with P-TEFb and the super elongation complex through MLL, consistent with its role in activating transcription elongation. In addition, cross-talk between H3K79me and H2Bub has been conserved in yeast and humans and seems to be associated with the efficiency of H3K79 methylation by Dot1L (23).

Here, we show that while H3 had approximately the same abundance in the regions near the TSS, the heavy-chain V, the 5′ end of Sμ, and the constant regions, while H3K79me2/3 was most abundant in the V region exon and Sμ that undergo high frequencies of AID-induced mutations. H3K79me2/3 was not abundant on the 5′ end of C region and was barely detectable in CD4, which is not expressed in these cells. Consistent with this, Dot1L was abundant in the V region and its 3′ flanking Sμ regions, but not in the C region. In contrast, H3K79me1 was relatively abundant in the C region. H2BK120ub is thought to cross-talk with Dot1L and facilitates its interaction with the H4 tail to stabilize the conformation of the enzyme on the nucleosome as transcription progresses from a poised to an active state, and changes the conformation of H3 to make K79 accessible (24, 60). H2BK120ub was also abundant on the coding portion of the V region but even more prominent in the C region than either Dot1L or H3K79me2/3. It is possible that the H2BK120ub associated with the C region compensates for the low levels of Dot1L, allowing H3K79me1 but not H3K79me2/3 to be generated there. Since we do not know exactly how Dot1L is recruited to specific sites, we cannot resolve this issue. Overall, these findings are consistent with higher levels of elongation in the coding regions of the V and S regions where mutation is occurring, and suggest that Dot1L and the unusual properties of this methyltransferase such as the cross-talk between H3K79, H2BK120ub, and H4, and its ability to change the conformation of H3 could be contributing some special modulation of the transcriptional process that targets and regulates AID activity.

It is important to note that these modifications and the targeting of Dot1L in Ramos cells do not require the expression of AID or the presence of ongoing mutations. In fact, wherever it has been studied, the molecules and factors involved in SHM and class switching all appear to preexist at the critical sites before AID is expressed (5). Since we do not have a truly inducible system for the onset of mutation in the V regions, we stimulated primary naïve mouse splenic B cells ex vivo to switch from IgM to IgG1. While the donor Sμ region is constitutively transcribed, 48-h stimulation turns on the sterile transcription of the relevant downstream S regions, increases the expression of AID and the initiation of AID-induced mutation of the downstream Sγ1, while suppressing the expression and mutation of Sγ3 (61). As can be seen in Fig. 1F, H3K79me2/3 is abundant on the very highly transcribed donor Sμ before stimulation but increases greatly on the Sγ1 at 48 and 72 h after stimulation. This is also relevant and supported by a very recent study showing either Dot1L depletion in mouse B cells or inhibition of Dot1L reduced ex vivo class switching (27). Equally important, there is virtually no H3K79me2/3 associated with Sγ3, which does not undergo AID-induced mutation.

Interestingly, the JH4 region, which is part of the many different rearranged V regions and undergoes somatic mutation in Peyer’s patch geminal center B cells (35), also has a high occupancy of H3K79me2/3. This is consistent with the idea that H3K79me2/3 marks regions that may ultimately undergo AID mutation but is not sufficient by itself to permit such mutations even in the presence of high levels of AID in the nucleus (62). These studies also suggest that H3K79me2/3 is not only playing a role in SHM in a human B cell line but also in primary mouse B cells.

The effects of the two inhibitors and the KO of Dot1L demonstrate that H3K79 methylation contributes to the AID mutational process. While the dose dependency is reassuring, the fact that there is only 30 to 40% inhibition by the inhibitors suggests that there is redundancy for the role of Dot1L, for example, by other elongation complexes, such as BRD4 and P-TEFb, as illustrated in this paper. Furthermore, it is clear that the inhibitor EPZ004777 only partially blocks the K79 methylation or predominantly blocks H3K79me2/3 since, even after treatment with the highest dose, there is still considerable monomethylation of chromatin genome-wide based on our Western analysis and of the H3K79me1 that is associated with the V regions. A very recent paper, where the methyltransferase activity of Dot1L was eliminated by site-directed mutagenesis, did not find any residual monomethylation (63). It should be noted that Dot1L can still associate with the V region when its enzymatic activity is blocked by the inhibitor. The inhibitors are useful in that they rule out clonal variation and the long-term effects of the loss of Dot1L. The greater inhibition of SHM by the KO of Dot1L provides independent confirmation of its role.

It is interesting that the reciprocal increase in monomethylation of the V region observed with the inhibitor is not present in the KO where Dot1L has been eliminated from IgH gene. The much greater loss of SHM in the KO could reflect some sort of additional scaffolding role for Dot1L. Consistent with this, a very recent study showed Dot1L itself may have a role in transcription elongation that is independent of H3K79 methylation by interacting with the other elongation complexes (63). Such a nonmethylation-dependent role is also consistent with the fact we did not observe a significant change of transcription level in the V region by RT-qPCR when using the inhibitor. Nevertheless, we do not know the relative roles of mono- and di/trimethylation or if the H3K79me1 is playing the same exact role as H3K79me2/3. For example, they could be recruiting the same factors to the transition from pausing to elongation but with different efficiencies. Furthermore, the fact that in the KO there is no H3K79me3 associated with the V region but still considerable SHM again suggests that there is considerable redundancy with other elongation complexes. This led us to examine whether other factors involved in elongation were associated with the V region chromatin. We found that there was an enrichment in the coding portion of the V region of BRD4 and CDK9. Furthermore, JQ1 that inhibits the BRD4 complex and DRB that inhibits CDK9 did in fact inhibit AID-induced somatic mutation in the Ramos cell line. These findings indicate that the BRD4 and P-TEFb elongation complex has an important role in the V region exon, which could be different from the pausing and release complex containing DSIF at the pausing site 20 to 40 bp downstream from the TSS in some genes. Different elongation complexes could also play distinct roles in the C region or the downstream superenhancers, which are far away from the V region (11–13).

The apparent role of all of these elongation factors increases our focus on whether the activation or licensing of AID (64) is increased by a transition from pausing to elongation of transcription. There is a decrease in nascent transcripts in the cells knocked out for Dot1L, but this is associated with a decrease in the abundance of Pol II and Pol II S2P that is involved in elongation, and of BRD4 and CDK9. The relative abundance of BRD4 and CDK9 decreases when the cells are treated with JQ1 but there is a relative increase in the abundance of both of them after treatment DRB, which is specific for only CDK9. It is possible that this increase reflects an attempt to compensate for the loss of CDK9 activity. In a more general sense, it is unclear why multiple elongation complexes are recruited to the V region. They could each carry out slightly different functions within the elongation process. However, the fact that there is an increase in CDK9 at the V region when Pol II is decreased, even though there is no overall change in CDK9 expression, suggests a strong pressure on the B cell to retain elongation at the V region. This apparent compensation also confirms redundancy among the different elongation complexes.

Taken together, these studies reinforce the idea that there is a “multilayered” (5) mechanism for specifically licensing the Ig V region to AID-induced mutations that draws heavily on the tight regulation of transcription elongation and perhaps the transition from pausing to elongation so as increase the likelihood of AID mutation but also to limit the processivity of that process to one mutation per cell division. Dot1L may be particularly important because of its specificity for H3K79 methylation and its potential to change the conformation of H3, thus increasing the restriction of this process through its effect on the local chromatin context. The participation of multiple elongation complexes may also contribute to the tight control of AID targeting and processivity.

Materials and Methods

Cell Culture and Inhibitor Treatment.

The human Burkitt’s lymphoma Ramos cell line was cultured with Iscove’s modified Dulbecco’s medium (BioWhittaker) supplemented with 10% (vol/vol) FBS (Atlanta Biologicals) and 100 U/mL penicillin-streptomycin (Mediatech). To induce AID-dependent mutations, the reporter Ramos cells described previously (34) were treated with 0.25 μM 4-Hydroxytamoxifen (4-OHT) for 7 d. The inhibitors EPZ004777 (Selleckchem, S7353), EPZ-5676 (Selleckchem, S7062), JQ1 (Selleckchem, S7110), and DRB (Sigma-Aldrich, D1916) were prepared in DMSO and the final concentration of DMSO was kept below 0.04%. When the various inhibitors were used, a separate DMSO treatment combining with 4-OHT was set as a control and the concentration of DMSO was the same in all of the samples. The inhibitor and 4-OHT–containing medium was freshly replaced every 2 to 3 d during the treatment. Mutation frequency represented by the loss of mCherry fluorescent cells was analyzed by flow cytometry.

Chromatin Immunoprecipitation.

ChIP assays were performed as previously described (65). In brief, 1 × 10⁷ cells were cross-linked with 1% formaldehyde for 10 min and followed by quenching with 125 mM glycine for 5 min. The fixed chromatin was isolated and sheared by sonication to an average length of 200 to 500 bp. After preclearing with Dynabeads Protein G (Thermo Fisher Scientific, 10004D), the same amount of chromatin was used for immunoprecipitations with 2 to 4 μg of specific antibodies or normal IgG, incubated overnight, and the DNA–protein–antibody complexes were incubated with Protein G Dynabeads for 2 h. After washing the beads, DNA were extracted by 10% Chelex-100 resin (Bio-Rad, 1421253) and digested with Proteinase K (Invitrogen, AM2546). The amount of immunoprecipitated DNA was quantified by real-time PCR and normalized to the input. The signal of the IgG control was lower than 0.08% and was subtracted. The information of primers and antibodies used for ChIP assays are listed in SI Appendix, Tables S1 and S2, respectively. The anti-H3K79me3 antibody (Abcam, ab2621) was used but since there is some evidence that it slightly cross-reacts with to H3K79me2 (66), we have labeled all of our results with it as “H3K79me2/3.”

UMI-Based Library Construction and Sequencing.

Genomic DNA was isolated from Ramos cells treated with 4-OHT+DMSO or 4-OHT+EPZ004777 using the DNeasy Blood & Tissue Kit (Qiagen, 69504) according to the manufacturer’s instructions. UMI-based libraries were constructed by two rounds of PCR reactions. Briefly, the first round of PCR using Q5 High-Fidelity 2X Master Mix (New England BioLabs, M0492S) started with 2-μg genomic DNA as template for four PCR reactions. A UMI-contained VH4-34 specific primer (SI Appendix, Table S3) was used and the PCR cycles were limited to three cycles to avoid reduplicative adding of UMIs to the V regions. After the first round of the PCR, the product was purified by the GeneRead Size Selection Kit (Qiagen, 180514) to remove the primers. The second round of PCR used the Q5 mix and the purified product from the first round of PCR and the Nextera Index Kit (illumine, 15055289) was used as primers to add adapters. The PCR cycles were limited to 16 cycles. The final libraries were purified with Generead Size Selection Kit and analyzed by Bioanalyzer. The second-generation sequencing was performed by Epigenomics core facility in Albert Einstein College of Medicine by using Illunima Mi-seq for pair-end 300-bp sequencing. For analysis for SHM, the raw data were processed and aligned using SHMprep: (http://www.ams.sunysb.edu/∼maccarth/software.html) provided by T.M. from Stony Brook University, with default parameters. The output from SHMprep was filtered to remove the sequences containing INDELS (insertion or deletion). The clean sequences Datasets S1–S3 were uploaded to and analyzed by SHMserver (http://sdev2.ams.sunysb.edu/site1/shmserver/session/create) to process the mutation data for Fig. 2B, and SI Appendix, Fig. S5 were generated by R scripts (https://github.com/Jun2BCR/BCR_analysis). The significance between two groups was analyzed using prop.test the methodology of Maccarthy et al. (67).

RT-qPCR for Nascent RNA and Total RNA.

Nascent RNA was extracted from the chromatin fraction of the cells by depleting mRNA as described in Casill et al. (40). Ramos cells were lysed by Cell Lysis Buffer (containing 25 mM Tris pH 7.9, 150 mM NaCl, 0.1 mM EDTA, 0.1% Triton-X, 1 mM DTT, and protease inhibitors) combined with Drosophila S2 cells as spike-in to obtain the nucleus fraction. Then the chromatin fraction was isolated from the nucleus by Nuclear Lysis Buffer (containing 20 mM Hepes pH 7.6, 300 mM NaCl, 7.5 mM MgCl₂, 1% Nonidet P-40, 1 mM DTT, and 1 M urea). The chromatin-bound total RNA was extracted using TRIzol (Invitrogen, 15596026) and mRNA was depleted by using Oligo dT (25) Magnetic Beads from NEBNext Poly(A) mRNA Magnetic Isolation Module (New England BioLabs, E7490L). After mRNA depletion, the nascent RNA was reverse-transcribed and quantified by real-time PCR. The quantified signal was double-normalized based on Casill et al. (40). Total RNA was extracted by TRIzol (Invitrogen, 15596026) and reverse transcription was performed using the Maxima First Strand cDNA Synthesis Kit (Thermo Scientific, K1641). The primers used for total RNA RT-qPCR can be found in SI Appendix, Table S3.

KO of Dot1L by Crispr/Cas9.

The gRNA targeting the 5′ exon the Dot1L was designed in the Benchling website and the one with highest efficiency and specificity was used. The sequence of the gRNA is: 5′-GCA​CTC​ACG​TAG​ACC​GGC​AG-3′. A vector containing Cas9, gRNA and GFP was made based on PX458 backbone (#48138, Addgene) and electroporated into healthy Ramos reporter cells. Two days after the electroporation, the viable GFP⁺ cells were sorted into 96-well plate with one cell for each well. After 3 wk of culture, the expanded clones were recovered for Western blot against Dot1L. The single clones giving low or no protein signal by Western blot compared with control were picked for further analysis. In the end, two independent Dot1L KO clones were screened by Western blot and then were confirmed by Sanger sequencing.

Statistical Analysis.

All statistical analyses were conducted using Graphpad Prism 8 software excepting the next-generation sequencing data. Error bars represent SD among independent experiments and P values were calculated using an unpaired Student’s t test. A P value < 0.05 was considered as statistically significant, and in the figures, *P < 0.05, **P < 0.01, and ***P < 0.001.

Additional materials and methods are described in SI Appendix.

Data Availability

The assembled datasets for clean sequences which were used for SHM analysis are listed as Datasets S1–S3. For ATAC-seq, all the datasets generated by this study have been deposited and are available in the Gene Expression Omnibus (GEO) database, https://www.ncbi.nlm.nih.gov/geo (accession no. GSE167873) (68). All other study data are included in the article and supporting information.