A 220-nucleotide deletion of the intronic enhancer reveals an epigenetic hierarchy in immunoglobulin heavy chain locus activation

We used core Eμ-deleted mice bred to a recombination activation gene (RAG) 2–deficient background (20) to study the chromatin and transcription state of the IgH locus just before initiation of recombination. We used chromatin immunoprecipitation to assay histone modification changes in the absence of Eμ. In primary B cell precursors that contain unrearranged IgH loci (pro–B cells), H3K9 acetylation (H3K9ac), a mark of gene activation, was severely reduced throughout the D_H-Cμ domain on Eμ⁻ alleles compared with Eμ⁺ alleles (Fig. 1 A). This included particularly high levels of H3K9ac at the J_H gene segments and the peak located close to DFL16.1, which is more than 50 kb 5′ of Eμ. Because the H3K9ac pattern in an Abelson virus–transformed pro–B cell line from Eμ⁻/RAG2⁻ mice (Fig. S1 A) was indistinguishable from that seen in primary cells, we extended the analysis to H4 acetylation (H4ac) in this cell line. H4ac levels were also substantially reduced in the absence of Eμ (Fig. S1 B). In contrast, a third activation-specific modification, dimethylation of histone H3 at lysine 4 (H3K4me2), was considerably less affected on Eμ⁻ alleles from bone marrow–derived pro–B cells (Fig. 1 B). We conclude that Eμ controls only a subset of tissue-specific positive histone modifications that mark the unrearranged IgH locus.

Dimethylation of lysine 9 of histone H3 (H3K9me2) is a mark of inactive chromatin. In pro–T cells, or nonlymphoid lineage cells, H3K9me2 is present throughout the IgH locus. In pro–B cells, H3K9me2 is replaced by H3K9ac in all parts of the D_H-Cμ region except the intervening DSP2 gene segments (7). We therefore investigated whether Eμ⁻ alleles reverted to H3K9me2 modification in the absence of H3K9ac. The pattern of H3K9me2 was indistinguishable between Eμ⁺ or Eμ⁻ pro–B cells (Fig. 1 C), revealing a discordance in the inverse relationship between H3K9ac and H3K9me2. We infer that loss of H3K9me2 is Eμ independent and gain of H3K9ac is Eμ dependent in primary pro–B cells. These observations indicate that Eμ⁻ alleles are in a partially activated state.

This idea was further corroborated by analysis of suppressive histone modifications in pro–B cell lines. We found that DQ52, J_H2, and J_H4 amplicons that were associated with particularly high H3K9ac in Eμ⁺ cells contained 2–3-fold higher levels of H3K9me2 in Eμ⁻ cells (Fig. 1 D). We note, however, that the locus was not restored to the fully repressed state seen in pro–T cells (Fig. 1 C) or to the level of H3K9me2 at the adjacent DSP2 repeats in Eμ⁺ pro–B cells. Dimethylation of lysine 27 of H3 (H3K27me2), another negative regulatory mark which has been proposed to be most evident in dividing cells (24), was also greatly elevated in the DQ52-Cμ region on Eμ⁻ alleles (Fig. S2). These observations emphasize the state of the Eμ-deleted locus as a transitional intermediate between a fully “open” and a fully “closed” configuration.

We further investigated the effects of Eμ deletion using a PCR-based DNase I sensitivity assay to probe Eμ⁺ and Eμ⁻ alleles in pro–B cell lines. To compare between samples, signals from IgH locus amplicons were normalized to an amplicon from the β-globin gene (25). Amplicons just 5′ or 3′ to the Eμ core were rapidly degraded in Eμ⁺ cells (Fig. 2, solid lines), which is indicative of the Eμ DNase I HS. Both amplicons were relatively insensitive in Eμ⁻ cells (Fig. 2, dashed lines), indicating that Eμ core deletion abrogated the HS (Fig. 2, Eμ-5′ and Eμ-3′). Classical DNase I hypersensitive site mapping by Southern blots confirmed this conclusion (unpublished data). Notably, DQ52 hypersensitivity was also substantially reduced in Eμ⁻ cells, suggesting that it was Eμ dependent. Loss of Eμ also reduced DNase I sensitivity within J_H gene segments and at Cμ but not at DFL16.1. The H3K9me2-bearing DSP genes remained DNase I insensitive in Eμ⁺ or Eμ⁻ cells. We conclude that Eμ controls local chromatin accessibility and is critical for generation of the DQ52 HS.

Because the DQ52 HS is Eμ dependent, Eμ-deleted alleles lack both known DNase I HSs in the germline D_H-Cμ region. The partially activated state of the locus that we observed may be the result of cryptic transcription factor binding sites that are revealed in the absence of Eμ. A more interesting implication is that initial locus opening (scored as H3K9 demethylation and H3K4 methylation) is mediated by cis sequences that are not marked by DNase I HS. Codependence of DNase I HS has been previously explored in the β-globin locus control region. In the mouse locus, loss of one HS does not affect formation of the others (3, 26); that is, these sites are generated independently. In contrast, deletion of individual HSs in BAC transgenics that carry the human locus substantially reduces the formation of other HSs (27). Loss of DQ52 HS on Eμ-deleted alleles resembles the latter observation in a germline rather than transgenic context.

We used quantitative RT-PCR with primers as indicated (Fig. 3, top line) to assay transcription in the absence of Eμ. Consistent with earlier results (18, 20), Cμ-encompassing transcripts were reduced ∼7–10-fold in Eμ-deficient primary pro–B cells. Additionally, sense-oriented transcripts, as assayed by the DQ52 amplicon, as well as all antisense transcripts assayed by all other D_H-amplicons (7, 8), were reduced 10–50-fold in Eμ⁻ pro–B cells (Fig. 3 A). Reduced transcript levels coincided with reduced RNA polymerase II density on Eμ⁻ alleles in pro–B cell lines (Fig. 3 B). We conclude that one function of Eμ is to recruit RNA Pol II to the IgH locus.

We also assayed the effect of Eμ deletion on transcription-associated histone modifications. H3K4me3 has been shown to be enriched at the 5′ ends and H3K36me3 at the 3′ ends of transcription units (28, 29). Eμ deletion reduced H3K4me3 levels to a third of that seen on Eμ⁺ alleles in primary pro–B cells (Fig. 3 C) and virtually eliminated both forms of modifications in the Eμ⁻ pro–B cell line (Fig. S3). We infer that low-level transcription in primary cells may be sufficient to induce H3K4me3; alternatively, this mark may build up as a result of its slow removal by histone demethylases. A plant homeodomain (PHD) in RAG2, which selectively binds H3K4me3, was recently shown to be required for efficient V(D)J recombination (30, 31). Our observations provide the first evidence that low levels of H3K4me3 present in Eμ⁻ primary pro–B cells may be sufficient to direct D_H to J_H recombination even in the absence of histone acetylation.

Because Eμ deletion affects V_H to DJ_H recombination, we examined sterile V_H gene transcription in Eμ⁺ and Eμ⁻ primary pro–B cells. By quantitative RT-PCR, we found that sterile transcripts of proximal V_H genes were substantially attenuated by loss of Eμ (Fig. 3 D); however, five amplicons representing upstream V_Hs were not significantly affected. For reference, V_H7183.1b (also know as V_H81X) is located ∼98 kb from DFL16.1 and Vox-1, a further 300 kb from V_H7183.1b. VGK2, the most 3′ gene in our set which is not affected by Eμ deletion, lies 420 kb 5′ of Vox-1. We conclude that Eμ influences transcription of gene segments located >400 kb away.

Earlier studies show that Eμ deletion results in five- to eightfold lower levels of D_H to J_H recombination (18, 20). These numbers were obtained from analyses of steady-state cell populations that are potentially subject to selection in vivo. To get an independent measure of D_H recombination efficiency, we expressed RAG2 in two Eμ⁺ and one Eμ⁻ RAG2-deficient pro–B cell lines by retroviral gene transfer and followed the levels of D_H recombination as a function of time. We amplified genomic DNA from infected cells with 5′ primers corresponding to either DFL16.1 or DSP gene segments and a common 3′ primer downstream of J_H3 (Fig. 4, top line). The amplified fragments were detected by Southern blotting after agarose gel fractionation. We observed easily detectable levels of DFL16-J_H and DSP-J_H rearrangements over the experimental time course in both Eμ⁺ cell lines (Fig. 4 A, lanes 1–4 and 13–16) but not in the Eμ⁻ cell line (Fig. 4 A, lanes 7–10). Transduced RAG2 gene levels were comparable between all three cell lines, and an amplicon from the β-globin locus was used to ensure equal loading of genomic DNA. The mean of three independent infections is quantitated in Fig. 4 B. We conclude that Eμ deletion severely impairs D_H to J_H recombination in this assay. The greater reduction of D_H to J_H recombination in cell lines compared with primary cells may be because the chromatin structure of the IgH locus in continuously cycling Eμ⁻ cells is skewed toward a suppressive state. This is reflected in lower levels of activating modifications H3K4me2 and H3K4me3 and higher levels of inactivating modifications H3K9me2 and H3K27me2 in these cells compared with primary cells. Alternatively, DJ_H junctions seen in the bone marrow of Eμ-deficient mice represents a gradual accumulation of recombinant alleles.

Our observations suggest an epigenetic hierarchy in the activation of the IgH locus. We propose that an Eμ-independent first step removes H3K9me2 and establishes the H3K4me2 mark (Fig. 5, line 2). This may permit transcription factors to access Eμ to convert a partially activated state to a fully activated one by Eμ-dependent induction of DNase I HS, histone acetylation, and transcription (Fig. 5 line 4). Though we cannot exclude the possibility that Eμ-independent and Eμ-dependent steps occur in parallel, we favor a sequential model based on studies in Saccharomyces cerevisiae demonstrating that histone acetyl transferases are recruited to promoters that are premarked with H3K4me2 (32). The coordinate control of acetylation and transcription in the IgH locus is in line with genome-wide association studies (28, 33). Because histone acetylation likely precedes transcription (34), we suggest that Eμ binding proteins recruit RNA Pol II and chromatin-modifying enzymes to generate the optimal substrate for transcription (Fig. 5, line 3). Subsequently, transcription-associated histone modifications, such as H3K4me3, are incorporated into the locus (Fig. 5, line 4), which, in the case of antigen receptor genes, may be particularly important in targeting the V(D)J recombinase via the PHD domain in RAG2.

Abelson-transformed RAG2^−/− and Eμ^−/−RAG2^−/− bone marrow–derived cell lines were cultured as previously described (7). CD19⁺ bone marrow cells were purified from RAG2^−/− and Eμ^−/−RAG2^−/− mice were purified as previously described (11). Thymi from the same animals yielded CD4⁻CD8⁻ thymocytes. All mouse experiments were approved by the Animal Care and Use Committees of the National Institute on Aging (Harvard University and the Immune Disease Institute).

ChIPs were performed as previously described (7). Formaldehyde–cross-linked and sonicated chromatin from 10⁷ cells was precleared with 5 µg of nonspecific IgG and immunoprecipitated with the requisite antibody (sources noted in Table S1) or an equal amount of nonspecific IgG. The coprecipitated DNA was purified and analyzed by real-time PCR using either previously described primers (7) or primers shown in Table S2.

Real-time PCR was performed as previously described (7) using the ABI Prism 7000 (Applied Biosystems). Abundance of target sequences in the immunoprecipitate relative to the input DNA (IN) was determined as previously described (35), where the relative abundance of the target sequence in the immunoprecipitate is 2^{−[Ct(IP)−Ct(IN)]}, and Ct is the cycle at which the sample reached a threshold value where PCR amplification was exponential.

20 × 10⁶ nuclei from RAG2^−/− and Eμ^−/−RAG2^−/− cells were treated with varying concentrations of DNase I. 25 ng of purified genomic DNA was used in quantitative PCR assays performed in duplicate with primer pairs shown in Table S2. The proportion of DNA for each amplicon was normalized to the amount of intact β-globin alleles at each DNase I concentration using the formula 2^{amp[Ct(0)−Ct(n)]}/2^{β[Ct(0)−Ct(n)]}, where Ct(0) is the cycle number for non-DNase I–treated samples at which an amplicon reaches the threshold for exponential amplification and Ct(n) is the cycle number for the sample treated with n units of DNase I. In this analysis, signal strength for β-globin amplicon would be 1 at all concentrations of DNase I. Sensitivity at the IgH locus was determined using two to three independent DNase I–treated samples from cells of each genotype.

Lentiviral particles expressing RAG2 were generated as previously described (31) by transiently transfecting 293T cells with pWPI-RAG2 along with helper plasmids pΔ8.2R and pVSVG using FuGENE 6 (Roche). The supernatant containing the virus was collected at 48 and 72 h after transfection and concentrated by ultracentrifugation for 2 h at 25,000 rpm and 20°C over a 20% sucrose cushion. 4–6 × 10⁶ cells were infected with freshly prepared control or RAG2-expressing lentivirus by spin inoculation in the presence of 10 µg/ml polybrene. Genomic DNA was isolated from the infected cells at different days after infection using the DNeasy tissue kit (QIAGEN). All procedures involving lentiviruses were performed under BSL2 conditions.

45 ng of genomic DNA from transduced cells or 9 ng of genomic DNA from C57BL/6 bone marrow was amplified by 33 rounds of PCR using the HotStarTaq polymerase (QIAGEN). The forward primers hybridized 5′ of either the DFL16.1 segment or the DSP2 segments and the reverse primer hybridized 3′ of J_H3 (Table S2). 25% of the PCR reaction was resolved by agarose gel electrophoresis, and rearrangement products were detected by Southern hybridization to an oligonucleotide probe that recognizes J_H3. An amplicon from the mouse β-globin locus was used to normalize across samples, and level of RAG2 was ascertained using a primer pair that amplified part of the Rag2 gene not present in RAG2⁻ cells (30 rounds of PCR amplification and 20% of the reaction used for Southern blotting).

Supplemental figures show the effects of Eμ deletion on histone modifications assayed in Abelson mouse leukemia virus–transformed cell lines derived from the bone marrow of Eμ⁺ and Eμ⁻ mice in a RAG2-deficient background. Table S1 lists the sources of antibodies and reagents used for ChIP and Table S2 lists primer sequences used in this work.

We thank Sebastian Fugmann, Cosmas Giallourakis, Yu Nee Lee, and Lei Wang for discussion and comments on the manuscript and Valerie Martin for editorial assistance.

T. Perlot is supported by a Boehringer Ingelheim Fonds PhD scholarship. F.W. Alt is an investigator of the Howard Hughes Medical Institute. This work was supported by National Institutes of Health grant AI2047 (to F.W. Alt) and by the Intramural Research Program of the National Institute on Aging (Baltimore, MD).

The authors have no conflicting financial interests.