Antigen receptor loci poised for V(D)J rearrangement are broadly associated with BRG1 and flanked by peaks of histone H3 dimethylated at lysine 4

Cell Lines. RAG2-deficient Abelson murine leukemia virus-derived Pro B cells (32) and RAG1-deficient, p53-deficient Pro T cells (33) were maintained in RPMI medium 1640 supplemented with 20% FBS and 0.05 mM 2-mercaptoethanol. NIH 3T3 cells were maintained in DMEM supplemented with 10% calf serum.

ChIPs. Approximately 2 × 10⁸ cells were used per ChIP and were prepared as described (34) (for additional details, see Supporting Materials and Methods, which is published as supporting information on the PNAS web site, www.pnas.org). DNA recovered from an aliquot of sheared chromatin precleared with protein A-Sepharose resin (Sigma) was used as the “input” sample. The remaining chromatin was incubated with one of five antibodies for 3 h at room temperature [1:500 anti-Acetyl H3 (K9/K14), Upstate Biotechnology (U.B.; Lake Placid, NY) catalogue no. 06-599; 1:500 anti-di-Me H3-K9, U.B. catalogue no. 07-212; 1:1,000 anti-di-Me H3-K4, U.B. catalogue no. 07-030; 1:1,000 anti-BRG1 (35), generously provided by Robert Kingston, Massachusetts General Hospital; and 1:2,000 anti-Phos H4-S1]. Anti-Phos H4-S1 is highly reactive toward histone H4, but because of sequence identity with the first five residues of histone H2A, minor cross-reactivity has been observed (S.D.T. and C.D.A., unpublished work). DNA recovered after IP and crosslink reversal was quantified by using picogreen fluorescence (Molecular Probes).

Real-Time PCR Analysis. Triplicate reactions containing 2 ng of input or IP DNA and 10 pmol of each PCR primer (Table 1, which is published as supporting information on the PNAS web site) were amplified in SYBR Green PCR Master Mix (Applied Biosystems) by using the Bio-Rad iCycler iQ. Fold enrichment values for each amplified DNA sequence were determined as described (36), with fold enrichment = R̂(Ct_input – Ct_IP), where R is the rate of amplification (ranging between 1.85 and 2) and Ct, the cycle threshold, an average of three values (see Supporting Materials and Methods for additional details).

Three different types of murine cell lines, representative of three different lineage and developmental contexts, were analyzed: a RAG2^–/– Abelson virus transformed Pro B cell line (32), a RAG1^–/– p53^–/– Pro T cell line (33), and NIH 3T3 cells, a fibroblast line. The use of cell lines ensured we were analyzing a homogeneous population of cells at a discrete developmental stage and that adequate material could be obtained for analysis. Representative analysis of ChIP samples from a second RAG2-deficient Pro B cell line confirmed the data for each of the modifications discussed below (data not shown). In the absence of recombinase activity, the Pro B and Pro T cells from RAG-deficient animals and the cell lines derived from them are blocked in development just before the first stage of rearrangement. Thus, the D and J segments at the IgH locus in the Pro B cells and the TCRβ locus in the Pro T cells are poised to rearrange (“active”) (15), whereas VH, Vβ, and light chain gene segments should be inaccessible in both cell lines (“inactive”). Indeed, on reintroduction of the absent RAG gene, DJ rearrangement at the IgH and TCRβ loci is observed in these Pro B and Pro T cell lines, respectively. No V to DJ rearrangement is detected in the Pro T cell line, whereas a low level of V to DJ joining occurs in the Pro B cell line (ref. 32; E. Oltz, personal communication; P. Sieh, personal communication). This latter observation is consistent with much previous work showing that V to DJ joining follows DJ joining in Abelson-transformed early B cells. All Ig and TCR loci are refractory to rearrangement in NIH 3T3 cells (37).

DNA recovered from each ChIP was analyzed by real-time PCR by using primer pairs specific for a collection of V, D, and J gene segments as well as C gene segments and regulatory enhancers (E) (Table 1). These primer pairs are distributed across four large (≈0.2–3.5 Mb) loci: IgH, TCRβ, Igκ, and Igλ (Fig. 1A), although the primary focus of our analyses is a comparison of the IgH and TCRβ DJ loci. A pair of primers specific for a segment of the promoter of the ubiquitously expressed CAD gene was also included (38).

Inverse Correlation of Acetyl H3 and di-Me H3-K9 at Antigen Receptor Loci. Localization of Acetyl H3 (Fig. 1B), an activating modification, was found to correlate inversely with that of di-Me H3-K9 (Fig. 2), a repressive modification, at Ig and TCR gene segments in the Pro B and Pro T cell lines (for a normalized comparison of these and other modifications, see Fig. 5). In the Pro B cell lines, the accessible region of IgH, as defined by hyperacetylation and low di-Me H3-K9, extends across a large (≈80 kb) region that contains the DH and JH segments as well as Eμ and Cμ. This is the same region found to be hyperacetylated in primary Pro B cells and other Abelson-transformed RAG-deficient cell lines (25, 26). Hyperacetylation and low di-Me H3-K9 are also seen at the two clusters of TCRβ D and J elements in the Pro T cell line (Fig. 2C). In addition, the level of acetylation is reproducibly higher over the J segments than over the D segments in active loci. IgH and TCRβ V segments are likely to be inaccessible in both the Pro B and Pro T cell lines, and they are indeed H3-K9 methylated but not acetylated in both cell types, providing a useful comparison to the active DJ loci. As expected, the IgH and TCRβ loci in NIH 3T3 cells are methylated but not acetylated. The relatively modest level of di-Me H3-K9 associated with inactive loci is in line with previous observations and is likely a reflection of the generally high overall levels of methylation across the genome (3). Finally, the active CAD gene is acetylated and undermethylated to a similar extent in Pro B, Pro T, and NIH 3T3 cells (Figs. 1B and 2 A and C).

The pattern of modified histones described above correlates well with the active and inactive regions of the IgH and TCRβ loci in the cell types analyzed and provides an independent set of criteria (in addition to DNase I sensitivity and sterile transcript production) for the presence of open and closed antigen receptor loci. Further, the acetylation pattern is generally consistent with prior work in both transformed and primary cells (25). Thus, these data provide an internal control and framework with which to compare the localization of other modifications and modifying activities.

Although the pattern described above is generally consistent with the known biology of developing B and T cells, one unexpected observation stands out. On the basis of acetylation [Fig. 1B, as well as BRG1 (Fig. 1C) and di-Me H3-K4 (Fig. 3)], the TCRβ enhancer appears open in Pro B cells, although TCRβ remains recombinationally silent. Perhaps one set of factors interacts with the TCRβ enhancer in Pro T cells to activate the locus, whereas a different set is used in Pro B cells to ensure a recombinationally inactive state. (Vβ14 is marked by similar chromatin modifications, but this may simply be due to its close proximity to Eβ.) We also note that a number of Vβ segments have been found to be acetylated in developing RAG-deficient thymocytes (28). However, the apparent difference in our findings might result from analysis of cells at slightly different developmental stages, differences in the particular Vβ segments analyzed, or differences between primary and immortalized cells.

Two Distinct Levels of di-Me H3-K9 at Inactive Loci in a Segment-Specific Pattern. Unexpectedly, two distinct levels of di-Me H3-K9 are found at certain inactive loci in a pattern that tightly correlates with segment type. In the Pro B cell lines, V segments and the enhancer at TCRβ are methylated at a “low inactive” level, which is comparable to that of IgH V segments (Fig. 2 A and B). In contrast, a “high inactive” level of methylation is found at the J and C gene segments at TCRβ. A similar pattern is seen at the inactive Igλ locus in the Pro B cell lines. Particularly striking is that this reproducible pattern is maintained despite the physical separation and intermingling of gene segments of different types across the TCRβ and the Igλ loci (Fig. 1 A). By statistical T test, the probability of the segment-specific “high inactive” and “low inactive” levels of methylation occurring by chance is <10^–6. In contrast, the inactive Igκ locus does not display this segment-specific pattern of di-Me H3-K9 (Fig. 2B). The presence of only “low inactive” H3-K9 methylation across the Igκ locus may reflect the different nature of control at this locus (39, 40) and may explain why some κ rearrangement can occur before complete IgH rearrangement (41).

BRG1 Is Broadly Distributed Across Accessible Loci in both Pro B and Pro T Cells. Although nucleosome-remodeling complexes are often recruited to specific target sites (42–45), we find here that association of BRG1 at IgH and TCRβ loci closely correlates with that of Acetyl H3 (see Fig. 5) and hence spans essentially the entire recombinationally accessible domain. In both the Pro B and Pro T cell lines, BRG1 is enriched at rearranging D and J segments relative to nonrearranging V segments (Fig. 1C) at levels of enrichment equivalent to that seen around promoter elements in other studies. (The ability to see such high levels of enrichment strongly suggests that BRG1 is present across the entire locus in most cells in the population, rather than at localized high levels in individual cells.) In contrast, the level of BRG1 at all loci in NIH 3T3 cells is essentially the same as the level at inactive V segments in Pro B cells. The levels of BRG1 at CAD are equivalent among all three cell types.

Peaks of di-Me H3-K4 Mark the Borders of Active Regions. Pronounced hotspots of di-Me H3-K4 are found at the ends of the DJ loci in cell lines where those loci are poised to rearrange (Figs. 3 and 5). At the IgH locus, these high peaks of di-Me H3-K4 (30- to 40-fold enrichments) are located at the first (most 5′) DH segment, DFL16.1 and downstream at the first C region (Cμ) (Fig. 3, see asterisks). These positions correspond to the ends of the active region as defined by acetylation in transformed (Fig. 1B and refs. 25 and 26) and primary cells (25) as well as BRG1 (Fig. 1C), di-Me H3-K9 (Fig. 2), and di-Me H3-K79 (46). More detailed analyses of the distribution of di-Me H3-K4 in the vicinity of these peaks show that the levels of methylation drop rapidly, within 2 kb of the central peak (Fig. 6A, which is published as supporting information on the PNAS web site). In comparison, the level of di-Me H3-K4 across the IgH locus in NIH 3T3 cells is undetectable.

Strikingly, a similar pattern of di-Me H3-K4 hotspots is observed in the Pro T cell line but now, in a lineage-appropriate fashion, at the TCRβ locus. At this locus, where there are two separate DJC clusters, three peaks are found that encompass these two regions (Fig. 3, see asterisks): one at the 5′ end of the first cluster (at Dβ1, a 30-fold enrichment); the second, a 70-fold enrichment, starting at Cβ1 and continuing into the 3-kb region that separates the two clusters; and the third (a 70-fold enrichment) downstream of the second cluster (starting at Cβ2). Enrichment of di-Me H3-K4 decreases between Cβ2 and Eβ, revealing two distinct peaks whose function may or may not be linked (data not shown).

As seen in Fig. 6B, the peaks of di-Me H3-K4 are not coincident with high levels of Acetyl H3 but rather are found where the levels of Acetyl H3 are at their minima in both loci. This pattern of discrete peaks of di-Me H3-K4 is in marked contrast with the previously observed correlation of di-Me H3-K4 with Acetyl H3 across active loci (3). However, the moderate elevation (≈6- to 10-fold) of di-Me H3-K4 seen here across active regions is more in keeping with this correlation of di-Me H3-K4 with active loci. In addition, it is elevated at the transcriptionally active CAD promoter in the three cell types analyzed. These two distinct patterns of di-Me H3-K4 distribution suggest that this modification may serve two roles: one as a mark of active loci and the other as a modification associated with the ends of active domains.

Histone H4-Serine 1 Phosphorylation at Antigen Receptor Loci in Nonlymphoid Cells. Antibodies specific for phosphorylated serine 1 of histone H4 (Phos H4-S1), a modification whose functional correlates are not yet known (7), yielded an interesting pattern. Phos H4-S1 is enriched across all IgH and TCRβ gene segments (Fig. 4), as well as at Igκ and Igλ gene segments (data not shown) in NIH 3T3 cells where no rearrangement occurs. Similarly high levels of enrichment were also observed in a mouse embryonic fibroblast line (data not shown). In contrast, relatively low levels of phosphorylation are present at these same gene segments in Pro B and Pro T cells, leading to the speculation that this modification may be involved in the tissue-specific regulation of V(D)J recombination. In addition, two sites with higher levels of Phos H4-S1 enrichment are present: one at the 3′ end of the V region and one downstream of the 3′ di-Me H3-K4 mark, at Cγ3 of the IgH locus in Pro B cells and, to a lesser extent, in Pro T cells (Fig. 4, see asterisks). The reduced levels of Phos H4-S1 at the transcriptionally active CAD gene in NIH 3T3 cells as well as Pro B and Pro T cells suggest that this modification may serve an inactivating role. Although the relative changes in phosphorylation are subtle by our analysis, the results are clear and reproducible. By analogy to di-Me H3-K9 (3), the relatively low enrichment of Phos H4-S1 may be the result of high overall levels of phosphorylation across the genome.

Histone Modifications Reflect the Limited Accessibility of the IgH Locus in Pro T Cells. The work presented here provides a molecular explanation, consistent with a previous suggestion (26), for why developing T cells permit a low level of rearrangement between IgH D and J segments. Marks of active chromatin, elevated Acetyl H3, and di-Me H3-K4 are present across the IgH DJC region in Pro T cells, albeit at much lower levels than in Pro B cells (compare Fig. 7, which is published as supporting information on the PNAS web site, with Figs. 1B and 3). [Elevated Acetyl H4 has also been observed at JH segments in another T cell line (26).] Unlike the Jβ segments in Pro B cells, which have a “high inactive” level of di-Me H3-K9, DH and JH gene segments in Pro T cells are associated with “low inactive” methylation, analogous to the “low inactive” level seen across the Igκ locus in Pro B cells (Fig. 2 B and C). Thus, the chromatin at the IgH D and J loci in Pro T cells appears to be in a partly recombinase-accessible state, with the differences in enrichment of the marks compared with B cells probably translating into a reduced amount of accessibility/rearrangement.

The presence of DNaseI hypersensitive regions has previously been used to identify regions of the antigen receptor loci that might harbor regulatory elements or might be accessible for transcription and recombination. Similarly, the presence of germline transcripts has served as a criterion of broad-scale openness, whereas the developmentally ordered series of rearrangements defined for IgH and TCRβ also provide a clear indication of the regions of chromatin likely to be accessible at any given stage. ChIP analysis provides an alternative means of identifying potential regulatory sites and additional layers of regulation. As discussed below, the work here has provided new insight into possible modes of regulation of V(D)J recombination and has uncovered a number of unanticipated features of the chromatin structure at Ig and TCR loci. It is noteworthy that very similar (but lineage-appropriate) patterns are found at the IgH locus in the Pro B cell lines and at the TCRβ locus in the Pro T cell line. These patterns are found in independently derived cell lines from different lineages that were transformed by different means, strongly suggesting that the insights gained here will reflect general mechanisms of regulation at antigen receptor loci.

Chromatin Modifications Define Subregions Within Active and Inactive Domains. Within domains that appear generally accessible or inaccessible, the enrichment of a modification is not always uniform across the entire domain. Rather, we find different levels of enrichment that fall into distinct reproducible patterns. First, two distinct levels of H3-K9 methylation, “high inactive” and “low inactive,” are found at the TCRβ and Igλ loci in Pro B cells. Second, lower levels of acetylation are associated with D compared to J segments at both IgH and TCRβ loci. A similar pattern is seen with histone H3-K79 methylation (46). Third, among the JH segments, we note a subtle but reproducible increase in Acetyl H3 levels from 5′ (JH1) to 3′ (JH4). It is possible that these differences in modification levels allow for modulation of accessibility and rearrangement within an otherwise “open” or “closed” locus. In support of this view, the low level of Ig DJH rearrangement observed in Pro T cells correlates with the low levels of “activating” histone modifications, including Acetyl H3 and di-Me H3-K4, associated with the DH and JH segments in these cells. Further, JH usage is biased toward 3′ J segments in some genetic backgrounds (47, 48).

Our observations underscore the complexity of regulation at antigen receptor loci and suggest that the loci are not just uniformly “open” or “closed.” Rather, the results indicate that both the maintenance of antigen receptor loci in an inaccessible state along with their conversion from silent to accessible conditions will involve different modifications used in combination. Both the types of histone modifications and the relative amounts of each modification are likely to be critical to the regulation of this process. As has been suggested by others (25, 49), the opening of an antigen receptor locus (or chromatin domain) might occur in a stepwise fashion, with an initial global opening of the locus followed by additional modifications to render particular regions or segments accessible for recombination. Perhaps the global enrichment of Phos H4-S1 at antigen receptor loci only in nonlymphoid cells is yet another example of this multistep process. High levels of Phos H4-S1 are widely distributed across IgH, TCRβ, Igκ, and Igλ loci in two different fibroblast cell lines, suggesting that Phos H4-S1 may function to regulate V(D)J recombination in a tissue-specific manner.

BRG1 Remodeling Activity at Large Domains of Accessible Chromatin. The distribution of the nucleosome remodeling factor BRG1 is distinct from that previously seen in studies of transcriptional regulation (42–45). BRG1 spreads across large regions within the IgH and TCRβ loci that contain recombinationally accessible gene segments, in a pattern similar to Acetyl H3. The preference of BRG1-like complexes for binding to acetylated nucleosomes (50–52) might explain the association of BRG1 across these large regions of chromatin. Indeed, the work presented here provides a physiological example of an in vitro phenomenon, the preferential association of SWI/SNF with acetylated nucleosomes. The broad distribution of BRG1 suggests that this remodeling activity (and perhaps others) may play a role in achieving an open chromatin structure that permits RAG cleavage, as suggested by in vitro analyses (31). The distinction from transcription is likely to reflect the differing requirements of the two processes. In the case of transcription, nucleosome remodeling locally around promoters may be sufficient to allow transcription complexes to initiate. In contrast, the recombinase machinery requires access to gene segments distributed over large regions at the antigen receptor loci, so nucleosome remodeling may have to be more widespread.

di-Me H3-K4, a Mark for Chromatin Boundaries? Sharp transitions between active and inactive regions as defined by chromatin modifications have been seen at other loci. At the mating-type locus in fission yeast silent heterochromatic domains, which are associated with elevated levels of di-Me H3-K9, are juxtaposed to euchromatin that is marked by high levels of di-Me H3-K4 (4). Similar patterns have been observed at the β-globin locus in chickens where Acetyl H3 and di-Me H3-K4 are tightly linked to active gene segments in a pattern that inversely correlates with di-Me H3-K9 localization (3).

Here, we have identified a specific chromatin modification, di-Me H3-K4, that is discretely localized at the apparent ends of active domains. These peaks of di-Me H3-K4 are not coincident with peaks of Acetyl H3. Rather, they are found where Acetyl H3 enrichments are approaching “inactive” levels (Fig. 6B). That these prominent discrete peaks of di-Me H3-K4 are seen surrounding the active DJ regions of both IgH and TCRβ strongly argues for the functional importance of this modification in the regulation of antigen receptor gene rearrangement and suggests that this modification may be a key part of setting up or maintaining the long-range organization of these loci.

The precise localization of di-Me H3-K4 suggests that specific regulatory elements must exist, perhaps analogous to the insulator elements of the β globin locus control regions, that will be recognized by DNA-binding proteins. These specific DNA-binding proteins would then recruit a lysine 4 methyltransferase to these positions. The intriguing location of these peaks of di-Me H3-K4 suggests that this modification may serve as a boundary marker, preventing the spread of “repressive modifications” into the active region or to direct modifying activities to the locus in a developmentally appropriate manner. It will be of great interest to determine whether insulator or enhancer blocking activities will be found associated with the sites of H3-K4 methylation, what factors are involved in the recruitment of the methyltransferase, and how the observed chromatin modifications change during lymphoid development.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: TCR, T cell receptor; IgH, Ig heavy chain; Igκ or Igλ, Ig light chains; ChIP, chromatin immunoprecipitation; C, constant; E, enhancers.

We acknowledge Joseph Geisberg and members of the Struhl lab for assistance with real-time PCR analysis and Laura Corey for help with MNase digestion procedures. We thank Mary Donohoe, Robert Kingston, Eugene Oltz, Kevin Struhl, and members of the Oettinger lab for helpful discussion and critical reading of this manuscript. This work is supported by a Predoctoral Award from the Massachusetts General Hospital Fund for Medical Discovery (to K.B.M.), the Leukemia and Lymphoma Scholars Program (M.A.O.), and National Institutes of Health Grants GM48026 (to M.A.O.) and GM63959 (to C.D.A. and S.D.T.).