Influence of a CTCF-Dependent Insulator on Multiple Aspects of Enhancer-Mediated Chromatin Organization

Abstract

Developmental stage-specific enhancer-promoter-insulator interactions regulate the chromatin configuration necessary for transcription at various loci and additionally for VDJ recombination at antigen receptor loci that encode immunoglobulins and T-cell receptors. To investigate these regulatory interactions, we analyzed the epigenetic landscape of the murine T-cell receptor β (TCRβ) locus in the presence and absence of an ectopic CTCF-dependent enhancer-blocking insulator, H19-ICR, in genetically manipulated mice. Our analysis demonstrated the ability of the H19-ICR insulator to restrict several aspects of enhancer-based chromatin alterations that are observed during activation of the TCRβ locus for transcription and recombination. The H19-ICR insulator abrogated enhancer-promoter contact-dependent chromatin alterations and additionally prevented Eβ-mediated histone modifications that have been suggested to be independent of enhancer-promoter interaction. Observed enhancer-promoter-insulator interactions, in conjunction with the chromatin structure of the Eβ-regulated domain at the nucleosomal level, provide useful insights regarding the activity of the regulatory elements in addition to supporting the accessibility hypothesis of VDJ recombination. Analysis of H19-ICR in the heterologous context of the developmentally regulated TCRβ locus suggests that different mechanisms proposed for CTCF-dependent insulator action might be manifested simultaneously or selectively depending on the genomic context and the nature of enhancer activity being curtailed.

INTRODUCTION

Transcriptional insulators regulate the enhancer-promoter communication that orchestrates the epigenetic landscape of specific loci to activate or repress genes in metazoan genomes. Enhancers can regulate their cognate promoters by diverse mechanisms (1, 2). These may involve direct contact with the promoter by looping and/or alteration of the epigenetic landscape of large domains that render them “open,” i.e., associated with chromatin modifications that make them accessible to trans-acting factors. Alteration of the large domains could be due to “tracking” of some proteins initially bound at the enhancer or, as proposed during “facilitated tracking,” the movement of enhancer itself along the chromatin. Insulators in the vertebrate genomes have been proposed to modulate these interactions by binding to CTCF, a multi-Zn-finger protein (3). Genome-wide analysis of CTCF demonstrates its crucial role in higher-order chromatin organization that can facilitate enhancer-promoter interactions as well as curtail them depending on the relative positions of these regulatory elements (4). Importantly, insulators block enhancer-promoter communication only when present between them. Several models have been proposed for enhancer blocking, keeping in view the varied mechanisms underlying enhancer activity (5). It has been postulated that the CTCF-bound insulator partitions the enhancer and promoter to topologically distinct chromatin loops and thus prevents their interaction. The insulator might also prevent spread of an “open” chromatin state initiated at the enhancer or act as a promoter decoy to affect enhancer blocking. The extent of overlap between these possibilities is not clear, since only a few enhancer-blocking insulators have been analyzed in a developmental context. An understanding of the enhancer action at the molecular level is useful to dissect insulator activity and vice versa.

Enhancer-dependent chromatin organization is also important for lineage- and developmental stage-specific regulation of VDJ recombination that leads to generation of functional genes encoding immunoglobulins (Ig) and T-cell receptors (TCR) in mice. Hence, antigen receptor (AgR) loci, including TCRα/δ, TCRβ, IgH, and Igκ, provide a very useful framework to explore the role of regulatory elements in defining the chromatin structure and organization. While the role of enhancers and promoters at these loci has been extensively investigated by genetic analysis (6, 7), the possible contribution of insulators is beginning to be elucidated. AgR loci are particularly rich in CTCF/cohesin binding sites that can potentially impact VDJ recombination by modulating locus contraction and/or by acting as insulators (8, 9). The CTCF-binding element IGCR1 at the IgH locus and Sis and Cer at the Igκ locus have been demonstrated to block the activity of enhancers Eμ and iEκ, respectively, by deletion analysis (10, 11), and TCRβ has been suggested to have a bimodal insulator that insulates the recombination center (12, 13). Additionally, insertion of H19-ICR of the imprinted Igf2/H19 locus to the TCRβ locus (Fig. 1) led to organization of an ectopic CTCF-dependent insulator that effectively blocked the Eβ activity and led to impaired transcription and recombination patterns in the mutant mice (14).

FIG 1.

Schematic diagram of murine Igf2/H19 and TCRβ loci and binding of CTCF to TCR-ins. (A) Endogenous Igf2/H19 locus showing the relative positions of Igf2 and H19 genes and endodermal enhancers (EE) that activate them. H19-ICR organizes a CTCF-dependent insulator and prevents endodermal enhancer-based activation of the Igf2 promoter on the maternal allele. H19-ICR and H19 genes are deleted in H19del13 alleles and replaced with the neomycin resistance gene (Neo-r). (B) (Top) Endogenous TCRβ locus showing relative positions of 31 V gene segments, enhancer Eβ, and promoters PDβ1 and PDβ2, which drive the expression of the DJCβ1 cluster (Dβ1-Jβ1.1-Jβ1.7-Cβ1) and the DJCβ2 cluster (Dβ2-Jβ2.1-2.7-Cβ2), respectively. Recombination signal sequences (12RSS or 23RSS) are located downstream of each V, upstream and downstream of each D, and upstream of each J gene segment but are not shown. (Bottom) TCRβ alleles used in this study. The maternally inherited allele was wild type, TCR-ins, or TCR-mut, and the paternally inherited allele was either TCRβ-del or TCR-cas, as specified for each experiment. TCR-ins has an insertion of H19-ICR, TCR-mut has an insertion of H19-ICR-mut (with all four CTCF-binding sites mutated), TCRβ-del harbors a deletion that spans the Jβ2.3-Cβ2 region of the TCRβ locus, and TCR-cas is a congenic strain that exhibits several SNPs in the region of interest (depicted as a string of numeral signs, not to scale). Use of TCR-cas or TCRβ-del as the paternal allele afforded an allele-specific analysis of the region for evaluation of various parameters. (C) Binding of CTCF to H19-ICR in TCR-ins as detected by ChIP. The region spanning the third CTCF-binding site of H19-ICR was analyzed for enrichment by ChIP in thymocytes of TCR-ins/TCRβ-wt, H19del13/H19del13 mice. Enrichments are representative of ChIPs from three biological replicates.

Organization of the ectopic insulator at the TCRβ locus by H19-ICR (14) provided a very useful system, for several reasons, to evaluate enhancer-promoter-insulator interactions and the impact of the insulator on various aspects of enhancer-mediated chromatin organization. First, enhancers at the AgR loci may arguably be more complex, as they regulate transcription as well as recombination. Second, Eβ regulates two promoters, PDβ1 and PDβ2, and their linked transcription units DJCβ1 and DJCβ2, respectively, and H19-ICR was positioned (Fig. 1B) such that the influence of the insulator could be unambiguously investigated in a position-dependent manner. Finally, it was particularly interesting that Eβ activates the TCRβ locus in a stepwise manner and exhibits “looping” with the cognate promoter as well as “opening” of a large chromatin domain during early stages of T-cell development (DN2/3 stage) (15). Biochemical and genetic analysis suggests that Eβ, loaded with associated transcription factors, acts as a long-range accessibility control element (ACE) in DN2/3 cells and generates a partially “open” chromatin structure, marked by acetylated histones. This is followed by direct interaction or “contact” by “looping” of the enhancer with PDβ1 and PDβ2, which regulate the chromatin accessibility in a more localized manner at Dβ1 and Dβ2. Since accessibility at DJCβ1 was at least partially generated by Εβ on PDβ1-deletion alleles, it was suggested that Εβ-based regulation involved some form of “tracking” in addition to enhancer-promoter “looping” (16, 17). Formation of this looping-dependent holocomplex additionally recruits factors required to initiate germ line transcription. The resultant chromatin structure facilitates access of RAG proteins to recombination signal sequence (RSS) elements flanking Dβ and Jβ segments, leading to Dβ-to-Jβ recombination and eventually Vβ-to-DβJβ recombination to generate a functional TCRβ gene. The TCRβ locus, in the wild-type and mutant TCR-ins and TCR-mut alleles (Fig. 1B) (14), provided a very useful system to dissect enhancer-promoter-insulator interaction in the wake of an enhancer activity of Eβ that differs from those of the enhancers at the Igf2/H19 locus.

At the Igf2/H19 locus (Fig. 1A), a common set of enhancers activates the Igf2 and H19 genes and exhibits contact-dependent “looping” interactions with target promoters (18). Parentally imprinted monoallelic expression of Igf2 is regulated by the CTCF-dependent activity of the H19-ICR insulator (19–23). H19-ICR binds CTCF and organizes a functional insulator only on the maternally inherited allele and thus prevents Igf2 expression from it. Since CTCF cannot bind to methylated CpG dinucleotides, the functional insulator is not organized on the paternally inherited allele whose H19-ICR is methylated. This leads to paternal allele-specific activation of Igf2. It has been proposed that the enhancer region scans the chromatin in search of the cognate promoter. In the presence of the CTCF-bound H19-ICR (insulator) on the maternally inherited allele, the enhancers are stopped from tracking any further and are unable to activate the Igf2 promoter (24). Unlike the case with the TCRβ locus and several other loci, enhancer-based activation of Igf2 does not involve modulation of chromatin structure in terms of histone modification of a large region (25).

Our analysis of H19-ICR in the heterologous context of the TCRβ locus demonstrates that a single type of insulator can curtail different aspects of enhancer activities that have been proposed to regulate genes in different developmental contexts. Our analysis also tests and supports the accessibility hypothesis of VDJ recombination with a novel approach of modulating the functional interactions between regulatory elements rather than deleting them.

MATERIALS AND METHODS

Mice.

All experiments used mice as approved by the Institutional Animal Ethics Committee. Ex vivo thymocytes were isolated from 4- to 6-week-old mice of specified genotypes.

ChIP assay.

The Low Cell chromatin immunoprecipitation (ChIP) kit (Diagenode; catalog no. kch-maglow-A48) protocol was used with minor modifications. Briefly, 1 × 10⁶ thymocytes were cross-linked (1% formaldehyde, 8 min, 25°C) and quenched with glycine. Their genomic DNA (gDNA) was sheared by sonication to an average size of 150 bp. ChIP-grade antibodies anti-H3K9-acetyl (catalog no. 07-352) from Diagenode; anti-H3K4-trimethyl (catalog no. 17-614), anti-H3K9-dimethyl (catalog no. 07-441), anti-H3K27-trimethyl (catalog no. CS200603), and anti-CTCF (catalog no. 07-729) from Upstate Biotech; and anti-RNA polymerase II (anti-RNAPII) (catalog no. SC-899x) and anti-CBP (catalog no. SC-369X) from Santa Cruz Biotech were used. Anti-Rag2 antibodies were a kind gift of David Schatz. For Rag2 ChIP, the cross-linking was in 1% formaldehyde (15 min at 25°C). Anti-rabbit IgG was used as a negative control for all ChIPs. The immunoprecipitated DNA was analyzed by PCR for anti-CTCF ChIP. For all others, the analysis was performed by quantitative PCR (qPCR) using ABI Prism SDS 7000 and SYBR green (ABI catalog no. 4309155) in triplicate. The average threshold cycle (C_T) was used for quantitation by the comparative C_T method. The relative enrichment of the test amplicons was calculated by the following expression: DNA after ChIP/DNA in input fraction = 2^{−(CT [ChIP] − CT [input])} = 2^−ΔCT.

The enrichment of test amplicons was normalized to the enrichment of the β-actin or γ-actin amplicon. Primers for allele-specific PCR were designed considering the single-nucleotide differences between Mus musculus domesticus (maternal) and Mus musculus castaneus (paternal) alleles (here referred to as domesticus and castaneus, respectively). Independent experiments confirmed that the amplification is specifically from the domesticus DNA (Table 1) for each allele-specific amplicon.

TABLE 1.

Specificity of amplification during allele-specific qPCR for ChIP analysis^a

Amplicon name	CT of gDNA
	D/D	C/D	C/C
A (also named Pr1)	24.89	25.81	Undetected
B	21.58	22.63	38.59
C	23.04	24.11	44.85
Pr2	21.07	22.56	33.6
En	22.14	23.06	35.38

^aGenomic DNA was derived from Mus musculus domesticus (D/D), Mus musculus castaneus (C/C), and the F₁ progeny (C/D) of a cross between the two. Allele-specific qPCR was carried out for the mentioned amplicons using equal amounts of gDNA as the template and SYBR green-based detection in real time. Relative C_T values for different gDNA templates were determined. For each amplicon, as expected for an allele-specific PCR that detects domesticus DNA, the C_T value of C/D gDNA was increased by approximately one cycle relative to that of D/D while the C_T value of C/C gDNA was increased by at least 10 cycles relative to that of D/D gDNA. Thus, each amplicon was suitable for domesticus allele-specific detection of DNA in the SYBR green-based qPCR assay.

Restriction enzyme accessibility assay.

Nuclei from 6 × 10⁶ thymocytes (isolated from 5 or 6 mice) were analyzed for the accessibility of restriction enzymes as described previously (26, 27). Briefly, 1 × 10⁶ nuclei were treated with 0 U, 1 U, or 10 U of RsaI or HaeIII at 37°C for 10 min. Purified DNA (200 ng) was used for ligation-mediated PCR (LM-PCR). Nested gene-specific primers were used for allele-specific qPCR using SYBR green on ABI Prism SDS 7000. Alternatively, the LM-PCR products were subjected to Snapshot analysis (ABI Snapshot multiplex kit, catalog no. 4323159) that incorporated specific fluorescently labeled dideoxynucleotides by single-nucleotide primer extension based on nucleotide differences between domesticus and castaneus alleles as described previously (14).

CpG methylation analysis.

Thymocyte gDNA, digested with HindIII, was used for sodium bisulfite-based conversion of cytosine to thymine in agarose beads followed by PCR as described previously (22). PCR products were cloned and sequenced to deduce CpG methylation status in the context of the parental origin of clones.

3C assay.

Chromosome conformation capture (3C) analysis was performed as described previously (28) with minor modifications. Briefly, 10 × 10⁶ ex vivo thymocytes (from 8 to 10 mice) were cross-linked with formaldehyde (1%, 10 min at 25°C), followed by quenching. Lysis of the cells was facilitated by 15 strokes in a Dounce homogenizer. Chromatin was digested with 800 U of HindIII overnight, diluted, and ligated. Ligated DNA was purified and used for an allele-specific 3C-qPCR assay that specifically amplified DNA originating from the domesticus allele in the 3C libraries (Table 2). Additionally, a 3C-bacterial artificial chromosome (BAC) library was prepared from TCRβ, Ercc3, and Igf2/H19 locus-derived BAC clones RP23-421M9, RP23-342K14, and RP24-251H17 (Bacpac Resources, USA) and MSMG01-426I12 (BRC, Riken, Japan). DNA from the BAC clones, mixed in an equimolar ratio, was digested with HindIII, ligated, and purified. The relative quantity of the test interacting regions in the 3C libraries, represented as cross-linking frequency between them, was estimated based on the standard curve generated by using various dilutions of the 3C-BAC library as the templates for each 3C-qPCR experiment. The amplification was monitored on ABI Prism SDS 7000 using PCR mix (ABI catalog no. 4304437) and custom-designed TaqMan probes. The enhancer-promoter interaction at the Ercc3 locus was also estimated in each 3C library using the 3C-BAC libraries. Finally, relative cross-linking frequency was calculated (cross-linking frequency of the test regions in the 3C library/cross-linking frequency of the Ercc enhancer to the promoter).

TABLE 2.

Specificity of amplification during allele-specific 3C-qPCR analysis^a

Amplicon	CT (3C library)
	D/D	C/C
Enh-P1	36.7	Undetected
Enh-P2	33.7	Undetected
Enh-C2	35.1	Undetected

^a3C libraries were constructed from Rag1-deficient TCRβ-wt/TCRβ-wt (D/D) and Rag1-deficient TCR-cas/TCR-cas (C/C) mice and therefore had domesticus or castaneus DNA at the TCRβ locus, respectively. Equal amounts of DNA were used as the template for 3C-qPCR. PCR amplification was monitored in real time using TaqMan probes. Relative C_T values were determined and compared for the 3C libraries for amplicons Enh-P1, Enh-P2, and Enh-C2. While D/D gave amplification, C/C-based amplification was undetected in each of these. Thus, these amplicons were suitable for domesticus allele-specific detection of 3C products in the 3C-qPCR assay. Primer “Ins” was located on H19-ICR. Therefore, all amplicons that used it (e.g., Ins-Enh, Ins-P1, etc.) were domesticus allele specific. Amplicons Enh-N1 and Enh-N2 were not verified for allele specificity, as their signals were very low in the 3C libraries.

Primer sequences for all assays are available upon request.

RESULTS

Insertion of H19-ICR into the TCRβ locus (Fig. 1B, TCR-ins) established a CTCF-dependent insulator (14). The inserted H19-ICR was able to bind CTCF when maternally inherited, as evidenced by ChIP analysis carried out in thymocytes of TCR-ins/+, H19del13/H19del13 mice (Fig. 1C). The observed enrichment unambiguously indicated the ability of H19-ICR of TCR-ins to bind CTCF, since H19-ICR of the Igf2/H19 locus has been deleted in the H19del13 alleles (29). Consistent with the ability of the insulator to block enhancer-promoter interactions when positioned between them, the Eβ-PDβ1 interaction-dependent transcription and Dβ-to-Jβ recombination were severely abrogated at the DJCβ1 gene cluster but the Eβ-PDβ2-dependent transcription and Dβ-to-Jβ recombination at the DJCβ2 cluster were unaffected by H19-ICR insertion. Further, mutations of the four CTCF-binding sites have been shown to abrogate CTCF binding in vitro and in vivo (30). Insertion of H19-ICR mutated at these sites (Fig. 1B, TCR-mut) exhibited transcription and recombination profiles similar to those of the wild-type allele and confirmed the CTCF dependence of insulator activity (14).

Using TCR-ins and TCR-mut alleles, we explored the insulator-based perturbation of enhancer-promoter communication. First, several parameters of chromatin structure of the TCRβ locus were investigated that are under the control of Eβ and hence could possibly be altered due to the ectopic insulator. Next, we analyzed the manner in which this CTCF-dependent insulator might abrogate Eβ activity.

Experimental design.

Enhancer blocking by H19-ICR was observed upon maternal inheritance of TCR-ins (14). Hence, ex vivo thymocytes from mutant mice that inherited the wild-type TCRβ or TCR-ins or TCR-mut allele maternally and either the TCRβ-del or TCR-cas allele paternally were used (Fig. 1B). The 15-kb deletion in paternal TCRβ-del (31) ensured that the PCR-based assays provide maternal allele-specific information. For analysis of regions not deleted by TCRβ-del, TCR-cas was used as the paternal allele, which has the DNA of Mus musculus castaneus origin at the TCRβ locus (14). The maternal alleles (wild type, TCR-ins, or TCR-mut) were of Mus musculus domesticus origin. Single nucleotide polymorphisms (SNPs) between the domesticus and castaneus strains of mice could, therefore, be used to derive allele-specific information.

All the mutant mice were additionally deficient for Rag1 and hence provided thymocytes which are arrested at the DN3 stage of development (32). This ensured that in the majority of the cells, Eβ was active and the prerecombination state of the TCRβ locus was maintained.

Alterations in activating histone modifications.

The chromatin landscape of the 25-kb region gets enriched in acetylated histone H3K9-acetyl prior to recombination due to Eβ activity (17, 33). We compared the association of H3K9-acetyl and that of H3K4-trimethyl, which signifies the transcriptionally activated status of chromatin. Although several regions of the Eβ-regulated domain exhibited enrichment in the wild-type TCRβ allele, the association of both these marks was notably high at PDβ1/Dβ1 and the Jβ segments of DJCβ1 as well as the DJCβ2 clusters (Fig. 2B and C, amplicons A to I).

FIG 2.

Influence of ectopic H19-ICR insulator on the presence of activating histone modifications H3K9-acetyl and H3K4-trimethyl on the TCRβ locus. (A) TCR-ins allele showing relative positions of the H19-ICR insertion on the TCRβ locus; regulatory regions PDβ1, PDβ2, and Eβ; and gene segments D, J, and C as well as PCR amplicons used for allele-specific ChIP analysis. Allele-specific PCR for amplicons A to C detected the enrichment of test regions on the maternally inherited domesticus allele (+, TCR-ins, or TCR-mut) and did not amplify the paternally inherited castaneus allele (TCR-cas) due to mismatched nucleotides in primers (represented as numeral signs below amplicons A to C). This was independently verified for each allele-specific amplicon (Table 1). Amplicons D to J detected the enrichment of test regions present exclusively on the maternally inherited allele (+ or TCR-ins) due to the deletion of the test region from the paternally inherited TCRβ-del (deletion shown below amplicons D to J). Thus, each amplicon (A to J) reported the enrichment specifically as being on the maternal allele. Genotypes of mice used for ChIP are defined in the sample key at the bottom of the figure. (B and C) Enrichment of H3K9-acetyl (B) and H3K4-trimethyl (C) on the TCRβ locus relative to β-actin on the maternally inherited wild-type TCRβ allele (+, light gray bars), mutant TCR-ins allele (dark gray bars), and mutant TCR-mut allele (white bars). The paternal allele was TCR-cas (left graphs) or TCRβ-del (middle and right graphs). Enrichment in ChIP with IgG (black bars) was mostly undetectable. Enrichments shown are means (plus standard errors of the means) from three biological replicates. Enrichments in TCR-ins and TCR-mut alleles were compared to the TCRβ-wt allele by Student's t test. *, P < 0.05.

The enrichment for the same histones was drastically reduced in the TCR-ins allele for regions located upstream of the inserted insulator (Fig. 2B and C, amplicons A to F). However, PDβ2, Jβ2, and Cβ2 segments located downstream of the insulator did not show significant differences in the presence of H3K9-acetyl or H3K4-trimethyl from the wild-type allele (Fig. 2B and C, amplicons G to J). Further, the CTCF dependence on the effect of the insulator was clearly evident from the high enrichment of H3K9-acetyl and H3K4-trimethyl in the TCR-mut allele at the PDβ1/Dβ1 region (Fig. 2B and C, amplicons A to C).

Repression of promoter PDβ1.

To confirm the repressed state of the promoter, we analyzed CpG methylation and association of repressive histone modifications. The TCR-ins allele exhibited an increase in enrichment in H3K9-dimethyl in the PDβ1 region (Fig. 3A) compared to the transcriptionally active wild-type allele, especially at the region encompassing core PDβ1 and Dβ1 (amplicon A). Some degree of enhancement of H3K27-trimethyl association on the TCR-ins allele was also apparent but showed considerable variation and was not statistically significant. The TCR-mut allele did not exhibit a repressed PDβ1.

FIG 3.

Alteration in the chromatin status of promoter PDβ1 as ascertained by presence of modified histones H3K9-dimethyl and H3K27-trimethyl as well as DNA methylation. (A) The schematic shows PDβ1 and the positions of amplicons A, B, and C used for allele-specific PCR that detected the enrichment specifically on maternally inherited alleles. Due to mismatched nucleotides in the primers (represented as numeral signs below amplicons A, B, and C), the paternal allele (TCR-cas) was not amplified (verified independently and shown in Table 1). Genotypes of mice used for ChIP are defined in the sample key at the bottom of the figure. The enrichment by ChIP using anti-H3K9-dimethyl and anti-H3K27-trimethyl relative to β-actin was analyzed on the maternally inherited wild-type TCRβ allele (light gray bars), mutant TCR-ins allele (dark gray bars), and mutant TCR-mut allele (white bars). Enrichment in ChIP with IgG (black bars) was undetectable in most cases. Enrichments shown are the means (plus standard errors of the means) from three biological replicates. Enrichments in TCR-ins and TCR-mut alleles were compared to the TCRβ-wt allele by Student's t test. *, P < 0.05. (B) The schematic shows the positions of eight CpG dinucleotides encompassing PDβ1, Dβ1, and a small region downstream of Dβ1. CpG5 and CpG8 (marked with numeral signs) are present on the domesticus allele (maternal) and absent on the castaneus allele (paternal) due to SNPs. The CpG methylation status in thymocyte gDNA is shown for both parental alleles in TCRβ-wt/TCR-cas, Rag1-deficient and TCR-ins/TCR-cas, Rag1-deficient mice. Each set of 8 circles represents a clone. Empty circles show unmethylated cytosines in the CpG dinucleotides examined, and filled circles show methylated cytosines. The data were derived from cloning of four PCR products originating from gDNA of two mutant mice in each case.

CpG methylation is a critical determinant of the repressed state. Eβ activity is responsible for demethylation of the region regulated by it (33). As expected, CpG residues in the PDβ1-Dβ1 region of wild-type alleles, i.e., both alleles in +/TCR-cas mice as well as the paternally inherited TCR-cas allele in TCR-ins/TCR-cas mice, were not methylated (Fig. 3B). In contrast, the PDβ1-Dβ1 region of the maternally inherited TCR-ins allele, which organizes a functional insulator, was hypermethylated.

These analyses confirmed that the functional insulator prevents the Eβ-based remodeling of the promoter PDβ1. The promoter stays repressed in terms of various epigenetic marks that were investigated.

Alterations in chromatin accessibility.

Eβ-based enhancement of H3K9-acetyl generates accessibility of chromatin for trans-acting factors necessary for transcription and recombination (17) and was estimated by the access of restriction enzyme RsaI or HaeIII to its recognition site (Fig. 4A). The accessibility of the RsaI site immediately upstream of PDβ1 (Fig. 4B, site R-1) was reduced in the TCR-ins allele compared to the wild-type allele. At Dβ1 (site H-1), the reduction was more remarkable and evident even when 10 U of HaeIII was used. Also, Jβ1.3 accessibility (site R-2) was enormously decreased in the TCR-ins mutant allele compared to the wild type (Fig. 4B and C). The sites downstream of the point of insulator insertion, i.e., in the vicinity of promoter PDβ2 (site R-3) and Jβ2.5 (site R-4), in TCR-ins were almost as accessible to RsaI as in the wild-type allele. The marginal reduction remains unexplained but clearly did not adversely affect transcription and recombination in TCR-ins (14).

FIG 4.

Alteration in restriction enzyme accessibility due to the presence of an ectopic insulator. (A) TCR-ins locus showing positions of RsaI sites (R-1 to R-4) and HaeIII site (H-1) used for the accessibility assay. The region encompassing R-1 and H-1 has SNPs (represented as numeral signs below) between domesticus and castaneus alleles, while the region encompassing R-2, R-3, and R-4 is deleted in TCRβ-del. (B) Relative accessibility of the wild-type and TCR-ins mutant alleles as determined by LM-PCR. The thymocyte nuclei used for estimating the accessibility at R-1 and H-1 were from +/TCR-cas and TCR-ins/TCR-cas mice. The LM-PCR-based amplified DNA was subjected to Snapshot analysis to estimate the relative contribution of the domesticus DNA to the total LM-PCR product [domesticus DNA/(domesticus DNA + castaneus DNA)], as shown for the maternally inherited wild-type TCRβ allele (light gray bars) and the TCR-ins allele (dark gray bars). Data are means (plus standard errors of the means) from four Snapshot reactions from two biological replicates. Accessibility at R-2, R-3, and R-4 was estimated in nuclei of +/TCRβ-del and TCR-ins/TCRβ-del thymocytes. The abundance of digested ends was calculated by qPCR in the LM-PCR for the TCR-ins allele (dark gray bars) relative to the wild-type TCRβ allele (light gray bars) after normalization with the accessibility of an RsaI site in a β-actin gene in each. Data are means (plus standard errors of the means) from three LM-PCR experiments from two biological replicates. The accessibilities of TCR-ins alleles were compared to that of the TCRβ-wt allele by Student's t test. *, P < 0.05. (C) LM-PCR products for R-2, R-3, and R-4 of a representative experiment as visualized by agarose gel electrophoresis.

Altered Rag2 binding pattern.

As reported for the J segments of other AgR loci (34), we observed that association of H3K4-trimethyl was very high at the Jβ segments. The H3K4-trimethyl mark on the histones is recognized by Rag2 (34), an important cofactor for Rag1 activity. A high degree of association of Rag2 was observed with the amplicons located in DJCβ1 as well as in DJCβ2 in the wild-type allele (Fig. 5, amplicons D, K, I, and L) but only with DJCβ2 amplicons (I and L) in the TCR-ins alleles. Consistent with a complete loss of H3K4-trimethyl due to insulator activity (Fig. 2C), the DJCβ1 amplicons (D and K) had practically no association with Rag2 in TCR-ins.

FIG 5.

Influence of the ectopic H19-ICR insulator on occupancy of the TCRβ locus by Rag2. (A) TCR-ins allele showing relative positions of the H19-ICR insertion on the TCRβ locus; regulatory regions PDβ1, PDβ2, and Eβ; and gene segments D, J, and C as well as PCR amplicons used for ChIP analysis using anti-Rag2 antibody. All amplicons (D, K, I, and L) detected the enrichment of test regions exclusively on the maternally inherited allele (+ or TCR-ins) due to deletion of the test region from the paternally inherited allele (TCRβ-del). Genotypes of mice are defined in the sample key. (B) ChIP-based enrichment of Rag2 on the TCRβ locus relative to γ-actin (g-actin) was analyzed on the maternally inherited wild-type TCRβ allele (light gray bars) and the TCR-ins mutant allele (dark gray bars). The enrichment of amplicon K for the TCR-ins allele was below detectable levels. The mock ChIP with IgG was analyzed for enrichment in all samples, which is shown when detected (black bars). Enrichments shown are representative of two biological replicates.

Changes in RNA polymerase II and CBP occupancy.

All the analysis described above demonstrated the outcome of insulator-based interference in several aspects of Eβ-mediated alteration of the chromatin structure in a position-dependent manner. The molecular basis of this interference was further investigated in the context of the possible “looping” interactions between the regulatory elements. Eβ was suggested to act as a loading platform for RNA polymerase II (RNAPII) and a transcriptional activator, CREB binding protein (CBP). Subsequently, these proteins are transferred to the target promoters PDβ1 and PDβ2 as the holocomplexes are stabilized due to Eβ-PDβ1 and Eβ-PDβ2 interactions established via looping of chromatin (16). Potentially, the insulator can interfere in formation of the chromatin loop or in the transfer of the RNAPII/CBP complex to the target promoter or exert its influence on chromatin structure at a later stage.

The possible interference of the insulator in transfer of RNAPII and CBP was investigated by ChIP. ChIP primers were designed to detect maternally inherited alleles (Fig. 6A and B, amplicons Pr1, Pr2, En, TH1, and TH2), while amplicons M and N could detect both parental alleles. As expected, in the wild-type TCRβ allele, the presence of both RNAPII and CBP was observed at the two promoters PDβ1 and PDβ2 and at the enhancer (Fig. 6A, amplicons Pr1, Pr2, and En). Surprisingly, a high level of RNAPII binding to a region 2.2 kb upstream of PDβ2 (amplicon M) was also evident, although amplicon N, located at about the same distance upstream of PDβ1, did not exhibit enrichment for RNAPII. The simplest explanation for this could be the proximity of this region to the 3′ end of Cβ1, although the possibility of its interaction with Eβ bound to RNAPII is also consistent with this observation.

FIG 6.

Binding of RNA polymerase II (RNAPII) and CREB binding protein (CBP) on the TCRβ locus as influenced by the presence of the ectopic H19-ICR insulator. (A) The locus diagram shows the wild-type TCRβ loci of the two alleles in +/TCR-cas mice with the position of PCR amplicons used for ChIP analysis. Maternal and paternal alleles are both wild type but have domesticus and castaneus DNA, shown in red and blue lines, respectively. ChIP samples were analyzed by allele-specific PCR primer pairs at amplicons Pr1, Pr2, and En (red arrowheads and type) that amplify DNA of domesticus origin (wild-type maternal allele), but primers for amplicons M and N (red plus blue arrowheads, denoted as M+M and N+N) detect DNA of both parental alleles. The allele specificity for amplification of Pr1, Pr2, and En was verified independently (Table 1). Graphs show the enrichment of RNAPII (left) and CBP (right) relative to the promoter of β-actin in +/TCR-cas, Rag1-deficient thymocytes (light gray bars). (B) Locus diagram showing the TCRβ alleles in TCR-ins/TCR-cas mice with the positions of amplicons used for ChIP analysis. TCR-mut/TCR-cas mice have similar configurations, except that the CTCF-binding sites of H19-ICR are mutated. Amplicons TH1 and TH2 are located at the 5′ and 3′ junctions of TCRβ with H19-ICR and were designed to detect the allele with an H19-ICR insertion. Thus, in TCR-ins/TCR-cas as well as TCR-mut/TCR-cas thymocytes, all primer pairs used for ChIP analysis (amplicons Pr1, Pr2, TH1, and TH2, shown as red arrows and red type) detect only the mutant maternal allele (TCR-ins or TCR-mut, domesticus origin). Graphs show the enrichment of RNAPII (left) and CBP (right) relative to the promoter of β-actin in TCR-ins/TCR-cas, Rag1-deficient thymocytes (dark gray bars) and TCR-mut/TCR-cas, Rag1-deficient thymocytes (white bars). RNAPII was undetectable at amplicons Pr1 and TH1 in TCR-ins/TCR-cas, Rag1-deficient mice. The enrichment in the mock ChIP assay with IgG (black bars) was mostly undetectable. Enrichments shown are means (plus standard errors of the means) from three biological replicates. Enrichments in TCR-ins alleles were compared to those in TCR-mut alleles by Student's t test. *, P < 0.05.

In the TCR-ins allele, PDβ1 was completely devoid of RNAPII and exhibited lowered association with CBP, although both these proteins were seen to bind PDβ2 (Fig. 6B). In the context of the observations reported earlier (16), this suggested that the insulator was preventing the Eβ-based transfer of the holocomplex to PDβ1. This interference in RNAPII and CBP binding to PDβ1 was CTCF dependent, as TCR-mut exhibited the presence of the two proteins at both the promoters.

Next, the RNAPII and CBP presence was analyzed across H19-ICR. In TCR-mut, but not in TCR-ins, RNAPII was present immediately downstream of Cβ1 at the 5′ end of the inserted H19-ICR (Fig. 6B, amplicon TH1). This may be expected since PDβ1-initiated DJCβ1 transcription is retained in TCR-mut but is lost in TCR-ins. However, in TCR-ins as well as in TCR-mut, RNAPII was also bound at the 3′ end of H19-ICR (Fig. 6B, amplicon TH2). The presence of RNAPII at more than 2 kb upstream of the PDβ2 promoter was rather surprising as RNAPII is not typically associated with DNA more than 0.5 to 1 kb upstream of the promoters (35, 36). This observation, however, is consistent with the possibility of the insulator preventing the progression of enhancer-bound tracking by RNAPII/CBP, as suggested by the facilitated tracking model (37). The CBP binding profile was qualitatively similar to that of RNAPII, except that CBP binding was lower in TCR-ins even at TH2. The reason for the lowered CBP at this 3′ end of the insulator remains currently unknown. Speculatively, interaction of the enhancer with the insulator might lead to loss of CBP.

Altered enhancer-promoter interactions.

To further understand how the inserted insulator might disrupt Eβ-PDβ1 interaction, the interactions between enhancer, promoter, and insulator elements (Fig. 7A) were investigated by chromosome conformation capture (3C) assay.

FIG 7.

Interactions between enhancer, promoter, and insulator elements ascertained by 3C-qPCR analysis. (A) Schematic representation of the TCR-ins locus showing the point of insertion of H19-ICR, HindIII sites (H-up and H1 to H15), and locations of primers (orange arrowheads close to HindIII sites) that report the interaction of PDβ1 (P1), PDβ2 (P2), Eβ (Enh), and the H19-ICR insulator (Ins) as well as nonregulatory regions (C2, N1, and N2). The 3C analysis was carried out in thymocytes of Rag1-deficient mice that maternally inherited the wild-type TCRβ allele (+/TCR-cas, Rag1 deficient), the TCR-ins allele (TCR-ins/TCR-cas, Rag1 deficient), or the TCR-mut allele (TCR-mut/TCR-cas, Rag1 deficient) as mentioned. The allele specificity of the 3C-qPCR for each interaction was verified independently (Table 2). (B and C) The interactions analyzed by 3C-qPCR (domesticus allele specific) using the enhancer as an anchor (Enh) or H19-ICR as an anchor (Ins) are shown as cross-linking frequency relative to the enhancer-promoter interaction at the Ercc3 locus. Interactions presented are an average (plus standard error of the mean) for three biological replicates for each genotype. X denotes a nondetectable interaction due to the absence of H19-ICR in the wild-type TCRβ locus. *, P < 0.05 in Student's t test compared to TCRβ-wt allele (B) or TCR-mut (C). (D) Schematic to show a model that that takes into account the possible interactions between regulatory elements and their outcomes in the wild-type, TCR-ins, and TCR-mut alleles. The Eβ-regulated domain (sand colored) in the wild-type allele exhibits an activated chromatin (green dashed line) and Eβ-PDβ1 and Eβ-PDβ2 interactions (arrows) that are detectable by 3C. The functional insulator, H19-ICR, in TCR-ins partitions the domain into two such that one is activated (sand colored) while the other is insulated from Eβ and remains nonactivated (gray). The insulator also exhibits some degree of interaction (dotted line) with the enhancer. In the absence of CTCF, H19-ICR does not organize a functional insulator, and hence, TCR-mut resembles the wild-type allele.

Interactions of Eβ with PDβ1 and PDβ2 were investigated by 3C-qPCR (allele specific) (Fig. 7B) in Rag1-deficient +/TCR-cas, TCR-ins/TCR-cas, and TCR-mut/TCR-cas thymocytes. 3C-qPCR was designed to detect 3C signal originating specifically from the maternally inherited domesticus allele (wild type, TCR-ins, or TCR-mut). Using Eβ as the anchor, a very high degree of interaction of Eβ with PDβ1 as well as with PDβ2, about 10-fold higher than the Ercc enhancer-promoter interaction, was observed on the wild-type allele (light gray bars). The observed interactions among Eβ, PDβ1, and PDβ2 were not merely a consequence of the proximity, as the interaction of Eβ with control region N1 was barely detectable (Fig. 7B, Enh-N1). We also observed a high degree of interaction of Eβ with a region upstream of Cβ2 (Eβ-C2), which may be attributed to scanning of the region by Eβ but might also arise due to the proximity of C2 to Eβ and/or PDβ2 as it is merely about 6 kb away from each of these elements. Control region N2, located upstream of PDβ1 beyond the currently known regulatory domain of Eβ, did not show any interaction.

Strikingly, Eβ-PDβ1 interaction was significantly reduced in a CTCF-dependent manner. The reduction was observed in TCR-ins (dark gray bars) but not in TCR-mut (white bars). Further, Eβ interactions with regions downstream of inserted H19-ICR, i.e., Eβ-PDβ2 and Eβ-C2, were not altered by the insulator presence. The enhancer also did not exhibit interactions with N1 or N2.

Earlier studies have reported the interaction of H19-ICR with the target enhancers and promoters (18, 24) based on qualitative analysis. Using the enhancer as the anchor, we observed a rather low degree of interaction of H19-ICR with the enhancer (Fig. 7B, Enh-Ins). To rigorously examine the possible interactions of the insulator with the regulatory regions, we used the inserted H19-ICR as an anchor in 3C-qPCR analysis. Although PDβ1, PDβ2, and Cβ2 are located rather close to the inserted H19-ICR, and therefore likely to give a very high degree of interaction, we were surprised that the interactions of these regions with H19-ICR were also significantly lower than Eβ-PDβ1 and Eβ-PDβ2 interactions (Fig. 7B and C), which could be compared, taking interactions at Ercc as a reference. Functional as well as nonfunctional insulators, in TCR-ins and TCR-mut alleles, respectively, exhibited similar degrees of interaction with these closely located regions, suggesting that proximity, rather than insulator function, was responsible for the detected Ins-P1, Ins-P2, and Ins-C2 signals in 3C-qPCR. In contrast, H19-ICR showed interaction with Eβ, located about 15 kb away, in TCR-ins but not in TCR-mut (Fig. 7C, Ins-Enh). Thus, the Ins-Eβ interaction in TCR-ins was clearly CTCF dependent and linked to insulator function. Although much lower than those with Eβ-PDβ1 or Eβ-PDβ2 interactions in the wild-type allele, the Ins-Eβ interaction-based signal detected by 3C-qPCR was nearly half of the Ercc enhancer-promoter interaction.

In sum, the H19-ICR-based functional insulator prevented the Eβ-PDβ1 interaction, plausibly by interacting with Eβ. The reduced Eβ-PDβ1 interaction was dependent on CTCF binding by H19-ICR. The interactions analyzed by 3C correlated well with the position dependence of the insulator and the observed functional readouts of the interactions (14) and with the analysis of chromatin structure described above.

DISCUSSION

The epigenetic state of the TCRβ locus is regulated precisely in a lineage- and developmental stage-specific manner by interaction of accessibility control elements (ACEs) PDβ1, PDβ2, and Eβ (38). Under the influence of an ectopic CTCF-dependent insulator, all aspects of Eβ-mediated chromatin organization were severely curtailed at the TCRβ locus.

The Eβ-regulated chromatin domain extends to about the 25-kb region encompassing Εβ, PDβ1, PDβ2, and their linked transcriptional units (38). The inserted insulator appeared to have partitioned the Eβ-regulated domain into two, as the entire PDβ1-DJCβ1 cluster, located upstream from the insulator, was missing all marks of Eβ-mediated activation (Fig. 7D). Our 3C analysis demonstrated that the inserted H19-ICR altered the chromatin loop organization in a CTCF-dependent manner. Compared to the wild-type allele, the TCR-ins allele exhibited an unaltered Εβ-PDβ2 interaction, a reduced Εβ-PDβ1 interaction, and a gain of Εβ-insulator interaction. In the TCR-ins allele, the PDβ1 promoter was repressed and did not bind RNAPII. Histone modifications H3K4-trimethylation and H3K9-acetylation were missing, and association of Rag2 with Jβ1 gene segments was also abrogated. Thus, PDβ1-DJCβ1 chromatin was inaccessible to transcription and recombination, consistent with our previous observations (14). Since insulators disrupt enhancer-promoter interactions only when positioned between them, Eβ continued to modify the chromatin of PDβ2-DJCβ2 as in the wild-type TCRβ allele. The activation status of PDβ2-DJCβ2, estimated by various parameters, was not enhanced compared to the wild type even when PDβ1-DJCβ1 was rendered inactive in TCR-ins. This suggests that even though PDβ1 and PDβ2 share the Eβ enhancer, they do not strictly compete with each other for being activated by the enhancer.

Insulators have been proposed to curtail enhancer activity by varied mechanisms, perhaps because of diversity in the mechanisms underlying enhancer activity. There is considerable evidence that direct contact between the enhancer and its cognate promoter is required for transcriptional activation, and CTCF-bound insulators interfere in this process (1, 5). While 3C-qPCR analysis has clearly established enhancer-promoter interactions to be more stable and necessary for promoter activation at many loci, enhancer-promoter-insulator interactions have not been previously investigated by such a quantitative analysis. Our 3C-qPCR analysis revealed the insulator-Eβ interaction to be much lower than Eβ-PDβ1 or Eβ-PDβ2 interactions, suggesting that H19-ICR is not likely to have a competitive advantage over the promoters PDβ1 and PDβ2 for looping interactions with Eβ. It may, therefore, be inferred to be a rather weak decoy for the promoter PDβ1. In view of this, it is plausible that even a less stable enhancer-insulator engagement is detrimental to enhancer activity as, despite showing relatively weak interaction in a 3C-based analysis, the insulator was very effective. Analysis of a few more enhancer-promoter-insulator sets will hopefully establish if this is also the case for other CTCF-dependent insulators. As suggested for other loci (18), the insulator-enhancer interaction detected by 3C may be interpreted to indicate that H19-ICR is acting as a “decoy promoter.” In contrast, the possibility of the H19-ICR acting as a “decoy enhancer” could not be examined due to the relatively close location of the inserted H19-ICR to each of the promoters.

Since “decoy” models cannot satisfactorily explain the ability of insulators to be effective only when positioned between the enhancer and promoter, insulators have also been proposed to act as barriers to the progression of “tracking signals” initiated at the enhancer or to the “facilitated tracking” of enhancers as they scan the chromatin until the cognate promoter is contacted (24, 37). The nature of the “tracking signal” for enhancer action is not clear, but histone modifications as well as RNAPII/CBP have been previously suggested to be the “tracking signals” (39). In this context, we observed a complete loss of accessibility of the PDβ1-DJCβ1 region in the presence of the insulator. If there was any involvement of “tracking” in Eβ activity as proposed previously (17), the insulator organized by the ectopic CTCF binding was able to curtail it effectively. The engagement of Eβ with H19-ICR and the association of RNAPII at the 3′ end (but not the 5′ end) of H19-ICR hint at the possibility of H19-ICR preventing facilitated tracking by Eβ. However, in the absence of direct evidence to determine the directionality in the dynamic action of Eβ or any other enhancers, the idea remains speculative. Our observations are also consistent with the possibility of Eβ scanning the chromatin by making several less-frequent/unstable random contacts with other elements within its target domain and restriction of the target domain by the intervening insulator. This is consistent with the emerging views that dynamic interactions of a variable frequency occur within chromatin domains; interactions between enhancers and their cognate promoters are most frequent and/or stable but may be modulated by CTCF-binding sites residing within the same topologically associated domains (TADs) (40).

If CTCF binding to the insulator can influence the contact-dependent interactions between enhancers and promoters by reorganizing the chromatin loops and their dynamics, the insulator can also be expected to influence interactions between other cis-acting elements that require juxtaposition. We did observe a dramatic alteration in the choice of Vβ segments for RSS-mediated V-to-DJ recombination upon insertion of H19-ICR in the TCRβ locus (14). This observation is consistent with the possibility that ectopic CTCF-binding sites promoted intradomain RSS interactions and hindered interdomain RSS interactions, just as they seem to influence enhancer-promoter interactions. However, the mechanistic basis for the altered Vβ usage is not yet clear. To understand the relevance of CTCF-mediated chromatin organization for transcription and recombination, interactions among endogenous CTCF/cohesin-binding sites on the TCRβ need to be mapped in wild-type and mutant alleles. CTCF-based chromatin loop organization is important for VDJ recombination at IgH, Igκ, and TCRα/δ loci (9). However, the mechanistic details have similar as well as distinctive features that are possibly dependent on the relative organization of regulatory elements and gene segments. The chromatin hub organization at AgR loci is likely to be rather complex since both transcription and recombination are regulated by it.

Irrespective of how the insulator curtailed the generation of chromatin accessibility, the correlations observed in our experiments between chromatin structure, transcription, and recombination under the influence of altered functional interactions between regulatory elements, rather than their deletion, additionally support the tenets of the accessibility hypothesis (41). Transcription-based acquisition of H3K4-trimethyl histones leads to recruitment of Rag2 and stimulates the activity of the RAG complex (34). Also, transcription displaces histones from the RSS flanking the V, D, and J segments and provides accessibility to the RAG complex (42, 43). These observations seem to provide the link between transcription and recombination. However, experiments using miniloci and chromatin templates assembled in vitro emphasize that chromatin remodeling, rather than transcription, is the basis for RAG-mediated recombination (44, 45). At the TCRα/δ locus, by insertion of a transcriptional terminator, a requirement of ongoing transcription for RAG-mediated recombination in vivo has been demonstrated (46). At the TCRβ locus, our observations of loss of Eβ-dependent transcription, reduced H3K4-trimethyl acquisition, and reduced Rag2 association at TCRβ upon insertion of a transcriptional insulator provide substantial support for the link between transcription and RAG-mediated recombination in vivo.

Organization of the H19-ICR insulator at the TCRβ locus afforded the analysis of its ability to restrict the Eβ enhancer activity. Even though the enhancers at the Igf2/H19 locus and the TCRβ locus appear to regulate the chromatin structure and activate the promoters differently, H19-ICR acts as an effective enhancer blocker in both cases. CTCF-bound insulators are known to influence the chromatin loop domain organization. In this context, our results suggest that within the framework of the altered chromatin loopscape that segregates the enhancer and promoter into separate loops, insulators interact with the enhancer as well as prevent the influence of the enhancer on the insulated domain. Consequently, they may simultaneously appear to be a promoter decoy as well as elements that prevent tracking of enhancers and/or of other epigenetic changes in case the enhancers employ these mechanisms to activate the locus. The varied mechanisms of insulators, as perceived, are an outcome of the specific type of enhancer activity that they might be preventing in a given developmental context.