CTCF-Mediated Human 3D Genome Architecture Reveals Chromatin Topology for Transcription

I. ChIA-PET is multifaceted for chromatin interaction mapping

In addition to the original ChIA-PET data deposited in the ENCODE project (ENCODE Project Consortium, 2012), we have generated new CTCF- and RNAPII-mediated chromatin interaction datasets using an improved ChIA-PET protocol (Figure S1A) for longer reads (2x150bp). Altogether, we collected 364 million uniquely mapped ChIA-PET reads in 12 ChIA-PET libraries from four human cell lines: GM12878, HeLa, K562 and MCF7 for analysis (Table S1).

A ChIA-PET experiment delivers paired-end-tag (PET) sequencing data from self-ligation and inter-ligation products (Figure 1A, S1B). The self-ligation PET data identify ChIP-enriched protein-binding sites. The clustered inter-ligation PET data detect enriched interactions mediated by the ChIP targeted protein factor, whereas the singleton inter-ligation data reflects higher-order topological proximity, similar to Hi-C data (Figure S1B-E). Therefore, in theory, the multifaceted ChIA-PET data is ideal for comprehensive 3D genome mapping.

Recently, a study using in situ Hi-C generated 4.9 billion contact reads in GM12878 cells and found the majority of chromatin interaction loops to be associated with CTCF-binding sites (Rao et al., 2014). We compared this dataset with our CTCF ChIA-PET and demonstrate that the two datasets displayed very similar contact patterns (Figure 1B). In addition, ChIA-PET identified specific CTCF loops with binding-site resolution at 100s bp level. Importantly, only a single HiSeq 2500 rapid mode run or even a MiSeq run was sufficient for ChIA-PET to reach the same output as in situ Hi-C, plus 10-fold higher resolution. Thus, ChIA-PET is cost-effective, inclusive and reproducible (Data S1, I; Extended Experimental Procedures) of generating multi-layer mapping data, capturing protein binding sites and enriched functional chromatin interactions, as well as non-enriched chromatin contacts for higher-order chromosomal conformations.

II. CTCF organizes chromatin contact domain into CTCF foci

CTCF-mediated chromatin interactions are pervasive across the entire genome in the four tested human cell lines (Figure S2A). Based on the CTCF interaction clustering scheme, we identified 53,741 high quality CTCF-mediated interactions in GM12878 cells (Figure S2B-E; Extended Experimental Procedures). Considering that cohesin protein complex is highly associated with CTCF in chromatin biology (Ong and Corces, 2014), we examined the CTCF-anchor sites for co-occupancy by cohesin using ChIP-Seq data of subunits RAD21 and SMC3. The vast majority (99%) of the CTCF interactions had cohesion co-occupancy in either one (n=10,952) or both anchors (n=42,297). Moreover, interactions with cohesin support on both anchors had significantly higher contact frequency than those with cohesin only on one anchor (Figure 2A, left). Since the sites with co-occupancy of CTCF and cohesin represent the most biologically relevant regions in our study, we focused further analyses on this subset.

We investigated the CTCF-motifs in anchors of CTCF loops identified in this study. Of the 42,297 interactions, ~83% (n=35,230) had CTCF motifs in both anchors with unique orientation. Among these, 33.1% were in tandem (i.e. tandem loop) and 64.5% (n=22,709) in convergent orientation (i.e. convergent loop) (Figure 2A right; Figure S2F; Table S2; Extended Experimental Procedures). Thus, CTCF-mediated chromatin loops have a clear orientation preference (convergent) that represents strong interactions with high contact frequency (Figure 2A, right), in agreement with in situ Hi-C data (Rao et al., 2014). In addition, tandem loops were present in significant numbers with intermediate interacting strength and span (Figure 2A, S2G), and most of them (82%) were positioned within convergent loops, perhaps representing a subgroup with possible supplemental function. A possible reason for the tandem loops discovered in our study but not in Hi-C is likely due to the power of specific enrichment in CTCF ChIA-PET experiments (Table S3; Extended Experimental Procedures).

To further understand how cohesin cooperates with CTCF in chromatin biology, we analyzed ChIP-Seq data of RAD21 and SMC3 and suggest that cohesin may surround the CTCF occupancy but have a preference toward the 3’ side of CTCF motif (Data S1, II). To more precisely define CTCF/cohesin co-occupancy, we performed ChIP-nexus (He et al., 2015) to identify the specific borders of DNA footprints bound by cohesin in relation to CTCF. The RAD21 ChIP-nexus data detected one 5’ border that is very close to the 5’ border of CTCF, and two 3’ border sites with the weak one matching exactly to the CTCF 3’ border and the stronger one being 40 bp downstream (Figure 2B; Data S1, II). Similar patterns were observed for SMC3 (Data S1, II). Collectively, these results suggest part of the cohesin ring complex directly overlap with the CTCF binding around the motif and embrace additional space downstream in the 3’ direction.

Many CTCF/cohesion-mediated loops often interconnect and continuously cover large chromatin segments (Figure 2C). Based on connectivity and contact frequency (Figure S3A; Extended Experimental Procedures), we identified 2,267 CTCF-mediated chromatin contact domains (CCDs) in GM12878 cells (Figure 2D, S3B). Comparison to Hi-C data showed that CTCF ChIA-PET and non-selective Hi-C were highly concordant in detecting chromatin domain structures (Figure 2C-D, S3C-D), indicating that CTCF-associated chromatin interactions are abundant in human 3D nucleome.

At the 4,534 boundaries of the 2,267 CCDs, the majority contained inward-facing CTCF-binding motifs (Figure 2E, S3E). Many CCD boundaries contained multiple CTCF-binding peaks with inward-facing motifs (Figure 2C inserts; Figure S3F), suggesting a possible double-knot-tie mechanics for tightening domain end structures (Figure S3G-I; Extended Experimental Procedure). In contrast, tandem loops were evenly distributed within the CCD space (Figure S3I), and showed high consistency in motif orientation (Figure 2F, S3J).

Interconnected CTCF binding and looping with convergent and tandem motif orientations constitute the finer details in chromatin topology. Figure 2G illustrates a CCD composed of multiple CTCF loops. Theoretically, CTCF dimerization with motifs could be either symmetric or asymmetric. Since the cohesin ring complex is most likely bound toward the 3’ downstream of CTCF motif (Figure 2B), we speculate that CTCF dimerization with two interacting motifs favors symmetric conformation. Therefore, a pair of convergent motifs would form a “hairpin loop”, while a pair of tandem motifs form a “coiled loop” (Figure 2H). This principle may have critical topological implications for the 3D structure of CCD and the overall chromosomal topology and genome organization. The example CCD in Figure 2G shows all 8 CTCF anchors (a-h) to be interconnected, thus, suggesting anchors in this CCD form an aggregated scaffold of “CTCF/cohesin focus” (Figure 2I). Consequently, the overall properties of genome organization would be collectively shaped by the 2,267 CTCF/cohesin-foci (2,267 CCDs) in GM12878 cells.

III. RNAPII transcription factories are spatially associated with CTCF/cohesin foci

To functionally characterize CTCF/cohesin-mediated chromatin topology, we overlaid functional genomics data (RNAPII ChIA-PET, chromatin state and RNA-Seq) with CCD footprint on the genome landscape (Figure 3A). The results indicated that most transcription activity occurs within CTCF-looped chromatin structures (Figure S4A). Most of the RNAPII-associated loops are smaller than CTCF loops (Figure S4B), and the vast majority of RNAPII-looping structures are included within CCD-defined genomic space (Figure S4B-C; Table S2).

To dissect the associations with transcription, we divided CTCF/cohesin-bound chromatin loop structures into “anchor” and “loop” regions, and then examined their epigenomic and transcriptional features (Extended Experimental Procedures). Unlike the loop regions, the anchors were enriched with active epigenomic markers, RNAPII occupancy and the presence of TSS (transcription start site) (Figure 3B), suggesting that CTCF-anchor regions are the foci for transcriptional activity. Unexpectedly, further detailed analyses focusing on anchors uncovered structure-function features related to transcription activity and directionality at three distinct levels. First, signals of active epigenomic markers tend to be higher towards 3’ direction of CTCF-motif for both convergent and tandem loops (Figure 3C). More strikingly, TSS at anchor regions showed clear strand-specificity and directional enrichment along with CTCF motif orientation, indicating that a sub-group of genes are pre-positioned within the CTCF-defined anchor regions with their promoters in harmony with CTCF motif orientation, thus dictating the directionality of transcription. Second, active epigenomic markers, RNAPII and TSS densities were highly enriched around the anchors of tandem loops compared to the convergent loops (Figure 3C). The same trend was also observed for B-cell specific transcription factors (TFs), e.g. ELF1 and ZEB1, but not for chromatin structural proteins CTCF, RAD21 and ZNF143 (Figure 3D). Third, the paired anchors involved in tandem loops also exhibited directionality: the “head” anchor tends to have higher signal intensities of active epigenomic markers, RNAPII and B-cell specific TF binding than the “tail” anchor (Figure 3C-D, S4D), while there was no such notable difference for the anchors of convergent loops. It implies that the head anchor in tandem loops may have more promoter property, whereas the tail anchor could possess more enhancer potential. To test this, H3K4me1 (an enhancer marker) and H3K4me3 (a promoter marker) ChIP-Seq data were used to assess the relative strength of promoter and enhancer. The ratio of H3K4me1/H3K4me3 at the tail anchor of tandem loops was significantly higher than the head anchor (Figure 3E right), indicating that the tail anchor is more likely involved in enhancer function, whereas the head anchor is more related to promoter. In contrast, no difference was observed for the paired anchors of convergent loops (Figure 3E left). Collectively, these data suggest that tandem loops possess distinctive features important for organizing transcription activity.

In GM12878 cells, thousands of genes and enhancers were found proximal to CTCF/cohesin anchors (i.e. anchor-genes/enhancers), and the rest were scattered within the CTCF/cohesin loop regions (i.e. loop-genes/enhancers) (Extended Experimental Procedures). We examined the anchor- and loop-genes based on their expression profiles in 56 different human tissues. Gene expression breadth analysis (Extended Experimental Procedures) indicated that anchor-genes were significantly less tissue-specific than loop-genes (Figure 3F left). Further analysis revealed that active anchor-genes were almost exclusively housekeeping (Figure 3F right; Figure S4E-F), emphasizing the notion that CTCF interaction anchors selectively enrich for constitutively expressed genes.

To investigate how active anchor-genes relate to active loop genes and enhancers, we examined their connectivity by RNAPII ChIA-PET, as described previously (Li et al., 2012). Using the newly generated RNAPII ChIA-PET data (Table S1), most active loop-genes were found connected to anchor-genes and/or anchor-enhancers (Table S2; Extended Experimental Procedures) through RNAPII-mediated interactions. Therefore, most active genes are connected, either directly or indirectly, to the anchors of CTCF loops (Figure 3G), suggesting that anchor-genes and anchor-enhancers could serve as nucleation points to aggregate related loop-genes towards corresponding CTCF/cohesin anchors for coordinated transcription. Figure 3H illustrates an example CCD with seven interconnecting CTCF anchors and a number of CTCF anchor-genes/enhancers and loop-genes/enhancers. RNAPII interactions indicated that these genes and enhancers were interconnected, which could be viewed (spatially) as a transcription factory docked to the chromatin structural base of CTCF focus. More examples are shown in Figure S4G-H.

Together, CTCF and RNAPII ChIA-PET analyses, along with chromatin state and RNA readout data, revealed that the basic topological units of chromatin looping structures and transcriptional function are highly cooperative for housekeeping functions and cell-type specificity.

IV. Haplotype variants alter CTCF-mediated chromatin structure and function

Allelic differences between two homologous chromosomes can influence inheritance characteristics in the human genome (McDaniell et al., 2010). The means by which allele-specific genetic variation affects chromatin organization has become an intriguing question (Leung et al., 2015). Hi-C analyses have demonstrated that nuclear proximity ligation is an efficient genome-wide approach for haplotype mapping of chromatin interactions (Selvaraj et al., 2013). Here, we sought to use ChIA-PET to investigate haplotype-specific chromatin interactions and subsequent structural and functional consequences.

We identified 65,718 phased PET reads mapped intra-chromosomally in the phased GM12878 genome, of which the vast majority (n=64,805, 98.6%) were cis-interacting PETs (Figure 4A, S5A-B; Extended Experimental Procedures). To investigate haplotype-specific chromatin interactions mediated by CTCF and RNAPII, we focused on phased PETs that were clustered to represent interactions enriched by these factors. Using this criterion, we identified 350 haplotype-biased anchors and 1,728 interactions mediated by CTCF, and 299 haplotype-biased anchors and 1,322 interactions mediated by RNAPII (Figure 4B-C; Extended Experimental Procedures).

Among the identified haplotype-biased chromatin interactions mediated by CTCF was the well-studied H19-IGF2 locus (Nativio et al., 2011) involved in genomic imprinting of autosomes (Figure S5C left, and additional example in Figure S5C right) thus demonstrating the robustness of our approach. We then explored whether allelic variations alter chromatin 3D structure between homologous chromosomes. Indeed, we found paternally biased super-long interactions (13Mb) mediated by CTCF on ChrX connecting the DXZ4 and FIRRE loci (Figure 4D), which has been reported previously (Horakova et al., 2012; Rao et al., 2014). In addition, we identified another super long-range interaction between FIRRE and G6PD (23Mb) exclusively on the same haplotype as indicated by both contact heatmaps and CTCF-enriched interactions. This interaction was further validated by DNA-FISH (Figure 4E).

Next, we investigated whether SNPs could directly alter chromatin topology and function at a finer scale (e.g. domain structure and individual loop). Others have demonstrated deletion and inversion of DNA fragments containing CTCF/cohesin binding sites could disrupt the nearby TAD structure and alter associated gene transcription (Dowen et al., 2014; Guo et al., 2015). Despite the success, CRISPR/Cas9-engineered regions in these studies involved from 100s to 1,000s bp, which could contain other sequence elements of unknown function. To avoid potential “collateral damage”, we exploited the SNP “genotype” as single nucleotide “perturbation” and evaluated the corresponding CTCF binding/looping property as the “phenotype” in different human individuals (cell lines) with either heterozygous or homozygous allele composition (Figure 5A). We first focused on the 39 allele-biased CTCF interaction anchors (i.e., 39 loci with phased SNPs as “naturally occurred” single nucleotide “perturbation”) located at CCD boundary regions. As shown in Figure 5B, a heterozygous SNP (maternal “T” and paternal “A”) was located on the 3’ boundary of a CCD. The maternal “T” exhibited strong CTCF binding (i.e. functional allele) while the paternal “A” allele showed weak binding (i.e. dysfunctional allele). We hypothesize that the loss of CTCF binding at the “A” allele in this locus would cause loss of CTCF-mediated looping and, in turn, alter CCD structure. To test this, we examined other individuals (cell lines) with homozygous “A/A” genotype at this locus. Indeed, in HeLa and MCF7 cells, no CTCF-mediated interactions originated from this locus and the corresponding CCD structures in these two cell lines were drastically different from GM12878 (See another case in Figure S6A). Together, these analyses validated that functional CTCF sites are necessary to maintain the proper CCD boundaries, and suggest that single nucleotide variations in CTCF binding sites could alter chromatin topology.

We then explored if “naturally occurred SNP perturbation” could alter CTCF tandem loops and consequently impact transcription inside CCDs. We identified 50 loci where allele-specific tandem loops were associated with genes harboring phased SNPs in the gene body, thus, testable for allelic expression bias. From them, 22 (44%) displayed significant allele-specific expression (P < 0.05; See examples in Figure S6B). In a particular example, the promoters of DENND2D and CHI3L2 reside at the opposite anchors of a CTCF tandem loop that is paternal-specific (Figure 5C). Consistently, the RNAPII occupancy and associated chromatin loops in this region are also paternal-biased, indicating this CTCF tandem loop is functionally involved in a paternal-specific transcriptional regulation. However, only CHI3L2 exhibited paternal-biased gene expression (see also Figure S6C) but not DENND2D. This example supports that, in an allele-specific manner, the enhancer at the tail anchor of a CTCF tandem loop could interact with gene promoters proximal to the head anchor of the loop in concordance with the motif orientation for transcription regulation (Figure 3C-E). Additionally, the phased SNP with allele-biased CTCF-binding coverage was located inside the CTCF-binding motif, and this SNP was the only nucleotide difference between the two homologs in kilobase-wide span. This observation impelled us to systematically examine how SNPs in the CTCF-motif would subsequently influence CTCF binding and looping.

Of the SNPs mapped within the 350 allelic-specific anchors bound by CTCF (Figure 4C), 70 reside in the core motif (Figure S6D). The alignment of each of the heterozygous SNPs within the CTCF-motif showed that alleles having strong CTCF-binding possess canonical motif consensus, whereas alleles with weak or no binding had deviated motif sequences, especially with position 14 (“G”) being the most affected (Figure 5D, S6E), suggesting that this “G” is critical for CTCF-binding affinity. For example, at position 14 of a CTCF motif in Chr17q21, the maternal and paternal alleles are G|C, respectively (Figure 5E). Despite this SNP being the only variable site in this locus and the surrounding kilobase region, the CTCF binding and interaction in maternal allele is 68-fold stronger than paternal, indicating that nucleotide “C” in the motif is disruptive for CTCF to bind to the paternal allele in this locus. Since SNP variation in CTCF motif could lead to such drastic alteration in CTCF binding and looping, we sought to determine whether changes in chromatin topology would link to human diseases.

We systematically assessed disease association of the disrupted CTCF-mediated interactions by examining the linkage disequilibrium (LD) between the 70 SNPs residing in CTCF motif (i.e. CTCF-SNPs) and GWAS identified disease associated SNPs (Extended Experimental Procedures). We hypothesized that disease-associated SNPs and dysfunctional CTCF-SNPs would be linked if they reside in strong LD blocks. In this setting, 32 of the 70 CTCF-SNPs were documented in dbSNP database, and 8 showed LD with disease associated SNPs in the tested populations (Data S1, IV; Table S4). Since GM12878 originated from an individual of European ancestry, we further focused on the CTCF-SNPs with disease association by LD in CEU population. One of the CTCF-SNPs found associated with asthma is SNP rs12936231 (Figure 5E). Of the two alleles (G|C), the dysfunctional “C” has been documented as a high-risk allele for asthma and autoimmune diseases, and suggested to alter chromatin remodeling and domain-wide transcription of certain genes (e.g. ZPBP2, GSDMB and ORMDL3; Verlaan et al., 2009). Separate eQTL data also suggested that this allele alters the expression level of GSDMB and ORMDL3 (Montgomery et al., 2010; Stranger et al., 2007). Interestingly, six additional asthma-associated SNPs were found in the same LD block (209kb) with this CTCF-SNP (rs12936231) in CEU having high pairwise correlation (D’>0.7), and the haplotype comprising these 7 risk alleles was frequently (0.422) observed in the CEU population (Figure 5E). Collectively, these results suggest that the disruption of CTCF motif by this “C” allele, which abrogates CTCF binding, looping, and chromatin topology, may be the primary molecular event leading to disease susceptibility, while the other 6 asthma-SNPs were likely non-causative “bystanders”. Moreover, these findings highlight the potential of allelic chromatin topology analyses to infer mechanisms by which SNPs are associated with disease and traits.

V. RNAPII-mediated chromatin interactions regulate allele-specific transcription

To investigate whether allelic variations affect RNAPII-mediated chromatin interaction and/or result in functional consequences, we identified 1,322 haplotype-specific RNAPII interactions that involved 299 haplotype-specific RNAPII interaction anchors in GM12878 (Figure 4C; Extended Experimental Procedures). Our haplotype analyses showed significant ChrX maternal-specificity of RNAPII interactions and gene expression (Figure 4C), consistent with the fact that paternal-X in GM12878 is imprinted (Rozowsky et al., 2011). Thus, at chromosomal scale, the connection between allele-specific RNAPII interaction and gene transcription is broadly established. To further validate that haplotype-specific chromatin interaction could lead to allele-specific transcription regulation (Figure 6A), we examined 40 TFs for their allelic binding specificity and found that the vast majority (95%) of the TF binding allele-biases were consistent with the RNAPII allele-specificity (Figure 6B; Extended Experimental Procedures). Furthermore, we identified 89 genes with allele-biased expression involved in allele-specific RNAPII interactions, and the majority (n=79) displayed the same allele-specificity in transcription as the allele-biased RNAPII interactions (Figure 6C-D; Table S5). For example, RNAPII binding and interactions over the SNPs at the XIST (X inactive specific transcript) locus showed highly consistent allele bias with haplotype-specific expression of XIST (Figure S5D). Such observations at the XIST locus demonstrate the accuracy of our haplotype chromatin interaction analysis.

Our data also reveals the haplotype effect of long-range enhancer-promoter interactions. For example, the promoters of LOC374443, CLEC2D and CLECL1 were in contact with an RNAPII-associated multi-gene complex (Figure 6E). It is observed RNAPII occupancy in the paternal allele at the upstream enhancer and promoter sites (300kb apart) were approximately 3-fold and 2.5-fold higher, respectively, than the matched maternal allele. In addition, the distal enhancer exhibited more than 3-fold higher paternal-biased binding by 3 B-cell specific TFs: BCL3, EBF1 and TCF12. We also observed 3.5-fold higher paternally biased expression in these three genes (Figure 6E). In contrast, nearby genes not directly involved in the RNAPII mediated interactions showed balanced expression between homologs. This example supports the notion that allele-specificity at distant enhancers is also effective in regulating allele-specific RNAPII activity at the target genes with high specificity.

VI. 3D genome models elucidate the human genome structure and function

Using an integrated 3D NucleOme Modeling Engine (3D-NOME) that builds a hierarchical tree structure to represent the 3D genome with increasing resolution (Figure S7A-C) (Szalaj et al., In preparation), we simulated the 3D genome models from the combined CTCF and RNAPII ChIA-PET data derived from GM12878. In these models, several known chromosomal features were captured, e.g. at the whole nucleome level, large chromosomes were preferentially projected in the periphery of nucleus, and small chromosomes were positioned in the inner nuclear space (Figure S7D-E; Extended Experimental Procedures). To gain further specificity at the level of individual chromosomes, we modeled Chr1 at different resolutions, and observed a putative conformation, whereby its two chromosomal arms bend and extend in the same direction (Figure 7A). Since our mapping data were derived from millions of cells, the predicted 3D model would reflect an average representation. It is possible that the body of Chr1 is fluidly changing between “open” and “closed” conformations (Figure 7B). Our 3D DNA-FISH results supported such speculation (Figure 7C; Extended Experimental Procedures).

From 3D genome simulation and DNA-FISH nuclear imaging, as well as the growing knowledge of chromatin topology (Boyle et al., 2011; Kalhor et al., 2012), a general feature of chromosome topology starts to emerge. Although much less condensed, chromosomes in interphase still maintain their core axes, comprised of mostly condensed heterochromatin segments (probably other matrix-filling material as well) (Figure 7D). The loosely organized “open” chromatin segments could extend laterally outward (as chromatin loops) around the chromosome axis, similar to the morphology of lampbrush chromosome (Morgan, 2002). We envision CTCF and cohesin (possibly with other factors e.g. Topoisomerase II, Liang et al., 2015) localizing on the surface of the core chromosome axis and defining the interface between the inner condensed (inside the chromosome axis core) and the outer loose domains. The condensed chromosome axis core could help to maintain the desired shape and physical properties for the overall chromosome territories, which are also impermeable as repressive or inactive compartments. On the contrary, the loose domains are open and permissive to molecules mediating nuclear functions.

A key component in our model is the spatial overlap of RNAPII associated transcription factories with the CTCF/cohesin contributed structural foci. To test this, we conducted super-resolution structured illumination microscopic (SIM) analysis of immunostaining using antibodies against CTCF and RNAPII in GM12878 cells (Figure 7E; Extended Experimental Procedures). To further increase the detection resolution, we performed Förster resonance energy transfer (FRET) assay using fluorescence lifetime imaging microscopy (FLIM) to detect the co-localization of CTCF and RNAPII (Figure 7F; Extended Experimental Procedures). Together, the SIM and FRET-FLIM analyses validated that the CTCF- and RNAPII-foci are indeed in close distance in the human nucleome.

Lastly, to test if CTCF prevalently locates along the chromosomal axes, we exploited lampbrush chromosomes. Although not ideal, lampbrush chromosome is still considered the classic model for chromatin looping morphology. We examined the lampbrush chromosomes isolated from newt nuclei using immunostaining to CTCF, and showed the CTCF signals predominantly along chromosome axes instead of the laterally extended chromatin loops (Figure 7G; Extended Experimental Procedure).

Long read ChIA-PET

Instead of using MmeI digestion as in the original ChIA-PET protocol (Fullwood et al., 2009), long read ChIA-PET uses Tn5 tagmentation to generate random size of PET templates for long tag sequencing reads (2X150bp) by HiSeq2500 (Extended Experimental Procedures).

ChIA-PET data processing

Short-read ChIA-PET data was obtained from ENCODE (Table S6) and processed using the original ChIA-PET Tool (Li et al., 2010). Long-read ChIA-PET data was processed by a customized ChIA-PET data processing pipeline. All data has been submitted to the NCBI GEO database (GEO accession: GSE72816).

ChIP-nexus

ChIP-nexus on RAD21 and SMC3 in GM12878 cells was performed as described in He et al., 2015.

3D DNA-FISH

DNA-FISH of GM12878 cells was performed according to Cremer et al., 2008 with minor modifications. The 3D images of all-chromosome painting were acquired with the Zeiss LSM 780 confocal microscope. The 3D Chr1 territory was obtained using Imaris (Bitplane) and Amira (FEI).

Immunostaining and SIM super-resolution microscopy

The CTCF and RNAPII immunofluorescence staining in GM12878 was performed according to routine procedures (Hall et al., 2015). Specimens were analysed using a Zeiss ELYRA PS.1 super-resolution system. The super-resolved images were generated using Zeiss Zen 2012 black edition software.

Co-localization detection of CTCF and RNAPII by FRET-FLIM

The FRET-FLIM analysis of the same specimen was performed following the same immunocytochemistry protocol described above. The measurement of fluorescence lifetime of the donor was performed on a Picoquant PicoHarp 300 Time-Correlated Single Photon Counting (TCSPC) system attached to Leica Sp8 confocal microscope, using 63x oil immersion objective (NA 1.4).

Lampbrush chromosome

Ovarian biopsies were performed on adult female newts. Germinal vesicles (nuclei) from stage V-VI oocytes were manually isolated. Lampbrush chromosomes were prepared as previously described (Penrad-Mobayed et al., 2010), and then immunostained with CTCF antibody subject for standard light transmitted and florescence microscopy. The fluorescence signals were measured on the chromosome axes and lateral loops.

01

6

7

8

02

03

04

1

2

3

4

5

1. Bickmore WA. The spatial organization of the human genome. Annu Rev Genomics Hum Genet. 2013;14:67–84. doi: 10.1146/annurev-genom-091212-153515. [DOI] [PubMed] [Google Scholar]
2. Boyle S, Rodesch MJ, Halvensleben HA, Jeddeloh JA, Bickmore WA. Fluorescence in situ hybridization with high-complexity repeat-free oligonucleotide probes generated by massively parallel synthesis. Chromosome Res. 2011;19:901–909. doi: 10.1007/s10577-011-9245-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
3. Cremer M, Grasser F, Lanctot C, Muller S, Neusser M, Zinner R, Solovei I, Cremer T. Multicolor 3D fluorescence in situ hybridization for imaging interphase chromosomes. Methods Mol Biol. 2008;463:205–239. doi: 10.1007/978-1-59745-406-3_15. [DOI] [PubMed] [Google Scholar]
4. Cullen KE, Kladde MP, Seyfred MA. Interaction between transcription regulatory regions of prolactin chromatin. Science. 1993;261:203–206. doi: 10.1126/science.8327891. [DOI] [PubMed] [Google Scholar]
5. Dixon JR, Selvaraj S, Yue F, Kim A, Li Y, Shen Y, Hu M, Liu JS, Ren B. Topological domains in mammalian genomes identified by analysis of chromatin interactions. Nature. 2012;485:376–380. doi: 10.1038/nature11082. [DOI] [PMC free article] [PubMed] [Google Scholar]
6. Dowen JM, Fan ZP, Hnisz D, Ren G, Abraham BJ, Zhang LN, Weintraub AS, Schuijers J, Lee TI, Zhao K, et al. Control of cell identity genes occurs in insulated neighborhoods in mammalian chromosomes. Cell. 2014;159:374–387. doi: 10.1016/j.cell.2014.09.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
7. ENCODE Project Consortium An integrated encyclopedia of DNA elements in the human genome. Nature. 2012;489:57–74. doi: 10.1038/nature11247. [DOI] [PMC free article] [PubMed] [Google Scholar]
8. Fullwood MJ, Liu MH, Pan YF, Liu J, Xu H, Mohamed YB, Orlov YL, Velkov S, Ho A, Mei PH, et al. An oestrogen-receptor-alpha-bound human chromatin interactome. Nature. 2009;462:58–64. doi: 10.1038/nature08497. [DOI] [PMC free article] [PubMed] [Google Scholar]
9. Guo Y, Xu Q, Canzio D, Shou J, Li J, Gorkin DU, Jung I, Wu H, Zhai Y, Tang Y, et al. CRISPR Inversion of CTCF Sites Alters Genome Topology and Enhancer/Promoter Function. Cell. 2015;162:900–910. doi: 10.1016/j.cell.2015.07.038. [DOI] [PMC free article] [PubMed] [Google Scholar]
10. Hall MH, Magalska A, Malinowska M, Ruszczycki B, Czaban I, Patel S, Ambrozek-Latecka M, Zolocinska E, Broszkiewicz H, Parobczak K, et al. Localization and regulation of PML bodies in the adult mouse brain. Brain Struct Funct. 2015 doi: 10.1007/s00429-015-1053-4. DOI: 10.1007/s00429-015-1053-4. [DOI] [PubMed] [Google Scholar]
11. He Q, Johnston J, Zeitlinger J. ChIP-nexus enables improved detection of in vivo transcription factor binding footprints. Nat Biotechnol. 2015;33:395–401. doi: 10.1038/nbt.3121. [DOI] [PMC free article] [PubMed] [Google Scholar]
12. Horakova AH, Moseley SC, McLaughlin CR, Tremblay DC, Chadwick BP. The macrosatellite DXZ4 mediates CTCF-dependent long-range intrachromosomal interactions on the human inactive X chromosome. Hum Mol Genet. 2012;21:4367–4377. doi: 10.1093/hmg/dds270. [DOI] [PMC free article] [PubMed] [Google Scholar]
13. Kalhor R, Tjong H, Jayathilaka N, Alber F, Chen L. Genome architectures revealed by tethered chromosome conformation capture and population-based modeling. Nat Biotechnol. 2012;30:90–98. doi: 10.1038/nbt.2057. [DOI] [PMC free article] [PubMed] [Google Scholar]
14. Leung D, Jung I, Rajagopal N, Schmitt A, Selvaraj S, Lee AY, Yen CA, Lin S, Lin Y, Qiu Y, et al. Integrative analysis of haplotype-resolved epigenomes across human tissues. Nature. 2015;518:350–354. doi: 10.1038/nature14217. [DOI] [PMC free article] [PubMed] [Google Scholar]
15. Li G, Fullwood MJ, Xu H, Mulawadi FH, Velkov S, Vega V, Ariyaratne PN, Mohamed YB, Ooi HS, Tennakoon C, et al. ChIA-PET tool for comprehensive chromatin interaction analysis with paired-end tag sequencing. Genome Biol. 2010;11:R22. doi: 10.1186/gb-2010-11-2-r22. [DOI] [PMC free article] [PubMed] [Google Scholar]
16. Li G, Ruan X, Auerbach RK, Sandhu KS, Zheng M, Wang P, Poh HM, Goh Y, Lim J, Zhang J, et al. Extensive promoter-centered chromatin interactions provide a topological basis for transcription regulation. Cell. 2012;148:84–98. doi: 10.1016/j.cell.2011.12.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
17. Liang Z, Zickler D, Prentiss M, Chang FS, Witz G, Maeshima K, Kleckner N. Chromosomes Progress to Metaphase in Multiple Discrete Steps via Global Compaction/Expansion Cycles. Cell. 2015;161:1124–1137. doi: 10.1016/j.cell.2015.04.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
18. Lieberman-Aiden E, van Berkum NL, Williams L, Imakaev M, Ragoczy T, Telling A, Amit I, Lajoie BR, Sabo PJ, Dorschner MO, et al. Comprehensive mapping of long-range interactions reveals folding principles of the human genome. Science. 2009;326:289–293. doi: 10.1126/science.1181369. [DOI] [PMC free article] [PubMed] [Google Scholar]
19. McDaniell R, Lee BK, Song L, Liu Z, Boyle AP, Erdos MR, Scott LJ, Morken MA, Kucera KS, Battenhouse A, et al. Heritable individual-specific and allele-specific chromatin signatures in humans. Science. 2010;328:235–239. doi: 10.1126/science.1184655. [DOI] [PMC free article] [PubMed] [Google Scholar]
20. Montgomery SB, Sammeth M, Gutierrez-Arcelus M, Lach RP, Ingle C, Nisbett J, Guigo R, Dermitzakis ET. Transcriptome genetics using second generation sequencing in a Caucasian population. Nature. 2010;464:773–777. doi: 10.1038/nature08903. [DOI] [PMC free article] [PubMed] [Google Scholar]
21. Morgan GT. Lampbrush chromosomes and associated bodies: new insights into principles of nuclear structure and function. Chromosome Res. 2002;10:177–200. doi: 10.1023/a:1015227020652. [DOI] [PubMed] [Google Scholar]
22. Nativio R, Sparago A, Ito Y, Weksberg R, Riccio A, Murrell A. Disruption of genomic neighbourhood at the imprinted IGF2-H19 locus in Beckwith-Wiedemann syndrome and Silver-Russell syndrome. Hum Mol Genet. 2011;20:1363–1374. doi: 10.1093/hmg/ddr018. [DOI] [PMC free article] [PubMed] [Google Scholar]
23. Ong CT, Corces VG. CTCF: an architectural protein bridging genome topology and function. Nat Rev Genet. 2014;15:234–246. doi: 10.1038/nrg3663. [DOI] [PMC free article] [PubMed] [Google Scholar]
24. Penrad-Mobayed M, Kanhoush R, Perrin C. Tips and tricks for preparing lampbrush chromosome spreads from Xenopus tropicalis oocytes. Methods. 2010;51:37–44. doi: 10.1016/j.ymeth.2010.01.014. [DOI] [PubMed] [Google Scholar]
25. Rao SS, Huntley MH, Durand NC, Stamenova EK, Bochkov ID, Robinson JT, Sanborn AL, Machol I, Omer AD, Lander ES, et al. A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping. Cell. 2014;159:1665–1680. doi: 10.1016/j.cell.2014.11.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
26. Rhee HS, Pugh BF. Comprehensive genome-wide protein-DNA interactions detected at single-nucleotide resolution. Cell. 2011;147:1408–1419. doi: 10.1016/j.cell.2011.11.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
27. Rozowsky J, Abyzov A, Wang J, Alves P, Raha D, Harmanci A, Leng J, Bjornson R, Kong Y, Kitabayashi N, et al. AlleleSeq: analysis of allele-specific expression and binding in a network framework. Mol Sys Biol. 2011;7:522. doi: 10.1038/msb.2011.54. [DOI] [PMC free article] [PubMed] [Google Scholar]
28. Selvaraj S, J RD, Bansal V, Ren B. Whole-genome haplotype reconstruction using proximity-ligation and shotgun sequencing. Nat Biotechnol. 2013;31:1111–1118. doi: 10.1038/nbt.2728. [DOI] [PMC free article] [PubMed] [Google Scholar]
29. Sims RJ, 3rd, Mandal SS, Reinberg D. Recent highlights of RNA-polymerase-II-mediated transcription. Curr Opin Cell Biol. 2004;16:263–271. doi: 10.1016/j.ceb.2004.04.004. [DOI] [PubMed] [Google Scholar]
30. Stranger BE, Nica AC, Forrest MS, Dimas A, Bird CP, Beazley C, Ingle CE, Dunning M, Flicek P, Koller D, et al. Population genomics of human gene expression. Nat Genet. 2007;39:1217–1224. doi: 10.1038/ng2142. [DOI] [PMC free article] [PubMed] [Google Scholar]
31. Szalaj P, Tang Z, Michalski P, Pietal M, Luo OJ, Ruan Y, Plewczynski D. 3D-NOME: an integrated 3-Dimensional NucleOme Modeling Engine for data-driven simulation of spatial genome organization. doi: 10.1101/gr.205062.116. In preparation. [DOI] [PMC free article] [PubMed] [Google Scholar]
32. Verlaan DJ, Berlivet S, Hunninghake GM, Madore AM, Lariviere M, Moussette S, Grundberg E, Kwan T, Ouimet M, Ge B, et al. Allele-specific chromatin remodeling in the ZPBP2/GSDMB/ORMDL3 locus associated with the risk of asthma and autoimmune disease. Am J Hum Genet. 2009;85:377–393. doi: 10.1016/j.ajhg.2009.08.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
33. Zhang Y, Wong CH, Birnbaum RY, Li G, Favaro R, Ngan CY, Lim J, Tai E, Poh HM, Wong E, et al. Chromatin connectivity maps reveal dynamic promoter-enhancer long-range associations. Nature. 2013;504:306–310. doi: 10.1038/nature12716. [DOI] [PMC free article] [PubMed] [Google Scholar]

Supplementary Materials