One surprising finding of the human genome sequencing project was that only 1.9% of the genome consists of protein-coding genes (1). Although they were previously thought to be "junk" DNA, several lines of evidence have shown that many "noncoding genomic dark-matter" regions are functional. First, whole-genome chromatin immunoprecipitation sequencing (ChIP-Seq) studies on transcription factors have revealed many transcription factor binding sites (TFBSs) that were not near gene promoters (2). Moreover, genome-wide association studies have indicated that noncoding regions can constitute significant risk factors for diseases, but the pathological mechanisms have remained unclear (3). These results raise the question of which noncoding genomic regions are functional; and if so, what is the mechanism by which these distal enhancers regulate transcription?
Studies at specific loci using technologies such as chromosome conformation capture (3C) fluorescence in situ hybridization (FISH) and variants (4) demonstrated that two or more chromatin regions that are separated by long distances within the chromosome, or present on different chromosomes, may be brought through specific protein tethers into close three-dimensional proximity, in a "chromatin interaction." These studies showed that certain distal TFBSs loop via chromatin interactions to other TFBSs that are proximal to gene promoters, indicating that these distal TFBSs are involved in gene regulation. However, because these previous methods were limited to partial genome detection at best, we needed new methods to understand the overall contributions that chromatin interactions and distal noncoding elements make toward regulating gene transcription in the entire genome (4).
Therefore, during my Ph.D. work, together with colleagues at the Genome Institute of Singapore, I developed a method called chromatin interaction analysis by paired-end tags (ChIA-PET), a genome-wide high-throughput technique for de novo detection of chromatin interactions (4). ChIA-PET involves the ligation of multiple chromatin fragments held in close proximity to each other, followed by the isolation of paired tags from different chromatin fragments, which are then sequenced with paired-end next-generation sequencing methods and mapped to a reference genome (Fig. 1A) (5). Chromatin interactions are read out as overlapping and enriched mapped paired distal sequences as opposed to noise, which remains as random sequences scattered throughout the genome (6) (Fig. 1A). Because the number of possible chromatin interactions is very high, we expected the complexity of the ChIA-PET libraries to be prohibitive for current sequencing efforts and hence used ChIP against specific transcription factors to both provide specificity to the identified interactions and reduce the complexity of the library (7). Initial efforts validated the feasibility of the method and showed that ChIA-PET could reveal not just global maps of TFBSs but also chromatin interactions between the TFBSs (8), representing a significant advance in methodology after ChIP-Seq for studying transcription factor mechanisms.
Subsequently, I used ChIA-PET to comprehensively characterize estrogen receptor-α (ER-α)–bound chromatin interactions in estrogen-treated human breast adenocarcinoma cells (MCF-7 cells) and generated the first ER-α chromatin interactome map. Complex intrachromosomal networks of chromatin interactions between ER-α TFBSs were common. Moreover, genes near ER-α TFBSs with chromatin interactions were significantly more associated with estrogen up-regulation than genes near ER-α TFBSs without chromatin interactions, indicating that in the context of ER-α, chromatin interactions are good predictors of gene activation (8).
Taken together, these results show that many noncoding dark-matter genomic elements function in gene regulation through extensive chromatin looping events. I postulate that chromatin interactions could constitute an efficient primary mechanism for ER-α transcription factor function: Dense cagelike networks of chromatin interactions and TFBSs could trap high local concentrations of oscillating ER-α and other transcriptional components (Fig. 1B), allowing cells greater control while reducing the number of transcription-associated proteins required to produce such responses (8). Such a mechanism also yields robustness through TFBS redundancy, so that point mutations at any one site would not totally abolish regulatory control, and TFBSs separated from their target gene through genomic structural variants could continue to regulate genes through chromatin interactions (8).
Furthermore, the presence of a dense network of high-order chromatin structures suggests that mutations in cells do not occur randomly, as previously thought. Chromatin structures could affect the accessibility of the genome to mutagens, and by bringing together different chromatin strands, chromatin interactions may facilitate genome rearrangements such as the TMPRSS2-ERG translocation found in many prostate cancers (9, 10). These results suggest that clinical biomarkers based on chromatin interactions, such as FISH probes, might indicate early changes that lead to disease, allowing for early therapeutic intervention.
To better understand the biology of chromatin interactions, many questions need to be addressed: Which transcription factors similarly employ chromatin looping? What are the dynamics of chromatin looping? How do chromatin loops change as cells undergo changes such as stem cell differentiation, or drug treatments? Do chromatin loops vary between different cells and species? Furthermore, what is the general pattern of all chromatin interactions in a cell? Hi-C is a complementary technique to ChIA-PET, which was recently developed by Job Dekker and colleagues (11). Hi-C uses a similar strategy as ChIA-PET but without ChIP enrichment in order to present a low-resolution map of general genome architecture (11). Moving forward, a technique that combines the high-resolution capabilities of ChIA-PET analyses, as well as the abilities of Hi-C, to capture all chromatin interactions in a high-resolution manner using the latest sequencing technologies such as Pacific Biosciences (12) would be the next breakthrough in chromatin structure analyses.
My Ph.D. work on ChIA-PET demonstrated that many TFBSs found in noncoding genomic regions could be functionally involved in gene transcription through the widespread mechanism of chromatin interactions. The development of ChIA-PET and related methodologies for genome-wide chromatin interactome analysis will facilitate new insights into chromatin interaction biology. Because chromatin interactions play fundamental roles in basic cellular processes and could be of clinical significance as possible biomarkers of diseases such as cancer, further research into chromatin interactome sequencing is warranted.
- International Human Genome Sequencing Consortium, Nature 431, 931 (2004).
- C. E. Massie, I. G. Mills, EMBO Rep. 9, 337 (2008).
- A. Visel, E. M. Rubin, L. A. Pennacchio, Nature 461, 199 (2009).
- M. J. Fullwood, C. L. Wei, E. T. Liu, Y. Ruan, Genome Res. 19, 521 (2009).
- M. J. Fullwood, Y. Han, C. L. Wei, X. Ruan, Y. Ruan, Curr. Protoc. Mol. Biol., Chapter 21, Unit 21, 15, 1 (2010).
- G. Li et al., Genome Biol. 11, R22 (2010).
- M. J. Fullwood, Y. Ruan, J. Cell Biochem. 107, 30 (2009).
- M. J. Fullwood et al., Nature 462, 58 (2009).
- R. S. Mani et al., Science 326, 1230 (2009).
- T. Misteli, Cell 128, 787 (2007).
- E. Lieberman-Aiden et al., Science 326, 289 (2009).
- J. Eid et al., Science 323, 133 (2009).