From association to mechanism in complex disease genetics: the role of the 3D genome

Functional complexity of the 3D genome

The human genome is organized into complex layers of intricate folds and loops that allow for proper gene expression regulation while fitting roughly three meters of histone-wrapped DNA into an interphase nucleus averaging 6 μm in diameter [11, 14, 15]. Each of the 23 homologous chromosomes are organized into specific regions of the nucleus, called chromosome territories, that restrict interactions between different chromosomes (Fig. 2a) [18, 19]. Each chromosome undergoes additional organization into active “A” and inactive “B” compartments [20–23]. B compartments contain densely packed regions of DNA, called heterochromatin, that are enriched with histone marks of inactivity [20]. A compartments of DNA are typically areas of open chromatin. A/B compartments are further organized to create thousands of megabase-sized sub-regions, called topologically associating domains (TADs), that promote chromatin interactions within the TAD and restrict interactions outside the TAD (Fig. 2a) [18, 20–24]. TAD boundaries are enriched with CCCTC-binding factors (CTCF) and cohesin proteins which facilitate loop formation through a process of loop extrusion (Fig. 2b) [13, 25, 26]. During loop extrusion, cohesin binds to and facilitates the “sliding” of DNA through the cohesin ring structure. The “sliding” on one side of the small loop tends to stop when a CTCF-bound sequence encounters the cohesin. The other side of the loop continues to “slide” and “grow” until another CTCF-bound sequence with convergent orientation reaches the cohesin. The two CTCF proteins homodimerize and create a stabilizing complex with cohesin [25]. The unknotted loop of DNA “extruding” from the newly established CTCF-CTCF–cohesin complex forms the TAD [25]. TADs are largely evolutionarily conserved and maintained during cellular differentiation and embryonic development [18, 27, 28]. In contrast, CTCF-CTCF–cohesion bound regions known as “insulated neighborhoods” organize dynamic enhancer–promoter interactions during cellular differentiation or in response to stimuli (Fig. 2a) [14, 15, 29, 30]. Typically, more than one dynamic insulated neighborhood is nested within a larger evolutionarily conserved TAD.

Characterizing the complex layers of organization that occur within the 3D genome and the regulatory mechanisms that dictate them have provided a framework from which current explorations of gene expression regulation are often based. Approximately 90% of identified enhancer–promoter loops occur within the CTCF-CTCF–cohesin boundaries of TADs and insulated neighborhoods [31]. As reported in several types of cancer, disrupting these boundaries can alter gene expression by relieving restrictions and allowing new loops to form between what was once an insulated enhancer and genes outside of the original loop [32]. What’s more, enhancer–promoter loops function not in isolation, but as regulatory networks where one enhancer has the potential to influence multiple genes and one gene can be influenced by multiple enhancers [16]. Given that a large majority of GWAS variants are located within enhancers that likely modulate distant, as well as neighboring, gene function, establishing detailed maps of enhancer–promoter loops and regulatory networks have the potential to more accurately predict the functional mechanisms influenced by causal GWAS variants and provide translational insights into how such alterations influence disease pathogenesis [5].

Investigating the 3D genome

The majority of techniques used to investigate the 3D genome are derivatives of the original chromatin conformation capture (3C) method, which uses a process called proximity ligation to capture interactions between two sequences of DNA that are in 3D proximity but are separated by linear distance (Fig. 3) [33, 34]. To capture a long-distance interaction, cells are first crosslinked, or fixed, to preserve interactions between the two regions of DNA and associated proteins, and then the entire genome is digested into small pieces using restriction enzymes. Because crosslinking keeps the interacting regions of DNA and associated proteins in close proximity after digestion, the remaining regions of DNA can be enzymatically ligated together to make a chimeric strand of DNA that, once de-crosslinked, can be used in downstream 3D chromatin applications. Building upon this basic methodologic framework, variations have been developed to facilitate both targeted and genome-wide 3D chromatin exploration.

Innovative modifications to traditional 3C, including methodologies such as 4C, 5C, and Capture-C (Table 1), have coupled 3C with microarray or high-throughput next generation sequencing (NGS) technologies to improve throughput with only minor reductions in resolution, but at the expense of quantitative capabilities [40–42]. Improved throughput has allowed for larger-scale targeted studies, like the Promoter Capture-C study by Hughes et al. that comparatively mapped the interactions between 6000 promoters and their regulatory elements in mouse embryonic stem cells and mature erythroid cells [42]. This study not only demonstrated the complexity of enhancer–promoter networks and that the promoter of a specific gene can be regulated by interactions with multiple regulatory elements, but also provided strong evidence that disease-associated risk variants are enriched in gene regulatory elements such as enhancers. Currently, most targeted methodologies still require over 100 million cells to get chromatin quantities necessary to obtain meaningful results, thus restricting use to immortalized cell lines [15]. Given the cell type- and context-specific nature of chromatin dynamics, restricted use of 3C-based technologies to immortalized cell lines has hindered functional characterization of GWAS variants in more relevant primary cell models.

To capture 3D chromatin interactions on a genome-wide scale, proximity ligation was coupled with NGS to create an innovative method called Hi-C (Table 1) [21]. Following proximity ligation, chimeric strands are sequenced and aligned to a reference genome to identify where the two interacting regions were originally located within the linear DNA sequence, thereby identifying the anchor points where chromatin organizing proteins form a DNA loop. Early investigations using Hi-C revealed, for the first-time, chromatin substructures, i.e., TADs, within the context of previously characterized chromosome territories [18]. Subsequent advances in Hi-C methodology have improved resolution from > 100 kb in 2012 to ~ 5–10 kb in 2017, allowing for the generation of 3D genome maps that are now widely used to predict enhancer–promotor interactions occurring within a population of cells at a fixed time [28, 43–45]. Javierre et al. used promoter capture Hi-C (PCHi-C) to map the interacting regions of 31,253 promoters in 17 human primary blood cell types [46]. Not only did this study successfully use primary human cells to perform PCHi-C, but also successfully demonstrated that active enhancers significantly and quantitatively contribute to cell type-specific promoter activity and subsequent gene expression [46].

Single cell Hi-C was first reported in 2013 to explore the cell-to-cell variability of chromatin structures using a single copy X-chromosome model in isolated mouse nuclei [47]. More recently, the use of nucleic acid barcodes to index single cell nuclei eliminated the need to isolate individual nuclei for Hi-C, thus providing a more streamlined approach [48]. As single cell technologies improve along with analytical methods, we anticipate rapid adoption of single cell 3D genome approaches for many experimental designs.

Protein-mediated genome-wide methodologies (Table 1) include additional steps to isolate regions of interacting DNA based on the architectural proteins that influence those interactions, such as loop boundary markers (CTCF, cohesin, etc.), epigenetic markers of enhancers (acetylation of histone H3 on lysine 27 (H3K27ac)), or transcription factors [15, 33, 49–51]. Targeting specific proteins involved in chromatin organization reduces the background signal and the required sequencing depth—number of sequencing reads—necessary to achieve meaningful semi-quantitative results. Furthermore, these improvements have significantly reduced the number of cells required, making it possible to now study DNA looping in primary cells. Recently, Mumbach et al. [52] reported using between 0.5 and one million cells to identify H3K27ac (a histone modification of active enhancers) looping profiles on naïve T cells, T-helper, and Th17 cells isolated from a primary T-cell population using HiChIP technology. The study demonstrated unique and differentially active enhancer loop clusters that corresponded with altered gene expression in each cell type [52], thus supporting current models suggesting different cell types and cell states adopt modified regulatory networks with specific enhancer–promoter loops to drive unique gene expression profiles.

3D genome-wide exploration generates tremendous amounts of sequencing data that require advanced algorithms and pipelines for processing, visualizing, and interpreting the functional significance of these 3D features. Fortunately, several robust software packages are publicly available and more are in development. Each algorithm uses different alignment strategies and filtering criteria to generate heatmaps based on interaction frequencies [43, 53] or looping diagrams that map protein-mediated DNA looping events in the context of linear chromatin [54]. Improvements to capture technologies that select for specific chromatin characteristics, such as histone marks or protein factors, and sequencing technologies that allow for sample barcoding and deeper sequencing continue to improve throughput and reduce background. Simultaneous improvements to the analysis pipelines that define the 3D genome continue to improve the base-pair resolution and quantitative capabilities of 3D technologies, allowing investigations of how disease-associated SNPs alter gene expression through modified 3D genome structures.

Application of 3C technologies to uncover new insights from GWAS in autoimmune disease

Autoimmune diseases, like most complex genetic diseases, result from the collective influence of multiple genetic variants on gene expression and responses to potentially damaging environmental conditions [2]. Investigating the functional significance of GWAS variants in the context of the 3D genome is essential for mechanistic understanding of how these variants, most of which are enriched in largely uncharacterized enhancer regions, influence disease pathology by reducing or enhancing interactions between enhancers and promoters within the enhancer regulatory network (Fig. 4a, b). For example, GWAS and fine-mapping revealed several autoimmune disease risk variants in the chromosome 6q23 locus, including a tandem pair of systemic lupus erythematosus (SLE)-associated polymorphisms, rs148314165 (−T) and rs200820567 (T > A) (referred to as the TT > A variants), located in an ENCODE-identified putative enhancer region located ~ 42 kb downstream of the tumor necrosis factor alpha-induced protein 3 (TNFAIP3) gene promoter [55, 56]. TNFAIP3 is a critical negative regulator of pro-inflammatory nuclear factor kappa B (NF-κB) signaling implicated in many autoimmune diseases, and therefore a suspected target gene of the identified enhancer [55–57]. Functional studies using the quantitative-PCR-based 3C method determined that this enhancer facilitated TNFAIP3 gene expression by bringing transcription factors, including NF-κB, to the TNFAIP3 promoter region via long-range enhancer–promoter interactions [55]. Importantly, the presence of the risk allele (−A/−A) in the enhancer was shown to significantly disrupt NF-κB binding and inhibit DNA looping of the enhancer to the TNFAIP3 promoter, effectively suppressing TNFAIP3 expression [57].

3D genome analysis is also a powerful method for identifying unsuspected candidate genes whose expression could be altered by risk variants in enhancers. For example, an enhancer harboring GWAS risk variants for rheumatoid arthritis was identified between oligodendrocyte transcription factor 3 (OLIG3) and TNFAIP3. In contrast to the well characterized role of TNFAIP3 in regulating inflammatory signaling pathways, OLIG3 is an important regulator of neuronal development and has no established role in the immune system [58], suggesting that TNFAIP3 was the likely target of this enhancer. To test this hypothesis, Capture Hi-C studies in both B and T cells from patients with rheumatoid arthritis were performed. Interestingly, these studies revealed that chromatin loops formed not only between the enhancer and the downstream promoter of TNFAIP3, but also with the promoters of interleukin 20 receptor subunit alpha (IL20RA) and interferon gamma receptor 1 (IFNGR1) located 180 kb upstream [59]. Functional follow-up studies using 3C confirmed that the presence of the risk allele of SNP, rs6927172, in the enhancer resulted in increased looping to the promoter and a concomitant increase in IL20RA gene expression [59]. Importantly, no long-range interactions were observed between the enhancer and the OLIG3 promoter, effectively eliminating this gene from further consideration. Together, these studies demonstrate the utility of targeted 3D genome applications to test and refine hypotheses regarding loop formation with enhancers harboring GWAS risk variants and their impact on gene expression.

Variants that disrupt anchor protein motifs, such as CTCF motifs that define TAD and insulated neighborhood boundaries, also have the potential to disrupt gene expression regulation by permitting aberrant boundary formation and allowing unrestricted enhancer activation of genes normally excluded from the neighborhood (Fig. 4c, d). An example of this occurs in T-cell acute lymphoblastic leukemia (T-ALL) where a mutation in a CTCF anchor motif disrupts the insulated neighborhood where the T-ALL-associated oncogene, TAL BHLH transcription factor 1 (TAL1), resides [60]. Importantly, this insulated neighborhood is devoid of a promoter, effectively inhibiting TAL1 expression. Disrupting this CTCF boundary allows the promoter of a nearby gene, STIL centriolar assembly protein (STIL), to reposition near TAL1 and activate expression [60]. This is one of many reported mutations in CTCF anchor motifs that have been shown to promote tumorigenesis by modifying looping activities around specific oncogenes [32, 61, 62].

3D genome technologies that typically require tens of millions of input cells have largely limited investigations into the potential implications of disrupting CTCF boundaries in primary immune cells involved in autoimmune disease pathogenesis. However, a recent report using 4C with NGS demonstrated that two asthma risk variants at the chromosome 17q21 locus, rs4065275 and rs12936231, individually altered CTCF binding motifs in CD4+ and CD8+ T cells [63]. In both cases, the 3D regulatory networks at this locus were significantly altered, resulting in increased expression of ORMDL sphingolipid biosynthesis regulator 3 (ORMDL3), a gene that facilitates cytokine production in the lung [63]. This study is one of the first to demonstrate that chromatin conformation methodologies can be used in primary cells to show that variants disrupting specific insulator boundaries can significantly alter the expression of genes implicated in disease pathogenesis. As more of these studies emerge in the future, we anticipate this to be a recurring theme, not only for cancer but complex diseases as well.