Histone modification is widely used as a means to classify enhancers according to their activity: H3K4me1 and the binding of trithorax-related mixed lineage leukemia (MLL) complex define primed or active enhancers; H3K27me3 is a key marker of poised or inactive enhancers; histone H3 lysine 27 acetylation (H3K27ac) is a hallmark of transcriptionally active enhancers (Figure 1) [11,12,13,14]. Recent trends highlight that rather than defining active enhancers with H3K27ac, different histone acetylation marks, such as simultaneous acetylation of histone H4 at both K5 and K8 (H4K5acK8ac) [15], Das et al., [submitted], histone H2B N-terminus multisite lysine (e.g., K5, K12, K16, and K20) acetylation (H2BNTac [16]), H3K122ac [17], and H4K16ac [18], to define active enhancers are emerging. A large number of histone modifications have been implicated in gene transcription, where H3K4me3 has been associated with gene promoter regions [19]. Table 1 summarizes the most well-characterized histone modifications and their involvement in E–P interactions.

Hyperacetylation of histone H4 at its N-terminal tail is essential for normal spermatogenesis [20,21] and occurs in several cancers, including nuclear protein in testis (NUT) midline carcinoma [22]. The bromodomain (BRD) and extra-terminal domain (BET) proteins, including BRD4, preferentially bind to histone H4 tails containing multiple acetylations within 1 to 5 amino acids, e.g., H4K5acK8ac, and only rarely bind to mono-acetylated histone species, including H3K27ac [23,24,25]. Although BRD4 primarily binds to multi-acetylated histone H4, most previous studies of BRD4 have focused on H3K27ac. BET proteins, including BRD4, have long been associated with large-scale control of the nuclear structure and higher-order chromatin organization [26,27]. For example, BRD4-NUT is an oncogenic fusion protein that drives NUT carcinoma [28]. NUT is normally expressed in post-meiotic spermatogenic cells, where it interacts with p300 and triggers genome-wide histone hyperacetylation [21]. Such interactions among BRD4-NUT, p300, and histone hyperacetylation lead to the formation of hyperacetylated nuclear ‘foci,’ corresponding to large chromosomal ‘megadomains’ (100 kb to 2 Mb) [22], which involve long-range acetylation-dependent inter- and intrachromosomal interactions [29]. This example illustrates how histone hyperacetylation (e.g., H4K5acK8ac, H2BNTac), targeted by readers or writers of BRD-containing proteins, facilitates long-range contact between enhancers and promoters and controls gene expression and how dysregulation of this process leads to disease, including cancer.

H3K27me3 is present at high levels in CpG islands, which are associated with the promoters of developmental genes in mammals [30,31]. In addition, along with H3K4me1, H3K27me3 is plentiful in poised enhancers (PEs), which become activated in human and mouse embryonic stem cells once these PEs are ‘marked’ with acetylation at H3K27 during the activation of associated genes [14]. In most cases, PEs are linked to developmental genes that are inactive in embryonic stem cells (ESCs) and that are expressed upon differentiation. Therefore, the PE chromatin signature is proposed to bookmark associated genes spatiotemporally and facilitate their activation once appropriate differentiation signals arise (Figure 1). Accordingly, loss of polycomb repressive complex (PRC) 2, which catalyzes the methylation of H3K27, induces a genome-wide redistribution of H3K27ac marking and activation of PEs [32]. The functional role of the polycomb-repressed or -poised state of enhancers remains unknown. Several studies suggest that polycomb-repressed enhancers may convert these regions to silencers [33,34], whereas PEs may be associated with the rapid activation of genes [35].

How do specific enhancers locate and contact promoters to accomplish transcriptional regulation? Remarkable progress has been achieved regarding the molecular principles and combinatorial logic underlying these processes [3,7], but chromatin-mediated E–P interactions remain incompletely understood. In general, chromosomal interactions create microenvironments, such as topologically associating domains (TADs, highly-interacting genomic regions defined as the population-level contact-frequency domains of higher interaction frequency within a region than between regions, [36]) and nuclear compartments that are characterized by the clustering of similar epigenetic marks, which constrain the search space and thereby increase the likelihood of interactions between enhancers and promoters in the genome [37]. In one model [7], interdependent layers of regulatory control cooperate to establish and maintain E–P interactions for robust cell-type–specific gene expression. Transcriptionally favorable cis-regulatory elements contain TFs, RNA polymerase II, and chromatin-remodeling and histone-modifying enzymes, whose combined activities induce chromatin accessibility. Specifically, proteins bound to enhancers and promoters may interact with each other non-randomly and preferentially [38,39], leading to cooperations that influence transcriptional regulation [7]. Chromatin loops that are marked with high levels of H3K27ac and low levels of H3K27me3 tend to change upon perturbation of PRC2, thus providing evidence that histone modification can alter the overall genomic architecture [40]. In cancer cells, H3K27ac dynamics modulate the interaction frequency between regulatory regions and can lead to allele-specific chromatin configurations that sustain oncogene expression [41].