10.1 Introduction

The epidermis at the surface of the skin provides a tractable and spatially hierarchical model to investigate the development of committed cells. The architecture of the epidermis provides the key structure for the physical barrier of the skin. Epidermal cells, or keratinocytes, in the most internal basal layer of the epidermis must strike a critical balance between self-renewal and differentiation in order to build a functional barrier across the entire body [1]. Keratinocyte self-renewal is characterized by parallel cell division within the basal layer. Expression of Keratin 5 (K5) and Keratin 14 (K14) marks these basal proliferating keratinocytes.

During differentiation, basal keratinocytes divide asymmetrically, giving rise to suprabasal keratinocytes that migrate outwards, entering the spinous layer [1,2,3]. Spinous and granular keratinocytes activate Keratin 1 (K1) and Keratin 10 (K10) expression concomitant with K5/K14 downregulation. During late terminal differentiation, the keratinocytes coordinately express many Epidermal Differentiation Complex (EDC) genes, including filaggrin (FLG) and FLG-like, late cornified envelope (LCE) , small proline-rich region (SPRR) , and S100 [4,5,6,7]. As the keratinocytes reach the outermost stratum corneum, they enucleate and are surrounded by a cornified envelope, the basic structural unit of the skin barrier, which is formed by transglutaminase-1-mediated cross-linking of scaffold proteins. The cornified envelopes are sealed together by keratinocyte-derived extruded lipids to form a semi-permeable barrier [2]. In mice, the pattern of functional barrier acquisition corresponds with maturation of the cornified envelopes, and proceeds from specific dorsal initiation sites at embryonic day (E)16, spreading to converge at the dorsal and ventral midline so that the whole embryo is impermeable by E17 [8].

Epidermal differentiation can be recapitulated in vitro by exposure of proliferating keratinocytes to high calcium levels [9, 10]. This process, called calcium switching, stimulates calcium receptor (CaR) and downstream phosphokinase C (PKC) signaling, which activates the Fos/Jun family of transcription factors that play important roles in keratinocyte differentiation (reviewed in [11]). Fos and Jun proteins form homo- or heterodimers that compose the AP-1 transcription factor complex [12]. In normal epidermis as well as in organotypic epidermal cultures, the expression pattern of AP-1 proteins is tightly regulated even within the differentiated layers [13]. Fos proteins are found in the nuclei of both basal and suprabasal keratinocytes. JunB and JunD are expressed in all layers of normal epidermis. Interestingly, c-Jun is expressed in the spinous layer, then disappears and reemerges in the outermost granular layer directly at the transition zone to the stratum corneum. As will be discussed in this chapter, many of the genes expressed in keratinocytes, in either proliferative or differentiated layers of the epidermis, contain AP-1 binding sites in their promoter regions. This suggests that specific combinations of AP-1 protein complexes bind to the enhancers for genes expressed at successive stages of epidermal differentiation.

Transcriptional regulation clearly plays a major role in development of the epidermis. So how does the keratinocyte decipher the genome to activate a transcriptional program specific to epidermal differentiation? Early studies of gene regulation focused on single gene promoters and the transcription factor binding sites contained within [14]. However, not all transcriptional activation is attributable to biochemical activity at the gene promoter, thus suggesting the contribution of other loci. Complete sequencing of the genomes of humans and model organisms has revolutionized our ability to further define, and ascertain the functions of, noncoding sequences in the regulation of gene expression [14,15,16]. Here we discuss the history of the conceptual advances in our understanding of the roles of enhancer elements in epidermal development. We discuss known functions for enhancer regulation of key genes that define the stages of epidermal development; the genetic, genomic and epigenetic features that allow us to identify enhancers; and the approaches that will enable a thorough understanding of the dynamic role of enhancers in gene activation.

10.2 What Is an Enhancer?

The concept of an “enhancer ” emerged in 1981. In that year, Pierre Chambon and George Khoury independently discovered a non-coding 72 bp tandem repeat sequence, upstream of the SV40 early gene promoters that was required for transcription [17, 18]. A subsequent study from Walter Schaffner identified the ability of the SV40 DNA sequence to “enhance” the expression of rabbit β-globin even when the SV40 sequence was placed thousands of base pairs away from the β-globin gene promoter in an expression vector [19]. Chambon observed similar results using the gene for conalbumin [20]. Further experiments performed by Paul Berg and Michael Fromm showed that the SV40 sequence was able to “enhance” transcription independent of its location (upstream or downstream of its target gene) and orientation (forward or reverse) [21]. This established the SV40 sequence as the prototype of a novel genetic element, an enhancer, and established the definition of a classical enhancer as a non-coding sequence that can modulate gene expression in a position- and orientation-independent manner.

The discovery of the SV40 enhancer paved the way for the identification of enhancers in other tissue types [15]. Often, searches for enhancers were prioritized and interrogated in the sequences surrounding the target genes that included upstream or downstream sequences, including 5′ and 3′ untranslated regions (UTRs), introns, and intergenic regions (reviewed in [22]).

In vivo studies of putative enhancers identified the spatiotemporal specificity of these sequences to ensure biologically relevant cell- and tissue-specific gene expression (reviewed in [23, 24]). Biochemical studies further identified clusters or arrays of transcription factor binding sites that act as “building blocks” of cis-regulatory modules that also show enhancer activity (reviewed in [15, 16, 25]). The enrichment of multiple transcription factor binding sites within an enhancer facilitates cell-specific expression largely attributable to combinatorial and differential binding of transcription factor family members in the context of different microenvironments. Based on these later studies, we more loosely define an enhancer as a non-coding sequence, containing clusters of transcription factor binding sites, that drives cell-, tissue-, or developmental stage-specific gene expression. Subsequent findings, described below, paved the way for the application of comparative genomics to enhancer discovery.

10.3 Transcriptional Regulation in Epidermal Development by Non-Promoter Sequences

Prior to the availability of whole genome sequences, biological insights into the transcriptional activation of key epidermal differentiation target genes initially focused on sequences immediately upstream of the transcription start site, relying on conservation of these sequences between mouse and human to identify putative regulatory regions [14]. These regions were cloned into reporter constructs and tested in vitro and in vivo to determine their ability to drive gene expression in the expected spatio-temporal pattern. Subsequent genetic deletion studies enabled the discovery of minimal promoter and regulatory core elements for activity and tissue specificity within segments of the cloned fragments. While the discovery of these promoter sequences identified key molecular players in gene activation and facilitated in vivo epidermal-specific genetic studies, it also highlighted the paucity of data to explain every nuance in gene expression for epidermal terminal differentiation.

10.3.1 Transcriptional Regulation of the Basal Keratin Genes (K5 and K14)

K5 and K14 are specifically expressed in mitotically active basal keratinocytes in most types of stratified epithelia. K5 and K14 proteins form stable heterodimers that are further cross-linked to provide structural support for cells [26, 27]. Their co-expression suggests that similar transcriptional mechanisms permit coordinated regulation of the two genes [28, 29].

Earlier studies identified marked similarities between the upstream sequences of K14 to viral and immunoglobulin enhancer elements [30]. Although 2.5 kb of 5′ upstream and 70 bp of 3′ downstream noncoding sequences of a cloned spliced human K14 gene are capable of directing gene expression in various cell lines, in vivo these regions drive epidermal-specific expression, demonstrated by reporter gene activity in transgenic mice that coincides with endogenous K14 expression [31]. An additional study of a 2 kb upstream sequence of K14 in transgenic mice also identified epidermal and outer root sheath hair follicle-specific reporter gene expression, conferred by a 700 bp sequence conserved between mouse and human, that exhibited keratinocyte-specific open chromatin as measured by DNaseI hypersensitivity [32]. The upstream cis-regulatory elements and promoter region of K14 contained binding sites for the transcription factor ETS, as well as AP-2, AP-1, and SP1 sites shared by the K5 regulatory region [32,33,34,35,36] (see below).

A similar study of the human K5 upstream sequence identified a 6 kb region that controls cell type-specific transcriptional activity [37]. A 90 bp sequence within this region is sufficient to activate expression specifically in the epidermis, hair follicles, and tongue. However, the pattern of expression was aberrantly switched from the basal layer to differentiated suprabasal cells. This suggested minimal promoter activity in the 90 bp upstream sequence for keratinocyte-specific expression and a requirement for other enhancers within the 6 kb upstream region for directing expression to the correct layers within stratified epithelia. The link between AP-2, SP1 and other unknown transcription factors for K5-tissue-specific activation was later determined based on the discovery of these transcription factor binding sites in the 6 kb sequence upstream of the human K5 transcription start site and corroborated with comparative studies of the bovine K5 promoter sequence [35, 37].

The mechanisms controlling coordinate expression of K5 and K14 were further elucidated by the discovery that mice lacking the transcription factor p63 fail to express K5 or K14 and do not develop a stratified epidermis [38, 39]. Biochemical studies revealed the presence of p63-responsive regulatory elements in an epithelial-specific enhancer 1.4 kb upstream of the K14 promoter [40,41,42,43] and in a transcriptionally active region upstream of K5 [44]. The shared and unique transcription factor binding sites harbored by the K5 and K14 enhancers suggest that both common and unique molecular mechanisms are involved in regulating their expression.

10.3.2 Transcriptional Regulation of the Suprabasal Genes

10.3.2.1 Early Differentiation: Regulation of K1 and K10 Expression

K1 and K10 are expressed in post-mitotic differentiating keratinocytes, and like K14 and K5, form heterodimers that are important for tissue integrity [45]. A 10.8 kb region surrounding the human K1 gene is sufficient to direct tissue-specific expression in transgenic mice during development, and is responsive to calcium induced differentiation in vitro [46]. Two calcium responsive elements were identified within a 1.7 kb enhancer in the 3′ flanking sequences of K1 [47, 48] including one element at the proximal 5′ end of the enhancer that is activated by AP-1 binding, and another that either suppresses or promotes the calcium response depending on the presence of vitamin D or retinoic acid, respectively [48]. Epidermal specific expression of K1 is directed by 207 bp of sequence at the distal 3′ end of this enhancer; however the transcription factor responsible for driving this expression has not been identified [47].

Insights into the transcriptional regulation of K10 expression were provided by analyses of mice deficient in the CCAAT/enhancer binding protein family member C/EBPβ, which exhibit decreased expression of K10 and K1 and mild epidermal hyperplasia [49]. This work was motivated by the observed increases in C/EBPα, C/EBPβ and C/EBPδ mRNA levels, as well as C/EBPα and C/EBPβ protein induction, upon calcium-induced differentiation of mouse primary keratinocytes [50]. DNaseI footprinting and gel-shift assays performed on the human K10 promoter, which had been identified via homology with the bovine sequence, revealed that C/EBP binding to three distinct elements was required for K10 expression [51]. While C/EBPα and C/EBPβ both activate K10, they are differentially expressed in stratified epidermis. C/EBPβ acts early in the basal and spinous layers of the epidermis and is later superseded by C/EBPα during keratinocyte differentiation and upward migration into the granular layer. Besides C/EBP binding to three intact C/EBP binding sites within the K10 promoter , full activation of K10 during differentiation also requires specific binding of AP-2 [51]. Consistent with co-expression of K1 and K10, consensus C/EBP binding sites are also observed in the upstream promoter regions of the K1 gene [52, 53].

10.3.2.2 Regulation of EDC Genes

Terminal differentiation in the epidermis is marked by the expression of four gene families that lie within the EDC and are coordinately activated at the transcriptional level: small proline-rich region (SPRR), late cornified envelope (LCE), filaggrin (FLG) and FLG-like (FLG-like), and S100 genes [4,5,6,7]. Pioneer studies to elucidate gene expression in relation to epidermal differentiation targeted the SPRR gene family members involucrin (IVL) and loricrin (LOR). These two important marker genes are distinctively expressed in terminally differentiated keratinocytes, and encode structural proteins that are cross-linked with many of the other proteins encoded by EDC genes to form the cornified envelope, the basic structural unit of the stratum corneum [54]. IVL is cross-linked early in the formation of the cornified envelope [55] and LOR is in turn cross-linked to the existing scaffolding containing IVL [56]. In the developing mouse embryo, Ivl and Lor transcripts are upregulated as early as E15.5 [57], and protein expression can be observed by E16.5, corresponding to the onset of skin barrier formation [6, 8]. The tight correlation of IVL and LOR expression with keratinocyte terminal differentiation makes them ideal candidates for studying the mechanisms that underlie the switch from a proliferating to a differentiating program.

10.3.2.3 Involucrin

A 3.7 kb upstream sequence of IVL directs keratinocyte-specific expression of a β-galactosidase reporter gene in transgenic mice [58]. Deletion analysis showed that this sequence comprises distal- and proximal-regulatory regions (DRR and PRR) or enhancers [59, 60], and contains an AP-1 binding site that is required for expression and synergizes with an adjacent SP1 binding site [61]. The DRR AP-1 site binds FRA-1, JunB, JunD, and p300, a histone acetyltransferase often associated with enhancers [62, 63], while the SP1 site interacts with SP1, SP3, and KLF4 transcription factors [59, 61, 64,65,66]. Because FRA-1 and KLF4 are both known to interact with p300 [65, 67], these data indicate that a complex of transcription factors forms on the DRR to drive IVL expression during keratinocyte differentiation.

10.3.2.4 Loricrin

Transcriptional activation of mouse loricrin expression was first localized to a 6.5 kb region spanning the loricrin gene [68]. Transgenic reporter mice in which LOR coding sequences were replaced by a β-galactosidase gene revealed that the remaining 1.5 kb of 5′-flanking sequence, a small noncoding exon, a 1.1 kb intron, a single coding exon, and 2.2 kb of 3′-flanking sequence from the mouse loricrin gene drive epidermal-specific, but not differentiation-specific expression. Minimal promoter activity, dependent on an AP1 site conserved between mouse and human, was mapped to a 60 bp sequence upstream of the transcription start site. In the case of the human LOR gene, enhancers located within 1.5 kb of 5′-flanking sequence and 9 kb of 3′-sequence are responsible for tissue- and differentiation-specific expression of a human LOR transgene in transgenic mice [69]. As few as 154 bp of 5′ sequence upstream from the cap site can direct expression specifically in cultured NHEK and HaCAT keratinocytes, in an SP1/c-Jun and p300/CREB-dependent manner [70]. Differential occupation of this site by SP3/CREB-1/CREMα/ATF-1/Jun B, and an AP-2-like protein (named keratinocyte-specific repressor-1 (KSR-1)) in a repressive state, and SP1/c-Jun/p300/CBP in an active state during differentiation, enabled LOR transcriptional resolution in stratified layers [70].

10.4 Identification of Enhancers by Comparative Genomics in the Post-Human Genome Era

The early studies described above identified the transcription factors responsible for transcriptional regulation of major epidermal differentiation genes. However, the molecular mechanisms underlying transcription factor-driven activation of these genes were less clear. Furthermore, these studies were limited to the analysis of proximal promoter and enhancer regions. The availability of complete genome sequences for a wide range of model organisms and animal species greatly facilitated the identification of putative enhancer and other regulatory sequences. Specifically, it shifted the discovery of enhancers toward a more systematic procedure – high-throughput and on a genome-wide scale. The discovery of putative enhancers was predicated on the identification of noncoding sequence conservation across multiple species and facilitated by multiple-sequence alignments [71]. Indeed, a study of conserved noncoding sequences present between human and pufferfish or ultraconserved between human and mouse found that many of these conserved noncoding sequences exhibit enhancer activities that are developmental- and tissue-specific in transgenic mouse assays [72]. Below we discuss the discovery of enhancers for relevant skin biology genes using comparative genomics.

10.4.1 SPRR Genes

The availability of whole genome data enabled the discovery of conserved noncoding sequences (CNSs and based on sequence alignments) in the SPRR locus that is clustered within the EDC [73]. Small proline-rich (SPRR) proteins are the primary constituents of the cornified envelope [74]. Many of the SPRRs are coordinately upregulated under stress conditions to rapidly build a temporary barrier [73]. Of the DNaseI hypersensitive sites (HSs) residing within CNSs, one demonstrated enhancer activity under conditions when the SPRR genes are coordinately upregulated, suggesting its potential as an enhancer region that coordinates SPRR gene expression.

10.4.2 PADI3

The family of peptidylarginine deaminases (PADs) is encoded by a cluster of 5 PADI genes on human chromosome 1p35–36 [75]. Peptidylarginine deaminase 3 (PADI3) encodes PAD3, which is involved in filaggrin metabolism, releasing individual filaggrin monomers that contribute to the natural moisturizing functions of the skin barrier. Using a comparative genomic approach, an enhancer located 86 kb from the PADI3 gene promoter was identified and determined to be calcium sensitive [76]. This enhancer was found to trigger expression of PADI3 upon epidermal keratinocyte differentiation, and links PADI3 expression to AP-1 transcription factors through chromatin opening and looping (see Sect. 10.5).

10.4.3 The Role of the CNE 923 Enhancer in Coordinate Regulation of the EDC

The coordinate regulation of EDC genes, as well as the synteny and linearity of the EDC locus across a wide range of mammalian species, suggest a molecular mechanism originating at the proximal genomic level. To delineate this, investigators screened for enhancer elements within the EDC [7]. In 2010, 48 conserved noncoding elements (CNEs) within the human EDC were identified from sequence alignments of orthologous EDC loci across eutherian (human, chimpanzee, macaque, mouse, rat, dog) and metatherian (opossum) mammals. Approximately 50% of these CNEs exhibited dynamic regulatory activity, and were thus identified as potential cis-regulatory elements or enhancers that might synergistically or independently coordinate EDC gene expression during skin barrier formation. Among these, human CNE 923, located approximately 923 kb from the transcriptional start site of the most 5’ EDC gene, S100A10, induced the highest luciferase reporter activity in proliferating and differentiated keratinocytes. CNE 923 exhibited DNaseI hypersensitivity in primary human keratinocytes, and was sufficient to drive reporter gene expression in the developing E16.5 epidermis in transgenic reporter mice. The activity of CNE 923 was monitored in an independent transgenic mouse line and was sufficient to drive β-galactosidase activity in the same dorsal to ventral pattern of barrier acquisition that coincided with EDC gene activation [57]. These studies provided compelling evidence for CNE 923 as an epidermal-specific enhancer and a potential LCR. CNE 923 was also noted to form dynamic chromatin interactions with a number of EDC genes, and this was sensitive to keratinocyte differentiation and dependent on the AP-1 transcription factor. The role of CNE 923 for mediating EDC chromatin architecture is discussed further in Sect. 10.7.

10.4.4 p63

A long-range cis-regulatory enhancer of p63 (p63LRE) spanning a 12 kb region in mice was recently identified by comparative genomics [77]. p63LRE comprises two evolutionarily conserved modules acting in concert to control tissue- and layer-specific expression of the p63 gene. Both modules are in an accessible and active chromatin state in human and mouse keratinocytes and in embryonic epidermis, and are strongly bound by p63. p63LRE activity is dependent on p63 expression in embryonic skin and also in the commitment of human induced pluripotent stem cells toward an epithelial cell fate. C/EBPα, C/EBPβ, and the POU domain-containing protein Pou3f1 repress p63 expression during keratinocyte differentiation by binding the p63LRE enhancer. The availability of these transcription factors in the outermost layers of the epidermis accounts for increased repression of p63, thereby relieving p63-mediated repression of EDC and keratinocyte differentiation genes in these layers, and limiting p63 activity to the basal layers. We discuss p63-bound sites and enhancer regions on a genome-wide scale in Sect. 10.7.4.

10.5 Epigenetics and Chromatin Remodeling

So far, we have discussed studies that make compelling arguments for transcriptional regulation in which the key determinants are transcription factor binding to individual nucleotide motifs. However, this view has been challenged by our increased understanding of the non-random packaging of linear DNA into histones, which alters the accessibility of DNA segments to transcription factor binding, and by the discovery of specific post-translational modifications to histone components that dynamically regulate DNA accessibility. Specifically, the identification of histone deacetylase (HDAC) and histone acetyltransferase (HAT) enzymes and their targets furnished the first direct evidence linking histone modification states to transcriptional regulation [78, 79]. These findings paved the way for investigations into higher order chromatin structure and genome compartmentalization. HATs catalyze the transfer of an acetyl group from acetyl CoA to the ε-amino group of lysine residues on histones. This mark is generally associated with active genes. Conversely, HDACs remove the acetyl group from acetyl-lysine (Ac-Lys) to regenerate the free ε-amino group, causing chromatin compaction and a transcriptionally repressive environment (reviewed in [80, 81]). The identification and characterization of other classes of histone modifying enzymes soon followed, implicating kinases [82, 83], lysine and arginine-specific methyltransferases [84,85,86], arginine deiminases [87, 88], ubiquitinases [89], deubiquitinases [90,91,92], and lysine- and arginine-specific demethylases (HDMs) in transcription regulation [93,94,95]. Collectively, these discoveries highlighted the role of chromatin modifications in governing eukaryotic gene expression and other DNA-dependent functions and ushered in a new era of chromatin-based epigenetic studies (reviewed in [96, 97]).

Within the context of the skin, p300 and CBP, two enhancer-associated HATs, have been implicated in the regulation of IVL and LOR (discussed earlier) [65, 70]. Epidermal-specific deletion of Actl6a, an essential component of HATs, resulted in de-repression of KLF4 and Brg1/Brm binding, thereby aberrantly activating epidermal differentiation genes and abolishing epidermal progenitor function [98]. HDACs also play important roles in epidermal development, specifically in preventing senescence of basal progenitor cells [99, 100]. Epidermal-specific deletion of both HDAC1 and 2 in embryonic epidermis resulted in a phenotype resembling the effects of loss of p63. This phenotype was associated with de-repression of ΔNp63-repressed target genes including the senescence gene p16 [101]. HDAC1/2 localizes to the promoter regions of ΔNp63-repressed targets in cultured human keratinocytes, and histones in these regions are hyper-acetylated following HDAC inhibition, indicating a requirement for HDAC1/2 in ΔNp63-mediated repression. Together, these data reveal essential roles for histone modifying enzymes in controlling the activities of key regulators of epidermal development (reviewed in [102]).

Ezh2, an essential component of the polycomb repressor complex 2 (PRC2) is also important for epidermal development [103]. Epidermal-specific loss of Ezh2 resulted in early epidermal differentiation owing to precocious recruitment of AP-1 transcription factor to the EDC. This indicated a role for Ezh2 in gene repression in proliferating keratinocytes by promoting histone H3K27 trimethylation (H3K27me3). However, loss of Ezh2, while decreasing H3K27me3 marks, was not sufficient to alter transcriptional status unless AP-1 was also recruited to the affected region, emphasizing an important and direct role for AP-1 in keratinocyte differentiation.

Recent cell biology studies have also highlighted the importance of chromatin remodeling in permitting efficient and coordinate regulation of clusters of genes in epidermal development. During epidermal development, the EDC locus moves away from the nuclear periphery and towards the nuclear interior prior to activation of EDC gene expression [104]. Ablation of either p63, a master regulator of epidermal development [38, 39, 105], or Satb1, a higher-order genome organizer that binds to the EDC in epidermal progenitor cells, caused altered chromatin conformation of the EDC, and loss of expression of genes that lie in the central domain of the EDC [106]. These findings identified Satb1 as an important downstream target of p63 required for proper establishment of higher-order EDC chromatin structure and coordinated gene expression [106]. Similarly, p63 and its direct target Brg1 are essential in remodeling the higher-order chromatin structure of the EDC and positioning the locus within the 3D chromatin landscape to allow efficient expression of EDC genes in epidermal progenitor cells during skin development [107, 108].

10.6 Methods to Identify Enhancers by ENCODE in the Post-Human Genome Era

Increased understanding of the chromatin state of the genome has forced us to re-examine initial models for the control of gene expression that focused entirely on the role of cis-regulatory elements (Sect. 10.3). Following completion of the human genome sequence, the National Human Genome Research Institute, recognizing the need to more fully understand the regulation of gene expression, launched the Encyclopedia of Non-Coding Elements (ENCODE), a collaborative public research project to identify and characterize the function of noncoding elements in the genome, and develop the tools and technology to achieve this goal [63]. Next-generation sequencing was instrumental in producing these genome datasets in a cost-effective manner. The ENCODE studies as well as work by others have greatly facilitated our ability to identify enhancers on a genome-wide scale based on chromatin modifications that are unique to these regulatory elements, such as DNaseI hypersensitivity (open chromatin), histone modification epigenetic marks (H3K27Ac, H3K4me1), and transcription factor binding (p300, activating TFs) associated with functional enhancers (reviewed in [15, 25, 109]). High-throughput chromatin immunoprecipitation linked with deep sequencing (ChIP-seq) for specific histone modifications has enabled the discovery of new enhancers and paved the way for further downstream functional analyses (reviewed in [15]). Table 10.1 lists signature genomic marks and methods to identify enhancers.

Table 10.1 Genomic features of enhancers and methods for prediction and functional validation
Full size table

Chromatin looping and tracking have been proposed as models to explain how distant enhancers are able to regulate their target genes [143, 144]. The first experimental demonstration of direct interactions between distantly located enhancers and target genes was made possible by the development of chromosome conformation capture (3C) techniques [125]. 3C was used to demonstrate that loop formation between the β-globin LCR enhancer and gene accompanied transcriptional activation [113,114,147], and established a paradigm that was later validated for numerous other loci, including the α-globin gene cluster, TH2, IFNG, MHC class II and IgH loci [148]. Transcription factor ChIP-chip studies also revealed that enhancers could be located even further from their target genes than previously thought, as far as 10–20 kbs to several Mbs away [149]. Often, these proximal and distal enhancers interact to co-regulate a target gene.

Recent improvements in chromosome conformation capture methods have allowed us to examine the chromatin interactions of genomic regions at varying levels of depth and resolution. 4C (circular chromosome conformation capture) detects all interacting sequences with a sequence of interest using a bait such as an enhancer [126]. 5C (chromosome conformation capture carbon-copy) is designed to detect many known interactions with numerous baits and typically within a gene locus [127], while the Hi-C approach is aimed at detecting all chromatin interactions [128]. Methods such as ChIA-PET (Chromatin interaction analysis with paired-end tag sequencing ) combine 5C and Hi-C methods to simultaneously identify genome-wide chromatin interactions and the proteins that bind interacting sequences [150]. Evidence for epigenetic modifications, chromatin looping, and the interplay between the two has been obtained relatively recently, and has provided new insights to our understanding of the biochemical aspects of enhancer-mediated transcriptional regulation.

10.7 Emerging Concepts for Understanding Enhancer-Promoter Interactions

10.7.1 Mechanisms Underlying Enhancer-Promoter Interactions and Topological Association of Chromatin Domains in the Regulation of Gene Expression

Apart from the aforementioned roles of p63, Satb1, Brg1, and AP-1 in transcriptional regulation during epidermal differentiation, additional molecular mechanisms underlying transcriptional regulation by enhancers are less clear. Here we discuss studies that have elucidated the mechanisms of enhancer-promoter interactions in epidermal biology, and lessons we can learn from other tissue models.

10.7.1.1 The Formation and Biology of Enhancer-Promoter Chromatin Loops

Formation of enhancer-promoter chromatin loops as a mechanism to drive gene activation has emerged as a major concept in the exploration of enhancer-promoter interactions. One of the best studied loci is the evolutionarily conserved β-globin locus that plays key roles in hematopoiesis (reviewed in [151]). The 5 globin genes (ε, Gγ, Aγ, δ and β) form a cluster, and are expressed in a developmental-stage- and tissue-specific manner, controlled by a LCR. The GATA-1 transcription factor and Ldb1 are required to form a chromatin interaction between the β-globin LCR and the β-globin promoter for transcriptional activation in erythroid cells [152, 153]. The requirement for formation of a chromatin loop for gene transcription was demonstrated using artificial zinc fingers (ZF) [154] to force chromatin loop formation by tethering Ldb1 to the β-globin locus control region in GATA-1 null erythroblasts. This was found to be sufficient to activate β-globin gene expression. This work was the first to demonstrate the causality of chromatin spatial interactions in promoting gene transcription.

10.7.1.2 Regulation of the EDC by CNE 923 Via AP-1 Mediated Chromatin Interactions

The organization of the EDC as a conserved cluster of genes with related functions [7] is reminiscent of the organization of the β-globin locus [151]. Moreover, the identification of the EDC enhancer CNE 923 as an epidermal specific enhancer (described in Sect. 10.4.3), suggests its potential function as an LCR that drives coordinate and concomitant EDC gene expression in a manner similar to the β-globin LCR. In line with this, 3C experiments performed with respect to the mouse orthologous CNE 923 sequence in proliferating primary mouse keratinocytes demonstrated that this sequence interacts with nine EDC gene promoters (Sprr2a1, Sprr2d, Sprr2f, Sprr1b, Sprr3, Ivl, Lce1b, Lce1a2, and Crct1 gene promoters) that lie as far as 500 kb from CNE 923, despite the lack of EDC gene expression in these cells, suggesting a poised and enhancer-mediated chromatin state [57]. After calcium-induced differentiation, CNE 923 interacted with the promoters of 11 EDC gene, including Lce3b, S100a6, Sprr2a1, Sprr2b, Sprr3, Sprr4, Ivl, Lce6a, Lce1b, Lce1e, and Crct1. The interaction between CNE 923 and S100a6, located 2 Mb apart, suggested that the EDC chromatin domain is compacted during differentiation to bring more linearly distal genes into close proximity with the enhancer. These results highlight dynamic chromatin looping interactions with 923 that are associated with concomitant EDC gene expression.

Bioinformatics analysis of human CNE 923 identified two highly conserved sequence blocks (PhastCons) that are required for enhancer activity. A consensus AP-1 transcription factor binding site within the most 5′ block is required for maximal enhancer activity, and pharmacological inhibition of AP-1 binding in calcium-induced keratinocytes represses EDC gene expression, and causes loss of c-Jun/AP-1 binding to 923 and aberrant chromatin remodeling. These observations identify a link between an epidermal-specific EDC enhancer and c-Jun/AP-1 transcription factor binding, and together with other studies [104, 106, 107], suggest that further analysis of the 3D structure of chromatin would aid our understanding of EDC regulation.

10.7.2 Higher Level Chromatin Architecture : Topologically Associated Domains (TAD) and Chromosome Territories (CT)

Topologically associated domains (TADs) were first identified by Hi-C [155, 156], and are defined as distinct clusters of enhancer-promoter interactions [157, 158]. At the highest order of chromosome organization, spatially proximal TADs compose a chromosome territory (CT), a compartment within the nucleus that is often segregated in a chromosome-specific manner (reviewed in [159]). Actively transcribed gene-rich loci that are in an open conformation are more likely to loop out of their CTs, suggesting that the space between CTs is important for genomic loci to access the transcription machinery (reviewed in [159]).

The importance of long range enhancer-promoter interactions in the context of a CT was demonstrated in studies of the developing limb bud [160], where differential expression of the Sonic hedgehog (Shh) gene is mediated by specific interactions between the Shh promoter and a long-range enhancer MFCS1. In the intermediate portion of the limb bud, which lacks Shh expression, the long-range enhancer is spatially and linearly distant from the Shh coding region. In anterior limb bud cells, the long-range enhancer interacts with the Shh coding region, but the interactors remain in a poised state within their CT. However, in cells of the zone of polarizing activity (ZPA) where Shh is actively expressed, 3D-FISH showed that the interacting regions relocate outside the CT.

A similar mechanism has been observed in the control of EDC gene expression in differentiating keratinocytes, where Satb1 binds to several sites across the EDC locus, compacting the EDC chromatin architecture into a densely looped structure that, upon relocalization of the locus into the nuclear interior by Brg1, enables efficient and coordinate activation of EDC genes [106, 107]. Similar mechanisms may be employed to control genes such as K1 and K10 which are coordinately activated as keratinocytes transition from basal to suprabasal layers of the epidermis.

10.7.3 Involvement of Cohesin and CTCF in Forming Active Chromatin Hubs

Enhancer activity can also be modulated by insulators that function as physical barriers to optimal enhancer-promoter formation for transcriptional activation. Here we discuss newly recognized attributes of enhancers and new direct roles for CCCTC-binding factor (CTCF) bound insulators in enhancer–promoter interactions and in broadly configuring the genome (Reviewed in [143, 161]).

Cohesin is a complex of proteins that holds sister chromatids together after DNA replication, until the sister chromatids separate at anaphase (reviewed in [162]). Analysis of the effects of mutations in cohesin subunits identified a role for cohesin in regulating enhancer-promoter interactions and gene expression (reviewed in [162]). Subsequently, it was discovered that mammalian cohesin complexes can be recruited to DNaseI hypersensitive sites and conserved noncoding sequences by the CTCF DNA binding protein [134,135,165]. CTCF often binds at insulators and at boundary elements to demarcate active chromatin hubs and limit the effect of enhancers [166], and cohesin contributes to CTCF’s enhancer blocking activity [163, 165]. Studies of the apolipoprotein gene cluster [167], the globin locus [168], and the T-cell receptor (Tcra) locus [169] demonstrated cooperation of CTCF and cohesin to mediate insulators corresponding to TAD boundaries , thereby maintaining proper chromatin loop formation and localization of transcriptional apparatus at the gene promoters to control gene expression.

These mechanisms are relevant to human disease, as chromosomal rearrangements of the conserved TAD-spanning WNT6/IHH/EPHA4/PAX3 locus that disrupt a CTCF-associated boundary domain within a TAD, cause limb malformations in humans. Mice harboring the equivalent disease-relevant rearrangements were generated using CRISPR/Cas9-genome editing, and displayed ectopic limb expression of a gene that lies within the locus but is not normally expressed in limb development, due to misplacement of a cluster of limb enhancers relative to TAD boundaries [158]. This finding demonstrates the functional importance of TADs for orchestrating gene expression via genome architecture and suggests the utility of analyzing disease-associated large-scale chromosomal rearrangements in delineating TAD boundaries.

Genomic studies in other tissue types and loci have also greatly advanced our understanding of the molecular mechanisms of CTCF-mediated chromatin looping events. Chromatin conformation Hi-C capture data with parallel CTCF ChIP-seq identified a nonrandom pattern of forward and reverse orientation for a given pair of CTCF binding sites involved within a chromatin loop [141,142,172]. Subsequent studies later identified a functional role for the directionality of CTCF binding sites to influence chromatin topology and enhancer-promoter function [173, 174]. CRISPR/Cas9-generated inversion of a genomic region spanning CTCF boundary elements in the P-cadherin enhancer altered chromatin topology [173]. Deletions of individual CTCF binding sites by CRISPR/Cas9 led to loss of CTCF, reduced cohesin binding, and reduced or abolished chromatin looping, while inversions restored CTCF and cohesin binding but reduced chromatin looping [174]. However, the impact of the presence or directionality of CTCF binding sites on proximal gene expression varied, suggesting the influence of additional factors. This was further evidenced by deletions of 4 CTCF binding sites in the casein locus that did not affect gene expression but instead caused more distal de novo Sultd1 activation by way of other nearby enhancers [175]. These studies demonstrated a novel governing principle for chromatin architecture in controlling gene expression, and holds great potential for more accurate and contextual prediction of the functionality of enhancers in the skin and other tissues.

10.7.4 The Role of Mediator and Super-Enhancers

More recently, a new class of enhancers called “super-enhancers” has been identified [176]. Super-enhancers are marked by high levels of Mediator coactivator complex occupation as determined by ChIP-seq and span much larger distances than typical enhancers (approximately 8.7 kb versus 703 bp). Mediator is a major component of the transcription pre-initiation complex (PIC) machinery with RNA polymerase II (RNA pol II) and is required for activator-dependent transcription in vitro and in vivo (reviewed in [177]). Reduced levels of Mediator specifically affect gene expression near super-enhancers [176]. This was convincingly demonstrated by the loss of enhancer-promoter loops of select genes upon deletion of Mediator [178, 179]. Mediator-occupied super-enhancers also exhibited enriched binding of transcription factors that are master regulators involved in cell-identity in ESCs, pro-B cells, T helper cells, myotubes, and macrophages, among other cell types. However, one can argue that Mediator and even cohesin binding may not be entirely necessary for gene activation as it was recently determined that chromatin looping between the globin LCR and the β globin locus via Ldb1 was established in the absence of Mediator and cohesion binding [180]. More recent work identified an emergent paradigm for super-enhancers in the direct biogenesis of master regulator miRNAs for tissue specificity [181].

A role for super-enhancers was recently identified in epidermal stem cells [182]. The target genes associated with epidermal stem cell-specific super-enhancers identified by H3K27Ac and Mediator ChIP-seq methods contain a high frequency of binding motifs for the transcription factors Sox9, Lhx2, Nfatc1 and Nfib, that are important for maintaining hair follicle stem cells [154,155,156,186]. ChIP-seq experiments showed that these transcription factors bind at high frequency to super-enhancers relative to typical enhancers. Lineage tracing during mouse epidermal development, wound-healing, and in cell culture showed that super-enhancers are remodeled according to their cellular environment, supporting the idea that enhancers are activated or silenced in a lineage-specific fashion. The dynamic behavior of these regulatory elements in human keratinocyte progenitors as well as in their differentiating progeny was further supported by the discovery of genomic profiles for superenhancers (H3K27ac) and typical enhancers (H3K4me1, H3K27ac, H3K4me3) [187] that were different across these distinct keratinocyte states. Binding of the p63 master regulator for keratinocyte differentiation was observed in a core set of super-enhancers and enhancers (shared between mouse and human) associated with keratinocyte-specific gene expression, but also notably in mouse-specific regions that underlie species-specific transcriptional differences [188].

Analysis of sites bound by Mediator thus enables identification of key transcription factors and enhancer sequences in a variety of cell types and their sensitivity to changing conditions, highlighting the potential of this approach as a tool to pinpoint important regulatory sequences involved in cell and tissue homeostasis, even without prior knowledge of the transcription factors or genes involved.

10.7.5 Non-coding RNAs

As discussed above, the establishment and maintenance of chromosome territories or TADs bring enhancer and promoter interactions into close proximity to facilitate gene activation (reviewed in [157]). In addition to the roles of cohesin and CTCF occupation in enhancer-promoter interactions, long non-coding RNAs (or lncRNAs) may also serve as a scaffold for the assembly of transcription factors and chromatin remodeling enzymes at the promoter [189]. The lncRNA HOTAIR, first discovered in adult skin [190], targets members of the polycomb repressive complex 2 (PRC2) to specific genomic loci, including Ezh2 [189] which, as discussed earlier, regulates epidermal differentiation [103]. HOTAIR also directly interacts with LSD1 [189], a protein complex that is associated with maintenance of epidermal stem cells in an undifferentiated state [191]. Moreover, PRC2 and LSD1 interact in a HOTAIR dependent manner [189], suggesting a potential role for HOTAIR in epidermal development.

More recently, evidence of transcription of functional enhancer RNA (eRNA) from non-coding enhancer sequences has emerged [192]. eRNAs are considered a separate class of non-coding RNA. Unlike lncRNAs, which are marked by H3K4me3 at their promoters and are frequently spliced and polyadenylated, eRNAs are transcribed from enhancer regions marked by H3K4me1, an absence of H3K4me3 histone modifications, and to a lesser extent, post-transcriptional modifications. During transcriptional activation, enhancer RNAs participate in a bi-directional enhancer-promoter activation feedback loop, whereby chromatin looping between the enhancer and promoter brings the bi-directionally transcribed eRNA near the target gene to drive gene expression, at the same time allowing the eRNA to stabilize the enhancer-promoter loop with the help of the Mediator complex (reviewed in [87]). This mechanism for eRNA mediated gene activation has been identified in the expression of various lineage-specific genes, including macrophage-specific genes [193], estrogen-regulated genes in breast cancer cells [194], and p53-regulated genes that induce cell cycle arrest [195]. To date, a role for eRNAs in epidermal development has yet to be identified, but should be kept in mind as a possible mode of gene regulation.

10.8 Conclusion

In this Chapter, we have outlined the history of scientific discoveries related to the transcriptional regulation of key genes involved in skin biology leading up to the concept of enhancers and the molecular mechanisms that orchestrate this process at the chromatin level in the post-human genome and post-ENCODE eras. Our current knowledge of the mechanisms regulating expression of genes involved in determining the proliferative or differentiated state of keratinocytes identifies common patterns in modes of regulation. These observations are consistent with mechanisms identified in other tissue systems. However, regulation of keratinocyte-specific genes appears to heavily utilize p63, AP-1, SP1, and C/EBP transcription factors in particular.

Earlier studies of proximal regions of epidermal genes including keratin genes, IVL, and LOR, and functional genetic approaches to delineate the sequences that drive transcriptional activities, underscored the role of the complex interplay of AP-1 proteins, SP1 and C/EBP transcription factors in epidermal gene regulation. However, we have yet to understand the larger regulatory landscape of the keratin gene clusters.

By contrast, studies of the EDC locus have benefited from the availability of bioinformatics tools and high-throughput methods, revealing that proper control of gene expression is an intricate hierarchy of events that depends firstly on appropriate post-translational regulation of histones to designate chromatin regions as accessible or inaccessible, followed by further organization and remodeling of the EDC together with proper nuclear spatial positioning [106, 107]. The establishment of regulatory landscapes and chromosome territories or topologically associated domains (TADs) brings regulatory enhancers and promoters into close proximity, allowing AP-1, together with other as yet unknown factors, to be poised for efficient coordinate activation of the EDC genes upon induction of epidermal differentiation [57].

The varied roles of AP-1 transcription factors provide an example of how different combinations of common transcription factors are brought together to form complexes to modulate or alter gene expression under different conditions such as spatial context and developmental stage. Distinct complexes may be formed by altering the occupancy of DNA binding proteins at arrays of enhancer elements in close proximity to gene promoters. Complex assembly may also be driven in part by long-range chromatin interactions that bring distal enhancers close to their target gene promoters, and this can be mediated in a myriad of ways. Within the field of skin biology, we have only recently begun to define the mechanisms by which enhancer-promoter interactions initially occur.

10.9 Future Studies

Further dissection of the regulatory principles underlying the gene expression patterns that accompany and drive epidermal development will require a two-pronged approach. The availability of whole genome sequences from increasing numbers of species [196], high-throughput techniques such as ChIP-seq, RNA-seq, ATAC-seq, and the development of bioinformatics tools to allow the integration of such data [197], have made it possible to approach the analyses of gene regulation at the genomics level. Specifically, these advances have enabled researchers to identify enhancer landscapes across the genome and further elucidate the principles that govern enhancer-promoter interactions and genome organization relevant to gene expression. Such an approach identified a surprising role for the DNA methyltransferases, Dnmt3a and Dnmt3b, in occupying enhancers and functionally linked them to epidermal stem cell function [198]. Furthermore, the discovery of skin disease sequence variants enriched in superenhancers highlights the clinical impact of this research and future studies in this area [187].

While several key mechanisms, such as the roles of chromatin looping, 3D genomic architecture, and non-coding RNAs were first demonstrated in non-epidermal tissues and cell-types, they are crucial in constructing a framework for understanding epidermal-specific enhancer-promoter interactions. For example, cohesin and CTCF have been shown to play a role in maintaining enhancer-promoter interactions in a multitude of cell-types and tissues [138,139,169, 199]. This strongly suggests that they are also likely to be major players in mediating the formation of chromatin loops in differentiating keratinocytes. This could be tested by identifying changes in CTCF and cohesin occupation in the different epidermal layers, as well as through functional genetic studies.

In order to completely understand the mechanisms driving enhancer regulated gene expression, we must continue to incorporate multi-disciplinary approaches and novel methods to approach the problem from genetic, molecular and cellular perspectives. Drawing from the expertise of the evolutionary biology field, we are now able to identify candidate regions with regulatory potential faster than ever before [71, 72, 200]. Additional enhancers that are not evolutionarily conserved can be identified using sequencing technologies (see Table 10.1 and [201]). Current advances in genome editing (CRISPR/Cas9, TALENs) have also made it easier to test the functions of endogenous enhancers [158, 173,174,204] and provide an improvement over the use of artificial transgenes that remove enhancers from their appropriate genomic context. The precision of these genome-editing methods enables us to directly test hypotheses regarding enhancer functions at specific locus/loci within the regulatory landscape. For example, CRISPR/Cas9 genome editing of several enhancers has demonstrated their requirement for gene expression [158, 173,174,204]. A high-throughput genetic screen for a targeted set of regulatory elements in the POU5F1 locus using CRISPR/Cas9 editing has also confirmed the functional role of enhancers for POU5F1 expression, while simultaneously revealing a new class of “TEMP” enhancers that are characterized by temporary loss of gene expression and weak reporter activity [205]. CRISPR/Cas9-mediated recombination of orthologous yet divergent enhancer sequences in mice has also provided a comparative functional assay to further assess the importance of transcription factor binding sites during development [204]. For instance, in vivo replacement of a mouse Sonic hedgehog enhancer with the orthologous snake-specific Sonic hedgehog enhancer led to a limb defect that was rescued by introduction of an ETS binding site that had been lost in the snake. In addition to modern genetics and genomics approaches, molecular tools, such as live-imaging and high-resolution microscopy and biomechanics studies will extend our understanding of the dynamics of enhancer-promoter interactions.

Elucidation of the molecular biology and biochemistry of enhancer-promoter interactions has set the stage for a new era of investigation into the mechanisms of transcriptional regulation. Armed with new methodologies for genome sequencing and editing and protein engineering to both discover enhancers and to rapidly test their functions, we are well placed to achieve a more comprehensive understanding of the principles of genome architecture that modulate cellular transcriptomes.