The first attempt to comprehensively map the human Transcriptome was carried out by the Encode Pilot Project [150]. The ENCODE pilot project selected DNA from 30 Mb (~1 %) of human genome. 15 Mb was manually selected based on the presence of well-studied genes or other known sequence elements and the existence of a substantial amount of comparative sequence data. The second 15 Mb was chosen from 30 to 500 Kb randomly selected regions demonstrating a good sampling of regions with varying content of genes and other functional elements [150]. Two types of tiling arrays were constructed that spanned these regions, oligonucleotide arrays and spotted arrays [150–152]. Repetitive elements were removed by RepeatMasker [153]. The study examined RNA from 11 cell lines and 20 tissues.

The next phase of ENCODE, which was completed in 2012 [51], introduced two significant changes. Now studies were extended to the entire genome (the remaining 99 %) and use of massively parallel or next generation sequencing (NGS; see Chap. 5 of this volume) was instituted. The latter greatly increased the sensitivity and through-put analysis of fragments of DNA from such techniques as ChIP, FAIRE, and DHS [154] which were now sequenced and mapped directly to the DNA of the human genome. Techniques used to directly evaluate RNA were likewise utilized with equally dramatic increases in sensitivity and through-put [28, 155]. In addition, a new research effort, Genome Wide Association Studies (GWAS), had begun the goal of which was to find genetic variations associated with a particular disease by rapidly scanning markers across the genomes of many people [156]. These studies are supported by a public-private partnership, the Genetic Association Information Network (GAIN) that includes the NIH, the Foundation for the National Institutes of Health, Pfizer Global Research & Development and others and independently by various individual NIH institutes.

With regard to transcription and assuming that the ENCODE 30 Mb were representative of the entire genome, one might have expected approximately 2 % of the region to have been transcribed. However, the study found that over all samples 92.6 % of bases represented in the ~30 Mb ENCODE region were transcribed into primary transcripts and 24.1 % into processed transcripts [150]. This was at least partly explained by the fact that 63 % of the transcripts mapped to regions outside of those for the annotated genes of this DNA and did not appear to code for protein [150]. Alternative splicing was a common feature occurring in 86 % of multi-exon gene loci which generated >5.4 transcript variants per locus on average [157]. This was further evaluated using 5′ rapid amplification of cDNA ends (RACE) and tiling arrays for the 399 genes located entirely within the ENCODE regions using RNA derived from 12 tissues. Investigators were able to detect RNA transcribed from 359 of these loci [150, 158]. 90 % of these genes contain a novel sequence some of which extended 50–100 kb 5′ to the annotated gene’s 5′ transcription start site. The studies demonstrated that a given gene may both encode multiple protein products and produce other transcripts that include sequences from both strands and from neighboring loci often without encoding a different protein [150, 158]. Transcription start sites (TSS) were assessed using two 5′-end-capture technologies, that is, cap analysis gene expression (CAGE) in which short (∼20 nucleotide) sequence tags originating from the 5′ end of full-length mRNAs are sequenced [159] and paired-end tag (PET) sequencing of cDNA [160]. 4,491 TSSs were found from these analyses, almost 10 times more than the number of established genes which was felt to potentially explain the extensive degree of transcription previously noted [150]. These findings were refined and extended in the subsequent phase of the ENCODE project which examined the entire human genome.

Other studies to characterize the genome in conjunction to RNA transcription were performed. ChIP-chip studies sought to identify promoter and enhancer sights by first treating chromatin with restriction enzymes to fragment it, next immunoprecipitating chromatin using antibodies to polymerase II and various transcription factors and then assaying the DNA fragments on microarrays [160]. Two other techniques used to obtain open regions of DNA likely to contain active regulatory sites were DNA isolated from DNase hypersensitive sites (DHSs) and from formaldehyde-assisted isolation of regulatory elements (FAIRE) [161, 162]. In the former assay chromatin is treated with low concentrations of DNase I and the fragments separated on an agarose gel. The low molecular weight fragments were isolated, amplified, and examined on tiling arrays. In the FAIRE technique, the chromatin is treated with formaldehyde to cross-link histones to other histones and to DNA. The chromatin is then fragmented by sonication and extracted with phenyl:chloroform. The DNA fragments from the open regions of chromatin without histones separate into the aqueous phase of the extraction and the histone:histone and histone bound DNA sequester to the organic phase. The DNA fragments recovered from the aqueous phase were amplified and examined on tiling arrays. Loci identified by these techniques were presumed to represent enhancer elements (transcription factor binding sites) and were designated regulatory factor binding regions (RFBRs). 65 % of these RFBRs were located within 2.5 kb of known or novel TSSs. Further analysis indicated a relationship between tissue specificity and unique TSSs and regulatory clusters (RFBRs) that were detected in the tissue [150].

Histones represent another important class of molecules associated with the control of DNA transcription [163]. Within chromatin, DNA is tightly wrapped around a disc-shaped core of eight histone proteins—two molecules each of histones H2A, H2B, H3, and H4. Double-stranded DNA that is 146 base pairs long is wrapped around this protein core to form the nucleosome. Another histone, H1 binds to the DNA as it exits the nucleosome. Nucleosomes are arranged initially as 10 nm diameter fibers [164] with some studies suggesting higher orders of super coiling into 30 nm or greater diameter fibers [46, 163] though the view of chromatin consisting of predominantly of 30 nm fibers is currently disputed [164]. Eukaryotic cells contain large multi-subunit proteins called chromatin-modifying complexes [163, 165] that posttranslationally modify the histones within the nucleosomes to cause the chromatin to be more or less compact (see Fig. 1.3; (modified after figure in Box 1 Histone Code in Sparmann et al. [166])). Modifications include acetylation, methylation, phosphorylation, and ubiquitination which operationally function by either disrupting chromatin contacts or by affecting the recruitment of nonhistone proteins to chromatin which can orchestrate the ordered recruitment of enzyme complexes to manipulate DNA [165, 167]. The histone modifications are created by pairs of enzymes that exhibit antagonistic effects toward each specific modification [168, 169]. For example, histone acetyltransferases (HATs) acetylate lysines, while histone deacetylases (HDACs) remove the acetyl groups of lysines [168, 169]. Acetylation of histones may in turn act as docking sites to stabilize or further recruit other protein complexes including chromatin remodelers which in turn can reposition or evict nucleosomes along the DNA in an ATP-dependent fashion thus creating nucleosome-free regions on enhancer sequences [170]. The ENCODE project tested for a number of these modifications applying ChIP-chip using the well-studied histone H3 and H4 modifications histone H3 acetylation at lysine 9, 14, (H3ac), histone H4 acetylation at lysine 5, 8, 12, 16 (H4ac), and histone H3 methylation at lysine 4 mono-, di-, and trimethylation (H3K4me1, H3K4me2, H3K4me3, respectively) [171]. The resulting maps and subsequent studies have indicated clear patterns of histone modifications. Chromatin DNA locations that are high in H3K4me1 but low in H3K4me3 have proven to be highly predictive of enhancer locations [170] and H3K4me1 is a mark of active enhancers. The TSSs of genes are closely associated with H3K4me3, H3K4me2, and H3ac modifications and H3K4me3 is a mark of gene promoters. TSS patterns differed between active and inactive genes. Expressed genes had distinct peaks of H3K4me2, H3K4me3, and H3ac modification downstream from the TSS. H3K27ac is a mark of transcriptionally active regions. The histone H3K27me3 is a mark of repressed regions and is generated by the Polycomb repressive complex 2 (PRC2) [172] discussed subsequently in section “Polycomb Group (PcG) Proteins”. Another mark of repressed regions is H3K9me3 which is a repressive mark associated with constitutive heterochromatin and repetitive elements [51, 125, 173]. The histone modification profiles show differences between cell lines associated with differences in gene transcription [125, 171].

Studies from the ENCODE project published in 2012 which examined the transcriptomes in 15 cell lines [126] found that, cumulatively, 74.7 % of the human genome was covered by primary transcripts and 62.1 % by processed transcripts. As in the ENCODE pilot project, genes expressed many splice variants simultaneously with an average of about 10–12 expressed isoforms per gene per cell line. One isoform dominates in a given condition usually capturing a large fraction of the gene’s total transcripts—at least 30 %. In a related study, transcripts of 492 protein coding genes on human chromosomes 21 and 22 were analyzed for locations of their 5′ and 3′ transcriptional termini [174]. For 85 % of these genes the boundaries extended beyond the current annotated termini most often connecting with exons of transcripts from other well annotated genes [174]. This finding could potentially cause problems for molecular pathologists since chimeric transcripts might be regarded as cancer specific [175]. This particular issue was addressed recently by Greger et al. [176] who examined RNA sequencing data from cell lines prepared from 462 individuals who participated in the 1000 Genomes project [177]. They identified 81 RNA tandem chimeric transcripts from the cell lines of these normal individuals. Six chimeric transcripts were intrachromosomal fusions of genes located on different strands and 15 were interchromosomal fusions. Six fusion transcripts had been regarded as cancer-specific [176]. None of the fusion transcripts are currently used clinically but the finding of these chimeric transcripts does raise questions with regard to the issue of adequate controls.

Alternative splicing is known to influence biological outcomes such as sex determination, neural differentiation, and programed cell death and can contribute to cancer progression [178]. Xiong et al. were able to identify 20,000 unique single nucleotide variants likely to affect splicing [178, 179]. The method correlates the presence of SNVs and the presence or absence of inclusion of a specific codon in the target transcript and does not take into account the presence or absence of a disease phenotype. Nevertheless, the method was successful in identifying misspliced genes with neurodevelopmental phenotypes in individuals with autism and expressed misspliced variants of MLH1 in patients with Lynch syndrome [179].

The compilation by ENCODE of all genes and transcripts identified in the ENCODE project is referred to as GENECODE [127]. Version 21 (June 2014 freeze) contains 196,327 transcripts and 60,155 genes [180]. Of these genes 19,881 are protein coding genes, 35,758 noncoding RNA genes with the remainder consisting of pseudogenes and immunoglobulin/T-cell receptor gene segments [180]. With regard to noncoding RNA genes 15,877 are classified as long noncoding genes (lnc) which is the term used for transcripts that are not associated with protein-coding loci with a minimum size of 200 bp. As has been noted earlier, not all genes encode for proteins the most obvious and abundant being genes for rRNA and tRNAs but also snRNAs, snoRNAs, miRNAs, and piRNAs. However, transcript function has generally been defined in terms of its role in protein expression. Mudge et al. [181] now suggest that in light of the abundant number of genes for noncoding RNAs and chimeric mRNAs the definition be broadened and a functional transcript be defined “as one that makes a contribution to phenotypic complexity, regardless of the mechanism by which this occurs”. Nonfunctional transcripts would then comprise all transcripts created by biological mechanisms (as opposed to technical artifacts) for which no such “contribution to phenotypic complexity” can be as yet determined.

Cellular phenotype reflects a cell’s growth requirements and functional capabilities which the cell manages through its pool of transcribed genes and transcriptome. As noted earlier, at least for protein encoding genes, gene transcription is highly dependent on association of the gene’s promoter region with enhancer elements and their bound transcription factors [52, 56, 57]. The number of transcription factor proteins are estimated to be 1,300–1,400 based on a survey of proteins with DNA binding domains and most of these are unannotated [55]. A gene expression study using Affymetrix U-133 GeneChips, which contain probes for 873 TFs, was performed on 32 normal tissues and showed expression of 510 TFs in at least one tissue. Approximately 1/3 (161 TFs) were present in all or most tissues with similar expression levels (ubiquitous TFs) while 2/3’s were selectively expressed in a few tissues (specific TFs). 172 TFs (34 %) were completely unannotated and were distributed similarly into the “ubiquitous” (69 TFs) and “specific” (103 TFs) categories [55].

While within the cell there is a one-to-one correspondence between a transcribed gene and its DNA sequence, there are on average tens to hundreds or more binding sites for each transcription factor [51, 52]. The ENCODE project provided two separate estimates of enhancers per cell one in the range of 400,000 “regions with enhancer-like features” [51] and the second 8.4 million “distinct DNase I footprints” [52]. When TFs bind to enhancer elements they cause the chromatin to decondense making the DNA accessible to cleavage by DNase I. However, the fragment of DNA actually bound to the TF is resistant to further DNase I cleavage leaving it as a “footprint” which can be isolated and sequenced [205]. If enough adjacent DNA sequence remains in the isolated “footprint” segment then the enhancer sequence can be assigned to a specific location within the human genome [52]. Using this approach applied to DNase I cleavage libraries from 41 diverse cell types, Neph and colleagues identified collectively 45,096,726 6–40 bp footprints across all cell types which they resolved to 8.4 million distinct footprint elements, each occurring in one or more cell type [52]. The number of footprints found per cell type ranged from 434,000 to 2.3 million. The ~400,000 “regions with enhancer-like features” is a value derived from a bioinformatics Hidden Markov Model (HMM) study based on inputted chromatin features such as DNase I hypersensitive sites, FAIRE data, and histone mark (ChIP-seq) data applied to nonoverlapping 200 bp segments of genomic DNA [206]. HMM attempts to assign a state (E.g. enhancer, promoter, TSS, and repressed) based on the inputted data [207] (see [208] for a nonbiological understandable example of HMM; see [209] for a bioinformatics example). The bioinformatics model presumably has use when applied to data across broad classes of metazoa [210] but will not be further considered here. The 8.4 million footprints were reduced to 683 unique motifs of which 394 were identified in experimentally-grounded motif models in three transcription factor databases (TRANSFAC [211, 212], JASPAR [213], UniPROBE [214]). Of the 289 novel motifs, all showed features of in vivo occupancy and evolutionary constraint similar to motifs for known transcription factors and showed cell-selective occupancy patterns highly similar with well-established TFs [52]. Neph et al. used the DNase I footprint data from the 41 cell lines to construct transcription factor regulatory networks and demonstrated that these networks were highly cell type specific reemphasizing the role of transcription factors in determining cell type specificity [57]. To construct the networks 475 transcription factor genes with well-annotated recognition motifs were identified using the three previously mentioned transcription factor databases [211–214] along with all DNase I footprints within a 10 kbp interval centered on the hub gene’s transcriptional start site (i.e., the hub gene’s proximal regulatory region or promoter region). To construct the network, each TF gene (hub) was connected to every other TF gene that the hub gene appeared to regulate by virtue of the presence of hub gene’s footprint in the other TF gene’s promoter region. These connections represented the regulatory interactions (edges) of the network. 475 transcription factors theoretically have the potential for 225,625 combinations of TF-to-TF regulatory interactions or network edges. However across all cells only a total of 38,393 unique, directed TF-to-TF edges were observed with an average of 11,193 TF-to-TF edges per cell. Regulatory interactions were highly cell-selective and were most frequently restricted to a single cell type, and collectively the majority of edges were restricted to four or fewer cell types. Only 5 % of edges were common to all cell types [57]. There was good agreement between the generated or “de novo” networks with TF-to-TF circuitry of known networks. The investigators computed for each cell type a normalized network degree (NND) vector [215] to capture the degree to which different cells type networks utilize similar transcription factors and clustered the cell types based on their NND vector. The resulting network clusters arranged the cell types into groupings that paralleled both anatomical and functional phenotypic characteristics [57]. These studies further verified that differences in patterns of activation of enhancer elements drive the gene expression patterns that are responsible for cellular differentiation and developmental identity [52–54, 56, 57].

The nucleus is the largest organelle within the eukaryotic cell measuring 5–10 μm in diameter [216, 217]. It is surrounded by two phospholipid bilayer membranes. The two membranes fuse at the nuclear pores through which RNA and proteins are transported between nucleus and cytoplasm. In many cells, the outer nuclear membrane is continuous with the rough endoplasmic reticulum, and the space between the inner and outer nuclear membranes is continuous with the lumen of the rough endoplasmic reticulum [216]. Beneath the inner nuclear membrane is a layer termed the nuclear lamina which is a mesh-work of type V intermediate filament proteins called Type A and B lamins [216–218]. The human genome contains about 1,300 discrete lamina-associated domains (LADs) that range in size from 80 kb to 30 Mb and together contain thousands of genes [219, 220]. The vast majority of lamina-associated genes are transcriptionally inactive and enriched in repressive histone marks such as H3K27me3 and H3K9me2 [219, 220]. The remainder of the interior of the nucleus is occupied by chromatin comprising the 46 human chromosomes, various nuclear bodies including the nucleolus [73, 89], Cajal body [221], nuclear speckles [222], various multi-subunit chromatin-modifying complexes [163, 165], proteins and lncRNAs responsible for RNA transcription [223] and DNA replication [224], and actin filaments and monomers [225, 226]. The nuclear space not occupied by the sub-inner membrane lamins, chromatin, and various nuclear body components is generally referred to as the nucleoplasm. The question regarding whether the nucleoplasm is filled by a “nuclear matrix”, that is, by some sort of nucleus-wide arborized network of filaments extending throughout the nucleoplasm has been intensely debated for the last 15 years [227–230] with the consensus seeming to not favor its existence ([230, 231] though this view is not universal [232, 233]). One source for such a matrix might originate from the nuclear lamins. In addition to being present in the nuclear lamina region, both types of lamins are also present in the nucleoplasm [218, 234]. In the nucleoplasm type A lamins are highly mobile while type B lamins are mainly immobile [234]. This suggests that nucleoplasmic type B lamins are either assembled into some type of structure or are tightly associated with other unknown immobile structural components [218, 234]. Recently, Belin et al. examined the nucleoplasm with probes for monomeric and filamentous actin [226]. Monomeric actin was detected in nuclear speckles, globular structures enriched in pre-mRNA splicing factors, which was said to be consistent with proposed interactions between actin and RNA-processing factors [222]. Filamentous actin was present in punctate structures throughout the interchromatin space and was excluded from chromatin-rich regions. Actin filament motion was quite slow and said to be “backtracking” in a manner characteristic of particles embedded in a viscoelastic medium such as a protein-based mesh similar to the proposed nuclear “matrix” [225].

The nucleus holds two copies of the human genome. The human genome consists of approximately 3.3 billion base pairs and in solution DNA has a dimension of 0.334 nm of length for each base [235]). Thus each human haploid genome measures approximately 1.1 m in length. Two copies of this DNA must be folded and arranged within the cell’s 5–10 μm diameter nucleus in a manner that allows transcription, replication, and cellular differentiation to occur. The end points of this process, that is, the initial folding of DNA and combining with histone proteins to form a 10 nm diameter chromatin fiber of nucleosomes (Fig. 1.3 above [163, 164]) and the localization of chromosomes into chromosome territories (CTs) within the interphase nucleus (Fig. 1.4 (figure 1B in [236])) [236, 237] are well documented. However, there are multiple unanswered questions regarding the higher order organization and arrangement of chromatin fibers within CTs as well as with regard to the proximity patterns of CTs themselves [238]. Proximity patterns of CTs, that is, their radial arrangements and neighborhood arrangements within nuclei appear to vary across cell types without any clear overarching generalizations in evidence. In spherical nuclei of lymphocytes the CT for gene rich chromosome 19 is predictably located near the center of the nucleus and that for gene poor chromosome 18 located in the periphery [237]. In the flattened oval nuclei of human fibroblasts the size and other physical parameters of the chromosomes appeared to have a greater impact on the location proximity patterns between chromosome CTs [236–238]. Gene-dense and/or highly expressed sequences were found equally distributed throughout their respective territories [237]. In other studies, while a given neighborhood arrangement of CTs was stable once established at the onset of interphase, this arrangement was not maintained following metaphase in daughter cells [237, 239]. Nuclear arrangements of CTs as well as chromatin order within CTs and genetic loci have been reported to undergo major changes during cell differentiation and upon certain functional demands such as erythroid differentiation, adipogenesis, hormonal stimulus, and when proliferating cells become quiescent following serum starvation [240–242]. One striking example of altered arrangement of chromatin within nuclei occurs in the nuclei of rod photoreceptor cells of nocturnal mammals. In these cells heterochromatin is localized in the nuclear center whereas euchromatin, as well as nascent transcripts and splicing machinery, line the nuclear periphery [243]. The inverted rod nuclei act as collecting lenses, and computer simulations indicate that columns of such nuclei channel light efficiently toward the light-sensing rod outer segments [243].

CTs are postulated to be built up from chromatin domains approximately 1 Mbp in size [237, 238]. This size was chosen on the basis of early studies using autoradiography indicating that in S-phase nuclei replication foci measured 1 Mbp in size on average and that these replication domains appeared to be a constant recurring feature of the chromatin over multiple cell cycles [238, 244–246]. Evidence for a functional chromatin domain size of ~1 Mbp also comes from a study by Kolbl et al. [247]. These investigators measured the radial nuclear position in a human Burkitt lymphoma cell line of three marker genes located on the short arm of chromosome 1 but separated by at least 10 Mbp. Expression levels as measured by qPCR of the three genes were the same whereas the total expression strength (TES) calculated as the sum of the transcription of all genes annotated within a surrounding window of about 1 Mbp DNA differed for each region. Radial nuclear position of the studied regions and genes correlated with total expression strength (TES) with highest TES occupying the most interior nuclear position [247].

Chromatin within the 1 Mbp domains is assumed to undergo varying degrees of compaction that enable it to fit within the dimensions of the CT and the nucleus (Fig. 1.5). Studies by Maeshima and others suggest that this compaction is accomplished through a process of irregular folding of the 10-nm fibers [46, 164]. The exact nature of the irregular folding is unclear and has been posited to be due to macro-molecular crowding [164, 248] associated with specific proteins such as cohesion and/or codensin II [164, 249, 250] or to formation of fractal globules from repeated crumpling (fractal model) that is a form of folding in which the clumped strands of chromatin within the globules avoid becoming entangled [46, 251, 252].

The arrangement of chromatin within CTs has been postulated to follow one of three patterns ([238] and references therein). In the chromosome territory-interchromatin compartment (CT-IC) model, CTs are built up from highly folded chromatin in chromatin domains (CDs) surrounded by a perichromatin region (PR) containing less condensed (decondensed) chromatin and a nearly DNA-free interchromatin compartment (IC) between the chromatin domains. The processes of DNA transcription, DNA replication, RNA splicing, and DNA repair take place in the perichromatin region (PR). In the interchromatin network (ICN) model euchromatin is made up from chromatin from chromatin fibers, which intermingle more or less homogeneously by constrained diffusion both in the interior of CTs and between neighboring CTs. In the giant loop field (GLF) model and long-range field (LRF) model transcription occurs on giant chromatin loops which expand from the surface of CTs and form a field of intermingling loops. When transcription ceases, the giant loops collapse back into condensed core domains of CTs. Using 3D structured illumination microscopy (3D-SIM) which detects targets of emitting fluorophores with an approximately eightfold improved resolution over conventional confocal laser scanning microscopy [253], Schermelleh et al. detected the presence of channels emanating from nuclear pores and extending through the lamina and into the heterochromatin [254]. Markaki et al. studied nuclei of C127 cells derived from a mouse mammary tumor using 3D-SIM and Hela cell nuclei with spectral precision distance/position determination microscopy (SPDM). The latter allows time-resolved single-molecular localization with a localization accuracy of a few nanometers in the lateral plane of the specimen [255]. Their studies examined nuclei for newly synthesized (nascent) RNA and DNA using BrUTP to detect the DNA and ATTO 488-dUTP to visualize the RNA. Immunofluorescence was used for detection of RNA polymerase II, histone H3K4me3, and H4K8ac the latter two of which are enriched at promoter regions of genes. Their studies clearly demonstrated the pattern of chromatin domains separated by interchromatin channels consistent with the CT-IC model. Chromatin domains showed decondensed chromatin along the edges of the IC which location was also the exclusive site of Polymerase II, nascent RNA, and nascent DNA. IC varied in width across a broad range from narrow channels to “lacuna” size ≥400 nm. Narrow IC channels could be filled with decondensed chromatin representing the PRs from two closely neighboring CDs whereas the interior of IC lacuna was home to splicing speckles but chromatin (giant loops or otherwise) was absent [241].

Methylation of DNA is one epigenetic mechanism controlling gene transcription and occurs with transfer of a methyl group to the cytosine of a CpG dinucleotide [278, 279]. CpG dinucleotides are concentrated in genomic regions called CpG islands ranging in size from 200 bp to several kilobases and typically located near gene promoters [278, 280]. DNA methylation is established de novo by DNA methyltransferase (DNMT) enzymes DNMT3a and DNMT3b and maintained during DNA replication by DNMT1 [279]. It is usually associated with gene silencing [279].

Polycomb group (PcG) proteins are critical regulators of normal differentiation [281] through modification of chromatin that induces gene silencing [166, 282, 283]. PcG proteins are organized into mainly two multi-protein complexes, e.g., Polycomb Repressive Complex 1 and 2 (PRC1 and PRC2) [172]. The Polycomb complex PRC1 consists of Pc, Ph, Psc, and dRing, but many additional proteins have been found to copurify with PRC1 ([284]. PRC2 includes Eed (Esc), Suz12, RbAp48, and the catalytic subunit, Ezh2 (E(z) [284]). In Drosophila in which these proteins were first studied, PRC1 and PRC2 are recruited to genes to be silenced via association with proteins that bind to specific DNA sequences called Polycomb response elements (PREs) located next to the target genes [172, 280]. However, in mammals this recruitment mechanism fails to account for most PcG protein occupied sites in vivo and protein occupancy correlates most precisely with broad domains delineated by unmethylated CpG islands of target genes which may act as PREs in vertebrates [280]. In addition, recruitment of PcG proteins has been shown to involve lncRNAs [172, 285, 286]. The central function of PRC2 is to methylate Histone H3 on K27 with the trimethylated product (H3K27me3) widely viewed as the operative chromatin mark that accompanies PcG induced gene silencing [172]. H3K27me3 contributes to PRC1 targeting and chromatin interaction. PCR1 ubiquitylates histone H2A on K119 which plays an important role in PCR1-mediated gene silencing. PRC1 also causes compaction of chromatin and may inhibit transcription by binding to general transcription factor TFIID (TBP) thereby blocking RNA polymerase II transcription activation [172].

Posttranslational modifications of histones H2A, H2B, H3, and H4 represent another epigenetic mechanism for regulating gene expression and cell fate and have been discussed earlier in section “Histone Modifications (Marks)” (see Fig. 1.3 [166, 287]). As noted a number of histone modifications are marks of repressed chromatin such as H3K27me3 and H3K9me3. How these histone marks and other states responsible for gene silencing are transmitted through DNA replication and mitosis is still unclear but the 90 KDa nuclear protein UHRF1 may be involved. UHRF1 can recognize the hemi-methylated state of newly replicated DNA and the methylation state of H3K9 and recruit DNMT1 and H3K9 methyltransferases to methylate each molecule respectively [168].