Epigenetics refers to the study of heritable changes in gene expression that occur without a change in DNA sequence. Epigenetic mechanisms therefore include all transcriptional controls that determine how genes are expressed during development and differentiation, but also in individual cells responding to environmental stimuli. The purpose of this review is to examine the basic principles of epigenetic mechanisms and their contribution to human disorders with a particular focus on fragile X syndrome (FXS), the most common monogenic form of developmental cognitive impairment. FXS represents a prototype of the so-called repeat expansion disorders due to “dynamic” mutations, namely the expansion (known as “full mutation”) of a CGG repeat in the 5′UTR of the FMR1 gene. This genetic anomaly is accompanied by epigenetic modifications (mainly DNA methylation and histone deacetylation), resulting in the inactivation of the FMR1 gene. The presence of an intact FMR1 coding sequence allowed pharmacological reactivation of gene transcription, particularly through the use of the DNA demethylating agent 5′-aza-2′-deoxycytydine and/or inhibitors of histone deacetylases. These treatments suggested that DNA methylation is dominant over histone acetylation in silencing the FMR1 gene. The importance of DNA methylation in repressing FMR1 transcription is confirmed by the existence of rare unaffected males carrying unmethylated full mutations. Finally, we address the potential use of epigenetic approaches to targeted treatment of other genetic conditions. © 2013 Wiley Periodicals, Inc.
When in 1942 Conrad H. Waddington coined the term “epigenetics,” the exact nature of genes and their role in heredity and in transcriptional regulation was not known; he used it as a conceptual model of how genes might interact with their surroundings to produce a phenotype [Waddington, 1942]. The term is a portmanteau of the words epigenesis (differentiation of cells from their initial totipotent state in embryonic development) and genetics. Only in 2008 at a Cold Spring harbor meeting, consensus was reached on the definition of epigenetic trait, as “stably heritable phenotype resulting from changes in a chromosome without alterations in the DNA sequence” [Berger et al., 2009]. The greek prefix epi- refers to features that are “on top of” genetics, that is, all those stable but reversible changes to DNA, RNA and proteins that regulate gene expression and allow cells with the same DNA to do different things at different times.
Broadly speaking, epigenetic mechanisms include all transcriptional controls that regulate gene expression, whether during development and differentiation or in mature cells responding to the environment. Epigenetic changes can be inherited (e.g., imprinting) and be relatively stable (e.g., chromosome X inactivation), but are often rapidly imposed or removed on a given locus according to cell needs. Four major layers of epigenetic controls have been identified: (1) cytosine methylation; (2) histone modifications; (3) chromatin remodeling; (4) RNA transcripts. Each of these transcriptional control levels is illustrated below, but it is important to appreciate that the last layer is probably the most critical since it targets the other chromatin modifying complexes to given loci in a sequence-specific manner [Mattick et al., 2009]. In a nutshell, dynamic changes to chromatin structure influence gene expression: genes are silenced when chromatin is condensed (heterochromatin), they are expressed when chromatin is “open” (euchromatin) and the appropriate transcription factors are available. Epigenetic changes can be induced by DNA damage [O'Hagan et al., 2008; Tabish et al., 2012] but also by the environment, thus providing a connection between the outer world and DNA, that is, the genetic library where all possible repertoires of molecular responses are stored. For example, food can alter the epigenetics of rats on different diets [Burdge et al., 2011] and changes in DNA methylation may occur as a result of low dietary levels of folate, methionine, or selenium, which can have profound clinical consequences such as neural tube defects, cancer and atherosclerosis [Friso and Choi, 2002; Ulrey et al., 2005].
Tinkering with the enzymes that add or remove the chemical tags to DNA or histones is an attractive alternative to modulate gene expression, especially if effective drugs can be developed for the purpose. Epigenetics is a rapidly evolving field and offers exciting opportunities for the diagnosis and treatment of complex clinical disorders, including cancer.
DNA METHYLATION AND DEMETHYLATION
DNA methylation is the most common epigenetic marker that cells use to switch off gene transcription. McGhee and Ginder  were the first who clearly demonstrated a correlation between DNA methylation and gene transcriptional repression. Using methylation-sensitive restriction enzymes they showed that the beta-globin locus was essentially unmethylated in cells that expressed beta-globin but methylated in other cell types. DNA methylation is also a defense mechanism suppressing the expression of endogenous retroviral genes and other harmful stretches of DNA that have been incorporated into the genome of the host over time. It is an important component of embryonic development, genomic imprinting, X chromosome inactivation and preservation of chromosome stability. It is therefore not surprising that errors in DNA methylation are linked to a variety of effects, including several human diseases (e.g. imprinting defect syndromes) and cancer.
DNA methylation is achieved by the enzymatic addition of a methyl group to the carbon at position 5′ of cytosine and typically occurs within CpG dinucleotides. Between 60% and 90% of all CpGs are methylated in mammals [Ehrlich et al., 1982]. Methylation near gene promoters (with or without a CpG island) varies considerably depending on cell type, with higher promoter methylation correlating with low or no transcription [Suzuki and Bird, 2008]. Typically, unmethylated CpG pairs are located in promoters of essential “housekeeping” genes and of expressed tissue-specific genes. Methylated cytosines spontaneously deaminate into thymines over evolutionary time; hence CpG dinucleotides steadily mutate to TpG dinucleotides, which is evidenced by the under-representation of CpG dinucleotides in the human genome (they occur at only 21% of the expected frequency) [International Human Genome Sequencing Consortium, 2001]. The conversion of cytosine into 5-methylcytosine is a post-replication process and is carried out by DNA methyltransferases (DNMTs) [Bestor, 2000]. Four different DNMTs have been identified: DNMT1 (involved in the maintenance of methylation pattern during DNA replication), DNMT2 (involved in tRNA methylation), DNMT3A and DNMT3B (both required for de novo methylation) [Bestor, 2000; Dong et al., 2001].
In animal models, the absence of Dnmt enzymes is lethal, while their overexpression is linked to a variety of cancers. Mutations in DNMT1 and DNMT3B cause human diseases: hereditary sensory neuropathy (HSNE type 1E) with hearing loss and dementia (OMIM #614116) [Klein et al., 2011] and cerebellar ataxia with deafness and narcolepsy (OMIM #604121) [Winkelmann et al., 2012] are autosomal dominant conditions caused by heterozygous mutations in the DNMT1 gene on chromosome 19p13; Immunodeficiency with Centromeric instability (of chromosomes 1, 9, and 16) and Facial anomalies syndrome 1 (ICF1) is an autosomal recessive condition due to homozygous or compound heterozygous mutations in DNMT3B gene on chromosome 20q11.2 (OMIM #242860) [Xu et al., 1999].
The effects of methylation on gene expression were directly tested in an experiment performed using 5-azacytidine (5-azaC) in mouse cells. This compound is an analog of the nucleoside cytidine. When it is integrated into growing DNA strands, it inhibits the action of the DNA methyltransferases. The inhibition of DNA methylation with 5-azaC allows to compare cells before and after treatment to check the impact that loss of DNA methylation has on gene expression and on cellular differentiation [Jones and Taylor, 1980].
Until a few years ago, DNA methylation was considered a stable modification that could be removed only if DNA replication was not followed by remethylation (passive demethylation). Recently, several studies have underlined the role of active DNA demethylation, which takes place independently from DNA replication and is mediated by enzymatic reactions. This active process is involved in demethylation of the paternal pronucleus of the zygote [Mayer et al., 2000], removal of the imprinting marks in primordial germ cells [Reik et al., 2001] and demethylation during embryonic development and somatic differentiation in order to initiate tissue-specific gene expression. There are two proposed mechanisms of active DNA demethylation: direct removal of 5-methylcytosine and removal of 5-methylcytosine via further modified cytosine bases. The first mechanism has a quite low thermodynamic probability, while the second one foresees several enzymatic steps including base excision repair (BER). The first step is the oxidation of the methyl group, which generates 5-hydroxymethylcytosine. This modified base can be either deaminated by AID/Apobec enzymes to give 5-hydroxymethyluracil [Guo et al., 2011] or oxidized into 5-formylcytosine and 5-carboxylcytosine by TET enzymes [Ito et al., 2011]. Both the deamination and the oxidation products are repaired by thymine DNA glycosylase (TDG) enzyme, a glycosylase which is involved in BER [Cortellino et al., 2011; Dalton and Bellacosa, 2012]. A schematic outline of DNA methylation and demethylation mechanisms is reported in Figure 1.
Histones are nuclear basic proteins which bind to DNA and form the structural units of chromatin, known as nucleosomes. Histones H2A, H2B, H3, and H4 form the central core of a nucleosome and 147 bp of DNA wrap around it, while histone H1 is the linker which binds the nucleosome at the DNA entry and exit sites [Luger et al., 1997]. Histones have an active role in several biological processes such as gene regulation, DNA repair [Rogakou et al., 1998], chromosome condensation (mitosis) and spermatogenesis (meiosis). This latter role is possible thanks to post-translational modifications, each of which is mediated by specific enzymes and alters the histone interaction with DNA and nuclear proteins. The H3 and H4 histones have long N-terminal tails protruding from the nucleosome, which can be covalently modified at several aminoacidic positions. Modifications of the histone tails are reversible (there are enzymes that add the extra group and other that remove it) and include methylation, acetylation, phosphorylation, ubiquitination, SUMOylation, citrullination, and ADP-ribosylation. Combinations of these epigenetic changes work as a “histone code” that is “written” by specific enzymes (e.g., histone acetyltransferases or HATs and histone deacetylases or HDACs) “read” by effector proteins recognizing these subtle changes [Strahl and Allis, 2000]. For example, the addition of up to three methyl groups to lysine has little effect on the histone chemistry since lysine charge is unchanged and a minimal number of atoms is added so that steric interactions are mostly unaffected. However, numerous nuclear proteins recognise lysine methylation with high sensitivity to the extent that, for example, mono-, di-, and tri-methylation of lysine 20 on histone 4 (H4K20) appear to have different “meanings.” Contrariwise, addition of an acetyl group has a major chemical effect on lysine as it neutralizes the positive charge. This reduces electrostatic attraction between histones and the negatively charged DNA backbone, loosening the chromatin structure; highly acetylated histones form more accessible chromatin and tend to be associated with active transcription. Lysine acetylation appears to be less precise in meaning than methylation, in that histone acetyltransferases tend to act on more than one lysine; presumably this reflects the need to alter multiple lysines to have a significant effect on chromatin structure.
Some examples of histone modifications relevant to transcriptional regulation are reported in Table I. In synthesis, tri-methylation of lysine 4 on histone 3 (H3K4Me3) at the promoter of active genes and tri-methylation of lysine 36 on histone 3 (H3K36Me3) in the body of active genes are histones modifications typically associated with actively transcribed genes (euchromatic configuration) [Strahl et al., 2002; Bernstein et al., 2005]; while tri-methylation of lysine 27 on histone 3 (H3K27Me3) and di- and tri-methylation of lysine 9 on histone 3 (H3K9Me2/3) are particularly associated with repressed genes (heterochromatic configuration) [Rea et al., 2000; Cao et al., 2002]. Based on the “histone code” specific for a particular gene or chromosome region, it might be possible to target a specific enzyme to revert a given histone modification. Otherwise, global histone acetylation levels can be increased, for example, with HDAC inhibitors.
|Type of modification||H3K4||H3K9||H3K14||H3K27||H3K36||H3K20||H2BK5|
Histone modifications go hand in hand with DNA methylation to ensure that a given DNA region is either accessible for transcription or silenced by chromatin condensation. Briefly, active chromatin regions have unmethylated DNA and high levels of acetylated histones, whereas inactive regions of chromatin contain methylated DNA and deacetylated histones. Methylated DNA can be recognized by several methyl DNA-binding proteins (such as MBD2 and MeCP2). These latter proteins can also bind histone deacetylases (HDACs), which remove the acetyl group from histone tails resulting in an heterochromatic configuration [Nan et al., 1998]. Mutations in the X-linked gene MECP2 typically cause a neurodevelopmental disorder that occurs almost exclusively in females, Rett syndrome (OMIM #312750) [Amir et al., 1999].
Chromatin remodeling is the enzyme-assisted process that facilitates access of nucleosomal DNA by remodeling the structure, composition and positioning of nucleosomes. Access to nucleosomal DNA is controlled by two major classes of protein complexes: covalent histone-modifying complexes (see the previous paragraph) and ATP-dependent chromatin remodeling complexes. The ATP-dependent proteins have a common ATPase domain and use energy from ATP hydrolysis to reposition (slide, twist or loop) nucleosomes along the DNA, expel histones away from DNA or facilitate exchange of histone variants, thus creating nucleosome-free regions of DNA for gene activation. Also, several remodelers have DNA-translocation activity to carry out specific remodeling tasks [Saha et al., 2006]. Each remodeler has a unique protein domain (e.g., helicase, bromodomain, etc.) that confers a specific biological function (apoptosis, DNA repair, regulation of gene expression, etc.). Chromatin remodelers are involved in several genetic diseases, for example, X-linked nuclear protein (XNP) and chromodomain helicase DNA-binding protein 7 (CHD7), which are mutated in ATRX syndrome (α-thalassemia X-linked mental retardation) (OMIM #301040) and CHARGE association (OMIM #214800), respectively [Gibbons et al., 1995; Lalani et al., 2006].
Non-coding RNAs (ncRNAs) appear to play a critical role in regulating the transcriptional status of the genome, both in the nucleus and in the cytoplasm [Castel and Martienssen, 2013], and cannot be considered just as “transcriptional noise” [Mattick et al., 2009; Morris, 2009]. Three classes of short ncRNAs must be mentioned: short interfering RNAs (siRNAs), microRNAs (miRNAs) and PIWI-interacting RNAs (piRNAs).
The RNA interference (RNAi) pathway, initially characterized in plants and in C. elegans [Fire et al., 1998], acts via siRNAs. These can be generated by processing longer double-stranded RNAs formed by complementary pairing of mRNAs and long antisense transcripts due to bidirectional transcription at the same locus [Morris, 2009]. The presence of long ncRNAs overlapping (at least in part) critical mRNAs has been proven in several imprinted loci such as those of Angelman and Beckwith–Wiedemann syndromes [Rougeulle et al., 1998; Lee et al., 1999] as well as those of Fragile X syndrome, spinocerebellar ataxia type 8 and Friedreich ataxia [Ladd et al., 2007; Chen et al., 2008; De Biase et al., 2009]. It is assumed that double-stranded RNAs are processed by the RNAi machinery into siRNAs that induce either degradation of complementary mRNAs (post-transcriptional gene silencing) or cause RNA-mediated DNA methylation (transcriptional gene silencing). Transcriptional gene silencing involves the recruitment of chromatin modifying complexes and DNA methyltransferases to specific loci by RNAs during differentiation and development [Mattick et al., 2009].
A similar regulatory role is played by miRNAs, short endogenous ncRNAs that adopt a hairpin conformation due to internal base complementarity, which regulate a large variety of biological functions [Wang et al., 2012]. So far, approximately 2000 miRNAs have been discovered in humans (www.mirbase.org) and each of them targets hundreds of messenger RNAs that are usually downregulated. Most of the downregulation of mRNAs occurs by inducing the degradation of the complementary mRNA, though some targeted mRNAs are simply prevented from being translated into proteins. It appears that about 60% of human protein-coding genes are regulated by miRNAs [Wang et al., 2012]. DICER1 is an RNase endonuclease essential in the production of miRNAs, as it cuts pre-miRNAs in the cytoplasm, generating double-stranded miRNA duplexes, eventually unwound into the active miRNAs. Recently, constitutional DICER1 haploinsufficiency has been shown to cause predisposition to a range of rare tumours (pleuropulmonary blastoma, cystic nephroma, Sertoli-Leydig cell tumors, and embryonal rhabdomyosarcoma), confirming the fundamental role of ncRNAs in the regulation of cell differentiation and proliferation [Slade et al., 2011; Doros et al., 2012].
A third class of regulatory ncRNAs includes piRNAs [Peng and Lin, 2013]. Initially identified as transcriptional suppressors of transposons in sperm cells, now they have been identified also in non-dividing neurons where they play an important role in the epigenetic control of memory-related synaptic plasticity [Rajasethupathy et al., 2012].
Long ncRNAs are often at least partly overlapping to protein-coding mRNAs (antisense transcripts) and modulate transcription of single genes or local chromosomal domains [Morris, 2009] with some notable exceptions: the 19-kb long XIST transcript is found on the entire length of the inactive X chromosome and silences transcription of the majority of X-linked genes. Its role is confirmed by the observations that deletion of the XIST gene in Xq13 prevents X-inactivation [Ørstavik et al., 2007], while certain mutations in its promoter cause preferential X-inactivation [Plenge et al., 1997]. Furthermore, recent experiments show that addition of a XIST transgene on chromosome 21 causes the cis inactivation of the entire hosting chromosome and even the formation of a “Barr body” composed by a condensed chromosome 21 [Jiang et al., 2013].
EPIGENETIC CONDITIONS: GLOBAL VERSUS LOCAL
Genetic and epigenetic changes are always intertwined. However we can make a distinction between those conditions where the epigenetic abnormality (epimutation) is global and those where it is local.
In the first case mutations in genes encoding for components of the epigenetic machinery result in genome-wide epigenetic effects, affecting many genes and altering the transcriptional profile of the cell. Several examples of these “epigenetic syndromes” have been recently reviewed by Berdasco and Esteller  and include Rett syndrome (caused by loss-of-function mutations in the MECP2 gene, encoding a methyl-DNA binding protein) and Rubinstein–Taybi syndrome (often due to deletions of the CREBBP gene, encoding the transcriptional coactivator CREB-binding protein).
In the second case the epimutation is localized to a specific gene or chromosomal region as in imprinting disorders, where microdeletions or mutations in the “imprinting centers” cause a local disturbance of gene transcription, resulting, for example, in Prader–Willi, Angelman or Beckwith–Wiedemann syndromes [Buiting, 2010; Choufani et al., 2010; Gurrieri et al., 2013]. As discussed below, also the hypermethylation of the FMR1 promoter is a local epigenetic defect, due to the expansion of a CGG repeat (genetic defect), and causing silencing of one gene. The same sequence of events (CGG repeat expansion, localized DNA methylation and silencing of a single gene) has been identified in several other loci and is always associated with the cytogenetic expression of a “folate-sensitive” fragile site. In fact, the expanded and hypermethylated CGG sequence is replicating so late that (under particular culture conditions) it is not duplicated at mitosis and cannot condense at metaphase. All these fragile sites (including FRAXA, FRAXE, FRAXF, FRA10A, FRA11A, FRA11B, FRA12A, and FRA16A) are located in promoters and their expression associates with silencing of the respective genes [Oberlé et al., 1991; Knight et al., 1993; Ritchie et al., 1994; Nancarrow et al., 1995; Jones et al., 2000; Sarafidou et al., 2004; Debacker et al., 2007; Winnepenninckx et al., 2007].
FRAGILE X SYNDROME
Fragile X syndrome (FXS; OMIM #300624) is the most common monogenic cause of developmental cognitive impairment with an estimated prevalence of around 1:4,000 males and 1:6,000 females [Crawford et al., 1999]. The clinical phenotype has been summarized in detail elsewhere [Pirozzi et al., 2011]. FXS is caused by loss-of-function mutation of FMR1 gene, located in Xq27.3, which corresponds to the position of the folate-sensitive fragile site FRAXA observed cytogenetically in affected males [Lubs, 1969]. The gene contains 17 exons spanning 38 kb with a CpG island and an unstable CGG repeat sequence in the upstream promoter region [Verkerk et al., 1991; Eichler et al., 1993] and it encodes an RNA-binding protein. In normal population this triplet is composed of 5–55 repeats, allowing transcription and translation of the gene; within this size range the gene is transmitted stably over generations. When the CGG repeat expands between 56 and 200 (premutation), the gene continues to transcribe (more) messenger RNA, although the premutated repeat becomes meiotically unstable [Fernandez-Carvajal et al., 2009]. The presence of AGG interruptions within the repeat region seems to stabilize the trinucleotide repeat [Eichler et al., 1993], while their absence results in increased size variability upon meiotic transmissions [Nolin et al., 2013]. Although premutation carriers do not have FXS, the significantly increased transcription of FMR1 mRNA [Primerano et al., 2002] sometimes results in a gain-of-function phenotype associated with early menopause in women [Allingham-Hawkins et al., 1999] and with a tremor-ataxia neurodegenerative syndrome (FXTAS; OMIM #300623) in males [Hagerman et al., 2001].
Premutated alleles can expand to over 200 repeats (full mutation) when maternally transmitted, thus causing transcriptional repression of FMR1 through epigenetic modifications, namely: cytosine methylation of the expanded sequence and of the CpG island, deacetylation of histones 3 and 4, demethethylation of lysine 4 on histone 3 (H3K4), methylation of lysine 9 on histone 3 (H3K9) and tri-methylation of lysine 27 on histone 3 (H3K27). Kumari and Usdin  also reported increased methylation of lysine 20 on histone 4 (H4K20) near the CGG expansion. All these epigenetic changes result in heterochromatinization of the FMR1 locus [Coffee et al., 2002; Tabolacci et al., 2005, 2008a], preventing transcription and resulting in absence of the FMRP protein. An exhaustive overview of FMRP functions is provided by Bagni and Oostra  in this same issue. Very few FXS patients have deletions or point mutations of the FMR1 gene [DeBoulle et al., 1993; Collins et al., 2010], resulting anyhow in FMRP loss-of-function.
Finally, a rare class of FMR1 alleles is represented by unmethylated full mutations (UFM) that have been exceptionally reported in phenotypically normal males who have a CGG expansion over 200 repeats completely devoid of cytosine methylation [Smeets et al., 1995; Pietrobono et al., 2005; Tabolacci et al., 2008a]. The characterization of cell lines derived from these individuals has revealed that FMR1 promoter DNA is completely unmethylated (in both the CGG expansion and the FMR1 CpG island), transcription is increased (as in premutation carriers), FMRP levels are approximately 30–40% compared to normal (due to ribosome stalling on the expanded FMR1 mRNA [Feng et al., 1995]) and histone marks are similar to those of euchromatin (acetylation of histones 3 and 4, methylation of lysine 4 on histone 3 and di-methylation of lysine 27 on histone 3) with the exception for partial methylation of lysine 9 of histone 3 (H3K9). The main epigenetic status of normal, FXS and UFM alleles is summarized in Figure 2. As discussed by Pietrobono et al. , these rare individuals suggest that the CGG expansion somehow causes a cascade of events (deacetylation of histones 3 and 4, methylation of H3K9, DNA methylation and eventually demethylation of H3K4) but, for unknown reasons, UFM carriers cannot complete the silencing process and maintain an active FMR1 locus.
The exact timing in which DNA methylation is established is also unknown. An immunohistochemical study on chorionic villi samples (CVS) demonstrated that FMRP is completely absent at 12.5 weeks of gestation, suggesting that FMR1 inactivation occurs at the end of the first trimester [Willemsen et al., 2002]. FMR1 remained inactive in iPS cells derived from FXS fibroblasts, retaining DNA methylation and histone modifications typical of inactive heterochromatin [Urbach et al., 2010]. Conversely, the FMR1 gene was expressed in embryonic stem cells (ESC) from a human FXS embryo, and underwent transcriptional silencing after ESC differentiation [Eiges et al., 2007]. Recently Sheridan et al.  observed morphological differences in iPS-derived neurons, with FXS cells having fewer and shorter neurites than controls, like in Fmr1 knockout mice [Huber et al., 2002] and in human post-mortem brains [Irwin et al., 2000]. Neuronal cells are crucial to investigate the effect of drugs acting on either the epigenetic status of FMR1 (5-azadC) or downstream pharmacological targets, like the mGluR pathway. Bar-Nur et al.  showed that 5-azaC was able to reactivate FMR1 in FXS-iPS cells and in their neuronal derivatives. A recent genome-wide methylation analysis revealed that aberrant methylation in FXS is specific to the FMR1 locus and no other differentially methylated loci were found throughout the rest of the genome [Alisch et al., 2013]. This methylated region extends over a 5.5 kb stretch upstream of the CGG repeat, in which a “methylation boundary” has been described [Naumann et al., 2009]. This boundary is lost in FXS, where spreading of DNA methylation over the entire FMR1 locus is observed [Naumann et al., 2009]. CTCF (CCCTC-binding factor), the first insulator protein found in mammals, is among the nuclear proteins binding the methylation boundary. CTCF also binds to the promoter and intron 2 of the FMR1 gene [Ladd et al., 2007]. We recently investigated the role of CTCF on FMR1 transcription and observed that CTCF depletion did not cause spreading of methylation over the FMR1 promoter of UFM cell lines, in spite of the presence of the CGG expansion in the full mutation range [Lanni et al., 2013]. We suggested that CTCF may stabilize a chromatin loop between sense and antisense transcriptional regulatory regions (located in the promoter and intron 2, respectively). In fact, coordinated bidirectional transcription has been demonstrated at the active FMR1 locus, overlapping the CGG repeats and exon 1 of the gene [Ladd et al., 2007]. Similar to FMR1, the antisense transcript (ASFMR1) is upregulated in individuals with premutations and is not expressed in full mutation alleles. We also demonstrated the presence of ASFMR1 in UFM males and its alternative splicing similar to that detected in premutation carriers [Lanni et al., 2013]. Although the exact role of the antisense transcript in the epigenetic balance of the FMR1 locus has to be elucidated, it is likely that ASFMR1 contributes to the premutation phenotype, but it may also play a key role in triggering transcriptional inactivation of full mutations since it extensively overlaps with FMR1 mRNA and is upregulated in the expanded alleles [Lanni et al., 2013].
TRANSCRIPTIONAL THERAPY OF FRAGILE X AND OTHER “EPIGENETIC” CONDITIONS
With the increasing knowledge accumulating on the many functions of the FMRP protein, especially at the synaptic level, numerous pharmacological trials have been performed to try and compensate for the altered function of specific neuronal receptors [reviewed in Bagni and Oostra, 2013]. One of the most advanced clinical trials used a metabotropic glutamate receptor antagonist with promising results [Jacquemont et al., 2011]. However, considering the large number of mRNAs targeted by FMRP and the various known dysregulated pathways, including the GABAergic pathway [reviewed in Bagni and Oostra, 2013], it would be better if one could turn on again the FMR1 gene silenced by the full mutation. In fact, the existence of rare UFM individuals, known since the report of Smeets et al. , suggested us the possibility of pharmacologically restoring FMR1 transcription. DNA demethylation can be obtained with 5′-azaC or, more efficiently, with 5′-aza-2′-deoxycytidine (5-azadC) that is incorporated as analog of deoxycitidine during cell replication and irreversibly blocks DNA methyltransferases [Jackson-Grusby et al., 1997]. In 1998 we first achieved in vitro reactivation of the FMR1 full mutation by treating fragile X lymphoblastoid cells with 5-azadC [Chiurazzi et al., 1998], detecting mRNA by RT-PCR and FMRP protein by immunocytochemistry in a fraction of treated cells (5–10%). The lower efficiency of mRNA translation, due to the CGG expansion [Feng et al., 1995], probably accounts for the apparent discrepancy between mRNA and protein levels. One year later [Chiurazzi et al., 1999] we combined 5-azadC treatment with various histone deacetylase (HDAC) inhibitors (butyrate, phenylbutyrate and trichostatin A) and observed a synergistic effect on FMR1 reactivation by semiquantitative RT-PCR. However, HDAC inhibitors alone were unable to induce any reactivation at all [Chiurazzi et al., 1999], suggesting that DNA methylation is dominant over histone hypoacetylation at the FMR1 locus, as also reported for other heavily methylated genes [Cameron et al., 1999]. These initial observations were confirmed by measuring FMR1 transcript with real-time RT-PCR and confirming DNA demethylation at the FMR1 promoter after 5-azadC treatment [Pietrobono et al., 2002]. We then investigated histone modifications in the promoter, exon 1 and exon 16 of FMR1 by chromatin immunoprecipitation (ChIP) before and after pharmacological treatment of FXS lymphoblasts with 5-azadC and acetyl-L-carnitine [Tabolacci et al., 2005]. This latter compound can efficiently increase histone acetylation, but is not sufficient for FMR1 reactivation when used alone [Tabolacci et al., 2005]. It is worth noting that, although acetylcarnitine does not reactivate FMR1 full mutations, it does have a significant clinical effect on fragile X patients and improves their adaptive and social behavior [Torrioli et al., 2008]. We also tried valproic acid (VPA), since it had been reported to increase histone acetylation and to demethylate DNA, but again only modest reactivation was obtained with apparently no DNA demethylation [Tabolacci et al., 2008b]. On the contrary, 5-azadC induced both histone acetylation and increased methylation of lysine 4 of histone H3 (H3-K4), while partly reducing methylation of lysine 9 of histone H3 (H3-K9) [Tabolacci et al., 2005]. These epigenetic changes, induced by 5-azadC, appeared to restore a euchromatic configuration of the FMR1 promoter in treated FXS cells (see Fig. 2), effectively transforming a methylated full mutation into an unmethylated full mutation (UFM) [Pietrobono et al., 2005; Tabolacci et al., 2008a]. 5-azadC effects increased with dose and treatment time [Chiurazzi et al., 1999; Pietrobono et al., 2002], but 4 weeks after treatment was discontinued we observed regain of DNA methylation in the FMR1 promoter and loss of transcription [Pietrobono et al., 2002]. Further follow-up experiments are warranted in order to understand if some clones actually remained demethylated or if all cells reverted to the heterochromatic status.
Ten years ago we first discussed the potential of pharmacological intervention in order to reactivate silenced genes or to increase the expression of specific genes in order to treat epigenetic disorders [Chiurazzi and Neri, 2003]. Some genes may simply require histone hyperacetylation to be turned on or to upregulate their transcriptional output. This the case of the ALDPL1 [Kemp et al., 1998] and SMN2 genes [Chang et al., 2001; Andreassi et al., 2004], involved in adrenoleukodystrophy and spinal muscular atrophy, respectively. Transcriptional activation could partially compensate for disease causing mutations in these two genes. Other genes that, like the fully mutated FMR1, have a methylated CpG island, require also DNA demethylation in order to resume transcription. We have actually shown that 5-azadC treatment is also effective in turning on the FAM11A gene associated with the Xq28 folate-sensitive fragile site, that is transcriptionally silent in FRAXF full mutations [Shaw et al., 2002]. The same effect might be expected for the other fragile sites due to a CGG repeat expansion followed by DNA methylation (e.g., FRAXE, FRA16A, etc.) mentioned earlier.
An obvious concern that arises when considering the clinical use of 5-azadC is its toxicity. In fact, while 5-azaC and 5-azadC are generally well tolerated in hematological malignancies [Gnyszka et al., 2013], the effects of a long-term treatment are unknown. A second obstacle is the apparent requirement for cell division in order to be effective; interestingly, at least two reports suggest that 5-azadC may require minimal or no incorporation in DNA to effectively reduce levels of the maintenance DNA methyltransferase DNMT1 [Ghoshal et al., 2005; Patel et al., 2010]. But the major objection to using drugs like 5-azadC or HDAC inhibitors is that their action is likely to be unspecific and genome-wide: however, a microarray screening of 10,814 genes by Suzuki et al.  showed that a very limited set of genes are actually transcriptionally upregulated by treatment with 5-azadC (51 genes) and/or trichostatin A (23 genes). Furthermore, our unpublished observations suggest that 5-azadC induces DNA demethylation only in limited regions and typically “respects,” for example differentially methylated regions of imprinted genes, as if certain regions were more susceptible to reactivation. However, in certain cases a genome-wide effect would be actually desirable as in cancer treatments [Gnyszka et al., 2013] or in treatment of polyglutamine expansion disorders, where global histone hypoacetylation seems to play a prominent role [Steffan et al., 2001; Bodai et al., 2012; Pirooznia and Elefant, 2013; Valor et al., 2013].
However, the ultimate therapy for “local” epigenetic disorders such as FXS or imprinting disorders may not rely on small molecules such as DNMT or HDAC inhibitors but on small RNAs designed for a targeted modulation of gene silencing or reactivation [Ackley et al., 2013]. In fact, if sufficient knowledge on the specific epimutation, efficient siRNAs and delivery vectors were available, siRNA-based therapeutics may eventually become a reality [Angart et al., 2013].
Basic research on the mechanisms of transcriptional regulation of the FMR1 gene is revealing the intricacies of epigenetic events leading to heterochromatin formation that include DNA methylation, histone modifications and RNA interference, possibly secondary to the interaction between FMR1 mRNA and its antisense transcript. Fragile X syndrome therefore represents an instructive model to study potential ways of modulating chromatin activation, eventually leading to a “transcriptional therapy.” Such a therapy may become available for those genetic disorders that are not due only to (genetic) DNA mutations but also to the (epigenetic) dysregulated expression of genes.
This review is dedicated to Giovanni Neri, our common (uncommon) mentor and guide in the field of genetics and a pioneer in research on fragile X syndrome. We thank him for his continuing support, his scientific insight and his enthusiasm while striving toward a cure for fragile X patients. We also gratefully acknowledge our financial supporters: Fondazione Telethon, FRAXA Foundation and the “Associazione Italiana Sindrome X Fragile ONLUS.”