The current definition of epigenetics relates to “the study of phenomena and mechanisms that cause chromosome-bound, heritable changes to gene expression that are not dependent on changes to DNA sequence” (Deans and Maggert, 2015). Epigenetic mechanisms actually regulate several DNA/RNA-mediated processes, including transcription, DNA repair and DNA replication, through modulation of the structure of chromatin, a macromolecular complex of DNA, RNA and proteins such as histones. Two major epigenetic mechanisms influence chromatin structure: histone modifications and DNA methylation (Allis and Jenuwein, 2016) (Figure 1).

In the nucleus, DNA is wrapped around core particles of chromatin, the nucleosomes, which are formed of octamers of histones, including H2A, H2B, H3, and H4. Histones are subject to various post-translational modifications (PTM) such as acetylation, methylation, phosphorylation and ubiquitylation. These histone modifications modulate the degree of compaction of nucleosomes, thereby affecting chromatin accessibility to various factors, particularly transcriptional regulators. (Jenuwein and Allis, 2001; Borrelli et al., 2008; Wang and Jin, 2010; Tsompana and Buck, 2014; Xu et al., 2014). Importantly, histone residues can be modified in a combinatorial, reversible, and targeted manner (Figure 1). For instance, histone acetyltransferases (HAT) and HDAC are involved in acetylation and deacetylation of specific histone residues, respectively (Figure 1). Similarly, methylases and demethylases regulate the addition and removal of methyl groups on histone residues (Figure 1). The combinatorial, reversible and targeted nature of histone modifications is the basis of the so-called ‘histone code’. It permits to achieve specificity in the outcome, through the action of proteins interpreting the code (e.g., “readers”) (Jenuwein and Allis, 2001; Bannister and Kouzarides, 2011; Jones et al., 2016). One major outcome is transcription. General rules of the “transcriptional” histone code are relatively well defined. For instance, histone acetylation, whatever the histone or the residue modified, promotes relaxed chromatin and is associated with transcriptional activation, whereas histone methylation, depending on the specific residue that is modified, can lead to transcriptional repression or activation. For instance, H3K9 methylation is associated with transcription repression, whereas H3K4 methylation correlates with transcription activation. Moreover, the genome comprises different regulatory regions, including promoters and enhancers (i.e., regulatory regions distal to promoters) playing specific roles in transcriptional regulation. These regions display specific histone modifications. For instance, promoters and enhancers of active genes are enriched in trimethylated H3K4 (H3K4me3) and in acetylated H3K27 (H3K27ac) respectively, further illustrating the targeting precision permitted by the histone code.

DNA methylation, another important epigenetic mechanism, consists in adding a methyl group to cytosines by DNA methyltransferases (DNMT), particularly at C5 position of cytosine in cytosine-guanine dinucleotide sequences (CpG), creating 5-methylcytosine (5-mC). DNMT1 is implicated in the maintenance of methylation patterns during DNA replication. In contrast, DNMT3a and DNMT3b have been associated with de novo DNA methylation (Jeltsch, 2006; Feng et al., 2010). Though DNA methylation was long considered as a stable process, it is now clear that post-mitotic cells can undergo active DNA demethylation, through a mechanism implicating TET proteins (Pastor et al., 2013). TET proteins induce hydroxylation of 5-mC (leading to 5-hmc) and are further involved in their oxidation. Oxidized 5-hmC are then processed by DNA repair mechanisms and converted back to their unmethylated state (Guo et al., 2011; Feng et al., 2015). 5-hmC is particularly extended in neurons, suggesting that regulation of DNA methylation is highly dynamic in these cells (Song et al., 2011; Szulwach et al., 2011). 5-mC are bound by several classes of methyl-binding proteins (such as MeCP2), which associate with other protein partners, including HDAC, forming co-repressor complexes (Urdinguio et al., 2009; Du et al., 2015). DNA methylation at gene promoters, generally enriched in CpG sites and forming so-called CpG islands, is an important mechanism involved in gene repression (Deaton and Bird, 2011) (Figure 1).

The collection of epigenetic modifications (e.g., the epigenome) can be assessed at genome-wide scale, using next generation sequencing-based techniques such as chromatin-immunoprecipitation-sequencing (ChIP-seq). Generation and integration of epigenomic data with transcriptomic and/or functional data to fully decode epigenomes at cell-type specific level is a major current challenge that requires the development of new techniques and analysis methods (Marconett et al., 2013; Shin et al., 2014). In particular, methods to process low cell numbers/single-cell need to be developed as well as powerful bioinformatics tools (Bock et al., 2016).

Transcriptional dysregulation plays a central role in HD pathogenesis (Seredenina and Luthi-Carter, 2012). Major transcriptional changes have been reported in the brains of HD patients (Hodges et al., 2006). This is also observed in HD mouse models, including transgenic and knock-in mice (Luthi-Carter et al., 2000; Luthi-Carter et al., 2002; Zucker et al., 2005; Roze et al., 2008). Transcriptomic studies using mice showed that transcriptional changes are progressive, CAG repeat-length-dependent and most extended in the striatum (Desplats et al., 2006; Kuhn et al., 2007; Becanovic et al., 2010; Langfelder et al., 2016). In the striatum of HD models, transcriptional changes occur in both directions: many genes are down- and up-regulated (Seredenina and Luthi-Carter, 2012; Francelle et al., 2014). Remarkably, down-regulated genes display a neuronal signature, since decreased genes in HD striatum are enriched in genes that define striatal neuron identity and function, such as Darpp32, Rgs9, Drd1 or Drd2 (Hodges et al., 2006; Kuhn et al., 2007; Vashishtha et al., 2013; Achour et al., 2015; Langfelder et al., 2016). In the other affected brain regions such as the cortex, fewer genes are down-regulated. However, they also present a neuronal signature, reflecting tissue identity (Vashishtha et al., 2013; Langfelder et al., 2016).

Interestingly, a recent study using RNAseq revealed a developmental signature associated with differentially expressed genes in post-mortem prefrontal cortex of HD patients. Specifically, Hox genes and additional homeobox genes were re-expressed in HD brains, suggesting that the transcriptome of HD neurons resemble that of immature neurons (Labadorf et al., 2015). Thus, these results support the view that the transcriptional program involved in the maintenance of neuronal identity is impaired in HD neurons. Up-regulated genes in the striatum or cortex of HD patients were also enriched in biological processes linked to inflammation (Hodges et al., 2006; Labadorf et al., 2015). However, up-regulation of immune response genes is not that clear in mouse models (Achour et al., 2015; Langfelder et al., 2016). Thus, down- and up-regulated genes in HD brain tissues display distinct functional signatures, suggesting that different mechanisms may operate.

The mechanisms underlying mutant Htt-induced transcriptional effects are unclear. However, they are thought to involve both altered regulation of transcription regulators and histone-modifying enzymes.

The hypothesis of an epigenetic origin of transcriptional dysregulation in HD, particularly implicating histone acetylation, has received increasing support over the years (Jaenisch and Bird, 2003; Lee et al., 2013; Glajch and Sadri-Vakili, 2015). The assumption that altered histone acetylation might contribute to HD was made in the early 2000s, when it was found that the HAT CREB-binding protein CBP is recruited into aggregates of mutant Htt, and that HDAC inhibitors (HDACi) improve phenotypes of drosophila and mouse models of HD (Nucifora et al., 2001; Steffan et al., 2001; Kazantsev et al., 2002). As a result, epigenetic strategies using HDACi to increase histone acetylation have been early considered for HD. Several preclinical studies have been performed using broad-spectrum HDACi (e.g., suberoxylanilide hydroxamic acid (SAHA), Trichostatin A (TSA), phenylbutyrate, sodium butyrate (NaB)] that target non-selectively HDAC of class I and II (Ferrante et al., 2003; Hockly et al., 2003; Gardian et al., 2005; Sadri-Vakili et al., 2007; Sharma and Taliyan, 2015b).

Histone deacetylase inhibitors improve some phenotypes of HD mice, including neuropathology and motor function. However, it is unclear whether beneficial effects require increased histone acetylation. Instead, studies suggest that the mechanism may involve acetylation of non-histone proteins. In support to this view, inactivation of a target of SAHA, Hdac 4, ameliorates neurodegeneration in HD mice through an apparently, transcription-independent mechanism, acting on mutant Htt aggregation process (Mielcarek et al., 2011). Non-histone-mediated beneficial effects of HDACi have also been documented in models of Parkinson disease (PD; Godena et al., 2014), suggesting common mechanisms between several neurodegenerative diseases.

New compounds have been developed in an attempt to generate more selective HDACi and with less toxic side effects (Herman et al., 2006; Thomas et al., 2008). The compound HDACi 4b, which was reported to ameliorate disease phenotype of HD mice, show high potency for inhibiting HDAC1 and HDAC3 (Thomas et al., 2008; Jia et al., 2012). However, physicochemical properties and metabolic profile of this compound were found suboptimal for investigation of HDAC inhibition in mice per oral administration (Beconi et al., 2012). The effect of RGFP966, an HDAC3-selective inhibitor, was recently investigated using HD mice (Jia et al., 2016). The results suggest that the compound limits glial cell response, diminishing markers of glial cell activation. Surprisingly however, a heterozygous inactivation of the Hdac3 gene in HD mice did not ameliorate disease-related phenotypes (Moumne et al., 2012), suggesting that more than 50% knock-down of the Hdac3 gene might be needed to see a beneficial effect. More recently, beneficial transgenerational effects have been reported using HDACi 4b in HD mice (Jia et al., 2015). Thus, despite some beneficial effects, the mode of action of HDACi in HD models remains elusive and may not systematically implicate histone- and transcription-dependent mechanisms.

It is still unclear whether expression of neuronal identity genes in HD brain tissues is restored upon HDACi treatment (Coppede, 2014; Wang et al., 2014; Sharma et al., 2015; Valor, 2015a). Clinical studies using HDACi are ongoing and the results are awaited. So far, Phase II studies provide indication for safety and tolerability of several compounds, including phenylbutyrate (Hogarth et al., 2007; Westerberg et al., 2015).

But what exactly is the status of the HD epigenome? Is acetylation the only histone modification impaired in HD affected tissues, and to which extent? Are acetylated histone residues all affected the same way by the HD mutation? Is DNA methylation also impaired? Is the chromatin structure globally altered and repressed or is it altered at specific genomic regions or gene loci? What is the underlying mechanism? What are the consequences of chromatin structures changes? Do they underlie transcriptional changes? Addressing these questions is certainly a prerequisite to the development of new epigenetic therapies for HD.

Many studies have already been performed that attempt to address these issues (Lee et al., 2013; Valor and Guiretti, 2014; Glajch and Sadri-Vakili, 2015). It is expected that the use of genome-scale approaches, which has remained so far rather limited in the HD field, will improve our knowledge of HD epigenome as well as provide insights into the mechanism responsible for altered epigenetic regulation in HD.

As mentioned above, altered histone acetylation in HD has been proposed to result from decreased activity of the HAT CBP, due to CBP recruitment into aggregates of mutant Htt, CBP depletion in neurons expressing mutant Htt and/or through an aberrant interaction of CBP with soluble mutant Htt (Steffan et al., 2000; Nucifora et al., 2001; Jiang et al., 2006; Seredenina and Luthi-Carter, 2012; Glajch and Sadri-Vakili, 2015) (Figure 4). The effect appears specific to CBP since CBP-related HAT p300 was unable to rescue cell toxicity in overexpression assays (Nucifora et al., 2001). However, although studies manipulating CBP levels in HD models support a role for CBP in pathogenesis (Steffan et al., 2000; Klevytska et al., 2010), including cognitive deficits (Giralt et al., 2012), it is still unclear whether altered CBP underlies HD neuronal transcriptional signature.

The activity of additional chromatin modulators was also found affected by Htt and/or mutant Htt. Htt was associated with polycomb repressive complex 2 (PRC2) in cell nucleus, which facilitated stimulation of PRC2 activity (Figure 4). Noticeably, CAG-expanded Htt further enhanced PRC2 activity in cell models (Seong et al., 2010). However, although H3K27me3 ChIPseq analyses revealed that Htt null mutation in ESC globally decreased H3K27me3, a result consistent with Htt-mediated stimulation of PRC2, the trend was opposite in NPC, suggesting that Htt may also be implicated in the process of H3K27me3 removal, and these different roles for Htt on H3K27me3 regulation may depend on cell differentiation state (Biagioli et al., 2015). Thus, the role of mutant Htt on genome-wide regulation of H3K27me3 needs to be further investigated, particularly in mature neurons. It is especially important to specify this issue since PRC2 deficiency in adult neurons leads to molecular, electrophysiological and behavioral phenotypes reminiscent to those seen in HD mice (von Schimmelmann et al., 2016). In fact, inactivation of PRC2 in striatal neurons resulted in re-expression of transcription factors involved in neuronal differentiation and in down-regulation of striatal identity genes, suggesting that transcriptional signatures resulting from PRC2 inactivation and from the expression of mutant Htt in the striatum share similarities (von Schimmelmann et al., 2016).

Additional enzymes modulating histone methylation were also deregulated in HD models. ESET/SETB1, a H3K9 methyltransferase, was increased in the striatum and cortex of HD R6/2 mice due to SP1/SP3-mediated transcriptional activation, which resulted in increased histone H3K9 methylation (Ryu et al., 2006) (Figure 4). Moreover, KDM5C/JARID1C, an enzyme involved in demethylation of H3K4me3, was up-regulated in the striatum and cortex of HD R6/2 mice and proposed to contribute to decreased H3K4me3 in HD brain tissues (Ng et al., 2013; Vashishtha et al., 2013) (Figure 4). Other studies further suggest that Htt and/or mutant Htt affect chromatin structure through modulation of the activity of enzymes regulating histone methylation, including H3K9 and H3K4 methylation (Ooi and Wood, 2007; Lee et al., 2013; Dietz et al., 2015).

As already mentioned, transcription factors also contribute to remodel the chromatin through the recruitment of histone modifying enzymes, including enzymes involved in histone acetylation and methylation. For instance, following binding to their cognate DNA response element, several transcription factors can interact with CBP in a stimulus-dependent manner, thereby increasing CBP concentration and histone acetylation at selective gene regulatory elements (e.g., enhancers and/or promoters) (Sterner and Berger, 2000). Thus, impaired level or activity of transcription factors in HD may also affect epigenetic regulations. Many transcription factors, including REST, SP1, TAF130, p53, were found deregulated in cells expressing the HD mutation, due to aberrant interaction with mutant Htt or altered transcriptional regulations (Nucifora et al., 2001; Bae et al., 2005; Ravache et al., 2010; Zuccato et al., 2010; Moumne et al., 2013; Langfelder et al., 2016) (Figure 3). It is tempting to speculate that some of these factors, particularly those regulating neuronal differentiation or maintenance of neuronal fate (this is for instance the case of REST), might contribute to HD epigenetic and transcriptional alterations.

Additionally, the ability of transcription factors to interact with chromatin-modifying enzymes depends in some cases on the activation of specific signaling pathways, which permits to optimize the coupling between epigenetic and transcriptional responses after stimulation. This is the case for CREB, which needs to be phosphorylated in response to activation of cAMP signaling pathway, to bind and recruit CBP on chromatin (Cardinaux et al., 2000). The Ras/MAPK/MSK1 signaling pathway is another pathway, where transcriptional and epigenetic responses are coordinated through phosphorylation events targeting transcription factors and histones (e.g., ELK1 and H3 at serine 10, respectively) (Brami-Cherrier et al., 2009; Bahrami and Drablos, 2016). These pathways, which control the dynamics of IEGs expression such as Fos, Egr1, and Arc, play a key role in the regulation of neuronal activity, including plasticity and excitability (Besnard et al., 2011; West and Greenberg, 2011; Lopez-Atalaya and Barco, 2014). Several studies suggest that these pathways are impaired in HD striatum, raising the hypothesis that the coupling between epigenetic and transcriptional mechanisms regulating neuronal activity is altered in HD (Besnard et al., 2011; Langfelder et al., 2016), which might contribute to altered epigenetic regulations (Chawla et al., 2003).

Despite these hypotheses, it remains unclear how mutant Htt leads to selective alteration of the chromatin associated with neuronal identity genes. The underlying mechanism may be the addition of direct effects of mutant Htt on the chromatin and transcription regulators described above (e.g., CBP, PRC2, ESET/SETB1, KDM5C/JARID1C, REST, SP1, CREB, TAF130, p53). Alternatively, it may be the result of an indirect mechanism caused by mutant Htt on neuronal homeostasis and activity that is yet to be identified.