The Chromosome Conformation Capture (3C) techniques
All 3C-based techniques rely on the use of a crosslinking agent to preserve the native three-dimensional structure of chromatin inside the nucleus, which is then subject to a restriction digestion followed by DNA ligation of fragments that are in close spatial proximity (Denker and de Laat, 2016). This results in the formation of a 3C library, which is a pool of hybrid DNA fragments between sequences that could be very distant from each other on the linear genomic scale. By interrogating and quantifying these fragments using the different techniques described below, interaction profiles can be generated for specific regions of interest, reflecting their spatial organisation. These techniques are mostly performed on a population of cells and therefore provide average-based information. 3C (one-vs-one) – interrogates a single pair of genomic loci, generally using qPCR primers to quantify their ligation (or ‘interaction’) frequency in the 3C library. 4C (one-vs-all) – interrogates interactions between one chosen locus and all other genomic loci. The 3C library is subject to a second digestion and ligation step, followed by an inverse PCR to amplify the unknown sequence ligated to the sequence of interest. 5C (many-vs-many) – interrogates all possible interactions within a region of interest (up to few megabases). A pool of oligos covering the region of interest and coupled to universal primers are hybridised to the 3C library. Ligation of oligos annealed to hybrid DNA fragments followed by deep-sequencing allows detection and quantification of the interaction events.
Hi-C (all-vs-all) – interrogates all possible interactions in the genome. During 3C library preparation, digested DNA fragments are labelled with biotin, and then ligated and sonicated. Streptavidin pulldown is used to enrich for ligation events and the enriched 3C library is coupled to universal primers and deep-sequenced. Other techniques have been developed based on these, including single-cell Hi-C, to investigate interactions in individual cells (Nagano et al., 2013) and ChIA-PET, which couples chromatin immunoprecipitation (ChIP) to Hi-C to interrogate interactions mediated by a protein of interest (Fullwood et al., 2009).
Topologically Associating Domains (TADs)
Hi-C and 5C techniques have revealed that chromosomes of a wide range of species, from bacteria to plants and humans, are organised in domains (Dekker and Heard, 2015). These domains can be identified at different scales in the hierarchical folding of chromosomes. At the multi-megabase scale, for example, domains are referred to as ‘A’ and ‘B’ compartments and correspond to the association of regions of active or inactive chromatin (Lieberman-Aiden et al., 2009). These compartments vary across cell types of the same species, reflecting their transcriptional status, and can be further divided at the sub-megabase scale into smaller domains, often cell-type invariant and called TADs (Dekker and Heard, 2015; Dixon et al., 2012; Nora et al., 2012). TADs seem to represent a functionally privileged scale among the continuum spectrum of hierarchical insulation domains that constitute chromosomes, at which optimal gene co-regulation can occur (Zhan et al., 2017). Increasing evidence supports the idea that TADs provide a structural basis for regulatory landscapes, modulating the communication between gene promoters and regulatory elements such as enhancers. TADs can on one hand allow promoters and enhancers to overcome large genomic distances and engage in frequent long-range contacts within the same domain, but also prevent ectopic, deleterious interactions between different domains (Lupiáñez et al., 2015; Symmons et al., 2016). The detailed molecular mechanisms underlying the formation and maintenance of these domains remain largely unknown but relies on the architectural protein CTCF (Nora et al., 2017).
THE TOPOLOGICAL LANDSCAPE OF THE Xic
The Xic has been historically defined as the minimal genetic region that is necessary and sufficient to trigger XCI (Brown et al., 1991a; Rastan, 1983; Rastan and Brown, 1990; Rastan and Robertson, 1985). On one hand the Xic responds to the levels of the pluripotency factors and XCI-activators, and on the other hand acts through its cis-regulatory landscape, altogether creating a window of opportunity for Xist activation in XX cells (Augui et al., 2011; Galupa and Heard, 2015). The full extent of this master regulatory locus remains unknown. Single-copy transgenes carrying Xist and surrounding regulatory neighbourhood up to 470kb fail to upregulate Xist upon differentiation in XX mESCs, where the necessary trans-acting environment is present (Heard et al., 1999). Crucial cis-acting elements are therefore missing from the transgenes tested. Recently, chromosome conformation capture techniques have provided new insights into this question. One of the seminal studies that led to the discovery of topologically associating domains (TADs) used 5C to characterise a 4.5Mb region centred on Xist (Nora et al., 2012). This analysis revealed that the Xist locus lies at the boundary between two TADs (Figure 1). Increasing evidence shows that TADs provide a structural basis for regulatory landscapes, allowing promoters and enhancers to overcome genomic distances and engage in long-range contacts more frequently (Lupiáñez et al., 2015; Symmons et al., 2016). Interestingly, while one TAD includes the promoter of Xist and its activators, the adjacent TAD includes its known negative cisregulators Tsix (Lee and Lu, 1999) and Xite (Ogawa and Lee, 2003). These two TADs could therefore represent the minimal Xic, a region of at least ~850kb, comprising all the required cis-acting elements for the timely and efficient upregulation of Xist.
The human XIC, however, shows a different topological landscape, with a TAD encompassing the XIST promoter and a boundary at the XIST locus as well, but no obvious adjacent TAD (Figure 1). The absence of a specific TAD harbouring the promoter of TSIX could be related to the fact that its function does not seem conserved in human (Migeon et al., 2001, 2002). In mouse, Tsix transcription blocks Xist upregulation in cis (Luikenhuis et al., 2001; Stavropoulos et al., 2001) and its heterozygous deletion in mESCs leads to complete non-random XCI upon differentiation, with the mutant allele always being inactivated (Lee and Lu, 1999). Homozygous deletion of Tsix leads to a “chaotic” pattern of Xist expression in differentiated cells, with a variable number of inactive X chromosomes (Lee, 2005). The evolution of Tsix in mouse might thus be related to its strict monoallelic Xist regulation. In both human and rabbit embryos, in which Tsix is either not functional or not present, respectively, biallelic Xist RNA clouds are common (Okamoto et al., 2011).
ESTABLISHING THE INACTIVE STATE
Once Xist is stably upregulated from only one X-chromosome, its lncRNA accumulates in cis and triggers a series of events that will ultimately lead to chromosome-wide inactivation and its maintenance. The silencing of genes is accompanied by a reorganisation of the architecture of the X-chromosome (see below), modification of histone tails (Chaumeil et al., 2002; Okamoto et al., 2004), recruitment of repressive complexes, incorporation of the histone variant macro-H2A (Costanzi et al., 2000; Mermoud et al., 1999), deposition of DNA methylation on CpG islands (Norris et al., 1991) and a shift to asynchronous replication timing (Takagi et al., 1982). All these features of the inactive X depend on the initial actions of the Xist RNA, but once the silent state is stabilised by epigenetic mechanisms (such as DNA methylation), the presence of Xist seems no longer required (Brown and Willard, 1994; Csankovszki et al., 1999; Wutz and Jaenisch, 2000). This might not be the case in all cell types and tissues, as suggested by a study in which Xist deletion from the hematopoietic compartment in adult mice was followed by female-specific pathologies associated with X-reactivation (Yildirim et al., 2013).
Formation of a Xist-induced repressive compartment
The exact order of events triggered by the Xist RNA and their causal relationships are still being determined. The earliest event reported so far upon accumulation of Xist on the X-chromosome is the formation of a repressive compartment from which RNA polymerase II and associated transcription factors are excluded (Chaumeil et al., 2006). During the inactivation process, genes are gradually recruited into this compartment. Initially located at the periphery of the Xist RNA domain, and actively transcribed, genes subject to XCI are found in more internal positions within the domain at later stages of differentiation, when they are no longer expressed (Chaumeil et al., 2008). Exceptions to this are genes that escape gene silencing (known as ‘escapees’, see below) and the Xic (from which Xist is expressed), which remain at the periphery or outside of the Xist RNA domain (Chaumeil et al., 2006). It is unclear whether recruitment into the Xist-induced silent compartment is necessary for gene silencing, or whether it is rather a consequence of that process, but there is a clear correlation between the location of a gene within the X-chromosome territory and its transcriptional status (Chaumeil et al., 2006; Dietzel et al., 1999). In somatic cells, genes were found at the periphery of the inactive X irrespective of their activity (Clemson et al., 2006), suggesting that at the onset of XCI, the repositioning of genes might be crucial for their initial silencing. Notably, recruitment of genes into the Xist RNA domain is impaired if Xist lacks its silencing domain (the conserved A-repeat (Wutz et al., 2002)). This might indicate that an inability to recruit genes into the domain impairs silencing, but it remains possible that genes are not recruited because they are not being silenced. As discussed later in this review, Xist RNA is able to reorganise the 3D architecture of the X-chromosome that it coats (Giorgetti et al., 2016; Splinter et al., 2011), and this is tightly associated with the A-repeat and its capacity to induce gene silencing (Giorgetti et al., 2016). Further dissection of the Xist RNA functions will be necessary to understand whether structural changes precede transcriptional silencing at the onset of XCI, or vice-versa.
The involvement of repetitive elements
The repressive compartment created by the Xist RNA consists mostly of repeat-rich regions, which are silenced before gene-rich regions and independently of the Xist A- repeat (Chaumeil et al., 2006; Chow et al., 2010; Clemson et al., 2006). A particular class of repetitive elements, LINEs, has been proposed by Mary Lyon to facilitate Xist RNA spreading and efficient silencing along the chromosome (‘the repeat hypothesis’) (Lyon, 1998, 2006). LINEs could correspond to the “way stations” suggested to explain why autosomes are less efficiently silenced by Xist compared to the X-chromosome (Gartler and Riggs, 1983). This model proposed that there are special sequences on the X-chromosome (the “way stations”) that help the propagation of the inactivation process, by boosting the inactivation signal through the chromosome. LINEs are indeed at least 2-fold enriched on the X-chromosome compared to autosomes (Bailey et al., 2000) and several studies have reported a correlation between efficiency of gene silencing and the presence of LINEs (see (Gendrel and Heard, 2014) for review). However, their exact role in XCI remains mostly enigmatic (Lyon, 2006). Some LINEs are silenced during XCI, participating in the formation of the Xist-induced repressive compartment, while expression of a subset of young LINEs, triggered on the X- chromosome at the onset of XCI, has been suggested to facilitate local propagation of silencing (Chow et al., 2010).
Table of contents :
RÉSUMÉ (in French)
Prologue: the paradigm of the X-inactivation centre
Review 1: X-chromosome inactivation: new insights into cis and trans regulation (review published in Current Opinion in Genetics & Development, 2015)
Review 2: Chromatin architecture and gene regulation during X-chromosome inactivation (review in preparation for Annual Review of Genetics, 2018)
Review 3: From promoters and enhancers to TADs and regulatory landscapes across development, disease and evolution (unpublished review)
Article 1: Predictive polymer modelling reveals coupled fluctuations in chromosome conformation and transcription (published in Cell, 2014)
Article 2: Evidence for cross-TAD communication during X-inactivation via the noncoding Linx locus (manuscript in preparation)
Article 3: Genetic dissection of TAD organisation and function at the X-inactivation centre (manuscript in preparation)
Methods for Article 2 and Article 3
DISCUSSION & PERSPECTIVES
Article 4: Xist-dependent imprinted X inactivation and the early developmental consequences of its failure (published in Nat Struct Mol Biol, 2017)