3.1 Introduction

The evolution of mammalian sex chromosomes started about 150 million years ago by mutations in the Sox3 gene that resulted in the new male sex determining gene Sry (Graves 2006). It is thought that after the birth of Sry, genes involved in male fertility evolved in close vicinity of Sry and that the accumulation of this block of heterologous genes blocked homologous recombination, which led to the degeneration of the Y chromosome. The loss of the ancestral genes on the new Y chromosome was compensated by a twofold upregulation of these genes on the remaining single X chromosome in male cells (Nguyen and Disteche 2006). However, this would have led to the overexpression of these genes in female cells, and to compensate for this, a silencing process coevolved in the female that ensured downregulation of the expression of X-linked genes. Currently, this silencing process, called X chromosome inactivation (XCI), entails cis inactivation of almost the whole X chromosome in most eutherians. XCI occurs early in the development of the female embryo, in mice already after the 4-cell stage (Mak et al. 2004; Okamoto et al. 2004; Okamoto and Heard 2006). Cells in the early mouse embryo always inactivate the paternally inherited X chromosome (Xp) and leave the maternally inherited X chromosome (Xm) active, which is referred to as imprinted XCI (Takagi and Sasaki 1975; West et al. 1977).

In the mouse, at 3.5 days postcoitum (dpc), imprinted XCI is reversed in the inner cell mass (ICM) of the blastocyst, resulting in reactivation of the Xp and subsequent initiation of random XCI around 5.5 dpc, whereas imprinted XCI is maintained in the extraembryonic tissue (Rastan 1982; Mak et al. 2004). Random XCI is also initiated upon differentiation of female embryonic stem (ES) cells derived from the ICM, providing a convenient model system to study XCI in vitro (Chaumeil et al. 2002; Navarro et al. 2008). Also in other eutherian species, including human, XCI is random and initiated early in embryonic development. However, it is unclear whether imprinted XCI is present in other eutherian species besides the mouse. Unlike imprinted XCI, in random XCI, both X chromosomes have an equal chance to be inactivated, causing ~50% of the cells to have an active Xp and ~50% of the cells to have an active Xm (Lyon 1961). Only one of the two X chromosomes should be inactivated because inactivation of all Xs, or even leaving both Xs active, is lethal to the cell (Marahrens et al. 1997; Lee 2002). Therefore, the number of X chromosomes present in the cell must be determined in the developing embryo. When a female cell has established that two X chromosomes are present, XCI is initiated on one of the two X chromosomes. Once random XCI is completed, the process is irreversible, and after each cell division, the inactivated X (Xi) will be clonally propagated, meaning that the same X remains inactivated in all daughter cells (Plath et al. 2002).

In the last few decades, several cis- and trans-acting factors involved in the regulation of the XCI process have been identified. The two main regulatory factors involved in XCI are Xist and Tsix (Penny et al. 1996; Marahrens et al. 1997; Lee et al. 1999), both located in a small region on the X chromosome, called the X-inactivation center (Xic, Fig. 3.1). Xist and Tsix encode functional ncRNAs. Xist expression and RNA spreading in cis is necessary for XCI to occur while Tsix represses expression of Xist in cis. Together, these two genes determine whether XCI occurs in cis on the X chromosome. Other elements, proteins, or genes that are involved in regulation of XCI are DXPas34, Xite, RepA, RNF12, OCT4, SOX2, NANOG, CTCF, and YY1, which seem to regulate Xist and/or Tsix expression and function, directly or indirectly, as described below.

Fig. 3.1
figure 1

Important players in XCI. Schematic representation of part of the X inactivation center including the Xist, Tsix, and Rnf12. Also shown is the localization of different repeats in Xist and the binding sites of different trans-acting factors involved in inhibiting XCI

Full size image

3.2 Cis-Regulatory Factors in XCI

The most important player in XCI is Xist, which is located on the X chromosome and encodes a 17 kb long noncoding RNA, which is spliced and polyadenylated (Borsani et al. 1991; Brockdorff et al. 1991; Brown et al. 1991). Prior to XCI, Xist expression is low and the transcript is unstable. However, upon initiation of XCI, Xist expression is upregulated on the future inactive X (Xi) and spreads along the X chromosome in cis, thereby directly or indirectly attracting chromatin modifiers involved in the chromosome-wide silencing process (Brockdorff et al. 1992; Brown et al. 1992). Many experiments have shown the importance of Xist in the XCI process. For instance, deletion of Xist from one X chromosome in XX female ES cells causes complete skewing of XCI toward the wild type X chromosome, while XY male ES cells are not affected (Penny et al. 1996). This is not a consequence of secondary selection in benefit of female cells inactivating the wild type X chromosome after completion of XCI, but the wild type X chromosome is always inactivated when Xist is deleted on one allele in female XX embryos (primary nonrandom XCI) (Marahrens et al. 1997, 1998; Gribnau et al. 2005). Furthermore, ectopic expression and spreading of Xist is enough to initiate chromosome inactivation, even on an autosome (Lee et al. 1996; Herzing et al. 1997; Lee and Jaenisch 1997). Silencing, at least partially, of a chromosome from which Xist is transcribed is irreversible after 3 days of differentiation in ES cells, as has been shown using an inducible Xist transgene. However, when Xist RNA is removed beforehand, the silenced state of genes is reversed (Wutz and Jaenisch 2000). Importantly, the expression level of Xist is one of the factors that determines skewing of XCI, as has been shown by changing the Xist transcription level on one of the two alleles by introducing a mutation or a deletion in the Xist promoter (Newall et al. 2001; Nesterova et al. 2003).

Xist contains different repeat sequences A–F, of which the A repeat is involved in gene silencing. Recent studies indicated that the A repeats form two stem loop structures, each containing four repeats, which attract the chromatin modifier complex PRC2 involved in gene silencing (Wutz et al. 2002; Maenner et al. 2010). The other sequences including repeats B–F play a redundant role in the proper localization of Xist to the X chromosome (Wutz et al. 2002). Comparison of the Xist genomic sequence across different eutherian species indicates that the Xist gene evolved very quickly and only revealed conservation of the promoter region and the different repeat structures (Nesterova et al. 2001). Recently, another smaller 1.6 kb ncRNA transcript, RepA , which partially overlaps with Xist and includes the A repeat, has been implicated to play a role in the initiation of XCI by locally attracting PRC2 prior to Xist spreading (Zhao et al. 2008). However, a clear function for RepA in the XCI process still needs to be established (Table 3.1).

Table 3.1 Noncoding RNAs and gene silencing
Full size table

Tsix is located 15 kb downstream from Xist and is transcribed in antisense direction of Xist. Tsix encodes a continuous antisense RNA of approximately 40 kb that spans all of Xist. Multiple transcription start sites for Tsix have been identified, and approximately 50% of the Tsix transcripts are spliced into various small isoforms of which the 3′ ends have an overlap with the promoter of Xist (Sado et al. 2001; Shibata and Lee 2003). Tsix is transcribed in male and female undifferentiated ES cells at a level about 10 to 100 times more than Xist, and during establishment of XCI from the allele that is to remain active in male and female differentiating ES cells. After completion of XCI, Tsix is downregulated (Lee et al. 1999; Shibata and Lee 2003).

Tsix is generally regarded as the major inhibitor of Xist and therefore as an important factor in XCI regulation. However, careful examination of the literature shows that overall antisense transcription through the Xist locus determines inhibition of Xist. For example, the loss of the major promoter of Tsix has no significant effect on the counting or initiation processes of XCI (Cohen et al. 2007). However, deletion of DXPas34, a CpG island located downstream of the Tsix transcription start site (TSS) from which antisense transcription is also initiated (Fig. 3.1), significantly decreases antisense transcription through the Xist locus and causes primary nonrandom inactivation of the targeted allele in female XX ES cells (Debrand et al. 1999; Vigneau et al. 2006; Cohen et al. 2007). Interestingly, the methylation status of DXPas34 coincides perfectly with the antisense transcription through Xist. The CpG island is hypomethylated when actively transcribed and hypermethylated when antisense transcription is downregulated (Prissette et al. 2001; Boumil et al. 2006). Antisense transcription is also initiated in a region ~10 kb upstream of Tsix, called Xite . Xite expression and the methylation pattern during XCI is similar to that of Tsix, and deletion of Xite results in reduced antisense transcription through the Xist locus and skewing of XCI toward inactivation of the targeted allele (Ogawa and Lee 2003; Stavropoulos et al. 2005, Boumil et al. 2006), implying a similar role for Xite in inhibition of Xist function as DXPas34 and Tsix. Furthermore, direct inhibition of antisense transcription by insertion of a polyA site between Xist and DXPas34 also causes primary nonrandom XCI in female ES cells and inappropriate XCI in male ES cells. Even more so, overexpression of antisense transcription on one allele results in primary nonrandom inactivation of the wild type allele (Luikenhuis et al. 2001). Finally, a 65 kb deletion encompassing not only Tsix but also Xite and DXPas34, thus abrogating all antisense transcription, shows not only complete primary nonrandom XCI of the targeted allele but also severe cell death in X0 and XY cells containing the deletion, invoked by improper XCI (Clerc and Avner 1998; Morey et al. 2004). Thus, inhibition of Xist seems to correlate with an increase in antisense transcription through the Xist locus.

3.3 Xist Versus Tsix

How does Tsix inhibit Xist expression? Different hypotheses have been proposed. First, Tsix may function by forming a double-stranded RNA heteroduplex with Xist, resulting in repressive small interfering RNA (siRNA), which functionally silences Xist in cis (Ogawa et al. 2008). However, overexpression of Tsix cDNA, which includes the homologous region with Xist on an allele with abrogated endogenous Tsix transcription by insertion of a polyA signal, does not restore Xist inhibition (Shibata and Lee 2004), arguing against RNA interference (RNAi)-based inhibition of Xist. Also, Dicer knockout mice and ES cells that have an impaired RNAi machinery exhibit correct XCI, although Xist is ectopically upregulated at later stages due to loss of DNA methylation at the Xist promoter (Nesterova et al. 2008; Kanellopoulou et al. 2009).

Secondly, Tsix and Xite might form a three-dimensional chromatin structure via DNA looping that enhances Tsix and Xite antisense transcription but excludes the Xist promoter and thereby inhibits Xist expression in cis. A chromosome-conformation-capture (3C) study has shown that Tsix and Xite interact over a long distance, while the Xist promoter seems to colocalize with the Jpx promoter when Xist is transcribed (Tsai et al. 2008). DXPas34 is a likely candidate for looping because deletion of DXPas34 causes a severely skewed phenotype in female ES cells and XCI in male ES cells (Debrand 1999; Vigneau et al. 2006; Cohen et al. 2007). Moreover, DXPas34 is bound by CTCF, a protein that is often implicated in the looping of DNA (Chao et al. 2002). However, the DXPas34 deletion does not significantly change the three-dimensional chromatin structure in male ES cells. Furthermore, it is hard to determine whether a specific three-dimensional chromatin conformation in cis is the cause or the consequence of the transcription profile of that allele (Tsai et al. 2008).

Finally, antisense transcription through the Xist locus may inhibit Xist upregulation through a transcription interference mechanism. How antisense transcription-based inhibition of Xist works mechanistically has not been shown but one can envision that promoter polymerase initiation complexes (PICs) will have more difficulty forming on a promoter when an elongation complex transcribing in the antisense direction coeXists at the locus (Shearwin et al. 2005). Furthermore, RNA polymerase II complexes of Xist and Tsix may collide during transcription elongation, causing a premature halt of Xist transcription and less Xist accumulation. Evidence for involvement of such a mechanism comes from studies that indicate a bimodal pattern of both Xist and Tsix transcripts, being highest at the transcription start site and gradually decreasing along the template (Shibata and Lee 2003; Marks et al. 2009). Alternatively, inhibition of Xist might be caused by alteration of the chromatin state of the Xist locus by the Tsix transcript. It has been postulated that Tsix transcription induces heterochromatin formation at the Xist promoter by Tsix-mediated recruitment of histone modifiers (Sado et al. 2005; Navarro et al. 2006). Recently, EED, a component of the PRC2 Polycomb complex, has been shown to work synergistically with Tsix in silencing Xist (Shibata et al. 2008). Furthermore, loss of antisense transcription through the Xist promoter causes reduction of CpG methylation and repressive histone modification marks, indicating that transcription from the Xist promoter is enhanced (Ohhata et al. 2008). However, findings of Sun et al. (2006) argue against this hypothesis by showing that activation of Xist on the future Xi is characterized by a transient heterochromatic state at the Xist promoter, perhaps induced by the silencing capacity of Xist itself and thus contradicting a functional role of chromatin modifications in the inhibition of Xist by Tsix. In conclusion, most evidence points toward a transcription or Tsix RNA-mediated mechanism of repression of Xist by Tsix, but the exact mechanism has yet to be established.

3.4 Trans-Regulatory Factors and Initiation of XCI

In the recent years, several trans-acting factors regulating XCI have been identified. Most of these factors are involved in suppression of XCI (XCI-inhibitors), either by repressing Xist or activating Tsix. Among the proteins involved in Tsix regulation are the insulator protein CTCF , and also the transcription factor Yin Yang 1 (YY1), for which several tandemly organized binding sites have been identified in the DXpas34 region, which is involved in Tsix regulation and in the Xite promoter. Knockout studies involving Yy1, or partial ablation of Yy1 and Ctcf through RNAi-mediated repression, revealed downregulation of Tsix expression and concomitant upregulation of Xist expression, supporting a role for YY1 and CTCF in activation of Tsix expression (Donohoe et al. 2007).

The pluripotency factors SOX2 , Nanog , and OCT4 have also been shown to be involved in the regulation of XCI by the silencing of Xist (Donohoe et al. 2007, 2009; Navarro et al. 2008). A binding site for all three factors has been identified in intron 1 of Xist, and binding of these factors is involved in the direct suppression of Xist. Interestingly, OCT4 and Sox2 also bind in the Xite enhancer, and OCT4 together with YY1 is recruited to Tsix downstream of the transcription start site and is involved in transcription activation of both Xite and Tsix. These factors therefore affect Xist expression through both Tsix-dependent and -independent pathways, indicating that different mechanisms act jointly in setting up the threshold that has to be overcome by Xist.

Autosomally encoded factors such as SOX2, OCT4, and Nanog play an important role in XCI. However, it can be excluded that sex-specific initiation of XCI is determined by these factors only because the concentration of these factors, if not regulated by (a) sex-chromosomal factor(s), will most likely be the same in male and female cells. Key to the XCI initiation process is therefore the presence of one or more X-encoded XCI-activators that are differentially expressed between male and female cells. Recently, the E3 ubiquitin ligase RNF12 has been identified as a dose-dependent X-linked activator of XCI (Jonkers et al. 2009). Additional copies of Rnf12 resulted in ectopic initiation of XCI in transgenic male cells and initiation of XCI on both X chromosomes in a high percentage of female cells. RNF12 may act through activation of Xist or suppression of Tsix, although the exact mechanism remains elusive so far. Also, Rnf12 cannot be the only XCI-activator because Rnf12 +/− female cells still induce XCI, albeit in a severely reduced percentage of cells, indicating that other X-encoded genes are involved in initiation of XCI (Jonkers et al. 2009).

Different mechanisms for counting the number of X chromosomes and initiation of XCI have been proposed. Most of these models explain XCI as a mutually exclusive process leading to one single Xi per female cell, for instance, through the protection of one X chromosome by an autosomally encoded blocking factor or pAiring and cross communication of both X chromosomes in female cells (Wutz and Gribnau 2007; Jonkers et al. 2009; Starmer and Magnuson 2009). However, recent studies indicate that XCI is more likely to be a stochastic process and that in female cells, both X chromosomes have a probability to initiate XCI (Monkhorst et al. 2008; Barakat et al. 2010). The probability to initiate XCI is determined by the nuclear concentration of the different XCI-activators and -inhibitors (Monkhorst et al. 2008, 2009). XCI-inhibitors set the threshold by suppression of Xist and activation of Tsix, which has to be overcome by the action of the XCI-activators. Only in female cells, the nuclear concentration of the XCI-activators is sufficient to boost enough Xist transcription, allowing spreading and initiation of XCI in cis. Because the XCI-activators are X-linked, initiation of XCI on one X results in rapid downregulation of the XCI-activator genes in cis, preventing initiation of XCI on the second X chromosome. Nonetheless, XCI can still be initiated on the remaining active X chromosome until enough XCI-activator protein is degraded after inactivation, which would lead to a female cell with two inactive X chromosomes. Indeed, a small percentage of female cells initiating XCI on both X chromosomes is found during the XCI process, and as expected when XCI-inhibitors are downregulated, or the XCI-activator Rnf12 is upregulated, this percentage of XiXi cells increases significantly. These results indicate that the regulation of XCI is determined by a tightly regulated balance of X-encoded activators and autosomally encoded inhibitors of XCI.

3.5 Establishment of the Inactive X

The first step in silencing the X chromosome is the spread of Xist RNA in cis over the X chromosome. Several redundant repeats of Xist are important for the localization of Xist RNA to the Xi (Wutz et al. 2002). Spreading of Xist causes depletion of RNA polymerase II and other components of the transcription machinery on the Xi within one day, and abrogates transcription of repeat and intergenic sequences, independently of the A-repeat (Chaumeil et al. 2006). However, silencing of X-linked genes is mediated by the A-repeat within Xist RNA and starts after 1–2 days, continuing until gene silencing is more or less completed after approximately 7 days of differentiation (Chaumeil et al. 2006; Lin et al. 2007). Silencing of genes is hypothesized to be associated with the relocation of active genes at the outer rim of the X chromosome territory toward the silent Xi territory invoked by the A-repeats (Chaumeil et al. 2006; Clemson et al. 2006; Lin et al. 2007).

After depletion of the transcription machinery from the Xi territory, the Xi chromatin is changed drastically (Fig. 3.2a, b). First, histone 3 lysine 27 trimethylation (H3K27me3) is acquired by transient localization to the Xi of the Polycomb repressor complex 2 (PRC2), which comprises protein subunits EED, EzH2, RbAp47/48, and Suz12, of which EzH2 has histone methyltransferase activity (Wang et al. 2001; Plath et al. 2003; Silva et al. 2003; Cao and Zhang 2004; de la Cruz et al. 2005). PRC2 is recruited by Xist RNA, as has been shown by either deletion of EED or conditional deletion of Xist, which both cause loss of H3K27me3 (Wang et al. 2001; Plath et al. 2003, 2004). PRC2 subunit EzH2 has been identified as the protein that targets the PRC2 complex to the A-repeat of Xist RNA (Zhao et al. 2008), although a more recent study indicated that SUZ12 may play a more important role in targeting PRC2 to Xist (Kanhere et al. 2010). Although PRC2 seems to be important for binding Xist to the Xi, it is not likely to be the only protein complex doing so because loss of PRC2 does not seem to affect random XCI in the embryo proper (Wang et al. 2001; Plath et al. 2003).

Fig. 3.2
figure 2

The landscape of chromatin modifications on the inactive X. ( a) On the left, the Xi in interphase is shown consisting of two distinct regions of heterochromatin, in pink and green. Xist RNA association, and H3K27me3, and ubH2A accumulation, among others, characterize the pink chromatin, whereas histone marks such as H3K9me3 and recruitment of HP1 characterize the green chromatin. The different chromatin states form a banded pattern on the inactive X chromosome in metaphase. On the right, the specific histone marks and other epigenetic features are depicted for the Xist associated pink chromatin (top) and green chromatin (bottom). (b) A large number of epigenetic changes are associated with the XCI process. The temporal changes, when induced by differentiation of female ES cells, are depicted along the timescale (days) and separated in color (pink or green) depending on which heterochromatin state the modification is associated with (as described in a). Changes associated with both heterochromatin states are shown in blue

Full size image

Apart from histone methylation, most cells also show accumulation of H2A lysine 119 ubiquitination (ubH2A) on the Xi after the onset of XCI, which is established by the Ring1A/B subunit of Polycomb repressor complex 1 (PRC1) (de Napoles et al. 2004; Fang et al. 2004; Plath et al. 2004). Ring1A and Ring1B have redundant functions in ubiquitination (de Napoles et al. 2004; Leeb and Wutz 2007), and only deletion of both Ring1 genes results in loss of ubH2A on the Xi (de Napoles et al. 2004). PRC1 recruitment to the Xi follows PRC2 recruitment, but is not solely mediated by H3K27me3, as has been shown in EED-deficient ES cells, but also by the 3′ end of Xist RNA, either directly through interaction with Xist or by indirect interaction with an Xist binding protein (Plath et al. 2004; Schoeftner et al. 2006). A potential candidate for targeting of the PRC1 complex to Xist RNA is the Polycomb homolog CBX7, which shows a high affinity for H3K27me3 and for RNA (Bernstein et al. 2006) and has been shown to interact with the Ring1 protein (Gil et al. 2004).

Another histone methylation mark associated with silenced chromatin, histone 3 lysine 9 trimethylation (H3K9me3), accumulates on the Xi just after H3K27me3 (Heard et al. 2001; Boggs et al. 2002; Mermoud et al. 2002; Peters et al. 2002; Rougeulle et al. 2004). H3K9me3 is most likely put in place by HMTase Suv39, and maintained by HP1, which is enriched on the Xi (Chadwick and Willard 2003, 2004), but other histone methyltransferases (HMTases) might also play a role.

H3K9me3 accumulation appears more or less simultaneous with the loss of acetylation of histone H3 and H4 (H3K9Ac and H4K5Ac, H4K8Ac and H4K12Ac, respectively) and trimethylation of histone H3 lysine 4 (H3K4me3) and histone H3 lysine 36 (H3K36me3), which are all hallmarks of euchromatin (Jeppesen and Turner 1993; Belyaev et al. 1996; Boggs et al. 1996, 2002; Keohane et al. 1996; Heard et al. 2001; Chaumeil et al. 2002; Chadwick and Willard 2003). Probably, a set of histone modifiers, including histone deacetylases (HDACs) and histone demethylases (HDMs), are attracted by H3K27me3 and Xist and colocalize with the Xi to direct the chromatin toward a heterochromatic state. Among the late epigenetic changes are macroH2A incorporation (Costanzi and Pehrson 1998; Mermoud et al. 1999), CpG island methylation, and late replication (Priest et al. 1967; Mohandas et al. 1981; Norris et al. 1991). MacroH2A is a H2A variant with a large C-terminal domain (Nusinow et al. 2007) that replaces H2A histones on the Xi after approximately 7 days of differentiation, forming a macrochromatin body (MCB) in a significant proportion of the cells (Costanzi and Pehrson 1998; Rasmussen et al. 2001). Xist expression is sufficient for initiation of H2A replacement by macroH2A and MCB formation (Rasmussen et al. 2001), and conditional deletion of Xist leads to loss of the MCB (Csankovszki et al. 1999). CpG methylation is also a late Xi mark and is put in place by de novo methyltransferase 3A (DNMT3A) (Hansen 2003) and maintained by DNMT1 (Sado et al. 2000).

Recently, several other factors have been shown to be involved in the maintenance phase of XCI. First, the DNA binding hinge-domain protein SmcHD1 plays a role in DNA methylation of the Xi. Loss of SmcHD1 results in depletion of DNA methylation at the X-linked CpG islands and reactivation of the Xi (Blewitt et al. 2008). It was postulated that SmcHD1 targets DNMT3A to the Xi, although no direct evidence in that direction was presented. Second, ATRX, encoded by an X-linked gene, has been shown to be involved in XCI. ATRX is a chromatin remodeler and a member of the SWI/SNF2 helicase family, which is enriched at the Xi, and the accumulation of ATRX can be regarded as a late mark of the Xi (Baumann and De la Fuente 2008). Interestingly, ATRX does repress not only X-linked genes on the Xi but also pseudo-autosomal genes that have translocated to an autosome, implicating that a (former) X chromosomal sequence is required to attract ATRX to a gene (Levy et al. 2008). Also, SATB1, which has been implicated in nuclear organization and involved in many forms cancers, has been identified as an important factor in Xist-mediated gene silencing (Agrelo et al. 2009). Expression of SATB1 allows Xist-mediated gene silencing even after the developmental window where Xist silencing is normally restricted to, indicating that SATB1 plays a key role in the establishment of the Xi. SAF-A is another factor involved in nuclear organization which plays an important role together with the tritorax protein Ashl, in the establishment of the Xi (Pullirsch et al. 2010). Both proteins, together with macroH2A, are involved in chromosome-wide histone H4 hypoacetylation, Interestingly, recruitment of most of the mentioned factors including components of PRC1 and PRC2, and SATB1, Ashl1, and SAF-A is not dependent on the A repeat of Xist, which is required for Xist mediated silencing of the Xi. This suggests that chromatin changes evoked by these proteins and protein complexes provide a repressive nuclear compartment, which may be required for subsequent gene silencing on the Xi mediated by the Xist A repeat. The recent discovery of these factors indicates that silencing of the Xi is more complex than initially thought and involves multiple factors, of which many are probably not yet revealed.

All these features of the Xi are important to lock-in the silenced state of the X chromosome. Together, they ensure that the Xi is nearly impossible to reactivate. The redundancy of the Xi hallmarks is demonstrated by conditional deletion of Xist after establishment of XCI, which causes loss of the macroH2A (Csankovszki et al. 1999) but still only leads to minor reactivation of the Xi, even when it is combined with loss of DNA methylation and inhibition of hypoacetylation (Csankovszki et al. 2001; Hernández-Muñoz et al. 2005).

3.6 Xist Spreading, Xi Organization, and Nuclear Organization

After Xist is upregulated on one of the two X chromosomes, it starts to spread in cis over the entire chromosome (Clemson et al. 1996; Hall and Lawrence 2003). Xist RNA is restricted to the inactivated X chromosome and does not localize to neighboring autosomes (Brown et al. 1992; Jonkers et al. 2008). Furthermore, studies on X:autosome translocations show that endogenously expressed Xist preferentially binds the X chromosomal part of the chromosome (Duthie et al. 1999; Keohane et al. 1999; Popova et al. 2006), and spreading into the autosome seems to be correlated with the density of LINE repeats (Popova et al. 2006). This observation has led to the LINE repeat hypothesis (Lyon 1998), in which it is stated that spreading of Xist is mediated by binding to LINE repeats. Indeed, LINE repeats are enriched twofold on the human X chromosome compared to autosomes, and the distribution of LINE repeats seems to correlate with the degree of XCI on the X chromosome (Boyle et al. 1990; Bailey et al. 2000; Ross et al. 2005). Also, computational studies of the DNA sequence surrounding genes escaping XCI compared to silenced X-chromosomal genes indicate a depletion of LINE repeats around escaping genes (Carrel et al. 2006; Wang et al. 2006).

Not all computational studies on the DNA sequence of the X chromosome find a clear correlation between LINE repeats and XCI (Chureau et al. 2002; Ke and Collins 2003). Also, Xist RNA does not spread over the X chromosome homogenously but appears to have a banded pattern when detected on a metaphase Xi and an open circle shape at the periphery of the Xi in interphase cells (Fig. 3.2a, left) (Duthie et al. 1999; Chadwick and Willard 2004; Smith et al. 2004). Curiously, this Xist RNA localization pattern does not seem to correspond to the density of underlying LINE repeats, but rather to the gene density on the X chromosome (Smith et al. 2004; Clemson et al. 2006). The banded pattern on the metaphase Xi of Xist RNA and gene rich regions can also be observed with histone marks H3K27me3, macroH2A and ubH2A, while histone marks H4K20me3 and H3K9me3 are enriched on the gene-poor regions of the Xi metaphase chromosome (Fig. 3.2a, b) (Gilbert et al. 2000; Chadwick and Willard 2004; Smith et al. 2004; Chadwick 2007). Therefore, the importance of the DNA sequence in the silencing process remains elusive as a direct interaction of LINE repeats or another specific DNA motive with histone marks and/or Xist RNA has not yet been reported.

Together, these data suggest a three-dimensional organization of the Xi, in which the gene-poor regions enriched in histone marks H4K20me3 and H3K9me3 are more internally located and the gene rich-regions, enriched in Xist RNA, H3K27me3, macroH2A, and ubH2A are present on the outer rim of the Xi territory (Chadwick and Willard 2004; Chaumeil et al. 2006; Clemson et al. 2006). Overall, the Xi becomes more spherical but retains a similar volume to the Xa (Eils et al. 1996). This Xi organization corresponds to DNA-FISH analysis of escaping and silenced X chromosomal genes, which shows that all analyzed genes are localized at the periphery of the Xi territory, but that active genes seem to “loop-out” of the chromosome territory (Dietzel et al. 1999; Chaumeil et al. 2006; Clemson et al. 2006). Early during the XCI process, Xist accumulation results in transcriptionally silent compartment devoid of RNA polymerase II and enriched for heterochromatin marks (Chaumeil et al. 2006). Interestingly, at this stage, only repetitive DNA is repressed and located within this silent compartment, and subsequent silencing of X-linked genes is accompanied by a shift in the localization of these genes toward a more internal localization. This change in localization and silencing of X-linked genes requires the presence of the Xist A repeat, in contrast to the RNA polymerase II excluded silent compartment that is also formed without the A repeat. Whether relocalization of X-linked genes upon XCI is the consequence of the XCI process itself or is directly involved in enforcing gene inactivation remains to be determined.

The Xi might not only have an intrinsic three-dimensional organization but is also specifically positioned within the nucleus. After inactivation, the Xi is preferentially located either at the periphery of the nucleus (Bourgeois et al. 1985; Belmont et al. 1986) or near the perinucleolar region (Bourgeois et al. 1985; Zhang et al. 2007). The specific positioning of the Xi could be mediated by the components of nuclear matrix. For instance, nuclear matrix scaffold protein SAF-A colocalizes with the Xi, which seems to be dependent on the RNA binding domain of the protein (Helbig and Fackelmayer 2003, Fackelmayer 2005). Furthermore, cells expressing mutated LaminA show depletion of heterochromatic marks H3K27me3 and H3K9me3 at the Xi, and the peripheral localization of the Xi is lost (Shumaker et al. 2006). These results indicate that the localization of the Xi in the nuclear periphery is either a consequence of its heterochromatic state or affects the heterochromatic state of the Xi (Shumaker et al. 2006; Fedorova and Zink 2008). However, the perinucleolar localization of the Xi is less easy to comprehend, especially because the Xi seems to preferentially colocalize with the perinucleolar region during S phase (Zhang et al. 2007). The S phase-specific localization is dependent on Xist, as autosomes containing an Xist transgene are also repositioned to the perinucleolar region in S phase, and conditional Xist knockout cells loose the preferential perinucleolar localization of the Xi. Interestingly, heterochromatin replication occurs late during S phase, at which point replication can only be observed around nucleoli and at the periphery of the nucleus (O’Keefe et al. 1992; Kennedy et al. 2000). Thus, perhaps, heterochromatin characterized by H3K27me3 needs a specialized nuclear compartment for replication and/or maintenance of the silenced state after replication.

3.7 Other Functional ncRNAs

The discovery of Xist provided a powerful model system to study the role and function of long ncRNA’s. Besides Xist, several other ncRNAs have been described to be involved in gene silencing in cis and in trans, and several parallels can be drawn between the action of these RNAs. Air and Kcnq1ot1 are two well-studied imprinted genes, both encoding noncoding transcripts involved in silencing in cis. Air encodes a 108-kb long unspliced transcript, which is transcribed antisense to the protein coding gene Igf2r (Lyle et al. 2000). Air expression is exclusively paternal, whereas Igf2r is maternally expressed. Besides Air-mediated silencing of the overlapping Igf2r gene, silencing also involves genes, including Slc22a3, located more than 200 kb away from Air, suggesting a direct role for the Air transcript in long range gene silencing. In cis silencing by Air involves the recruitment of G9A, required for H3K9 mono and dimethylation, and similar to Xist spreading (although less robust) the Air RNA appears to form a silent nuclear domain that envelops the paternal Slc22a3 locus. Interestingly, G9A appears to be needed for the silencing of Slc22a3 but not for the repression of Igf2r (Nagano et al. 2008). This finding indicates that different mechanisms may be involved in the regulation of antisense transcribed overlapping genes (Air/Igf2r) and long range gene silencing in cis (Air/Slc22a3). This is reminiscent of findings obtained with the regulation of the Xist/Tsix locus and silencing of X-linked genes in cis, which also supports the presence of different mechanisms involved in these processes.

Expression of Kcnq1ot1 is also imprinted, and the 91 kb paternally expressed gene is transcribed antisense to, and partially overlaps with Kcnq1 (Fitzpatrick et al. 2002; Pandey et al. 2008). Kcnq1ot1 is involved in the regulation of a cluster of imprinted genes on mouse chromosome 7 (Mancini-Dinardo et al. 2006). In cis silencing of Kcnq1ot1 spans a region of 400 kb in the embryo and 780 kb in the placenta and involves recruitment of several chromatin modifiers including G9A and PRC2 (Pandey et al. 2008; Redrup et al. 2009). Similar to Xist and Air, RNA FISH studies indicate that Kcnq1ot1 appears to form a silent nuclear domain, which is larger in the placenta than in the embryo. Interestingly, the Kcnq1 domain is also found in close proximity to the nucleolus in a high percentage of cells, suggesting a lineage-specific localization of this locus with the nucleolus (Pandey et al. 2008). Whether there is a functional role for this localization close to the nucleolus and whether the localization is dependent on spreading of Kcnq1ot1 remain to be determined.

In contrast to Kcnq1ot1 and Air, which invoke silencing in cis, HOTAIR has recently been identified as an ncRNA involved in silencing in trans (Rinn et al. 2007). HOTAIR is a 2.2-kb RNA expressed from the HOXC locus, which represses transcription of different Hox genes in the HOXD locus, which is located on a different chromosome. HOTAIR associates with PRC2, which mediates silencing of HOXD genes in trans, through H3k27me3 of target genes. HOTAIR expression is increased in several tumors, and loss of HOTAIR expression inhibits cancer invasiveness (Gupta et al. 2010). Interestingly, overexpression of HOTAIR results in genome-wide changes in the targeting of PRC2 and increased cancer invasiveness. This indicates that HOTAIR plays a much broader role in targeting of PRC2, besides regulation of the HOXD locus.

Recently, a genome-wide study indicated the presence of more than 3,300 large ncRNAs (Guttman et al. 2009; Khalil et al. 2009). About 20% of these ncRNAs associate with PRC2 and other chromatin modification complexes, indicating that findings with Xist and other well-studied ncRNAs including Air, Kcnq1ot1, and HOTAIR may be extrapolated to explain the function of these newly identified ncRNAs.

3.8 Conclusion

The discovery of Xist exemplified the importance of ncRNAs in cellular function. Xist was the first identified mammalian large ncRNA involved in gene silencing, providing a powerful model system to study RNA-mediated gene silencing. New advances in RNA sequencing indicate that many more ncRNAs will soon be identified as functional RNAs, and unraveling the role of Xist in XCI will help in understanding the function of these ncRNAs in the regulation of gene expression. Nevertheless, despite a lot of progress in understanding the role of Xist in XCI, many questions remain. For instance, how is the binding specificity of RNA binding proteins and complexes generated, which proteins are involved in fixing Xist to the chromatin, and why are so many proteins implicated in XCI dispensable for the XCI process? We are hopeful that the quickly advancing technology allows these questions to be addressed in the near future.