Current view of the potential roles of proteins enriched on the inactive X chromosome

X chromosome inactivation (X-inactivation) is triggered by X-linked noncoding Xist RNA, which is expressed asymmetrically from one of the two X chromosomes in females and coats it in cis to induce chromosome-wide silencing. Xist RNA is thought to play a role as a platform in recruiting proteins involved in gene silencing and heterochromatinization, which mediate serial changes in epigenetic modification of the chromatin. During the last two decades, many proteins have been shown to be enriched on the inactivated X chromosome in mouse and human. Although the biological significance of most of them for X-inactivation has not been fully established, extensive studies of these proteins should provide a better understanding of the molecular basis of how X-inactivation mediated by Xist RNA is regulated. Here, we review the potential roles of some of these proteins in the stepwise process of Xist RNA-mediated chromosome silencing.


INTRODUCTION
During early development in female mammals, one of the two X chromosomes becomes transcriptionally silenced, so that the dosage difference of X-linked genes between the sexes is compensated for (Lyon, 1961). This X-inactivation is a phenomenon which, in addition to the t-haplotype (Sugimoto, 2014), has attracted the attention of mouse geneticists for many years. The X chromosome is also intriguing in the context of hybrid male sterility (Oka and Shiroishi, 2014). In the process of X-inactivation, an X-linked noncoding RNA, Xist, is asymmetrically upregulated on either X chromosome as cells differentiate, and subsequently associates in cis with the X chromosome from which it is transcribed to induce chromosome silencing. Although the molecular mechanisms are still largely obscure, Xist RNA is thought to recruit proteins that are involved in gene silencing and heterochromatinization. Available evidence also suggests that the association of Xist RNA with the inactive X chromosome (Xi) is mediated not by hybridization with DNA to form a heteroduplex but by an interaction with chromatin proteins (Clemson et al., 1996). Given that Xi is often found at the nuclear periphery, and that Xist RNA remains associated with the nuclear matrix or scaf-fold even after chromatin proteins and genomic DNA are removed, it is likely that Xist RNA serves as a bridge between chromatin and other structural components. Although many proteins such as chromatin proteins and epigenetic regulators have been shown to localize to Xi (reviewed in De La Fuente et al., 2011), it is unclear in most cases what impact Xist RNA has on their localization to Xi. Indeed, the biological significance of most of these proteins for X-inactivation is not fully understood. Detailed studies of these proteins, however, should help us to understand the molecular basis of how Xist RNAmediated X chromosome silencing is initiated, established and maintained, and it is therefore timely to summarize our current knowledge about proteins that are enriched on Xi. Here, we provide an overview of several of these proteins ( Fig. 1) in terms of their potential role in the stepwise process of Xist RNA-mediated gene silencing.

macroH2A AND Parp1
macroH2A was the first protein found to localize to Xi (Costanzi and Pehrson, 1998). It is a variant of histone H2A with a large C-terminal non-histone domain known as the macro domain, and is three times the size of canonical histone H2A (Pehrson and Fried, 1992). There are three isoforms of macroH2A: macroH2A1.1 and 1.2 are produced from a single gene, H2afy, by alternative splicing (Pehrson et al., 1997;Rasmussen et al., 1999), while Edited by Hiroshi Iwasaki * Corresponding author. E-mail: tsado@nara.kindai.ac.jp macroH2A2 is encoded by H2afy2. These are enriched on Xi in a chromosome-wide fashion and form a macro chromatin body (Chadwick and Willard, 2001;Costanzi and Pehrson, 2001;Mietton et al., 2009). Their recruitment occurs at the morula stage during mouse development (Costanzi et al., 2000), and at a relatively late phase in the process of X-inactivation during embryonic stem (ES) cell differentiation (Mermoud et al., 1999;Pullirsch et al., 2010). Conditional depletion of Xist RNA in female mouse embryonic fibroblasts (MEFs), while not compromising the X-inactivation state, results in a loss of macroH2A1.2 from Xi (Csankovszki et al., 1999;Wutz et al., 2002). Xist expression from an autosomal transgene, which apparently induces silencing of neighboring genes at the integration site, is accompanied by the enrichment of macroH2A1 in differentiating ES cells (Rasmussen et al., 2001). On the other hand, mutant Xist RNA is still capable of recruiting macroH2A1 although it is defective in silencing because it lacks the A-repeat, one of the conserved repeats present in the 5' region of the RNA and essential for Xist's function in silencing (Wutz et al., 2002). These studies demonstrate that although the deposition of macroH2A on Xi depends on the presence of Xist RNA, it is not sufficient for either triggering Xinactivation or maintaining the inactivated state. However, depletion of macroH2A in combination with treatment with inhibitors of DNA methylation and histone deacetylases resulted in reactivation of an X-linked GFP transgene in a portion of MEFs, suggesting that macroH2A, DNA methylation and histone hypoacetylation act synergistically in the maintenance of the Xinactivation state (Hernández-Muñoz et al., 2005).
Given the specific localization of macroH2A not only to Xi but also to the imprinting control regions of the repressed alleles, macroH2A, especially the macro domain, is likely to be involved in gene silencing. It has been shown that the macro domain mediates an interaction between macroH2A1.1 and poly(ADP-ribose) polymerase 1 (PARP1), which also localizes to Xi (Nusinow et al., 2007;De La Fuente et al., 2011). PARP1 belongs to a group of nucleosome binding proteins that affect chromatin structure, and, when catalytically inactive, it inhibits transcription by binding to chromatin. macroH2A may therefore act as a transcriptional silencer by associating with PARP1 to repress its enzymatic activity through the macro domain (Ouararhni et al., 2006;Nusinow et al., 2007). However, targeted disruption of H2afy and Parp1 in mice revealed that mutants homozygous for either gene are viable and fertile, demonstrating their nonessential roles in chromatin regulation during normal development (Wang et al., 1995;Changolkar et al., 2007). In addition, female ES cells depleted of both macroH2A1 and macroH2A2 are capable of undergoing Xinactivation upon differentiation (Tanasijevic and Rasmussen, 2011). Furthermore, although the details have not yet been provided, it has been reported that H2afy and H2afy2 double knockout mice are viable and fertile (Gaspar-Maia et al., 2013). We have also found that female mice deficient for both H2afy and H2afy2 are viable and fertile (unpublished observations). It is thus likely that macroH2As are dispensable for normal development as well as for the initiation and establishment of X-inactivation. A recent study, however, demonstrated that macroH2A acts as a barrier to nuclear reprogramming and that its depletion facilitates reactivation of Xi in somatic cells (Pasque et al., 2011), suggesting that macroH2As serve as an epigenetic lock against reversion of the chromatin state from the differentiated to the pluripotent cell type. ? nuclear organization Proteins enriched on the inactive X chromosome

SAF-A/hnRNP U
Xist RNA exerts its effect on gene silencing by coating the chromosome from which it originates. The mechanism by which Xist RNA attaches to the X chromosome in cis, however, is still largely obscure. It has been reported that Xist RNA remains associated with the nuclear matrix after removal of chromatin by DNase digestion and salt extraction, suggesting that it constitutes part of a non-chromatin nuclear structure (Clemson et al., 1996). Interestingly, scaffold attachment factor A (SAF-A), also known as heterogeneous nuclear ribonucleoprotein U (hnRNP U), has been shown to localize to Xi (Helbig and Fackelmayer, 2003;Fackelmayer, 2005). SAF-A/hnRNP U is a prominent component of the nuclear matrix or scaffold and has been suggested to be involved in chromatin organization, transcriptional regulation and RNA processing. It has two functional domains, a SAFbox near the N terminus (Kipp et al., 2000) and an RGGbox near the C terminus (Kiledjian and Dreyfuss, 1992), and an Spla and Ryanodine receptor (SPRY) domain of unknown function (Ponting et al., 1997). The SAF-box and RGG-box mediate binding to AT-rich scaffold or matrix attachment regions (S/MARs) and to RNA, respectively. As is the case for Xist RNA, SAF-A/hnRNP U stably associates with the territory of Xi through the RGG-box even after chromatin is removed (Helbig and Fackelmayer, 2003). Given these molecular characteristics, this protein is a plausible candidate anchor point for Xist RNA on Xi. In fact, siRNA-mediated knockdown of SAF-A/hnRNP U results in the dissociation of Xist RNA from Xi (Hasegawa et al., 2010). Furthermore, the SAFbox and RGG-box are both essential for tethering Xist RNA to Xi. A recent computational analysis based on physicochemical properties of nucleic acid and amino acid side chains has predicted with high confidence that the SAF-box binds to the A-repeat region (Agostini et al., 2013). It has been shown, however, that a silencingdefective Xist RNA lacking the A-repeat region is still capable of recruiting SAF-A/hnRNP U as well as macroH2A and Polycomb group proteins (Pullirsch et al., 2010), suggesting that deposition of these proteins is insufficient for inducing X-inactivation. Furthermore, given that SAF-A/hnRNP U apparently begins to associate with Xi at the time of the transition to the maintenance phase of X-inactivation (Pullirsch et al., 2010), the recruitment of Xist RNA to Xi should precede that of SAF-A/hnRNP U. This appears to contradict the idea that SAF-A/hnRNP U plays a role in anchoring Xist RNA on the X chromosome to induce X-inactivation. Depletion of SAF-A/hnRNP U in differentiating female ES cells, however, has been shown to compromise the initiation of Xinactivation (Hasegawa et al., 2010). Although hypomorphic mutants for SAF-A/hnRNP U, which die at around embryonic day (E) 9.5, have been described (Roshon and Ruley, 2005), it is unclear whether or not there are any sex-specific differences in the time at which developmental defects become apparent. It will be important to study the effect of functional loss of SAF-A/ hnRNP U using null mutant mice.

SATB1 AND SATB2
Special AT-rich sequence binding protein 1 (SATB1) is another plausible factor that may participate in the inactive structure of Xi. SATB1 is a transcription factor that is predominantly expressed in thymocytes and binds to a core unwinding region within the S/MARs (Kohwi-Shigematsu and Kohwi, 1990). It contains a homeodomain at the C terminus and two Cut domains containing Cut-like repeats in the center as potential DNA binding domains (Nakagomi et al., 1994;Dickinson et al., 1997). SATB1 displays a cage-like distribution surrounding heterochromatin in thymocyte nuclei and this network regulates the expression of target genes through higher-order packaging of chromatin structure (Cai et al., 2003).
The ability of Xist RNA to induce gene silencing is likely to be restricted to a certain window during differentiation (Wutz and Jaenisch, 2000). The induction of Xist RNA coating can cause gene silencing in undifferentiated ES cells but not in differentiated cells such as MEFs. However, the intriguing finding that Xist RNA is still capable of inducing gene silencing in pre-B cells and pre-T cells (Savarese et al., 2006) suggests that not only embryonic cells but also pre-B and pre-T cells share a primary initiation factor(s) that triggers Xist RNA-mediated gene silencing (Agrelo et al., 2009). SATB1 and its paralog, SATB2, were found to be expressed in ES cells as well as pre-B and pre-T cells and downregulated in differentiating ES cells (Savarese et al., 2009). Furthermore, ectopic expression of either SATB1 or SATB2 in MEFs could induce gene silencing on the chromosome coated with Xist RNA, suggesting these proteins as the presumptive initiation factors (Agrelo et al., 2009). Notably, however, although mutants doubly deficient for SATB1 and SATB2 are embryonic lethal, MEFs that can be derived from double-mutant female fetuses contain a typical Xi, indistinguishable from that in wild-type female MEFs (Alvarez et al., 2000;Nechanitzky et al., 2012). This result suggests that SATB1 and SATB2 are dispensable for the initiation of gene silencing on Xi in mice. Although SATB1 and SATB2 are attractive factors in that they confer on differentiated cells the ability to respond to Xist RNA and initiate X-inactivation, further studies are required.

POLYCOMB REPRESSIVE COMPLEXES 1 AND 2
(1) PRC2 Polycomb group (PcG) proteins play an important role in developmental gene regulation and cell fate commitment. Trimethylation at lysine 27 of histone H3 (H3K27me3), one of the histone modifications that effects gene repression, is mediated by polycomb repressive complex 2 (PRC2), which consists of four core PcG protein components, Ezh2, Eed, Suz12 and RbAp46/48, in combination with accessory factors such as Jarid2 and PCL2 (MTF2). Although Ezh2 is a catalytic subunit of canonical PRC2, its paralog Ezh1, present in noncanonical PRC2, is also capable of catalyzing H3K27me3 and is responsible for residual H3K27me3 in cells lacking Ezh2 (Shen et al., 2008). Eed and Suz12 seem to be indispensable components for the activity of either canonical or noncanonical PRC2.
Enrichment of H3K27me3 is one of the earliest epigenetic changes that happen to the X chromosome undergoing inactivation following the accumulation of Xist RNA on it. Eed was the first protein whose mutation was shown to cause a phenotypic difference in embryonic development between the sexes, attributable to a defect in X-inactivation (Wang et al., 2001). In the trophoblast of female embryos deficient for Eed (Eed -/-), where X-inactivation is normally imprinted, the inactivated paternal X chromosome becomes progressively reactivated during postimplantation development due to a failure in the maintenance of the inactive state. Although the deficiency of Eed results in embryonic lethality in both sexes, it makes the females die earlier than the males. It is unclear, however, if Eed is required for the initiation of imprinted paternal X-inactivation in preimplantation embryos, because Eed protein and/or mRNA stored in the oocytes, although consumed by the late blastocyst stage (E4.5) (Kalantry and Magnuson, 2006), is initially available even in Eed -/embryos. In contrast to imprinted Xinactivation in the extraembryonic tissues, random Xinactivation in the embryonic tissues is properly initiated and maintained in Eed -/female embryos after implantation, indicating that Eed is dispensable for random Xinactivation (Kalantry and Magnuson, 2006). Suz12, like Eed, is also required for the enzymatic activity of PRC2. Functional disruption of Suz12 in the mouse resulted in embryonic lethality at the early postimplantation stages, with a global loss of H3K27me3 (Pasini et al., 2004). Given the indispensable role of Eed and Suz12 for PRC2 activity, it is likely that H3K27me3 is essential for the maintenance of imprinted X-inactivation in the trophoblast but for neither the initiation nor the maintenance of random X-inactivation in the embryonic lineage.
It has been reported that a short RNA transcribed from the Xist locus, RepA, which shares its sequence with the 5' part of Xist RNA, interacts with Ezh2 and recruits PRC2 to the Xist locus, from which H3K27me3 propagates along the chromosome (Zhao et al., 2008). Other long noncoding RNAs (lncRNAs) such as HOTAIR have also been shown to interact with PRC2 and to facilitate H3K27me3 at their target loci (Rinn et al., 2007;Tsai et al., 2010). These findings have provided the basis of the concept for the role of one class of lncRNA to modulate chromatin and gene expression. Although this is an attractive and plausible model, another study based on a biochemical analysis measuring binding constants demonstrated that the binding affinities of PRC2 for lncRNAs that have been shown to interact with PRC2 and for irrelevant RNAs are comparable (Davidovich et al., 2013). Such promiscuous binding of PRC2 to various RNAs raises an issue about the specificity of the association between Ezh2 and Xist RNA and argues for the need for reevaluation.
Jarid2, a founder member of the Jumonji family of proteins, was recently shown to be transiently enriched on the X chromosome undergoing inactivation via an interaction with Xist RNA prior to the recruitment of PRC2, suggesting a role for Jarid2 in targeting PRC2 to Xi (da Rocha et al., 2014;Kaneko et al., 2014). Depletion of Jarid2, however, only delayed the recruitment of PRC2 to the X chromosome, and PRC2 was eventually loaded and catalyzed H3K27me3 on the X chromosome (Li et al., 2010). Therefore, while Jarid2 appears to play a role in the initial efficient recruitment of PRC2 to the X chromosome, its significance in the process of X-inactivation is still obscure. It should be noted that since homozygous mutants of Jarid2, although embryonic lethal, can develop until the midgestation stage (Takeuchi et al., 1995), X-inactivation would not be significantly affected.
(2) PRC1 The inactivated X chromosome is also characterized by monoubiquitination at lysine 119 of histone H2A (H2AK119ub1), although the biological significance of this modification is not fully understood. This is catalyzed by the E3 ligases Ring1A and Ring1B (de Napoles et al., 2004), and either of them, in combination with Mel18 and Cbx proteins, among others, is incorporated into another class of PcG protein complex, PRC1. While Ring1A-deficient mice are viable and fertile, those deficient for Ring1B are developmentally arrested after implantation with gastrulation defects and die before E10.5 (Voncken et al., 2003). H2AK119ub1 on Xi is maintained in cells lacking Ring1A or Ring1B, but is lost if both are depleted (de Napoles et al., 2004). It is not clear whether depletion of H2AK119ub1 compromises Xinactivation status, as the fact that cells lacking both Ring1A and Ring1B are selected against precludes the analysis of a long-term effect of H2AK119ub1 depletion. Although the model that H3K27me3 mediated by PRC2 forms a binding site for PRC1 to facilitate H2AK119ub1 (Cao et al., 2002;Kuzmichev et al., 2002;Fischle et al., 2003;Min et al., 2003) has been generally accepted, a subsequent study demonstrates that PRC1 on its own can target Xi in the absence of H3K27me3, suggesting direct targeting of Xist RNA by PRC1 (Schoeftner et al., 2006).
Recently, a noncanonical PRC1 complex, which con-tains RYBP in place of CBX chromodomain proteins, has been shown to become loaded at its targets independently of H3K27me3 and to mediate H2AK119ub1 as efficiently as canonical PRC1 containing CBX proteins (Tavares et al., 2012). PRC1 containing RYBP was also found to be enriched over the territory of Xist RNA expressed from an autosomal transgene. These findings suggest that PRC1 containing RYBP can be recruited to Xi in response to Xist RNA coating and can catalyze H2AK119ub1 independently of H3K27me3. The molecular basis of this noncanonical PRC1 recruitment to the Xist RNA territory will be an interesting subject of future studies.

SmcHD1
SmcHD1 was first identified as a protein whose functional disruption caused female-specific lethality during mouse embryonic development (Blewitt et al., 2008). It belongs to the SMC family and contains an N-terminal ATPase domain and a C-terminal SMC hinge domain. Given that the postimplantation development of female embryos is significantly compromised if X-inactivation does not occur properly, and that the gross morphology of homozygous SmcHD1 mutants observed by Blewitt et al. (2008) was indistinguishable from that of their wild-type littermates at the early postimplantation stages, it is likely that X-inactivation had been initiated and established in these mutants, at least to an extent that was sufficient to support early postimplantation development. However, derepression of X-inactivated genes became apparent in the mutants by the midgestation stage, with an occasional decrease in DNA methylation at nearby CpG islands. These initial findings suggested that SmcHD1 was involved in DNA methylation at X-linked CpG islands and contributed to the maintenance of Xinactivation. Recent transcriptome analyses, however, revealed that only about 10% of the genes on Xi in SmcHD1-deficient female MEFs were affected, and that the derepression was not necessarily associated with hypomethylation at the CpG islands (Gendrel et al., 2013;Mould et al., 2013). These findings suggest that hypomethylation at the CpG islands is not a direct cause of the derepression of the X-inactivated genes in the SmcHD1 mutants, but rather is one of the consequences. It is therefore likely that SmcHD1 exerts its effect on the maintenance of X-inactivation via different types of epigenetic regulation. It is well established that PcG proteins play important roles in maintaining the repressed chromatin state in various contexts including Xi, probably by means of PRC2-mediated H3K27me3 and PRC1mediated H2AK119ub1. The fact that Xi still retains H3K27me3 in SmcHD1-deficient MEFs implies that SmcHD1 and PcG proteins may function in independent pathways.
HBiX1, recently identified as a protein that coimmuno-precipitates with heterochromatin protein 1 (HP1), has been shown to interact with SMCHD1 in human cultured cells (Nozawa et al., 2013). These three proteins appear to form a complex and localize to Xi. Genome-wide analyses by ChIP-seq have revealed that SMCHD1 and HBiX1 tend to associate with H3K27me3 and H3K9me3, respectively, on Xi. Depletion of SMCHD1 or HBiX1, although affecting the distribution of neither H3K27me3 nor H3K9me3 when examined by ChIP-seq, caused an apparent decompaction of Xi, which became barely detectable in interphase nuclei by immunofluorescence using an antibody against H3K9me3. Although derepression of genes on Xi was not observed in cells upon depletion of SMCHD1 or HBiX1, these results suggest the existence of an SMCHD1-HBiX1 pathway to maintain higher-order chromatin structure. Although neither HP1 nor H3K9me3 is detected on a mouse Xi by immunofluorescence analysis in interphase nuclei, it would be of interest to study the impact on X-inactivation during development if HBiX1 were deleted in the mouse.

CONCLUDING REMARKS
Xist RNA is considered to serve as a platform for proteins that are involved in the initiation and maintenance of chromosome-wide heterochromatinization by coating the X chromosome undergoing inactivation. Although many proteins have been shown to be enriched on Xi, conclusive studies of how they are targeted to Xi and whether they interact with Xist RNA have not yet been performed in most cases. Although RNA immunoprecipitation and cross-linking immunoprecipitation have been powerful techniques for revealing the interaction between small RNA and proteins, they are not invariably applicable to lncRNA and proteins. It will be important, if challenging, in the future to employ more biochemical approaches to address the interactions, and their molecular basis, between Xist RNA and the proteins that are enriched on Xi.