X inactivation is the mechanism by which mammals adjust the genetic imbalance that arises from the different numbers of gene‐rich X‐chromosomes between the sexes. The dosage difference between XX females and XY males is functionally equalized by silencing one of the two X chromosomes in females. This dosage‐compensation mechanism seems to have arisen concurrently with early mammalian evolution and is based on the long functional Xist RNA, which is unique to placental mammals. It is likely that previously existing mechanisms for other cellular functions have been recruited and adapted for the evolution of X inactivation. Here, we critically review our understanding of dosage compensation in placental mammals and place these findings in the context of other cellular processes that intersect with mammalian dosage compensation.
Sex determination in mammals is based on the heteromorphic sex chromosomes X and Y. It is believed that the X and Y chromosomes evolved from a pair of homologous autosomes. The evolutionary decay of genes on the Y chromosome results in a different number of gene‐rich X‐chromosomes being found in males and females, and, therefore, a large‐scale genetic imbalance between the sexes (Graves, 2006). To equalize this dosage difference, an epigenetic compensatory system has evolved. Similar XY sex chromosome systems have arisen elsewhere in the animal kingdom; however, the mechanisms by which dosage compensation is achieved in flies and worms, for example (reviewed by Lucchesi et al, 2005), vary from the mammalian mode, which is the focus of this review.
In mammals, dosage compensation is achieved by inactivating one of the two X chromosomes in females. Haplo‐insufficiency, resulting from the shut down of an entire chromosome, poses a potential problem; therefore, the presence of a single active X chromosome (Xa) must have conferred an evolutionary advantage and, hence, enabled its conservation in the process of natural selection. There is now evidence to support a second form of dosage compensation in mammals, which involves the upregulation of the Xa to balance the differential dosage between the X chromosome and the autosomes (Gupta et al, 2006; Nguyen & Disteche, 2006). So, what advantage did X inactivation confer and how did it arise in mammals?
Mammalian X inactivation seems to come in two forms: imprinted and random X inactivation. Imprinted X activation occurs in early mammals, such as marsupials, which diverged from placental mammals approximately 180 million years ago. This form of dosage compensation is achieved by inactivating the paternally inherited X chromosome. Imprinted X inactivation can also be found in the extra‐embryonic tissues of a subset of placental mammals. The cells forming the embryo in placental mammals undergo random X inactivation, through which either the paternal or the maternal X chromosome is inactivated. This novel mechanism is based on a noncoding RNA called Xist, which is unique to placental mammals and has not been found in marsupials or other vertebrate genomes (Duret et al, 2006). The evolutionary force behind random X inactivation and the emergence of the regulating Xist RNA is unclear; however, it could have been a combination of environmental circumstances and the selection pressure for placental mammals.
It has been suggested that the expansion of placental mammals was a consequence of the change in oxygen levels over the past 200 million years (Falkowski et al, 2005). The placental system requires a high ambient oxygen concentration owing to its inefficient transport from the maternal circulation to the developing embryo. This system became more efficient with the rise in global oxygen levels and resulted in a greater reproductive demand on mammalian females. We speculate that as a consequence of this increased demand on females, the selection pressure for the evolution of advanced mammalian traits became limited to males. The erosion of the Y chromosome, and the consequent emergence of a single X in males, provided a haploid region of the genome that is subject to efficient and stringent selection. Consistent with this, genes that are expressed in brain, muscle and germ cells are enriched on the human X chromosome (Graves, 2006).
Hemizygosity of X‐linked genes makes males more prone to the adverse effects of X‐linked mutations (Franco & Ballabio, 2006). This situation also occurs in female marsupials, in which imprinted inactivation of the paternal X chromosome unmasks mutations on the maternal Xa. By contrast, random X inactivation in female placental mammals generates a mosaic of cells with either a paternal or maternal Xa, thereby masking the effect of deleterious mutations and increasing reproductive fitness (Franco & Ballabio, 2006). We therefore postulate that the benefit obtained in protecting the female against potentially deleterious mutations puts random X inactivation at a selective advantage and, hence, has become the mechanism of choice in placental mammals.
Random X‐inactivation consists of an ordered series of processes. Each cell ensures, in a random manner, that only one X chromosome remains active and that the other X chromosome is inactivated. The differential treatment of the two X chromosomes results in an Xa and an inactive X (Xi), both of which are present in the same female nucleus. The Xist RNA, which is expressed exclusively from, and coats, the Xi (Fig 1), is used to establish the two functionally distinct forms of the X chromosomes. Although RNA components have been commonly used throughout evolution in the regulation of gene expression, Xist is unique in its ability to spread over and encompass the Xi. The proposed mechanisms that underlie Xist function are presented in this review. We also discuss how the Xi can be identified from its epigenetic marks, which represent a chromatin decoration feature that is commonly found in heterochromatic regions in the nucleus. Furthermore, we discuss the role of Polycomb group (PcG) proteins—notorious for maintaining transcriptional repression of developmental control genes in species ranging from flies to mammals—in the faithful maintenance of the Xi throughout cellular divisions.
From meiotic pairing to counting and choosing the X
Random inactivation poses an additional problem compared with other dosage‐compensatory mechanisms because the two X chromosomes are present in different states—active and inactive—within the female nucleus. In imprinted X inactivation, the X chromosomes are distinguished by their parental origin. The inactive state of the paternal X chromosome has been proposed to be a carry‐over effect from meiotic sex‐chromosome inactivation in the male germline. Asynapsed chromosome segments, resulting from the non‐homologous XY chromosome pair, are subject to transcriptional silencing in order for meiosis to proceed (Namekawa et al, 2006; Turner et al, 2006). In mice—in which random X inactivation occurs—a pre‐inactivated paternal X arrives at the zygote, which is consistent with the idea of imprinted X inactivation being the ancestral form of random X inactivation (Huynh & Lee, 2003). This paternal Xi is subsequently reactivated in the embryo, followed by random X inactivation in a Xist‐dependent manner (Mak et al, 2004; Okamoto et al, 2004).
At the onset of random X inactivation, the number of X chromosomes relative to the number of autosomes is counted, allowing one X to remain active per diploid chromosome set (reviewed by Heard & Disteche, 2006). Counting and choice are controlled by a single locus on the X, known as the X‐inactivation centre (Xic), which contains multiple regulatory elements including the Xist gene (Fig 2A). Ectopic Xic transgenes can trigger the initiation of X inactivation in male cells (Chow et al, 2005; Herzing et al, 1997; Lee et al, 1996). The function of Xic has been studied extensively by using gene‐targeting techniques in mice. A deletion spanning 65‐kbp at the 3′ end of Xist leads to ectopic X inactivation in cells that have only one X chromosome (Clerc & Avner, 1998). Another choice regulator, Tsix RNA, overlaps with the Xist gene and is transcribed in the antisense orientation (Lee et al, 1999). Tsix is initially expressed on both X chromosomes and is downregulated on the Xi before inactivation; conversely, Tsix expression persists longer on the Xa. Heterozygous deletion of Tsix in female cells causes the deletion‐bearing chromosome to be inactivated, suggesting that Tsix has a role in choice (Lee & Lu, 1999). Tsix is developmentally regulated by enhancers contained in the Xite and DXPas34 elements. Differential methylation of Xite and the CCCTC‐binding factor (CTCF)‐binding sites on DXPas34 correlate with X chromosome choice in mice (Boumil et al, 2006). The CTCF protein is associated with chromatin boundaries and has been proposed as a candidate factor involved in choice, although functional evidence is lacking (Chao et al, 2002). Consistent with this, deletion of DXPas34 results in ectopic X‐inactivation (Vigneau et al, 2006).
In male cells, Tsix is not required for silencing of the Xist promoter. The existence of a ‘competence factor’, which is produced in cells that have more than one X chromosome and predisposes females to initiate X‐inactivation by favouring Xist expression, has therefore been proposed (Lee & Lu, 1999). Consistent with this, an ectopic promoter inserted upstream of Xist has been shown to predispose the chromosome to be chosen for inactivation (Nesterova et al, 2003). Conversely, deletions within Xist also affect choice and cause inactivation of the other X chromosome (Marahrens et al, 1998). This indicates that Xist is regulated by positive signals and negative elements, which possibly form an epigenetic switch within Xic (Lee, 2005; Vigneau et al, 2006).
Recently, a model involving physical pairing of the Xic loci to mediate inter‐chromosomal communication has received experimental support from an analysis of the nuclear position of the Xic loci in cells undergoing X inactivation (Marahrens, 1999; Bacher et al, 2006; Xu et al, 2006). The Xic loci come within close proximity of each other just at the time when X inactivation is initiated (Fig 2B). This implies that pairing of the Xic loci might provide the necessary communication needed to establish the differential treatment of the homologous X chromosomes. Remarkably, ectopic insertions of Xite and Tsix transgenes can initiate de novo pairing, suggesting a role for these regulatory elements in trans (Xu et al, 2006). Whether the mechanism behind Xic pairing is related to the meiotic pairing of homologous chromosomes is unknown. It is also interesting to note that imprinted X inactivation in the extra‐embryonic lineages of some placental mammals depends on Xist (Okamoto et al, 2005). It is open to speculation whether Xist‐dependent imprinted X inactivation could have arisen after the evolution of random X inactivation in placental mammals or is an evolutionarily older form.
Using the Xist RNA to shut down an X chromosome
Once the choice is made, expression of the long functional Xist RNA is upregulated on the X chromosome to be inactivated (Chow et al, 2005). Xist RNA molecules then accumulate over the chromosome and initiate silencing. Gene targeting in mice has shown that Xist is required for both imprinted and random X inactivation (Marahrens et al, 1997; Penny et al, 1996), and that ectopic Xist expression in the absence of other Xic sequences can initiate chromosome‐wide silencing (Wutz & Jaenisch, 2000). However, initiation of silencing by Xist is restricted to the early stages of differentiation, implying that it is developmentally regulated. Initially, gene silencing is reversible and dependent on Xist; however, at a later stage in differentiation, X‐inactivation becomes independent of Xist and irreversible (Csankovszki et al, 1999). Hence, Xist is crucial for initiating silencing, but has a minor role in maintaining the Xi.
Xist localization does not require X chromosome‐specific sequences, and ectopic Xist expressed from autosomal transgenes can localize to autosomes and cause silencing, albeit to differing extents (Herzing et al, 1997; Lee et al, 1996; Wutz & Jaenisch, 2000). The observation that Xist spreading and the maintenance of silencing is less efficient outside the X chromosome instigated the idea of ‘way stations’ or ‘boosters’ at intervals along its length (Fig 3A). Owing to the high density of long interspersed elements (LINEs) on the X chromosome, these regions have been proposed as candidate boosters (Lyon, 2003). A recent study using cells with a t(X;4)37H X;autosome translocation showed that Xist spreading stops close to the X‐autosomal translocation point (Popova et al, 2006) and correlates with a severe drop in LINE density on chromosome 4. However, the mechanism underlying the function of distinct booster elements needs to be clarified.
The analysis of functional motifs within the mouse Xist RNA can provide insights into how Xist causes gene silencing. The association of Xist with the chromosome and its ability to trigger silencing are functionally separable (Wutz et al, 2002). Initiation of silencing depends on sequences at the 5′ end of Xist, which contains 7.5 repeats of a motif that is predicted to fold into an RNA structure comprising two stem loops. This structural motif is conserved among all placental mammals and might provide a binding platform for factors that act in gene repression. Xist RNA lacking this motif remains stable and accumulates within the chromosome territory, but no longer induces silencing. Hence, the ability to silence is not a prerequisite for Xist localization. The spreading of Xist RNA along the chromosome is mediated by functionally redundant sequences that are dispersed throughout the remainder of Xist and act synergistically (Wutz et al, 2002).
We have proposed a model based on the idea that these redundant sequences consist of several weak binding sites that facilitate binding to the chromosome and other factors in a cooperative manner. The binding of one factor allows easier binding of the next, resulting in a stable complex. This cooperative interaction drives the spread of Xist from its site of transcription, where a high concentration of Xist molecules is predicted to nucleate chromatin attachment (Fig 3B). Experimental evidence for the cooperative binding of Xist complexes is currently lacking. Nonetheless, this model is consistent with an added layer of complexity in an interphase nucleus. Considering that the Xic is positioned at the periphery of the chromosome (Chaumeil et al, 2006), Xist spreading must be directed towards the centre of the chromosomal territory (Fig 3C). Therefore, a spreading model based on the linear sequence of the X chromosome might be an oversimplification.
Xist is exceptional among cellular RNAs owing to its ability to encompass and silence an entire chromosome. Intriguingly, dosage compensation in flies involves the roX1 and roX2 RNAs, which target the dosage‐compensation complex to the single male X and mediate transcriptional upregulation (Lucchesi et al, 2005). RNA has been recently shown to be involved in the activation of Polycomb‐repressed genes in the fly (Sanchez‐Elsner et al, 2006). It seems possible that such an RNA might have been adapted at the onset of mammalian evolution for the emergence of Xist, although no ancestor has been detected to date (Duret et al, 2006).
Organization of the chromosome territory of the Xi
One way to ensure the recognition of only the elected Xi, and no other chromosome, is to physically reorganize and decorate the chromatin with distinguishable marks. During differentiation, the Xi forms a condensed and highly compacted heterochromatic structure that is characterized by the absence of intronic RNA and RNA polymerase II (Okamoto et al, 2004). Nongenic sequences, such as centromeric and genomic repeats, have been detected within the territory of the Xi, whereas genes are predominantly found at the periphery (Clemson et al, 2006). Recently it has been shown that Xist creates a transcriptionally silent nuclear compartment, which consists of intergenic and repetitive DNA, and that the 5′ end of Xist RNA is required for the organization of repressed genes into the nongenic silent domain (Chaumeil et al, 2006).
Reorganization of the Xi territory is also reflected in the recruitment of factors and chromatin modifications (Table 1; Chow et al, 2005; Lucchesi et al, 2005). Interestingly, scaffold attachment factor‐A (SAF‐A), which is known to bind satellite DNA and scaffold attachment regions/matrix attachment regions (SARs/MARs), is enriched on the Xi (Fackelmayer, 2005). SARs and MARs have been implicated in higher‐order chromatin organization. In the presence of nucleic acids, SAF‐A proteins form multimers; this process is a prerequisite for their specific binding to SAR sequences and indicates a cooperative binding mechanism. Xist RNA and SAF‐A are retained in nuclear matrix preparations after DNA and chromatin have been extracted (Fackelmayer, 2005). Intriguingly, both LINEs and MAR sequences are (A+T)‐rich, and some MARs have been shown to overlap with LINE repeats. Together, this evidence suggests roles for SAF‐A and the nuclear scaffold in X‐inactivation.
Finally, the Xi might be not only spatially but also temporally separated from other chromosomes. Replication asynchrony of the Xi relative to other chromosomes is observed with the Xi replicating early in S‐phase in preimplantation embryos and replicating late in embryonic cells after implantation (reviewed in Chow et al, 2005). Late replication of the Xi is observed after the onset of embryonic stem cell differentiation, but not in Xist‐expressing undifferentiated embryonic stem cells (Wutz & Jaenisch, 2000). Full reorganization of the chromatin of the Xi therefore takes place during cellular differentiation.
Setting up stable silencing with Polycomb
Once the silent state of the Xi is created in early development, the repressed status is maintained throughout subsequent cell divisions. Xist only has a minor role in the maintenance of the Xi, whereas multiple epigenetic marks, including DNA methylation, late replication and hypo‐acetylation of histone H4, act synergistically in the maintenance of X‐inactivation (Fig 4A; Lucchesi et al, 2005). The mechanism that underlies the transition between the initiation and maintenance phases is poorly understood, although recent findings suggest the involvement of PcG proteins in this process.
PcG proteins are conserved from flies to mammals and maintain transcriptional repression of developmental control genes (Fig 4B; Bantignies & Cavalli, 2006; Ringrose & Paro, 2004). Two biochemically distinct PcG protein complexes—Polycomb‐repressive complex 1 (PRC1) and PRC2—have catalytic activity. PRC1 and PRC2 mediate histone H2A Lys119 ubiquitylation (H2AK119ub1) and histone H3 Lys27 trimethylation (H3K27me3), respectively. As H3K27me3 enhances PRC1 binding to chromatin, PRC2 has been proposed to have a recruitment function (Bantignies & Cavalli, 2006).
PRC1 and PRC2 are recruited to the Xi early in X‐inactivation, establishing H3K27me3 and H2AK119ub1 marks along the Xi (de Napoles et al, 2004; Fang et al, 2004; Plath et al, 2003). PcG localization to the Xi and, as well as the persistence of the associated histone marks, is strictly dependent on Xist, irrespective of the developmental state of the cells (Kohlmaier et al, 2004; Schoeftner et al, 2006). Notably, a silencing‐deficient Xist RNA can recruit PRC1 and PRC2, showing that their recruitment alone is not sufficient to initiate X‐inactivation (Kohlmaier et al, 2004; Plath et al, 2003). These observations are in contrast to the traditional role of PcG proteins in the maintenance of gene repression, but suggest a function for the PcG complexes in the establishment of a chromatin structure on the Xi for the maintenance of X‐inactivation (Fig 4C).
The function of PRC2 in X inactivation has been studied by disrupting the embryonic ectoderm development (Eed) gene, which is an essential component of PRC2. In the absence of PRC2 function, H3K27me3 is no longer enriched on the Xi; PRC2 is not required for the initiation or maintenance of random X inactivation (Kalantry & Magnuson, 2006; Schoeftner et al, 2006). Moreover, PRC2 is required for the maintenance of imprinted X‐inactivation in differentiating trophoblast cells, but is dispensable for Xi maintenance in most cells of the extra‐embryonic lineages (Kalantry et al, 2006). An interesting observation is that Ring1b, but not Mph1 and Mph2, which are all members of the PRC1 complex, can be recruited by Xist in the absence of PRC2 and mediates H2AK119ub1 (Schoeftner et al, 2006). This implies that components of PRC1 and PRC2 can be recruited in parallel. Given the essential roles of PRC1 and PRC2 during early embryogenesis, as shown by the lethality caused by the disruption of Ring1b and Eed in mice, we propose that both complexes could act redundantly in X‐inactivation. Alternatively, PRC1 function could be crucial for X inactivation in embryonic cells, whereas PRC2 might be of less importance. How PcG proteins are recruited is yet to be resolved, and X‐inactivation might provide a good model to study the mechanism of Polycomb silencing from a new angle.
During the maintenance phase of X inactivation, Xist alone is not sufficient for the recruitment of PcG proteins; an additional factor, which we have introduced as chromosomal ‘memory’, is required (Kohlmaier et al, 2004). The molecular nature of this ‘memory’ is unclear. Xist expression early in differentiation establishes memory, and both Xist and memory are required for H3K27me3 and H2AK119ub1 on the Xi later in differentiation (Kohlmaier et al, 2004; Schoeftner et al, 2006). The establishment of memory is observed when the state of gene silencing becomes stable and irreversible. This suggests a role for memory in the transition from the initiation to the maintenance of X inactivation. A function for PcG complexes in the establishment of memory has been proposed, but awaits experimental confirmation.
X inactivation is a model for the developmentally controlled formation of silent chromatin. A number of pathways act in a stepwise manner to establish a stable state of repression. Progress has been made in understanding the molecular details of X inactivation, but the method by which Xist initiates silencing remains unknown. It is also unclear how the process of X‐inactivation intertwines with cellular differentiation, and the molecular details of the timing of counting and the restriction of Xist in initiating silencing remain to be defined. Future studies of X‐inactivation might therefore also lead to a better understanding of stem cells and of the process of cellular differentiation.
We apologize to all the authors whose work could not be cited. We thank L. Klein for critically reading the manuscript. This work was supported by the Institute of Molecular Pathology (IMP) through Boehringer Ingelheim, and by grants from the Austrian Ministry of Science Genome Research in Austria (GEN‐AU) project and the Austrian Science Fund (FWF).
- Copyright © 2007 European Molecular Biology Organization
Martin Leeb, Anton Wutz, Karen Ng & Dieter Pullirsch