The evolution of unequally sized X/Y sex chromosomes posed a major mathematical dilemma for mammals. While females have two X chromosomes (XX), males have one X and one Y chromosome (XY). Because the Y chromosome contains just a small fraction of the 2,000 genes on the X chromosome, this system of sex determination endows the female with nearly twice the number of sex chromosome genes as the male. As first described by Mary Lyon 40 years ago, X-chromosome inactivation (XCI) evolved to compensate for this dosage imbalance by silencing one whole X chromosome in XX individuals. Dosage compensation is essential for proper embryonic development, as embryos that fail to achieve it differentiate poorly and die at the time of uterine implantation. This presented a challenging problem for the mammalian cell, because it required that the cell be able to mark epigenetically two chromosomes that otherwise share a nucleoplasm and that all genes on one X be inactivated without affecting genes on the other X. Since the earliest days of its discovery, XCI was known to be controlled by a master activity switch on the X. To understand how the process is controlled, we have taken a genetic approach in mice, coupling a whole-organism model with an embryonic stem cell model that is capable of recapitulating X inactivation in culture.

Two forms of XCI have been described. In stochastic XCI, either the mother's or father's X chromosome can be silenced. Just after implantation, cells that form the embryo proper count the number of X chromosomes and undergo XCI only if there is more than one X chromosome. In XX cells, a choice mechanism designates one active and one inactive X, with each X having an approximately equal chance of being inactivated. In contrast, during imprinted XCI, the father's X chromosome is inactivated in all cells. This form is believed to have evolved first and is still found in extant marsupials, such as the kangaroo, and in the extraembryonic tissues of some placental mammals such as the mouse. In contrast to random XCI, the imprinted process bypasses counting and choice so that the only chromosomes capable of being silenced are those inherited from the father. We are interested in understanding both the mechanism underlying XCI and how XCI and imprinting evolved in mammals during the past 300 million years.

Mechanisms: Antisense RNA, RNA-Directed Silencing, and X-Chromosome Pairing
XCI is controlled by an X-linked inactivation center (Xic), a region of ~100 kb that harbors several noncoding genes, including Xist, Tsix, and Xite. Chromosome-wide silencing requires Xist RNA, which spreads along the X and recruits silencing factors. In female cells, Xist RNA envelopes the inactive X, appearing as a "cloud" of transcripts by RNA fluorescence in situ hybridization. Xist is regulated in cis by its antisense partner, Tsix, which is responsible for keeping the X chromosome active by blocking the expression of Xist on the future Xa (active X). Tsix's down-regulation on the future Xi (inactive X) enables Xist RNA to be induced for the first time and spread along the X. Tsix is in turn regulated by Xite, an upstream locus harboring multiple intergenic transcripts and a Tsix-specific enhancer.

We are beginning to understand how Tsix and Xite work together in cis to inhibit the action of Xist. Following counting and choice, Tsix's action on the linked Xist allele determines whether X inactivation will initiate on that chromosome. We have shown that Xist is controlled by Tsix RNA-directed chromatin change in two ways. First, Tsix mediates a type of RNA-directed DNA methylation. On the future Xa, Tsix transcription across the Xist locus recruits the de novo DNA methyltransferase, Dnmt3a, to the linked Xist allele. The second mechanism of action is surprising. On the future Xa, Tsix expression induces euchromatin in Xist, which paradoxically represses Xist. On the future Xi, Tsix down-regulation induces a transient heterochromatic state in Xist, followed paradoxically by high-level Xist expression. These "flip-flop" chromatin dynamics preempt and predict asymmetric Xist expression. These data indicate that Xite and Tsix control Xist expression through complex cis interactions at the X-inactivation center. Our current work is directed at understanding the molecular details.

Although XCI is known to be regulated in cis by Xite, Tsix, and Xist, the two X chromosomes must in principle also be regulated in trans to ensure distinct X-chromosome fates—one on and the other off in a mutually exclusive manner. We have demonstrated that interchromosomal pairing mediates this communication. Pairing between the X chromosomes occurs transiently at the onset of XCI and is specific to Xic sequences, rather than the whole X chromosome. Deleting Xite and Tsix perturbs pairing and counting/choice, while their autosomal insertion induces de novo X-autosome pairing. Ectopic X-autosome interactions inhibit endogenous X-X pairing and block the initiation of XCI. Thus, the noncoding Tsix and Xite genes work also in trans by nucleating X-X pairing and coordinating counting and mutually exclusive choice. We do not know how the X chromosomes are brought together and what message is communicated between the two. These problems are the subject of further investigation.

Hypothesis: Preinactivation as a Mechanism of Imprinting
For much of the past 30 years, it was thought that female embryos begin development with two active X chromosomes and X inactivation does not take place until uterine implantation. In this classical view, female embryos have twice the dose of the X chromosome than male embryos throughout the preimplantation period. However, because some observations have been difficult to explain by the classical model, we have used technologies not available in the 1960s and 1970s to reexamine various aspects of the traditional view.

We have now shown that one X chromosome is already silent at zygotic gene activation (2-cell stage). The chromosome is paternal in origin and is not uniformly silent. Genes close to the X-inactivation center show the greatest degree of silencing; more distal genes can partially, if not entirely, escape silencing. After implantation, imprinted silencing in the extraembryonic tissues becomes globalized and more complete on a gene-by-gene basis. These results argue that imprinted XCI is biphasic in nature: preimplantation embryos exhibit an incomplete form of silencing and postimplantation embryos display more complete silencing.

We propose that the female embryo is in fact dosage compensated from the time of conception along much of the X chromosome. Furthermore, we suggest that this imprinted form can be traced back to meiotic sex chromosome inactivation in the male germline, in which the X and Y chromosomes are transcriptionally silenced by virtue of their being unpaired at pachytene. Indeed, further work has shown that the spermatid X chromosome remains silent after meiosis and that it remains so throughout the postmeiotic period. This postmeiotic sex chromatin (PMSC) forms a distinct subnuclear compartment in the spermatid and adopts a distinct heterochromatic character. Thus, there appears to be a continuity of silencing from the paternal germline to the female conceptus. We argue that the silent paternal X in early embryos originates in the male germline and is inherited by the zygote as a presilenced chromosome.

In view of this hypothesis, we suggest that the marsupial form of imprinted XCI is mechanistically related to the preinactivation process evident in the mouse. A preinactivation mechanism based on meiotic silencing of unpaired chromosomes would have easily solved the problem of dosage compensation in the earliest mammals. This type of mechanism requires no counting or choice and silences the supernumerary X (from the father) in the female zygote (the males are untouched because they inherit a Y from the father). We are exploring the nature and character of the imprint in the paternal germline and the evolutionary trajectory of imprinting in mammals.