Review of:

Role of the Xist Gene in X Chromosome Choosing

York Marahens, Jan Loring, and Rudolf Jaenisch,

The process of fertilization normally leaves female mammals with 2 copies of the X chromosome. Sometime during the development of each somatic cell, one of these X chromosomes is inactivated, by conversion to heterochromatin that is not expressed. Normally this process is random, but there are various theories about how the cell might choose one chromosome or the other. This paper looks at experimental evidence and explores some models of a mechanism for the inactivation of one X chromosome in female mammals.

With this background in mind, the general theory of X chromosome inactivation involves the following nomenclature, given by the authors in their background. The x-controlling element (Xce) is thought to be the site involved in inactivation. This site has been mapped to the x inactivation center (Xic), a region of the chromosome required in two copies per cell for inactivation (par1). (par2). X chromosomes with the same Xce allele have an equal chance of being chosen, but in heterozygous cases certain alleles are more likely to be chosen than others. The XIST gene, which has been mapped into the Xic but seems to be separate from the Xce, is necessary for inactivation and codes for a 15kb untranslated RNA which coats the inactive chromosome. An X chromosome with an Xce but no Xist can be selected for inactivation, but cannot inactivate, indicating that Xist acts downstream from Xce. Xist has also been shown to be required for heterochromatin formation.

Rodents have the fortunate trait whereby their extraembryonic tissue always inactivates the paternal X chromosome. The authors use this fact in earlier research to prove that Xist acts downstream of the imprinted selection, because mice with a paternally inherited Xist deletion fail in embryogenesis due to the presence of 2 active X chromosomes.

They also found that females who receive the altered Xist gene from the maternal side grew up with the wild type paternal chromosome only inactivated. Since the paternal gene was wild type, it could be inactivated and the embryos survived emryogenesis, with its imprint selection. However, the development of somatic tissues should have used random selection of an X chromosome, according to the leading models before the publication of this paper. This model would explain this result by saying that wherever the mutant chromosome was selected, the presence of 2 X chromosomes in the cell would cause the loss of this cell from developing embryos, which might survive despite massive cell loss. Exactly this has been shown to happen in the case of SearleÕs X-autosome translocation T(X;16)16H), with these females being born at a low birth weight because of the cell loss during differentiation.

The paper under review seeks to show that mice heterozygous for a maternally inherited Xist mutation have undergone "primary nonrandom inactivation", with wild type chromosome being selected and inactivated in every cell. Their first evidence of this is that the pups from the cell line with Xist+/- females are exactly the same weight as their wild type sibs. This is in stark contrast to female pups born which are heterozygous for SearleÕs translocation, which are all 79% of the weight of their siblings (see figure 1).

Figure 2 shows a comparison of Xist +/- and wt embryos at various stages of development. It seems sufficent to show that the two types of embryos are the same size during development. One might doubt that people could really tell the difference between embryos, even those differing by only 21%, as with the SearleÕs translocation. For this reason, it would be more convincing to show control pictures the SearleÕs low birth weight embryos. I am, however, willing to accept their conclusion that the mutant embryos are the same weight as wild type embryos.

Thinking that loss of possible XX cells from the Xist+/- cells could still be very gradual, but occurring anyway, the authors perform RNA FISH on embryonic cells after 7.5 days of gestation. E7.5 cells were chosen because at this point of gestation, 85% of normal cells are "high-level monoallelic", expressing the Xist RNA product from only one chromosome (coating and causing inactivation) while 15% are "differential biallelic", still expressing two forms of Xist, with one very major and one very minor. The authors state that this means that inactivation is complete at E7.5, a claim that cannot be directly supported by the information in this paper. However, they may have other reasons to believe that high expression of Xist occurs on inactivated chromosomes, and it is not crucial to the paper since it at least indicates that a particular chromosome has been set down the path of inactivation.

The authors interpret their FISH data as follows: control Xist^- males are "low-level monoallelic", or express a small amount of mutant Xist RNA, because a probe to the full Xist gene binds to a small portion of the cell, as in Figure 4a. This is the same for wild type males (not shown in paper). Control Xist +/+ females produce the expected mixture of high-level monoallelic Xist and differential biallelic Xist, because a full Xist probe bound in this fashion (Figure 4c). Experimental Xist+/- females have differential bialleleic expression and high-level monoallelic expression, because a full probe binds in that pattern (Figure 4d). Since the authors have proven elsewhere that Xist RNA cannot coat chromosomes and cause the large signal as in differential biallelic and high level monoallelic expression, this panel should show that the mutant Xist bearing chromosome was never selected for inactivation. To prove this, the authors conduct more control experiments, this time with "deletion probes", which will only recognize wild type Xist transcripts. If cells underwent the random choice mechanism, we would expect cells half of the cells to have high-level monoalleleic expression of wild type Xist, and the other half to have low-monoallelic expression. To demonstrate that their method could detect low-level monoallelec expression, they exposed the deletion probe to wild type male embryos and were able to see a small dot (not shown). It is acceptable that this data is not shown if the authors did not see anything similar in the experimental results. To demonstrate for sure that the deletion probe would not bind to the deleted Xist gene, the authors show in Figure 4b that there is no binding using Xist^- male embryos. This is not a false negative due to mistaken proceedure because a probe to Pgk-1 binds. The authors say that using the deletion probe on Xist +/- cells shows only high-level monoallelic expression. This is the expected result from primary non-random slection, proving that this is what has occurred in the cells.

The photographs figure 4 do indeed support the authorsÕ conclusion about the binding pattern of the probe to the Xist gene. The only dubious aspect of the photographs could be panel E, which seems to show a tiny dot of pink next to the large expression area. This could be indicative of differential bialleleic expression, but the dot is small enough that it could in fact be a printing error, or artifact of the FISH method. This seems more likely as a similar dot is present in a male cell in panel A, and the dubious dot is distinct from the small signal from low-level monoallelic expression. The only other dubious piece of information is the finding that 466/470 Xist +/- cells gave high-level monoalleleic expression using the deletion probe. The thery would have predicted that 100% of the cells would give high-level monoalleleic expression, and while the authors state that none of the 4 deviant cells was differential biallelic, it would be nice to know what had happened in those cells. However, they seem statistically insignificant, (99% instead of 100%) and so I accept the authorsÕ conclusion.

The evidence discussed above indicates that the authors have found a positive element that is required for chromosome inactivation. This disproves the previous theory which held that chromosomes were chosen by a negative element which blocked one chromosome from being inactivated and caused the other to be inactivated instead.

Different deletions in the Xist gene have different effects on the cell. A different group of researchers found that a large deletion in the Xist gene, from 3Õ to exon one (the promoter) to part way through exon one. This deletion, which causes no transcription the Xist locus to take place, does not interfere with random choosing. Therefore, the authors propose that the random choice mechanism depends on a DNA element which involves one of the introns that is deleted in their mutant Xist gene.

Important to the proposed model is the fact that female cells carrying more than two X chromosomes will inactivate all but one. There seems to be a typographical error in the paper, however, with the conflicting statements on page 661 that "tetraploid (2n) female cells inactivate two X chromosomes and keep twoÉ active", but "XXXX individuals [must produce] three [initiation factors]". The authorsÕ proposed mechanism can only make sense for supernumary X chromosome cells if all but one X chromosome is inactivated. Using this assumption, the authors propose a mechanism involving a single inactivation blocker which binds to the Xist gene, switching back and forth between chromosomes, and several initiation factors for inactivation. These 2 factor types bind to different portions of the Xist gene, but are mutually exclusive to each other. The authors do not make it clear that these factors must bind to different points in the gene, but it seems that this is required for the mechanism to make sense. In the authorsÕ deletion, the initiation factor would be unable to bind to the mutant gene, and so be forced to be in high competition for the site on the wild type chromosome. This would leave only the mutant allele open for the binding of the blocking factor.

The authors also point out the failings of a simpler model that uses only a single initiation factor to select chromosomes for inactivation. This model cannot explain the inactivation of more than one chromosome, as does occur in nature.

I think that the model which the authors propose is plausible, but it seems that it could be tested by further research. I propose that the next step would be to identify possible factors associated with the model. Affinity chromatography could be used with portions of the Xist gene bound to oligos. Researchers could characterize the factors which bind to the DNA. Cells could be pulse labeled before the collection of the initiation factors, using radioactive Sulfur if the factors turned out to the proteins, or other radioactive materials if the factors turned out to be of other nature. These radioactive purified products could be used to probe a yeast expression library (yeast because the factors come from mammals, and yeast would provide the most similar environment in which to create the library) in order to clone the genes responsible for creating the factors. Methods could also be employed to determine which portions of the Xist gene bind to which factors. Affinity chromatography could be employed as discussed above.

Direct correspondence to: bebuxton@davidson.edu