Species Differences in TSIX/Tsix Reveal the Roles of These Genes in X-Chromosome Inactivation

Abstract

Transcriptional silencing of the human inactive X chromosome is induced by the XIST gene within the human X-inactivation center. The XIST allele must be turned off on one X chromosome to maintain its activity in cells of both sexes. In the mouse placenta, where X inactivation is imprinted (the paternal X chromosome is always inactive), the maternal Xist allele is repressed by a cis-acting antisense transcript, encoded by the Tsix gene. However, it remains to be seen whether this antisense transcript protects the future active X chromosome during random inactivation in the embryo proper. We recently identified the human TSIX gene and showed that it lacks key regulatory elements needed for the imprinting function of murine Tsix. Now, using RNA FISH for cellular localization of transcripts in human fetal cells, we show that human TSIX antisense transcripts are unable to repress XIST. In fact, TSIX is transcribed only from the inactive X chromosome and is coexpressed with XIST. Also, TSIX is not maternally imprinted in placental tissues, and its transcription persists in placental and fetal tissues, throughout embryogenesis. Therefore, the repression of Xist by mouse Tsix has no counterpart in humans, and TSIX is not the gene that protects the active X chromosome from random inactivation. Because human TSIX cannot imprint X inactivation in the placenta, it serves as a mutant for mouse Tsix, providing insights into features responsible for antisense activity in imprinted X inactivation.

Introduction

As a result of X inactivation, only one X chromosome is transcriptionally active in diploid somatic cells of both sexes (Lyon 1962); additional X chromosomes are inactivated. Transcriptional silencing is initiated during embryogenesis and can be recapitulated in murine embryonic stem (ES) cells (Martin et al. 1978; Lee et al. 1996). The key control region on the X chromosome is called the “X inactivation center” (XIC human, Xic mouse). It contains the X-inactive specific transcript gene (XIST human, Xist mouse [MIM 314670]) (Brockdorff et al. 1992; Brown et al. 1992), which programs the X chromosome for inactivation. The accumulation of XIST transcripts at the critical stage of embryonic development is essential for the cascade of events that ultimately silences the chromosome from which it is transcribed (Penny et al. 1996; Marahrens et al. 1997). The noncoding XIST RNA binds to the future inactive X (Beletskii et al. 2001), to propagate transcriptional silencing by the inducing of changes in chromatin (Jeppesen and Turner 1993; Wutz and Jaenisch 2000).

Having no functional X chromosome—or more than one (Takagi and Abe 1990)—is lethal for diploid cells of both sexes; therefore, one—and only one—X chromosome must be protected against XIST-induced inactivation. This is done by repressing the XIST allele on one X chromosome. In males, the single XIST allele is turned off. In females, the XIST allele is repressed on only one X chromosome, designating it the active X. The other X chromosome(s) in the cell is inactivated via the XIST-induced inactivating signal(s). The choice of parental XIST allele to repress is usually random in eutherian mammals. However, in mouse placenta, X inactivation is imprinted (the paternal X chromosome is inactive; the maternal X chromosome is active) ( Takagi and Sasaki 1975); the maternal Xist allele is always turned off, so that the maternal X chromosome is protected from inactivation (Lee 2000).

The prime candidate for the gene that represses the murine Xist allele in cis is the Tsix gene (MIM 300181), because (1) it encodes an antisense transcript that overlaps the entire Xist locus (Lee et al. 1999) and (2) its promoter includes a CpG island that is methylated on the active X chromosome (Debrand et al. 1999). Like Xist, Tsix lacks open reading frames, and the RNA remains in the nucleus; however, unlike Xist, the Tsix RNA is associated with the future active X chromosome (Lee et al. 1999). Tsix seems to be an antagonist of Xist in the mouse placenta, where X inactivation is imprinted. When Tsix RNA is eliminated from the maternal X chromosome by induced recombinant-based deletions that alter the Xic, the Xist allele on that chromosome is not repressed; this leads, in some cells, to inactivation of both X chromosomes (Lee and Lu 1999; Lee 2000; Sado et al. 2001). On the basis of such observations, Lee ( 2000) and Sado et al. ( 2001) have proposed that Tsix is the factor that protects the maternal X chromosome from imprinted X inactivation in placental tissues (by interacting with the Xist transcript on that chromosome). On the basis of deletion-induced skewing of X inactivation, Lee and her colleagues have further proposed that Tsix has a role in random inactivation as well, in that it chooses the active X chromosome (Boumil and Lee 2001; Stavropoulos et al. 2001), and that it is, along with the transcription factor CTCF, the regulatable epigenetic switch for X inactivation (Chao et al. 2002). It should be noted that these induced deletions are not Tsix-specific deletions but are genomic deletions that not only alter the Tsix transcript but also may alter other, yet-unknown regulatory elements as well. Therefore, although it is likely that murine Tsix has a role in imprinted X inactivation, a role for Tsix in random inactivation has yet to be established.

We have recently identified the human counterpart of murine Tsix, within the human XIC (Migeon et al. 2001). Like the mouse gene, TSIX is well expressed in embryo-derived cells, initiates downstream of the 3′ end of XIST, and produces an untranslated RNA. TSIX is transcribed (1) from human XIC transgenes in mouse ES cells and (2) in human embryoid body–derived (HED) cells. The TSIX RNA is transcribed from the strand opposite that from which XIST is transcribed and is, in part, antisense to XIST, overlapping the XIST locus at its 3′ end.

However, the human TSIX gene differs from its murine counterpart in that it lacks the CpG island that is essential for the function of murine Tsix (Lee 2000; Sado et al. 2001) (fig. 1). The antisense TSIX transcript is abbreviated; it overlaps only exons 5–8 of the XIST transcript, so that exons 1–4 and the XIST promoter are not covered (Migeon et al. 2001) (fig. 1). Such differences suggest that TSIX is unable to function like the murine gene—a supposition that is consistent with the fact that X inactivation is random, not imprinted—in human placental tissues (Migeon and Do 1979; Migeon et al. 1985).

Figure 1.

Partial maps of (a) human and (b) mouse XIC regions, comparing locations of XIST and TSIX transcripts and of CpG islands. Blackened boxes indicate XIST exons; unblackened boxes indicate CpG islands; and arrows show the transcripts and the direction of transcription. a, Human XIC. The relevant 18–80 kb of the U80460 DNA sequence (GenBank) is shown. h, m, and u are primer sets, and the gray-shaded boxes indicate the FISH probes. b, Mouse XIC. The relevant 60 kb is shown.

On the basis of RT-PCR assays, both XIST and TSIX have been shown to be transcribed in ES cells carrying six copies of the human XIC transgene (Migeon et al. 2001). Since both transcripts originate from the transgene—and, hence, form the same chromosome—it seemed likely that TSIX did not repress XIST expression. Yet, the possibility that XIST is transcribed in some cells and that TSIX is transcribed in others could not be excluded. Furthermore, we did not know the location oftranscripts in human cells. RNA FISH analysis permits us to determine the origin of these RNA transcripts at the cellular level. By simultaneously hybridizing probes for human XIST and TSIX to RNA in interphase nuclei from human fetal cells, we could see (1) which X chromosome(s)—active, inactive, or both—was the source of the TSIX transcript and (2) whether one transcript had an effect on the other.

Material and Methods

Cell Lines

Human

The HED cells were two 46,XX cell lines originating from female primordial germ cells at 5 and 11 wk after fertilization (and hereafter designated “LVEC” and “SDEC,” respectively). These fibroblast cultures, established by Dr. John Gearhart, are no longer totipotent but transcribe some genes expressed early in development (Shamblott et al. 2001). Our use of these HED cells was approved by the Johns Hopkins University committee concerned with patents and by Geron, Inc., according to the licensing agreement.

The somatic cells of human fetal origin were fibroblasts derived from discarded products of conception, at various gestational stages. Aliquots of frozen cells obtained for previous studies were the source of proliferating cells for the present study. Specimens are described in table 1 and include (1) chorionic villi cultures of normal newborn placentas (five female and one male), (2) fetal fibroblast cultures from 5–18 wk gestation (six female and one male), and (3) fibroblasts from a hydatidiform mole (K1248), with two copies of the identical paternal X chromosome (Jacobs et al. 1980). Also analyzed were fibroblasts obtained, with Institutional Review Board approval, from individuals ranging, in postnatal age, from 2 mo to 25 years (eight females and one 49,4XY male). Clones originating from single cells were obtained by cell dilution (Migeon et al. 1985), and parental origin of the active X chromosome was determined on the basis of parental X-chromosome markers.

Table 1.

TSIX Expression in Human Somatic Cells, on the Basis of RNA FISH and RT-PCR Analysis

		FISH Signal [No. of Signals per Cell] ofc(%)		Status with TSIX Primer Setd,e
Source and Agea	Reference Lineb	XIST	TSIX	h	m	u
Embryoid bodies:
5 wk	LVEC, XX	72	40	+	+	+
11 wk	SDEC, XX	76	32	+	+	+
Fetal fibroblasts:
5–8 wk	ET-2, XX	92	86	+	+	+
	K-19, XX	74	52	+	+	+
	Fe18, XX	82	36	+	+	+
	K-1, XY	0	0	−	−	−
Hydatidiform moles:
10 wk	K1248, XpXp	88	68	+	+	Trace
Chorionic villi:f
Birth	CV-5, XX	78	42	+	+	+
	CV-4, XY	0	0	−	−	−
	PG, XXX	82 [2]	64 [2]	+	+	+
Postnatal fibroblasts:
Birth	TW-1, XX	62	40	ND	ND	ND
2 mo	TW-2, XX	94	32	+	Trace	+
	Clone 1g	92	46	ND	ND	ND
	Clone 2h	86	30	ND	ND	ND
2 years	HJ, XXXXY	72 [3]	6 Trace [3]	Trace	−	−
4 years	TaW, XX	80	26 Trace	+	Trace	−
4 years	CW, XX	79	0	ND	Trace	−
8 years	AB, XX	84	22 Trace	+	Trace	Trace
20 years	PS, XX	86	0	−	−	−
24 years	CC, XX	90	0	−	−	−
25 years	MF, XX	82	0	ND	−	−

^a Age is from conception, as estimated from LMP and fetal measurements.

^bIncludes sex-chromosome complement of specimen, X and Y; Xp = paternal X chromosome.

^cPerformed with XIST- and TSIX-specific probes simultaneously. For most postnatal fibroblasts, the TSIX signals were present in trace amounts. A few cells had a TSIX signal in the absence of an XIST signal (average 0.3% [range 0%–8%]).

^dRT-PCR primers are as shown in figure 1. Strand-specific priming was used for XIST exon 5 primer set h; primer sets m and u are TSIX specific.

^e+ = Present; − = absent; ND = not done.

^fFrom full-term placentas.

^gInactive maternal X chromosome.

^hInactive paternal X chromosome.

Mouse

The ES cells with the human XIC transgene (ES-10) were derived from J1 ES cells (Migeon et al. 1999). ES-10 has six copies of the 480-kb transgene that includes the human XIST locus, ∼70 kb flanking the 3′ end of XIST, and ∼380 kb flanking 5′ XIST inserted into mouse chromosome 11. The ESch-10 somatic cells were fibroblasts from an 18-d-old stillborn XY chimeric mouse 949-1, carrying the ES-10 transgene (Migeon et al. 1999).

RNA FISH

The procedure was as described elsewhere (Clemson et al. 1996). In brief, interphase cells on slides were permeabilized, fixed in paraformaldehyde, and dehydrated. Then, without denaturation, the slides were hybridized, for 3 h, with labeled probes, along with human COT-1 and salmon sperm DNA. The DNA probes were labeled, by the nick-translation kit (Vysis), with either biotin-11-dUTP or digoxygenin-16-dUTP (Roche), and the signals were detected with either avidin/biotinylated anti-avidin or anti-digoxygenin-rhodamine, respectively, with DAPI as the counterstain.

FISH Probes

The probes were long-range PCR products that were cloned into the pCR2.1-TOPO vector (Topo TA kit; Invitrogen): All probes come from nonoverlapping segments of the XIST and TSIX genes. XIST exon 1 is specific for XIST and contains the ∼2.5-kb DNA segment of nt 24658–27215 in the PAC U80460 sequence (GenBank). TSIX-specific probes are TSIX-70 (∼1.7 kb; nt 69086–70779) and TSIX-57 (2.9 kb; nt 55348–58274), both in U80460.

RT-PCR

RNA was obtained from ES cells before and after differentiation into embryoid bodies. Total RNA was isolated with TRIZOL (Invitrogen) and was treated with DNaseI. To screen for transcripts, RT-PCR was performed (1μg of RNA; 35 cycles [Migeon et al. 1999]). The cDNA was generated by random priming, for sequences outside the XIST transcription unit, whereas strand-specific priming was used for the portion of TSIX that overlaps XIST (Migeon et al. 2001).

Primers

Sequence data for TSIX and the XIC transgene are provided in the GenBank entry for PAC 92E23, which has been sequenced in two parts, U80459 and U80460. The locations of primers used for RNA analysis are shown in figure 1 and are based on the U80460 sequence: for primer set h, the start of the antisense primer is at nt 41838 (136-bp product); for primer set m, the forward primer starts at nt 57793 (482-bp product), and, for primer set u, the start is at nt 70420 (377-bp product. For strand-specific RT-PCR, first-strand cDNA was synthesized by either sense (XIST) or antisense (TSIX) primers, as described elsewhere (Migeon et al. 2001).

Results

TSIX Coexpressed with XIST from the Inactive X Chromosome in ES Cells

Figure 1a shows the location of TSIX and XIST transcripts in the GenBank U80460 sequence, as well as the FISH probes used for these studies. These probes were gene specific, since they were homologous to the nonoverlapping segments of XIST and TSIX. When hybridized independently, the two TSIX probes, TSIX-57 and TSIX-70, gave similar signals, but only the studies with TSIX-70 are shown. The results of RNA FISH, using simultaneous hybridization of digoxygenin-labeled XIST and of biotin-labeled TSIX probes, were the same for mouse ES cells carrying the ES-10 human transgene, ES-10, and for human female HED cells of germ-cell origin (LVEC). (fig. 2a). In each case, there was only one XIST signal (red) and one TSIX signal (green) per cell. The XIST signal was large, resembling that in human adult female cells (fig. 2a, “CC” row). The TSIX signal was also substantial, indicating many molecules per nucleus. It was as well focused as the signal of XIST; however, it was usually smaller than the signal of XIST (fig. 2, “TSIX” and “Merge” lanes), irrespective of the nature of the fluorochromes used. When the TSIX and XIST images were merged, the two signals always coincided (yellow signals denote overlap), as would be expected if the two RNA samples originated from the same XIC. Both signals in mouse ES-10 cells were seen before and after differentiation into embryoid bodies and came from the mouse chromosome 11 carrying the human XIC transgene. In the human cells, both signals came from the inactive X chromosome, visualized as a Barr body in DAPI-stained interphase cells (not shown). The frequency of TSIX signals was usually less than that for XIST signals (table 1), reflecting some variability in TSIX expression. Rare (frequency <5%) cells had a single TSIX signal in the absence of XIST. Figure 2a (“CC” row) shows that TSIX was not detectable in somatic cells from an adult female.

Figure 2.

Simultaneous hybridization, in situ, of XIST- and TSIX-specific probes to nuclear RNA. In panels a, b, and d, the probes are XIST exon 1 (red) and TSIX-70 (green); overlapping signals (yellow) are seen in merged images. a, TSIX and XIST coexpressed in mouse ES-10 cells carrying the human transgene ( row ES-10), human HED cells (row LVEC), and 46,XX adult human female fibroblasts (row CC). b, Persistence of TSIX transcripts in human female cells from fetus, placenta, and neonate: 5–6 wk gestation (row ET-2), 6.5–8 wk gestation (row Fe18), 2 mo after birth (row TW-2), and full-term placenta, 47,XXX (row PG). c, TSIX expressed from the inactive X chromosome. Sequential TSIX RNA/XIST DNA hybridization showing three XIST alleles (red) but only two TSIX RNA signals (green) in PG (47,XXX) placenta. The arrow indicates the active X chromosome lacking the TSIX signal. d, TSIX expressed from the human inactive X chromosome, of either parental origin, in hydatidiform mole (K1248, with inactive paternal X chromosome) and in clones 1 and 2, from specimen TW-2, that have either the maternal X chromosome (Xi_mat) or the paternal X chromosome (Xi_pat) as the inactive X chromosome, respectively.

Persistence of TSIX Expression in Somatic Cells

Before X inactivation, murine Tsix is transcribed, to a modest extent, from both X chromosomes; at the time of inactivation, the allele on the inactive X chromosome is repressed, whereas that on the future active X chromosome is up-regulated transiently, disappearing from ES cells 2–9 d after their differentiation into embryoid bodies and from chimeras 12–15 d after conception (Lee et al. 1999; Sado et al. 2001). However, the human TSIX signal was present in 56% of cells from a chimeric mouse fetus 18 d after conception (data not shown).

TSIX expression also persists in human fetal somatic cells long after X inactivation is established. Figure 2b shows the monoallelic expression of TSIX in 46,XX fetal specimens (ET-2 and Fe18) with gestational age 5–8 wk. Characteristically, the TSIX signal was associated with an XIST signal at the same location. Most (74%–92%) cells had an XIST signal, and many (36%–86%) also had a TSIX signal. This signal varied in size, like that for XIST, but, in merged images, it usually did not cover the entire XIST signal (see fig. 2, “Merge” lane). Both signals overlie the sex chromatin body.

In addition, TSIX was expressed from the inactive X chromosome in female placental tissues at term (table 1 and fig. 2b, “PG” row). As would be expected for an inactive X-associated transcript, TSIX was never seen in male placental cells. The placental cells from a newborn with a 47,XXX karyotype had two TSIX and two closely associated XIST signals per cell, consistent with the presence of two inactive X chromosomes (fig. 2b, “PG” row). Sequential TSIX RNA hybridization followed by XIST DNA hybridization in these cells showed three X chromosomes but only two TSIX RNA signals, confirming that the active X chromosome did not express TSIX (fig. 2c, “PG” row).

Because TSIX was expressed in all the female fetal specimens and in chorionic villi at birth, we examined postnatal specimens to determine when the TSIX transcript disappears. Significant expression persisted in skin fibroblasts of a 2-mo-old female (fig. 2b, “TW-2” row). TSIX RNA was also present in trace amounts in specimens from 4–8-year-old females and from a 2-year-old 49,XXXXY male, but it was extinguished in young adults (table 1).

TSIX Expression—Monoallelic but Not Parentally Imprinted

That TSIX was monoallelically expressed and that there were fewer TSIX signals than XIST signals (table 1) raised the possibility that TSIX expression might be imprinted—coming only from the maternal X chromosome, as in the case of mouse placental cells. Therefore, we looked at TSIX expression in cells from a hydatidiform mole, a placental tumor, whose chromosomes come exclusively from sperm. Our previous analysis of these 46,XX cells showed that both X chromosomes were of paternal origin and, in fact, had been derived from duplication of a single paternal X chromosome (Jacobs et al. 1980). The single XIST signal in these cells showed that one of the identical X chromosomes was transcriptionally inactive. It was associated with a TSIX signal in 68% of the cells (table 1 and fig. 2d). We also studied two clones obtained from TW-2 skin fibroblasts—one with the maternal inactive X chromosome and the other with the paternal inactive X chromosome, in all the cells (see the “Material and Methods” section). The TSIX signal was seen in both clones (fig. 2d, “Clone 1:Xi mat” and “Clone 2:Xi pat” rows), showing that TSIX is expressed from an inactive X chromosome of either parental origin.

RT-PCR Analysis of TSIX Expression

Many of the same specimens were analyzed by RT-PCR using either primer sets unique to TSIX (see fig. 1) or strand-specific priming of the cDNA (fig. 3). The results in table 1 show that TSIX RNA is in fetal specimens for as long as 2 mo after birth. TSIX RNA was detected neither in adult control cells (human male and female) nor in the fetal or placental cultures established from males (table 1 and data not shown). Some of the postnatal specimens had traces of RT-PCR products, confirming the results of the FISH studies.

Figure 3.

Results of RT-PCR analysis, showing persistence of TSIX expression in human female cells of fetal and placental lineages: strand-specific priming of cDNA and amplification with primer set h, for XIST exon 5 (136-bp product) in the presence (+) or absence (−) of reverse transcriptase (RT). Top, Antisense (TSIX). Bottom, Sense (XIST). Specimens are from an adult male (lane 1), an adult female (lanes 2 and 3), mouse ES-10 cells with the transgene (lanes 4 and 5), a female HED cell, LVEC (lanes 6 and 7), placentas CV2 (lanes 8 and 9) and CV5 (lanes 10 and 11), and fetal fibroblasts at 12 wk (lanes 12 and 13) and at 18 wk (lane 14). Extra bands for ES-10 are of mouse origin.

Discussion

We have studied the expression of human TSIX in cells analogous to those used for the study of the mouse Tsix gene, employing similar methods and conditions. The results of our analysis show differences between the two antisense transcript genes. They differ in respect to (1) the chromosome from which the gene is transcribed (Tsix, active X chromosome; TSIX, inactive X chromosome) and (2) the ability to repress XIST/Xist (Tsix, yes; TSIX, no). Our observations provide compelling evidence that human TSIX is not the gene that protects the human active X chromosome against XIST-induced inactivation, because it cannot repress the XIST allele in cis. The TSIX locus is coexpressed with XIST from one and the same inactivation center in ES-10 cells and from the inactive X chromosome in human germline–derived HED cells (fig. 2a) and in fetal somatic cells of females (fig. 2b). In addition to its failure to inhibit XIST expression, TSIX does not prevent inactivation of the chromosome from which it is transcribed; in fact, TSIX is expressed uniquely from an inactive X chromosome and clearly does not interfere with the silence of this chromosome. In contrast to the tightly regulated control of murine Tsix, the human gene is subject to prolonged and variable expression, persisting throughout gestation and even beyond. TSIX expression certainly is not harmful to cell viability. Because TSIX in normal human fetal cells is transcribed from the inactive X chromosome and not the active one, and because its abundant transcript does not prevent XIST expression, it seems that Tsix repression of murine Xist has no counterpart in humans.

These observations, considered along with the fact that human TSIX lacks the critical elements of murine Tsix (i.e., the CpG island and complete overlap with XIST), strongly suggest that human TSIX not only differs from its mouse counterpart but also is a defective gene. Most likely, TSIX is an evolutionary vestige of an ancestral gene and never functions in X inactivation in human cells. What remains to be explained is why, during embryogenesis, TSIX is monoallelically transcribed exclusively from the inactive X chromosome. TSIX expression is not parentally imprinted, because transcripts come from the inactive X chromosome of either parental origin (fig. 2d). Perhaps chromatin changes responsible for—or induced by—XIST transcription allow the expression of this evolutionary vestige; alternatively, the chromatin changes on the future active X chromosome, responsible for repression of XIST, may also inhibit the expression of this ineffective antisense transcript. More difficult to explain is the gradual repression of the TSIX allele on the inactive X chromosome, after birth. Conceivably, this is due to concomitant changes in XIST expression and/or chromatin confirmation of the inactive X chromosome that have not been recognized until now.

Species differences in details of developmental events provide powerful tools for the dissection of these complex processes. Because human TSIX is unable to repress XIST, it is mutant with respect to its murine counterpart, and it affords insights into mechanisms of Xist repression by Tsix and of antisense function in parental imprinting. The mechanism most compatible with our observations is that Tsix repression of Xist in the mouse placenta requires interaction between Tsix transcripts and the Xist promoter, as well as the presence of the Tsix CpG island. The absence of the CpG island in the human TSIX gene, considered along with the inactivating effect of similar deletions induced in the murine gene, indicates that this Tsix element is essential for antisense activity. On the basis of a survey of imprinted genes on human chromosome 11, Onyango et al. (2000) have suggested that imprinting requires the presence of two CpG islands, one of which is associated with the antisense transcript. That the abbreviated TSIX transcript does not overlap the XIST promoter or its CpG island undoubtedly contributes to its impotence as an XIST inhibitor. Experimental evidence supporting this conclusion comes from studies by Luikenhuis et al. (2001), who have shown that truncation of the Tsix transcript, even when the CpG island is intact, interferes with the ability of the mutant Tsix to inhibit Xist expression in mouse ES cells.

If TSIX is unable to repress the XIST locus in cis, then it cannot be the key gene that protects the active X chromosome against programmed inactivation in human cells. Evidence that mouse Tsix may also not be sufficient for random X inactivation comes from studies of ES cells with a 65-kb genomic deletion that removes ∼15 kb of Tsix DNA, eliminating the Tsix RNA; targeted site-specific replacement of Tsix in these ES cells does not restore random X inactivation (Morey et al. 2001). However, the deletion is large, and restorations may perturb the chromatin in the region. Nonetheless, another candidate (or other candidates) for this function needs to be identified. It is difficult to envision how an X-chromosomal gene could, by itself, regulate the function of an XIST allele on only one of the members of the X-chromosome pair. The ultimate regulator of XIST/Xist (and of Tsix) is probably encoded by an autosomal gene (Lyon 1971; Jacobs and Migeon 1989), is limited in quantity, and is subject to cooperativity. Such a trans-blocking factor needs to find only one X chromosome at random, and, once a sufficient number of molecules bind to one X chromosome, the XIST allele on that chromosome will be blocked. One might predict that more of this trans-blocking factor might protect more than one X chromosome, and this is the case in triploid specimens (i.e., three sets of autosomes and either an XXY or XXX sex-chromosome complement), in which two X chromosomes can remain active (Weaver et al. 1975; Jacobs et al. 1979; Migeon et al. 1979).