X Inactivation Analysis and DNA Methylation Studies of the Ubiquitin Activating Enzyme E1 and PCTAIRE-1 Genes in Human and Mouse

Abstract

Previously reported data on the X inactivation status of the ubiquitin activating enzyme E1 (UBE1) gene have been contradictory, and the issue has remained unsettled. Here we present three lines of evidence that UBE1 is expressed from the inactive X chromosome and therefore escapes X inactivation. First, by RNA in situ hybridization, UBE1 RNA is detected from both the active and inactive X chromosomes in human female fibro-blasts. Second, UBE1 is expressed in a large panel of somatic cell hybrids retaining inactive human X chromosomes, including two independent hybrids that did not require UBE1 expression for survival. And third, sites at the 5′ end of UBE1 are unmethylated on both active and inactive X chromosomes, consistent with the gene escaping inactivation. In order to address whether other genes that escape inactivation map to the same region of the X chromosome, we have also examined the expression of genes mapping adjacent to UBE1. The gene for PCTAIRE-1 (PCTK1) maps within 5 kb of UBE1 and similarly escapes X inactivation by the somatic cell hybrid assay, whereas six other genes that are within 1 Mb of UBE1 in Xp11.23 are silenced on the inactive X chromosome. Comparative mapping studies of the homologous loci in mouse establish that Ube1-x and Pctk1 are also within close physical proximity on the murine X chromosome, and expression studies of the Pctk1 gene determine that, similar to Ube1-x, it is subject to X inactivation in mouse. Methylation of CpG residues at restriction sites at the 5′ end of both genes on the murine inactive X chromosome is consistent with both genes being subject to X inactivation in mouse, in contrast to their expression status in humans.

Introduction

X chromosome inactivation transcriptionally silences most, but not all, genes on one X chromosome in mammalian females as a mechanism of dosage compensation between males and females (reviewed in 1, 2). This inactivation requires the presence in cis of a region called the X inactivation center. The XIST gene maps to this region and is exclusively transcribed from the inactive X chromosome, and developmental studies suggest a direct role for XIST in the initiation of X inactivation (3–6). Other aspects of the X inactivation process, such as the mechanism by which inactivation spreads across the entire X chromosome, remain largely unknown. Insight may come from studying the genes that appear immune or resistant to inactivation and are expressed from both the active and inactive X chromosomes.

Genes that escape X inactivation are not rare, and to date at least 16 genes have been identified that are expressed from the human inactive X chromosome (reviewed in 7). Most of these genes cluster in or lie adjacent to the pseudoautosomal region. Pseudoautosomal genes, by definition, are identical on the X and Y chromosomes, and therefore in females expression of the inactive X allele is required for proper dosage equivalence. Five additional genes map adjacent to the pseudoautosomal region in Xp22.32 and have been found to escape inactivation (8–11). However, dosage analysis of the one gene in this region that has been tested, STS, indicates only partial inactive X expression at approximately one-third the level of the active X chromosome (12). Y-linked sequences have been identified for three of these genes, but they are or are expected to be pseudogenes (9, 10, 13). For these reasons, escape from inactivation in this region may simply reflect the fact that, during evolution, the pseudoautosomal boundary has relocated from a position once proximal to these genes (14). Notably, no genes have been identified in this 8 Mb region that are subject to inactivation.

Such clustering argues strongly that escape from inactivation, at least at the distal tip of the X short arm, is controlled by regional mechanisms. However, whether this portion of the chromosome represents a special case is unknown. To determine whether escape from inactivation is regionally controlled or is locus-specific must be independently addressed for the other genes that escape inactivation and are interspersed among genes that are inactivated (15 and reviewed in 16) by assessing the inactivation status of genes that lie within very close physical distances.

The A1S9T gene was initially identified as a human X-linked gene that was capable, in somatic cell hybrids, of complementing a defect in the mouse tsA1S9 cell line, which carries a temperature-sensitive cell cycle mutation (17). Growth at the non-permissive temperature was restored regardless of whether the human X was active or inactive, suggesting that A1S9T escapes X inactivation (18, 19). The gene that confers this complementing activity was subsequently cloned and sequenced (20) and is homologous to ubiquitin activating enzyme E1 (UBE1) (21). UBE1 maps to the proximal short arm in Xp11.2 (22, 23), a region that was predicted to contain genes that escape inactivation because it remains early-replicating on the generally late-replicating inactive X chromosome (24) and contains acetylated histone H4 (a marker of active chromatin) on the essentially underacetylated inactive X chromosome (25).

Data on tsA1S9×human somatic cell hybrids notwithstanding, doubt has been raised as to the inactivation status of the human UBE1 gene. By Northern analysis, UBE1 transcription is not proportional to the number of X chromosomes carried in different cell lines (i.e. UBE1 RNA levels did not differ in 49,XXXXX, 47,XXX and 45,X cell lines) (20), an approach used to demonstrate that other X-linked genes are fully expressed from the inactive X chromosome (26–29). Based on these data, Zacksenhaus et al. concluded that UBE1 was subject to X inactivation (20). To settle this discrepancy, here we reevaluate UBE1 expression and demonstrate by three criteria that the gene is expressed at significant levels from the inactive X chromosome, and thus escapes X inactivation. Further, using UBE1 as a model gene that escapes inactivation, two directly adjacent genes have been identified whose inactive X chromosome expression states are assayed to address chromosomal mechanisms of escape from inactivation.

Results

UBE1 is expressed from active and inactive X chromosomes

Previous evidence that UBE1 escapes inactivation was derived from a functional assay in mouse-human somatic cell hybrids in which expression of the human gene was required to rescue a murine temperature-sensitive mutation (18). Since only hybrid clones expressing UBE1 survive at the non-permissive temperature, the possibility remained that inactive X expression was a consequence of this selection, perhaps reflecting reactivation of an inactivated gene in a subpopulation of these cells. Therefore, it was necessary to determine whether UBE1 escapes inactivation by independent methods.

To directly demonstrate that UBE1 is expressed from the inactive X chromosome, UBE1 RNA transcripts were detected by in situ hybridization in interphase nuclei from a 46, XX fibroblast cell line. A cosmid containing UBE1 was cohybridized with a plasmid containing the XIST gene, using conditions under which hybridization is specific to RNA but not DNA. Hybridization to XIST RNA serves as a control to identify the location of the inactive X chromosome, since previous data demonstrated that XIST RNA is both exclusively transcribed from and specifically localized to the inactive X at interphase (30, 31). Since all X chromosomes in excess of one are inactivated (32), there is one inactive X chromosome and therefore one XIST signal in a 46, XX cell (Fig. 1). However two RNA signals were detected by the UBE1 cosmid in each nucleus (>20 cells scored). Because a UBE1 signal is associated with the XIST signal, we conclude that UBE1 is transcribed from the inactive X chromosome.

Figure 1

RNA in situ hybridization in a 46,XX cell. XIST RNA is identified by a probe labeled with biotin, and detected with rhodamine-conjugated avidin (red signal). UBE1 RNA is identified by a probe labeled with digoxigenin and detected with anti-digoxigenin antibody coupled to fluorescein (green signal). Nuclei are counterstained with DAPI. A clear UBE1 signal from both the active X and inactive X (arrow) is visible. The relative intensity of the two signals was variable from cell to cell, with no consistent difference between the two (unpublished data).

Methylation at the 5′ end of UBE1 and Ubel-x

Methylation patterns at CpG islands of X-linked housekeeping genes correlate with expression (reviewed in 33), as genes that are inactivated are methylated on the inactive X chromosome (e.g. 34), while genes that escape inactivation are unmethylated (29, 35, 36). Since UBE1 was demonstrated to escape inactivation in the somatic cell hybrids (18), it is important to determine whether its methylation pattern is consistent with these expression results, and whether this pattern is conserved in human diploid cells.

In order to examine methylation, it was initially necessary to identify and characterize the genomic region at the 5′ end of the gene to determine whether UBE1 was associated with a CpG island. The 5′ end of the UBE1 cDNA has been published (21). Our data, based on additional cDNA cloning, are consistent with those data, but extend the cDNA sequence 5′ by 67 bp. A 5 kb genomic HindIII fragment was identified by Southern blots as the most upstream fragment to hybridize with cDNA probes generated from the published UBE1 sequence (21). This HindIII fragment was subcloned from a UBE1- containing cosmid, and a restriction map that includes this clone is shown in Figure 2B. The 196 bp first exon of UBE1 is entirely untranslated, with the start methionine encoded by the first three nucleotides of exon 2. The 500 bp sequence in and downstream of the first exon was identified as a CpG island by standard criteria (37, 38). Fragments from the 5 kb HindIII subclone were used to hybridize to Southern blots with DNA from males, females, and hybrids with active or inactive X chromosomes which had been digested with methylation-sensitive restriction enzyme whose sites are present in the island. The autoradiographs in Figure 2A reveal complete digestion of all sites in males, females, active X, and inactive X hybrids, indicating that all sites examined within the CpG island are unmethylated on both active and inactive X chromosomes. However, other sites outside of the island are not similarly unmethylated on both chromosomes. Two additional hybrids and two other lymphocyte samples gave similar results (data not shown). These methylation results at the CpG island are consistent with the gene escaping X inactivation in both human lymphocytes and somatic cell hybrids.

Figure 2

(A) Methylation at the 5′ end of UBE1. DNA from normal male and female lymphocytes or from somatic cell hybrids was digested with PstI and further digested with methylation-sensitive restriction enzymes if indicated. The active X hybrid (Xa)=A23-1aCl5, and inactive X hybrid (Xi)=LT23-1E2Buv5Cl26-7A2. The mouse cell line, A9, is included to establish that the additional bands present in the somatic cell hybrids (marked with white dots) represent cross-hybridization to the rodent homologue in the hybrids. For each hybrid and lymphocyte sample digested with the methylation-sensitive enzymes, there are faint bands in addition to the prominent bands, representing products of digestion. The sizes of these fragments were used to determine that bothEagI and both AvaI sites within the CpG island are unmethylated on active and inactive X chromosomes. (B) Restriction map of 6 kb of genomic sequence (including the 5 kb HindIII subclone) containing the 5′ end of UBE1. The black bar represents the 700 bp BglI fragment used to probe Southern blots above. The only additional site outside of this HindIII fragment that was analyzed is the upstream BssHII site. The 500 bp CpG island is 69% C+ G and has an Observed/Expected (Obs/Exp) ratio of 0.72 [using the definition of an island as having a % C + G > 0.5 and an Obs/Exp ratio < 0.6, where Obs/Exp=(number of CpG × number of bp)/(number of C × number of G) (38)]. Restriction enzyme abbreviations: H=HindIII, P=PstI, Sc=SacII, A=AvaI, E=EagI, Bs=BssHII, Sm=SmaI.

Multiple restriction sites within the human UBE1 island identify CpG sites that are unmethylated on both X chromosomes. To determine whether the methylation pattern of these sites specifically correlates with the expression state of the gene, the homologous sequences were analyzed at the murine Ube1-x locus, which has been demonstrated to be subject to X inactivation (39, 40). Male and female mouse DNA was digested with the same restriction enzymes that were used to analyze the human locus, and hybridized at reduced stringency to the same genomic fragment from the human CpG island. DNA from the male mouse is completely digested with the three enzymes examined (Fig. 3). The lack of hybridization in the AvaI lane likely indicates that there are multiple unmethylated sites, generating fragments too small to be seen on these blots. For female DNA, an uncut band remains, in addition to digested bands for each of the enzymes that are the same sizes as in the male DNA. Additional experiments, using other human probes from this region and a mouse clone, identified the same PstI band, ensuring that these probes were hybridizing to the mouse Ube1-x homologue (data not shown). These data suggest that the active X is unmethylated at these sites, while the inactive X is methylated, consistent with the known expression state of the mouse gene. Together, these data suggest that methylation of sites within the island in both mouse and human is an accurate indicator of the expression state of the gene in both species.

Genomic mapping at the UBE1 locus

Because the above experiments demonstrated that UBE1 is expressed from the inactive X chromosome, it is possible to use UBE1 to address whether escape from inactivation (at least at this locus) is controlled by regional mechanisms. During the course of mapping to define the genomic sequences associated with the 5′ end of UBE1, a number of methylation-sensitive restriction sites were identified downstream, prompting us to extend the genomic map more thoroughly in the 3′ direction. A CpG island was located no more than 5 kb from the 3′ end of UBE1 (Fig. 4). The gene associated with this island was fortuitously identified as the Schizosaccharomyces pombe cdc2-related protein kinase, PCTAIRE-1 (PCTK1) (41), which had been previously broadly mapped to Xp11 (42). The published genomic map of the PCTK1 locus extends almost 20 kb upstream of the 5' end of the PCTK1 transcript and showed complete agreement of HindIII, BamHI, SfiI, and XhoI restriction sites over the region that overlapped with the UBE1 genomic map (42). This placed PCTK1 downstream of UBE1, in the same orientation. Two restriction sites, SacII and BssHII, are present at the 5′ end of the PCTK1 cDNA, and correspond to sites within the PCTK1 CpG island. According to the published PCTK1 restriction map (42), approximately one-half of the 3.1 kb cDNA is represented in the map shown in Figure 4. Southern blotting with a small cDNA probe for UBE1 that ends 40 nucleotides upstream of the poly A tail, places the 3′ end of UBE1 within the 2 kb EcoRI-HindIII fragment upstream of PCTK1. These mapping data indicate that the 3′ end of UBE1 and the 5' end of PCTK1 lie a minimum of 3 kb and a maximum of 5 kb apart from each other. These data significantly refine the results from a recent paper that concluded that UBE1 and PCTK1 colocalize to the same 420 kb YAC (43).

Figure 3

Methylation of the mouse Ube1=x gene. The human Bgl 700 probe that was hybridized to Southern blots in Figure 2 was used to probe DNA from C57BL/6J male and female mouse livers. DNA is digested with PstI, and further digested with restriction enzymes listed. Southern was washed at reduced stringency in 2× SSC at 65°C.

An additional cDNA from this region was isolated as a direct selection clone from a cosmid containing UBE1 and corresponds to the locus DXS8237E (44). A cDNA PCR product that is close to the 3′ end of the DXS8237E gene maps to an XhoI-SfiI fragment ∼8 kb upstream of the UBE1 CpG island (Fig. 4). An XhoI site is also present at the very 5′ end of this PCR product, whose direction was established by comparing sequence in cDNA and DNA and determining the orientation of consensus splice sites. Because the cDNA fragment hybridizes to an XhoI-SfiI fragment, the orientation of DXS8237E with respect to UBE1 could be established with the two genes transcribed in the same direction. Additionally, other DXS8237E cDNA probes that map towards the 5′ end of the gene hybridize further upstream (JMD and HFW, data not shown). The 5′ end of this gene has not been determined, but current data indicate that it is upstream of this existing restriction map. The map in Figure 4 is built upon and confirms, integrates, and extends the existing genomic maps around UBE1 (20) and PCTK1 (42).

Expression studies of UBE1, PCTK1, DXS8237E, and other Xp11.23 genes

Since these genes map within close proximity, it is important to determine whether the two genes that flank UBE1 also escape X inactivation and, similarly, to establish what the inactive X expression patterns are for other nearby genes within Xp11.23. Expression was tested by PCR of reverse-transcribed RNA (RT-PCR) of cDNA in somatic cell hybrids retaining human active or inactive X chromosomes. Many of the somatic cell hybrids in this laboratory were made with the mouse tsA1S9 cell line and use the UBE1 selection scheme to ensure retention of X chromosomes (16, 17). Therefore, although expression for each gene was tested in a total of eight hybrids, including five carrying inactive X chromosomes, the results shown in Figure 5A specifically focus on two inactive X hybrids, LT23-1E2Buv5Cl26-7A2 and L23-4B, that did not use this selection system and, therefore, do not require expression of the UBE1 gene for survival. In addition, the somatic cell hybrid, A23-1aCl5, is isogenic to the LT23-1E2Buv5Cl26-7A2 hybrid, carrying the identical X chromosome, but in its active state (data not shown). Analysis of these hybrids eliminates the possibility that expression levels vary because of allelic differences. For each gene tested, no consistent or significant differences in levels of expression were seen among different active X hybrids or different inactive X hybrids (data not shown).

Figure 4

Genomic organization of DXS8237E, UBE1, and PCTK1. Genomic map is on top, with cDNAs at the bottom. Abbreviations for restriction enzymes: H=HindIII, E=EcoRI, B=BamHI, Xh=XhoI, Sf=SfiI, and the methylation-sensitive restriction enzymes; Ea=EagI, Sc=SacII, Bs=BssHII, Sm=SmaI, Nr=NruI. This region was mapped using sequential cDNA probes against genomic and YAC DNAs digested with the enzyme(s) indicated on the restriction map; subclones from the cosmid and a genomic phage at the 5′ end of the PCTK1 gene which were used for further mapping of CpG islands are indicated by thick black bars on the genomic map. SmaI sites were only mapped in the subclones and are indicated specifically to show the abundance of these methylation sensitive sites within islands, but does not preclude additional sites outside these fragments. Enzymes in bold indicate methylation-sensitive sites present within a CpG island. Patterned bars on cDNA maps identify the location of specific probes used for genomic mapping, whereas patterned bars on the genomic map identify the restriction fragment to which the cDNA probe hybridizes.

Figure 5

(A) X-linked gene expression from active (Xa) or inactive (Xi) X chromosomes in somatic cell hybrids. Negative image of ethidium bromide stained RT-PCR products. Amplification of cDNA from human female lymphoblast (H), but not from the mouse A9 cell line (M) confirms that each primer pair specifically detects a human transcript. cDNA from somatic cell hybrids: lane 3, t60-12; lane 4, L23-4B; lane 5 A23-1aCl5; lane 6, LT23-1E2Buv5Cl26-7A2. The somatic cell hybrids are not under selection to retain UBE1. Additionally, the hybrids in lanes 5 and 6 are isogenic, having the same human X chromosome in either its active (lane 5) or inactive (lane 6) state. PCR primers for each gene span introns and generate a specific sized cDNA product. (B) Inactive X expression of genes within Xp11.23 (Fig. 5a; CJ Brown unpublished data; LC and HFW unpublished data). Distances and gene locations are based on a YAC contig (46). OnlyUBE1 and PCTK1 are expressed (+) from the inactive X chromosome, while the other genes examined are subject to inactivation (−). Two additional genes in the region, PFC and SYN, have tissue specific expression patterns, and their inactivation status can not be addressed using the somatic cell hybrid system.

For both UBE1 and PCTK1, expression is seen from all hybrids with either an active or inactive X chromosome (Fig. 5A), indicating that PCTK1, like UBE1, escapes inactivation. In contrast, for the other genes tested, there is no PCR product from the hybrids containing an inactive X chromosome, demonstrating that DXS8237E, ELK1, and ARAF1 are subject to X inactivation. For DXS8237E, these data confirm, in additional somatic cell hybrids, the result reported earlier (44). In addition, these experiments confirm that UBE1 does escape X inactivation in non-selected hybrids and is well expressed from the inactive X chromosome.

Figure 5B shows the chromosomal location of eight genes within Xp11.23 whose expression states have been tested (Fig. 5A; ref. 45, C. J. Brown, unpublished data; LC and HFW, unpublished data). Only two, UBE1 and PCTK1, are expressed from the inactive X chromosome. Map distances have been determined by a YAC contig that spans the region, and are based on sequence tagged site (STS) content mapping (46). This distance, however, may be an underestimate, as the region between TIMP1 and UBE1 on this map is ∼550 kb, yet others have reported that this distance is greater than 600 kb (47). The chromosomal orientation of the UBE1, PCTK1, and DXS8237E genes was determined by STS content mapping within YACs previously anchored in this contig. Because DXS8237E is present in the YAC yWXD1909 which contains both UBE1 and ZNF157, whereas PCTK1 is not (data not shown), we conclude that PCTK1 is distal and that the chromosomal orientation of these genes is as shown in Figure 5B.

Mapping and X inactivation of the mouse Pctk1 gene

The possibility that murine genes mapping to the syntenic region of human Xp11.23 escape inactivation is suggested by the finding that both the human and mouse chromosomal locations contain acetylated histone H4 (a marker of gene activity) on the otherwise generally underacetylated inactive X chromosome (25). However, the only murine gene analyzed that maps to the A2 region of the mouse X chromosome, Ube1-x, is inactivated (39, 40). Therefore, it was of interest to determine whether the genomic organization of Pctk1 and Ube1-x was conserved in mouse, and if so, to test for inactive X expression of Pctk1.

The physical distance between the mouse Ube1-x and Pctk1 genes was determined by genomic pulsed field gel electrophoresis. High molecular weight DNA from a male mouse was digested with rare-cutting restriction enzymes that had been determined by conventional Southern analysis to map within or outside of the CpG islands associated with either gene. Figure 6 shows autoradiographs of the pulsed field blot hybridized with probes for both genes. Four common bands were identified by hybridization with either probe (indicated by arrows). For the enzymes MluI and NruI these shared fragments are ∼18 and 23 kb, indicating that the genes are quite close. Previous studies indicated that there are SmaI and SacII sites within each island, and (as predicted, if the genes are in any orientation other than tail to tail and transcribed towards each other) hybridization to DNA digested with these enzymes identifies different sized fragments with each probe. These data, together with the presence of common EagI sites, suggest that the genes are very close. To rule out the possibility of polymorphic differences between the strains used for conventional mapping experiments (C57BL/6J) and the (C3H × 101) F1 used for pulsed field mapping, the pulsed field gels were repeated using DNA from a C57BL/6J mouse and a similar hybridization pattern was seen (data not shown).

Figure 6

Physical mapping of the mouse Ube1-x and Pctk1 genes. Pulsed field gel of genomic DNA from a (C3H ×101) F1 male mouse, digested with enzymes indicated. Blots were hybridized with cDNA probes designed from the published mouse sequences for each gene. The Ube1-x probe is a PCR product corresponding to nucleotides 3091–3428 (sequence from 85) and the Pctk1 probe includes nucleotides 1502–2040 (sequence from 53). White arrows identify common bands that hybridize with each probe.

Since the above experiments demonstrated that both mouse genes are quite close, it was of interest to test whether the murine Pctk1 gene was expressed from the inactive X chromosome. Expression was tested in an inter-subspecies F1 female resulting from the mating of a T(X;16)H female (T16H) to a Musmusculus castaneus male. Females carrying the balanced T16H translocation show non-random inactivation of their normal X chromosome (48, 49). In F1 females, this normal X can be distinguished by polymorphic differences between alleles in the two mouse subspecies. This non-random inactivation system (Fig. 7A) (39, 50–52) was used to determine whether the murine Pctk1 gene is subject to inactivation. To identify polymorphic sites that would allow differentiation of the T16H and Mcastaneus alleles, a PCR fragment from the 3′ end of the gene was sequenced in both mouse strains. A single nucleotide difference was identified that changes the M. castaneus allele from A to G at position 1542 (sequence from 53) and generates a PvuII restriction site.

Expression was tested by RT-PCR from cDNA made from liver RNA. PCR products were digested with PvuII after amplification to differentiate the two alleles. As seen in Figure 7B, both alleles of Pctk1 are expressed in cDNA from a (C57BL/6J × M. castaneus) F1 female mouse characterized by random X inactivation. However, in the non-randomly inactivated (T16H × M. castaneus) F1, a PCR product is only present from the T16H allele (on the active X) and not the M. castaneus allele. These data indicate that the Pctk1 gene, like the adjacent Ube1-x gene, is subject to inactivation in mouse.

Methylation of the PCTK1 and Pctk1 genes

To analyze methylation of the human PCTK1 gene, a genomic probe from the 5′ portion of the CpG island was used to hybridize to Southern blots made from male and female lymphocyte DNA. Multiple AvaI and SacII fragments were identified and restriction sites correspond to sites mapped in a phage subclone. Identical digestion patterns were seen in males and females, suggesting that all sites are unmethylated on both active and inactive X chromosomes (Fig. 8A). Similar studies were performed with the mouse gene using this same probe, but at reduced stringency, on Southern blots with DNA from male and female C57BL/6J mice cut with a variety of methylation-sensitive restriction enzymes. A number of restriction sites were identified that are partially digested in females, but are entirely cleaved in males (Fig. 8B). All digested fragments were of similar size, suggesting that these restriction sites are clustered and most likely found within the CpG island. These mouse data suggest that the active X chromosome is digested and unmethylated, and the inactive X is methylated. Therefore, methylation patterns of both the mouse and human PCTK1/Pctk1 genes are consistent with the expression of the genes in the two species.

Figure 7

(A) Interspecific cross used to analyze X inactivation in mouse. Only one of the derivative X chromosomes carries the X inactivation center (XIC) and can be inactivated. However, cells in which this 16^X chromosome is inactivated are dosage unbalanced due to inactivation of adjacent chromosome 16 sequences and failure to inactivate X sequences on the X¹⁶ chromosome. Selection against these inviable cells results in complete non-random inactivation of the normal X chromosome (39, 50, 51). (B) The murine Pctk1 gene is subject to X inactivation. Negative image of ethidium bromide stained PCR products. Pctk1 was amplified from liver cDNA from the parental strains and both F1 animals. All lanes are digested with PvuII after amplification. In DNA from the (T16H × cast) F1, the absence of amplification products the same size as cDNA indicates that the primers span an intron. The two larger faint bands present are the DNA products from both alleles, confirming that the animal is heterozygous, despite, as seen in cDNA, expressing only the T16H Pctk1 product.

Figure 8

Methylation of the CpG islands at the human and mouse PCTK1/Pctk1 genes. Both blots are hybridized using a 2 kb human genomic HindIII-BssHII fragment, which includes the 5′-most sites within the human CpG island. (A) Methylation of human PCTK1 gene. DNA was digested with PstI, and further digested with enzymes indicated. (B) Methylation of the murine Pctk1 gene. Southern of mouse DNA was digested with HindIII initially, and other enzymes as noted. The Southern blot was washed at reduced stringency to 2× SSC at 65°C.

This region is interesting from an evolutionary standpoint, as UBE1 does not have a Y homologue in humans (17, 20, 54), yet does in mouse (39, 54). However, Ube1-y, which is a candidate for the spermatogenesis factor Spy, is expressed only in testes. Y-specific sequences are also conserved in marsupials (55). To determine whether the murine Pctk1 gene shares any homology to sequences on the Y chromosome, two different cDNA probes (PCR products extending from nucleotides 678–1377 and 1502–2040) (sequence from ref. 53) were hybridized at reduced stringency to Southern blots containing DNA from male and female mice. For four restriction sites examined, no male–specific bands were detected (data not shown). A similar experiment was performed for the human PCTK1 gene using a cDNA probe generated from nucleotides 781–1415 of the human sequence (from ref. 41). This probe also failed to detect Y-specific bands in DNA from males or a somatic cell hybrid carrying the Y chromosome (data not shown). These data suggest that it is unlikely that there is a functional Y homologue in either species unless the genes are quite diverged. If there is an evolutionary breakpoint between Ube1-x and Pctk1 which determines the sequences that have homologues on the Y chromosome, at least in mouse, this breakpoint must lie in the small distance between these genes.

Discussion

UBE1 has been demonstrated to escape X inactivation in somatic cell hybrids, including two that do not require the expression of UBE1 for survival. Additionally, hypomethylation of five sites within its CpG island on active and inactive X chromosomes in somatic cell hybrids and lymphocytes provides indirect evidence consistent with expression of the gene from the inactive X. In addition to these data, RNA in situ hybridization to interphase nuclei in fibroblast cell lines using a UBE1 cosmid demonstrates that there are transcripts from both the active and inactive X chromosomes (Fig. 1). The cosmid containing UBE1 also contains the DXS8237E gene (see Fig. 4). Because DXS8237E was demonstrated in somatic cell hybrids to be subject to X inactivation, it is unlikely that the inactive X RNA detected with this cosmid probe represents transcripts other than from UBE1. Therefore, together, these experiments demonstrate in diploid cells and in hybrids that UBE1 escapes X chromosome inactivation.

There are a number of possibilities that could explain why UBE1 transcript levels on Northern blots did not increase proportionately with the number of X chromosomes in cell lines (20). First, expression may be reduced from the inactive X chromosome, but not at levels detectable by our assay. This would be similar to the STS gene, which is expressed from the inactive X chromosome at about 30% of active X levels (12). Alternatively, in cell lines carrying multiple X chromosomes, the transcript could be dosage compensated to decrease expression levels from all X chromosomes by a mechanism other than transcriptional inhibition. Additionally, UBE1 transcription is cell cycle dependent (20), and transcript levels could be altered by culture conditions. In support of the idea that inactive X transcription may be decreased, it has been observed that UBE1-selected somatic cell hybrids with inactive Xs tend to accumulate multiple X chromosomes in each cell, whereas active X hybrids do not (18; V. Powers and HFW, unpublished data). If transcript levels from each X are suboptimal, there may be a growth advantage for cells carrying additional inactive X chromosomes. Further experiments will be necessary to establish why UBE1 transcription levels do not correlate with the number of X chromosomes in cell lines. Regardless, none of these possibilities is inconsistent with our conclusion that UBE1 escapes X inactivation.

We have examined the expression status of eight genes mapping within 1 Mb in Xp11.23. Two of these genes, UBE1 and PCTK1, escape inactivation and are located within 5 kb of each other at the telomeric end of this region, while six other genes are subject to inactivation. That two genes escaping inactivation are positioned within very close proximity, while the others are inactivated, seems most consistent with models in which regional mechanisms control escape from inactivation. At least seven genes on the X short arm (outside of the pseudoautosomal or adjacent region, whose genes, as discussed above, should be considered independently) have been identified that are expressed from the inactive X chromosome (26, 56, 57; APM and HFW unpublished data). These data indicate that escape from inactivation in this region is not rare. However, not enough genes have been analyzed to determine whether the similar expression pattern of two genes within 5 kb is indicative of coordinate control or simply reflects a chance occurrence in a region where genes that escape inactivation are relatively frequent. To differentiate between these possibilities would require analysis of a much larger number of genes within a specified distance.

Clues to the mechanism of gene control in this area may be found by determining why DXS8237E is subject to X inactivation. Do the expression differences between UBE1 and DXS8237E imply that these genes lie on either side of a boundary that delimits a domain containing genes that escape inactivation from one whose genes are inactivated? Or is DXS8237E structurally different from the other two genes (e.g. with regard to promoter strength or CpG density)? These levels of regulatory control may not be mutually exclusive alternatives, since a similar expression study of genes within Xp11.21 (albeit with a much greater physical distance between the genes that were analyzed) identified three clustered genes that escape inactivation, with one separated from the others by two genes that are inactivated (15), perhaps indicating that both regional and promoter-specific mechanisms control the expression of these genes.

To date, six genes that escape inactivation on the human X chromosome have been analyzed in mouse, and only one of these murine genes, Xe169, is expressed from the inactive X chrzzomosome (57, 58). This argument has been used to explain the difference in viability of XO females in humans and mice (50), perhaps reflecting a true difference in the process of X inactivation between mouse and human. However, these data could also indicate differences in the promoter sequences, chromosomal context or evolutionary conservation of these genes in the two species. For example, the humanXE169 and DXS423E genes in Xp11.21 map within ∼300 kb (15; APM and HFW, unpublished data), and both escape X inactivation (56–58); however in mouse, although the physical distance between these genes is conserved (data not shown), their inactive X expression patterns are not (52, 58, 59). If chromosomal elements control escape from inactivation, these data could suggest that a boundary between a domain that escapes inactivation and one that is inactivated has shifted between mouse and human. Because histone acetylation data on the mouse inactive X chromosome suggest that the region near Ube1-x might contain expressed genes (25), Pctk1 was analyzed to see if inactive X expression patterns are identical in human, but discordant in mouse. However, both mouse genes, Pctk1 and Ube1-x, are inactivated, suggesting that if regional elements control escape from inactivation, both of these genes have been excluded from this active domain. Further gene detection and analysis will be necessary to determine whether murine genes in this region do escape X inactivation.

Chromosomal mechanisms are also likely to be responsible for controlling gene expression of autosomal imprinted regions (60). Several clusters of imprinted genes have been identified: the H19/Igf2 region of mouse chromosome 7 and human chromosome 11, and the Prader-Willi/Angelman region on human chromosome 15 (61, 62). Replication timing shows that asynchronous replication correlates with expression patterns across imprinted domains (63). In support of regional imprinting control mechanisms, targeted transgenic mice with a maternal deletion of the H19 gene and 10 kb of 5′ flanking sequence disrupt imprinting of upstream genes resulting in biallelic expression of both Igf2 and Ins2 (64). Whether this deletion allows Igf2 expression by eliminating competition for shared enhancers (65) or disrupts a nearby sequence required for the imprinting of genes in this region remains to be determined. In addition, several familial cases of Angelman and Prader-Willi Syndromes have been reported that carry putative imprinting mutations that affect methylation and gene expression across at least 1.5 Mb (66, 67). However, at least for the human IGF2 gene, specific promoter usage within the same tissues can direct either bi-allelic or imprinted expression (68). These data suggest that even within an imprinted domain, local sequences determine whether a gene, or in this case, a promoter-specific transcript, will ‘escape’ imprinting.

DXS8237E and UBE1 are physically the closest mapped genes that have been identified that show discordant X inactivation patterns. If these two genes do identify distinct expression domains, there may be control sequences or boundary elements that lie in the 8 kb that separates the 3′ end of DXS8237E from the 5′ end of UBE1. A number of sequences have been identified that may serve as chromosomal boundary elements. One class of elements includes the Drosophila sequences designated as specialized chromatin structures, scs and scs′, which insulate encompassed DNA fragments from nearby chromatin effects (69, and reviewed in 70). A similar element has been shown to function in mammalian systems (71, 72), suggesting that these sequences may be conserved. Other sequences that are hypothesized to allow domain specific expression patterns are matrix attachment sites (73), or methylation centers (74, 75), which control methylation patterns of adjacent or distant sequences without conferring other regulatory information.

Materials and Methods

RNA in situ hybridization

The probes used for hybridization experiments were G1A, a plasmid containing 10 kb of genomic DNA from the 3′ end of the XIST gene (30, 31) and a cosmid, ICRFc100E0981 (76), containing ∼13 kb of the 25 kb genomic UBE1 locus (Fig. 4) and the DXS8237E gene. Because DXS8237E has been demonstrated to be subject to inactivation in somatic cell hybrids retaining inactive X chromosomes (Fig. 5; ref. 44), RNA hybridization with this cosmid specifically tests inactive X UBE1 transcription. The cell line used for hybridization was a 46, XX epithelium cell line, ATCC CCL 75 (WI-38), from the American Type Culture Collection. Preparation of cells for hybridization and hybridization conditions were as previously described (31, 77). Hybridization was to non-denatured cells (such that cellular DNA was not accessible for hybridization). A total of >20 cells were selected and scored on the basis of successful hybridization of both probes.

Sequencing

PCR products or miniprep DNA samples were purified using Wizard DNA clean up system (Promega). Taq based cycle sequencing reactions using fluorescent dideoxy terminators were run and automatically read on an ABI 373 sequencer (Applied Biosystems). Subsequently sequences were analyzed using DNASTAR (DNASTAR, Inc.) software.

Library screenings

All clones were isolated from established commercial libraries. cDNAs were screened from human fetal brain or heart and mouse lung libraries (Stratagene). The UBE1 cosmid was obtained by screening a gridded X chromosome cosmid library [ICRF (78)]. The human genomic phage library was a flow sorted X genomic library (LAOXNL01) from the American Type Culture Collection. YACs were acquired from both Research Genetics and ICRF.

Genomic DNA analysis

Genomic DNA was prepared by phenol extraction or salted out (79). Conditions for DNA digestion, probe labeling, and hybridization for Southern blots has been previously described (23, 80). Blots were washed unless otherwise noted to 65°C, 0.1× SSC and 0.1% SDS.

For genomic pulsed field gels, agarose plugs were prepared from the thymus of 1 week old mouse, and plug preparation and pulsed field conditions and blotting were performed exactly as described (81). Digested plugs were electrophoresed with a 10–90 s pulse, 120° angle, at 6 V/cm, 14°C for 20 h. Gels were transferred to Nytran⁺ membranes (Schleicher and Schuell), hybridized at 65 °C in Church buffer (82), and washed at 65 °C to 2x SSC.

Inactive X expression assays

The X inactivation status of human X-linked genes was determined by assaying expression in the mouse-human somatic cell hybrids t60–12 and LT23-1E2Buv5Cl26–7A2 which solely retain an active and inactive human X chromosome respectively (16, 45) and the hybrids L23–4B, that carries an inactive X chromosome in addition to six human autosomes, and A23–1aCl5, that carries an active X chromosome in addition to twelve autosomes. The X chromosome in A23–1aCl5 is isogenic to the inactive X in the LT23-1E2Buv5Cl26–7A2 hybrid (data not shown).

To assay X inactivation in mouse, T(X;16)16H females and M.m.castaneus mice were obtained from the Jackson laboratory. C57BL/6J and F1 interspecific animals were bred in house. Balanced translocation carriers were identified by fluorescence insitu hybridization using probes proximal and distal to the translocation breakpoint. Additionally, females carrying the translocation were confirmed by detecting non-random inactivation of the Hprt gene.

RNA from cell lines or mouse tissues was prepared with the phenol/guanidinium isothiocyanate solutions, RNAzol (Cinna-Biotecx) or TRIzol (Gibco-BRL), according to recommended procedures. The RNA was quantitated spectrophotometrically and 1 or 5 µg was used for each reverse transcription reaction, which were performed as described (45). 1/20th of each reverse transcription reaction was used for subsequent RT-PCR reaction. PCR primers used for expression studies are listed in a 5′ to 3′ orientation followed by the size of the cDNA product that is amplified. UBE1 1: GAGCGGGGACTTTGTCTCCT and 7: CTTTGACCTGACTGACGAT (150 bp cDNA product) (sequence from ref. 20), PCTK1 h4: CACGCCAACATCGTTACGCT and h7: TGGGATTGACTTGGCTCGG (286 bp) (sequence from ref. 41), DXS8237E A: CAGCTCCCTGCGAGATGACG and B: CTCGATAGGCGTTACAATGC (650 bp) (sequence from ref. 44), ELK1 A: GGACCTAGAGCTTCCACTCA and B: AGAGCATGGATGGAGTGACC (388 bp) (sequence from ref. 83), ARAF1 B: TCAGCAAAATCTCCAG-CAAC and 3: TGGAGATGGAGGAGCTCCCA (482 bp) (sequence from ref. 84). The primers used to analyze mouse Pctk1 expression were 1: CAGGTATCCTGTCCAATCAG and 2: ATTCCAATGGTTGGGTTGACA (538 bp) (sequence from ref. 53). The human DXS8237E primers were amplified using a 65 °C annealing temperature. All other primers were amplified using an annealing temperature of 55 °C.

Acknowledgements

We thank Drs Carolyn Brown and Bernadette Holdener for helpful discussions and Drs Jonathan Derry and Uta Francke for providing sequence from the ZNF157 gene prior to publication (86). This work was supported by NIH research grant GM45441 (to HFW).

Author notes

⁺

This work was performed in partial fulfilment of the requirements for a Ph.D. degree to L.C. from the Department of Genetics at Stanford University