An X-linked human collagen transgene escapes X inactivation in a subset of cells

In mammals, dosage compensation of X-linked genes is achieved by inactivation of one of the two X chromosomes in female cells during early embryogenesis (Lyon, 1972, 1988; Gartler and Riggs, 1983; Grant and Chapman, 1988). Studies of X-autosome translocations and rearrangements suggest that the X inactivation process is probably initiated from a single site, XIC (X chromosome inactivation center) in man and Xce (Xcontrolling element) in mouse (Russell and Cacheiro, 1978; Takagi, 1980; Rastan, 1983), and spreads to the adjacent regions. In man, the XIC has been localized cytologically to band Xq13.1 (Mandel et al., 1989; Brown et al., 1991a), and the mouse Xce has been mapped to the distal portion of the D region (Keer et al., 1990). Alleles at the Xce locus influence the probability of a X chromosome to be inactivated (Johnston and Cattanach, 1981).

The nature of the signal and the mechanism of initiation and spreading of the inactive state are not known. A human gene, XIST (X inactivation specific transcript), that is only expressed from the inactive X chromosome has recently been identified (Brown et al., 1991b). Interestingly, the XIST and its murine homologue, Xist, are localized to the XIC region in human and Xce region in mouse (Brown et al., 1991b; Borsani et al., 1991; Brockdorff et al., 1991a,b) suggesting that these transcripts may have a role in X inactivation.

Three lines of evidence suggest that DNA methylation may play a role in the maintenance of the inactive state. Cytosine residues in CpG sequences (Bird, 1986) up-stream of some X-linked genes are hypermethylated on Xⁱ, the inactive chromosome, and are hypomethylated on X^a, the active chromosome (Lock et al., 1986, 1987; Mullins et al., 1987). Also, treatment of cells with 5-azacytidine, which causes heritable hypomethylation of cytosine in the DNA, can cause reactivation of genes on Xⁱ (Lock et al., 1986). Finally, DNAs isolated from Xⁱ versus X^a show different transformation efficiencies (Liskay and Evans, 1980; Chapman et al., 1982), indicating that Xⁱ and X^a DNA fragments are differentially modified.

Although most X-linked genes are believed to be subject to X inactivation, the number of genes is growing that have been demonstrated to escape the inactivation process. Initially, genes mapped to or near the peudoautosomal region were found to escape X inactivation partially or completely (Race and Sanger, 1975; Shapiro et al., 1979; Migeon et al., 1982; Goodfellow et al., 1984; Jones et al., 1989). More recently, genes located more proximal on Xp such as the ZFX gene (Schneider-Gädicke et al., 1989), the A1S9T (Brown and Willard, 1990) and the RPS4X gene (Fisher et al., 1990) on Xq13.1 close to the XIC have been shown to escape inactivation, suggesting that escape from X inactivation is not limited to the pseudoautosomal region of the human X chromosome. There is also a striking correlation between escape from X inactivation and the existence of homologs on the Y chromosome which have similar expression patterns. For instance, genes that escape X inactivation in human, such as ZFX and RPS4X, have a closely related homologue on the Y chromosome (Schneider-Gädicke et al., 1989; Fisher et al., 1990). In contrast, the murine Zfx and Rps4 genes, which undergo normal X inactivation, either have no homologue on the Y chromosome (in the case of Rps4 gene) or the Y homologous gene has a different expression pattern (in the case of Zfx) (Adler et al., 1991; Ashworth et al., 1991; Zinn et al., 1991). These results raise the question whether cis-acting regulatory sequences, involved in regulating gene activity, also influence the susceptibility of a given gene to X inactivation. So far, sequence analyses on some X-linked genes have failed to reveal any sequence motifs common to genes, which are either subject to or escape from X inactivation (Goldman et al., 1987). If such regulatory sequences do exist, they must lack common motifs or they must be localized outside of the sequenced regions. An alternative approach to address this question is to study whether an autosomal gene will escape X inactivation upon integrating into the X chromosome, or whether an X-linked gene will be subject to inactivation when inserted into autosomes.

We have derived a transgenic mouse line (HucII) that carries a single copy of the human α1(I) collagen gene (COL1A1) on the X chromosome (Wu et al., 1990a). In this study, we have determined the site of transgene integration and tested whether the state of X inactivation affects the expression of the transgene.

High relative molecular mass DNAs were isolated from tissue culture cells or embryo tissues following digestion with 400 U/ml proteinase K (Wu et al., 1990a).

For genotype analysis, DNA was digested with the restriction enzyme EcoRI and separated on an 0.8% agarose gel. After transferring to Zetabind™ membrane, blots were probed with a 14 kb mouse proα1(I) collagen clone pI (Harbers et al., 1984), which hybridizes to a 23 kb EcoRI fragment of the human collagen transgene and to a 14 kb EcoRI fragment of the endogenous mouse gene. To determine the sex of the animals and the cell lines, blots were rehybridized to the Y chromosome specific probe pY2 (Lamar and Palmer, 1984).

To test the methylation status of the transgene on active and inactive X chromosomes, 5 μg of DNA from each cell clone or animal were digested with either MspI or HpaII (in 10-fold excess). After transferring, filters were hybridized with a 2.3 kb fragment derived from the 5′ region of the human proα1(I) collagen clone pHCH2.3 (Wu et al., 1990a). The same filters were also probed with a 1.3 kb XbaI-XbaI fragment containing the 5′ end of the mouse proα1(I) collagen gene (Wu et al., 1990a).

Total RNA was prepared according to the method of Auffray and Rougeon (1980). S1 analysis was performed to test the steadystate mRNA levels from tissue culture cells or embryonic tissues. To measure the human α1(I) collagen mRNA, a 1.1 kb EcoRI-TaqI fragment was isolated from pHCH2.3. The 5′ end of the fragment was labelled with γ-³²P by using T4 kinase. As an internal control, a 0.54 kb 3′ end-labelled BstEII-StuI fragment of mouse α2(I) collagen was used (Wu et al., 1990b). Hybridization was carried out with 5 μg of RNA and probes described above at 50°C for 16 –20 hours. Gels were dried, exposed directly to Kodak X-OMAT AR films and quantified on a Betagen betascope.

Embryos were isolated at day 14 of gestation. Cells were dispersed by treatment with trypsin and plated on tissue culture dishes.

For establishment of 6TG^r cells, primary cells were cultured in DME medium containing 10% FCS and 6-thioguanine (10 μg/ml, Sigma) for at least 40 days. To derive permanent cell lines, the above culture were transformed with a retroviral vector containing SV40 large T antigen cDNA and neo gene (Jat and Sharpe, 1986). After 6TG and G418 selections (10 μg/ml and 500 μg/ml, respectively), individual clones were isolated and genotyped as described above.

Cells were grown on gelatin (0.2%)-coated slides for two or three days before fixing with 4% paraformaldehyde (pH 7.3). An 43 base oligonucleotide probe derived from the first exon of the human COL1A1 gene (Wu et al., 1990a) was labelled with ³⁵S-dATP by terminal transferase (BRL). Approximately 1.5×10⁶ cts/mimute of labelled oligo were added to each slide and hybridized at 42°C for 14 to 16 hours in a solution containing 4× SSC, 50% formamide, 1×Denhardt’s, 0.2 M sodium phosphate, 10% dextran sulfate, 0.2 M DTT, 0.25 mg/ml yeast t-RNA, and

0.3 mg/ml herring sperm DNA. Slides were washed in 1×SSC at 55°C. After dehydration and air drying, slides were dipped in emulsion and kept at 4°C for 10 days before developing. Finally, the cells were stained with hematoxylin.

A primary fibroblast culture was established from lung and kidneys of a 12.5-day-old fetus and were grown in DMEM supplemented with 10% calf serum and antibiotics. After one subculture, the cells were exposed to colcemid for 2 hours and to ethidium bromide (10 μg/ml) for 1 hour. They were then trypsinized, spun and resuspended in the prewarmed hypotonic 0.75M KCl and fixed in 3:1 methanol : acetic acid for several hours. After three washes with fresh fixative, the cell suspension was dropped onto wet precleaned microscope slides and air-dried. The slides were stored in a desiccator for 2 to 3 weeks before in situ hybridization.

The cosmid probe (pJB8) was labelled by nick-translation with three tritium-labelled nucleotides ([³H]dATP, [³H]dCTP and [³H]dTTP) to a specific activity of 2.8×10⁷ cts/minute/μg, and hybridized to chromosome preparations of HucII fetal cells as described previously (Harper and Saunders, 1981; Francke et al., 1986). Probe concentrations in the hybridization solution were 25 and 50 ng/ml. Slides were exposed to photographic emulsion for 11 days. Post-hybridization G-banding was produced as described (Francke et al., 1986). The locations of labelled sites on the X-chromosome were recorded on a standard ideogram (Nesbitt and Francke, 1973).

We have shown previously that the human collagen transgene in HucII mice was integrated on the X chromosome. This transgene, including the 18 kb coding region and 1.6 kb 5′ and 20 kb 3′ flanking sequences, was expressed tissue-specifically and was as efficient as the endogenous gene (Wu et al., 1990a). To test whether the transgene was expressed when carried on the inactive X chromosome, S1 nuclease protection analysis was performed. A 5′ end-labelled, 1.1 kb EcoRI-TaqI fragment was used as shown in Fig. 1B. Since the TaqI site is present only in the human proα1(I) collagen gene, this probe gives rise to a 204 bases protected fragment from the human-specific proα1(I) mRNA but not from the endogenous mouse proα1(I) mRNA. A 3′ end-labelled mouse α2(I) probe, which generates a 122 bases protected fragment, was used as an internal control as described (Wu et al., 1990b). Fig. 1A shows the expression levels of the transgene in hemizygous males (lanes 1 and 2), homozygous females (lanes 3 and 4) and heterozygous females (lanes 5 and 6). The relative HucII expression levels in Fig. 1A were quantified on a Betagen betascope and expressed as a ratio of human α1(I) (204 bases protected fragment) to mouse α2(I) (122 bases protected fragment) (Table 1). If the level of expression seen in hemizygous males was set as 100%, homozygous and heterozygous females expressed approximately 125% and 62%, respectively. Assuming that the transgene on the active X chromosome in females was expressed at the same level as in males, these results suggest that the transgene on the inactive X was not silent but expressed at a level of about 20 –25%.

To establish genetically that the level of expression of the HucII gene was influenced by X chromosome inactivation, we utilized a mouse strain bearing one copy of a balanced reciprocal X:autosome translocation (T(X;16)16H) termed Searle’s translocation. In females heterozygous for the translocation, the normal X chromosome is always inactivated (Takagi, 1980; Rastan, 1983). It is thought that initially inactivation is random, but that selection eliminates cells that either have an X or an autosomal imbalance. X chromosomal imbalance will occur because X^T, not carrying Xce, cannot be inactivated. Autosomal imbalance can occur because inactivation initiated on 16^T which carrys Xce, may spread into the autosomal region (Takagi, 1980; McMahon and Monk, 1983; Rastan, 1983). The HucII transgene in females with Searle’s translocation will, therefore, be on the inactive X chromosome.

Males hemizygous for the HucII transgene were bred to heterozygous females carrying Searle’s translocation as shown in Fig. 2A. All female offspring with T16H carried the human collagen gene on their normal X chromosome as confirmed by Southern analysis (data not shown). RNA was isolated from the limbs of newborn females and Hprt isozyme analysis was performed on liver extracts (Kratzer et al., 1983). Normal XX females express the A and B allele of Hprt and can be distinguished from females carrying the translocated chromosome which express the B isozyme only (Fig. 2A). Fig. 2B shows the results of S1 nuclease protection analysis of nine newborn females. A 204-base protected fragment was present in all but lane 10 where the wild-type mouse RNA was used as a negative control. As expected, chromosomally normal heterozygous females (lanes 2, 7 and 9), which display random X inactivation, have a high level of HucII gene expression. In contrast, expression of the transgene in heterozygous females carrying Searle’s translocation (lanes 1, 3-6 and 8) was reduced to approximately 18% when compared to the expression level in hemizygous males (quantified by Betagen betascope, Table 1).

The incomplete X inactivation described above could result from reduced levels of transgene expression in every cell carrying the HucII gene on Xⁱ. Alternatively, it could be due to high levels of transgene expression in a subset of cells carrying the HucII gene on the Xⁱ, with the gene in the remaining cells completely inactivated. In an attempt to distinguish between these two possibilities, we bred a homozygous HucII female with a male carrying the Hprt^bm3 mutation, a deletional mutation in the Hprt gene (Hooper et al., 1987; Thompson et al., 1989). All female offspring from this cross are heterozygous for the HucII transgene and for the Hprt^bm3 mutation. Cells from a day-16 female embryo were explanted in tissue culture and selected in medium containing 6TG (6-thioguanine) for at least 40 days (see Materials and methods). Cells surviving 6TG selection (6TG^r) are expected to have the X chromosome carrying the HucII transgene inactivated and an active X chromosome carrying the Hprt mutation (X^a(Hprtbm3,+) X^{i(Hprtb,HucII)}). RNA was extracted from control cells and 6TG^r cells and HucII expression was quantitated by S1 analysis. Fig. 3 shows that transgene expression was reduced to approximately 5% of the level found in untreated cells (compare lanes 2 and 3) which was lower than the level observed in animals carrying the HucII gene on the Xⁱ.

To determine transgene expression at the single cell level, individual cell clones were isolated from a 6TG^r culture transformed with a retroviral vector containing SV40 large T antigen and the neo gene (Jat and Sharpe, 1986). The presence of the transgene was confirmed by Southern analysis (data not shown). Transgene expression was not detected in nine out of nine such clones while α2(I) collagen mRNA was expressed at normal levels (Fig. 3, lanes 4 to 12). The lack of HucII gene expression in cell clones indicated that the transgene carried on the inactive X can be completely inactivated in individual cells, suggesting that the low level of transgene expression in the animal may be due to a subset of cells expressing the transgene from the inactive X chromosome.

To test this hypothesis, we performed in situ hybridization on the SV40 transformed 6TG^r cells using a ³⁵S-labelled human COL1A1-specific oligonucleotide (Wu et al., 1990a, see Materials and methods). Nonspecific background was evaluated by hybridization of the same probe to a mouse fibroblast line which did not carry the human transgene (Fig. 4C). Unselected cells revealed grains over approximately 50% of the cells as expected for an X-linked gene (Fig. 4A). A similar number of grains was detected in a small number of cells in the 6TG^r cell population (Fig. 4B). To quantify the fraction of cells expressing the transgene, ten adjacent microscopical fields were screened for cells with grains. Approximately 3% of cells were positive (25 out of 785 cells), indicating that a subset of cells expresses the transgene from Xⁱ at a level which is comparable to the level expressed from the X^a.

The putative CpG-rich island in the human COL1A1 gene is located within a 2.5 kb DNA fragment, extending from 5′ non-transcribed region into the first intron, and contains more than sixteen HpaII sites. To investigate whether expression of the transgene correlated with DNA methylation, DNA was isolated from the cell populations or clones described in Fig. 3, and analyzed by Southern blots after digesting with HpaII (sensitive to methylation, H) or its isoschizomer MspI (insensitive to methylation, M). A 2.3 kb HindIII fragment (Wu et al., 1990a), extending from a position 0.85 kb 5′ of the transcription start site into the first intron of the human proα1(I) collagen gene, was used as a probe. As shown in Fig. 5A, two bands indicated by arrows in every “M” lane represent complete digestion products of the HucII gene. All bands of higher relative molecular mass represent incomplete digestion products (lanes H) due to partial methylation of the transgene. Significantly higher methylation levels were detected in cells with low (6TG^r cells, lane 3) or no (lanes 4 to12) transgene expression than in cells expressing the gene (untreated cells, lane 2). No cross-hybridization of the probe to the endogenous mouse collagen gene was found when the same amount of mouse genomic DNA was used as a control (lane 1).

To test whether the differential methylation of the 5′ region of the transgene observed in cell populations and individual cell clones reflected the methylation status of the transgene in animals, we performed similar analyses with DNA from transgenic males and heterozygous females carrying either two normal X chromosomes or one normal and one T16H translocated X chromosome. As shown in Fig. 5B, the HucII gene was undermethylated when carried on the male X chromosome (lane 1) but was highly methylated in females carrying Searle’s translocation (lanes 2, 4 and 5). The methylation level seen in normal females (lanes 3 and 6) was lower than in females carrying Searle’s translocation, which would be expected if the transgene is on the Xⁱ in all the cells in females carrying the translocation but only in about half of the cells in normal females. Thus, the higher methylation of the transgene in females carrying Searle’s translocation correlated well with the reduced expression of the gene when localized on the inactivated X chromosome.

The differential methylation was specific for the 5′ region of the transgene since no significant difference in DNA methylation levels were seen when the same filters were hybridized with a human COL1A1 cDNA probe (data not shown).

Completeness and stability of X inactivation of a given gene could be influenced by its position near the pseudoautosomal boundary or its distance from the Xce. To determine the localization of the HucII gene on the X chromosome, we carried out in situ hybridization by using the cosmid probe present in the HucII transgene to metaphase chromosomes of primary fetal fibroblast culture.

The fetal HucII culture had a male karyotype of 40, XY. Of 54 spreads analyzed, 27 were diploid, 4 triploid, 22 tetraploid and 1 octaploid. After autoradiography, a total of 80 X chromosomes were scored for silver grains and 25 labelled sites were noted. The nonrandom distribution of label with equal-sized peak over bands D and E (Fig. 6, left) suggests that the transgene is located in either band D or E near the D/E border. This corresponds to a location between the breakpoints in T(X;16)H16 Searle’s translocation and the Is(7;X)Ct1 insertion, distal to but close to Xce (Fig. 6, right).

The availability of mice carrying an autosomally derived transgene on the X chromosome provides an opportunity to investigate molecular mechanisms involved in X inactivation. The HucII strain carries a complete copy of the human COL1A1 gene on its X chromosome with flanking sequences exceeding the span of known regulatory regions of the gene (Wu et al., 1990a). The transgene in the HucII females was expressed at a relative level of 125%, and 62% in homozygous and heterozygous females, respectively compared with HucII hemizygous males. Furthermore, when expression was analyzed in animals with Searle’s translocation, the transgene carried on the normal X chromosome was found to be expressed at a highly reduced level (approximately 18% of that in hemizygous males). It has previously been shown that the normal X chromosome is inactivated in all cells of postnatal mice carrying this X-autosome translocation (Adler et al., 1991). Our results, therefore, raised the question of whether the partial inactivation observed in animals was due to (1) incomplete selection against cells in which the translocated X is inactivated; or (2) incomplete inactivation of the HucII locus either in all cells or in a subset of cells.

To address these questions, transgene expression was studied in individual cells carrying an Hprt mutation on one X chromosome allowing selection for cells which carried the transgene on the Xⁱ (6TG^r cells). A low level of transgene expression was detected in such 6TG^r cell populations. This is consistent with the notion that transgene expression in females with Searle’s translocation may be due to incomplete inactivation rather than incomplete selection. The analyses of transgene expression in individual cell clones as well as by in situ hybridization suggested that expression was completely repressed in most cells, but a subset of cells expressed the transgene at a normal level. Therefore, the partial escape from inactivation observed in animals is likely due to a fraction of cells that are capable of expressing the transgene on Xⁱ at a similar level as when carried on X^a. The level of transgene expression in females with Searle’s translocation was higher than would have been predicted if only 3% of the cells in the animal expressed the transgene from Xⁱ. We consider two possibilities to explain this discrepancy: (i) the transgene on the Xⁱ is expressed in a higher fraction of cells in animal tissues than in explanted cultured fibroblasts or (ii) the transgene, when carried on Xⁱ is inactivated in most or all cells, but becomes reactivated at later stages of development in a fraction of the cells. To distinguish between these two possibilities would also illuminate another unsolved problem, i.e., whether the inactivation process “skips” over a transgene which does not respond to the inactivation signal. Alternatively, a transgene, after being subjected to the sequential process of X inactivation, might become reactivated due to inefficient maintenance of the inactive state. Both modes have been observed for X-linked genes: (i) a number of genes escaping X inactivation are interspersed with inactivated genes mapping close to XIC or to the pseudoautosomal region (see Introduction) or (ii) reactivation of the Oct gene, which is located close to the centromere, occurs in liver of aged animals (Wareham et al., 1987). Analysis of transgene expression in tissues from animals of different ages would clarify this issue.

We asked whether the specific location of the transgene could be invoked to explain its occasional expression on the inactive X chromosome. Our chromosomal in situ hybridization data localized the human collagen transgene near the interface of bands XD and XE. With the T(X;16)16H translocation breakpoint cytogenetically assigned to band XD (Francke and Taggart, 1979), the transgene is located distal to the breakpoint in a region close to Xce (Brockdorff et al., 1991). Thus, sheer physical distance from the inactivation center, a factor that has been considered in explaining the reactivation of the subcentromic Oct gene in aged mice (Wareham et al., 1987), is unlikely to play a role. The transgene, on the other hand, may be located very close to Xist, that is consistently and exclusively transcribed from the inactive X. Thus, one could speculate that whatever mechanism is responsible for Xist expression may occasionally spill over to the transgene.

The maintenance of the inactive state appears to involve DNA methylation. It has been amply demonstrated that genes carried on the inactive X are hypermethylated in their 5′ regions as compared to their homologs on the active X. Furthermore, inactivated genes on the Xⁱ can be reactivated by interfering with maintenance of DNA methylation. We have compared the methylation pattern of the transgene when carried on the Xⁱ with that on the X^a. The 5′ end of the gene was undermethylated in cells expressing the HucII gene and highly methylated in cells not expressing the transgene. A similar correlation was also seen in the animals. No correlation between DNA methylation and gene expression has been seen with the endogenous Col1a1 and Col1a2 genes, i.e., the sequences flanking the promoter never became hypermethylated regardless of whether the gene was active or not (McKeon et al., 1982; Jähner and Jaenisch, 1985 and data not shown). When COL1A1 was integrated on Xⁱ, however, its promoter region became hypermethylated and transcriptionally inactivated supporting the hypothesis that the maintenance of the inactive state on X chromosome involves DNA methylation.

To date, the analyses of different autosomal genes when integrated into the X chromosome as transgenes have yielded different patterns of expression: the chicken transferrin gene was expressed when carried on either Xⁱ or X^a (Goldmann et al., 1987) while the AFP (Krumlauf et al., 1986) and the human collagen gene described in our work were subject to inactivation when carried on the Xⁱ. In the latter cases, a partial escape from inactivation was observed in the extraembryonic (in the case of AFP) or the somatic tissues (in the case of the human collagen transgene). With an alternative approach, Pravtcheva et al. (1991) have studied the expression patterns of an X-linked Pgk-1 gene when integrated into different regions of autosomes. They found that the late activation of the paternal Pgk-1 locus is not transgene specific but depends on its linkage to the X chromosome. Data derived from above transgenic mice suggest that cis-acting elements, involved in regulating gene activity, may not be major factors influencing the susceptibility of a given gene to X inactivation. Conclusive data, however, are lacking because so far only expression of single transgenes in one location on the X chromosome have been compared. To test more directly whether some regions on the X chromosome render genes more susceptible to inactivation while others do not, animals carrying the same autosomal genes in different positions on the X chromosome, or animals carrying different autosomal genes in the same location must be derived. Such experiments are technically feasible using the recently established methods of homologous recombination that allow similar-sized fragments of DNA to be placed in defined chromosomal regions.

We thank David Page and Andrew Zinn for comments on the manuscript, and Dr Rui-zhu Ling for statistic calculation. We thank Brigitta E. Foellmer for expert technical assistance in cytogenetics. This work was supported by National Institute of Health Grants PO1 HL41484 and (OIG) 5R35-CA44339 to R. Jaenisch and GM26105 to U. Francke. R. F. was supported by Max Kade Foundation. U. F. is an investigator of the Howard Hughes Medical Institute.