Inactivation of an X-linked transgene in murine extraembryonic and adult tissues

Dosage compensation in mammals is achieved by inactivation of one of the two X chromosomes in female cells resulting in a single active X chromosome being present in both XX females and XY males (Lyon, 1961). This process takes place during early embryogenesis and seems associated with cell differentiation (Monk and Harper, 1979). Xinactivation in the mouse first occurs in cells of the trophectoderm and the primitive endoderm at 3.5 and 4.5 days of gestation, respectively. In these cells that will participate in the formation of the extraembryonic tissues, the X chromosome of paternal origin is preferentially inactivated (Lyon and Rastan, 1984; Takagi and Sasaki, 1975; West et al., 1977). X-inactivation then occurs between 5.5 and 6 days of gestation in the primitive ectoderm (epiblast), which will give rise not only to the three germ layers of the fetus but also to the extraembryonic mesoderm (which is part of the yolk sac). In all these tissues, X-inactivation occurs at random (Gardner and Lyon, 1971; Rastan, 1982; Takagi, 1974). Once it has occurred, X-chromosome inactivation is both highly stable and heritable in a clonal fashion.

This process is thought to be under the control of a region, acting in cis, called the X-inactivation center (Xic) located by genetic and cytogenetic studies on the X chromosome distal to the T(X:16)16H translocation breakpoint within band X_D (Avner, 1989). The random inactivation observed in adult tissues can be slightly influenced by different alleles of a putative locus called Xce (X controlling element), which maps to the same span as the postulated inactivation center itself (Cattanach, 1975). Recently, a human gene, XIST (inactive X-specific transcript) and its murine homologue Xist have been identified and localised to the XIC candidate region in humans and mouse (Borsani et al., 1991; Brockdorff et al., 1991; Brown et al., 1991). Xist displays an unusual pattern of transcription in that it is transcribed exclusively from the inactive X chromosome. The location of this sequence to the region of Xic and its unique pattern of expression has led to suggest that it may play an important role in the inactivation process, and may possibly correspond to the X-inactivation center itself.

Although the fundamental mechanism of this regulatory process involving cis control over long genetic distances is not yet understood, it has been shown that in adult tissues, genes present on the X chromosome are subject to specific methylation, linked to the inactivation process. Housekeeping genes such as Hprt and Pgk-1 are hypermethylated in the 5′ portion of the gene when present on the inactive X (Hansen and Gartler, 1990; Keith et al., 1986; Lock et al., 1986; Wolf et al., 1984; Yen et al., 1984). This methylation process seems to occur after the initial steps of chromosome inactivation (Lock et al., 1987) as a secondary mechanism involved in the maintenance of the inactivation in adult tissues.

Additional information concerning the inactivation process has come from the analysis of autosomal material either inserted or translocated into the X chromosome. An apparent spreading of inactivation into autosomal sequences was shown, although this spreading can be limited or incomplete (Lyon, 1983; Russell, 1983).

A number of transgenes located on the X chromosome have been used as models to analyse the inactivation process. The results obtained are not completely coherent: whilst the chicken transferrin transgene studied by Goldman et al. (1987) appears to escape inactivation in adult tissues, both a mouse α-fetoprotein transgene (Krumlauf et al., 1986) and a rat vasopressin transgene (R. Behringer, unpublished results) show correct regulation of expression when present on the inactive X chromosome in adult tissues. One hypothesis advanced to explain the inactivation escape of the chicken transferrin gene was that the non mammalian chicken genome (in which X-inactivation does not occur) either lacks specific signal sequences or is resistant to X-inactivation in the mouse. It was therefore of interest to study the inactivation sensitivity of a non eukaryotic gene which had been integrated into the X chromosome. One such transgene is the bacterial neomycin resistance gene (neo) present in the transgenic mouse strain MTKneo1 (Stewart et al., 1987).

We analysed the regulation of expression of this transgene on the inactive X chromosome in both adult somatic tissues and extraembryonic tissues; our results show that inactivation of the neo transgene as well as methylation of its internal CpG sites are similar to those of endogenous Xlinked genes. Furthermore, we have shown that, at the level of transcription, paternal inactivation of the transgene seems complete in the extraembryonic endoderm.

The study of this neo transgene clearly establishes the capacity of non eukaryotic sequences to respond to the Xchromosome regulation not only in adult somatic cell lineages but also in extraembryonic lineages.

The M-TKneo1 transgenic strain obtained after retroviral infection of embryo cells was derived from the ICR strain (Stewart et al., 1987). The retroviral vector MMCV-neo (Wagner et al., 1985) is present as a single copy on the X chromosome in this strain (see map in Fig. 3C) and carries the neomycin phosphotransferase gene from Tn5 under the control of an internal herpes simplex virus thymidine kinase (TK) promoter. This vector also contains a v-myc gene under the control of the LTR sequences. All MTKneo1 animals are homozygous for the transgene.

The T(X;16)16H translocation (Searle’s translocation) mouse strain, the XO female mice and the C3H/He-Pgk-1a mouse strain were obtained from the breeding facility of the Pasteur Institute.

In mice carrying the Searle’s translocation, the reciprocal translocation has occurred between the X chromosome and chromosome 16: in cells from adult females heterozygous for this translocation, both segments of the rearranged X are active, whereas the normal X chromosome is inactive.

The pSV2neo plasmid was labeled by nick translation to a specific activity of 1.0×108 disints/minute/μg. Lymphocytes from homozygous M-TKneo1 female mice were cultured under standard conditions in the presence of Concanavalin A for 72 hours at 37°C, 5-bromo-deoxyuridine being added for the final six hours of culture to ensure good post-hybridisation chromosomal R-banding.

Metaphase cells were hybridised as previously described (Dautigny et al., 1986). Slides were covered with Kodak NTB2 nuclear track emulsion and exposed at 4°C. After development the chromosome spreads were first stained with buffered Giemsa solution and cell metaphases photographed. R-banding was then performed using the fluorochrome-photolysis-giemsa (F. P. G.) technique and the metaphases rephotographed prior to analysis.

Total RNA from adult organs was homogenised in guanidinium thiocyanate and purified by centrifugation on a cesium chloride cushion according to described methods (Chirgwin et al., 1979).

For northern analysis, 20 μg of total RNA were denatured at 68°C for 10 minutes in 50% formamide, 2.2 M formaldehyde, 5 mM sodium acetate, 1 mM EDTA, 20 mM MOPS pH 7.0 and separated by electrophoresis in 1% agarose, 2.2 M formaldehyde gel in MOPS buffer (Sambrook et al., 1989). The RNAs were transferred to a nylon membrane (Hybond N, Amersham) in 10× SSC (1× SSC is 0.15 M NaCl, 0.015 M sodium citrate) and crosslinked to the dry filter by UV irradiation (0.16 kJ/m2) according to Church and Gilbert (1984).

Genomic DNA was isolated from adult organs as described (Sambrook et al., 1989). After restriction enzyme digestion, DNA was separated on 0.8% agarose gels and Southern blotted to nylon membranes (Hybond N, Amersham). DNA was crosslinked to the dry filters by UV irradiation.

The neomycin probe used for northern and Southern analysis was a 1.1 kb BamHI-BglII fragment from pSV2neo. The probes were labeled with α[32P]dCTP (Amersham) using the random priming method (Feinberg and Vogelstein, 1984). The β₂-microglobulin probe is a 300 bp EcoRI fragment from a mouse β₂-microglobulin cDNA clone (Daniel et al., 1983) kindly provided by G. Gachelin (Institut Pasteur, Paris).

Prehybridisation and hybridisation were performed in Church buffer (Church and Gilbert, 1984; 1 mM EDTA, 0.5 M NaHPO₄, pH 7.2, 7% NaDodSO₄) at 65°C overnight. Blots were washed in 2× SSC for 15 minutes at room temperature, then 2× SSC for 30 minutes at 65°C and 0.4× SSC for 30 minutes at 65°C and exposed to Kodak XAR-5 films with an intensifying screen at −70°C.

Reciprocal crosses between transgenic M-TKneo1 animals and C3H/He-Pgk-1a mice were established, females tested daily for the presence of vaginal plugs and embryos taken at 14.5 days post coitum. These embryos were sexed under the microscope by analysis of the genital ridges. Individual female yolk sacs were rinsed in PBS and incubated for 2 hours at 4°C in 2.5% pancreatin, 0.5% trypsin, 0.5% polyvinylpyrrolidone, the reaction being stopped by a 10 minute incubation in DMEM medium supplemented with 10% fetal calf serum. Mesoderm and endoderm were then separated using watchmaker’s forceps (Hogan et al., 1986). Pools of female mesoderm or endoderm were flash frozen in liquid nitrogen and kept at −80°C.

Total RNA was extracted with acid phenol/chloroform from pools of 5 endoderms or mesoderms homogenised in guanidinium thiocyanate according to described methods (Chomczynski and Sacchi, 1987). After precipitation with isopropanol, samples were treated with RNase-free DNase for 15 minutes at 37°C and reprecipitated after phenol/chloroform extraction. The average yield of total RNA from a single yolk sac varied between 20 and 40 μg. This method of RNA extraction proved to be more efficient with small quantities of material than the guanidinium thiocyanate/cesium chloride method.

The BamHI-HindIII fragment from the pSV2neo plasmid was subcloned into Bluescript KS. This plasmid was linearised with NcoI and in vitro transcribed with T7 polymerase (Promega Biotec), using α[32P]CTP (Amersham): the labeled antisense RNA probe is 633 nt long (including 55 nt from the multiple cloning region of Bluescript).

This probe (2×10⁵ disints/minute/sample) was hybridised with 20 μg of either yolk sac endoderm or mesoderm RNA in 20 μl of 75% formamide, 0.5 M NaCl, 1 mM EDTA, 20 mM Tris Ph 7.5 overnight at 45°C after a denaturation step of 15 minutes at 85°C. Hybrids were treated with RNase A (Sigma, 40 μg/ml) and RNase T1 (Sigma, 2 μg/ml) in 0.3 M NaCl, 5 mM EDTA, 10 mM Tris pH 7.5 at 37°C for 30 minutes. After proteinase K treatment (0.3 mg/ml in 0.1% NaDodSO₄ at 37°C for 30 minutes), phenol/chloroform extraction and ethanol precipitation, the samples were loaded on 8 M urea 6% acrylamide gels in borate buffer.

10 μg of yolk sac endoderm or mesoderm RNA were reverse transcribed with 10 units of AMV reverse transcriptase and 2 μg of random hexamer primers according to Kawasaki (1990). Samples were treated with DNase-free RNase (10 μg) for 15 minutes at 37°C and ethanol precipitated. For each RNA sample, a control experiment was carried out in parallel omitting the reverse transcriptase. Serial dilutions were prepared from each cDNA sample (500, 50 and 5 ng) and amplified for 20 cycles using Taq polymerase (Cetus) and 0.2 μmol of the primers (92°C, 20 seconds; 65°C, 30 seconds; 70°C, 1 minute) in a total volume of 50 μl.

30 μl of each reaction product were electrophoresed on a 2% agarose gel, transferred to Hybond N+ (Amersham) and hybridised to random primed neo or actin probes (Pst1 fragment of pAL41 β actin) (Alonso et al., 1986). These PCR conditions (low number of cycles and transfer of the amplified products to a membrane, followed by hybridisation) are such that amplification of cDNAs was linear and dependent on the concentration of the initial RNA.

The neo primers used were:

5′GTCCTGATAGCGGTCCGCCA3′ and 5′GTGTTCCGGCTGTCAGCGCA3′ as previously described (Kim and Smithies, 1988).

The actin primers used were:

5′ACCCACACTGTGCCCATCTAC3′ and 5′CTTCATGGTGCTAGGAGCCAG3′ (Alonso et al., 1986).

In situ hybridisation using the neo probe was performed on metaphase spreads of lymphocytes from transgenic MTKneo1 animals. Only metaphases showing silver grains on the X chromosome were analysed. The majority of the grains, i.e. 19 out of a total of 26 (approximately 73%) were detected on bands XD and XE and correspond to the probable localisation of the transgene on the X chromosome (Fig. 1).

The retroviral vector MMCV-neo (Wagner et al., 1985) is present as a single copy in the M-TKneo1 strain. Previous studies have shown a tissue-specific expression of the neomycin resistance gene in male and female homozygous transgenic mice (Stewart et al., 1987); a 2.5 kb transcript presumably originating from the internal TK promoter was detected in M-TKneo1 heart, lung, muscle, kidney, testis and brain but not in liver, intestine, spleen, thymus and pancreas.

In order to investigate if the neo transgene was correctly inactivated in adult tissues, a cross between females heterozygous for the T(X;16)16H translocation (Searle’s translocation) and M-TKneo1 males was established. Both X,Xneo and TX:16,Xneo female progeny were obtained. Since the T16H translocation stock carries the Tabby (Ta) coat color marker on the non translocated X, the female progeny carrying the translocation can be easily distinguished by their non Tabby coat color.

In these females (TX:16,Xneo), both segments of the rearranged X resulting from the reciprocal exchange between the X chromosome and chromosome 16 remain active and the normal Xneo chromosome derived from the transgenic stock is uniformly inactivated in all cells.

Northern analysis of total RNA from several tissues of these females showed the absence of neo expression (Fig. 2, upper panel), whereas the presence of a 2.5 kb transcript was detected in control homozygous M-TKneo1 lung, heart and kidney but not liver RNA, as previously described.

These results show that in those adult tissues in which it is expressed, the neo transgene is subject to X-inactivation as other endogenous genes.

The restriction map of the transgene was established using several enzymes including the methylation-sensitive restriction enzyme AvaI. Three AvaI sites were localised, one in the neo coding region, one in the TK promoter region and one in the myc coding region (Fig. 3C).

In order to establish a possible correlation between methylation and the X-inactivation state of the transgene, spleen and brain DNAs from several transgenic animals were digested with AvaI and HindIII. Southern blot hybridisation of spleen DNAs to the neo probe is shown in Fig. 3A.

1. In both hemizygous M-TKneo1 male mice and XO females carrying the neo transgene (XneoO), a major 1500 bp DNA fragment is detected, due to complete digestion at the AvaI sites in the neo coding region and TK promoter region. The 2000 bp fragment detected in the spleen DNA of these animals results from partial digestion at the AvaI site located in the TK promoter and may be linked to a tissue-specific methylation pattern (see below).

2. In the TX:16,Xneo female carrying the neo transgene on the normal X chromosome (always inactive in this cross), a single 9000 bp DNA fragment appears (HindIIIHindIII fragment), indicative of complete methylation of the three AvaI sites of the transgene present on the inactive X chromosome.

3. In the homozygous M-TKneo1 female, subject to random X inactivation, the 1500 and 2000 bp fragments from the active X chromosome are detected, as well as the 9000 bp fragment from the inactive X chromosome.

Similar results were obtained using two other restriction enzymes, SalI and XhoI, which contain a CpG dinucleotide in their recognition site (data not shown).

DNA from spleen (in which the transgene is not expressed) and from brain (in which it is expressed) gave identical AvaI methylation patterns except for the 2000 bp fragment. Figure 3B shows the Southern blot analysis of M-TKneo1 male DNA from these tissues. The 2000 bp fragment present in the spleen is absent from brain DNA, suggesting that methylation of the AvaI site located in the TK promoter could be involved in the observed tissuespecific expression of the transgene.

These results suggest that specific methylation of the transgene is associated with inactivation of the X chromosome and in particular that a bacterial gene such as the neomycin resistance gene can be subject, in the context of the X chromosome, to the same DNA methylation modifications as normal X-linked eukaryotic genes.

Extraembryonic tissues display a different pattern of regulation compared to embryonic tissues and show preferential inactivation of the paternal X chromosome. In order to study the expression of the neo transgene in extraembryonic tissues, yolk sacs of 14.5 day female embryos from a cross between male and female M-TKneo1 mice were dissected into endoderm (derived from the primitive endoderm) and mesoderm (derived from the epiblast) and total RNA was prepared from these tissues. Since neo mRNA levels in yolk sac tissues were undetectable by northern analysis, we performed the more sensitive RNase A mapping technique using a 633 nt neo antisense RNA probe synthetised in vitro by T7 polymerase, of which 578 nt are complementary to the neo RNA. Neo mRNAs were present in both endoderm and mesoderm of M-TKneo1 animals.

Densitometry analysis (Fig. 4B) of the bands detected on the gel revealed a twofold lower level of RNA in yolk sac mesoderm and a fivefold lower level in yolk sac endoderm compared to the amount in adult tissues, in accordance with the difficulty of detection by northern analysis.

In order to determine if the neo transgene was inactivatable in extraembryonic tissues, we set up a cross between a T16H female and an M-TKneo1 male and analysed the endoderm and mesoderm from female 14.5 day embryos. Since both X,Xneo and TX:16,Xneo female embryos are obtained from this cross, it was first necessary to analyse RNA from whole embryos in order to identify the TX:16,Xneo embryos characterised by the absence of neo expression and the X,Xneo embryos characterised by its expression. Litter sizes were very small and from the three female embryos obtained, only one embryo was found to be TX:16,Xneo. RNA from the yolk sac endoderm and mesoderm of this embryo was analysed by RNase A mapping: no expression of the neo transgene was detected in these tissues (Fig. 4, lane TE and TM) indicating that the transgene was inactivatable in extraembryonic tissues.

As a control for expression in these tissues of an endogenous X-linked gene, Pgk-1^a/Pgk-1^b isozyme analysis was performed. No expression of the paternally inherited Pgk1^b isozyme was observed in the endoderm cells of animals from cross Pa, whereas expression of both isozymes was detected in the mesoderm cells with predominant expression of the maternally transmitted gene, as previously described, possibly due to the Xcec allele present in the C3H/He-Pgk-1^a strain (West et al., 1977) (data not shown).

Total RNA from these tissues was then analysed by RNase A mapping. The 578 nt long protected neo mRNA fragment was present in the yolk sac endoderm of embryos from cross Ma (Fig. 4, lane MaE), but not in the yolk sac endoderm of the embryos from cross Pa (lane PaE). The yolk sac mesoderm of embryos from both crosses was also analysed as a control, since X-inactivation occurs at random in these tissues (lane MaM and PaM) and showed as expected the presence of neo mRNA.

In order to confirm this result and analyse more precisely the amount of neo mRNA present in the yolk sac endoderm and mesoderm, a quantitative RT-PCR experiment was performed on RNA from the Pa and Ma crosses, using neo primers and actin primers as a control. No neo expression was detected in the YS endoderm RNA when the transgene was carried on the paternal chromosome, whereas neo mRNA was present in the endoderm of the Ma cross and in the YS mesoderm of both crosses (Fig. 5).

From these results using RNase mapping and the highly sensitive RT-PCR technique, we conclude that the neo transgene is subject to preferential paternal inactivation as other X-linked endogenous genes in the extraembryonic endoderm. Furthermore, the result obtained from the RTPCR experiment shows for the first time that transcriptional activity of a gene carried by the paternal X chromosome seems apparently totally repressed.

Previous studies on transgenes inserted into the X chromosome have shown correct inactivation in adult somatic tissues of the mouse α-fetoprotein, SV40 TAg and rat vasopressin genes (Behringer et al., 1988; Krumlauf et al., 1986; R. Behringer, unpublished results). Only the chicken transferrin gene appears to escape inactivation (Goldman et al., 1987). This observation has been explained in several ways. The gene could have been inserted into a non-inactivated region of the mouse X chromosome. Several human genes such as MIC2 (Darling et al., 1986), ZFX (Schneider-Gädicke et al., 1989), RPS4 (Fisher et al., 1990) and A1S9 (Brown and Willard, 1989) have for example been shown to escape inactivation. Although the murine homologs Zfx, Rps4 and A1s9 are subject to inactivation in mouse (Adler et al., 1991; Ashworth et al., 1991; Zinn et al., 1991), it cannot be excluded that some murine X-linked genes may escape inactivation. The chicken genome (not subject to inactivation) could also lack the correct signals for inactivation in the mouse or could contain repressor sequences inhibiting X-inactivation, therefore suggesting the presence of specific domains or sequences involved in inactivation. The third hypothesis that the inserted unit of 187 kb is too large to be subject to the signals of X-inactivation is unlikely since this is in contradiction to studies on X-autosome translocations and insertions. Autosomal segments much larger than the 187 kb unit of the transferrin gene can, if linked to that part of the X chromosome carrying the inactivation center, be inactivated, probably due to a spreading phenomenon (Lyon, 1983; Russell, 1983).

The transgene present in the mouse strain M-TKneo1 is a retroviral vector carrying the bacterial neomycin resistance (neo) gene under the control of an internal herpes simplex virus thymidine kinase (TK) promoter. It has been previously shown that the neo gene in this transgenic strain was efficiently expressed under the control of this promoter in several male and female adult tissues (Stewart et al., 1987).

Our experiments were designed to determine the localisation of the neo transgene on the X chromosome, to analyse its expression in different tissues and to study the level of DNA methylation when present on the inactive X chromosome.

The analysis of the transgene insertion by in situ hybridisation provided evidence of its localization in the D or E band of the X chromosome. Preliminary breeding experiments have shown that the neo transgene and the Pgk-1 gene are tightly linked; this result is in agreement with the cytologically defined position since Pgk-1 lies proximally to the DXPas2 marker known to locate to the XD/XE band interface. This region of the X chromosome is known to undergo global X-inactivation. The isolation of mouse sequences flanking the transgene will allow us to determine more precisely the site of integration of the vector and define its positional relationship to other endogenous genes.

The use of a cross between the transgenic animal and a T16H female enabled us to study the regulation of expression of the neo transgene in adult tissues. In somatic cells from females heterozygous for the translocation, both segments of the rearranged X are active (Disteche et al., 1981; Lyon et al., 1964; Ohno and Lyon, 1965), whereas the normal X chromosome carrying the neo transgene is inactive. In these TX:16,Xneo females, no expression of neo is found in tissues such as lung, heart and kidney (in which the gene is normally expressed in M-TKneo1 animals) therefore showing that, in adult tissues, the prokaryotic neo sequence can be subject to X-inactivation.

It has been suggested that inactivation spreads outward from the inactivation center in a segmental fashion, using relays or ‘way stations’ on the chromosome to maintain the inactive state in adult somatic tissues (Gartler and Riggs, 1983).

If this is true, then the M-TKneo1 transgene would have integrated into the X chromosome in a segment that is under the control of these putative relays and would therefore be correctly inactivated at least in adult tissues. This correct regulation also implies that the prokaryotic derived DNA does not contain any sequence capable of inhibiting inactivation.

DNA methylation has been suggested as a maintenance mechanism involved in the X-inactivation process. Since the neo transgene was correctly inactivated in adult tissues, we investigated the methylation pattern of this exogenous gene and have shown that the AvaI restriction sites located within the transgene are methylated when present on the inactive X. This point is particularly interesting since it shows that a bacterial gene, normally only subject to restriction modification methylation, can not only be subject to inactivation like endogenous murine genes on the X chromosome but can also undergo specific methylation of CpGs like other genes on the inactive X.

Analysis of the methylation pattern in brain (in which neo is expressed) and spleen DNAs (in which neo is not expressed) shows a partial methylation of the AvaI site located upstream of neo in the TK promoter of the transgene in spleen DNA. This site had previously been shown to be involved in expression of the thymidine kinase gene (Christy and Scangos, 1982). Since this analysis was performed on male DNA, therefore excluding the involvement of methylation due to X-chromosome inactivation, we suggest that this particular methylation may be linked to the lack of expression seen in the spleen of transgenic animals and in some way related to tissue-specific transcriptional control.

Studies on several transgenes have shown the existence of a germline-specific imprinting of the transgenic locus, which is correlated with a differential methylation of the transgene according to its paternal or maternal origin. This genomic imprinting or methylation status of the transgene does not however seem to reflect the methylation status of the chromosomes, in the case of transgenes integrated in autosomal loci. It has been suggested that the transgene itself, by disrupting a chromosomal structure specifies its own methylation pattern (Reik et al., 1990; Sasaki et al., 1991). All transgenes, with the exception of one (Sapienza et al., 1987) show higher methylation of the transgene of maternal origin: it has been suggested that the absence of methylation in sperm is maintained in subsequent somatic divisions in the egg.

It is interesting to note that in the case of the neo transgene, the methylation status of the transgene is not linked to the parental origin of the chromosome; in XO females, the transgene is of paternal origin, whereas in the XY males, the transgene is of maternal origin; in both cases, the transgene is unmethylated. In the TX:16,Xneo females, the transgene is of paternal origin and is methylated. It is therefore clear that the methylation pattern seen with the neo transgene is specifically linked to X inactivation and not to a parental imprinting.

It can be concluded from our results that in adult tissues a transgene located on the X chromosome can be subject both to inactivation and to the de novo methylation which possibly acts as a maintenance mechanism on segmental domains of the X chromosome after it has been inactivated (Lock et al., 1987).

In the yolk sac of the 14.5 day embryo, endogenous Xlinked genes show a preferential paternal inactivation in extraembryonic endoderm, whereas these genes are subject to random inactivation in extraembryonic mesoderm (originating from the epiblast). It has previously been suggested that in the mouse the extraembryonic tissues carry a parental imprint at the time of inactivation leading to preferential paternal inactivation. The epiblast tissues, in which inactivation occurs later, undergo random inactivation possibly due to the loss of this imprint (Kaslow and Migeon, 1987).

For the analysis of the expression of the neo transgene in the yolk sac extraembryonic tissues, we used the RNase A mapping technique and RT-PCR detection, since the levels of neo mRNA were too low to be easily detected by northern analysis.

A comparison of the amounts of neo mRNA present in adult and extraembryonic tissues has for instance shown a fivefold reduction in the level of mRNA in the yolk sac endoderm compared to adult tissues. This may reflect a tissue-specific transcriptional control of the transgene in adult and extraembryonic tissues and could possibly be linked to the site of integration of the transgene in the X chromosome.

The cross established between a T16H female and the transgenic male enabled us to determine that the neo transgene is inactivatable in extraembryonic tissues, since it is not expressed either in yolk sac endoderm or mesoderm of TX:16,Xneo embryos.

The transgene is expressed in the mesoderm (of embryonic origin) of 14.5 day embryo yolk sac from both reciprocal crosses Ma and Pa, in which the transgene is maternally or paternally derived, suggesting that in this tissue neo is subject to random inactivation. In contrast, in the endoderm of these yolk sacs (of extraembryonic origin), the transgene is only expressed when maternally inherited. The T16H and these results taken together suggest therefore that not only is the neo transgene inactivated in extraembryonic tissues as are endogenous X-linked genes but also that it is subject to parental imprinting and shows preferential paternal inactivation in these extraembryonic tissues.

Many authors have previously described this preferential paternal inactivation; in particular, cytogenetic studies (Rastan and Cattanach, 1983; Takagi and Sasaki, 1975) and quantitation of Pgk-1 allozymes (Bücher et al., 1985) have been performed on extraembryonic tissues. Our experiments performed on yolk sac endoderm RNA clearly show that not only preferential but total paternal inactivation of the neo transgene is observed in this tissue, within, of course, the limit of sensitivity of the RT-PCR technique.

The neo transgene, despite its non eukaryotic bacterial origin, is therefore correctly inactivated in extraembryonic and adult tissues and can be both subject to X-inactivation and imprinted in the germ line as other paternally inherited endogenous X-linked genes. This is in contrast with another X-linked transgene, the α-fetoprotein transgene, which is submitted to X-inactivation in neonatal tissues but not in extraembryonic endoderm when present on the paternally derived X chromosome (Krumlauf et al., 1986).

Several explanations can account for the difference in expression of the neo and the α-fetoprotein transgenes. (i) Expression in the yolk sac endoderm may be due to the specific mode of regulation of the α-fetoprotein gene: since the endogenous gene is itself very strongly expressed in the yolk sac, it is possible that transcription factors have access to the transgene and override the signals of inactivation in these tissues (Krumlauf et al., 1986).

(ii) Since functional differences exist between the inactive X chromosome in extraembryonic versus adult tissues (Grant and Chapman, 1988), the inactivation mechanism may be affected by the nature of the insert: the α-fetoprotein transgene consists of several tandem copies of the minigene, whereas the neo transgene is present as a single copy retroviral insertion.

(iii) Alternatively, the differences of expression of these transgenes may be linked to a failure of imprinting rather than to X-inactivation itself. Since neither the neo nor the α-fetoprotein transgene would be expected to contain intrinsic X-chromosome imprinting signals, this suggests that the sites of integration of transgenes may play a critical role in the regulation of their expression in extraembryonic tissues and that target sequences involved in the control of inactivation and imprinting of X-linked genes may be distinct.

The finding that the neo transgene can be correctly inactivated in both adult somatic cells and extraembryonic lineages has important implications for the development of test systems for characterising the various genetic elements involved in the X-inactivation process, including candidate sequences for Xic itself.

We thank Richard Behringer for unpublished information on X-linked transgenes, Maria Luisa Gaspar and Sandra Pellegrini for many helpful discussions and Laurence Amar, Charles Babinet and Jean-Pierre Rousset for critical reading of the manuscript. The densitometry scans were kindly performed by Gerard Masson.

This work was supported by grants from the French Muscular Dystrophy Association (A. F. M.) and the Ministère de la Recherche et de l’Espace (M. R. E.).