Temporal and regional changes in DNA methylation in the embryonic, extraembryonic and germ cell lineages during mouse embryo development

Methylation of the pyrimidine base cytosine in DNA may be one of the mechanisms underlying differential programming of cell lineages in mammalian development. In somatic tissues approximately 2-5 % cytosines are methylated and the vast majority of these occur in the dinucleotide CpG and hence symmetrically on both DNA strands. Hemimethylated DNA produced by DNA replication becomes fully methylated by the action of a maintenance methylase. Differential patterns of methylation of CpG sites in the genome therefore provide stable and heritable information capable of regulating expression of specific genes (reviewed in Doerfler, 1983), of chromosomes domains (Naveh-Many & Cedar, 1981) and of whole chromosomes (e.g. the active and inactive X chromosomes in female cells, reviewed in Monk, 1986).

Mechanisms must exist that establish different patterns of methylation during embryonic development. Generally the idea of a methylated ‘ground state’ for development has been favoured (Singer, Robert-Ems, Luthardt & Riggs, 1979; Razin & Riggs, 1980; Jâhner & Jaenisch, 1984). This would be in keeping with demethylation being associated with the onset of expression of specific genes. In addition, direct measurement has suggested high levels of methylation in preimplantation mouse (Singer et al. 1979) and rabbit (Manes & Menzel, 1981) embryos, as well as in teratocarcinoma cells (Singer et al. 1979). Furthermore, Jâhner, Stuhlmann, Stewart, Harbers, Lôhler, Simon & Jaenisch (1982) have argued for the presence of a de novo methylase activity during preimplantation development and in embryonal carcinoma cells (Stewart, Stuhlmann, Jâhner & Jaenisch, 1982), to explain the methylation of Moloney murine leukaemia virus sequences following integration into the host genome.

Differential methylation has also been proposed as a basis for the onset of X-chromosome inactivation in female embryos. During development, inactivation of an X chromosome occurs first in extraembryonic lineages as they differentiate and later in embryonic tissue (Monk & Harper, 1979; Takagi, Sugawara & Sasaki, 1982). The preferential inactivation of the paternal X chromosome in the early extraembryonic lineages (Takagi & Sasaki, 1975; West, Freis, Chapman & Papaioannou, 1977; Harper, Fosten & Monk, 1982) suggests that paternal and maternal genomes are distinguishable (imprinting). During ontogeny of the female germ line the inactive X chromosome is reactivated (Monk & McLaren, 1981). Could differential methylation and changing patterns of methylation be the molecular bases for these events?

In this work we have developed a highly sensitive technique, using only several hundred cells, for looking at genomic DNA methylation. We have examined methylation in oocytes and sperm, preimplantation mouse embryos, different lineages dissected from postimplantation embryos, and purified fetal germ cells, and we have established the timing and pattern of changes in methylation during mouse embryo development.

All samples were collected in TNE (100 mM-NaCl, 50 mM-Tris pH 7·5, 5mM-EDTA) containing 2% sodium dodecyl sulphate and 200 μg ml^-1 protease K and stored at -70°C. When sufficient tissue was collected; samples were pooled for DNA preparation.

Sperm was collected from the testis and epididymis. Eggs were collected from female MFI mice (OLAC) superovulated by intraperitoneal injection of 5 i.u. of PMS (pregnant mare serum) followed 45 to 48 h later by 5 i.u. of HCG (human chorionic gonadotrophin). The following day approximately 1500 ovulated eggs were collected from the oviducts of unmated females. The eggs were treated with hyaluronidase (300i.u. ml^-1 in PB1 medium; Whittingham & Wales, 1969) for 5 min to remove cumulus cells. Residual adhering cumulus cells were removed by finely drawn pipettes.

From mated females approximately 500 8-cell embryos were flushed from the oviducts on the third day of pregnancy, and around 200 blastocysts from the uterine horns on the fourth day of pregnancy. Approximately 60 post-implantation embryos at and days’ gestation were dissected from the uteri of naturally mated females. The egg cylinders were treated with trypsin and pancreatin to loosen the cells layers. The -day egg cylinders were dissected into epiblast, extraembryonic ectoderm and primary endoderm as described in Monk & Harper (1979), and the -day egg cylinders into embryonic and extraembryonic (chorion) portions.

From -days’ gestation gonads can easily be recognized as male or female; male gonads show developing testicular cords. At days, embryos must first be sexed by examination of amnion cell nuclei for sex chromatin (Burgoyne, Tam & Evans, 1983) to identify gonads as male or female (part of this work was kindly performed by Mr Phillip Lee). At least 30 gonads, together with attached mesonephros tissue, were isolated from postimplantation embryos from to days’ gestation and the gonad and mesonephros were dissected apart. Embryonic kidneys, extraembryonic placentae and yolk sacs were also isolated. The mesoderm and endoderm components of the yolk sac were separated as described in McMahon, Fosten & Monk (1983).

Approximately 90 % pure populations of germ cells were obtained from isolated - and -day female and male gonads as follows. Gonads were incubated in phosphatebuffered saline, containing 0·4 % polyvinylpyrrolidone and 0·2% EDTA, at room temperature for 30 min. The germ cells were released by rupture and squeezing of the gonads and purified by gently wafting them away, in a stream of medium, from gonadal somatic cells and red blood cells, keeping them always grouped together. They were collected into microcaps for storage at —70°C. Germ cells were collected in this way from between 150 and 200 gonads for each sample. We are grateful to Susan Lindsay for help with the collections.

For the larger amounts of tissue, DNA was isolated as described in Flavell, Kooter, de Boer, Little & Williamson (1978). DNA was prepared from sperm nuclei according to Bird, Taggart & Gehring (1981). For the samples of low cell numbers, DNA was isolated by ultracentrifugation as follows. Samples stored at -70°C were pooled, the volume made up to 2 ml with TNE (2 % SDS, 200 μg ml^-1 protease K, see above), incubated at 37°C for Ih (to allow protease K digestion), phenol-extracted two times and two volumes of absolute alcohol added to the aqueous phase. Following ultracentrifugation (25 000 rev min^-1, 18 h) the deposit was gently washed with 70% alcohol, resuspended in 30 μl of Tris 10mM pH7-4, EDTA ImM, RNase 25,μgml^-1, and stored at —20°C ready for use.

To ensure the reproducibility of results, collections of biological material and DNA preparations were made on at least two separate occasions for each stage of embryonic development, except the pure germ cell isolations that were performed once only.

DNA preparations were digested for 1 h with MspI or HpaII (enzymes from BioLabs). Completeness of digestion was assured by the use of excess of enzyme (about 100-fold) on the minute quantities of DNA involved. Completeness of digestion and lack of degradation was supported by the reproducibility of the results from replicate experiments with independent DNA preparations, and the appearance of bands representing repetitive sequences or mitochondrial DNA (see Results). The digests were end-labelled with [α-³²P]dCTP using the Klenow fragment of DNA polymerase I (Maniatis, Fritsch & Sambrook, 1982). Uncut (mock-digested) DNA was end-labelled in each experiment to ensure the high molecular weight of the starting DNA preparations and verify the absence of degradation (e.g. see Fig. 1). A HindIII digest of 5 ng of λ DNA was end-labelled to serve as markers. The end-labelled DNA was subjected to electrophoresis (16 h, 25 V) on a 0-7 % agarose gel (0-5 to Ing DNA per slot) and Southern blot (Southern, 1975) of the labelled fragments. The fragment size distribution could be observed following exposure of the filters to X-ray film (about 3 days). Transfer of DNA from the 0·7 % gels to the filters was complete since only background counts were detectable in the gel following blotting.

Densitometry was performed with a Joyce Loebl Chromoscan 3. Variation of degree of methylation is correlated with both the total number of fragments (labelled ends) as well as the shape of the distribution of fragments at the tops of the lanes. Comparisons of total counts as an indication of number of fragments proved unreliable due to uneven loading of the minute quantities of DNA in some of the gels and the differences being confined mainly to the tops of the lanes. However, a measure of the shapes of the Hpali fragment distributions of the different DNAs could be obtained from the ratio of number of fragments at positions 10 and 3 kb. The 10/3 ratios for the different DNAs were normalized to the 10/3 ratio of a reference DNA on each gel. Results determined in this way were reproducible for replicate experiments for independent isolates of a particular tissue DNA.

DNA preparations were digested with Mspl and Hpali. Completion of digestion was monitored by digestion of A phage DNA added to a sample of the main digest. Digests were loaded at 0·1 μg per lane on 0·7 % agarose gels and electrophoresed for 16 h at 25 V. Following transfer of the DNA fragments to nitrocellulose by Southern blot, they were hybridized to ³²P-labelled nick-translated probe (pMR134 or mitochondrial probe, 200 ng, approximately 2x10⁸ dissintsmin^-1μg^-1) for 48 h, the filters were washed and exposed to X-ray film for approximately 1 week.

DNA was isolated from mature sperm, ovulated unfertilized oocytes, 8-cell embryos, blastocysts and separated embryonic and extraembryonic lineages of postimplantation conceptuses as described in the Materials and methods. To study methylation, DNA preparations were digested with the isoschizomeric enzymes Mspï and Hpali which cut the DNA at the four-base sequence, CCGG. Mspl will cleave the DNA at this sequence whether the internal cytosine is methylated or not, whereas Hpali will only cut the unmethylated sequence (Waalwijk & Flavell, 1978a). A comparison of Mspl and HspH fragment sizes therefore gives an indication of the extent of methylation of the total genomic DNA. Fragment-size distributions could be observed by end-labelling the DNA fragments with [a-³²P]dCTP, using the Klenow fragment of DNA polymerase I, followed by electrophoresis, blotting to nitrocellulose filters and autoradiography. The method is extremely sensitive and global methylation can be examined in 1 ng of DNA (from approximately 200 cells). It must be borne in mind in interpreting these experiments that the endlabelling procedure is an indication of numbers of fragments of a particular size and not the molar amount of DNA in the different regions of the gel. The high M, DNA is therefore under-represented since ends are so few. Nevertheless the results are reproducible and the technique allows comparisons between different DNA isolates.

Fig. 1 shows a representative autoradiograph of end-labelled uncut samples of the DNA preparations used in this work. The low level of labelling at the top of the lanes indicates large fragments and few ends and that therefore the starting DNA is high M_T. Very slight degradation (labelled ends of low M_r fragments at the bottom of the lanes) is seen for some of the DNAs.

Fig. 2 shows a representative experiment with endlabelled embryonic DNAs following Mspl and Hpali digestion. The heavy labelling at the bottom of all the lanes is due to the high proportion of fragments derived from the islands of clustered CpG sequences known to occur upstream from many genes (Bird, Taggart & Gehring, 1985). These islands are unmethylated (except on the inactive X chromosome). An indication of the degree of methylation of the total DNA of a particular tissue may be observed by a comparison of Mspl and HpaU digests at the tops of the lanes; methylated DNA is indicated by a distribution of the label skewed towards the higher M_r in the Hpall lanes. Densitometry tracings were made from all replicate gels and a measure for comparison of fragment-size distribution in the different DNAs was obtained from the ratio of fragments at positions 10 and 3 kb (indicated in Fig. 2). According to this criterion (see below and Table 1) the sperm and 8-cell DNA are more methylated than oocyte DNA, and the blastocyst DNA shows very low methylation indeed.

For the separated lineages of the postimplantation conceptuses, also shown in Fig. 2, the DNA of the -day embryonic region (epiblast) shows only a slight skewing of label towards high M_r methylated fragments. One day later, the DNA of the embryonic region of the -day conceptus is highly methylated. The DNAs of the extraembryonic lineages in the postimplantation conceptuses, the primary endoderm (and extraembryonic ectoderm, data not shown in Fig. 2) at -days’ gestation, and the chorion at -days’ gestation, are more methylated than blastocyst DNA, yet undermethylated compared with the -day embryonic tissue DNA (and see Table 1).

The repetitive DNA fragments seen in the oocyte DNA, and to a lesser extent in the 8-cell and blastocyst DNAs in Fig. 2, are Mspl fragments from mitochondrial DNA. Fig. 3 shows an Mspl digest of liver DNA hybridized with a mixture of three probes together covering the entire mitochondrial DNA (Kearsey, Flannagan & Craig, 1980). The band pattern agrees well with that observed for the repeated fragments in the end-labelled total DNA of the oocytes and preimplantation embryos.

During mouse fetal development the primordial germ cells reach the developing gonads at around 11 days’ gestation and by 13 days’ gestation female germ cells have entered into the prophase of the first meiotic division. The previously inactive X chromosome is reactivated by about 13 days’ gestation (Monk & McLaren, 1981). At this time male germ cells have entered mitotic arrest; further rounds of mitosis and subsequent entry into meiosis will begin after birth.

In order to examine the methylation status of the fetal germ cell DNA during the period of X-reactivation and initiation of meiosis in the female germ cells, DNA was prepared from female and male fetal germ cells at and days’ gestation and from whole gonads, containing both germ cells and supporting somatic cells from to days’ gestation. In addition, DNA was prepared from -day embryonic somatic tissue, the mesonephros tissue (dissected apart from the developing gonad) and kidneys, from extraembryonic tissues (yolk sac endoderm, yolk sac mesoderm and placenta), and from an embryo-derived stem cell line (HD14, Evans & Kaufman, 1981).

Fragment size distributions are shown in Fig. 4 for Mspl and Hpall digests of the DNA from - and -day female and male germ cells, from DNAs of the separated yolk sac endoderm and mesoderm components and from embryo-derived stem cells. Once again the majority of ends are generated by the Hpall tiny fragments from the methylation-free islands of CpG sites in both the Mspl and the Hpall digests. Densitometry traces were taken from all replicate gels and the traces measured at positions 10 and 3 kb. This analysis (see below, and Table 1) shows that the fetal germ cell DNAs are markedly undermethylated compared to the yolk sac DNAs, and all other DNAs.

Quantification of absolute levels of methylation in the different tissues is not possible due to the bulk of the fragments from the CpG islands. Nevertheless comparisons between tissues could be made by analysis of ratios of fragments at positions 10 and 3 kb in the Hpall lanes. After logarithmic transformation to reduce the dependence of variance on the mean, the data were analysed by a sequential modification of the Q method for analysis of variance (Snedecor & Cochran, 1967). Table 1 summarizes the data obtained from replicate analyses of the DNAs with the log mean values arranged in descending order of degree of methylation. The data are reproducible and show different levels of DNA methylation with respect to skewing of the Hpall distributions towards heavier fragments. The most methylated DNA is -day embryonic DNA. DNAs from -day epiblast, from extraembryonic portions of - and -day conceptuses and from 8-cell embryos are significantly less methylated than -day embryonic DNA. DNAs from oocytes, blastocysts and fetal germ cells are strikingly undermethylated and significantly less methylated than all the other DNAs, including sperm DNA.

There were no significant differences between tissues for the ratios of fragments at 10 and 3 kb in the Mspl digests. In addition to the comparison of Hpall lanes alone, a further analysis was done to compare values for the Hpall ratio divided by the Mspl ratio from the accompanying Mspl lane for each DNA in each experiment. Although this analysis showed greater variation in replicate pairs of lanes than did the analysis of the Hpall lanes alone, the Hpall 10/3: Mspl 10/3 values showed the same rank order as the Hpall ratios in Table 1.

The fetal germ cells are markedly undermethylated and this could be in part the molecular basis for the X-chromosome reactivation. In this respect, the interesting question is whether there is a specific demethylation event in the germ line. There is evidence from dosage of hypoxanthine phosphoribosyl transferase activity (Monk & McLaren, 1981) and expression of phosphoglycerate kinase isozymes (McMahon, Fosten & Monk, 1981), that the inactive X chromosome may already be reactivated by days’ gestation. To analyse methylation of DNA from pure germ cells earlier, at days’ gestation, would be difficult, not only because each embryo must be individually sexed by sex chromatin in the amnion, but also due to the difficulty in isolating sufficient germ cells. We therefore decided to look at methylation of the DNA in whole gonads in the region of a dispersed repetitive sequence, MIF (mouse interspersed family recognized by probe pMR134, Bennet, Hill, Pietras, Woodworth-Gutai, Kane-Haas, Houston, Heath & Hastie, 1984, approximately 3x10⁴ copies of 5-7kb per haploid genome). The probe was kindly provided by Nick Hastie (MRC Clinical and Population Cytogenetics Unit, Edinburgh).

Fig. 5 shows hybridization of probe pMR134 to the DNA preparations of the different tissues of the female and male conceptuses and also to sperm DNA, and DNA from an embryo stem cell line. The Mspl lanes show two bands of hybridization whereas Hpall digests of DNAs from somatic tissues, female and male embryonic kidneys and mesonephros, show a high molecular weight smear of hybridizing DNA. Therefore the MIF dispersed repeat sequence detected by pMR134 is methylated in somatic tissues. Placental DNA and yolk sac endoderm DNA show a //pall pattern indicating reduced methylation with a smear extending down to clearly visible bands corresponding to those in the Mspl lanes. Yolk sac mesoderm DNA however is methylated to a similar extent to somatic DNA. These results confirm work already published by others (Chapman, Forrester, Sanford, Hastie & Rossant, 1984) and are included here for comparison of methylated and undcrmcthylated DNAs with the gonadal DNA and to show that tissue-specific differences in methylation estimated from end-label analysis are supported by the findings with the repetitive sequence.

Both female and male gonad DNA from -day embryos show the unmethylated bands as well as methylated DNA at the top of the lanes. At days’ gestation the female gonads continue to show an unmethylated component in the Hpali digests. The unmethylated bands appear fainter for the -day male gonads and for mature sperm DNA. Embryoderived stem cell (EK) DNA shows an undermethylated pattern similar to yolk sac endoderm DNA. Similar tissue-specific differences in methylation were obtained for a satellite repetitive sequence (data not shown; Sanford, Forrester, Chandley & Hastie, 1984).

To show that the unmethylated component of the gonad DNA was due to the germ cells, some of the purified germ cell DNAs were used to hybridize to the MIF probe. Fig. 6 shows that the purified germ cell DNAs show predominantly the unmethylated bands with probe pMR134. There is little hybridization at the top of the lanes in the absence of DNA from the somatic component of the gonads. Having thus established that the unmethylated MIF sequences in the gonad could be ascribed to the germ cells we analysed DNAs from gonads isolated from embryos from through days’ gestation. The results are shown in Fig. 7. Although the hybridization of the -day DNAs is poor due to paucity of DNA and long exposure to X-ray film it is nevertheless clear that -day germ cells are already unmethylated for these repeat sequences.

We have developed techniques for the isolation of micro-amounts of DNA from early embryonic lineages and subsequent analysis of overall genomic DNA methylation by end-labelling. The end-label technique allows comparisons of overall genomic DNA methylation in the DNA of as few as 200 cells and could be extended to other areas where biological material is limiting such as stem cell systems and small pieces of tissue in fate-mapping studies. The quantification of the shape of the Hpall fragment distributions by ratios of fragments at 10 and 3 kb is to some extent arbitrary but nevertheless it is reproducible and allows an objective measure of tissue- and stage-specific differences. The validity of the quantification of methylation is supported by similar tissuespecific differences in DNA methylation obtained by hybridization techniques (this work; Chapman et al. 1984; Sanford et al. 1984; Sanford, Chapman & Rossant, 1985) and by direct measurement (Razin, Webb, Szyf, Yisraeli, Rosenthal, Naveh-Many, Sciaky-Gallili & Cedar, 1984).

The experiments described in this paper relate to several areas in development: imprinting of maternal and paternal genomes, demethylation and remethylation in early development, independent methylation of embryonic and extraembryonic lineages, X-inactivation and reactivation and the origin of the germ line. These aspects are discussed below.

Sanford et al. (1984) previously demonstrated undermethylation of immature oocyte DNA and methylation of sperm DNA for the dispersed repetitive sequences (MIF) and undermethylation of satellite sequences for both genomes. We have confirmed and extended their results to show global genomic undermethylation of oocytes and (relatively) global genomic methylation of sperm. In addition we have used ovulated oocytes that have completed the first meiotic division and therefore represent more accurately the maternal genome at the time of fertilization. Sperm DNA is known to be highly methylated in the region of a number of unique sequences (Waalwijk & Flavell, 1978b; Mandel & Chambón, 1979; Rahe, Erickson & Quinto, 1983). On the other hand sperm DNA appears to be unusual in that satellite sequences are undermethylated (Sturm & Taylor, 1981; Sanford et al. 1984; Ponzetto-Zimmerman & Wolgemuth, 1984; Monk, unpublished data). Overall, in our analysis, sperm DNA is undermethylated compared to DNA of somatic tissues (see also Razin et al. 1984), similar to DNA of extraembryonic tissue, and methylated compared to DNAs of oocytes, fetal germ cells and blastocysts. The difference in the degree of methylation of the maternal and paternal genomes at the onset of development suggests methylation as a good candidate for differential imprinting.

DNA methylation in the 8-cell embryo is compatible with a mixture of undermethylated maternal and methylated paternal DNA. There may be some maintenance methylase inherited in the egg cytoplasm but this appears to be lost or diluted out during cleavage since blastocyst DNA has very little methylation. The loss of sperm genome methylation could be one aspect of the activation of the paternal X-linked hypoxanthine phosphoribosyl transferase (HPRT) gene during preimplantation development (Kratzer & Gartler, 1978; Monk & Harper, 1978).

It has been proposed that a demethylation process is associated with the delineation of extraembryonic tissues (Manes & Menzel, 1981; Razin et al. 1984; Sanford et al. 1985). In this study we show that trophectoderm, and probably primary endoderm, are delineated at a stage when there is very little of the DNA methylation inherited in the sperm and oocyte genomes remaining. The blastocyst trophectoderm DNA, and also the primary endoderm DNA, must however become further methylated subsequently. Therefore, their patterns of methylation occur independently and potentially differently from those in the embryonic genomes. Overall genomic methylation in these extraembryonic lineages however does not occur to the final extent observed in the fetal somatic tissues. This has been previously reported in the mouse for total DNA (Razin el al. 1984) and for repetitive sequences (Chapman et al. 1984; also this paper) as well as for a number of unique gene sequences (Razin et al. 1984; Sanford et al. 1985) and in human, where global methylation of DNA of the placenta is less than that for somatic tissue DNA (Lindsay, Adams and Monk, unpublished data).

The onset of methylation in the fetal lineage probably begins around the time of implantation and occurs progressively over several cell generations. Recent work has shown that DNA from the ICM of the blastocyst of the mouse on the fourth day of pregnancy is not detectably methylated (Monk, Boubelik & Maidens, in preparation) but that onset of methylation is detectable one day later and increases progressively thereafter (this work). It has been suggested (Monk, 1986) that ‘reinforcing’ methylation may be superimposed on prior events regulating activity of genes or chromatin domains so as to stabilize or irreversibly inactivate regions that will never be required for a particular differentiated cell lineage. In fetal genomes much of the methylation occurs postgastrulation and hence methylation could be involved in the differential programming of the definitive germ layers.

The early X-inactivation in the extraembryonic tissues does not appear to be linked to a global methylation event. Rather, these tissues arise at a time when there is little methylation remaining. The preferential inactivation of the paternal X chromosome could be related to the higher methylation of the sperm genome compared to that of the oocyte, and residual methylation differences in the paternally inherited X chromosome at the blastocyst stage.

The initiation of X-chromosome inactivation in the embryonic region may coincide with the onset of de novo methylation in this tissue but further methylation follows and the primary X-inactivation event may well be initiated by other mechanisms.

The extremely low methylation in fetal germ cells may be one aspect of the molecular basis of X-chromosome reactivation in that the low methylation condition may allow, or enable, reactivation. However, a specific demethylation event as the-basis for X-chromosome reactivation does not appear to occur; as early as days’ gestation, prior to X-reactivation, fetal germ cells are already undermethylated.

Fig. 8 shows a working model for stage- and tissuespecific demethylation and de novo methylation in early development. Strikingly undermethylated DNA is depicted by a single line, and increasing and decreasing methylation as changing thickness of line. The two alternative paths for the origin of the germ line are shown as interrupted lines - one prior to de novo methylation, the other involving methylation and demethylation. We favour the hypothesis that the germ line delineates independently prior to the delineation of the three germ layers, and prior to any extensive methylation of the fetal precursor cell genomes. First, at the earliest stage we have examined, germ cell DNA is already undermethylated. Second, on theoretical grounds, the germ line genome must not be subjected to ‘reinforcing methylation’ since it must remain totipotent, or permissive to reprogramming to developmental totipotency.

In early mouse development, following loss of methylation of the paternal genome DNA, methylation occurs independently and to different extents in the primary germ layers, the extraembryonic lineages and in the germ line. The onset of methylation in development will be regulated by the timing of activation of the embryonic genes for maintenance methylase and the extent of methylation may be determined by an interplay of lineage-specific chromatin structure and methylase activities. Extensive methylation may be a mechanism of reinforcing or stabilizing active and inactive chromatin (e.g. the inactive X chromosome) and the germ cell lineage may well be delineated before this occurs.

We thank Anne McLaren, Robin Lovell-Badge and Peta Maidens for their interest and their critical reading of the manuscript, Anne McLaren for advice on statistical analysis of the data and Peta Maidens for help with densitometry and photography.