Methylation levels of maternal and paternal genomes during preimplantation development

Imprinting of each parental genome has been proposed to explain the requirement for the participation of both parental genomes in order to achieve full development in the mouse (Surani et al. 1984; McGrath and Solter, 1984). Many of the studies that point to the existence of genomic imprinting have been done in mice although there is increasing evidence for its involvement in a variety of human diseases (reviewed Reik, 1989; Hall. 1990). Genomic imprinting is assumed to occur independently during gametogenesis such that by the time of fertilisation epigenetic differences in the two genomes have been established. In this way the parental chromosomes have become imprinted as being of maternal or paternal origin. Genetic data suggest that only certain chromosomal domains are subject to overt imprinting, such that inheriting both copies of such regions from only one parent results in a recognisable phenotype which are often complementary (Searle and Beechey, 1985; Cattanach, 1986). Two endogenous genes, the insulin-like growth factor II and its receptor, have recently been identified as undergoing imprinting (Barlow et al. 1991; De Chiara et al. 1991). The means by which parental alleles are distinguishable and functionally non-equivalent are important and unresolved problems.

Since it is difficult to look at the two parental alleles separately, the random introduction of transgenes into the mouse genome has been used as a method for probing imprinted chromosomal regions by serendipity (Surani et al. 1988). It is then possible to follow a hemizygous transgenic locus in offspring that is inherited from one parent or the other. Independent observations of several different transgenic mouse strains have suggested that methylation may indeed be involved in the imprinting process, although it is far from clear whether or not it is the primary imprinting mechanism (Reik et al. 1987; Sapienza et al. 1987; Swain et al. 1987). In the majority of documented cases of ‘imprinted’ transgenes, the transgene was more highly methylated when inherited from the mother than when paternally inherited. This was sometimes associated with differences in expression, with the potential for expression being conferred by paternal transmission. Overall, sperm shows a lower proportion of methylcytosine than does any other adult somatic tissue (Gama- Sosa et al. 1983) and this is reflected in the hypomethylated state of both the major and minor satellite sequences (Sanford et al. 1984; Ponzetto-Zimmerman and Wolgemuth, 1984); however, other lower copy repeat sequences (Chapman et al. 1984; Sanford et al. 1987) and many unique gene sequences (Waalwijk and Flavell, 1978; Mandel and Chambón, 1979; Rahe et al. 1983) are highly methylated. Foetal and neonatal oocytes are believed to be drastically undermethylated both globally and at various endogenous repeat sequences (Sanford et al. 1984, 1987; Monk et al. 1987) and the ovulated oocyte is globally undermethylated (Monk et al. 1987). By extrapolation, the belief has been that the mature oocyte genome is still also undermethylated at repeat sequences and perhaps also at single copy sequences. Thus it is an attractive idea that methylation may play a role in establishing differences between parental chromosomes. Intermediate levels of methylation of several different repeat sequences observed in the early embryo were interpreted as reflecting the combination of the undermethylated maternal egg DNA and highly methylated paternal sperm DNA (Sanford et al. 1987). Hence, it was suggested that differential methylation is retained through the early cleavage divisions, prior to the large scale de novo methylation events at gastrulation (Monk, 1988), and this could contribute to the markedly different development of androgenetic and parthenogenetic embryos at these early stages.

As a means of assessing the methylation levels of the two parental genomes separately, we decided to compare parthenogenetic embryos, which contain only maternal chromosomes, with normal, fertilised ones. Ideally we would have included androgenetic embryos in this analysis, unfortunately they are difficult to make and more importantly develop very poorly (Surani et al. 1986). If indeed the oocyte genome was drastically undermethylated and remained so following fertilisation, parthenogenetic embryos would have been predicted to show very low levels of methylation. However, contrary to this prediction we found that the Murine Urinary Protein (MUP) and Intracisternal A Particle (IAP) families were highly methylated and LI repeat fairly methylated in the early parthenogenetic embryo. This led us to a systematic analysis of the methylation levels in sperm, oocytes, eggs and both normal and parthenogenetic embryos. We report here our results that suggest that methylation of several repeat sequences occurs prior to fertilisation in both parental genomes. In the case of the LI sequence where sperm shows higher levels of methylation than do eggs, methylation at the paternal sequence is lost during the first two cell cycles down to levels equivalent to that of the maternal sequences. Following this, methylation levels of each of these sequences are maintained until the 8- to 16- cell stage when there begins a gradual demethylation over the next 2 –3 cell cycles of the LI and IAP sequences.

C57BL/6J × CBA/Ca F₁ mice were used throughout. Female mice 4-6 weeks old were superovulated by intraperitoneal injection of Pregnant Mare’s Serum (PMS: Folligon, Intervet). Germinal-vesicle-stage oocytes were recovered 40 –44 h later from ovaries by puncturing with fine forceps and the oocytes released from the surrounding cumulus by careful pipetting up and down within fine bore pipettes. Oocytes were washed through PBI plus Bovine Serum Albumin (BSA, Sigma; Whittingham and Wales, 1969) and carefully scrutinised for adhering cells before collection. For eggs and embryos injection of PMS was followed 48 h later by injection of human Chorionic Gonadotrophin (hCG: Chorulon, Intervet), and for fertilised eggs females were caged overnight with Ft males and examined for the presence of vaginal plugs next morning (designated day 1). All unfertilised or fertilised eggs were recovered 18 –20 h post-hCG as described previously (Howlett et al. 1987) and cultured further in vitro.

Unfertilised eggs to be collected and analysed directly were dezonaed by brief exposure to warm Acid Tyrode’s solution, pH2.5 (Nicholson et al. 1985), washed in PBI plus BSA and carefully scrutinised for adhering cumulus cells. This procedure resulted in metaphase-II-arrested oocytes that were mostly devoid of first polar bodies.

For parthenogenetic embryos, eggs were activated 18.5 –19h post-hCG by exposure to 7% ethanol in T6 containing 4 mg ml ⁻¹ BSA (Howlett et al. 1987) for 4.5 min, washed and transferred to T6 plus BSA containing 0.5 mg ml ⁻¹ Cytochalasin B (Sigma) for 4h as described previously (Surani et al. 1984). Fertilised and diploid parthenogenetic embryos were cultured in T6 plus BSA under paraffin oil (BDH) at 37 °C in 5 % CO₂ in air until collection. Note that the concentration of sodium lactate in T6 should be 23.7 mM and not 0.045 mM as printed in error in Howlett et al. (1987).

Inner cell masses (ICMs) were isolated from day 5 blastocysts by immunosurgery (Solter and Knowles, 1975) as described previously (Barton et al. 1985).

The DNA polymerase inhibitor Aphidicolin (Sigma) was used at 2 μg ml ⁻¹ in T6 plus BSA from a stock stored at −20 °C at 2 mg ml ⁻¹ in dimethyl sulphoxide (Howlett, 1986). Cytochalasin D (Sigma) was used at 0.5 ¼g ml⁻¹ in T6 plus BSA to inhibit cell division from a stock stored at 0.5 mg mL in water at −20 °C.

All oocytes, eggs and embryos were collected in 200 –400 μl of a buffer containing 38mM NaCl, 15 mM Tris, pH8 and 5MM EDTA plus 0.5% sodium dodecyl sulphate and 100 μg ml ⁻¹ proteinase K and incubated for 1 –3 h at 55 °C. To aid precipitation of such small quantities of embryonic DNA 10 μg of E. coli tRNA was added as carrier before the mixtures were extracted twice with phenol and once with chloroform, the aqueous phase was then made 250 mM with NaCl and the DNA precipitated by addition of 2 volumes of absolute ethanol. Samples were stored at −70 °C. Sperm from Fj males was squeezed from isolated epididymides, collected in PBS and the DNA prepared as above except that 70 mM β- mercaptoethanol and 0.5 mg ml ⁻¹ proteinase K was used in the lysis solution. DNA was recovered by spinning at 14 000 revs min ⁻¹ for 10 min in an Eppendorf centrifuge, washed once with 200 μl of 70% alcohol, dried and redissolved in 10 mM Tris, pH 8, 1 MM EDTA. Restriction enzymes MspI and HpaYL were used under the conditions recommended by the suppliers (NBL). Restriction digest samples were electrophorcsed through 0.8% agarose gels, alkali-transferred to Gene Screen Plus or Hybond N filters and hybridised with LI-specific fragments labelled with a³²PdCTP by oligo-priming (Feinberg and Vogelstein, 1983). Some blots were washed in 1 % SDS, 0.1 SSC at 90 °C for 2 h and rehybridised with labelled IAP- (Lueders and Kuff, 1980) or MUP-specific fragments (Kuhn et al. 1984) many blots were washed and rehybridised with a mitochondrial probe to check for complete HpaH. digestion (Hecht et al. 1984). Blots were scanned on a Joyce Loebel chromoscan to enable methylation levels to be quantified. Methylation levels are given (indicated on relevant lanes of figures) such that 100 % means that all the label is found in the limited mobility band at the top of the gel, 50% indicates that half of this material has been digested down into smaller fragments.

Oocyte, egg or embryo samples were collected in 20 μl of sample buffer and run on 7.5 % polyacrylamide gels (Laemmh, 1970). Proteins were electrophoretically transferred to nitrocellulose paper in a buffer containing 20% methanol at 100 mA overnight at 4 °C. Blots were then washed and blocked with 3% BSA in PBS containing 0.1 % Tween-20 (Sigma; PBS-Tween) for 2h, washed 3 × 10min with PBS- Tween and incubated with a monoclonal methylase antibody (a subclone of M2B10, gift from T. Boehm; Pfeifer et al. 1985) for 4 –6h at a concentration of 10 –20 μg ml ⁻¹. After washing, a second layer peroxidase-conjugated rabbit anti-mouse antibody (Miles) was used at 1:400 dilution in PBS for 45 –60min and visualisation achieved with 4-chloro naphthol as substrate (as described previously; Jones, 1986). Photographs of the blots were scanned on a Joyce Loebel Chromoscan to enable amounts of protein in different embryo samples to be semi-quantitated.

In experiments leading to the observations reported here, we had observed rather high levels of methylation of LI sequences in parthenogenetic embryos (see below). This was surprising since, based on the previous findings of Sanford et al. (1984, 1987) that these sequences were virtually unmethylated in dictyate-stage foetal and newborn oocytes, we had expected to find very low levels of methylation in parthenogenones, which contain only maternally inherited chromosomes. We thus decided to determine the methylation status of oocytes nearer to the time of fertilisation and so ovulated, metaphase II eggs were collected. To avoid somatic contamination from cumulus cells, ovulated eggs were collected after removal of the zonae pellucidae. Thus, pure populations of eggs could be analysed and compared with sperm. DNA was prepared and restricted with the methylation-sensitive enzyme Hpall and the distribution of the LI repeat sequences compared. The level of methylation of the LI sequences in metaphase II eggs was found to be lower than in sperm but not as low as would have been predicted previously (Fig. 1). From scans of many independent blots, we estimate the average level of methylation in eggs to be 42%. In previous experiments, the level of methylation of LI sequences in dictyate-stage oocytes was found to be 5 % (Sanford et al. 1987). Using a probe recognising the intraci stern a IA particle (IAP) and murine urinary protein (MUP) families of repeats, both of which had been found to be undermethylated in dictyate-stage oocytes, we found these sequences to be completely methylated in ovulated eggs and hence indistinguishable from sperm (Figs 2 and 3). To confirm earlier results of undermethylation of these sequences in dictyate-stage oocytes, DNA was extracted from day 15 foetal ovaries (Fig. 4), subjected to methylation analysis using the LI and IAP probes and compared with foetal testes. From scans we find that about 8 % and 15% of LI and IAP sequences are unmethylated in foetal testes and ovaries, respectively (Fig. 4). Assuming that all of the unmethylated fraction is contributed by the germ cells in both testes and ovaries, we must conclude that both germ cell components are completely unmethylated at these sequences (see legend to Fig. 4 and Monk et al. 1987; Sanford et al. 1987). We also isolated immature oocytes from ovaries three days after birth and found LI and IAP sequences to be virtually unmethylated (data not shown). These observations suggested therefore that a considerable amount of de novo methylation occurs during the growth and/or maturation phase of oogenesis.

Using a monoclonal antibody against the methylase protein complex for Western blot analysis (Pfeifer et al. 1985), we detected extremely high levels of methylase protein in the ovulated egg which were already present by the GV stage (Fig. 5). This is in agreement with the suggestion that the egg possesses high levels of methyiase activity, as measured by an in vitro assay (Monk et al. 1991).

We were interested to see how methylation levels on maternal and paternal chromosomes would behave during preimplantation development. In particular, it was of interest to see whether or not the differences detected at the LI sequences were going to persist for a number of cell divisions following fertilisation. Methylation was therefore compared for the three sequences in normal fertilised and parthenogenetic embryos. For the LI sequences, methylation in the 1-cell fertilised egg was considerably higher (72 %) than in the ovulated egg (47%) or parthenogenetic 1-cell (42%, Fig. 6, compare lane b with lanes a and c). This is consistent with high methylation levels persisting on the paternal chromosomes whereas maternal chromosomes remain at about 40-45 % methylation. However, by day 2 in the late 2-cell fertilised embryo, it appears that most of the sperm-specific methylation has been lost, so that fertilised embryos show very similar levels of methylation to that of eggs and parthenogenetic 1- or 2-cell embryos (Figs 1 and 6). This same intermediate level of methylation remains in fertilised embryos on day 3 and begins to drop on day 4 around the time of cavitation reaching about 13–15 % in the blastocyst on day 5 (Fig. 1). The rates of preimplantation development of parthenogenetic embryos parallels that of fertilised embryos so that cavitation, for example, occurs in both embryos at about 32 cells and at the same time from fertilisation or activation (S.K.H., unpublished observation). Care was taken to analyse the two types of embryos when at the same developmental stages, which was achieved by collecting fertilised embryos 6h ahead of parthenogenones (6h being the estimated time between that of in vivo fertilisation and experimental ethanol activation). It was, however, often found that demethylation occurred more slowly in parthenogenones than in fertilised embryos, such that methylation appeared to remain slightly higher on days 4 and 5 in parthenogenones (Fig. 1). This observation may well be due to the slight developmental slowing of parthenogenones in culture that is first apparent on day 5 (S.K.H., unpublished observations).

The level of methylation of the two other repeated sequences, the IAP and MUP families, was also analysed comparing normal and parthenogenetic embryos. Both the IAP and MUP families were very highly methylated in the early embryo and there was no difference between normal and parthenogenetic embryos. Some demethylation of the IAP sequences was apparent in the blastocyst (Fig. 2). Unfortunately the signal with the MUP probe was not strong enough to detect any evidence of demethylation in the blastocyst. Whether or not other repeated gene families and indeed single copy sequences become demethylated during preimplantation development remains to be seen, but the overall methylation of the blastocyst genome is clearly very low (Monk et al. 1987).

The levels of methyiase protein were also determined in the two types of embryos at different stages (Fig. 5). Interestingly, the loss of methylation at the LI sequences was paralleled by a marked decrease in the levels of methyiase in the embryo as has been shown for methyiase activity (Monk et al. 1991). 8-cell embryos still contained 50–60% of the levels found in the egg, and this level decreased to about 30% on day 4 to be virtually undetectable in the day 5 blastocyst (Fig. 5). Thus, on a per embryo basis there is about a 20-fold loss of the protein between the egg and the blastocyst which would correspond to about a 2000-fold decrease per cell. Again we noted that enzyme levels tended to remain slightly higher in the parthenogenetic embryo.

The process of demethylation of the LI sequence was investigated in more detail by the use of drugs to prevent either DNA replication or cell division. Blocking DNA replication in the 8-cell embryo with aphidicolin allows cleavage to 16 cells but not DNA synthesis during the 4th cell cycle (Smith and Johnson, 1985). Embryos were put into aphidicolin as 8 cells and cultured for a further 24 h when they were arrested at 16 cells; analysis of their DNA showed that demethylation had been almost completely prevented by inhibiting DNA replication (Fig. 7, compare lane b with a and c). 8-cell embryos were also placed into cytochalasin D (CCD) to prevent cell division but not DNA replication. They were cultured for a further 24 h, by which time the nuclear number had increased to 16. The extent of demethylation in such CCD-arrested embryos was similar to that of control day 4 embryos and higher than in aphidicolin-treated embryos (Fig. 7, compare lanes e and f with d, g and h). This suggested that DNA replication but not cell division per se is required for demethylation and that it begins during the S phase of the 4th cell cycle.

By the blastocyst stage, the two lineages of troph- ectoderm cells and inner cell mass (ICM) cells have been established. The trophectodermal and primitive endodermal lineages are known to be markedly undermethylated for both centromeric and dispersed repeat sequences and structural genes while methylation of the same sequences is high in the primitive ectodermal lineages of the embryo itself (Chapman et al. 1984; Rossant et al. 1986). Therefore, it was of interest to see if the low methylation levels of LI sequences in the blastocyst solely reflected the trophectodermal contribution. Thus, ICMs were isolated on day 5 and compared with whole blastocysts. Methylation levels of isolated ICMs were also found to be low and appeared very similar to the pattern in the whole embryo (Fig. 7, compare lanes i and j). It appears therefore that there is a fairly uniform low level of methylation of LI sequences in the blastocyst across both cell lineages. This is at variance with results from rabbit embryos, where high levels of methylation were found in the ICM as compared to the trophectoderm (Manes and Menzel, 1981). In the mouse embryo, global methylation levels seem to increase in the ICM shortly following implantation, and continue to rise during gastrulation (Monk 1988).

Previous experiments demonstrated that a number of repeated sequences in the mouse genome are undermethylated in foetal germ cells. Thus, LI, MUP and IAP sequences are undermethylated in female foetal germ cells around days 13 – 15 (Sanford et al. 1987; Monk et al. 1987 and Fig. 4), LI (Monk et al. 1987) and IAP (Fig. 4) sequences are also undermethylated in male foetal germ cells. However, in mature spermatozoa as well as in spermatids and pachytene spermatocytes, these sequences are highly methylated, suggesting that there are major de novo methylation events in the male germline, perhaps associated with the resumption of mitosis in germ cells after birth (Sanford et al. 1987). Our results suggest that there is also de novo methylation during oogenesis that most probably occurs during growth of the oocyte such that by the time of ovulation (and probably even by the germinal vesicle stage, our unpublished observations) the female genome has become methylated at several repeat sequences. The extent to which different repeat sequences become methylated in the oocyte varies, with IAP and MUP sequences apparently becoming as methylated as in sperm, whereas the LI sequence reaches a level where about half of the sequences are methylated. This contrasts with the previously determined level of 5 % in dictyate-stage oocytes (Sanford et al. 1987). As a consequence, the two parental genomes are perhaps more similar in terms of methylation status at the time of fertilisation than might have been predicted. This observation seems to make it unlikely that genome-wide methylation is a means by which imprinting of the maternal and paternal chromosomes during preimplantation development is achieved. In the only other analysis of methylation in oocytes and eggs, it was found that the oocyte genome was globally less methylated than the sperm genome (Monk et al. 1987). It is therefore quite possible that there exist other repeat and unique sequences that are differently methylated in egg and sperm. Also, the methylation status of low copy sequences, while thought to be relatively high in sperm (Waalwijk and Flavell, 1978; Mandel and Chambón, 1979; Sturm and Taylor, 1981; Ponzetto-Zimmerman and Wolgemuth, 1984), remains to be determined in eggs. Despite the obvious absence of any genome-wide differences in our experiments, it is still quite possible that certain sex-specific methylation differences exist and hold the key to the parental identity of alleles (see below). So far it is not clear whether imprinted transgenic sequences show any methylation differences between the egg and the sperm copy at the time of fertilisation. However, our observation of de novo methylation during oogenesis may be pertinent to transgene imprinting, as the great majority of imprinted transgenes are repressed following maternal inheritance (Surani et al. 1988).

We made use of parthenogenetic embryos as a convenient way of looking at the behaviour of the maternal chromosomes in isolation and compared them with normal fertilised embryos where both parental chromosomes are represented. Our results suggest that the high levels of methylation at the MUP and IAP sequences that are found in both egg and sperm are retained through early development and that there begins to be signs of demethylation of both maternal and paternal copies in the blastocyst at least for the IAP family. For the LI repeat the same levels of about 40 % methylation of the maternal chromosomes are retained through to the morula stage and following cavitation there is a marked decrease down to very low levels in the blastocyst. The highly methylated sperm-derived LI sequences, however, lose nearly half of their methylation during the first two cell cycles such that by the late 2-cell stage (after 2 rounds of DNA replication) they show an identical level of methylation as do the maternal chromosomes, they then behave in a similar manner to the maternal copies with another marked demethylation beginning after the third cell cycle. This second demethylation at the LI sequences seems to coincide with the time of onset of cavitation. Indeed observations of differences in timing of this demethylation in different strains of mice whose embryos cavitate with different cell numbers suggest that there may be a link between demethylation and cavitation (S.K.H., unpublished observations). Since our experiments with inhibitors show a dependence of the demethylation on DNA replication and the earliest morphogenetic events in the preimplantation mouse embryo including compaction and cavitation seem to depend on the number of nuclear divisions (Surani et al. 1980; Smith and Johnson, 1985), the link may simply reflect the number of replication cycles that the genome experiences.

High global levels of methylation were found in earlier studies in the preimplantation mouse (Singer et al. 1979) and also rabbit (Manes and Menzel, 1981) embryo. In the rabbit the overall level of methylation decreased to about half from the 2-cell to the early blastocyst stage (Manes and Menzel, 1981) and a similar decrease in the global level of methylation has been found in the mouse preimplantation embryo (Monk et al. 1987). This is accompanied by a drastic reduction in the levels of methylase protein per cell during preimplantation development. However, the activity and specificity of the methylase is unclear since methylation levels of some sequences is maintained but not others - for example, the sperm-derived LI sequences during the first two cell cycles. It will be interesting to determine the methylation status of unique genes during the first few cycles to see whether the sperm LI demethylation can be extrapolated to single copy genes; if so, it may explain how paternal genes become expressible by the late 2-cell stage (for example, Sawicki et al. 1981). Why the egg should accumulate such abnormally high levels of methylase remains obscure, but it may explain why viral sequences introduced during early preimplantation development become highly methylated (Jahner et al. 1982). When analysing the methylation of repeated sequences in the genome, it is impossible to tell whether any intermediate levels of methylation reflect differences between individual copies of the repeats, or between individual cells in the embryo, or both. The apparent requirement for continuing DNA replication for demethylation (and not to a demethylating activity as has been proposed; Gjerset and Martin, 1982) suggests that one component of the demethylation process could involve the loss of activity of maintenance methylase. Loss of maintenance methylation can potentially generate differences in methylation between cells, which may contribute to the mechanism of mosaic expression recently observed in transgenic mice (McGowan et al. 1989).

The direct comparison between mature sperm and egg genomes shows that there is a fair degree of similarity in the methylation levels of several repeat sequences at the time of fertilisation, and furthermore it appears that where differences still exist (LI repeats) they may be removed following the first two rounds of DNA replication such that both paternal and maternal genomes then undergo the same developmental changes in methylation through into the blastocyst. However, there is an indication that subtle but significant differences may persist for longer. The ratio of the two major digestion products of LI, at 3.5 kb and at 5 kb (Fig. 1), was significantly different in normal and parthenogenetic embryos at least until the blastocyst stage (Fig. 8). Since the 3.5 kb and 5 kb MspI bands presumably reflect two subfamilies of the LI repeat, it is possible that one of these retains methylation slightly longer on paternal and the other one on maternal chromosomes. This indication will, however, require more attention in further experiments.

Our observations highlight the occurrence of major methylation and demethylation events in mouse gametogenesis and early embryogenesis and the apparently tight developmental control. The functional significance of these changes, however, remains largely elusive.

We thank all the members of the Reik and the Surani labs, and in particular Azim Surani, for help, discussions and comments on the manuscript and G. P. Pfeifer and T. L. J. Boehm for the kind gift of the methylase antibody. We thank L. Notton and D. Styles for their patience during the preparation of this manuscript. W.R. is a fellow of the Lister Institute of Preventive Medicine. S.K.H. is supported by a TAP grant from the AFRC, in addition work was supported by a grant from Combat Huntington’s Chorea.