Altered imprinted gene methylation and expression in completely ES cell-derived mouse fetuses: association with aberrant phenotypes

Culture and manipulation of mammalian preimplantation embryos can affect their subsequent development at fetal and perinatal stages. Pronounced effects have been observed in sheep and cattle where culture, nuclear transfer and asynchronous embryo transfer frequently result in the birth of larger than normal offspring with increased mortality and malformations (i.e. the ‘large calf syndrome’, reviewed in Walker et al., 1996). In the mouse, culture of preimplantation embryos may influence growth during fetal development (Bowman and McLaren, 1970), and also micromanipulation can have long-term influences on growth and gene expression (Römer et al., 1997). Phenotypic effects have also been observed in murine embryonic stem (ES) cells, which are derived from the inner cell mass of the blastocyst. Early-passage ES cells can be used to produce completely ES cell-derived fetuses (Nagy et al., 1990). However, upon prolonged culturing the potential of many ES cell lines becomes impaired, resulting in ES-derived fetuses with abnormalities such as increased size and mass, polydactyly, swollen edematous skin and perinatal death (Nagy et al., 1990, 1993; Wang et al., 1997). High passage stem cells may give rise to abnormalities even in chimeras, and frequently result in postnatal death of chimeras with high ES contributions (Nagy et al., 1990). It has been hypothesized that this loss of developmental potential results from the accumulation of epigenetic alterations in the ES cells and that these in particular affect imprinted genes (Nagy et al., 1993). Such ‘epimutations’ could also potentially explain the abnormalities observed in the large calf syndrome (Moore and Reik, 1996; Walker et al., 1996).

DNA methylation is one of the epigenetic features involved in the regulation of gene expression (Bird, 1995). One reason for believing that methylation may play a role in the changes observed on culture of stem cells and embryos, is that the genome is undergoing extensive changes in both global and gene-specific methylation patterns at precisely the stages when ES cells are being derived and manipulated. During preimplantation development, the bulk of the DNA in the genome becomes demethylated. As a consequence, only low levels of methylation persist at the blastocyst stage and hence in ES cells. Most genes acquire their fetal levels of methylation after implantation through de novo methylation (Monk et al., 1987; Frank et al., 1991; Kafri et al., 1992). For imprinted genes, genes whose expression is parent-of-origin-dependent, the developmental regulation of methylation seems to be rather different. All analysed imprinted genes show parental allele-specific methylation (John and Surani, 1996), and studies on embryos deficient in methyltransferase have established that this methylation is involved in at least the somatic maintenance of allele-specific gene expression (Li et al., 1993). In contrast to non-imprinted genes, imprinted genes already have considerable levels of methylation at the blastocyst stage which, for the insulin-like growth factor 2/cation-independent mannose-6-phosphate-receptor gene (Igf2r), (a region upstream of) the H19 gene, the U2af1-rs1 gene and the Snrpn gene, is at least partly derived from the gametes (Stöger et al., 1993; Tremblay et al., 1995; Feil et al., 1997; Shibata et al., 1997; Shemer et al., 1997). In other imprinted genes, such as the insulin-like growth factor 2 (Igf2) gene, these allelic patterns become established during or shortly after implantation (Brandeis et al., 1993). Although the complex regulation of imprinted gene methylation in the early embryo needs further elucidation, these data suggest that imprinted genes might be particularly prone to epigenetic mutation on derivation and culturing of stem cells. In order to investigate this, we selected a number of imprinted genes to characterise in vitro-induced alterations in DNA methylation, and determined the effects of such alterations on gene expression. For our analysis we chose two maternally expressed genes, H19 and Igf2r, and two paternally expressed genes, Igf2 and U2af1-rs1. The Igf2, H19 and Igf2r genes are known to be involved in the regulation of fetal and postnatal growth through the IGF signalling pathway, whereas U2af1-rs1 encodes a splicing factor, the biological function of which is presently unknown (John and Surani, 1996).

The methylation and expression patterns of these four imprinted genes were examined in interspecific hybrid ES cell lines and completely ES cell-derived fetuses, using sequence polymorphisms that allowed us to distinguish the parental alleles. This study confirmed that, after derivation and subsequent passaging, stem cell lines can give rise to developmentally compromised fetuses. The main finding is that, in all ES lines analysed, allelic methylation changes had occurred in the four imprinted marker genes. These epigenetic alterations persisted in ES fetuses in which they were associated with aberrant imprinted gene expression.

For the derivation of ES lines SF1-1, SF1-3 and SF1-8, (C57BL/6 × CBA/Ca) F1 eggs were in vitro fertilized with M. spretus sperm and cultured to the blastocyst stage, in KSOM medium with bovine serum albumin (Lawitts and Biggers, 1993). ES cells were derived as described before and cultured in ES medium (Allen et al., 1994). For ES cell line SF1-G, C57BL/6 female mice were mated with M. spretus males. Morula-stage embryos were flushed and ES cell derivation was performed as for the other lines. During the first passages (< passage 4), ES cells were cultured on STO feeder cells. Subsequent culturing was without feeders on gelatinized plates. All lines were karyotyped and chromosomes were counted before being used for the derivation of fetuses. Parthenogenetic ES lines PR4, 5, 9, 10, 13 and 14 were derived as before (Feil et al., 1997) and were grown in ES medium in the absence of feeder cells.

ES fetuses were produced by microinjection of 10-15 ES cells into tetraploid blastocysts, a modification of a methodology developed by Nagy et al. (1990). Initially, tetraploid blastocysts of the mouse line ROSA26 (which has a ubiquitously expressed lacZ reporter gene; a gift of P. Soriano) were used to determine the efficiency of the injection method. 28 day-11 fetuses were produced in which all the cells of the embryo proper were non-LacZ staining and hence ES cell-derived. The percentage contribution of diploid cells in all subsequently derived ES fetuses was studied by Southern blotting. In total, 40 day-11 ES fetuses (out of 125 transferred blastocysts), 58 day-13 to -14 fetuses (out of 255 blastocysts), and 3 day-15 ES fetuses (out of 40 blastocysts) were produced (only live animals with a beating heart were scored). Chimeric fetuses were made by injection of hybrid ES cells into naturally fertilized diploid blastocysts. Statistical comparisons between groups of chimeric fetuses were performed using ANOVA.

DNA extraction, endonuclease digestion, electrophoresis and Southern blotting were performed as described before (Feil et al., 1994). RNA isolation, electrophoresis on 1% formaldehyde gels, transfer to Hybond N⁺ filters (Amersham) and northern hybridisation were also as described before (Feil et al., 1994). The following probes were used: U2af1-rs1 probe 1 (550 bp EcoRI fragment; Feil et al., 1997); a 1094 bp SfuI-MluI fragment from region 2 of Igf2r (Stöger et al., 1993); a 1.8 kb Igf2r cDNA probe spanning the middle part of the cDNA; Igf2 probe 6 (0.9 kb KpnI-BamHI fragment; Feil et al., 1994); a 0.5 kb H19 fragment from the 3′ part of exon 1, and probe pY353/8, corresponding to a Y-chromosome repeat sequence (Bishop et al., 1983). Qualitative expression analysis was performed by RT-PCR amplification; primers used were: Igf2, 5′-GGCCCCGGAGAGACTCTGTGC-3′ and 5′-TGGGGGTGGGTAAGGAGAAAG-3′ (annealing at 60°C, 30 cycles); Igf2r, 5′-ATGATGACAGCGACGAAGACC-3′ and 5′-GAATTCGGCGCCACATGGTGTTCAAGAAG-3′ (57.1°C, 30 cycles); U2af1-rs1, 5′-TGTGGTACGGCCAGCCTATG-3′ and 5′-GATCAGACATACTGCGGATA-3′ (60°C, 35 cycles); H19, 5′-TGCTCCAAGGTGAAGCTGAAAG-3′ and 5′-GTAGGGCATGTTGAACACTTTATG-3′ (65°C, 30 cycles).

We derived (M. musculus × M. spretus) F1 ES cell lines of which four (SF1-1, SF1-3, SF1-8 and SF1-G) were used to produce ES cell-derived fetuses (ES fetuses). The four ES cell lines became stable in culture after an initial 6-8 passages, when most of the cells appeared morphologically undifferentiated. They were analysed for their chromosomes and found to have 40 chromosomes in >80% of chromosome spreads analysed. Southern hybridisation with a Y chromosome-specific probe showed that lines SF1-1, SF1-3 and SF1-8 were male, whereas line SF1-G was female (data not shown). ES fetuses, produced by injection of ES cells (passage 9-12) into tetraploid blastocysts, were analysed on day 13-14 of development, at which stage all fetal tissues were completely ES cell-derived (see Materials and methods). Allelic methylation and expression was determined for the Igf2, Igf2r, U2af1-rs1 and H19 genes in ES cells and ES fetuses (Figs 1 and 2; summarized in Table 1).

The U2af1-rs1 gene is paternally expressed in fetuses and adult tissues (Hatada et al., 1993), in which the gene and its upstream regulatory sequences are methylated on the maternal chromosome (Feil et al., 1997). Methylation was studied by digestion of a NotI site in the 5′ untranslated region (Fig. 1A) and levels of methylation at this restriction site were found to represent those of the entire gene (not shown). The NotI site was highly methylated on the maternal allele in ES cell line SF1-1, whereas in lines SF1-3, SF1-8 and SF1-G, it was unmethylated on both parental chromosomes. Most of the SF1-1 derived fetuses showed complete maternal methylation, whereas a minority showed partial maternal methylation, and in one fetus (fetus 172), U2af1-rs1 was unmethylated. In all fetuses derived from lines SF1-3, SF1-8 and SF1-G, both parental alleles of U2af1-rs1 were unmethylated. Qualitative expression was analysed by RT-PCR (Fig. 2A). In ES line SF1-1, expression was predominantly paternal, and most derived ES fetuses showed (maternal U2af1-rs1 methylation and) expression from the paternal chromosome (e.g. fetus 170). In other SF1-1-derived fetuses (e.g. fetus 171), in which only a proportion of the maternal chromosomes was methylated, equal expression was observed from both parental chromosomes. One SF1-1 fetus was obtained (fetus 172; with both U2af1-rs1 alleles unmethylated) which showed biallelic expression, albeit more strongly from the maternal than from the paternal chromosome. Biallelic expression was also observed in SF1-3, SF1-8 and SF1-G ES cells and ES fetuses, in which U2af1-rs1 was unmethylated on both parental chromosomes. In all instances, therefore, allelic expression of U2af1-rs1 was inversely correlated with allelic methylation.

Methylation in the H19 gene was assessed by analysing the allelic digestion of eight HpaII restriction sites located within (the first three exons of) the gene and its promoter (Fig. 1B). These HpaII sites are paternally methylated in normal embryonic and adult tissues (Bartolomei et al., 1993). Allelic H19 expression was analysed by RT-PCR amplification (Fig. 2B). SF1-G ES cells and ES fetuses displayed the normal (Sasaki et al., 1995) paternal methylation in the H19 gene and expression was exclusively maternal (Table 1). ES cell lines SF1-1, SF1-3 and SF1-8 were predominantly (∼85%) paternally methylated, and these three cell lines and most of the derived fetuses, had predominantly maternal (∼85%) expression (Figs 1B, 2B). One SF1-8 derived fetus (fetus 25) had higher levels of maternal H19 methylation and displayed predominantly paternal expression. Therefore, as with U2af1-rs1 expression, H19 expression was inversely correlated with methylation.

Methylation in the paternally expressed Igf2 gene was assayed in the ‘differentially methylated region 2’ (DMR2), a paternally methylated region in the coding portion of the gene (Feil et al., 1994). We analysed the digestion of three HpaII restriction sites in the DMR2 (Fig. 1C). In all four (M. musculus × M. spretus)F1 ES cell lines, these HpaII sites were methylated on both parental chromosomes. In the (livers of) SF1-G derived fetuses, the HpaII digestion pattern was as in normal control (M. musculus × M. spretus)F1 fetuses, suggesting paternal DMR2 methylation (Table 1). In many of the fetuses derived from ES lines SF1-1, SF1-3 and SF1-8, in contrast, the HpaII digestion pattern was similar to that of ((M. musculus × M. spretus) × M. musculus)F2 fetuses, suggesting the presence of maternal methylation. Indeed, in these fetuses, a 2.1 kb partial digestion product was apparent, which indicated maternal DMR2 methylation (Fig. 1C). This allelic alteration did not affect the tissue-specificity of DMR2 methylation, which we have previously shown to be present in Igf2-expressing tissues, particularly in the liver (Feil et al., 1994). This is shown for SF1-1 derived fetuses 170-172 (Fig. 1C), in which high levels of methylation were detected in liver, whereas the remainder of the fetus displayed much reduced methylation levels. We next analysed Igf2 expression by RT-PCR amplification (Fig. 2C). In all SF1-G fetuses, expression was exclusively paternal (Table 1). In contrast, the ES fetuses with maternal DMR2 methylation, derived from the three other ES cell lines, showed predominantly maternal Igf2 expression. In one ES fetus (fetus 64) equal expression was detected by RT-PCR from both parental chromosomes, and in this fetus, the DMR2 methylation analysis suggested biallelic methylation. Although methylation patterns in ES cells (biallelic in all lines) did not persist on differentiation into ES fetuses, the allele-specific expression patterns did. Therefore, allelic Igf2 expression did not correlate with DMR2 methylation in ES cells, but did so in ES fetuses.

The maternally expressed Igf2r gene (Barlow et al., 1991) contains an intronic region (‘region 2’) that is maternally methylated in embryonic and adult tissues (Stöger et al., 1993). We found HpaII restriction sites in region 2 to be maternally methylated in all the ES cell lines, and this was retained by most of the ES fetuses (Fig. 1D). Igf2r expression was analysed by RT-PCR (Fig. 2D) according to Villar et al. (1995). In all four ES cell lines, expression was biallelic (albeit more strongly from the maternal allele). In the majority of the ES fetuses only maternal Igf2r expression was observed (e.g. fetuses 61-65). Some of the ES fetuses derived from lines SF1-1 and SF1-8 had predominantly maternal expression, but also showed some paternal expression (e.g. fetus 22). In these animals region 2 showed some DNA methylation on the paternal chromosome. Biallelic Igf2r expression was observed only in one fetus (171) where region 2 was methylated on both parental chromosomes. Therefore, expression of Igf2r did correlate with methylation in the ES fetuses, but not the ES cells.

We did not observe changes in the overall levels of genomic methylation, measured by analysing methylation in the Line 1 repeat element (data not shown). In the imprinted genes analysed, quantitative changes were observed in U2af1-rs1 only (Fig. 1A), but it was unclear whether these had arisen on derivation or during culture of the ES cells. To address whether culturing on its own can give rise to quantitative methylation changes, we subjected six newly derived parthenogenetic ES cells to prolonged culture, and determined levels of U2af1-rs1 methylation (Fig. 3). After derivation (passage 2-3) these lines displayed the expected (Shibata et al., 1997) complete (maternal) methylation. On subsequent culture, however, maternal methylation was almost entirely lost in four of the lines (PR5, PR9, PR13, PR14). In the two other lines (PR4 and PR10), quantitative changes were also observed, but these displayed respective decreases and increases in methylation during the culturing up to passage 12.

Northern analysis did not reveal major differences in Igf2 expression between fetuses derived from different cell lines (Fig. 4). In contrast, H19 expression was found to be virtually absent in many of the SF1-1, SF1-3 and SF1-8 derived ES fetuses. This was unexpected since, in most ES fetuses, normal maternal H19 expression had been detected in the qualitative RT-PCR analysis (Fig. 2B). A systematic comparison revealed that in all ES fetuses with predominantly maternal Igf2 expression, H19 was strongly repressed (Table 1). For example, SF1-3 derived ES fetuses 61-65 had comparable levels of Igf2 expression (Fig. 4), whether expression was predominantly (61-63), biallelic (64) or paternal (65) (Fig. 2C). No H19 expression was detected in the fetuses with maternal Igf2 expression (fetuses 61-63), whereas the other two fetuses (64 and 65) displayed high levels of H19 expression (Fig. 4). We were not able, by northern analysis, to quantify U2af1-rs1 expression in the fetuses, and the poorer quality of some of the RNAs did not allow comparative expression analysis of the Igf2r transcript.

Since the fetuses which showed no expression of H19 had mostly unaltered paternal methylation in the gene and its promoter (Fig. 1B, Table 1), we were interested in determining whether alterations had occurred further upstream of the gene, in a region which carries a paternal methylation mark that is already detected in the gametes (Tremblay et al., 1995). HhaI restriction sites in this upstream region were methylated on both parental chromosomes in all the fetuses which did not express H19 (Fig. 5). For example, SF1-3 derived fetuses 61-63, which had no H19 expression and maternal Igf2 expression, showed biallelic methylation of this upstream region. SF1-3 derived fetus 64, in which H19 showed high levels of expression (and Igf2 was expressed biallelically), had full paternal and partial maternal methylation in the H19 further upstream region. Fetus 65, the only SF1-3 derived fetus with exclusive paternal Igf2 expression and high levels of H19 expression, had exclusively paternal methylation in the H19 upstream region. Importantly, the aberrant methylation in fetuses without H19 expression was already present in the ES cells, with ES lines SF1-1, SF1-3 and SF1-8 displaying almost complete methylation on both parental chromosomes. Unaltered (Tremblay et al., 1995) exclusively paternal methylation was detected in ES cell line SF1-G and all derived fetuses, and this correlated with unaltered qualitative and quantitative expression of Igf2 and H19 (Fig. 4).

We next sought to address which proportion of genes might be expressed at abnormal levels during early fetal development in the ES fetuses. Protein expression was analysed by high-resolution 2-dimensional electrophoresis (2-DE) at day 11 of development, at which developmental abnormalities were not (yet) apparent (Fig. 6A). Of more than 3000 proteins analysed, only seven (0.2%) quantitative differences were apparent, but none of these were consistently altered in ES fetuses derived from the same ES line (data not shown).

The previous analysis demonstrates that the imprinted genes examined showed alterations of allelic methylation and allelic expression in ES cells, and that these alterations were maintained to fetal stages. A phenotypic characterisation of the ES fetuses was therefore carried out to see whether aberrant imprinting status was associated with altered phenotypes. ES fetuses derived from lines SF1-3 and SF1-8 had external developmental anomalies, the most striking of which was polyhydramnios, observed in all SF1-3 and SF1-8 derived fetuses (Fig. 6C). SF1-3 and SF1-8 derived fetuses also showed poor development of the mandible and in about 80% of these fetuses, interstitial bleeding was apparent (Fig. 6D,E). Lines SF1-1 and SF1-G yielded fetuses in which external morphology and the extraembryonic tissues appeared to be normal (Fig. 6B), although on closer examination we found that about 20% of these did have interstitial bleeding, albeit less severely so than the SF1-3 and SF1-8 derived fetuses.

Given the low recovery (approx. 23%) of live ES fetuses at days 13-14 of development, and therefore the limited and variable number of ES fetuses in each recipient mouse, the ES fetus system did not allow simple comparison of ES lines for their fetal growth potential. To address this important question we derived chimeric fetuses by injection of ES cells into diploid host blastocysts. We compared day-14 chimeras derived from line SF1-1, which had produced externally normal ES fetuses, and ES line SF1-8, which produced fetuses with developmental abnormalities. Chimeras were obtained at frequencies of 90% and 80%, for lines SF1-1 and SF1-8, respectively. For line SF1-1, there was no significant correlation between the percentage of chimerism (measured by Southern blot analysis using RFLPs) and fetal mass or crown-rump (C/R) length. In contrast, the mass of day-14 SF1-8 chimeras (and C/R length) correlated significantly with the percentage of chimerism, and on average these fetuses were about 30% heavier than SF1-1 chimeras (Fig. 7). Interestingly, 18 of the 24 SF1-8 chimeras appeared morphologically normal. This contrasted with the aberrant developmental phenotype of the ES fetuses derived from this stem cell line, and demonstrates rescue of phenotypic abnormalities by the host embryo.

Using an interspecific ES cell system we considered whether, on derivation and culture of ES cells, epigenetic alterations accumulate in imprinted genes and whether such epimutations are associated with altered gene expression in ES fetuses. Indeed, aberrant methylation patterns were detected in all imprinted genes analysed, and these either corresponded to primary epigenetic alterations or reflected such alterations. Hence, methylation changes in U2af1-rs1 and H19 were already present in ES cells and persisted on differentiation into ES fetuses. We assume the same is true of Igf2r, but because fetuses with aberrant Igf2r imprinting were less frequent, methylation changes in ES cells would have gone undetected in bulk culture. Alterations in Igf2 became apparent only in fetuses. Interestingly, the methylation changes we observed were of the three possible types: loss of methylation (U2af1-rs1), gain of methylation (H19 upstream region; region 2 of Igf2r) and allelic change in methylation (DMR2 in Igf2). Our observation that normal methylation patterns do not become re-established on differentiation into ES fetuses extends earlier findings on ES cells homozygous for a disruption of the DNA methyltransferase gene (Li et al., 1993). Expression of a methyltransferase cDNA in these targeted ES cells restored normal levels of methylation in non-imprinted genes, but not in imprinted genes (Tucker et al., 1996). The stem cells in our study had been cultured for a number of passages before they were used for the production of ES fetuses and it was therefore unclear whether the methylation changes had occurred during the derivation, or the subsequent culturing of the cells. For U2af1-rs1, we could address this question further by analysing parthenogenetic ES cell lines, which all had full maternal methylation after derivation. Methylation in U2af1-rs1 was highly unstable and frequently lost on culturing of these cells. This observation extends a study by Szabó and Mann (1994), who described quantitative methylation changes in Igf2 and H19 in monoparental ES cells. How specific these methylation changes are to imprinted genes is not known. However, CpG islands in non-imprinted genes do not become methylated in ES cells (Antequera et al., 1990; Frank et al., 1991), suggesting that methylation patterns in imprinted genes might be particularly unstable in ES cells.

Altered methylation patterns were associated with qualitative and, in the case of H19, quantitative changes in gene expression. In the U2af1-rs1 gene, loss of maternal methylation correlated with biallelic expression. In ES cells and ES-derived fetuses from line SF1-1, U2af1-rs1 was maternally methylated in a proportion of the cells, which correlated inversely with levels of maternal U2af1-rs1 expression. Allelic expression of Igf2r was mostly unaltered in the ES cells and altered in few of the derived fetuses, and this stability of the region 2 imprint has been reported for 129/Sv ES lines as well (Labosky et al., 1994). Although the maternal region 2 methylation is present throughout preimplantation development (Stöger et al., 1993), it has been shown that repression of the Igf2r allele occurs only after implantation (Lerchner and Barlow, 1997) and paternal anti-sense transcription originating in region 2 has been implicated in this repression (Wutz et al., 1997). Our finding of maternal methylation and biallelic Igf2r expression in all four ES cell lines agrees with these studies.

In most of the ES fetuses produced from lines SF1-1, SF1-3 and SF1-8, the normally paternally methylated and expressed Igf2 gene showed maternal DMR2 methylation, which correlated with unequal biallelic expression strongly biased in favour of the maternal allele. Unexpectedly, fetuses with strong activation of the maternal Igf2 allele showed no expression of H19. This biallelic repression of H19 was not associated with methylation changes in its promoter, but with biallelic methylation of sequences further upstream of the gene. It has been shown that these sequences function as a silencer in Drosophila (Lyko et al., 1997). Our study suggests that in the mouse this imprinting element may be involved in long-range chromatin interactions. Based on our findings, we propose that, when on the maternal chromosome and unmethylated, the H19 upstream element allows downstream enhancers to interact with the H19 promoter via appropriate looping of chromatin, and thereby insulates the flanking Igf2 gene from interactions with such regulatory sequences. On the paternal chromosome, in contrast, this region is fully methylated, and its conformation might not allow higher order folding required for H19 expression, thus leading to Igf2 expression. Such a mechanistic, methylation-dependent, involvement of the H19 further upstream element agrees with the enhancer competition model proposed by Bartolomei et al. (1992), in that only one of the two genes can be expressed from a single chromosome. This model does not, however, explain our finding that fetuses with biallelic methylation in the H19 upstream element expressed Igf2 predominantly from the maternal chromosome. This strong activation of the maternal gene may have arisen because of additional alterations in methylation and/or chromatin, possibly within the Igf2 gene itself. It might be relevant that these fetuses had substantial levels of maternal methylation in the DMR2, a normally paternally methylated region that we proposed to be involved in silencing of transcription (Feil et al., 1994).

Our data show that epigenetic alterations arise on derivation and culturing of ES cell lines, do not become corrected during postimplantation development, and are associated with aberrant imprinted gene expression in the fetus. Therefore, this in vivo differentiation system might constitute an experimental model for deregulation of imprinting in developmental disorders and embryonal tumours. In sporadic Wilms’ tumours of the kidney, for example, healthy tissue adjacent to the tumour already shows biallelic IGF2 expression in association with biallelic repression and methylation of H19, suggesting that epigenetic alterations occurred during early development (Okamoto et al., 1997). In the human fetal overgrowth syndrome Beckwith Wiedemann Syndrome, germline or early embryonic errors may occur in the IGF2-H19 domain without DNA mutations (Reik et al., 1995), and also in Angelman Syndrome, imprinting defects have been detected without apparent genetic alteration (Burger et al., 1997). We have previously shown that developmental abnormalities resulting from micro-manipulation can be transmitted to the next generation (Römer et al., 1997). The question that arises is whether altered methylation patterns in imprinted genes are heritable as well. Although we did produce adult chimeric animals for three of the ES lines (not shown), germline transmission could not be achieved because of hybrid sterility, and this intriguing question should be addressed on a different genetic background.

From this and other studies on ES cells one might also anticipate that in in vitro cultured preimplantation embryos, imprinted genes can undergo epigenetic alterations. Indeed, Sasaki et al. (1995) found that culturing of preimplantation mouse embryos may lead to biallelic H19 expression in extraembryonic tissues. Earlier studies suggested that culture and manipulation of preimplantation mouse embryos may result in aberrant growth and developmental abnormalities (Bowman and McLaren, 1970; Reik et al., 1993). In cattle and sheep, culture and manipulation of preimplantation embryos can lead to enhanced fetal growth and developmental abnormalities such as increases in bone lengths, facial/head abnormalities and polyhydramnios (Walker et al., 1996; Kruip and den Daas, 1997), and high birthweights, perinatal death and reduced development of the lower jaw have been reported in sheep cloned by transfer of nuclei from cultured fibroblasts (Schnieke et al., 1997). In ES fetuses derived from two of the stem cell lines we observed possibly similar abnormalities, which included polyhydramnios, interstitial bleedings and reduced mandible development, and growth-enhancement was apparent in chimeras. Our observations make it likely that these defects are at least partly caused by aberrant expression of imprinted genes. Future research should aim to link specific components of the aberrant phenotypes with specific epigenetic alterations in gene expression.

We thank N. D. Allen for assistance at the onset of this study, G. Kelsey, D. P. Barlow, T. Forné and S. Khosla for reading the article, J. Walter for Igf2 sequence information, D. Brown for statistical analysis, and D. P. Barlow and A. Collick for genomic probes. This research was supported by the Ministry of Agriculture, Fisheries and Food, BBSRC and Action Research. R. F. is a Babraham Research Fellow.