Human Reproduction Update Advance Access originally published online on March 15, 2006
Human Reproduction Update 2006 12(3):233-242; doi:10.1093/humupd/dmk005
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Biparental hydatidiform moles: a maternal effect mutation affecting imprinting in the offspring
1 Department of Obstetrics and Gynecology, 2 Department of Molecular and Human Genetics, Baylor College of Medicine, Houston, TX, USA and 3 Department of Obstetrics and Gynecology, Assiut University, Assiut, Egypt
4 To whom correspondence should be addressed at: Department of Obstetrics and Gynecology, Baylor College of Medicine, 1709 Dryden, Suite 1100, Houston, TX 77030, USA. E-mail: iveyver{at}bcm.tmc.edu
Submitted on August 18, 2005; resubmitted on December 3, 2005; accepted on January 3, 2006
 |
Abstract |
|---|
TOP
Abstract
IntroductionHighly recurrent hydatidiform moles (HMs) studied to date are not androgenetic but have biparental genomic contribution (BiHM). Affected women have an autosomal recessive mutation that causes their pregnancies to develop into HM. Although there is genetic heterogeneity, a major locus maps to chromosome 19q13.42, but a mutated gene has not yet been identified. Molecular studies have shown that maternal imprinting marks are deregulated in the BiHM trophoblast. The mutations that cause this condition are, therefore, hypothesized to occur in genes that encode transacting factors required for the establishment of imprinting marks in the maternal germline or for their maintenance in the embryo. Although only DNA methylation marks at imprinted loci have been studied in the BiHM, the mutation may affect genes that are essential for other forms of chromatin remodelling at imprinted loci and necessary for correct maternal allele-specific DNA methylation and imprinted gene expression. Normal pregnancies interspersed with BiHM have been reported in some of the pedigrees, but affected women repeatedly attempting pregnancy should be counselled about the risk for invasive trophoblastic disease with each subsequent BiHM.
Key words: embryology / genetic disorders / germ cells / imprinting / pregnancy
|
Introduction |
|---|
A hydatidiform mole (HM) is an abnormal pregnancy with excessive proliferation of placental villi but severely stunted or absent embryonic development. HM can be classified into complete HM (CHM) or partial HM (PHM). Most CHM are androgenetic (AnCHM) with a diploid genome that is entirely paternally derived. PHM have a triploid genome, with three copies of each chromosome, two of which are paternally inherited (Szulman and Surti, 1978
). When two of the three copies of each chromosome are of maternal origin, triploid pregnancies have an underdeveloped placenta. Hence, an excess of paternally inherited genes results in molar pregnancy, which suggests that deregulated expression of imprinted genes plays an important role in this abnormal development of the human trophoblast. HM is considered a sporadic disorder with 1% recurrence risk, but several familial and sporadic cases of women with highly recurrent CHM (and sometimes PHM) have been described (Wu, 1973; Ambani et al., 1980; Federschneider et al., 1980; Nalini et al., 1989; Kircheisen and Schroeder-Kurth, 1991; Narayan et al., 1992; Sunde et al., 1993; Seoud et al., 1995; Helwani et al., 1999; Moglabey et al., 1999; Sensi et al., 2000; Al-Hussaini and Abdel-Alim, 2001; Ozalp et al., 2001; Fisher et al., 2002; Judson et al., 2002; Al-Hussaini et al., 2003; Fallahian, 2003; Hodges et al., 2003). Where studied, the molar pregnancies of these women are of normal diploid biparental inheritance (BiHM) and do not contain excess of paternally inherited genetic material (Helwani et al., 1999) but show abnormal expression (Fisher et al., 2002; Hodges et al., 2003) or epigenetic marking of imprinted genes (Judson et al., 2002; El-Maarri et al., 2003). The pedigrees are consistent with the affected women having an autosomal recessive mutation (Moglabey et al., 1999; Sensi et al., 2000; Judson et al., 2002; Hodges et al., 2003; Panichkul et al., 2005; Slim et al., 2005) that appears to disturb normal imprinting in their pregnancies. An active search for the mutated gene, hypothesized to encode an essential transacting factor for the regulation of imprinting in oocytes and/or the preimplantation embryo, is ongoing in several laboratories. Here, we review current research progress on BiHM and speculate on possible functions of the candidate gene, based on recently accumulated knowledge about epigenetic processes in germ cells, fertilized zygotes and preimplantation embryos. |
Aetiology and genetics of sporadic CHMs and PHMs |
|---|
The incidence of HM is variable, complicating from 1/1000 to 1/2000 pregnancies in the United States (Benirschke and Kaufmann, 1990
; Szulman, 1999) and to 1/125 in some high-incidence areas in Southeast Asia. In the Middle East, where many familial cases of highly recurrent BiHM originate, the incidence is about 1/500 (Al-Hussaini and Abdel-Alim, 2001). CHM is characterized by embryo degeneration probably between 15 and 31 days post conception (dpc) (Szulman, 1999) and varying degrees of abnormal proliferation of both the cytotrophoblastic and the syncytiotrophoblastic layers. The initially oedematous and hypercellular mesenchymal villar stroma develops central cisternae that coalesce into larger cystic cavities compressing the stroma and central blood vessels. The result is a mass of clear vesicles that vary in size from microscopic to a few centimetres and no discernible fetus or amnion. One should differentiate between true CHM and hydropic degeneration, which is not considered gestational trophoblastic disease (Berkowitz et al., 1991; Szulman, 1999). The term PHM is used when the molar changes are focal and less advanced, and fetal development is preserved. PHM have slowly progressive avascular chorionic villi, whereas vascular villi with a functioning fetal–placental circulation are spared (Shapter and McLellan, 2001).
Haplotype analysis with polymorphic markers has determined that most CHM are AnCHM (Kajii and Ohama, 1977; Wake et al., 1978). The majority have a 46,XX karyotype, resulting from monospermic fertilization of an oocyte with ‘inactivated genome’ (also referred to as ‘empty oocyte’) with duplication of the male haploid genome contributed by the fertilizing sperm (Figure 1a). Although a 46,YY conceptus could also be formed from these events, this karyotype has never been observed in CHM, suggesting that absence of the X chromosome is not compatible with early development. About 20% of AnCHM result from dispermic fertilization of an oocyte with an inactivated genome and can have a 46,XY or 46,XX karyotype (Szulman and Surti, 1978; Kovacs et al., 1991) (Figure 1b). Most PHM are diandric triploid (69,XXY, 69,XXX or 69,XYY) and are formed by dispermic fertilization of a haploid oocyte, resulting in a conceptus with two paternal and one maternal haploid set of chromosomes (Szulman and Surti, 1978; Daniel et al., 2001) (Figure 1c). Although PHM are associated with longer development of a viable fetus, triploidy is a highly lethal chromosomal abnormality and most affected embryos die within a few weeks of conception. Of those surviving longer, more than 70% have severe symmetrical intrauterine growth restriction, and 93% have structural anomalies including three to four syndactyly of hands, hydrocephalus, heart anomalies and micrognathia (Jauniaux, 1998). Digynic triploidy with two maternal haploid sets of chromosomes results in a small underdeveloped placenta (McFadden and Kalousek, 1991). Interestingly, ovarian teratomas, which are tumours that originate from parthenogenetically activated female germ cells, contain a wide variety of tissues but never trophoblast (Varmuza and Mann, 1994). These data clearly support an imprinting effect wherein the parental genomes contribute unequally to the development of the placenta: paternally expressed genes promote trophoblast proliferation and differentiation, and maternally expressed genes are essential to counterbalance this effect (Falls et al., 1999).
|
|
BiHMs |
|---|
Kindreds wherein several women have recurrent HM have now been reported (Ambani et al., 1980
; Kircheisen and Schroeder-Kurth, 1991; Sunde et al., 1993; Seoud et al., 1995; Helwani et al., 1999; Moglabey et al., 1999; Sensi et al., 2000; Fisher et al., 2002; Judson et al., 2002; Al-Hussaini et al., 2003; Fallahian, 2003; Hodges et al., 2003; Panichkul et al., 2005). Most molar gestations of these affected women are CHM, but some PHM have been described. In three of the reported pedigrees (Ambani et al., 1980; Sunde et al., 1993; Seoud et al., 1995; Fallahian, 2003; Fisher et al., 2004), the women also had liveborn offspring, but in the other families and some sporadic cases of recurrent HM, they only have large numbers of consecutive HM pregnancies (Wu, 1973; Ozalp et al., 2001; Judson et al., 2002; Al-Hussaini et al., 2003; Fisher et al., 2004; Panichkul et al., 2005). Using short tandem-repeat markers spread throughout the genome, Helwani et al. (1999) studied the parental genomic contribution of molar pregnancy tissues from affected women in a large consanguineous pedigree described by Seoud et al. (1995). They established that the pregnancies were diploid with one paternal and one maternal allele at each of the informative loci (Figure 1d). Using different methods of analysis, Sunde et al. (1993) had already found support for maternal and paternal genomic contribution in the recurrent HM of another member of this pedigree (Helwani et al., 1999). Hence, although phenotypically these pregnancies were like AnCHM (Helwani et al., 1999) or diandric PHM (744; Sunde et al., 1993), they were biparentally inherited (BiHM). This observation led the authors to suggest that the HM phenotype may be caused by the deregulation of one or more imprinted genes in those pregnancies (Sunde et al., 1993; Helwani et al., 1999). |
Evidence for an autosomal recessive mutation in women with BiHM |
|---|
The familial cases of BiHM are extremely rare, but the inheritance pattern in several reported pedigrees support that BiHM is an autosomal recessive Mendelian disorder (Ambani et al., 1980
; Kircheisen and Schroeder-Kurth, 1991; Seoud et al., 1995; Sensi et al., 2000; Al-Hussaini and Abdel-Alim, 2001; Ozalp et al., 2001; Fisher et al., 2002; Judson et al., 2002; Al-Hussaini et al., 2003; Fallahian, 2003; Hodges et al., 2003; Slim et al., 2005). The family structure of these kindreds often shows consanguinity in the parents of the affected women as well as between these women and their partners. If the BiHM would have resulted from an autosomal recessive mutation in the conceptus, transmitted by heterozygote carrier parents, only 25% of pregnancies would be expected to be affected. In the majority of described pedigrees (8/11), the pregnancies of affected women are only confirmed HMs (Al-Hussaini et al., 2003), spontaneous miscarriages that are not all pathologically verified or occasionally a stillbirth. In others (3/11), there are some normal pregnancies interspersed with the BiHM (Ambani et al., 1980; Sunde et al., 1993; Seoud et al., 1995; Fallahian, 2003; reviewed in Fisher et al., 2004). Furthermore, in some families, a few of the affected women have recurrent molar pregnancies with different partners, who themselves were fathers of healthy offspring from previous or subsequent relationships (Fisher et al., 2000; Al-Hussaini et al., 2003). These combined observations indicate that the women themselves are affected with the autosomal recessive mutation that causes the recurrent BiHM and that the paternal genotype does not contribute to the pathogenesis. However, it is currently not well understood why in some pedigrees, not all pregnancies of these women are BiHM (Slim et al., 2005).
The first linkage study to map the gene mutated in women with BiHM was performed in a large Middle-Eastern pedigree with extensive consanguinity (Seoud et al., 1995) as well as a second northern European pedigree with affected sisters (Kircheisen and Schroeder-Kurth, 1991). These genetic linkage studies revealed that the affected women were homozygous by descent for a 15.2 cM region on chromosome 19q13.3-19q13.4, flanked by markers D19S924 and D19S890 (Moglabey et al., 1999). This interval was first narrowed by 2.8 cM at the telomeric end in an unrelated family (Sensi et al., 2000) and confirmed in a fourth familial case (Fisher et al., 2002). A more recent study refined the candidate region to a 1.1 Mb interval in 19q13.42 between centromeric marker D19S418 and a new marker at the telomeric end (AAAT11138) (Hodges et al., 2003). Our own data on four newly described Middle-Eastern pedigrees are consistent with this region being a major candidate locus (Panichkul et al., 2005). In other pedigrees, linkage to chromosome 19q13.42 could not be established, supporting that there is genetic heterogeneity for this condition. Additional candidate loci for these other families have not been reported (Judson et al., 2002; Slim et al., 2005).
The original 15.2 cM critical interval encompassed the imprinted PEG3/ZIM2 locus (paternally expressed gene 3/zinc-finger gene 2 from imprinted domain) (Kim et al., 2000; Van den Veyver et al., 2001). PEG3 is the human homologue of the imprinted mouse Peg3 gene, is highly expressed in placenta and is silenced from the maternal allele in both human and mouse (Kuroiwa et al., 1996; Hiby et al., 2001). Along with X-linked loci, Peg3 has been implicated in the placental hypertrophy found in the inter-species hybrid crosses between the monogamous Peromyscus polionotus (PO) and the polyandrous Peromyscus maniculatus (BW) mice (Vrana et al., 1998, 2000). The subsequent fine mapping has now entirely excluded this imprinted locus from the candidate region (Hodges et al., 2003; Panichkul et al., 2005). However, altered imprinting of PEG3 may still contribute to the placental phenotype of the BiHM via a mutation in a transacting factor that deregulates maternal epigenetic marking (Judson et al., 2002; El-Maarri et al., 2003) of this and other imprinted genes (see below).
|
BiHM have a global defect in reprogramming or maintenance of maternal imprinting marks |
|---|
Most imprinted genes lie in clusters of interspersed maternally and paternally expressed genes, which often contain cytosine–guanine (CpG) islands, groups of CpG dinucleotides. Cytosine residues within CpG dinucleotides can become methylated at position 5 of the pyrimidine ring. CpG islands located in imprinted gene clusters are marked by differentially methylated regions (DMRs) on either the maternally or the paternally inherited chromosome, and the disruption of DNA methylation at such DMRs has been used to diagnose patients with defects in genomic imprinting (Kubota et al., 1996
). CpG-island methylation is typically associated with silencing in cis of the nearby gene, but some DMRs act as long-range imprinting control centres and may show more complex relationships between methylation and gene expression. In regions with more than one regulatory DMR, the primary imprints which are established in the developing gametes may influence the CpG methylation status of secondary imprints. Secondary imprints can occur from the post-fertilization to the blastocyst stage (Weinstein et al., 2004).
Because genomic imprinting was implicated in the pathogenesis of BiHM, two groups have evaluated CpG methylation of DMRs at several imprinted loci in tissues from a total of five BiHM pregnancies (Judson et al., 2002; El-Maarri et al., 2003). Judson et al. (2002) studied tissues derived from a patient with familial recurrent BiHM that did not map to the 19q13.42 region. In the second study, two tissues were obtained from the original pedigree that demonstrated linkage to 19q13.42, whereas two were derived from sporadic BiHM (El-Maarri et al., 2003).
Abnormal CpG methylation at imprinted genes after passage through the maternal germline was found by both, but subtle differences were seen at specific loci. Judson et al. (2002) found loss of cytosine methylation in BiHM of the DMRs at KCNQ1OT1, PEG1, PEG3/ZIM2 and 5' SNRPN that are normally methylated on the maternal allele and unmethylated on the paternal allele. Similarly, El-Maarri et al. (2001) demonstrated a loss of methylation at the PEG3 and 5' SNRPN DMRs (Figure 2a). In the mouse, all of these sites are primary imprints that normally become methylated in the oocyte. In humans, it is not clear whether 5' SNRPN is a primary DNA methylation imprint or is established as a secondary imprint in the embryo (El-Maarri et al., 2001). Single-nucleotide polymorphism analysis of the 5' SNRPN DMR on DNA from parents and from the BiHM revealed that CpG methylation was indeed lost from the maternal allele (El-Maarri et al., 2003).
|
CpG methylation at a subset of the DMRs at the GNAS1 imprinted locus was also examined (Figure 2b). GNAS1 encodes the heterotrimeric G protein
-subunit, GS, the overlapping longer XLs protein, and the unrelated chromogranin-like protein NESP55. GNAS1 has a complex structure and imprinting pattern (for review, see Beaudet, 2004; Weinstein et al., 2004). Recent data suggest that there are two imprinting control regions at GNAS1 (Bastepe et al., 2005; Liu et al., 2005a), and we will discuss the reported findings at the analysed GNAS1 DMRs in BiHM in this context. GS is biparentally expressed in most tissues but paternally silenced in some (Weinstein et al., 2004). The DNA at the GS promoter at exon 1 is unmethylated on both alleles, but in mice there is differential histone H3 lysine 4 methylation (H3K4me) at this exon (Sakamoto et al., 2004). The primary imprinting mark for tissue-specific paternal imprinting of GS is the 2.5 kb upstream exon 1A. CpG methylation of its maternal allele is established in oogenesis. The unmethylated paternal allele expresses a non-coding RNA (1A or A/B transcript) that may participate in the regulation of imprinting of GS but not of the other GNAS1 transcripts (Williamson et al., 2004; Liu et al., 2005a). The latter depends on the other primary maternal methylation imprinting mark, 47 kb upstream, at the NESPAS/XLs promoter which controls the expression of the XLs mRNA and NESPAS, an antisense RNA to NESP55 (Coombes et al., 2003). The first exon of XLs (35 kb upstream of GS exon 1) is also methylated on the maternal allele, but this is established later in development than at the NESPAS/XLs promoter, and some randomness in methylation is seen in mouse blastocysts (Liu et al., 2000). The XLs mRNA and the NESPAS antisense RNA are transcribed from the unmethylated paternal allele (Weinstein et al., 2004). There is evidence from the analysis of mouse models as well as DNA from patients with primary pseudohypoparathyroidism type IB that there can be dissociation between the imprinting at the first exon of XLs and the NESPAS/XLs promoter (Liu et al., 2005b). The NESP55 DMR is methylated on the paternal chromosome and unmethylated and transcribed from the maternal chromosome. This is a secondary imprint, established in post-implantation development that appears co-regulated with and dependent on the primary imprinting mark at NESPAS/XLs (Weinstein et al., 2004).
Judson et al. found that exon 1A is unmethylated in BiHM, which is as expected if this is a primary maternal imprint that fails to be established. A few of the control chorionic villus samples (2/10) and 1/3 AnCHM showed some aberrant methylation patterns. This is currently difficult to explain but does not detract from the significance of the finding that there is a failure to establish a primary maternal methylation mark at this exon in BiHM. The NESPAS/XLs promoter also behaved as expected, with loss of the primary maternal methylation mark, whereas the data at the first exon of XLs showed an unexpected variable mixed pattern of methylation. This is not completely understood but could be related to the fact that this is a later developing and not primary methylation mark (see above). The NESP55 DMR was analysed in both studies and showed complete methylation on both alleles, i.e. a paternal epigenotype. Polymorphism analysis confirmed that the gain of methylation was on the maternally inherited allele (El-Maarri et al., 2003). Judson et al. concluded that this gain occurred, because maternal lack of methylation at the NESP55 DMR (as well as presence of methylation at the NESPAS/XLs DMR) is a secondary mark dependent on the primary maternal methylation mark at exon 1A, while El-Maarri et al., who also found some maternal methylation gain at the H19 DMR (discussed below), concluded that there is a primary gain of methylation. Even though more recent data now show that DMR methylation at NESPAS/XLs, at least in mice, appears to be a primary mark with which NESP55 is co-regulated and independent of exon 1A methylation (see above), both interpretations are still plausible.
Both studies also examined the H19 DMR, which carries a primary CpG methylation imprint on the paternal allele (Figure 2c). In the familial BiHM that was not linked to 19q13.42, CpG methylation at H19 was completely normal (Judson et al., 2002), as expected if the underlying defect would only cause failure to establish primary maternal methylation imprints. However, in contrast to the Judson study, El Maarri et al. found that the H19 DMR demonstrated variable CpG methylation (El-Maarri et al., 2003). Two of four BiHM samples aberrantly acquired methylation on the maternal allele: one showed a mixed variable pattern and a sporadic BiHM had normal methylation. These results suggested to the authors that there is not only a loss of methylation on primary maternally methylated alleles but also a gain of primary paternal-specific methylation on maternal alleles in the samples they analysed.
The DNA methylation defect in BiHM appears to be specific to imprinted genes and not the result of a global disruption of DNA methylation, because there is normal methylation at non-imprinted CpG sites (Judson et al., 2002) and at an X-linked gene that is subject to X chromosome inactivation (El-Maarri et al., 2003). Expression analysis by immunohistochemistry on BiHM molar trophoblast tissues for p57KIP2, which is normally expressed from the maternally inherited copy and silenced from the paternally inherited copy, showed markedly decreased expression (Fisher et al., 2002; Hodges et al., 2003). This is also expected, because imprinted expression of this gene depends on the maternal methylation of the KCNQ1OT1 DMR. Finally, the DNA methylation status of investigated DMRs at imprinted genes is normal in the mothers of the affected pregnancies (El-Maarri et al., 2005), further supporting the observation that the imprinting defects arise de novo during female gametogenesis.
What might cause the different results at the H19 DMR between these studies? One explanation might be genetic heterogeneity: the investigated tissues were from BiHM that were not linked to 19q13.42 (Judson et al., 2002), linked to this region (El-Maarri et al., 2003), or from sporadic BiHM cases of unknown genetic linkage (El-Maarri et al., 2003). Thus, it is possible that different basic components of the machinery that establishes or maintains imprinting marks may be disrupted in different cases: one resulting in loss of DNA methylation at DMRs of primary maternal methylation imprints, whereas another also causes gain of DNA methylation on the maternal allele of DMRs that have a primary methylation imprint on the paternal allele. There may also be influence from genetic modifiers or environmental factors on the phenotypic expression of the underlying defect (Slim et al., 2005). Other potential factors explaining the different results could be the use of short-term culture of the studied trophoblast (Judson et al., 2002) or unreported pathological confirmation of the diagnosis in sporadic cases. This information was not provided by El-Maarri et al. (2003), and one of the BiHMs in their study appeared to have normal methylation marks at several of the analysed loci. Hence, although those differences in the results of methylation analysis at imprinted loci between the two studies should ideally be confirmed on additional independent samples, they may have implications for hypotheses on the nature and timing of the molecular defects resulting in BiHM.
|
Chromatin dynamics at imprinted genes |
|---|
It has become clear in recent years that allele-specific cytosine methylation at CpGs of DMRs is only one aspect of the multiple chromatin modifications associated with imprinted gene expression. The methylated DMRs of imprinted loci have other hallmarks of repressive chromatin, such as post-translational modifications of core histone proteins that make up the nucleosomes around which the DNA double helix is wound. Methylated DMRs also show hypoacetylation of histones H3 and H4, dimethylation of H3 at lysine 9 (H3K9me2) and, in at least one case, trimethylation of H3 at lysine 27 (H3K27me3). The converse has been shown for the unmethylated alleles of DMRs, which contain hyperacetylated H3 and H4 and H3K4me, all of which are hallmarks of transcriptionally permissive DNA (Fournier et al., 2002
; Perk et al., 2002; Soejima and Wagstaff, 2005). Experiments in plants (Arabidopsis) and yeast (Neurospora crassa) indicate that in these organisms, the repressive histone modification H3K9me2 may induce cytosine methylation. The absence of histone modifications in mutant mouse models with inactivated DNA methyltransferase enzymes and the reduced DNA methylation in mice with inactivation of the major H3K9 methyltransferases, Suv39h1 and Suv39h2, suggests that this relationship is conserved in mammals (Soejima and Wagstaff, 2005). Furthermore, H3K9 methylation has been shown to be necessary for CpG methylation and allele-specific expression at Snrpn (Xin et al., 2003). The observation that for some genes, imprinted expression in placenta occurs without DNA methylation, but in the presence of H3K9 methylation, has led to the suggestion that H3K9 methylation may be the initial chromatin modification, sufficient to maintain imprinting in the short lifespan of the placenta, but requiring stabilization by DNA methylation in the longer-lived embryo proper (Lewis et al., 2004; Morgan et al., 2005). This is interesting and implies that in the BiHM pregnancies, deregulated DNA methylation at imprinted loci could be secondary to a defect of another type of chromatin modification essential for imprinted gene expression. Unfortunately, histone modifications have not yet been studied in BiHM. Such experiments may be difficult to accomplish considering the scarceness and archival nature of most samples. |
Molecular mechanisms and timing of reprogramming of imprinted genes |
|---|
Although imprinting at specific loci is not universally conserved between humans and mice [e.g. Igf2r and Mash2/Ascl2 are imprinted in mouse, but not in humans (Riesewijk et al., 1996
; Miyamoto et al., 2002)], the basic molecular mechanisms and relative order of events are most likely conserved between the two species. Most of our current knowledge about the critical steps required to correctly establish and maintain imprinted gene expression derives from analysis of various mouse models but is pertinent to this review. There are four critical steps in this process that overlap with dramatic genome-wide DNA methylation and epigenetic modifications of chromatin-associated proteins (Figure 3): (i) erasure of previous imprints in the developing germ cells, (ii) re-establishment of parent-specific imprints either in the germ cells or in the early embryo, (iii) postzygotic maintenance of imprints in the preimplantation embryo and (iv) maintenance of imprints after lineage commitment of the embryonic and extra-embryonic tissues.
|
In the mouse, primordial germ cells (PGCs) are set aside from the epiblast at 7.2 dpc (Hajkova et al., 2002), proliferate and begin to migrate into the genital ridge by 8.5 dpc, where they start arriving at about 10.5–11.5 dpc. This timing corresponds to the 5- to 11-week human embryo (Onyango et al., 2002). During PGC migration from the genital ridge, complete demethylation of all CpG dinucleotides, including those at imprinted DMRs, occurs in both male and female germ cells (Labosky et al., 1994; Tada et al., 1997, 1998; Kato et al., 1999; Hajkova et al., 2002; Lee et al., 2002; Onyango et al., 2002; Sato et al., 2003; Yamazaki et al., 2003). This process appears to be encoded within the gamete itself and is completed by 13.5 dpc; the speed by which it occurs suggests an as yet unidentified active demethylase enzyme (Hajkova et al., 2002; Morgan et al., 2005). The number of germ cells increases to about 25 000 at 13.5 dpc, the time when gender-specific differences become apparent in the primitive gonads and female PGCs are arrested in meiotic prophase I, whereas male PGCs undergo mitotic arrest in G0 (Labosky et al., 1994; Hajkova et al., 2002). Subsequent re-establishment of the epigenetic marks associated with parent-of-origin–specific imprinting initiates once the germ cells start differentiating (Ferguson-Smith and Surani, 2001) and occurs earlier in the male gamete than in the female gamete with variable timing for specific imprinted genes in gametes of both sexes (El-Maarri et al., 2001; Obata and Kono, 2002; Li et al., 2004; Lucifero et al., 2004). In male gametes, it occurs at the prospermatogonia stage before the start of meiosis when the cells are in G0 arrest (Davis et al., 1999, 2000; Sato et al., 2003) and is usually complete by 18.5 dpc but can continue until the mature sperm stage. In contrast, female gametes initiate remethylation later (after birth in the mouse) in the growing oocyte, and this may continue after fertilization in the female pronucleus of the zygote (Obata et al., 1998; Davis et al., 2000; El-Maarri et al., 2001; Ferguson-Smith and Surani, 2001; Lucifero et al., 2004).
A few essential proteins are known to be necessary for the correct reprogramming of DNA methylation imprints. Conditional inactivation in germ cells of the de novo DNA methyltransferases, Dnmt3a and Dnmt3b, has revealed that Dnmt3a is important for both maternal and paternal imprinting (Kaneda et al., 2004) while Dnmt3b may primarily regulate DNA methylation of retrotransposons and pericentromeric heterochromatic regions. Mouse DNA methyltransferase 3-like protein (Dnmt3l), a homologue of Dnmt3a and Dnmt3b that lacks the catalytic domain, is required for the establishment of imprinting in female, but not male gametes. In the male, Dnmt3l is necessary for the methylation of retrotransposons and pericentromeric heterochromatic regions in prospermatogonia, and its absence leads to meiotic arrest and sterility (Bourc’his et al., 2001; Hata et al., 2002; Bourc’his and Bestor, 2004; Webster et al., 2005). Dnmt3l directly interacts via its C-terminal domain with the catalytic regions of various isoforms of Dnmt3a and Dnmt3b. In cell-culture experiments, interactions enhance the efficiency of Dnmt3a and Dnmt3b to methylate DNA in vitro and in vivo and may participate in targeting these proteins to the DMRs of imprinted genes slated for de novo methylation (Chedin et al., 2002; Suetake et al., 2004; Chen et al., 2005). Only the Dnmt3a2 and the Dnmt3b1 isoforms appear co-expressed with Dnmt3l during the critical stages of germ cell development, but the relevance of this co-expression for asymmetric DNA methylation of imprinted DMRs is unknown.
In the oocyte, not all imprinted genes acquire their methylation marks at the same time. For example, the Peg3 DMR acquires DNA methylation earlier than the Peg1 DMR (Obata and Kono, 2002; Lucifero et al., 2004). Within the developing oocyte, there is also asynchrony between the parental alleles: gametes with the maternally inherited Snrpn allele acquire DNA methylation faster than those with the paternally inherited Snrpn allele (Lucifero et al., 2004). This observation supports the existence of other, not yet characterized, epigenetic modifications at imprinted loci that endow them with an epigenetic parent-of-origin–specific memory despite the total erasure of DNA methylation in PGC.
During the first cleavage divisions, the genomes are once more subject to massive changes in chromatin organization and DNA methylation. Immediately after fertilization, the paternal pronucleus undergoes reprogramming whereby the DNA-bound protamines are exchanged for highly acetylated histones, which immediately acquire H3K4, H3K9 and H3K27 monomethylation. During the transition from paternal protamines to maternal histones, the paternal genome rapidly undergoes active demethylation of CpG dinucleotides within the first 6 h after fertilization, except at methylated DMRs of imprinted genes, which appear protected and remain methylated. The maternal genome also becomes demethylated, but much slower and over several cell divisions, in a manner dependent on DNA replication. This suggests a passive process, caused by the lack of maintenance methylation. Recent data also indicate that different histone modifications in paternal and maternal genomes (Kourmouli et al., 2004; Santos et al., 2005) might explain their different timing of global demethylation in the preimplantation embryo. Maternally methylated DMRs at imprinted genes are also protected during global demethylation of the maternal genome (Morgan et al., 2005). Mice with a mutation that selectively removes Dnmt1o, encoding an oocyte-specific isoform of the primary maintenance methyltransferase Dnmt1 that is only in the nucleus at the 8-cell stage of embryonic development (Mertineit et al., 1998), show loss of methylation of maternally methylated DMRs with consequent bi-allelic expression of maternally imprinted genes. This supports that Dnmt1o is essential for protecting the maternally methylated DMRs during global demethylation of the genome in the preimplantation embryo (Howell et al., 2001). However, because the mutation in several BiHM families does not map to the genomic region containing DNMT1, it is likely not sufficient. Hence, all the mechanisms or molecular modifications that maintain the differential methylation of imprinted DMRs are unknown.
|
Hypotheses for the molecular defect underlying the formation of BiHM |
|---|
Current data show abnormal methylation of maternally methylated DMRs in BiHM and support that the defect is not in erasure but rather in the establishment or maintenance of maternal imprints (Fisher et al., 2002
; Judson et al., 2002; El-Maarri et al., 2003). It cannot be unequivocally excluded that maternal imprints are maintained in the early embryo but are subsequently lost in the trophoblast as it proliferates along its molar phenotype. However, this has been considered a less likely hypothesis, because there is relative homogeneity of the pathological and molecular findings in affected CHM samples (El-Maarri et al., 2003, 2005). Although Judson et al. (2002) concluded that the primary defect may lie in a failure to acquire or maintain primary DNA methylation imprinting marks in the maternal germ line, El-Maarri et al. (2003) in their study of imprinting in different BiHM samples also found evidence for a gain of methylation on the maternally inherited alleles of a primary paternal methylation mark. Samples that can be analysed are scarce, and, until the gene has been found and animal models can be created, this will remain an open question. What are the most likely attributes of the candidate gene(s) mutated in women with recurrent BiHM? They should be expressed either in the female gamete during the critical window when imprinting marks are established or in the preimplantation embryo when imprinting marks have to be protected from genome-wide reprogramming and CpG demethylation. The known pedigrees strongly suggest a true maternal-effect mutation. Hence, even though the candidate gene may encode a protein that carries out its function in the zygote or the embryo during the first-cleavage divisions, the mRNA should be produced from the maternal genome. The absence of a male reproductive phenotype in the pedigrees does not preclude the gene’s expression in male gametes but argues that its function is not essential for their normal development and function.
For those cases where the mutation does not map to the major locus on chromosome 19q13.42, one can only evaluate candidate genes based on their known role in imprinting mechanisms. Analysis of DNMT1o, DNMT3L, DNMT3a and DNMT3B in a BiHM family not linked to 19q13.4 did not reveal any mutations (Hayward et al., 2003). This is not surprising considering newer information on the function of these proteins. DNMT3L has no methyltransferase activity on its own but co-operates with DNMT3A to establish imprinting marks. Although theoretically an excellent candidate, recent data suggest that DNMT3L also participates in other de novo methylation events, as well as in the establishment of paternal DNA methylation imprints. DNMT3A is also a de novo DNA methyltransferase for many other regions of the genome (Chen et al., 2004), and its DNMT3A2 isoform is expressed in germ cells and specifically affects imprinting, but not in a maternal allele-specific manner (Kaneda et al., 2004). Because the establishment and maintenance of DNA methylation imprints may be preceded by the methylation of specific lysine residues on core histones, it is possible that proteins that perform these histone modifications or other chromatin remodelling factors are mutated in BiHM. Yet, the major histone methyltransferases, G9A, SUV39H1 and SUV39H2, are probably not involved, because they have a more widespread role in the genome (Tachibana et al., 2002; Rice et al., 2003). A novel histone-modifying enzyme with a unique specificity towards imprinted loci on the maternal genome or a factor that specifically targets more general histone-modifying enzymes to these regions would be a good candidate.
The currently established minimal candidate region on chromosome 19q13.4 contains more than 60 transcripts. By in silico analysis, a subset is uniquely or primarily expressed in complementary DNA libraries from germ cells, embryonic stem cells or ovaries. The presence of a transcript in the candidate region that encodes a protein with a SET domain, which is found in many known or putative histone methyltransferases, is very interesting. However, no mutations in the coding region of this gene have been identified in women affected with BiHM (Shao and Van den Veyver, unpublished data). Furthermore, the murine homologue of this gene has been shown to be a H4K20 methyltransferase that primarily acts on pericentromeric heterochromatin, and a role of this protein for histone methylation at imprinted regions has not been described (Schotta et al., 2004; van der Heijden et al., 2005).
|
Clinical implications for patients with recurrent HM |
|---|
BiHM have a similar clinical course and management as AnCHM. Patients with CHM classically present with vaginal bleeding, abnormally high levels of maternal serum HCG and uterine size greater than expected for the gestational age. The wide use of ultrasonography and serum HCG levels now allows for early diagnosis in asymptomatic women. The sonographic features of a uterine cavity filled with multiple sonolucent areas of varying size and shape (‘snowstorm’ appearance) and absence of embryonic or fetal structures combined with an elevated serum HCG level are highly indicative of the presence of CHM, even before the final histopathological diagnosis is confirmed (Jauniaux, 1998
). Medical complications of CHM include pregnancy-induced hypertension, anaemia, hyperemesis, thyrotoxicosis, embolization and the development of ovarian theca lutein cysts. Treatment options range from the simple suction evacuation with follow up to chemotherapy, hysterotomy or hysterectomy depending on the patient’s age, risk factors and completion of their family. Patients with CHM have an increased risk of persistent gestational trophoblastic disease, invasive mole and choriocarcinoma development, requiring treatment with methotrexate or combination chemotherapy. Although the total number of studied cases is limited, our own observations of women from two consanguineous pedigrees and other published reports are consistent with an increased risk for invasive trophoblastic disease after a BiHM that is at least as high as in sporadic AnCHM (Al-Hussaini et al., 2003), indicating that their clinical courses are largely comparable. These observations also support that it is not the apparent homozygosity for recessive mutations in AnCHM that predisposes to malignant transformation but rather the unbalanced expression of imprinted genes in both AnCHM and BiHM.
For patients with recurrent HM, reproductive options are currently very limited. In some pedigrees, pregnancies with normal outcomes have been reported in affected women with BiHM (Moglabey et al., 1999), but in the majority of families, including those we study, there are no normal pregnancy outcomes. Attempts to avert recurrent molar pregnancies with IVF combined with donor sperm and preimplantation genetic diagnosis have had poor success (Edwards et al., 1990, 1992; Reubinoff et al., 1997). In the reported cases, the affected pregnancies were not genotyped for biparental inheritance, as most were done before the reports of biparental inheritance of recurrent HMs. In one case, there appeared to be a higher incidence of triploid embryos (Pal et al., 1996) based on morphological analysis of the fertilized oocytes, which indicated that they contained three pronuclei. This suggests that some cases of recurrent HM may have a different aetiology than BiHM, but it remains to be proven whether they result from a defect in the paternal gametes for which it would be valid to attempt insemination with donor sperm as a therapy. Others have suggested that because some affected women have normal pregnancies interspersed with the recurrent moles, the disease may be under similar influences from environment or genetic modifiers as sporadic AnCHM and propose that once identified, these factors could potentially be modulated and open avenues for treatment in the future (Slim et al., 2005). Until then, patients with highly recurrent BiHM, who may repeatedly attempt spontaneous pregnancies (Al-Hussaini et al., 2003), should be counselled about the risk for persistent trophoblastic disease or invasive choriocarcinoma with each molar pregnancy. Once the underlying genetic defect is identified and is unequivocally proven to be oocyte-specific, a potential alternative strategy for those women with recurrent BiHM could be IVF with donor oocytes, especially in those families where affected women have only HMs and never a normal pregnancy outcome.