In many epigenetic phenomena, covalent modifications on DNA and chromatin mediate somatically heritable patterns of gene expression. Genomic imprinting is a classical example of epigenetic regulation in mammals. To date, more than 100 imprinted genes have been identified in humans and mice. Many of these are involved in foetal growth and deve lopment, others control behaviour. Mono-allelic expression of imprinted genes depends on whether the gene is inherited from the mother or the father. This remarkable pattern of expression is controlled by specialized sequence elements called ICRs (imprinting control regions). ICRs are marked by DNA methylation on one of the two parental alleles. These allelic marks originate from either the maternal or the paternal germ line. Perturbation of the allelic DNA methylation at ICRs is causally involved in several human diseases, including the Beckwith–Wiedemann and Silver–Russell syndromes, associated with aberrant foetal growth. Perturbed imprinted gene expression is also implicated in the neuro-developmental disorders Prader–Willi syndrome and Angelman syndrome. Embryo culture and human-assisted reproduction procedures can increase the occurrence of imprinting-related disorders. Recent research shows that, besides DNA methylation, covalent histone modifications and non-histone proteins also contribute to imprinting regulation. The involvement of imprinting in specific human pathologies (and in cancer) emphasizes the need to further explore the underlying molecular mechanisms.
Mammalian genomic imprinting is an epigenetic marking phenomenon leading to mono-allelic expression of a subset of genes . Mono-allelic expression depends entirely on the parental origin of the gene. Thus some imprinted genes are expressed only from the maternally inherited allele, whereas others are expressed exclusively from the paternal allele. Approximately 130 autosomal imprinted genes have been identified in the mouse so far (see http://www.mousebook.org/catalogue.php?catalog=imprinting). Many of these are imprinted in humans as well (http://igc.otago.ac.nz/home.html). The majority of the known imprinted genes are arranged in clusters of several tens up to thousands of kilobases (kb) in size. Imprinted gene expression across these evolutionarily conserved clusters is regulated by ICRs (imprinting control regions), essential DNA sequence elements that are up to several kilobases in size. ICRs are CpG-rich regions that are methylated only on one of the two parental alleles. Therefore they are also referred to as DMRs (differentially methylated regions).
Imprinted genes are involved in various biological processes. Not surprisingly, therefore, perturbation of their expression can cause embryonic or postnatal lethality, aberrant growth and, in some cases, abnormal behaviour. In humans, genetic alterations at imprinted domains include UPDs (uniparental disomies), in which one pair of chromosomes is inherited from one of the parents only. This alters the levels of expression of the imprinted genes concerned, causing various complex diseases. UPDs affecting entire chromosomes, or portions of chromosomes, are causally involved in the neuro-behavioural disorder PWS (Prader–Willi syndrome) and in the BWS (Beckwith–Wiedemann syndrome), associated with aberrant foetal growth (see below).
In genomic imprinting, covalent modifications are put on to specific loci during male or female gametogenesis. These epigenetic marks, which confer a ‘memory’ of the parental origin to these loci, are called ‘imprints’. Detailed molecular and genetic studies have determined that imprinting is controlled by cytosine methylation at ICRs. After fertilization, DNA methylation imprints at ICRs are maintained throughout embryonic development and postnatal life . In the developing embryo, these marks confer parental-allele-specific expression to close-by genes. In the primordial germ cells of the developing embryo, however, all pre-existing methylation at ICRs is erased, and new imprints are re-established at later stages, thus completing the imprinting cycle  (Figure 1).
DNA methylation in mammals is regulated by members of the DNMT (DNA methyltransferase) family . Several laboratories discovered that DNMTs play essential roles in the establishment and maintenance of methylation imprints. Factors that interact with DNMTs, or are associated with DNA methylation, are important for the imprinting process as well. In the first part of this chapter, we summarize the different factors known to be involved in the acquisition and maintenance of DNA methylation imprints. In the second part, we provide examples of human diseases that are caused by genetic or epigenetic perturbation of imprinted genes (summarized in Table 1).
Factors involved in the acquisition and maintenance of imprints
In the mouse, establishment of DNA methylation imprints in the developing male germ cells initiates at E (embryonic day) 14.5–15.5. Paternal ICRs become fully methylated several days later (at E17.5–18.5) when final sperm cell differentiation ensues. Establishment of methylation imprints in the female germ cells takes place in adult life only, during the final stages of oogenesis. The de novo DNMT Dnmt3a and a related co-factor called Dnmt3L are required for imprint establishment in both of the germ lines [2,3]. Although Dnmt3L does not have methyltransferase activity, it brings about the catalytic action of Dnmt3a, most likely by recruiting Dnmt3a to the chromatin . Another de novo methyltransferase, Dnmt3b, is also required for embryonic development, but it is not essential for imprint establishment.
Previous work suggests that histone modifications also influence imprint establishment in germ cells. Dnmt3L selectively binds to histone H3 in vitro, but only when H3 Lys4 (H3K4) is not methylated . Therefore H3K4 methylation could prevent acquisition of de novo DNA methylation. In agreement with this hypothesis, several maternal ICRs are enriched in H3K4 dimethylation (H3K4me2) in male germ cells (where they do not become methylated) . More recently, it was reported that H3K4 demethylation by KDM1B, a lysine demethylase, is an essential first step in the acquisition of DNA methylation imprints at some ICRs in oocytes  (Figure 2). Besides the absence of H3K4 methylation, the presence of other specific histone modifications could facilitate Dnmt3a recruitment to its target loci. Symmetrical dimethylation of histone H4 Arg3 (H4R3me2s) is one possible candidate. Using the β-Globin locus as a model, it was recently shown that Dnmt3a recognizes Prmt5-mediated H4R3me2s through its PHD (plant homeodomain) motif. This recognition triggers acquisition of new DNA methylation at this locus in developing blood cells .
Features of the DNA sequences of ICRs might contribute to the acquisition of DNA methylation as well. Structural studies on the Dnmt3a–Dnmt3L complex provided insights into how these proteins might recognize and methylate DNA . Specifically, the characteristics of the tetrameric protein structure formed of two Dnmt3a and two Dnmt3L proteins suggest that DNA sequences with a series of CpG dinucleotides at 8–10 bp intervals could be the preferred substrate. Interestingly, such a regular spacing between CpGs has been detected at many ICRs .
Non-histone proteins also contribute to the specificity of imprint establishment. A recent mouse targeting study showed that Zfp57, a zinc-finger domain protein, is involved in oocyte-specific imprint establishment at the Snrpn ICR . Furthermore, a screen for genetic mutations in familial cases of BWS led to the identification of NLRP2 (NALP2), another protein that contributes to the establishment of maternal imprints . The molecular function of NLRP2 is not clear. Genetic mutations in a related gene, NLRP7 (NALP7), had been associated before with familial cases of complete hydatidiform mole, a severe developmental condition in which epigenetic abnormalities are apparent at all of the maternal ICRs .
Somatic maintenance of parental imprints through development
Faithful maintenance of imprints is essential for normal development. Its perturbation is causally involved in different human pathologies and is thought to be an early contributing factor in cancer. This raises the question of how ICRs maintain their allelic DNA methylation in somatic cells and tissues. Continuous expression of Dnmt1 is clearly essential, as well as its level of expression. Conditional targeting of Dnmt1 has shown that, during the critical pre-implantation period, maternal and zygotic Dnmt1 are sufficient to maintain methylation imprints . Np95 (also known as Uhrf1) binds to methylated DNA through its SRA (SET and RING finger-associated) domain and forms complexes with Dnmt1 to mediate Dnmt1 loading to replicating heterochromatic regions. Consequently, Np95 maintains global and local DNA methylation and thereby represses imprinted genes .
There is growing evidence for involvement of other proteins, which are not part of the DNA methylation machinery. Zfp57, for instance, contributes to embryonic maintenance of maternal imprints. In embryos lacking Zfp57, methylation is largely lost from the Snrpn ICR and partial losses of methylation occur at other ICRs as well . Support for ZFP57 involvement in human disease has come from studies on patients with TNDM (transient neonatal diabetes mellitus) . In approximately one-fifth of cases, TNDM is caused by loss of DNA methylation at the ICR of the imprinted PLAGL1 gene (also called ZAC1) on chromosome 6q24. This subgroup of TNDM patients shows concomitant methylation losses at other ICRs, suggesting that ZFP57 interacts with multiple ICRs, probably through recruitment of repressive protein complexes. In one genetic study, the ZFP57 locus was found to be genetically homozygous in patients, but not in unaffected family members. Subsequent identification of deleterious genetic mutations in the gene confirmed its causal involvement in TNDM. Another key protein involved in imprint maintenance is PGC7/Stella, a maternal factor which is essential for early development and protects against loss of DNA methylation at several imprinted loci .
Genomic imprinting, assisted reproduction and human disease
Several developmental and growth disorders in humans are caused by epigenetic perturbations of imprints that probably arise during the early stages of development. Although the underlying mechanisms are unknown, the occurrence of these ‘epimutations’ could be linked to the parental genomes’ global reprogramming in the zygote at pre-implantation. Perhaps DNA methylation patterns are not yet firmly fixed in the non-committed cells of early embryos unlike differentiated cells at later developmental stages. Indeed, manipulation and culture of pre-implantation embryos and embryonic stem cells can readily perturb imprints, affecting imprinted expression and phenotype later in development [16,17]. Concordantly, the incidence of certain imprinting diseases appears to be increased following assisted reproduction. Different ARTs (assisted reproduction technologies) can affect the epigenetic regulation of imprinted genes in the babies born . ART had been long suspected to increase the occurrence of classical imprinting syndromes and recent studies have indeed reported their increase, particularly of the BWS and AS (Angelman syndrome), in babies born following ARTs [17–19]. The underlying causal mechanisms remain unknown. However, since methylation defects can occur concomitantly at several imprinted loci in individual patients, there could be involvement of factors regulating multiple ICRs.
Prader–Willi syndrome and Angelman syndrome
PWS and AS were the first reported imprinting disorders in humans. Genetically, they map to human chromosome 15q11-q13, to the so-called PWS/AS chromosomal region (Figure 3A). This region comprises a large cluster of imprinted genes which are causally involved in these syndromes. All imprinted genes in this cluster are expressed highly in the brain, from the paternally inherited chromosome, except one, UBE3A, which is expressed from the maternal allele. PWS affects one in 10000 children and is characterized by neonatal muscular hypotonia, hyperphagia and early childhood obesity, hypogonadism, short stature, small hands and feet, and mild mental retardation . Often this pathology is caused by paternal deletion or maternal UPD of the PWS/AS region, leading to loss of expression of all of the paternally expressed genes in the cluster.
AS has rather opposite clinical features and is characterized by developmental delay, ataxia, absence of speech, severe mental retardation, jerky arm movements and inappropriate laughter . In contrast with PWS, AS is often caused by maternal deletion of 15q11-q13. Whereas PWS is thought to be a contiguous gene syndrome with several imprinted (paternally expressed) genes causing the disorder, AS is caused by loss-of-expression of a single imprinted gene, UBE3A, which encodes an ubiquitin ligase involved in brain development.
The promoter and first exon of the SNRPN gene constitute the ICR of the PWS/AS domain. This ICR has a maternally derived DNA methylation imprint. Human studies indicate that this imprint becomes established after fertilization only. Thus the de novo DNA methylation machinery somehow recognizes the right parental allele in the zygote. Based on transgenic mouse models, it was proposed that this recognition involves a maternally derived chromatin signature . In the mouse, the Snrpn ICR acquires its methylation imprint at an earlier stage, during the maturation of the oocyte, and this involves both Dnmt3a and Dnmt3L. If Dnmt3L is absent during oogenesis (in Dnmt3L−/− females), the imprint is not established. However, ICR methylation is later detected in some of the offspring of Dnmt3L-deficient females. This intriguing observation supports the idea that, also in the mouse, there could be a chromatin signature which is inherited from the oocyte and mediates imprint acquisition at the Snrpn ICR during early development .
Silver–Russell and Beckwith–Wiedemann syndromes
The foetal growth disorders SRS (Silver–Russell syndrome) and BWS map to chromosome 11p15 and have opposite phenotypes. Whereas SRS is characterized by intra-uterine growth restriction, BWS shows foetal overgrowth . Chromosome 11p15 comprises two neighbouring imprinting clusters that are crucial for the control of foetal growth. The most proximal one is the large KCNQ1 cluster that comprises CDKN1C, a negative regulator of cellular proliferation. The other cluster is small and comprises IGF2 (insulin-like growth factor 2), a major regulator of foetal growth in mammals. SRS and BWS are most frequently caused by epimutations at the ICRs that control the KCNQ1 and IGF2 domains (Figure 3B), whereas 11p15 duplications and other genetic changes are infrequent [19,22,23,24].
SRS is a sporadic disorder with intra-uterine and postnatal growth retardation, facial dysmorphism, fifth finger clinodactily, feeding difficulties and, often, body asymmetry. Its incidence is 1 in 50000 births . Recent studies show that in approximately half of the patients, DNA methylation is lost at the ICR controlling the IGF2 imprinted domain. This leads to loss of IGF2 expression, explaining the severe growth retardation. Approximately 10% of SRS patients have maternal UPD of chromosome 7. It is unclear which imprinted gene(s) could be causally involved in these patients, but GRB10 (growth-factor-receptor-binding protein 10) is one of the candidates .
Conversely, BWS is characterized by overgrowth and increased risk of childhood cancers including Wilms’ tumour of the kidney. Its diagnostic features are high birth weight and length, macroglossia, midline abdominal wall defects, ear creases or ear pits, and neonatal hypoglycaemia after birth . It is mostly a sporadic disease and approximately 10% of BWS have the exact opposite epigenetic alteration at the ICR of the IGF2 locus than the one observed in SRS. These patients show gain of methylation at this ICR, on the parental allele which is normally unmethylated. This induces bi-allelic expression (and increased protein levels) of IGF2 . BWS arises not only from epigenetic alterations at the ICR of the IGF2 locus, but also through altered DNA methylation at the ICR which controls the neighbouring KCNQ1 domain. In approximately half of the patients, methylation is lost at this intronic ICR, which leads to bi-allelic silencing of CDKN1C, causing the observed overgrowth in this class of patients.
Transient neonatal diabetes mellitus
TNDM occurs in growth-retarded neonates who show persistent hyperglycaemia within the first 6 weeks of life. It has an incidence of ∼1 in 400000 births. Affected babies require exogenous insulin therapy for a mean duration of 3 months while endogenous insulin levels are either extremely low or undetectable. Recovery is usually complete by 18 months of age, although some patients relapse and develop diabetes later in life. Paternal UPD of chromosome 6 and large paternal duplications of 6q24 have been linked to TNDM. This led to the hypothesis that TNDM is caused by overexpression of an imprinted gene located in a ∼5.4 Mb region on chromosome 6q24. The disease was found subsequently to involve an imprinted locus that comprises the paternally expressed transcription factor gene PLAGL1. This locus has a putative ICR with a maternal DNA methylation imprint. In the majority of patients with TNDM, there is pronounced loss of methylation at this ICR, leading to bi-allelic PLAGL1 expression . Interestingly, PLAGL1 regulates the expression of other imprinted genes, including IGF2 and CDKN1C, and influences the transcriptional activity of KvDMR1, the ICR that controls the KCNQ1 domain . It is unknown which factors induce loss of DNA methylation at the PLAGL1 ICR, which may occur in BWS as well . However, a recent study shows that the zinc finger protein ZFP57 contributes to the maintenance of methylation at this (and other) ICRs .
Pseudohypoparathyroidism and Albright’s hereditary osteodystrophy
The GNAS locus on chromosome 20q13.2-13.3 is a complex locus that produces multiple transcripts through the use of alternative promoters and splicing. Genetic and epigenetic mutations at the GNAS locus are involved in different forms of PHP (pseudohypoparathyroidism), a disorder characterized by lack of responsiveness to PTH (parathyroid hormone) and additional hormones. Patients have short stature, rounded faces and, often, mild mental retardation. There are two main subtypes of PHP. PHP-type 1a patients show resistance to PTH and display a combination of physical features called AHO (Albright’s hereditary osteodystrophy). In these patients, the disease is mostly caused by heterozygous inactivating mutations in the GNAS gene. The parental origin of the mutations determines the severity of the hormone resistance. PHP type-2 patients do not have AHO, and the great majority of cases are caused by loss of DNA methylation at one of the ICRs of the GNAS locus.
Uniparental disomy 14
Maternal UPD of chromosome 14 (UPD14mat) causes pre- and post-natal growth retardation, congenital hypotonia, joint laxity, motor delay and mild mental retardation. Paternal UPD14 (UPD14pat) has a more severe phenotype, including polyhydramnios, thoracic and abdominal wall defects, growth retardation and severe developmental delay. Cases of segmental UPD14 indicate that the distal portion of chromosome 14q is the critical region. The 14q32.2 region comprises an imprinting cluster that includes the paternally expressed DLK1 and RTL1 and the maternally expressed GTL2 and RTL1as (RTL1 antisense) genes, all under the control of a single ICR. The clinical phenotypes linked to maternal and paternal UPD14 appear to be caused by deregulation of these imprinted genes through several mechanisms. Gene targeting studies in mice suggest that the ICR has an important cis-acting regulatory function on the maternally inherited chromosome, and that excessive RTL1 and decreased DLK1 and RTL1 expression are relevant to UPD14pat and UPD14mat phenotypes respectively . The imprinted 14q32.2 region is also associated with increased risk of Type 1 diabetes .
In this overview on imprinting and its deregulation in disease we focused on ICRs, the key regulatory elements involved. Epigenetic imprints at ICRs are established in one of the two germ lines and, after fertilization, are maintained throughout development. In the newly formed germ cells of the developing embryo, however, epigenetic imprints are removed to allow subsequent establishment of new imprints for the next generation. These sequential events – establishment, maintenance and erasure – constitute the developmental cycle of imprinting (Figure 1). Importantly, imprinting-related pathologies are often associated with epigenetic changes (Table 1). The discussed examples of such epimutations all concern the maintenance of imprints in the early embryo. Although maintenance mechanisms remain poorly understood, recent data indicate that this process also involves non-histone proteins and is linked to specific patterns of histone methylation. Future research will identify the specific histone methyltransferases (and demethylases) involved. Their functional importance in the regulation of imprints needs then to be explored in cells and animals, in a similar manner as has been done for DNMTs.
There is a growing interest in whether assisted reproduction could affect imprinted gene loci and thereby the development and health of the babies born. This important question emerged after it was demonstrated in mice and in ruminant species, that embryo culture can heritably affect imprinted gene expression. Various studies link assisted reproduction to increased frequencies of imprinting syndromes. These disorders occur at low frequencies in the general population and are still very rare following assisted reproduction. Other risks, including low birth weight and poor postnatal development, are significantly higher. In the coming years it would be relevant to explore at the genome-wide level to what extent DNA methylation and other epigenetic modifications are altered by assisted reproduction.
During the coming decade, genetic studies in humans and targeting experiments in mice will undoubtedly provide new important insights into the sequence elements and the protein factors that control imprinted genes. Detailed biochemical studies on the protein complexes involved and on how they modify chromatin and DNA will be required to better understand why epigenetic mutations arise in cultured cells and embryos, and sometimes during early development, leading to diverse pathologies.
• Genomic imprinting is an epigenetic mechanism leading to mono-allelic expression of certain genes depending on the parental origin of the allele.
• Imprinted genes are organized in conserved clusters. Their allelic expression is controlled by ICRs.
• ICRs are marked by parental allele-specific DNA methylation that is established during either oogenesis or spermatogenesis.
• In somatic cells, ICRs also show allele-specific histone methylation.
• After fertilization, the allelic chromatin imprints at ICR are maintained throughout development, in all the cell lineages.
• Perturbation of the epigenetic imprints at ICRs leads to perturbed gene expression and is causally involved in several human diseases.
• Environmental stresses, such as embryo and cell culture, or assisted reproduction technologies, can also readily perturb imprints, with long-term consequences for gene expression and phenotype.
- © The Authors Journal compilation © 2010 Biochemical Society