Keywords

1 Introduction

One of the remarkable discoveries of the twentieth century was the finding that some genes in mammals do not obey the rules of Mendelian inheritance. Accordingly, “imprinted” genes classically exhibit monoallelic expression in a parent-of-origin dependent manner [1,2,3]. Some imprinted genes are monoallelically expressed in all expressing cell types, while others show partial, tissue-specific, and/or temporal imprinting. Their most critical and fundamentally important characteristic is that parent-of-origin expression is dependent on passage through the parental germline [1]. Functional differences between the parental genomes were first experimentally demonstrated using pronuclear transfer experiments to artificially generate diploid mouse embryos whose genomes were exclusive of either maternal or paternal origin [4,5,6,7,8,9]. Development of these monoparental embryos was initially relatively normal. However, embryos consisting of two maternally-derived genomes became progressively growth-restricted and died around embryonic day (E) 10, with particularly limited development of extra-embryonic tissue. Conversely, embryos with two paternal copies were both developmentally delayed and growth restricted, dying around E8.5 with an abundance of extra-embryonic tissue. Similarly, monoparental embryonic stem cells in chimeras with wild-type cells allocate to essentially reciprocal regions in chimeric embryos [10, 11], highlighting the functional differences and requirement for both parental genomes. Studies of uniparental disomic (UPD) embryos delineated parent-of-origin functional differences to specific chromosomal regions [12], with these regions found to contain domains of maternally and paternally expressed genes [13,14,15].

The discovery of imprinted genes, along with their phylogenetic distribution with an increasing knowledge of their functions, has led to considerable speculation concerning the evolution of genomic imprinting with two prevalent explanations being genetic conflict over maternal resources [16, 17] and maternal-offspring adaptation [18]. These, and many other theories that attempt to ascribe the importance of genomic imprinting, account for the intimate and complex relationship between the mammalian mother and her offspring with increasing investment toward viviparity dependent on an elaborate placenta and considerable postnatal nurturing [19] (Fig. 8.1). An important aspect of imprinting is the control of gene dosage by epigenetic marks inherited from the parental germline. Epigenetic marks under some circumstances can undergo modifications in response to environmental cues such as low protein diet in pregnancy [20], with a potential to affect fetal growth and other phenotypic consequences later in life. The consequent flexibility conferred by the epigenetic mechanism may be advantageous for supporting adaptation to different environments, but may also be responsible for diseases. Understanding the mechanisms that establish, maintain, and erase imprinting is therefore important for our understanding of some human diseases.

Fig. 8.1
figure 1

Genomic imprinting and the increasing investment in viviparity in mammals. Within vertebrate lineages, genomic imprinting has only been observed in marsupials and, more prominently, in eutherians with prolonged fetal development in utero before birth. Mya—million years ago

Full size image

2 Establishment of Canonical Imprinting

Canonical imprinting resulting in monoallelic gene expression depends on the establishment of heritable DNA methylation marks in either the maternal or paternal germline, and these differentially methylated regions (gDMRs) are propagated in post-zygotic cells [1, 21, 22] (Fig. 8.2). In mice de novo DNA methylation at gDMRs is initiated in the male germline before birth [23,24,25] and detectable as DNA methylation more broadly at intergenic sequences and transposons [26]. For the two paternally imprinted loci, H19-DMR and IG-DMR, changes in histone modifications and high transcriptional read-through precede DNA methylation [27] by the de novo DNA methyltransferase DNMT3A and a DNMT-like protein DNMT3L; the latter lacks enzymatic activity [28,29,30,31]. De novo paternal methylation of the rodent-specific Rasgrf1 gDMR requires components of the PIWI-interacting RNA pathway which normally silences transposable elements in the male germline [32] and a second de novo DNA methyltransferase, DNMT3B [24]. For the H19-DMR and IG-DMR, these sequences acquire DNA methylation as part of a global mechanism methylating the mature sperm genome to an average level of ~90% [26]. A recent study showed recruitment of the DNMT3A/3L complex extensively to the genome of male germ cells by histone H3 lysine 36 dimethylation (H3K36me2) marks broadly deposited by the histone methyltransferase NSD1 [33].

Fig. 8.2
figure 2

The imprinting cycle of establishment, maintenance, spreading and erasure. DNA methylation imprints are initially established in the male and female germlines. After fertilization, imprints evade the waves of genome-wide reprogramming to become germline differentially methylated regions (gDMRs). These gDMRs act as imprinting centers from which the imprints “spread” to neighboring genes, establishing domains of maternally and paternally expressed genes. These gDMRs are maintained for the lifetime of the animal along with the monoallelic expression of a subset of imprinted genes. For many genes, monoallelic gene expression does not persist and/or can be tissue-specific

Full size image

De novo DNA methylation in the female germline occurs after birth, during the growth phase of oogenesis [34]. Between birth and adulthood, the average genomic DNA methylation level goes from ~2% in non-growing oocytes, to nearly 40% in fully grown, germinal vesicle oocytes [26, 35]. DNA methylation is preceded by the loss of histone H3 lysine 4 di/trimethylation (H3K4me2/3), and gain of trimethylation at lysine 36 (H3K36me3) [36]. These domains with the H3K36me3 marks are established over transcribed gene bodies by the histone methyltransferase SETD2, which associates with the elongating RNA polymerase II [37, 38]. The recruitment of DNMT3A/3L complex follows to the H3K36me3-marked regions to catalyze de novo DNA methylation [28,29,30,31, 34, 39,40,41], resulting in a characteristic oocyte DNA methylome with a strong correlation between DNA methylation and the levels of transcription [26, 42]. All maternal gDMRs discovered to date are CpG-rich promoter elements [43], and each of them is covered by a transcript initiating from nearby oocyte-specific promoters [42]. Transcription through gDMRs has been shown experimentally to be functionally required for de novo DNA methylation at 6 maternal germline DMRs: Nespas/Gnasxl and 1A, KvDMR1, PWS-IC, Zac1 igDMR, and Peg3DMR [42, 44,45,46,47,48]. The generality of this mechanism for all maternal gDMRs is furthermore supported by the loss of all maternal imprints in oocytes deficient for DNMT3A, DNMT3L, or SETD2 [29, 35, 37].

3 Maintenance

After fertilization, both parental genomes undergo extensive passive (maternal) and active (maternal and paternal) DNA demethylation [49]. However, DNA methylation at gDMRs survives this reprogramming, which is not seen for the rest of gametically methylated regions. The protection of DNA methylation at gDMRs is, in part, mediated by the zinc-finger proteins ZFP57 and ZFP445, which bind to a methylated TGCCGC consensus motif within gDMRs [50,51,52,53,54]. These KRAB zinc finger proteins recruit TRIM28 (also known as KAP1) and the histone methyltransferase SETDB1. Consequently, the recruitment of this protein complex at the methylated gDMRs introduces H3K9me3 marks in the region, which promotes preferential recruitment of the DNMT1 DNA methylation maintenance machinery. DNA methylation at gDMRs is subsequently propagated for the lifetime of the individual through the repeated action of the maintenance DNA methyltransferase DNMT1 [55,56,57] in conjunction with UHRF1 (also known as NP95) [58].

4 Spreading

gDMRs operate as imprinting centers (ICs; also known as imprinting control regions) to establish domains of maternally and paternally expressed genes [13,14,15]. In mice, there are estimated to be 21 gDMRs originating during oogenesis, while three of them are inherited from sperm. These epigenetic marks regulate the allelic expression of around 100 non-coding transcripts and protein-coding genes [15, 59, 60]. There is strong evidence that approximately fifteen of these domains are conserved in humans, while some other are unique to humans, notably with uniparental expression of genes restricted to the placenta [61]. While not all gDMRs have been functionally demonstrated to act as imprinting centers in mice, there are six that have been studied through targeted deletion and have been found to regulate imprinting of the domains within which they reside: Airn/Igf2r, Igf2/H19, Snrpn, Kcnq1ot1/Cdkn1c, Dlk1/Gtl2, and Gnas cluster [62,63,64,65,66,67]. Both spreading and maintenance of imprinted expression must be carried out in cis because the two alleles exist in the same nuclear environment. Some domains contain long non-coding RNAs (lncRNAs) [68] with the expression of the unmethylated parental allele. Some of these lncRNAs have been shown to have a role in the imprinting of a domain containing several imprinted loci, while others lncRNAs appear not to play such a direct epigenetic role [68,69,70,71,72,73,74,75,76,77].

5 Erasure

Erasure of imprints is prerequisite in each germline cycle in preparation for the establishment of new epigenetic marks according to the sex of the embryo. The erasure of DNA methylation imprints occurs in the precursors of the germline, the primordial germ cells (PGCs). Genome-wide analyses of DNA methylation in mouse PGCs showed that not all genomic regions follow the same demethylation kinetics and that near-complete erasure of the genome occurs in two distinct waves. In phase I of reprogramming, from E8.5 to E10, the genome of migrating PGCs is gradually demethylated by passive dilution of 5mC marks [25, 78]. Specific genomic sequences—called late-demethylating—are largely protected from this first wave of passive demethylation. These include germline-specific genes, X-linked CpG islands (CGIs) in female embryos, specific families of repetitive elements, and imprinted gDMRs [25]. The beginning of phase II of demethylation coincides with the colonization of the genital ridges by PGCs [79]. It is characterized by a rapid conversion of 5mC to 5-hydroxymethylcytosine (5hmC) by the ten-eleven translocation dioxygenase TET1, followed by a passive loss of this epigenetic mark, which is not maintained by DNMT1 [80,81,82,83]. From E10.5 to E13.5, the methylation levels of imprints fall from 70% to ~10% on average, while the genome reaches its lowest level of global DNA methylation known, at close to only 3% [25]. A strikingly similar dynamic of demethylation has also been observed in human PGCs, suggesting that epigenetic reprogramming of the PGC genome is a conserved process, but the temporal sequence and possibly the mechanism might differ from the observations in mice [84]. Failure to erase imprints in gonocytes leads to abnormal expression of imprinted genes and embryonic lethality in the next generation because of the direct inheritance of abnormal grand-parental imprints [81, 85].

So far, TET1 is the only factor known to be implicated in phase II reprogramming in gonocytes in mice. A careful analysis of the progeny from Tet1-null mice revealed a broad spectrum of embryonic and postnatal phenotypes, suggesting abnormal expression of imprinted genes [85]. The analysis of DNA methylation patterns at imprinted gDMRs revealed that loss of TET1 in one of the parents is associated with abnormal biallelic DNA methylation at gDMRs in the offspring [85, 86]. These important results established the essential function of TET1 in imprint erasure since they are consistent with a failure to erase grand-parental imprints, which are then abnormally passed on to the progeny: perdurance of grand-paternal DNA methylation marks would be manifested when inherited maternally from Tet1-null females, and the converse for grand-maternal marks. The precise timing and mechanism of imprints erasure in the human germline merits further investigation.

6 Portrait of an Imprinted Domain: Mouse Distal Chromosome 7

The imprinted region on mouse distal chromosome 7/human chromosome 11p15 has been the focus of considerable study in both humans and mice due to links with the imprinting disorders Beckwith–Wiedemann Syndrome (OMIM 130650; BWS) and Silver-Russell Syndrome (OMIM 180860; SRS), which both show parent-of-origin features. Evidence for the presence of imprinted genes at this chromosomal location came from studies on UPD7 mice which suggested the presence of more than one imprinted gene [87, 88]. Maternal UPD of the distal chromosome 7 region resulted in significant impairment of fetal and placental growth, with fetal death by embryonic day (E) 17.5 [88]. Paternal UPD of the same region resulted in delayed embryonic development with lethality by E10.5, alongside significant developmental abnormalities in the placenta [88,89,90]. We now know that this region contains two adjacent but mechanistically distinct imprinted domains (Fig. 8.3A). Moreover, these domains best exemplify what is known about the mechanisms that establish domain-wide imprinting.

Fig. 8.3
figure 3

The imprinted domains on mouse distal chromosome 7. (A) The imprinted regions on mouse distal chromosome 7 showing the organisation and direction of imprinting of genes within the IC1 and IC2 domains. (B) IC1 domain imprinting involves differential conformation of the paternal and maternal alleles dependent on methylation-sensitive binding of CTCF. (C) IC2 domain imprinting requires the transcription of the long, non-coding RNA Kcnq1ot1 from the paternal allele, which is thought to attract the chromatin complexes PRC1 and PRC2 that add silencing histone marks along the domain, and may involve physical tethering to nucleoporins

Full size image

7 The IC1 Imprinted Domain

In mice, the IC1 domain spans over 100 kilobases and contains the paternally expressed protein-coding genes Igf2 [87, 91] and Ins2 [92], and the maternally expressed non-coding RNA H19 [93]. Contained within H19 is a functionally important imprinted microRNA, miR-675 [94]. The gDMR within the IC1 domain, called H19-DMR, is located approximately 2 kb upstream of the H19 promoter. This region acquires DNA methylation exclusively in the male germline, resulting in paternal-allele-specific DNA methylation in offspring [95]. Other regions of differential DNA methylation in the IC1 domain include the promoter of H19 on the silenced paternal allele, three other DMRs within the body of the Igf2 gene also on the paternal allele which, in the case of Igf2, is the expressed allele. These sites in sperm are unmethylated but they acquire methylation after implantation and are referred to as somatic DMRs (sDMR) [96]. Maternal inheritance of targeted deletions of H19-DMR results in biallelic Igf2 and Ins2 expression alongside loss-of-expression of H19, even when the H19 promoter is left intact [63, 97]. The physical arrangement of paternally and maternally expressed genes at this locus and the observation of similar expression patterns for Igf2 and H19 in the developing embryos led to the idea that these genes compete for the same enhancers [97]. H19-DMR contains binding sites for zinc finger protein CCCTC-binding factor (CTCF) that is conserved amongst mammals [98,99,100]. Binding of CTCF to H19-DMR only occurs when this region is unmethylated [98,99,100]. Accordingly, the binding of CTCF is thought to organize the domain into a specific chromatin conformation [101,102,103,104] which permits expression of only H19 from the maternal allele [102, 105, 106]. In the absence of CTCF binding, a different conformation is adopted, resulting in the expression of Igf2 and Ins2 but not of H19 (Fig. 8.3B). Similar mechanism involving a structural conformation dependent on methylation-sensitive CTCF binding regulates imprinted expression of the paternally methylated Dlk1-Dio3 domain [107].

8 The IC2 Imprinted Domain

The IC2 domain spans approximately 800 kb and contains a number of protein-coding genes with preferential expression of the maternally inherited allele, including Phlda2 [108], Cdkn1c [109] and Ascl2 [110], and the paternally-expressed lncRNA, Kcnq1ot1 (aka Lit1) [65] (Fig. 8.3A). The gDMR within the IC2 domain is known as KvDMR1 (Kv-differentially methylated region 1) [65, 111, 112]. KvDMR1 is methylated on the maternal allele and contains the promoter for Kcnq1ot1 [111]. Paternal inheritance of a targeted deletion of this region in mice results in biallelic expression of the normally maternal-expressed genes upstream and downstream of the gDMR, and the loss of repressive histone modifications that normally coat the paternal allele [65, 75, 113,114,115]. Unlike in the IC1 domain, transcription of the lncRNA Kcnq1ot1 rather than the DMR itself, is required to establish and maintain paternal silencing of the IC2 domain [75, 76]. All the protein-coding IC2 domain genes require DNA methylation in cis on the maternal KvDMR1 allele for their maternal expression [28,29,30, 116,117,118,119]. One of the consequences of Kcnq1ot1 expression is the acquisition of a somatic DNA methylation mark on the silent paternal allele of Cdkn1c [120]. In mice deficient for the HELLS helicase (aka LSH), this somatic DMR is not appropriately established resulting in the biallelic expression of Cdkn1c [121]. Furthermore, the maintenance of this somatic DMR by DNMT1 is required for the continued silencing of the paternal allele of Cdkn1c during development [120, 122, 123]. This additional layer of epigenetic gene silencing suggests the importance of sustaining imprinting of Cdkn1c into adulthood.

The loss of domain-wide imprinting at the IC2 domain mirrors the earlier findings for the imprinted locus spanning the maternally-expressed Igf2r gene, whereby truncation of the lncRNA Airn resulted in biallelic expression of Igf2r in the embryo and placenta, and of the downstream genes Slc22a2 and Slc22a3, normally imprinted and maternally expressed in the placenta [69]. Various models have been proposed to explain how the expression of paternal lncRNAs mediates domain-wide imprinting. The presence of lncRNAs overlapping with and oppositely transcribed to protein-coding genes, as observed for Kcnq1ot1/Kcnq1 and Airn/Igf2r, suggested the possibility of transcriptional interference [124], supported by the finding that Airn transcription through the Igf2r promoter was necessary to silence Igf2r [73]. However, while this could account for imprinted genes with promoters overlapping with lncRNAs, for the majority of loci there are imprinted genes that lie upstream of the lncRNA and ones that are too far downstream to be overlapping with the transcription of the lncRNA itself. An alternative idea was that lncRNAs function by disrupting the promoter-enhancer interactions required for active gene transcription, although there are no enhancers for Slc22a2 or Slc22a2 within the Airn transcribed region [125], and enhancers for Cdkn1c lie outside the region spanned by Kcnq1ot1 [126]. LncRNAs may instead prevent the expression of nearby genes in cis by forming RNA “clouds” that coat the domain and recruit inactivating histone-modifying complexes [114, 127,128,129] (Fig. 8.3C). LncRNAs such as Kcnq1ot1 may additionally function by physically localizing and tethering the paternal allele to the nuclear periphery [127, 130, 131]. There is some evidence that tethering involves complexing with nucleoporins which are the main components of nuclear pore complexes embedded in the nuclear membrane [132]. While posttranscriptional knockdown of Kcnq1ot1 by RNAi does not impact imprinted gene expression in stem cells [133], conditional deletion of KvDMR1/IC2 in the early mouse embryo results in loss of both paternal gene silencing and acquisition of somatic DNA methylation [134], which suggests differences in the epigenetic mechanisms involved in short and longer-term silencing of the IC2 domain.

9 Non-canonical Imprinting

Recent genome-wide next-generation sequencing screens have identified genes exhibiting parent-of-origin behavior that lie outside imprinted domains and are not clearly associated with gDMRs [22, 135, 136]. These are termed non-canonical imprinted genes and their imprinted expression is mostly transient, generally restricted to the early embryo and extra-embryonic lineages. These genes rely on allelic histone modification, particularly histone H3 lysine 27 trimethylation (H3K27me3), following direct inheritance from the oocyte [137,138,139]. These observations revealed that inheritance of DNA methylation is not the only epigenetic mark for the direct inheritance from the mature gametes to regulate the expression of imprinted genes during development. The functional relevance of non-canonical imprinting in humans merits further investigation [140, 141].

10 Evolution of Genomic Imprinting

The evolution of genomic imprinting remains a fascinating subject for discussion since imprinting leads to monoallelic expression of specific genes with critical roles in development. Why would evolution favor a system that is vulnerable to recessive mutations [16, 17], and why is this specific phenomenon observed in mammals and flowering plants [142]. Clues may come from the observation that genomic imprinting leading to parent-of-origin-specific expression in not universal to all mammals and has not been reported in egg-laying mammalian species, such as monotremes, platypus and echidnas [143]. The emergence of imprinting in mammals predates the bifurcation between metatherians (marsupials) and eutherians (placentals), with some imprinted genes being shared amongst them. This suggests a link to the emergence of viviparity and fetal development in utero, leading to new energetic demands on the pregnant female [144] (Fig. 8.1). A few “classical” imprinted genes appear to be well conserved in mammals, but several species-specific differences have also been observed.

With respect to mechanisms, two different mechanisms that might have contributed to the emergence of species-specific imprinted genes, both related to transcription-coupled de novo DNA methylation in oocytes, have been described. The first mechanism implicates the generation of a new gene via retrotransposition within the intron of a host gene. Since this host gene is expressed in growing oocytes, the inserted retrogene acquires a maternal gDMR and becomes imprinted and paternally expressed in the progeny. Four different examples have been documented, one of which is rodent-specific (Zrsr1/U2af1-rs1), while the other three are more evolutionarily ancient, and also imprinted in human (Mcts2, Nap1l5, Inpp5f_v2) [145,146,147,148]. Although the retrotransposition event might have triggered the formation of a new imprinted gene, it is also possible that the necessary signals were already present at the ancestral host gene, although these are not imprinted themselves.

A second mechanism is based on the insertion of a new oocyte-specific promoter next to a host gene promoter. Here, the oocyte promoter is provided by a long terminal repeat (LTR) retrotransposon. Specific families of LTR elements evade silencing mechanisms operating in oocytes and act as promoters for diverse transcripts in growing oocytes. Since LTR elements are highly polymorphic and different families of elements have colonized different lineages during evolution, they have been shown to be responsible for species-specific differences in the DNA methylome of oocytes, which is guided by transcription [149]. Notably, specific LTR elements drive oocyte transcription through a number of species-specific gDMR. Indeed, a comparison of mouse and human maternal gDMRs has identified 4 mouse-specific and 17 human-specific examples of paternally-expressed imprinted genes acquiring DNA methylation in oocytes as a consequence of a nearby LTR promoter [61]. These results suggest that novel imprinted genes can be generated during evolution at least in part via the insertion of an active LTR promoter through retrotransposition. The targeted deletions of the upstream LTR at the imprinted genes Slc38a4 and Impact lead to loss of imprinting and biallelic expression, providing functional support for this model [61].

Once established, further elaboration of canonical imprinted genes and arrangements in domains can be envisaged. Regulated by gDMRs, three essential requirements must be met for functional imprinting to evolve: (i) acquisition of a DNA methylation imprint in one germ line; (ii) maintenance of the gDMR through the preimplantation phase of epigenetic erasure; and (iii) prevention of acquisition of DNA methylation on the unmethylated allele during postimplantation waves of de novo methylation [150].

Endogenous retroviruses have also been implicated in the regulation of allelic expression at non-canonical imprinted genes, which are all paternally expressed and show imprinted expression only in preimplantation embryos and extra-embryonic lineages. More importantly, imprinting in this case is independent of gametic DNA methylation and relies instead on direct inheritance of repressive H3K27me3 histone marks from the oocyte [137], but these imprints are transient and do not persist after implantation. The expression of paternally inherited non-canonical imprinted genes in the extra-embryonic ectoderm showed that the maternal allele acquires DNA methylation as a somatic DMR, via DNMT3A/3B activity [138, 139]. So the germline imprint at those genes with a histone mark (H3K27me3) is eventually replaced by a DNA methylation mark for maintenance of imprinted expression in the extra-embryonic tissues. Furthermore, this epigenetic mechanism preferentially targets endogenous retroviruses, specifically ERVK LTRs [138]. These repetitive elements, which can act as promoters or enhancers in extra-embryonic tissues, appear to guide this unusual form of tissue-specific imprinting by protecting the paternal allele from DNA methylation-mediated silencing, which is the default pathway in embryonic lineages. Since endogenous retroviruses are implicated in non-canonical imprinting, we can expect this mechanism to regulate mostly species-specific imprinted genes.

11 The Function of Imprinted Genes in Mouse Development

Several imprinted genes function in a dosage-sensitive manner to regulate fetal growth and placental development early in life, and to influence both metabolic and behavioral processes later in life including those relating to mothering, as has been extensively and elegantly reviewed [22, 151,152,153,154,155,156]. As with imprinting mechanisms, the function of imprinted gene located within the mouse distal chromosome 7 epitomize the function of imprinted genes more generally.

12 IC1 Domain Genes

Igf2 was one of the first imprinted genes to be identified in mice [87, 91], along with H19 [93] and Igf2r [157]. Igf2 encodes a fetal growth factor structurally related to insulin that normally signals to promote cell proliferation, growth, differentiation, and survival via IGF1R. IGF2 also binds to IGF2R, which sequesters and degrades excess IGF2 [158]. The finding that IGF2R is a maternally-expressed imprinted gene [157] was instrumental in supporting the idea that genomic imprinting evolved in response to a parental “tug-of-war” imposed by the development of mammalian offspring in utero [16, 17]. While deletion studies were important for understanding the normal function of IGF2, manipulating the dosage of Igf2 can provide greater insight into the function of imprinting. In mice, loss of imprinting (LOI) of Igf2 results from disruption of H19-DMR. Expression of both Igf2 and Ins2 occurs at approximately twice the normal level in this model with concomitant loss of H19, resulting in increased birth weight of between 8 and 30% together with placental overgrowth [97, 159]. Fetal growth is supported by the placenta which functions both in the transport of nutrients and in securing the availability of nutrients through the action of placental hormones on the mother [160]. IGF2 positively regulates the development of the region of the placenta involved in nutrient transport [161, 162]. Together, these experiments identify Igf2 as a paternally-expressed imprinted gene that normally functions to promote fetal growth and instruct changes in the placenta required to enhance nutrient transport.

H19 was the first noncoding transcript to be identified as an imprinted gene, and one of the most abundant polyadenylated RNAs in the developing mouse embryo [93, 163]. The precise function of H19 has been challenging to study because of the mechanistic and reciprocal link with Igf2 imprinting. Accordingly, a deletion to facilitate the loss of H19 resulted in upregulation of Igf2 alongside an overgrowth phenotype despite not disrupting the gDMR [159]. Mice lacking the H19 transcript appear to develop normally [164] but there is evidence that H19 functions as a tumor suppressor with the loss of H19 associated with faster progression in mouse tumorigenesis models [165]. H19 function may, in part, be mediated by the microRNA mir-675 which is processed from the first exon of the H19 transcript in both mice and humans [94, 166, 167]. In mice, loss-of-function of Ins2 alone does not lead to developmental or metabolic dysfunction [168, 169] but when combined with ablation of the related Ins1 gene, pups are born small and die shortly thereafter from neonatal diabetes [168].

13 IC2 Domain Genes

Loss of imprinting of the IC2 domain, in which all paternally-silenced IC2 domain genes become biallelically expressed, results in both fetal and placental growth restriction [65]. Within this domain, the function of Cdkn1c, Phlda2, and Ascl2 has been studied most intently. Cdkn1c encodes a cyclin-dependent kinase inhibitor, which normally suppresses proliferation and promotes differentiation in a wide range of tissues. Ablation of Cdkn1c in mice results in fetal overgrowth, placentomegaly with expansion of placental endocrine lineages, abdominal wall defects, cleft palate, renal dysplasia, adrenal cytomegaly, maternal preeclampsia, and premature birth [170,171,172,173,174,175]. Transgenic mice expressing Cdkn1c at twice the normal level are growth restricted identifying Cdkn1c as the major regulator of embryonic growth within the IC2 domain [176]. These experiments in mice identified Cdkn1c as a maternally-expressed imprinted gene that normally functions to restrain fetal growth and placental development. Imprinting of Cdkn1c (reduced paternal expression) would therefore be predicted to enhance fetal and placental growth, and potentially support longer gestation.

Phlda2 encodes a PH domain-only protein normally maternally expressed in the mouse and human placenta where expression is highest [108, 177]. Phlda2 functions to limit expansion of a major placental endocrine lineage in mice [178,179,180,181]. Precisely regulated expression of Phlda2 is necessary for normal fetal growth with both loss- and gain-in-expression resulting in growth restriction [179, 181, 182]. Just two-fold expression of Phlda2 driven by a transgene resulted in >10% reduction in birthweight with relative sparing of the head, neonatal hypoglycemia followed by catch-up growth [182]. Rather than playing a direct role in regulating fetal growth, fetal growth restriction occurs as a consequence of the placental endocrine insufficiency induced by the reduction in placental endocrine cells. In addition to regulating placental development and fetal growth, Phlda2 plays a critically important role in instructing maternal caregiving behavior through controlling the production of placental hormones that act on the mother. Wild-type dams carrying and caring for offspring with different doses of Phlda2, behave atypically toward their offspring [183]. Dams carrying and caring for Phlda2 knockout offspring with an expanded placental endocrine compartment engage in more pup-focused behaviors whereas dams carrying and caring for Phlda2 transgenic offspring with a reduced placental endocrine compartment neglect their pups and prioritize nest building [183]. A direct role for imprinted genes in regulating maternal behavior has previously been highlighted by studies on the paternally expressed genes Peg1 and Peg3 but in both these examples, the dam was genetically altered [184, 185]. In the case of the Phlda2 studies it is the genetically altered offspring that influence their mother’s behavior before birth, consistent with the function of placental hormones in programming the maternal brain in pregnancy. High-quality maternal care is important for the offspring’s later life behavior, and both transgenic Phlda2 offspring and their wild-type siblings sharing the same abnormal environment were found to exhibit anxiety-like symptoms, mild depression, atypical social behavior, and reduced cognitive abilities as adults [186]. Together, these studies illustrate the far-reaching roles of imprinted genes both within and across generations.

Ascl2 encodes a bHLH transcription factor imprinted in the mouse placenta but not the fetus [110]. Ascl2 plays a critically important role in the placenta supporting the development of the endocrine lineages with full ablation resulting in embryonic lethality at E9.5 [110, 187, 188]. More modest reductions in the expression level of Ascl2 are also associated with placental defects alongside fetal growth restriction [189, 190]. Conversely, overexpression of Ascl2 results in reduced development of several placental endocrine lineages and fetal growth restriction [191]. These experiments indicate that a single dose of Ascl2 in the placenta is required for the proper allocation of progenitors to the different placental endocrine lineages required to support optimal fetal growth.

Cdkn1c, Ascl2, and Phlda2, through limiting the developing of the placental endocrine lineages, all function to limit the production of placental hormones. Consequently, alterations in their dosage can impact both fetal growth and maternal adaptations to pregnancy which in human populations may manifest as classic complications of pregnancy [192].

14 Imprinted Genes Influencing Adult Physiology and Behavior

In addition to these fundamentally important roles during gestation, imprinted genes also function to influence behavioral [22] and metabolic processes [193] critically important for the survival of mammals. Briefly, both Igf2 and Cdkn1c are expressed in the developing and adult nervous system and both genes are important for adult neurogenesis, with IGF2 functioning both as a paracrine and autocrine factor [194] while Cdkn1c is required for neural stem cell quiescence and differentiation [195]. Cdkn1c is also required for the maturation of midbrain dopamine neuronal cells [196]. Consistent with important roles during brain development and into adulthood, manipulating the expression of these genes results in alterations in behavior. Increased dosage of Cdkn1c has been linked to altered reward behaviors and social functions in mice [197,198,199] and Igf2 has been shown to be important for memory consolidation and cognitive function [200]. There is evidence for expression of the normally silenced allele of these two genes in the adult brain with the maternal expression of Igf2 in the choroid plexus and leptomeninges [201], functionally important for neurogenesis [194], and paternal expression of Cdkn1c in the brain important for neocortical development [202]. In addition to the nervous system, both Igf2 and Cdkn1c are expressed during development in tissues with important metabolic functions. For example, Igf2 and Cdkn1c are both relatively highly expressed in the developing pituitary [203] where Cdkn1c regulates progenitor proliferation but not differentiation [204]. Igf2 null mice possess more brown adipose tissue just prior to their birth [205] while Cdkn1c is both required for the proper differentiation of intrascapular brown adipose and stimulates “browning” within white adipose depots in mice [206]. Igf2 functions in different lineages to regulate both the size and function of the pancreas [207] while Cdkn1c regulates cell cycle exit of progenitors during the early stages of pancreas formation [208]. These multi-fold, multi-tissue roles mean that precisely regulated expression of Igf2 and Cdkn1c is key not only for early development but also for later life metabolism and behavior.

15 Variations in Genomic Imprinting and Human Disease

15.1 Genomic Imprinting Disorders

A number of classic congenital disorders are associated with alterations in imprinted domains, involving both genetic and epigenetic abnormalities in patients (Table 8.1) [209, 210]. These disorders classically exhibit parent-of-origin transmission, where transmission is possible, with clinical features commonly affecting growth, metabolism, and behavior consistent with the identified functions of many imprinted genes.

Table 8.1 Overview of eight different imprinting disorders in humans
Full size table

16 Beckwith–Wiedemann Syndrome (BWS; OMIM #130650)

One of the most common imprinting disorders is BWS, which is associated with a variety of genetic and/or epigenetic alterations localized to human chromosome 11p15.5 [210] (Fig. 8.4). BWS is estimated to occur in one in every 15,000 births and primarily involves overgrowth features [211]. BWS is usually diagnosed based on the presence characteristics including macrosomia (birth weight >97th percentile), macroglossia (unusually large tongue), neonatal hypoglycemia, ear creases or pits and abdominal wall defects, hemihypertrophy (one side of the body or a part of one side of the body is larger than the other), visceromegaly, nervus flammeus (port-wine stain), cleft palate, cardiac abnormalities, advanced bone age, enlarged placenta and abnormalities in placental vasculature. There is a high incidence of premature birth for BWS infants, sometimes in combination with polyhydramnios (excessive amniotic fluid) and gestational hypertension [212], with some BWS mother’s suffering the potentially life-threatening disorder of preeclampsia with HELLP (hemolysis, elevated liver enzymes, and low platelet count) [213]. Infants are at increased risk of developing congenital/childhood tumors such as Wilms’ tumor, adrenocortical carcinoma, hepatoblastoma, and neuroblastoma, although this risk decreases with age. Five to ten percent of BWS patients exhibit gain of methylation at H19-DMR (IC1), predicted from mouse studies to result in overexpression of IGF2, while 40–50% of patients have loss of DNA methylation at KvDMR1 (aka IC2) predicted to result in loss of expression of all the IC2 domain protein-coding genes. Although several BWS patients may sporadically lose DNA methylation at IC2 as an epimutation, recent evidence points to cis-acting genetic causes in some cases. Since IC2 is located in an intron of the KCNQ1 gene, implicated in long QT syndrome-1 [111], de novo establishment of the maternal DNA methylation imprint at IC2 is predicted to be guided by transcription through the region during oocyte growth, possibly from a KCNQ1 transcript initiating at its canonical start site. Such a model is in fact supported by studies of mouse mutants carrying the insertion of a transcriptional termination sequence adjacent to IC2, which prevents extension of a Kcnq1 transcript across IC2 [46]. That such a mechanism is conserved in humans is supported by rare BWS patients, also affected by long QT syndrome, in which transcription from the KCNQ1 promoter through IC2 is perturbed by maternally inherited translocations, promoter deletions, splice variants, or duplications within the KCNQ1 locus [214,215,216,217]. Patients with familial BWS carry inactivating germ line mutations in the coding sequence of the maternally inherited CDKN1C allele [218, 219], further highlighting the pathological contribution of CDKN1C in BWS. Functional studies in mice have been highly informative for our understanding of the gene changes underlying BWS [220]. While overgrowth in BWS could result from either too much IGF2 or too little CDKN1C, mice that overexpress IGF2 do not exhibit defining features of BWS such as cleft palate and abdominal wall defects which are observed in response to loss of murine Cdkn1c [170, 171]. It seems likely that either or both alterations have potential to contribute to BWS.

Fig. 8.4
figure 4

Contribution of imprinted genes to both rare and highly common human diseases exemplified by human chromosome 11p15. Genetic and epigenetic disruptions to imprinted domains are associated with rare imprinting disorders while variations in the expression levels of individual genes or IGNs may underlie highly common conditions such as low birth weight, type 2 diabetes, obesity, neurodevelopmental and behavioral disorders

Full size image

17 Silver-Russell Syndrome (SRS; OMIM #180860)

SRS is a very rare imprinting disorder with approximately 70% of patients having alterations affecting human chromosome 7 or 11, with 30% of unknown origin [210] (Fig. 8.4). SRS is diagnosed with an approximate frequency of 1 in 300,000 but may be far more common. SRS is defined by fetal growth restriction and failure to thrive postnatally and some or all of the following: normal head circumference, triangular-shaped face with a large protruding forehead, clinodactyly, undergrowth of one side of the body (hemihypotrophy), fasting hypoglycaemia, night sweats and excessive thinness into adulthood. An international consensus statement summarizing recommendations for clinical diagnosis, investigation, and management was published in 2017 [221]. As with BWS, the cognitive and behavioral characteristics of SRS are less well established but there are reports of specific learning difficulties [222,223,224,225], hyperactivity [226], attention deficits [224], autistic regression [227], and eating difficulties [228,229,230,231]. Maternal uniparental disomy of chromosome 7 is present in 5–10% of SRS patients (mUPD7; two maternal copies). Within the duplicated region there are two imprinted domains one of which contains the maternally-expressed GRB10 gene which has been implicated in both growth and behavior in mice [232, 233] and one of which spans MEST (aka PEG1) which has similar growth regulatory properties in mutant mice [184]. The larger proportion of SRS cases (30–60%) have hypomethylation of H19-DMR predicted, from mouse studies, to be associated with loss of expression of IGF2. A few SRS patients carry maternal microduplications of the 11p15 IC2 domain with the minimally duplicated region spanning CDKN1C, KCNQ1, PHLDA2, and SLC22A18 [234,235,236]. These patients are predicted to have twice the normal level of CDKN1C expression. Rarely, patients diagnosed with SRS possess mutations within CDKN1C thought to increase the stability of the protein [237,238,239]. Maternally inherited dominant missense mutations primarily within the PCNA-binding domain of CDKN1C have been reported in IMAGe syndrome (Intrauterine growth retardation, metaphyseal dysplasia, adrenal hypoplasia congenita, and genital anomalies; # 614732) [240, 241] (Fig. 8.4). IMAGe syndrome is a very rare condition combining intrauterine growth restriction, metaphyseal dysplasia, adrenal hypoplasia congenita, and genital anomalies and is clinically distinguished from SRS by the presence of adrenal insufficiency. As with BWS, mouse studies provide support for too little IGF2 or too much CDKN1C in SRS cases linked to chromosome 11p15. However, mice with loss-of-expression of Igf2 do not exhibit the more defining features of SRS such as lack of body fat and altered behaviors, which are seen in mouse models with gain-in expression of Cdkn1c [197,198,199, 206, 242].

18 Multilocus Imprinting Disorders (MLID)

A number of patients with BWS have epigenetic alterations affecting additional imprinted regions with the first suggestion coming from the finding that some patients diagnosed with transient neonatal diabetes mellitus (TNDM) also have hypomethylation at KvDMR1 [243]. It is now clear that a significant proportion of patients with a diagnosed imprinting disorder have an MLID, whereby more than one imprinted domain is impacted [244, 245]. In BWS 20–50% of patients with loss of KvDMR1 (IC2) methylation have an MLID, and MLID have been reported in approximately 15% of SRS cases [246]. One widely held explanation for simultaneous epigenetic defects at multiple ICs is a failure to maintain DNA methylation at these sites during the genome-wide reprogramming that occurs shortly after fertilization. Evidence for such a mechanism comes from studies on TNDM patients with DNA methylation defects at other imprinted loci, many of whom carry recessive mutations in ZFP57 [247], one of the proteins that maintain allelic DNA methylation and H3K9me3 at imprinted DMRs in mouse preimplantation embryos [52, 53]. Although mutations in ZFP57 have not, as yet, been reported in BWS or SRS patients with MLID [248, 249], a case of BWS with MLID has been described in the progeny of a mother homozygous for mutation in NLRP2 [250]. Interestingly, mutations in the related NLRP5 gene, coding for a component of the oocyte subcortical complex, are associated with MLID [251]. Given the multiple epigenetic modifiers involved in establishing and maintaining genomic imprinting, it is likely that other epigenetic regulators (such as ZNF445, DNMT1, UHRF1, SETDB1, or TRIM28) could be found to be involved in the DNA maintenance defects characteristic of MLID.

19 Imprinted Genes and More Common Complications of Pregnancy

Imprinting disorders are relatively rare (Table 8.1) but low birth weight (LBW), defined by The World Health Organisation as birth weight <2500 g at any gestational age (United Nations Children’s Fund and World Health Organisation 2004), is one of the most common complications of pregnancy, affecting up to 19% of all births in the developing world and between 5 and 7.5% of births in developed countries [252]. Several genes within the 11p15.5 IC1/IC2 domain are known to play an important role in determining birth weight in mice, either by intrinsically regulating fetal growth potential or through regulating placental development to extrinsically impact fetal growth (Fig. 8.4). Alterations in the expression of IC1/IC2 domain genes have been reported in association with fetal growth disorders in human pregnancy. Placental expression of IGF2 generally positively correlates with birthweight consistent with studies in mice [91, 253]. For example, elevated placental IGF2 has been reported in large for gestational age infants [254, 255] and lower placental IGF2 in small for gestational age infants [256, 257]. Conversely, placental expression of CDKN1C generally negatively correlates with birthweight [234, 235, 256, 258,259,260], consistent with the finding that birth weight decreases with increasing levels of Cdkn1c in mice [176, 206]. IGF2 and CDKN1C are not the only fetal growth restriction genes located on human chromosome 11p15. In mice elevated Phlda2 drives late fetal growth restriction resulting in LBW followed by catch-up growth [182]. Abnormally elevated placental PHLDA2 is a highly common alteration linked to LBW [261] with a prevalence that may be as high as 25% in confirmed cases of fetal growth restriction [262]. Increased placental expression of both PHLDA2 and CDKN1C has been observed in small-for-gestational-age (SGA) infants, highlighting the potential for domain-wide loss of imprinting [256].

In addition to regulating fetal growth, imprinted genes within the IC1/IC2 domain are critically important for placental development in mice, involved in both nutrient transport and the regulation of placental hormone production. Consequently, aberrant expression of imprinted genes in the fetally-derived placenta may contribute to disorders impacting the mother during pregnancy. For example, loss of function of Cdkn1c in the offspring has been linked to preeclampsia-like symptoms in the mouse dams [174], consistent with the finding of HELLP syndrome in human mothers of BWS infants with maternally inherited CDKN1C mutations [213]. Similarly in mice increased expression of fetal/placental Igf2 mediated by disrupted IC1 domain imprinting impacts maternal glucose management with potential relevance to gestational diabetes [263]. Changes in the expression of Phlda2 in the mouse placenta have been linked to alterations in the behavior of mothers toward their newborns [154] with potential relevance to pregnancy-related mood disorders [264].

20 Imprinted Gene Networks

In addition to the domain-wide regulation of allelic expression, some imprinted genes have been observed to interact within imprinted gene networks (IGNs) with important roles in development [265]. A number of imprinted genes located in different domains possess a strikingly similar pattern of expression suggesting interactions, with Cdkn1c and Plagl1 (aka Zac1) providing a key early example [243]. The interaction between imprinted genes in networks was further highlighted through the bioinformatic analysis of genes co-expressed with Plagl1 [266]. Disruption of these networks can result from changes in the expression of individual imprinted genes as shown for H19 [267] or disruption of epigenetic regulators, as shown for Trim28 [268]. To add further complexity, mutations disrupting networks can have more than one phenotypic outcome as a consequence of IGNs adjusting stochastically. Bimodal phenotypes have been reported for a number of mouse models in which individual imprinted genes have been genetically modified, further exemplified by studies on Trim28 mutants, where haploinsufficiency disrupted IGNs in mice manifesting an obese phenotype but not those of normal weight [268]. Evidence for similar stratification of IGNs in human populations was also reported in this study. Disrupted expression of IGNs may therefore contribute to the considerable clinical overlap noted for several imprinting disorders [210]. Disruptions to IGNs may contribute to a wide range of common human diseases (Fig. 8.4).

21 Imprinted Genes and Early Life Adversity

Imprinted genes may respond to external exposures at the individual, domain-wide, or IGN level contributing to human disease. Early life adversity is a major threat to human health because of the harmful outcomes for offspring in the short and longer term, a phenomenon described as fetal programming or developmental origins of health and disease [269, 270]. The clear role of imprinted genes in regulating fetal growth and postnatal phenotypes along with the predicted flexibility of epigenetic marks involved in the establishment, spreading, and maintenance of their expression has led to many researchers asking whether imprinted genes respond to adversity. Such studies are dependent on the demonstration of loss of monoallelic expression mediated by changes in epigenetics marks which can be challenging to demonstrate. Exposure to a maternal low-protein diet drives persistent loss-of-imprinting of at least one gene within the IC2 domain [20]. In this study, Cdkn1c luciferase reporter mice were used in which gene expression is visualized as a bioluminescent signal in living mice. Exposure to low-protein diet from conception resulted in loss of normal paternal silencing alongside loss of DNA methylation at the somatic Cdkn1c-sDMR but without impacting methylation of the IC2 domain gDMR, consistent with previous studies reporting that DNA methylation at gDMRs is relatively resistant to dietary adversities [271, 272]. Expression and methylation changes occurred before birth and persisted into adulthood even under conditions of a normal diet. Critically, supplementation with the methyl donor folate in pregnancy prevented loss of silencing. While this study provides definitive experimental evidence that imprinted genes can respond epigenetically to a dietary exposure, it remains to be determined how many genes are responsive, under what adversity conditions and to what extent the mechanisms underpinning responsiveness are conserved. Given that different mechanisms lead to the establishment and maintenance of imprinting, it seems likely that multiple pathways could disrupt the epigenetic regulation of imprinted loci. Outcomes of individual adversities may, however, be similar due to disruptions in IGNs.

22 Conclusion

Imprinted genes are associated with rare imprinting disorders such as BWS and SRS, but animal models suggest that imprinted genes have a greater potential to contribute more widely to human diseases including low birth weight, and chronic health conditions such as obesity, type 2 diabetes, and mental health disorders (Fig. 8.4). Imprinted genes may also, at least in part, be associated with the well-established relationship between early life adversity and human disease. A major challenge is how to translate functional outcomes from genetically modified mice to humans. Many studies have focused on comparing DNA methylation levels at gDMRs as a proxy for imprinted gene expression levels. However, the majority of imprinted genes are not directly spanned by gDMRs, and multistep epigenetic processes are required to establish and maintain their monoallelic expression. Developments in genome-wide approaches, single-cell technologies, combined with mathematical modeling are required to fully establish the extent to which variation in the expression of imprinted genes might contribute to wide-ranging human disease.