Imprinted silencing is extended over broad chromosomal domains in mouse extra-embryonic lineages.

Gene silencing in imprinted gene clusters is established by an epigenetic initiator that is often a long non-coding (lnc) RNA. The clustered organization of known imprinted genes indicates that the initiator extends imprinted silencing over broader chromosomal domains in extra-embryonic lineages compared to the embryo. We propose that extension of imprinted gene clusters may result from known epigenetic differences between extra-embryonic and embryonic lineages that alter the behavior of epigenetic initiators. New RNA sequencing technology will enable the full extent of imprinted silencing in embryonic and extra-embryonic lineages to be defined, but appropriate analysis and cell systems are required, which we define here based on a review of recent studies.

Current Opinion in Cell Biology 2013, 25:297–304

This review comes from a themed issue on Cell nucleus

Edited by Edith Heard and Danesh Moazed

For a complete overview see the and the

Available online 13th March 2013

0955-0674/$ – see front matter, © 2013 Elsevier Ltd. All rights reserved.

Introduction

Mammalian development requires both maternal and paternally inherited genomes due to the presence of imprinted genes that are expressed from only one parental allele [1,2,3]. Imprinted expression is regulated by genomic imprinting, a classic example of epigenetics defined as a heritable change in gene expression caused by mechanisms other than changes in the underlying DNA sequence and widely considered to be based on chemical modifications of DNA or associated molecules [5]. Epigenetic regulation is likely to be a stepwise process where an initial epigenetic signal is recognized by an initiator [6] that then targets changes in gene expression to specific loci. This gene expression state could then be maintained by additional factors, enabling modulation of the epigenetic signal at different steps in the process. In genomic imprinting the initial epigenetic signal is differential DNA methylation of the imprint control element (ICE) established during gametogenesis by the DNMT3A/3L de novo DNA methyltransferase complex [7-9], and maintained on the same parental allele in all subsequent cell divisions by the DNMT1 maintenance methyltransferase [10]. Targeting of de novo DNA methylation to the ICE has been attributed to overlapping transcription at the Gnas ICE in the oocyte and to discrete sequence elements within the Igf2r ICE [11]. An ICE is associated with at least one, but usually a cluster of imprinted genes (Figure 1a) [13]. DNA methylation is regarded as a repressive epigenetic mark, but DNA methylation of the ICE is found on the same parental chromosome as the expressed imprinted protein-coding allele. This is because imprinted silencing is triggered by recognition of the unmethylated ICE by an epigenetic initiator, whereas recognition is blocked on the other allele by ICE methylation. The initiator can be a macro long non-coding (lnc) RNA whose expression is controlled by the unmethylated ICE (reviewed in [14]), or a CTCF (CCCTC-binding factor) insulator complex bound to the unmethylated ICE (reviewed in [15]). Once imprinted gene silencing is established somatic DNA methylation can be recruited and may serve to lock in the silent state; however, this is a relatively rare event [13,16]. In contrast to allele-specific somatic epigenetic modifications that are a consequence of imprinted gene silencing, allele-specific ICE methylation arises during gametogenesis and is found in all somatic cells irrespective of imprinted expression [13].

Figure 1

Tissue-specific silencing of imprinted genes. (a) Multiple-lineage (ML) imprinted silencing: the epigenetic initiator (often a macro long non-coding (lnc) RNA) recognizes the unmethylated imprint control element (ICE) and triggers imprinted silencing of sensitive genes. (b) Tissue-specific imprinted silencing. Top: Loss of differentially methylated region (DMR) by hypermethylation: the unmethylated ICE allele becomes methylated preventing recognition and silencing by the initiator. Middle: Lack of initiator: the initiator is absent in this cell type so imprinted genes are not silenced. Bottom: Altered sensitivity to the initiator: e.g. additional genes become sensitive to silencing by the initiator as occurs for extra-embryonic lineage (EXEL) specific imprinted expression. (c) A 12.5 days post coitum (dpc) mouse embryo. Left: photo with the embryo outlined and the extra-embryonic membranes and placenta highlighted. Right: cartoon of the embryo highlighting the extra-embryonic membranes, placenta and maternal contributions to the placenta from infiltrating blood vessels and the closely associated decidua. VYS visceral yolk sac; PYS parietal yolk sac.

A characteristic of imprinted genes is that they show tissue-specific and developmental stage-specific regulation of imprinted expression [17,18]. Thus although genomic imprinting is thought to affect a small subset of mammalian genes, the true number has not yet been identified as only a limited range of tissues and developmental stages have been examined [4]. In some cases tissue-specific imprinted expression results from a locus-specific change affecting only a single gene. However, in other cases a tissue-specific response is likely to be involved as genes from multiple loci are affected. Tissue variation in imprinted expression is also likely to arise from modulation of epigenetic steps subsequent to the ICE methylation, as this is a constitutive mark present in all somatic cells, at all developmental stages. Three mechanisms controlling tissue-specific imprinted expression have emerged (Figure 1b). First, imprinted expression can be lost by gaining DNA methylation on the unmethylated allele of the ICE in specific tissue types, as occurs for Dlk1 and Cdh15 [19]. Widespread de novo methylation during gametogenesis leads to a large number of differentially methylated regions (DMRs) in the preimplantation embryo [20], most of which are lost by the postimplantation stage due to a gain of methylation on the other allele [20]. However, a distinguishing ICE feature is their protection from gain of methylation on the unmethylated allele [20], indicating that tissue-specific loss of imprinted expression by this mechanism is rare. Second, genes can escape imprinted expression because the initiator is not present in a certain tissue. For example, neuron-specific loss of Airn lncRNA leads to biallelic expression of Igf2r in neurons, and a similar mechanism may account for the loss of Ube3a imprinted expression in glial cells [21-23]. For the well-studied Igf2r, Kcnq1 and Gnas clusters the lncRNA identified as the initiator is expressed in almost all cell types, indicating that loss of imprinted expression due to absence of the initiator is also rare. Third, a change in sensitivity to the initiator can cause a loss or gain of imprinted expression in certain tissues. For example, human Igf2 loses imprinted expression in adult liver due to the use of an alternative promoter [24]. This promoter presumably also uses alternative enhancers that are not blocked by insulator activity on the maternal allele, as H19 retains imprinted expression demonstrating that the initiator (insulator complex) is present [25]. Increased sensitivity to the initiator has been suggested to explain gain of imprinted expression of flanking genes in extra-embryonic cell lineages [26] (Figure 1b). This represents a tissue-specific rather than a locus-specific mechanism and therefore is likely to affect the largest number of genes. Indeed genes showing extra-embryonic lineage (EXEL) specific imprinted expression form the majority of genes showing tissue-specific imprinted expression [26].

Epigenetic features of extra-embryonic lineages may influence imprinted expression

Extra-embryonic tissues are the placenta, umbilical cord and the yolk sacs that surround the mouse embryo and together provide a maternal interface to supply nutrients, remove waste and send signals that pattern the early postimplantation embryo (Figure 1c) [28]. Cell lineages that contribute to extra-embryonic tissues are derived at different development stages. Trophectoderm and primitive endoderm differentiate in the preimplantation embryo at 3.5 and 4.5 days post coitum (dpc) respectively and are collectively known as the extra-embryonic lineages. The trophectoderm contributes to cell lineages in the placenta and parietal yolk sac, while the primitive endoderm contributes to the endoderm layers of the parietal and visceral yolk sacs (Figure 1c). Other extra-embryonic cell types are derived from the epiblast lineage after gastrulation at 7.5 dpc in the postimplantation embryo [29]. EXEL specific imprinted expression has only been reported in the placental labyrinth and spongiotrophoblast and visceral yolk sac endoderm [26]. Extra-embryonic lineages appear to retain features of the early embryo that are lost in the epiblast lineage. For example, postfertilization the genome undergoes global DNA demethylation, although in imprinted clusters methylation of the ICE is protected, before global DNA methylation increases again in the postimplantation embryo [30]. The extra-embryonic lineages are derived before implantation and maintain a low level of DNA methylation [31,32]. In addition, the early cleavage-stage embryo undergoes imprinted X-inactivation before the paternal X becomes reactivated in the epiblast lineage and then undergoes random X-inactivation in the postimplantation embryo [33]. In contrast, the extra-embryonic lineages retain imprinted X-inactivation throughout embryonic development [34,35]. These examples show that extra-embryonic lineages retain some of the epigenetic features of the early embryo, which may relate to regulation of EXEL imprinted expression. This is supported by the example of Sfmbt2, which shows imprinted expression in the blastocyst and throughout the early postimplantation embryo that later becomes restricted to extra-embryonic lineages [36]. The repressive polycomb and EHMT2 histone methyltransferase complexes have been shown to be required to maintain silencing of some EXEL imprinted genes in placenta [37-39], but why the initiator causes silencing of these genes only in extra-embryonic lineages remains unclear. A comprehensive genome-wide examination of active and repressive histone modifications of an extra-embryonic lineage has not yet been done, but low DNA methylation is associated with open chromatin which may allow the initiator to silence more genes in extra-embryonic lineages by extending its influence over a larger chromosomal domain.

Imprinting silencing is extended in extra-embryonic lineages

Genes reported to show EXEL specific imprinted expression have been associated with five imprinted clusters (Figure 2), and also include a number of solo imprinted genes that are not associated with a known ICE [27]. In these five clusters, genes showing EXEL imprinted expression are non-randomly distributed lying further away from the ICE than other imprinted genes (Figure 2). For example, in the Igf2r cluster the Airn lncRNA expressed from the paternal allele acts as the initiator, overlapping and silencing the Igf2r gene in all cell types where it is expressed, but also silencing only in extra-embryonic lineages the Slc22a2 and Slc22a3 genes whose promoters lie between 159 and 237 kb upstream [40]. Similarly in the Kcnq1 cluster the Kcnq1ot1 lncRNA acts as the initiator silencing Kcqn1, Cdkn1c, Slc22a18, and Phlda2 in multiple cell lineages, while Tssc4, Cd81, Tspan32, Aslc2, Th, Nap1l4 and Osbpl5 that lie further away are only silenced in extra-embryonic lineages (Figure 2) [41]. The Igf2 cluster, where the initiator is a CTCF insulator complex [42], also shows this pattern with H19 and Igf2 showing imprinted expression in most cell types, while Ins2, which is more distant from the ICE, shows imprinted expression only in EXEL tissues [43]. The Grb10 and Peg10 clusters follow the same pattern with EXEL genes lying further away from the ICE than other imprinted genes, although there are some exceptions. For example, Asb4 shows imprinted expression in multiple embryonic and extra-embryonic tissues [44], but on the linear chromosome lies further from the putative ICE for the Peg10 cluster than the EXEL genes Ppp1r9a, Pon2 and Pon3 (Figure 2). However, although linear position on the chromosome may indicate proximity it is unclear how the genes are positioned relative to the ICE in three-dimensional space in the nucleus. Collectively the data from known EXEL genes in imprinted clusters indicate that imprinted silencing is extended over broader chromosome domains in extra-embryonic lineages than the embryo.

Figure 2

Imprinted gene clusters containing genes known to show extra-embryonic lineage (EXEL) specific imprinted expression. Clusters are shown to scale with a key.

Defining the extent of imprinted silencing in extra-embryonic lineages

The distribution of known EXEL imprinted genes in the genome indicates that imprinted silencing is extended in extra-embryonic lineages compared to the embryo. However, only five clusters have been shown to have EXEL genes whereas over 20 gametic DMRs that are proven or potential ICEs have been identified [20], indicating that more EXEL imprinted genes could be discovered. To determine the extent of imprinted silencing in extra-embryonic lineages a genome-wide unbiased approach is required. Imprinted expression can be directly detected by RNA sequencing (RNA-seq) of F1 crosses between genetically distant mouse strains and identifying single nucleotide polymorphisms (SNPs) that show the same parental-allele expression bias in reciprocal crosses (Figure 3a). This approach was used to investigate EXEL imprinted expression in placenta, and also to investigate tissue-specific imprinted expression in embryo and at different stages of brain development [47,48]. The placental studies and the majority of the other tissue-specific studies reported the discovery of a small number of new imprinted genes with the exception of one that reported over 1300 new imprinted transcripts in embryonic and adult brain [49]. However, two subsequent studies that repeated this work failed to confirm most of these novel imprinted transcripts, indicating that the majority were false positives, and providing lessons on the technical limitations of RNA-seq and appropriate analysis that also apply to studies to uncover the extent of EXEL imprinted expression [48].

Figure 3

Imprinted gene discovery by RNA sequencing. (a) Reciprocal crosses are made from genetically distinct strains of mice and tissue collected from F1 animals. This material is subject to RNA sequencing and parental-allele expression of a gene is determined by detecting allelic expression of single nucleotide polymorphisms (SNPs) obtained from a reference database. Imprinted expression is distinguished by showing a reciprocal bias in expression, for example, where the maternal allele is preferentially expressed in both crosses. (b) Biallelic expression of a gene in multiple cell types can obscure detection of imprinted expression in a single cell type in mixed tissues. This problem can be overcome by isolating a homogenous population of cells.

False positives in detecting imprinted expression by RNA-seq can arise from technical issues such as sequencing errors or misalignment of reads, or due to biological variation. Therefore, a robust statistical approach is required to identify high-confidence imprinted gene candidates. Statistical methods assuming random independent sampling were shown to not accurately model imprinted expression detected by RNA-seq, indicating that the false positive rate must be determined empirically to account for systematic biases [48]. To achieve this first an ‘imprinting score’ was devised that combines deviation from biallelic expression with sequencing coverage to give a statistical confidence value for imprinted expression of each SNP. Second, a false positive rate was set by determining an imprinting score at which the SNPs detected in a mock comparison between F1 crosses of the same genotype, where no imprinted expression should be detected, was 5% of the number detected in the reciprocal cross [48]. However, despite implementing this analysis pipeline only 6 of the 11 top candidates identified by RNA-seq were validated, emphasizing the importance of biological replicates and validation by independent techniques to confirm imprinted expression [48]. Supporting evidence for imprinted expression can be gained from genome-wide parental allele-specific maps of DNA methylation and histone modifications of F1 crosses [53]. The repressed allele of some imprinted genes acquire somatic DNA methylation, but as the majority do not, it is a poor predictor of imprinted expression [13,16]. However, differential H3K4me3 and H3K27ac that mark active promoters and H3K36me3 that covers the gene body of expressed genes have been shown to predict imprinted expression [53]. Current methods for direct validation of imprinted expression are all based on quantification of a reverse transcription polymerase chain reaction (RT-PCR) product containing a SNP in F1 animals, which is a laborious process if a large number of SNPs need to be validated. Restriction fragment length polymorphisms and Sanger sequencing can validate strong biases in parental allele expression, but the more sensitive quantification provided by pyrosequencing and Sequenom MassARRAY sequencing may be required to confirm subtle biases [47,48].

Determining the extent of imprinted silencing in EXEL tissues also requires an appropriate study system. Placenta contains several lineages that can show EXEL imprinted expression but may also contain lineages that do not. Thus EXEL imprinted expression may be masked or obscured by biallelic expression in other cell types within an organ (Figure 3b). Placenta also contains maternal cells from infiltrating maternal blood vessels and blood as well as the tightly associated decidua that cannot be completely removed (Figure 1c). This maternal contribution can create the false impression of maternal imprinted expression, requiring extra validation steps [26]. These problems could be overcome by isolating a pure EXEL cell population, which has only been described for the visceral yolk sac endoderm [26].

Concluding remarks

The distribution of known EXEL imprinted genes indicates that imprinted silencing by epigenetic initiators is extended in extra-embryonic lineages compared to the embryo. The most parsimonious explanation for increased sensitivity to the initiator in extra-embryonic lineages is that the initiator acts by the same mechanism as in the embryo, but is able to silence a larger chromosomal domain, perhaps due to a different chromatin environment or downstream process. For example, in the Igf2r cluster the Airn lncRNA is the initiator that silences the Igf2r promoter by transcriptional interference independent of the lncRNA product [40,56]. In contrast, in placenta the Airn RNA product recruits and targets the repressive EHMT2 H3K9me2 methyltransferase to the Slc22a3 promoter, which is necessary for its silencing [37]. These results could be reconciled by a model whereby Airn transcription initiates silencing of EXEL genes by interfering with the formation of enhancer-promoter interactions, then as a secondary step EHMT2 could maintain imprinted silencing [14]. Understanding how more genes become subject to genomic imprinting in EXEL tissues will not only offer insights into the biology of extra-embryonic lineages, but could identify epigenetic mechanisms that operate to lesser degrees in all somatic tissues.

References and recommended reading

Papers of particular interest, published within the period of review, have been highlighted as:

• of special interest

•• of outstanding interest