Role of transcription in imprint establishment in the male and female germ lines

Imprinted genes in the soma follow the instructions that arise in the male or female germline

A recent article reports the role of transcription in the establishment of paternal genomic imprinting [1]. This prompts the authors to discuss the germline-specific transcription events that mark the chromosomes at imprinted regions as ‘paternal’ or ‘maternal’.

Imprinted genes are characterized by their monoallelic expression pattern from one parental chromosome only, the one inherited from the father or from the mother [2,3]. In the soma, parental allele-specific expression of imprinted genes in an imprinted domain, is controlled by a shared imprinting control region (ICR; Figure 1A). The function of the ICR (as a promoter or an insulator) is restricted to one parental chromosome by DNA methylation in the other parental allele. The gametic mark, with some exceptions, consists of DNA methylation at paternal (PAT) or maternal (MAT) germline differentially methylated regions (gDMRs; Figure 1B). At fertilization, gDMRs are carried into the zygote in the methylated form by the sperm or the oocyte, respectively, and are subsequently maintained methylated through cell divisions in somatic cells in the paternally or maternally inherited chromosome. Maintenance of DNA methylation at the gDMRs requires the action of ZF protein factors [4–7]. The gDMR is critical in setting up secondary somatic DMRs and is required for parental allele-specific expression of the imprinted genes in the domain.

The gametic marks at gDMRs are the products of the male and female germlines. Both germlines initiate as primordial germ cells, which carry both PAT and MAT methylation at gDMRs and exhibit allele-specific expression similar to somatic cells [8]. The imprint marks are erased in primordial germ cells [8,9] before the new gamete-specific marks are established in the germ cells according to the gonad's sex (Figure 2). Due to the erasure and reestablishment processes, the two alleles become functionally similar and the imprinted genes become biallelically expressed or biallelically silent in the germ cells [8,10] in both germlines. The establishment of methylation imprints at PAT and MAT gDMRs are independent and asynchronous processes. PAT gDMRs are established in mitotically arrested germ cells of the male fetus (prospermatogonia) [11–13], whereas MAT gDMRs are established after birth in the growing oocytes of the juvenile female [14–16].

The molecular machinery that sets up the new imprints includes enzymes that work on DNA and histones. DNMT3A, and its cofactor, DNMT3L, carry out de novo methylation at imprinted gDMRs in both male and female germlines [17–22]. De novo methylation occurs in prospermatogonia at sites of unmethylated or diminishing H3K4m2/3 [23,24] and the same is true in growing oocytes [25]. This requires demethylating H3K4me2/3 in the germ cells before DNA methylation can take place. The H3K4 demethylase KDM1B is required for establishing MAT gDMRs in growing oocytes [25,26] by removing methyl groups from H3K4, in turn allowing the interaction with the ADD domain of DNMT3L [27], and de novo methylation activity at those DNA sequences. PAT gDMR methylation is slightly delayed by active chromatin marks in the allele inherited from the mother [12]. MAT gDMR establishment requires histone deacetylation by HDAC1/HDAC2 [28]. MAT gDMR establishment also requires SETD2 H3K36me3 methyltransferase in growing oocytes [29]. H3K36me3 attracts the PWWP domain of DNMT3B [30,31], and likely also of DNMT3A. The establishment of PAT gDMRs does not require SETD2, but it requires NSD1, an H3K36me2 methyltransferase [32].

The male or female gamete-specific marking of PAT and MAT gDMRs in the male versus female germlines is the key to genomic imprinting. One important question is whether there is a general mechanism that specifies the location for all MAT gDMRs versus all PAT gDMRs in the female versus male germline. It is not known why de novo DNA methylation by DNMT3A and its cofactor, DNMT3L, mark MAT DMRs in the female germline but PAT DMRs in the male germline, even though these enzymes do not distinguish DNA sequences. The histone-modifying enzymes KDM1B, HDAC1/2, SETD2 and NSD1 do not have sequence specificity, either. One part of the Rasgrf1 gDMR has a special piRNA-mediated mechanism for methylation establishment [33] and it is methylated by DNMT3C [34]. piRNA-mediated de novo DNA methylation is sequence-specific targeting but is not a general mechanism for all PAT gDMRs.

Transcription across MAT gDMRs facilitates MAT imprints in oocytes

The genomic locus-specificity of MAT gDMR establishment is provided by transcription events in the germline. Transcription runs across MAT gDMRs in growing oocytes [16,35]. These transcripts are often initiated in alternative oocyte-specific 5′ promoters (Figure 3A) [21,36] or in retroviral elements [37]. Mouse genetic studies have revealed that transcription across MAT gDMRs in growing oocytes is required for the establishment of imprinting in the female germline [35–39]. Proper gDMR methylation establishment in the oocyte ensures that the MAT chromosome is methylated while the PAT chromosome is unmethylated in the soma (Figure 3B). MAT gDMRs function as promoters. In certain imprinted domains, a noncoding transcript expressed from a MAT gDMR exclusively in the PAT unmethylated allele regulates other imprinted genes; it suppresses genes in the PAT chromosome, while the genes in the MAT chromosome are allowed to be transcribed (Figure 3B). DNA methylation fails to occur in growing oocytes when the oocyte-specific transcript is truncated before entering the MAT gDMR [35,38,39] or when the oocyte-specific promoter is deleted [36,37]. Imprinted MAT gDMR establishment, therefore, depends on the transcript from the oocyte-specific upstream alternative promoter (Figure 3C). In response to failure to establish MAT gDMR methylation, imprinted expression in the domain is lost in somatic cells (Figure 3D). SETD2 and H3K36me3 are also required for de novo MAT gDMR methylation. Transcription across MAT gDMRs in growing oocytes correlates with H3K36me3 [37] and facilitates H3K36me3 [36], suggesting that transcription is the primary event and H3K36me3 is the secondary in recruiting the DNA methyltransferase to MAT gDMR sequences. Specific examples follow when the role of transcription in growing oocytes was genetically tested at MAT gDMRs.

Transcripts were detected traversing several MAT gDMRs of imprinted loci (Gnas, Igf2r, Grb10, Kcnq1, Zac1 and Impact) in growing oocytes prior to and at the time of de novo methylation in the female germline [35]. In these cells, transcription did not occur at the control H19/Igf2 PAT gDMR, which does not attain methylation in oocytes but does in prospermatogonia. The role of transcription in establishing maternal germline methylation marks was proposed and directly tested by Chotalia and colleagues at the Gnas locus. When the authors truncated the Nesp transcript, which runs across the domain in growing oocytes, they observed a disruption in methylation establishment at the MAT gDMRs (Nespas/Gnasxl and 1A) of this domain in the oocyte. In response, there was loss of methylation at the two MAT gDMRS in the F1 neonatal brain and misregulation of allele-specific expression of imprinted genes along the domain in the brain and liver of newborn pups, which had very poor postnatal survival.

Joh and colleagues showed that methylation establishment failed at the Zrsr1 MAT gDMR in the absence of Commd1 transcription through the gDMR in growing oocytes when a poly(A) signal cassette was inserted between the Commd1 promoter and the Zrsr1 gene [38]. The Zrsr1 gene was twofold upregulated and biallelically expressed in the brain of adult mice that inherited the MAT gDMR hypomethylation from the oocyte. The gDMR hypomethylation effect was not reversible after it had occurred in the oocyte even when the RNA truncation cassette was removed subsequently in the zygote. These results importantly showed that transcription-dependent methylation at the MAT gDMR occurs specifically in the growing oocyte, but not during early development or postnatal growth.

Singh and colleagues generated a truncation of the Kcnq1 transcript to eliminate transcription in oocytes across the KvDMR [39]. They found that methylation establishment at this MAT gDMR was abolished in the growing oocytes. When the hypomethylation was passed from the female germline into the embryos, this resulted in biallelic Kcnqot1 ncRNA expression and the repression of maternally expressed genes, such as Cdkn1c and Ascl2, in this imprinted domain.

Smith and colleagues [40] used BAC transgenes in mice containing 100 kb of sequence upstream of Snrpn exon 1, which directed the appropriate imprinted expression of Snrpn and was sufficient to direct appropriate methylation of the Snrpn MAT gDMR following maternal but not paternal transmission. This region contains alternative U exons, which are expressed in oocytes and splice into Snrpn exon 2. The U exon transcription is found prior to DNA methylation imprint establishment in the growing oocyte, but it is undetectable at an earlier time when fetal oogonia are entering meiotic prophase. Deletion of the U exons from the BAC transgene resulted in DNA hypomethylation at the MAT gDMR and biallelic Snrpn expression. Interestingly, an additional transgenic line lacked the U exons but still displayed maternal silencing. In this line, transcription initiated upstream of the transgene insertion site through the MAT gDMR in juvenile ovaries, suggesting that whereas transcription across the gDMR is essential, it is not necessary to occur from the U exons.

Veselovska and colleagues carried out deep RNA sequencing in oocytes and found that transcription, often from alternative, oocyte-specific promoters that may originate in retroviral elements, correlates with DNA methylation and accounts for up to 90% of the methylome. Deletion of the alternative, oocyte-specific promoter (Zac1o), eliminated transcription across the Zac1 MAT gDMR in growing oocytes [36]. This resulted in failure of DNA methylation establishment in the female germline with consequences to allele-specific DNA methylation and gene expression in the offspring: the Zac1 gDMR, which serves as somatic promoter to Zac1, was hypomethylated and Zac1 was twofold upregulated and biallelically expressed in all neonatal tissues.

Bogutz and colleagues proposed that oocyte transcription initiating in lineage-specific endogenous retroviruses is likely responsible for methylation establishment at 4/6 mouse-specific and 17/110 human-specific imprinted gDMRs [37]. Indeed, when they deleted the murine-specific endogenous retroviruses (ERV) upstream of Impact or Slc38a4, they found that MAT gDMR establishment failed in the oocyte. Corresponding aberrant biallelic expression of Impact and Slc38a4 was found in the fetuses that inherited the epimutation from the oocyte.

Transcription across a PAT gDMR facilitates imprint establishment in prospermatogonia

Very deep RNA sequencing experiments [23] revealed that broad, low-level transcription runs across all three known PAT gDMRs specifically in 15.5 dpc prospermatogonia (Figure 4), where de novo DNA methylation is initiated [11,41]. Transcription along PAT gDMRs was also detected by reverse transcription PCR, but the broad extent and very low level of these transcripts were not recognized at that time [24]. The role of the broad, low-level transcript in PAT gDMR establishment was tested at the H19/Igf2 ICR using mouse genetics (Figure 5A) [1]. The reciprocal maternal and paternal allele-specific expression of the H19 and Igf2 genes in mouse chromosome 7 depends on the ICR, which behaves as a maternal allele-specific CTCF-mediated insulator in normal somatic cells (Figure 5B) [42]. When the broad, low-level transcript was truncated at the entry point to the ICR, this resulted in hypomethylation of the gDMR in prospermatogonia (Figure 5C) [1]. When the hypomethylated gDMR was inherited to the offspring in the PAT chromosome, it caused the misregulation of parental allele-specific gene expression along the imprinted domain (Figure 5D). These findings suggest that broad, low-level transcription across the H19/Igf2 PAT gDMR serves as a locus-specific molecular mechanism for imprint establishment in the male germline.

Broad, low-level transcription across PAT gDMRs could be a unifying molecular mechanism for imprint establishment in the male germline

The intergenic Dlk1-Dio3 DMR in mouse chromosome 12 is a PAT gDMR [11,41,43,44] and it is located between the protein-coding, paternally expressed Dlk1 gene and the maternally expressed lncRNA Gtl2 (Meg3) and controls the allele-specific expression along the entire domain all the way to the Dio3 gene. Mouse genetic [45,46] and epigenetic editing studies in cell culture [47,48] characterized the roles of the bipartite intergenic Dlk1-Dio3 DMR. One side of the DMR, a CpG-rich region, carries the DNA methylation mark from the sperm and determines the secondary somatic Gtl2 DMR methylation and silencing at the Gtl2 promoter in the paternal allele. The embedded tandem repeats maintain methylation in the paternal allele in the soma [46] through binding of ZFP57 [4]. The other side of the DMR maintains the unmethylated and active state of the Gtl2 promoter in the maternal allele. It is not known what specifies gDMR establishment for the intergenic Dlk1-Dio3 DMR in the male germline.

The third known PAT gDMR regulates Rasgrf1, in chromosome 9 [49]. In neonatal brain cells, DMR methylation is required for the expression of the active Rasgrf1 paternal allele by inhibiting CTCF insulation [50]. A tandem repeat is required for establishing methylation of this PAT gDMR in the male germline and protecting it in the preimplantation stage embryo, but not for PAT DMR maintenance later in the soma [51]. The Rasgrf1 DMR is unique among PAT gDMRs in that it is methylated by DNMT3B in addition to requiring DNMT3A and DNMT3L in prospermatogonia [11]. It is also methylated by DNMT3C [34]. De novo DNA methylation of this gDMR involves the piRNA pathway, and it is targeted at an RMER4B repeat element at two piRNA sites [33]. The targeting piRNAs originate in chromosome 7 in another RMER4B element. The MILI and/or MIWI2–piRNA complex targets the nascent pit-RNA that starts out of the tandem repeat. Removing the above RMER4B sequences in chromosome 9 or chromosome 7 affects DNA methylation in one part of the DMR, but hypomethylation is not inherited to the offspring [52]. One part of the Rasgrf1 gDMR depends on NSD1-mediated H3K36me2 marks [32]. Additional mechanisms must be in place at full gDMR establishment.

Similar to the H19/Igf2 gDMR, RNA deep sequencing experiments detected specific transcription through these PAT gDMRs in prospermatogonia along the Dlk1/Dio3 and Rasgrf1 loci at the time when de novo methylation is initiated [23]. Based on the role of transcription at the H19/Igf2 gDMR, the authors propose that broad, low-level transcription across PAT gDMRs could be a unifying molecular mechanism for imprint establishment in the male germline. It is likely required for the establishment of the Dlk1/Dio3 and the full Rasgrf1 PAT gDMR in prospermatogonia.

Conclusion

Transcription across gDMRs in both female and male germlines is emerging as the key unifying mechanism for imprint establishment. However, the transcripts that run across imprinted MAT versus PAT gDMRs are specific to the germline of the female versus male individuals. Importantly, the timing of transcription across MAT versus PAT gDMRs is asynchronous; it matches the time and cell type of de novo methylation in the female versus male germline being exclusive to growing oocytes of postnatal juvenile female, versus mitotically arrested germ cells of the male fetus, respectively. In addition, the genomic location and the nature of the transcripts that run across MAT gDMRs in growing oocytes (strong transcript of a coding gene from alternative upstream promoter) are different from the transcripts that cross PAT gDMRs in prospermatogonia (broad, low-level transcription). All these differences in transcription events can explain the difference in targeting de novo DNA methylation to PAT versus MAT imprinted gDMRs in the male versus female germ cells.

Future perspective

Mouse genetic studies will test whether transcription across the Dlk1/Dio3 and the Rasgrf1 PAT gDMR specifies de novo methylation in prospermatogonia for the establishment of PAT imprint at those imprinting control regions. The mechanism of how broad, low-level transcription across PAT gDMRs instructs DNA methyltransferase activity will be revealed. The knowledge on imprinting mechanisms will facilitate better management or possibly the prevention of imprinting diseases.