Trim28 Is Required for Epigenetic Stability During Mouse Oocyte to Embryo Transition

See allHide authors and affiliations

Science  23 Mar 2012:
Vol. 335, Issue 6075, pp. 1499-1502
DOI: 10.1126/science.1216154


Phenotypic variability in genetic disease is usually attributed to genetic background variation or environmental influence. Here, we show that deletion of a single gene, Trim28 (Kap1 or Tif1β), from the maternal germ line alone, on an otherwise identical genetic background, results in severe phenotypic and epigenetic variability that leads to embryonic lethality. We identify early and minute epigenetic variations in blastomeres of the preimplantation embryo of these animals, suggesting that the embryonic lethality may result from the misregulation of genomic imprinting in mice lacking maternal Trim28. Our results reveal the long-range effects of a maternal gene deletion on epigenetic memory and illustrate the delicate equilibrium of maternal and zygotic factors during nuclear reprogramming.

Imprinting, the process resulting in gene expression specific to the parent of origin, is based on differential epigenetic modifications on chromosomes inherited from the father and mother. Imprinted gene loci are marked by differentially methylated regions (DMRs), disruptions of which can cause human defects or cancer (1). For instance, aberrant expression from the Igf2/H19 cluster is involved in multiple cancers, the dwarfism Silver-Russell syndrome and the overgrowth Beckwith-Weidemann syndrome (25).

Protection of the inherited, germ line–derived methylation at imprinted loci is vital, especially during the oocyte-to-embryo transition (OET) when egg- and sperm-derived genomes undergo extensive epigenetic reprogramming to a totipotent state. Demethylation of the paternal genome is a key event during OET (6); however, it is incomplete, as germline imprints and some other sequences retain their methylation state, which ensures inheritance from germ line to soma (7, 8). Proteins such as PGC7 are known to protect the maternal genome from active demethylation during OET and to prevent DMR demethylation at several imprinted regions (9). Maternal-zygotic deletion of the Krüppel-associated box domain (KRAB)–zinc finger protein ZFP57 also results in loss of methylation at multiple imprinted loci, yet maternal deletion alone creates no phenotype in embryos, owing to paternal gene rescue (10).

A cohort of KRAB–zinc finger proteins mediate the interaction of TRIM28 (KAP1 or TIF1β), the central component of an epigenetic modifier complex, with specific genomic loci. TRIM28, in turn, recruits chromatin-modification and remodeling factors that are associated with the formation of repressive chromatin (11). We show that loss of maternal Trim28 alone results in a highly pleiotropic, 100% lethal phenotype and demonstrate its requirement for maintaining genomic imprints and supporting a proper epigenetic environment during the OET.

Trim28 is highly expressed in oocytes and early embryos (Fig. 1, A and B). To address its maternal function, we used a Zp3-Cre mating scheme (fig. S1) to delete Trim28 from oocytes, which were then fertilized by wild-type males. Embryos derived from this mating lack both Trim28 RNA and protein until transcription from the paternal allele ensues after zygotic gene activation (ZGA) at the early two-cell stage. TRIM28 protein becomes detectable from the four-cell stage onwards.

Fig. 1

Expression and maternal deletion of Trim28. (A) Immunofluorescence using antibody against TRIM28 (green) and phalloidin (red). Fully grown oocytes, two-cell–, four-cell–, and eight-cell–stage control and maternal mutant embryos are shown. Scale bar, 50 μm. (B) QPCR performed on control and maternal mutant oocytes and different preimplantation stage embryos (FGO, fully-grown oocyte; MII, metaphase II oocyte). (C) Quantification of growth restriction in maternal Trim28 mutants compared with controls by measuring dry weight (mg) (Student’s t test; *P < 0.05). (D) Mutant E15.5 embryos displaying large array of phenotypes and growth defects compared with a control embryo. Scale bar, 5 mm.

Despite normal development to the blastocyst stage, viable offspring were never found. Detailed analysis at any postimplantation stage showed a highly pleiotropic phenotype, with 40 to 70% of the embryos being resorbed, which indicated partial postimplantation embryonic loss (table S1). Embryos surviving gastrulation undergo many normal aspects of embryonic development but never live after birth. Most are significantly growth restricted (Fig. 1C) and display great phenotypic variability, including edemas, craniofacial malformations, hemorrhage, and complete and hemi-anopthalmia (Fig. 1D). Often, such phenotypic variability is attributed to segregation of genetic modifiers on mixed genetic backgrounds, yet we backcrossed the Trim28 floxed allele mice into C57BL/6J and confirmed the purity of their background by single-nucleotide polymorphism (SNP) genotyping (table S2). Hence, all experimental embryos are genetically identical. Zygotic heterozygous Trim28 embryos do not display haploinsufficiency (12, 13), which establishes that maternal Trim28 is fundamental to proper embryonic development. Yet, its loss causes phenotypes long after OET and ZGA.

To address the molecular consequences of maternal Trim28 deletion, we performed microarray analysis and found Igf2 down-regulated in four of six mutants on embryonic day 12.5 (E12.5). Igf2as is also repressed in the affected embryos, whereas H19 is up-regulated (Fig. 2A and fig. S2, A to C). All three genes are members of the imprinted H19/Igf2 cluster, whose germline DMR (H19 DMR) is paternally methylated, which promotes Igf2/Igf2as expression from the paternal and H19 expression from the maternal allele (1419) (Fig. 2B). Paternally inherited deletion of, or mutations and/or epimutations in, the H19 DMR causes loss of Igf2 expression and up-regulation of H19 (20, 21), as observed in maternal Trim28-null embryos.

Fig. 2

Expression and methylation state of imprinted genes in maternal Trim28 mutants. (A) Heat map and clustering of selected genes from two imprinted loci across six maternal mutant (M1 to M6) and four control (C1 to C4) embryos analyzed by microarray. (B) Schematic representation of the H19/Igf2 imprinted locus and its regulation on maternal and paternal inherited alleles, respectively. (C to G) Methylation levels of DMRs examined by quantitative pyrosequencing after bisulfite conversion. Each dot represents an individual fetus (round, mutant; square, control). Analysis of the H19 germline (C), H19 promoter (D), IG- (E), Snrpn (F), and Peg3 DMRs (G) is shown. Bar indicates median in each group.

We determined the methylation state of the H19 DMR by quantitative pyrosequencing after hydrogen sulfite (bisulfite) conversion. All control embryos show 50% methylation at the DMR (Fig. 2C). However in maternal Trim28 mutants, H19 DMR methylation varies between 50% and almost complete absence (Fig. 2C), a finding confirmed by combined bisulfite restriction analysis (COBRA) and by cloning and sequencing (fig. S3, A and B). Notably, the H19 DMR methylation state correlates tightly with the Igf2 expression levels in each of the respective mutants (fig. S3, C to E). Similar observations were also made at later stages (fig. S4). A secondary DMR in the H19 promoter is established during preimplantation development and is subject to the germline H19 DMR methylation state (14, 22). Methylation levels of both DMRs correlate in all embryos (Fig. 2D and fig. S3, D and E). These findings show that after fertilization, maternal TRIM28 protects the H19 DMR on the paternal chromosome from aberrant DNA demethylation.

To determine whether the effect is H19/Igf2-specific, we examined other imprinted genes. The intergenic (IG)–DMR is also paternally methylated but responsible for Dlk1/Dio3 cluster regulation. Microarray-detected expression of genes in the Dlk1/Dio3 imprinted cluster shows only marginal changes in the maternal Trim28-deleted embryos tested. Accordingly, differential methylation of the IG-DMR is just slightly affected (1 out of 16) (Fig. 2E). The Prader-Willi/Angelman syndrome cluster, normally methylated in the maternal germ line, contains a number of paternally expressed genes, such as Snrpn, Mkrn3, Magel2, Ndn, and Peg12 (23). These are close to twofold up-regulated in one of six mutants (Fig. 2A and fig. S2, D to F). Hypomethylation at the Snrpn DMR can be detected in maternal mutants, albeit to a lower extent and frequency than at the H19 DMR (Fig. 2F). The maternally methylated Peg3 DMR is unaffected in embryos surveyed at E12.5 (Fig. 2G). Thus, correlating with the pleiotropic phenotype, we find a loss of methylation at DMRs, which is highly variable both between and within individual embryos (fig. S5). Notably, many embryos do not show demethylation of all DMRs at a given locus, which implies that they are chimeric for normal and aberrantly imprinted cells, a condition likely contributing to the highly variable phenotype.

To ascertain a relation between TRIM28, ZFP57, and the DMRs, we performed chromatin immunoprecipitations (ChIPs) followed by quantitative real-time fluorescence polymerase chain reaction QPCR in wild-type E12.5 embryos with normal methylation levels (Fig. 3 and fig. S6). TRIM28 and ZFP57 are enriched at the H19 DMR, which indicates interaction. SETDB1, a component of the TRIM28 complex, catalyzes trimethylation of lysine residue 9 on histone 3 (H3K9me3), which is also significantly enriched at the H19 DMR. Extending this analysis to other imprinted regions, we also find TRIM28, ZFP57, and H3K9me3 enriched at the Snrpn, IG-, and Peg3 DMR. In conclusion, the TRIM28 complex binds all tested DMRs, including those whose methylation state we find only slightly affected or unaffected by the loss of maternal Trim28. To address binding of TRIM28 to the H19 DMR in maternal mutants, we queried seven individual E12.5 embryos. TRIM28 was enriched at the H19 DMR in four mutants, which show normal H19 DMR methylation, whereas no enrichment was detected in three others, which show severe hypomethylation (Fig. 3, E and F). Therefore, absence of maternal Trim28 results in loss of DNA methylation and lack of paternal TRIM28 binding at later stages. This suggests that DNA methylation is required at the DMR for effective binding of the TRIM28 modifier complex, an observation further supported by recent ZFP57-binding studies in embryonic stem cells (24).

Fig. 3

TRIM28 binding to DMRs is methylation dependent. (A to D) ChIP shows enrichment of TRIM28, ZFP57, and H3K9me3 at the H19 (A), Peg3 (B), Snrpn (C), and IG-DMR (D), respectively. Depicted is the average enrichment over input in percent, with or without antibody, from five (TRIM28 and H3K9me3) or three (ZFP57) individual control E12.5 embryos (*P < 0.05). (E) Interaction of TRIM28 with the H19 DMR in individual maternal mutants analyzed by ChIP. (F) DNA methylation analysis of the H19 DMR by quantitative pyrosequencing of individual maternal mutants used for ChIP in (E). (Pos., position of CpG-island within the amplicon)

Our evidence suggests that maternal Trim28 prevents DNA demethylation during OET, yet the global methylation state in mutant zygotes is not altered according to 5-methyl or 5-hydroxymethyl cytosine (9, 25) immunofluorescence staining (fig. S7). We next addressed specifically DMR methylation in preimplantation embryos. We chose the H19 DMR as the candidate most likely to display hypomethylation in pools of embryos, as it is most frequently affected in survivors of the first wave of lethality, at implantation. As H19 DMR is paternally imprinted, it is not methylated in oocytes, but, at the two-cell stage, controls show methylation at about half the alleles (Fig. 4). In mutants, however, H19 DMR methylation is reduced, a loss even more pronounced at the four- and eight-cell stages. Conversely, the Peg3 DMR is maternally imprinted and fully methylated in control and mutant eggs (fig. S8). Thus, Trim28 is not required to maintain this maternal imprint in the growing egg. After fertilization, the paternal, unmethylated allele is detectable, yet overall Peg3 DMR methylation is not visibly affected in mutant two-cell– or eight-cell–stage embryos.

Fig. 4

Loss of H19 DMR methylation in mutant preimplantation embryos. Combined bisulfite restriction analysis of the H19 DMR methylation state in pooled (A) two-cell–, (B) four-cell–, and (C) eight-cell–stage controls and maternal mutants (UN, uncut; D, Dra I; B, Bst UI). Dra I digestion tests the efficiency of bisulfite conversion; the restriction site for Bst UI is protected from bisulfite mutagenesis if the CpG nucleotides within the site are methylated (u, unmethylated; m, methylated). Two individually analyzed embryo pools are shown for mutants and controls, respectively. (D) Model of TRIM28 function after fertilization and consequences of the maternal deletion. Symbols represent individual methylated (filled) or unmethylated (unfilled) DMRs. Colored outlines represent protective TRIM28 complexes at DMRs containing maternal (green) and later zygotic (red) TRIM28.

It is unclear why imprinted loci are not equally affected. Redundant mechanisms such as PGC7-mediated protection or locus-specific accessibility for modifiers could play potential roles. Additionally, early lethality at implantation and pooling of embryos at preimplantation stages could mask the detection of more affected DMRs. Indeed, when we analyzed individual E4.5 embryos for both H19 and Peg3 DMR methylation, we also found infrequent, yet complete, Peg3 DMR hypomethylation (fig. S9). Such embryos appear unlikely to survive beyond implantation, as we detect only unaffected Peg3 DMRs at later stages.

In maternal Trim28 mutant embryos, demethylation of DMRs occurs slowly and inefficiently, although it still creates sufficient impact to cause lethality. This delay may be mediated by factors such as PGC7 and/or ZFP57 playing additional protective roles. This creates an environment in which paternal TRIM28, if present above a given threshold, can prevent complete DMR demethylation. However, once demethylated, the TRIM28 complex can no longer bind and recover this loss (see ChIP in Fig. 3E). Therefore, the molecular and phenotypic variability in maternal Trim28-deleted mutants emerges from a random combination of stochastically affected TRIM28 target loci (26). Because of ongoing cell division, this can further vary between blastomeres within an individual embryo. As a result, the embryo is a mosaic in which the degree of mosaicism and extent of gene disregulation determines the time and mode of embryonic and fetal lethality (Fig. 4D). Our findings illustrate the exquisite temporal and spatial balance between maternal and zygotic factors in the early embryo that are required to maintain epigenetic states. Perturbing this balance can cause a wide spectrum of phenotypic variability.

Supporting Online Material

Materials and Methods

Figs. S1 to S9

Tables S1 and S2

References (2729)

Array Data

  • * No longer at Singapore Institute for Clinical Sciences.

References and Notes

  1. Acknowledgments: This research was financed by A*STAR, Singapore and by grants from the Wellcome Trust and UK Technology Strategy Board. We thank H. Wollmann, S. Balu, D. Tham, and K. Yamazawa for their help. Microarray data and accession nos. can be found in the Supporting Online Material.
View Abstract

Stay Connected to Science

Navigate This Article