REVIEWS Box 2 I Transgenerational influence of epigenetic alterations in germ cells in mammals because expression levels of the imprinted genes, which include ny important developmental Recent studies have suggested that exposure to chemicals and malnutrition conditions can affect not only the children of the affected individuals, but also genes, are unbalanced in such embryos. When the their grandchildren. This might be attributable to epigenetic alterations that occur imprinted genes are appropriately modified by genetic engineering and developmental manipulation, however, disruptors, the number of spermatogenic cells decreased in the F1 generation. This it was possible to derive adult femalemice with two mater effect was transmittable through the male germ line to subsequent generations, nal genomes and no paternal complement 66(BOX 3) and this was correlated with altered DNA-methylation patterns. In another As the method involves genetically engineered animals example, exposure to methyl-donor supplementation during midgestation and highly complex nuclear-transfer technologies, affected the epigenetic s of fetal germ cells. The mouse A y gene, which its direct application to livestock seems difficult. influences coat colour, is regulated by the DNA methylation status of an intra- cisternal A particle(lAP)retrotransposon inserted at pseudoexon 1A of the gene The methyl-donor supplementation shifted the coat colour of the F2 generation to - Pigenetic silencing of retrotransposons darker one. This suggests that the methyl donor directly or indirectly affected the Only germ cells can transmit genetic information to epigenetic status of Ay in fetal germ cells. Finally, epidemiological studies have the next generation. Therefore, transposons, which indicated that grandchildren of malnourished women show low birth weight4, mobilize in the genome and might cause insertional that grandchildren of men who were well-fed before adolescence have a greater nutations, have to be strictly controlled in these cells. Approximately 40-50% of the mammalian genome is have low cardiovascular mortality In these cases, it is possible that the nutrition occupied by retrotransposons, which mobilize through status caused epigenetic alterations in germ cells, but further studies are needed an RNA intermediate, although many of them are to confirm this possibility. truncated or have accumulated mutations Mammalian retrotransposons include short interspersed nuclear elements(SINEs), long interspersed nuclear elements with a 10-nucleotide CpG interval are a preferred sub- (LINEs) and endogenous retroviruses(long terminal strate,and these are found in many imprinted loci. repeat(LTR)-type retrotransposons However, there are many other regions in the genome One way to control transposable elements is through with the same CpG spacing. Another study showed that epigenetic mechanisms?. In the male germline, al DNMT3L interacts with unmodified H3K4 (REF. 61), retrotransposon sequences undergo de which might restrict targets to regions without H3K4me. methylation during the fetal prospermatogonium Together, both nucleosome modification and CpG spac- stage, concomitant with the de novo methylation ing might provide the basis for the recognition of the of the imprinted loci(FIG. 1). Gene-knockout studies imprinted loci by DNMT3A-DNMT3L(REF. 60)(FIG. 3). in mice showed both common and differential target The differential methylation of the imprinted loci in the specificities of DNMT3A and DNMT3B with respect male and female germlines might require additional to these sequences: SINEBl is mainly methylated factors. In the case of paternally imprinted H19, pro- by DNMT3A, whereas LINEl and IAP are methyl X-chromosome inactivation tection of this locus from de novo DNA methylation in ated by both DNMT3A and DNMT3B (. 50). By The process that occurs in oocytes requires CCCTC-binding factor(CTCE), which contrast, DNMT3L is required for methylation of is known to bind to an unmethylated H19(H19 fetal all these sequences, indicating the crucial function liver mRNA)regulatory region and broad specificity of this factor in de novo dNA the pair of x chromosomes is X-chromosome inactivation in female mice is imprinted methylation((FIG 3) ownregulated to match the in pre-implantation embryos and the extra-embryonic The functional importance of DNA methylation in tissues of post-implantation embryos, and in both cases retrotransposon silencing and germ-cell development that is present in males. The the paternal X chromosome is preferentially inactivated. was first seen in Dnmt3L knockout mice. The LINE and inactivation process involves This imprinted X inactivation depends on both an acti- AP retrotransposons, of which de novo methylation a range of epigenetic vating imprint on the maternal X chromosome and an was prevented by Dnmt3L mutations, were highly tran mechanisms on the inactivated inactivating imprint on the paternal X chromosome. scribed in spermatogonia and spermatocytes' hanges in dNa methylation As mentioned above, the inactive X chromosome is mutations also caused meiotic failure with widespread histone modifications. reactivated in female PGCs, but maintenance of the non-homologous chromosome synapsis and progressive active state of the maternal X chromosome beyond fer- loss of germ cells by the mid-pachytene stage[TABLE 1) Chromosome synapsis tilization requires an imprint. Nuclear transplantation This resulted in complete azoospermia in older ani wo pairs of sister chromatids experiments showed that this maternal imprint is set on mals. The non-homologous synapsis could arise from the X chromosomes during the growth of the oocyte, illegitimate interactions between dispersed retrotrans- chromosomes)that occurs at as with the imprints at autosomal loci. As the maternal poson sequences that were unmasked by demethyla X chromosome from Dnmt3a/Dnmt3b double-mutant tion or from single-or double-strand breaks that were oocytes seems to have normal imprints, the mecha- produced during replicative retrotransposition Argonaute proteins are the nism of this imprinting might be different from that of Recently, a link between a small-RNA pathway and central components of RNA. autosomal imprinting DNA methylation of retrotransposons was discovered silencing mechanisms. They k Parthenogenesis, which is a successful development of (FIG 3).MILL, a member of the Piwi subfamily of Argonaute roteins, Is and, if it were possible in mammals, would provide a way early as E12.5(REF. 70) and interacts with a class of small and the catalytic activity for to produce clones of livestock animals. However, imprint- RNAs called piwi-interacting RNAs(piRNAs) In ing is a major barrier to parthenogenetic development Mili-mutant testis, LINEl and IAP retrotransposons were NATURE REVIEWS GENETICS @2008 Nature Publishing GroupX-chromosome inactivation The process that occurs in female mammals by which gene expression from one of the pair of X chromosomes is downregulated to match the levels of gene expression from the single X chromosome that is present in males. The inactivation process involves a range of epigenetic mechanisms on the inactivated chromosome, including changes in DNA methylation and histone modifications. Chromosome synapsis The association or pairing of the two pairs of sister chromatids (representing homologous chromosomes) that occurs at the start of meiosis. Argonaute proteins Argonaute proteins are the central components of RNA￾silencing mechanisms. They provide the platform for target-mRNA recognition by short guide RNA strands and the catalytic activity for mRNA cleavage. with a 10-nucleotide CpG interval are a preferred sub￾strate, and these are found in many imprinted loci60. However, there are many other regions in the genome with the same CpG spacing. Another study showed that DNMT3L interacts with unmodified H3K4 (Ref. 61), which might restrict targets to regions without H3K4me. Together, both nucleosome modification and CpG spac￾ing might provide the basis for the recognition of the imprinted loci by DNMT3A–DNMT3L (Ref. 60) (FIG. 3). The differential methylation of the imprinted loci in the male and female germlines might require additional factors. In the case of paternally imprinted H19, pro￾tection of this locus from de novo DNA methylation in oocytes requires CCCTC-binding factor (CTCF), which is known to bind to an unmethylated H19 (H19 fetal liver mRNA) regulatory region62. X-chromosome inactivation in female mice is imprinted in pre-implantation embryos and the extra-embryonic tissues of post-implantation embryos, and in both cases the paternal X chromosome is preferentially inactivated. This imprinted X inactivation depends on both an acti￾vating imprint on the maternal X chromosome and an inactivating imprint on the paternal X chromosome. As mentioned above, the inactive X chromosome is reactivated in female PGCs, but maintenance of the active state of the maternal X chromosome beyond fer￾tilization requires an imprint. Nuclear transplantation experiments showed that this maternal imprint is set on the X chromosomes during the growth of the oocyte63, as with the imprints at autosomal loci. As the maternal X chromosome from Dnmt3a/Dnmt3b double-mutant oocytes seems to have normal imprints64, the mecha￾nism of this imprinting might be different from that of autosomal imprinting. Parthenogenesis, which is a successful development of unfertilized eggs, is observed in many vertebrate species and, if it were possible in mammals, would provide a way to produce clones of livestock animals. However, imprint￾ing is a major barrier to parthenogenetic development in mammals because expression levels of the imprinted genes, which include many important developmental genes, are unbalanced in such embryos. When the imprinted genes are appropriately modified by genetic engineering and developmental manipulation, however, it was possible to derive adult female mice with two mater￾nal genomes and no paternal complement65,66 (BOX 3). As the method involves genetically engineered animals and highly complex nuclear-transfer technologies, its direct application to livestock seems difficult. Epigenetic silencing of retrotransposons Only germ cells can transmit genetic information to the next generation. Therefore, transposons, which mobilize in the genome and might cause insertional mutations, have to be strictly controlled in these cells. Approximately 40–50% of the mammalian genome is occupied by retrotransposons, which mobilize through an RNA intermediate, although many of them are truncated or have accumulated mutations. Mammalian retrotransposons include short interspersed nuclear elements (SINEs), long interspersed nuclear elements (LINEs) and endogenous retroviruses (long terminal repeat (LTR)-type retrotransposons). One way to control transposable elements is through epigenetic mechanisms67. In the male germline, all retrotransposon sequences undergo de novo DNA methylation during the fetal prospermatogonium stage50, concomitant with the de novo methylation of the imprinted loci (FIG. 1). Gene-knockout studies in mice showed both common and differential target specificities of DNMT3A and DNMT3B with respect to these sequences: SINEB1 is mainly methylated by DNMT3A, whereas LINE1 and IAP are methyl￾ated by both DNMT3A and DNMT3B (REF. 50). By contrast, DNMT3L is required for methylation of all these sequences50, indicating the crucial function and broad specificity of this factor in de novo DNA methylation (FIG. 3). The functional importance of DNA methylation in retrotransposon silencing and germ-cell development was first seen in Dnmt3L knockout mice. The LINE and IAP retrotransposons, of which de novo methylation was prevented by Dnmt3L mutations, were highly tran￾scribed in spermatogonia and spermatocytes53,54,68. The mutations also caused meiotic failure with widespread non-homologous chromosome synapsis and progressive loss of germ cells by the mid-pachytene stage (TABLE 1). This resulted in complete azoospermia in older ani￾mals. The non-homologous synapsis could arise from illegitimate interactions between dispersed retrotrans￾poson sequences that were unmasked by demethyla￾tion or from single- or double-strand breaks that were produced during replicative retrotransposition53. Recently, a link between a small-RNA pathway and DNA methylation of retrotransposons was discovered69 (FIG. 3). MILI, a member of the Piwi subfamily of Argonaute proteins, is expressed in the male and female gonads as early as E12.5 (REF. 70) and interacts with a class of small RNAs called piwi-interacting RNAs (piRNAs)69,71. In Mili-mutant testis, LINE1 and IAP retrotransposons were Box 2 | Transgenerational influence of epigenetic alterations in germ cells Recent studies have suggested that exposure to chemicals and malnutrition conditions can affect not only the children of the affected individuals, but also their grandchildren. This might be attributable to epigenetic alterations that occur in fetal germ cells. When gestating female rats were exposed to some endocrine disruptors, the number of spermatogenic cells decreased in the F1 generation. This effect was transmittable through the male germ line to subsequent generations, and this was correlated with altered DNA-methylation patterns112. In another example, exposure to methyl-donor supplementation during midgestation affected the epigenetic status of fetal germ cells113. The mouse Avy gene, which influences coat colour, is regulated by the DNA methylation status of an intra￾cisternal A particle (IAP) retrotransposon inserted at pseudoexon 1A of the gene. The methyl-donor supplementation shifted the coat colour of the F2 generation to a darker one. This suggests that the methyl donor directly or indirectly affected the epigenetic status of Avy in fetal germ cells. Finally, epidemiological studies have indicated that grandchildren of malnourished women show low birth weight114, that grandchildren of men who were well-fed before adolescence have a greater risk of mortality from diabetes, and that descendants of men who suffered famine have low cardiovascular mortality115. In these cases, it is possible that the nutrition status caused epigenetic alterations in germ cells, but further studies are needed to confirm this possibility. R E V I E W S nature reviews | genetics volume 9 | february 2008 | 133 © 2008 Nature Publishing Group