Histone H3 lysine 9 trimethylation is required for suppressing the expression of an embryonically activated retrotransposon in Xenopus laevis

Transposable elements in the genome are generally silenced in differentiated somatic cells. However, increasing evidence indicates that some of them are actively transcribed in early embryos and the proper regulation of retrotransposon expression is essential for normal development. Although their developmentally regulated expression has been shown, the mechanisms controlling retrotransposon expression in early embryos are still not well understood. Here, we observe a dynamic expression pattern of retrotransposons with three out of ten examined retrotransposons (1a11, λ-olt 2-1 and xretpos(L)) being transcribed solely during early embryonic development. We also identified a transcript that contains the long terminal repeat (LTR) of λ-olt 2-1 and shows a similar expression pattern to λ-olt 2-1 in early Xenopus embryos. All three retrotransposons are transcribed by RNA polymerase II. Although their expression levels decline during development, the LTRs are marked by histone H3 lysine 4 trimethylation. Furthermore, retrotransposons, especially λ-olt 2-1, are enriched with histone H3 lysine 9 trimethylation (H3K9me3) when their expression is repressed. Overexpression of lysine-specific demethylase 4d removes H3K9me3 marks from Xenopus embryos and inhibits the repression of λ-olt 2-1 after gastrulation. Thus, our study shows that H3K9me3 is important for silencing the developmentally regulated retrotransposon in Xenopus laevis.

retrotransposons representing 33-49% of these repetitive elements 6 . Furthermore, transposition of the human or mouse L1 retrotransposons was found to occur in transgenic mice and rats during embryogenesis, leading to somatic mosaicism 5 . Additionally, LTRs of retrotransposons can function as alternative promoters and first exons during mouse embryogenesis 4 , forming chimeric transcripts of retrotransposons and protein coding genes. Chimeric transcripts formed by this means are not expressed at the same developmental stages as their conventional non-chimeric counterparts 4 , suggesting that retrotransposons can control developmentally regulated expression of these transcripts.
Some studies found that the expression of retrotransposons in early mouse embryos is important for embryonic development. For example, murine endogenous retrovirus-like (MuERV-L), a mouse LTR retrotransposon that is transcriptionally activated 8-10 h after fertilization, is needed for the development of mouse embryos to the 4-cell stage 7 . Moreover, MuERV-L was found to form chimeric transcripts in mouse embryos 4 and marks the sub-population of embryonic stem cells, which carry the property of 2-cell stage embryos 8 . Another retrotransposon that seems to be important for the embryonic development of mice is LINE-1. The microinjection of morpholino antisense oligonucleotides targeting LINE-1 irreversibly arrests embryonic development of mice 9 . Since the expression of specific TEs in early embryos is essential for normal development, it is important to understand the mechanisms leading to TE activation in early embryonic stages and subsequent TE silencing. DNA demethylation does not seem to play a major role in TE activation in mouse embryos 10 . Instead, histone modifications were shown to regulate expression of TEs 1 . In mouse embryos, retrotransposons are marked by the repressive histone mark H3K9me3 and by the activating histone mark histone H3 lysine 4 trimethylation (H3K4me3), which is lost in parallel to transcriptional repression of retrotransposons 6 .
In contrast to regulation of retrotransposons in mammalian embryos, almost nothing is known about epigenetic mechanisms of retrotransposon regulation in Xenopus embryos. It has been shown that Xenopus has very few bivalent domains marked by both H3K4me3 and histone H3 lysine 27 trimethylation (H3K27me3) to regulate expression of developmentally important genes, being different from mammals and zebrafish embryonic chromatin 11,12 . Therefore, it is possible that different regulation mechanisms of embryonically active retrotransposons may exist in Xenopus. Moreover, Xenopus embryos allow the collection of a large number of in vivo embryonic cells when zyogotic transcription starts. In gene expression database several retrotransposon transcripts of Xenopus laevis could be identified. In addition, the expression of a few retrotransposons was reported to be developmentally regulated 13,14 . Based on this knowledge, we further investigated the expression of retrotransposons in Xenopus laevis and found retrotransposons whose expression is upregulated in early embryos. These retrotransposons are marked by H3K4me3 and H3K9me3 on their LTRs at different embryonic stages. Enzymatic removal of H3K9me3 marks from Xenopus embryos leads to the enhanced expression of a specific retrotransposon, demonstrating a role of H3K9me3 in the transcriptional repression of the developmentally regulated retrotransposon.

Results
The transcript levels of the retrotransposons 1a11, λ-olt 2-1 and xretpos(L) are upregulated after midblastula transition and downregulated after the gastrula stage. To identify retrotransposons that are specifically activated during early embryonic development, we first performed RNA-sequencing (RNA-seq) analysis of gastrula embryos (stage 10.5-11.5) and calculated RPKM (reads per kilobase per million mapped reads) values of putative retrotransposons, whose transcript sequences are available in NCBI database (Supplementary Table S1). Their transcripts were detected in gastrula embryos (Supplementary Table S1). Among them, the expression of 10 different putative retrotransposons was examined in eggs and embryos (stage 1-37/38) by reverse transcription-quantitative polymerase chain reaction (RT-qPCR) (Fig. 1a,b, Supplementary Fig. S1). Using this method, we are able to determine overall transcript levels of retrotransposons existing in numerous copies at different genomic loci (see Discussion). Three retrotransposons were found, whose transcript levels are upregulated after the midblastula transition and downregulated after the gastrula stage (Fig. 1b,c). These three retrotransposons are 1a11 13 (GenBank accession number: L11263.1), λ-olt 2-1 14 (GenBank accession number: AF145965.1) and xretpos(L) 14 (GenBank accession number: AF057166.1). λ-olt 2-1 and xretpos(L) show 95% nucleotide sequence identity in 1595 nucleotides 14 . These retrotransposons start to be activated soon after the midblastula transition. Then, the transcript levels of 1a11, λ-olt 2-1 and xretpos(L) in whole Xenopus laevis embryos increase from the blastula stage (stage 8) to the gastrula stage (stage 12) and gradually decrease after the late gastrula stage (stage 12.5) (Fig. 1b). For this analysis all values were normalized to the total RNA concentrations of each samples. Normalization to the transcript level of a constantly expressed control gene leads to similar results ( Supplementary Fig. S2). Furthermore, we asked if these three retrotransposons are transcribed from both strands. Antisense transcription was detected at the gastrula stage (Fig. 1d, St12.5 at upper graphs), but the sense transcripts represent the vast majority of transcripts (Fig. 1d, red bars compared to almost invisible blue bars at lower graphs).
During embryonic development the cell number increases rapidly from one cell (stage 1) to 150,000 cells at the tail bud stage (stage 26) (Supplementary Fig. S3). Therefore, it is possible that the increase of the retrotransposon transcripts after stage 8 only reflects the rapid cell proliferation, whereas the transcript level per cell stays rather constant. In order to exclude this possibility, the transcript level per cell was calculated based on the total cell number estimation ( Supplementary Fig. S3). These results also . pwp1 is a constantly expressed gene that was used as control (n = 1-4). All values were normalized to the overall RNA concentration of the respective sample. Sample RT-is a negative control. Error bars represent SEM. **P = 0.005, *P = 0.021. (c) Schematic diagram of the accumulated transcript levels of 1a11, λ-olt 2-1 and xretpos(L) during embryogenesis, explaining that their expression is upregulated after midblastula transition and shows the highest level at the late gastrula stage. (d) 1a11, λ-olt 2-1 and xretpos(L) are mainly detected as sense transcripts although antisense transcription can also be found. Forward primers were used for reverse transcription to analyze the expression of antisense transcripts from retrotransposons (blue bars), while reverse primers were used as a control (red bars). Upper graphs summarize antisense transcription at stages 1 and 12.5. Lower graphs represent the comparison between sense and antisense transcription although antisense transcription is almost invisible due to its much weaker expression than sense transcription. Relative transcript levels were compared to the transcript levels at the one-cell stage (St1 = 1) (n = 4). showed that the transcript level per cell increases after stage 8 and decreases after stage 12.5 (Fig. 1e). Thus it can be concluded that retrotransposons 1a11, λ -olt 2-1 and xretpos(L) show dynamic changes in transcription activities during embryonic development and the highest transcript levels of these are detected at the late gastrula stage.
The LTR of λ-olt 2-1 is found in a transcript, which contains a high homology sequence with mars2. It was previously shown that retrotransposons can regulate gene expression in early mouse embryos by functioning as alternative promoters and first exons 4 . Therefore, we asked if the three retrotransposons 1a11, λ-olt 2-1 and xretpos(L) also form chimeric (fusion) transcripts in early Xenopus embryos. To investigate this, a blast search for the three retrotransposons against a transcriptome assembly mapped to the Xenopus laevis genome 6.1 15 was performed. The transcriptome was generated from the RNA-seq analysis of gastrula embryos (stage 10.5-11.5) using paired-end sequencing with Illumina HiSeq2000. We found a transcript consisting of three exons, with the first exon and its surrounding region showing 97% sequence identity to the LTR of λ-olt 2-1 (Fig. 2a). The third exon shows no sequence identity to any reported gene. The second exon of this transcript, however, shows 87% sequence identity to the third exon of a gene called mars2. This mars2 gene encodes a mitochondrial methionyl-tRNA synthetase, which is needed for the translation of mitochondrial genes. Interestingly, tRNA is involved in the transposition of retrotransposons 16 . Hereafter, the transcript containing the LTR of λ-olt 2-1 as first exon (Fig. 2a, middle transcript) is referred to as mars2-like. The mars2 transcript found under the NCBI Reference Sequence NM_001086369.1 (Fig. 2a, lower transcript) is referred to as conventional mars2. The splicing junction between λ-olt 2-1 and mars2-high homology sequence was validated by continuous sequencing reads over the splicing junction ( Supplementary Fig. S4). Moreover, we amplified mars2-like transcripts using primers located at the beginning of exon 1 and in exon 2, and sequenced the PCR product. This sequencing analysis further validated that mars2-like contains both retroelement λ-olt 2-1 (exon1) and mars2 homology sequence (exon 2) ( Supplementary Fig. S5). mars2 expression during embryonic development. Previous studies using mouse have shown that chimeric transcripts containing retrotransposons are specifically expressed in early embryos 4 . We therefore asked if the expression pattern of mars2-like, which contains the LTR of the retrotransposon, during embryonic development differs from that of the conventional mars2. To test this, different pairs of qPCR primers were designed to specifically detect either mars2-like or conventional mars2 (Fig. 2a). The transcript level of conventional mars2 (primer 2) is highest at the first three embryonic stages and then gradually decreases until the blastula stage (stage 9) (Fig. 2b). At this stage, the transcript level of mars2-like (primer 1) starts to increase (Fig. 2b). Interestingly, it seems that mars2-like transcript is substituting the transcript of conventional mars2 (Fig. 2b, the bottom merged graph). The expression of mars2-like (primer 1) resembles the expression of λ-olt 2-1 itself (Fig. 1b). These results suggest that λ-olt 2-1 may regulate the expression of mars2-like transcript.

1a11, λ-olt 2-1, xretpos(L) and mars2-like are newly transcribed by RNA polymerase II.
Oocytes, direct precursors of eggs, contain some messenger RNAs (mRNAs) that have no or only very short poly(A) tails. During oocyte maturation and early embryonic development these short poly(A) tails are elongated, which facilitates translation [17][18][19] and the amount of poly(A) tailed RNAs increase 20,21 . Since we were using oligo-dT primer for RT-qPCR ( Fig. 1a), mRNAs without poly(A) tails or with too short poly(A) tails would not be reverse transcribed. If the transcripts of 1a11, λ-olt 2-1, xretpos(L) and mars2-like had such short poly(A) tails in eggs, we would therefore not be able to detect them until these short poly(A) tails were elongated. Thus, it is possible that the low transcript level of 1a11, λ-olt 2-1, xretpos(L) and mars2-like at the early embryonic stages is only an artifact, caused by the incapability of amplification by reverse transcription. In order to test if 1a11, λ-olt 2-1, xretpos(L) and mars2-like are newly transcribed or if their transcripts undergo changes in lengthening their poly(A) tails, α -amanitin was injected into fertilized one-cell embryos at two different concentrations (2 ng and 0.1 ng) (Fig. 3a). At the lower concentration (0.1 ng injection) α -amanitin can be used as a specific inhibitor for RNA polymerase II, while the higher α -amanitin concentration such as 20 μ g/ml also inhibits RNA Polymerase III 22 . Injected, normal-looking embryos were collected at the blastula stage (stage 9) and samples for qPCR were prepared (Fig. 3a). As a control, the transcript level of sox17 and pwp1 was measured. sox17 expression, which is embryonically activated, was significantly inhibited by the addition of α -amanitin (P = 0.00002, Fig. 3b), indicating that α -amanitin treatment inhibits embryonic transcription. In contrast, the transcript level of pwp1 was not influenced by the injection of α -amanitin (Fig. 3b), suggesting that pwp1 transcripts are maternal transcripts. Even at the low α -amanitin concentrations the transcript levels of 1a11, λ-olt 2-1, xretpos(L) and mars2-like were significantly (P < 0.05) reduced by the injection of α -amanitin (Fig. 3c). In contrast, the transcript level of conventional mars2 does not seem to be influenced by the injection of α -amanitin. These results indicate that all three retrotransposons and mars2-like are newly transcribed by RNA polymerase II in embryos, while conventional mars2 seems to be a maternal transcript. transcription during embryogenesis (Fig. 1b, c). It was shown that retrotransposons in mouse embryos are marked by the active histone mark H3K4me3 and by the repressive histone mark H3K9me3 6 . Repression of those retrotransposons was accompanied by the loss of H3K4me3, rather than the gain of H3K9me3 6 . Therefore, we tested if LTRs of 1a11, λ-olt 2-1 and xretpos(L) are also marked by H3K4me3 and H3K9me3 before and after the gastrula stage, at which transcriptional levels of those retrotransposons become highest. We collected embryos at the blastula stage (stage 8), the early gastrula stage (stage 10/11), the late gastrula stage (stage 12.5) and the neurula stage (stage [16][17][18] to perform chromatin immunoprecipitation (ChIP) analysis. Non-specific rabbit immunoglobulin G (IgG) was used as a negative control and precipitated 0.15-0.006% of input DNA (Supplementary Fig. S6a) while histone H3 as a positive control precipitated up to 27% of input DNA (Supplementary Fig. S6b). Both H3K4me3 and H3K9me3 were enriched on LTRs of retrotransposons ( Supplementary Fig. S6a). Relative enrichment of H3K4me3 to total histone H3 indicates that H3K4me3 levels increased up to stage 12.5 on all three retrotransposons (Fig. 4a, blue bars, *P < 0.05). A significant increase of H3K9me3 during developmental progression was observed in λ-olt 2-1 (Fig. 4a, red bars, **P = 0.00008). These results suggest that the increase of H3K9me3 levels is correlated with transcriptional repression of retrotransposons and therefore might be important for the suppression of retrotransposon expression.   Fig. 6b). All error bars represent SEM. **P = 0.00008, *P < 0.05. Blue = H3K4me3, red = H3K9me3, and green = IgG. (b) KDM4d mRNA was injected into fertilized embryos to remove H3K9me3 during embryogenesis. At the gastrula stage (stage 10) and at the neurula stage (stage 17) embryos were collected and prepared for western blotting or qPCR. Myc-GFP mRNA was used as a control. KDM4d overexpression removes H3K9me3 and releases transcriptional repression of λ-olt 2-1 at the neurula stage. Since the results of the ChIP analysis indicate that H3K9me3 is enriched on retrotransposons during embryonic development (Fig. 4a), we asked if this enrichment of H3K9me3 is required for retrotransposon repression in early and late embryonic stages. In order to reduce H3K9me3, mRNA of lysine-specific demethylase 4d (KDM4d), which specifically removes histone H3 lysine 9 trimethylation 23,24 , was injected into fertilized embryos. KDM4d mRNA-injected embryos were collected at the early gastrula (stage 10/10.5) and neurula (stage 16/17) stages and subjected to western blotting (Fig. 4b). The results showed that, in comparison to myc-GFP mRNA-injected control embryos, H3K9me3 was strongly reduced in KDM4d mRNA-injected embryos (Fig. 4c). Thus, H3K9me3 mark is removed in the gastrula and neurula embryos by KDM4d overexpression. Surprisingly, the removal of H3K9me3 mark did not cause any apparent phenotype in Xenopus embryos at least up to the neurula stage (stage 17).
Since mars2-like expression seems to be driven by λ-olt 2-1 (Fig. 2), we also measured the transcript level of mars2-like in embryos depleted of H3K9me3. We found that the KDM4d mRNA injection significantly (P = 0.040) enhances mars2-like transcription in neurula embryos (stage 17) in good agreement with the transcriptional upregulation of λ-olt 2-1 (Fig. 4d). In contrast, the expression of conventional mars2 was not affected by KDM4d mRNA injection (Fig. 4d). These results suggest that mars2-like is also regulated by H3K9me3 and support the idea that λ-olt 2-1 is involved in transcriptional regulation of mars2-like.
Even though H3K9me3 was also detected on λ-olt 2-1 LTR at early gastrula stages (stage 10) (Fig. 4a), KDM4d mRNA injection only had an effect on the expression of λ-olt 2-1 and mars2-like during the neurula stage (stage 17). A possible reason for this difference might be that some factors, which are associated with H3K9me3 and needed for heterochromatin establishment, are not yet expressed during gastrulation. These factors could for example be the Heterochromatin binding protein 1, which functions in heterochromatin mediated gene silencing and directly binds to H3K9me3 26,27 . By western blotting we showed that HP1β is not detectable in embryos at early gastrula stage (stage 10), but is expressed during neurula stages (stage 17) (Fig. 4c). It is therefore possible that the lack of HP1β during gastrulation may explain the insensitivity of λ-olt 2-1 and mars2-like expression to H3K9me3 removal in gastrula embryos (Fig. 4d).

Discussion
We examined the expression of ten putative retrotransposons during Xenopus embryogenesis and found that the expression of three of these retrotransposons (1a11, λ-olt 2-1 and xretpos(L)) is upregulated after midblastula transition and downregulated after the gastrula stage. The LTR of λ-olt 2-1 also forms a fusion transcript with a sequence that showed high similarity to mars2. All three retrotransposons are marked by the histone modifications H3K9me3 and H3K4me3. The KDM4d mRNA injection, which results in H3K9me3 removal, causes an increase of the λ-olt 2-1 and mars2-like transcript levels at the neurula stage, indicating that H3K9me3 plays a role in the repression of λ-olt 2-1 after gastrulation.
All three retrotransposons, 1a11, λ-olt 2-1 and xretpos(L), have been found as LTR retrotransposons in Xenopus laevis 13,14 . 1a11 encodes a 71 kDa protein, which shows similarity to the viral structure protein Gag. It was estimated that the Xenopus genome contains hundreds of sequences related to the LTR of 1a11, and 20-30 sequences related to its open reading frame 13 . The published expression analysis data also indicate transcriptional repression in late embryonic development, in good agreement with our qPCR data 13 (Fig. 1b). The Xenopus genome seems to contain approximately 200 copies of xretpos(L) LTR, 15 copies of the xretpos(L) open reading frame, and 50 copies of the complete xretpos(L) 14 . Again, the published expression analysis data of xretpos(L) support our qPCR data 14 (Fig. 1b). Moreover, it was shown by RNA and antisense RNA injections that xretpos(L) has a posterior-ventralizing activity during Xenopus embryogenesis 28 . All three retrotransposons have in common that they show no amino acid homology to the conserved pol gene. Since the pol gene encodes proteins needed for transposition (protease, reverse transcriptase, and integrase), it is likely that 1a11, λ-olt 2-1 and xretpos(L) have lost their ability to transpose 14 .
We found that the LTR of λ-olt 2-1 can function as the first exon and forms a transcript with a sequence that has a high homology with mars2 during Xenopus embryogenesis (Fig. 2a, called as mars2-like). However, we were not able to determine whether λ-olt 2-1 is located near conventional mars2 gene due to an incomplete understanding of Xenopus genome. The second exon of mars2-like transcript shows high nucleotide sequence homology to the third exon of conventional mars2. Conventional mars2 encodes a mitochondrial methionyl-tRNA synthetase. Mutations of the human homologue of this gene cause a neurodegenerative disease called Autosomal Recessive Spastic Ataxia with Leukoencephalopathy (ARSAL). It is likely that retroelements-derived repetitive sequences surrounding human Mars2 (e.g. Line 1 and Line2 elements) are mediating genomic rearrangements that cause ARSAL 29 . The methionyl-tRNA synthetase encoded by mars2 is needed for the translation of mitochondrial genes. Interestingly, tRNAs are also used as primer for the reverse transcription of retrotransposons or retroviruses. Methionyl-tRNA for example functions as a primer of yeast and plant retrotransposons like Ty1, Ty2, Ty3, Ty5 and Ta1 16 . It is also suggested that xretpos(L) is using arginine-tRNA as primer for reverse transcription 28 . Therefore, the developmentally regulated mars2 expression might be related to activities of transposable elements in embryos.
Our RT-qPCR data suggest that λ-olt 2-1 is driving mars2-like transcription (Fig. 2b). In this case, the LTR of λ-olt 2-1 might be used as a promoter to allow mars2-like expression (Fig. 2b). Several cases have been reported where LTR sequences act as alternative promoters 4,8,30 . One of the most relevant cases has been reported in mouse embryos at the 2-cell stage, where LTR sequences of MuERV-L regulate stage-specific expression of some protein coding genes 8 . However, in the case of mars2-like, it is unlikely that this gives rise to a meaningful protein since many stop codons are found in all reading frames. We therefore assume that mars2-like is more likely to be a long non-coding RNA (lncRNA) or a pseudo-gene product. lncRNAs are defined as non-coding RNAs longer than 200 nucleotides. These lncRNAs take on various functions, such as recombination, cell-cell signaling, post-transcriptional regulation, or regulation of chromatin states 31 . The transcript level of conventional mars2 is strongly reduced from stage 9 to stage 10; at the same time the transcript level of mars2-like strongly increases (Fig. 2b). This implies that mars2-like RNA might be involved in the regulation of conventional mars2 expression.
1a11, λ-olt 2-1 and xretpos(L) are marked by H3K9me3 as well as H3K4me3 (Fig. 4a). Our findings show that repression of λ-olt 2-1 after gastrulation is accompanied by the gain of H3K9me3, whereas the loss of H3K4me3 is not observed in λ-olt 2-1. In pre-implantation mouse embryos the repression of retrotransposons (LINE-1, SINE B2 and the LTR retrotransposon IAP) is accompanied by H3K4me3 loss, but not H3K9me3 gain 6 . It was suggested that the loss of H3K4me3 precedes the establishment of heterochromatin via H3K9me3 expansion. In more differentiated cells, H3K9me3 and DNA methylation play a central role in repression of repetitive elements [32][33][34][35] . Interestingly, our results suggest that some of Xenopus retrotransposons, especially λ-olt 2-1, seem to utilize heterochromatinization for early silencing. This notion is also supported by the appearance of HP1β when transcriptional repression starts (Figs 1b,4c) and by the functional test of H3K9me3 removal by KDM4d (Fig. 4d). The gradual increase in the H3K9me3 level during Xenopus embryonic development 36 (Fig. 4a) may influence such differences between mouse and Xenopus. Moreover, differential epigenetic regulation between mouse and Xenopus has been reported. For example, Xenopus only has very few bivalent histone marks of H3K4me3 and H3K27me3 unlike mice 11,12 and a very low level of H3K27me3 is found in early embryos 36 . H3K9me3 may have a prevalent function to silence genes by promptly recruiting heterochromatin proteins in Xenopus embryos.
We have shown an epigenetic mechanism of retrotransposon regulation in Xenopus laevis embryonic development. Although expression patterns of retrotransposons have been characterized in many different species, little is known about differential mechanisms of their regulation. It would be an interesting question to ask whether differential retrotransposon regulation affects their activities. Depending on the species, transposon diversity and frequencies of transposition are different 37 . In the case of Xenopus, one third of the Xenopus tropicalis genome consists of transposable elements, which is comparable to mammalian genomes 38 . However, it shows a high diversity of L1 non-LTR retrotransposons and LTR retrotransposons 38 . Such difference in transposon activities may result from a variety of gene regulation mechanisms for retroelements depending on species. Thus, exploring of the regulation mechanisms of transposable elements will provide insight into development and genomic diversity.

Animals.
All experiments using frogs were carried out following requirements of the UK Home Office.
All experimental protocols were approved by the UK Home Office.
RNA extraction, reverse transcription and qPCR. Three oocytes, eggs or embryos per sample were collected and frozen at − 80 °C. For RNA extraction, RNeasy Mini Kit (QIAGEN) was used according to the manufacturer's protocol. RNA was eluted in 35 μ l of the provided RNase-free water. For reverse transcription, 12 μ l of the RNA solution was mixed with 0.5 μ l 100 μ M Oligo-dT-Primer and 0.5 μ l RNase Inhibitor Murine and incubated at 65 °C for 5 min. All samples were then immediately placed on ice for more than 1 min. Afterwards, 6.6 μ l of a reverse transcription mix (4 μ l of 5 x First Strand buffer, 1 μ l of 0.1 M DTT (dithiothreitol), and 1.6 μ l of 10 mM dNTP-mix) were added to the RNA solution. For negative control (RT-), 4 μ l of this reaction mix were transferred to a new tube. After adding 0.5 μ l of Superscript III Reverse Transcriptase to the reaction mix, all samples were first incubated at 50 °C for 60 min and then at 70 °C for 15 min. All samples were diluted by 7.25-fold with ddH 2 O. For Fig. 1d, gene-specific forward primers (reverse primers as a control) were used for reverse transcription in order to examine whether retroelements are transcribed from both strands. Primer sequences are listed in Supplementary Table S2 For qPCR, 7300 Real Time PCR System (Applied Biosystems) was used. A reaction mix, consisting of 8.36 μ l of ddH 2 O, 0.14 μ l of 10 μ M Primer mix (forward + reverse), and 12.5 μ l of SYBR Green Jump Start Taq Ready Mix for High Throughput QPCR (Sigma-Aldrich) per sample, was used. This reaction mix (21 μ l per sample) was then transferred to a 96 Well PCR plate. Afterwards, 4 μ l of the template were added. The cycling conditions were used according to the ABI standard: 1 min at 50 °C (x1), 10 min at 95 °C (x1), 15 sec at 95 °C and 1 min at 60 °C (x40). For quantification, a standard curve was created using a serial dilution of embryonic cDNA or, in case of ChIP-samples, of genomic DNA. When we analyzed expression changes of retrotransposons during early embryonic development, values of one-cell stage (stage 1) were set as 1. In some experiments when one-cell stage values are undetectable, the following stage was set as 1.
Xenopus laevis RNA-Seq. RNA-Seq libraries of Xenopus laevis embryos at the gastrula stage were prepared with the TruSeq RNA kit and sequenced on an Illumina HiSeq 2000 instrument as paired ends at 50 base length. This resulted in 88.60 million paired end reads which were filtered for low quality reads (< Q20) followed by trimming of low quality bases from the ends of the reads (< Q20). Reads of good quality where a paired read was of low quality were kept. Adaptors were removed from both pairs using Trim Galore! (based on cutadapt).
Genome based expression analysis. Xenopus laevis genome build 6.1 from Xenbase 15 was used as a reference genome. Transcript sequence locations on the genome from a previous assembly were used as a background 39 . Repeat and transposable element sequences were obtained by searching the NCBI database for keywords along with Xenopus laevis. The resulting sequences were added to the genome and to the junction file for Tophat 2.0.6 40 . Tophat was used to map the RNA-Seq reads to the genome resulting in 77.75 million paired end reads mapped. Read counts were then generated for each of the repeat and transposable elements. These counts were normalized by transcript length and total read counts in order to calculate RPKM values.

Identification of mars2-like.
To identify the mars2-like transcript a BLAST search was performed, using a RNA-Seq dataset from Xenopus embryos at stage 10.5-11.5. The transcript sequence of mars2-like was originally found in transcriptome assembly mapped to the Xenopus laevis genome 6.1. The mars2-like map including exon information shown in Fig. 2a was captured from Xenopus laevis 6.1 genome. Moreover, mars2-like transcript was amplified using primers binding to exon 1 (Primer sequence: GGACAAGACATGTGAGGAATG) and to exon 2 (Primer sequence: ggccacgcgtTTCTCTAC-TACCCTTTTCCT or ggccacgcgtAGTAAATAGCATGGAAGAAT). The resulting PCR product was used for DNA sequencing.
Injection of α-amanitin or KDM4d-encoding mRNA into fertilized embryos. Eggs were in vitro fertilized by distributing them in a solution of smashed testis in L15-Medium. After 10 min 5 ml of ddH 2 O was added to the solution and the eggs were incubated for further 20 min. Fertilized eggs were then dejellied in 40 ml of 2% cysteine solution (40 ml of 0.1 x MMR, 0.8 g of Cystein, and 640 μ l of 10 M NaOH) for 3 min. The dejellied, fertilized eggs were washed 10 times with 0.1 x MMR (100 mM NaCl, 2 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , and 5 mM HEPES pH 7.4) to carefully remove the cysteine solution. Then stage 1 embryos were transferred to injection medium 41 (6% Ficoll and 0.4 x MMR) and incubated at 14 °C for at least 15 min. For injection of 4.6 nl of α -amanitin (2 ng or 0.1 ng) or 4.6 nl of KDM4d or myc-GFP encoding mRNA (1 mg/ml), the Drummond Nonoject II Injector was used. Mouse KDM4d was cloned into pCS2 vector and the linearized vector was subjected to mRNA synthesis using MEGAscript SP6 (Life Technology).
Polyacrylamid gels (12% SDS) were loaded with 20 μ l of this supernatant. Gel electrophoresis and western blots were performed according to standard protocols. For blotting PVDF membranes, a