DNA methylation plays a crucial role in epigenetic gene regulation in many organisms (1). In most eukaryotes, cytosines in the symmetric CpG or CpNpG context are methylated (2) and methylation is propagated by maintenance methyltransferases (3,4) on the newly synthesized DNA strand shortly after replication. In addition, other methyltransferases may place new marks on DNA by de novo methylation (5–8), including the modification of asymmetric sites (9). Methylated cytosines may recruit specific chromatin components and initiate heterochromatin formation by histone deacetylation and methylation (10–12). On the other hand, modified histones may recruit DNA methyltransferases and mediate C-methylation (13,14). Though DNA methylation is found in organisms as diverse as animals, plants and fungi, it appears to be absent in Schizosaccharomyces pombe (15,16), Caenorhabditis elegans and other organisms where epigenetic silencing is mediated by histone modification alone.

It was first detected in tobacco that RNA can direct the methylation of homologous sequences in the genome (17,18). In plants and yeast (10) silencing of chromatin was found to be mediated by small RNAs that originate from the RNA interference pathway. RdDM (RNA-directed DNA methylation) is thus another branch of regulation by small RNAs. Recently, several models have been presented as to how the RNA guides for DNA modification may be generated and which components of the cellular machinery are involved (19–21). Targets for epigenetic chromatin silencing are frequently transposons and retrotransposons that are a source for naturally occurring siRNAs (22,23). Transposons are reactivated (22) and may be mobilized when components of the RNAi pathway are disrupted (24).

The presence of a functional RNAi machinery (25), the cloning of numerous siRNA-like molecules from Dictyostelium containing sequences of retrotransposons (F. Söderbom and V. Ambros, manuscript in preparation) and the suggestion that pericentric heterochromatin contained clustered retrotransposons (26) raised the question whether RNA-mediated DNA modification may be involved in epigenetic gene regulation in Dictyostelium. Similar to the situation in Drosophila, previous work had suggested that Dictyostelium DNA was not methylated (27). The discovery of a methyltransferase gene of the Dnmt2 family in the Drosophila genome and the subsequent detection of low levels of cytosine methylation (12) prompted us to reinvestigate this issue in Dictyostelium. In the recently completed genome of the social amoeba (26), we found a Dnmt2-like gene that had also been independently detected by Ponger et al. (28).

Our data show that Dictyostelium contains a functional DNA methyltransferase enzyme that methylates specific C-residues in the retrotransposons DIRS-1 and Skipper. Skipper silencing depends on both DNA methylation and a functional RNAi machinery.

For growth and development cycles, cells were plated out on SM agar in association with K.aerogenes and allowed to grow and develop. Spores were harvested after 7 days and used to inoculate a fresh plate.

Clonal isolates were obtained by plating ~500 cells on an SM plate with K.aerogenes. Cells were picked from clones into Costar plates containing HL5 medium with the appropriate selective drug. After several days, cells were transferred to suspension culture.

The product was cloned via EcoRI and BamHI into pDd-GFP to generate a C-terminal GFP fusion. For expression of the Myc-tagged protein, the same PCR product was cloned via EcoRI and HindIII into pDEX_RH. Two complementary oligonucleotides encoding the Myc-tag were inserted into the BamHI and HindIII sites to obtain a C-terminal Myc fusion.

PCR-primers for fragment amplification used for random-prime Klenow-labelling. DnmA 801 bp (+183 bp intron): Forward, TTAATGAGTCCACCATGTCAAC; reverse, CCTGTACCTTCAATAAATTTTCC; DIRS-1-RT-LTR 726 bp: forward, GTATGCCCTGTTCGCCACCTTGC; reverse, AACATTTATTTATTTGAATTTCCC; Skipper GAG 400 bp: forward, TGAAGCTAAAACCATTGACGC; reverse, CTAATTGAACTTCAGCAGTACC; eriA 660 bp: forward, GGATCCGATGTCCACAACAACAACATC; reverse, GGATCCTTTACTGATTTCATTGTTGAAAC.

For the thioredoxin control: forward, GAACGAGCTCCATGGCCAATAGAGTAATTCATG; reverse, CGCGGATCCTTATTTGTTTGCTTCTAGAGTACTTC.

Aliquots containing 10 µg of DNA were digested with the restriction enzymes indicated, separated on 1% agarose gels, blotted to nylon membrane and hybridized with PCR fragments labelled with [α-³²P]dATP by random priming. Primers for PCR amplification are listed as control primers in ‘bisulfite sequencing’ and in ‘northern blotting’.

Primers for DIRS-1 LTR PCR. Control PCR: forward, ATCAAATTGTTTTAGTTTTTAGTG; reverse, AACATTTATTTATTTGAATTTCCC. Bisulfite PCR: forward, ATCAAATTATTTTAATTTTTAATA; reverse, AATATTTATTTATTTGAATTTTTT.

Primers for Skipper RT. Control PCR: forward, CTGTTACCTTAGTGAAGATGGG; reverse, GGGCATCTATTGTCTTATGACATGG. Bisulfite PCR: forward, AAATCTTACATATATTATCAATAAA; reverse, AATAATTGAGTAGTATGTTGGGT.

Primers for Skipper LTR: Control PCR: forward, GTTAGAGACTCAAAACTAAATTA; reverse, TGTAAAAGTCACTCACACTAATC. Bisulfite PCR: forward, ACTCAAAACTAAATTAATTTAAAATTAA; reverse, ATGAAGGATAGAAAGAGTGAAAGAT.

PCRs were carried out on bisulfite treated and on untreated DNA with both sets of primers. Experiments were continued when only the appropriate primers yielded products on the different DNA samples. PCR products from cloned DIRS-1 LTR and Skipper RT fragments were also subjected to bisulfite sequencing to confirm that all cytosine residues could be converted.

In some experiments, DNA for the retrotransposon fragments was enriched by restriction digestion, gel electrophoresis and preparation of the appropriate size fragment for bisulfite treatment. The results were, however, not significantly improved in comparison with those obtained from bisulfite treatment of total DNA. PCR fragments were purified by MicroSpin™S-400HR columns before sequencing (MWG Biotech).

Cells expressing the DnmA-c-Myc fusion protein were blocked with PBG buffer (1× PBS containing 3% BSA and 0.045% cold water fish gelatin; Sigma) for 1 h at 37°C. Primary mAb 9E10 (directed against the c-Myc epitope) was added in a 1:50 dilution for 3 hours. Cells were washed three times with 1× PBS, then the secondary TRITC coupled polyclonal antibody (Dianova, Hamburg) was added at a 1:1000 dilution in PBG for 1 h.

Microscopic analysis was carried out with a Leica DM IRB inverted fluorescence microscope. For image acquisition a Leica DC 350F digital camera and IM50 software were used. Images were processed in Adobe Photoshop.

Corning ultra GAPS arrays containing probes for 450 defined genes, 5400 non-redundant expressed sequence tags (ESTs), and positive and negative controls were pre-hybridized and subsequently hybridized with the dye coupled cDNA preparations (41). After washing, arrays were scanned by Scan-Array^R 4000XL and analysed by Scan Array Express Version 2.1 (Perkin Elmer Life Sciences, Wellesley, USA). Furthermore, the R project for Statistical Computing (http://www.r-project.org) and Significance Analysis of Microarrays (SAM) Version 1.21 (http://www-stat.stanford.edu/~tibs/SAM/) were used.

Figure 3 (A) Semi-quantitative RT–PCR on dnmA gene transcripts from wild-type Ax2 cells and the dnmA disruption mutant. Capillary electrophoresis on DNA from wild-type cells (B) and the dnmA disruption strain. (C) The position of 5-methyl cytosine is indicated. (more ...)

Figure 1 Amino acid sequence alignment of Dnmt2 homologues from D.discoideum, Arabidopsis thaliana, Homo sapiens and Danio rerio. The 10 characteristic Dnmt2 motifs (44) are indicated by roman numbers. TRD, target recognition domain.

RT–PCR showed that dnmA RNA expression was high in vegetative cells and decreased during development (Figure 2A). N-terminal fusions with GFP or a c-Myc-tag were predominantly found in the nucleus (Figure 2B), thus confirming the in silico predicted nuclear localization by PSORTII (44) and LOCtree (45).

Figure 2 Expression of dnmA. (A) Semi-quantitative RT–PCR was performed on total mRNA isolated from vegetatively grown cells (veg) and from cells after 6 and 16 h of development on filters. Expression of the thioredoxin (trx) gene family is shown as a (more ...)

We generated a knock-out strain of the dnmA gene by homologous recombination. The gene disruption was confirmed by PCR, Southern blot (data not shown) and RT–PCR (Figure 3A). No obvious phenotypic alterations were observed during growth or in development.

In order to identify specific targets for DNA methylation, we analysed RNA derived from the dnmA knock out and from wild-type cells by microarray. Setting a threshold of at least 2-fold difference, 7 genes were found to be overexpressed in the KO strain and 12 were underexpressed. Among the upregulated genes, the ESTs SLD246 and SLE355 contained sequences encoding the gag and RT genes, respectively, of the retrotransposon Skipper (47). We concentrated initially on the analysis of Skipper, as this was of special interest since transposable elements are frequently targets for methylation (22). Interestingly, the ESTs SSL485 containing non-coding Skipper sequences and SLC703 at the very end of the RT/integrase sequence were not overexpressed in the mutant. Skipper has 20 copies in the Dictyostelium genome and consists of direct long terminal repeats and three open reading frames (ORFs) encoding GAG (matrix protein), PRO (protease) and RT (reverse transcriptase), POL (polymerase) and IN (invertase) (47). Skipper is expressed at very low levels during growth and development. Bisufite sequencing of a 294 bp region in the RT gene bottom strand showed that 12 out of 21 C-residues were methylated. Methylation was mostly at asymmetric sites and could not be correlated with any obvious sequence context (Figure 4A). By bisulfite sequencing of the RT fragment derived from the dnmA knock-out strain no methylated C-residues (with the exception of two that are likely to be artifacts, see below) could be detected. This indicated that methylation within this sequence depended on DnmA.

Figure 4 (A) Bisulfite sequencing of 294 bp in the Skipper RT gene (bottom strand). The top part shows the organization of the retrotransposon with GAG, PRO and RT genes and LTRs. ESTs that were included in the microarray are indicated below. (B) Bisulfite sequencing (more ...)

Bisulfite sequencing of 214 bp of the Skipper LTR bottom strand revealed no methylation of the 12 C-residues within this sequence (Figure 4B).

We then analysed the retrotransposon DIRS-1 that was not covered by the microarray. DIRS-1 occurs in 40 complete and ~200 incomplete copies in the genome. It consists of inverted long terminal repeats and three overlapping open reading frames encoding protein1, reverse transcriptase and a recombinase. In addition, a heat-shock-induced transcript in antisense orientation termed E1 is derived from the right LTR of the locus. Transcription of DIRS-1 is developmentally regulated and expression increases substantially after the onset of starvation (48–50).

We sequenced 128 bp of the right LTR (bottom strand) and found that 7 out of 23 C-residues in this region were consistently methylated (Figure 4C). Similar to the results from Skipper, methylation sites were mostly asymmetric and no obvious sequence context could be correlated with methylation.

Using DNA derived from the knock-out strain, no DNA methylation (with the exception of four that are likely to be artifacts, see below) could be detected, thus indicating that DnmA was responsible for all C modification of DIRS-1, as we had seen for Skipper.

Overall, eight non-converted C-residues were found by bisulfite sequencing in the dnmA mutant strain (two in Skipper RT, two in Skipper LTR and four in DIRS-1 LTR). This is a very small proportion of the unambiguously sequenced C-residues and we assume that this is due to an incomplete bisulfite reaction, or to PCR or sequencing errors. Importantly, most of these cytosine residues were never found to be methylated in the wild type and thus likely represent an in vitro artifact. However, we cannot exclude that an additional minor, non-conventional methyltransferase is active in the knock-out strain.

Figure 5 (A) Northern blots with specific Skipper (top) and DIRS-1 (bottom) probes on RNA isolated from wild-type Ax2 and dnmtA− cells after a short period of growth in axenic medium and after 16 passages through the developmental cycle (indicated by asterisk). (more ...)

In S.pombe siRNAs are involved in chromatin remodelling (10,51) and in plants(18), de novo DNA methylation is mediated by small RNAs (18). We therefore examined if Skipper and DIRS-1 expression was impaired in mutants affecting the RNAi pathway. As shown in Figure 5B, Skipper expression was not only upregulated in the dnmA knock-out but also in strains where one of the three RdRP genes (25) was disrupted. A minor degree of overexpression was also observed in dicer A (drnA) and the heterochromatin protein gene hcpA (M. Kaller, U. Euteneuer and W. Nellen, manuscript submitted) knock-outs. In contrast, DIRS-1 expression was only affected in the rrpC mutant. The data showed that components of the RNAi pathway are involved in suppression of both retrotransposons but that DIRS-1 and Skipper are regulated by distinct mechanisms.

Figure 6 RNAs derived from the small RNA library are aligned on the schematic depiction of DIRS-1: sense strand (A) and antisense strand (B). The inverted repeats are marked in yellow. The ORFs I, II and III and the E1 antisense transcript are depicted as colored (more ...)

Figure 7 (A) Southern blot with Skipper and DIRS-1 probes on wild-type Ax2 and dnmA− mutant cells after 0 and 16 rounds (indicated by asterisk) of the developmental cycle. From left: ethidium bromide staining, hybridization with an eriA probe, hybridization (more ...)

Until now, Dictyostelium DNA was considered not to be modified by cytosine methylation (27). With the completion of the genome sequencing project (26), a gene, dnmA, highly homologous to the methyltransferases of the Dnmt2 family, was detected. We show here that the dnmA gene product is responsible for methylation of C-residues in the retrotransposons Skipper and DIRS-1. As might be expected, a disruption of the methyltransferase gene results in transcriptional activation of Skipper and, probably as a consequence, in mobilization of the retrotransposon. Surprisingly, DIRS-1 expression is not affected by a loss of DNA methylation. We assume that the different organization and the different strategies of transposition of the elements may cause this discrepancy: While DIRS-1 has inverted terminal repeats that contain methylated C-residues, Skipper has direct terminal repeats that are not methylated. Furthermore, there are many incomplete copies of DIRS-1 in the genome and the element frequently transposes into its own copies while Skipper has almost exclusively complete and separate copies in the genome. DIRS-1 appears to constitute centromeres in Dictyostelium and is also clustered in the vicinity of the telomeres while Skipper is more spread out over the genome (50). Though these characteristics argue for mechanistic differences in transposition, they cannot functionally explain the susceptibility of Skipper but not DIRS-1 expression to DNA methylation. It is noteworthy that the promoter for Skipper transcription is probably located in the LTR and promoter methylation usually causes transcriptional silencing. In the case of Skipper, however, coding sequences rather than the LTR are methylated. We therefore propose that DNA methylation causes chromatin remodelling over the entire retroelement and thus blocks accessibility for the transcription machinery. The component(s) mediating chromatin remodelling remain elusive since a search for methyl-C binding proteins in the Dictyostelium database did not reveal any candidate gene containing significant similarity to methyl binding domains. This may not be surprising since conventional MBDs bind to the symmetric CpG motif in a certain sequence context (52) and most methylated C-residues are in a non-symmetric context in Dictyostelium. In contrast, Drosophila contains a putative methyl-C binding protein MBD2/3 that associates with specific regions in the DNA (53) but it has not been shown to bind to the methylated, asymmetric C-residues.

In yeast, siRNAs are required to recruit heterochromatin components to specific sites in the genome (10) and in plants DNA methylation is targeted by small RNAs (18). We therefore investigated whether siRNAs complementary to retrotransposon sequences could be found. Again, completely different results were obtained for the two retroelements: while no siRNAs corresponding to the complete Skipper element could be detected, more than 600 different molecules were found for DIRS-1. Since siRNAs covered essentially the entire DIRS-1 sequence, all methylation sites in the short segment that we sequenced by the bisulfite method had a corresponding siRNA. However, not all siRNAs covered C-residues that were methylated. Since methylation sites are mostly asymmetric, cytosine modification has to occur de novo as would be predicted for a Dnmt2-like enzyme (54). Our observations strongly suggest that siRNAs are involved in DNA methylation. However, their role in the regulation of DNA methylation cannot be determined at this point. In fact, it has been shown in Neurospora that de novo DNA methylation is independent of RNAi (55).

With the detection of siRNAs for DIRS-1, it was obvious to examine if components of the RNAi machinery were involved in silencing of the retrotransposons. The results were unexpected since Skipper silencing was strongly dependent on functional RdRPs and the dicer A gene even though no siRNAs had been found. In contrast, DIRS-1 expression was induced only when the RdRPC gene (rrpC) was knocked out. In a previous study (25), we had shown that rrpC was not required for post-transcriptional gene silencing by RNAi. However, rrpC is necessary for PTGS mediated by antisense RNA, indicating differences between gene silencing by single stranded (antisense) and double-stranded (RNAi) RNA molecules (H. Martens and W. Nellen, unpublished data). It is tempting to correlate this with the fact that the antisense transcript E1 is transcribed from the right DIRS-1 LTR. The observation that a knock out of the Dictyostelium Dicer homologue drnA had no effect on DIRS-1 expression could be explained by the activity of the second Dicer gene drnB. However, since Skipper was clearly upregulated in a drnA knock-out strain, this provides evidence that the two dicer genes are not completely redundant.

In a transposition active transcript of DIRS-1, the inverted repeats can form a double strand of more than 300 bp and the formation of a 1.7 kb double strand with the antisense transcript E1 (49,56) is feasible. However, siRNAs are also found in regions that do not form putative double strands. They may therefore originate from RdRP copies of the primary transcript and would thus constitute secondary siRNAs (57). Alternatively, an antisense strand could be generated by an RNA polymerase IV-like activity that has recently been detected in Arabidopsis (19,21). However, a polymerase IV-like enzyme has not yet been identified in Dictyostelium. Though we do not know whether the abundant siRNAs that we found for DIRS-1 are functionally involved in silencing, it is still surprising that most knock outs of RNAi components (except for rrpC) did not show an effect on DIRS-1 activity. Possibly, DNA methylation followed by chromatin remodelling as proposed above and antisense regulation provide redundant pathways for DIRS-1 silencing and both alone are sufficient. In contrast, Skipper, responded to the gene knock outs as if it was silenced by the RNAi machinery. Though no Skipper siRNAs were found, they could still be generated but escape detection due to rapid turnover or low abundance.

Various mechanisms to silence transposons and retrotransposons in one organism have been observed (19,22). In Arabidopsis, different classes of transposable elements are regulated by different, though overlapping mechanisms that require DNA methylation, histone modification and the RNAi system. Remarkably, different components contribute to various degrees to silencing of specific transposable elements. In addition, transcriptional and post-transcriptional mechanisms contribute simultaneously to the silencing machinery.

In an independent approach, Shaulsky and co-workers detected low levels of DnmA dependent DNA methylation in Dictyostelium (M. Katoh, T. Curk, Q. Xu, B. Zupan, A. Kuspa, G. Shaulsky, in press). Using methylation sensitive restriction enzymes, they identified methylated CpG sites in DIRS-1 and in the guaB gene.

Taken together, our data establish Dictyostelium as a model system to study mechanisms of epigenetic gene regulation by DNA methylation and RNA interference. Both mechanisms appear to overlap in function to silence the retrotransposons DIRS-1 and Skipper but further experiments are required to elucidate how the two pathways interact.

Supplementary Data are available at NAR Online.

The authors thank Michael Wassenegger (Centrum Grüne Gentechnik, Neustadt) for providing DNA from tobacco infected with PSTVd. Harald Rühling (Cell Biology, University Kassel) is acknowledged for support with microscopy. The authors thank Gad Shaulsky for communicating data prior to publication. This work was supported by grants of the Deutsche Forschungsgemeinschaft [SPP 1129, Ne285/8 and by the EU (STREP Fosrak)]. B.E.B. was in part supported by a stipend from the Otto-Braun-Fonds. Funding to pay the Open Access publication charges for this article was provided by the Deutsche Forschungsgemeinschaft.

Conflict of interest statement. None declared.