Title: Pivotal role of AtSUVH2 in heterochromatic histone methylation and gene silencing in Arabidopsis
Abstract: Article17 March 2005free access Pivotal role of AtSUVH2 in heterochromatic histone methylation and gene silencing in Arabidopsis Kathrin Naumann Kathrin Naumann Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Andreas Fischer Andreas Fischer Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Ingo Hofmann Ingo Hofmann Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Veiko Krauss Veiko Krauss Department of Genetics, University of Leipzig, Leipzig, Germany Search for more papers by this author Sameer Phalke Sameer Phalke Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Kristina Irmler Kristina Irmler Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Gerd Hause Gerd Hause Biocenter, Martin Luther University Halle, Halle, Germany Search for more papers by this author Anne-Cathleen Aurich Anne-Cathleen Aurich Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Rainer Dorn Rainer Dorn Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Thomas Jenuwein Thomas Jenuwein Research Institute of Molecular Pathology, The Vienna Biocenter, Vienna, Austria Search for more papers by this author Gunter Reuter Corresponding Author Gunter Reuter Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Kathrin Naumann Kathrin Naumann Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Andreas Fischer Andreas Fischer Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Ingo Hofmann Ingo Hofmann Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Veiko Krauss Veiko Krauss Department of Genetics, University of Leipzig, Leipzig, Germany Search for more papers by this author Sameer Phalke Sameer Phalke Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Kristina Irmler Kristina Irmler Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Gerd Hause Gerd Hause Biocenter, Martin Luther University Halle, Halle, Germany Search for more papers by this author Anne-Cathleen Aurich Anne-Cathleen Aurich Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Rainer Dorn Rainer Dorn Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Thomas Jenuwein Thomas Jenuwein Research Institute of Molecular Pathology, The Vienna Biocenter, Vienna, Austria Search for more papers by this author Gunter Reuter Corresponding Author Gunter Reuter Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany Search for more papers by this author Author Information Kathrin Naumann1,‡, Andreas Fischer1,‡, Ingo Hofmann1, Veiko Krauss2, Sameer Phalke1, Kristina Irmler1, Gerd Hause3, Anne-Cathleen Aurich1, Rainer Dorn1, Thomas Jenuwein4 and Gunter Reuter 1 1Institute of Genetics, Biologicum, Martin Luther University Halle, Halle, Germany 2Department of Genetics, University of Leipzig, Leipzig, Germany 3Biocenter, Martin Luther University Halle, Halle, Germany 4Research Institute of Molecular Pathology, The Vienna Biocenter, Vienna, Austria ‡These authors contributed equally to this work *Corresponding author. Institute of Genetics, Biologicum, Martin Luther University, Weinbergweg 10, 06120 Halle, Germany. Tel.: +49 345 552 6300/303; Fax: +49 345 552 7294; E-mail: [email protected] The EMBO Journal (2005)24:1418-1429https://doi.org/10.1038/sj.emboj.7600604 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions Figures & Info SU(VAR)3–9 like histone methyltransferases control heterochromatic domains in eukaryotes. In Arabidopsis, 10 SUVH genes encode SU(VAR)3–9 homologues where SUVH1, SUVH2 and SUVH4 (KRYPTONITE) represent distinct subgroups of SUVH genes. Loss of SUVH1 and SUVH4 causes weak reduction of heterochromatic histone H3K9 dimethylation, whereas in SUVH2 null plants mono- and dimethyl H3K9, mono- and dimethyl H3K27, and monomethyl H4K20, the histone methylation marks of Arabidopsis heterochromatin are significantly reduced. Like animal SU(VAR)3–9 proteins SUVH2 displays strong dosage-dependent effects. Loss of function suppresses, whereas overexpression enhances, gene silencing, causes ectopic heterochromatization and significant growth defects. Furthermore, modification of transgene silencing by SUVH2 is partially transmitted to the offspring plants. This epigenetic stability correlates with heritable changes in DNA methylation. Mutational dissection of SUVH2 indicates an implication of its N-terminus and YDG domain in directing DNA methylation to target sequences, a prerequisite for consecutive histone methylation. Gene silencing by SUVH2 depends on MET1 and DDM1, but not CMT3. In Arabidopsis, SUVH2 with its histone H3K9 and H4K20 methylation activity has a central role in heterochromatic gene silencing. Introduction Epigenetically established changes in chromatin structure define the gene expression potential during development. These processes are controlled by complex DNA and histone modification systems. Modifications at the highly conserved N-terminal histone tails include acetylation, methylation, phosphorylation and ubiquitination, and show characteristic differences between active and repressed chromatin states (Stahl and Allis, 2000; Jenuwein and Allis, 2001). Lysine methylation at H3K4, H3K36 and H3K79 marks transcriptionally active chromatin, whereas methylation of H3K9, H3K27 and H4K20 defines repressed chromatin domains (Lachner et al, 2001; Fischle et al, 2003). Mono-, di- and trimethylation states of histone lysine residues extend the coding potential of the 'epigenetic histone code' and specific states of H3K9, H3K27 and H4K20 methylation define repressed heterochromatic domains in mouse and Drosophila (Peters et al, 2003; Schotta et al, 2004). In mouse H3K9 trimethylation, H3K27 monomethylation and H4K20 trimethylation index pericentric heterochromatin, whereas in Drosophila these regions show H3K9 di- and tri-, H3K27 mono-, di- and tri-, and H4K20 trimethylation (Ebert et al, 2004). Plant development in contrast to animals is rather plastic and considerably affected by environmental factors. Therefore, subtle changes in chromatin structure might be required for fine-tuning of gene expression and probably for this reason multi-gene families for DNA and histone modification systems are found in plants. In Arabidopsis, three different classes of DNA methylases (Cao and Jacobsen, 2002), 12 putative methylcytosine-binding proteins (Zemach and Grafi, 2003), 37 SET domain proteins (Baumbusch et al, 2001), 18 putative histone deacetylases and 12 putative histone acetylases have been identified (Arabidopsis Genome Initiative, 2000; Pandey et al, 2002). A predominant role in establishment of epigenetically stable active or repressed chromatin domains is attributed to histone lysine methylation by SET domain proteins. These proteins can be assigned to four groups typified by their Drosophila homologues E(Z), TRX, ASH1 and SU(VAR)3–9 (Jenuwein et al, 1998). In Arabidopsis, several of these genes were identified by developmental phenotypes of mutations. Like animal E(Z), CLF (Goodrich et al, 1997) and MEA (Grossniklaus et al, 1998) act as negative regulators and are involved in control of flower and endosperm development, respectively. The homeotic effects of clf and the function of MEA in imprinting of paternal genes first indicated an important role of SET domain proteins for chromatin regulation and epigenetic inheritance in plants (Goodrich et al, 1997; Vielle-Calcada et al, 1999). Similar to TRITHORAX in animals, Arabidopsis ATX1 acts as an activator of homeotic genes (Alvarez-Venegas et al, 2003). Epigenetically stable transmission of a condensed and transcriptionally inert chromatin state is characteristic for heterochromatin and heterochromatic gene silencing. With the evolutionary conserved SU(VAR)3–9 SET domain protein (Tschiersch et al, 1994; Ivanova et al, 1998; Aagaard et al, 1999) and demonstration of its function in histone H3K9 methylation (Rea et al, 2000), a basic factor of heterochromatin formation has been identified. In plants, in general, a large fraction of the genome is heterochromatic and extensive heterochromatic silencing processes are found. Involvement of histone H3K9 and DNA methylation in heterochromatin formation and heterochromatic gene silencing was documented in studies of the SU(VAR)3–9 homologue SUVH4 (KYP) and DNA methylation defective mutations of the MET1, CMT3, DRM and DDM1 genes (Cao and Jacobsen, 2002; Gendrel et al, 2002; Jackson et al, 2002; Tariq et al, 2003). In contrast to mammals, plants contain a large number of SU(VAR)3–9 homologues (Baumbusch et al, 2001). In Arabidopsis, 10 different SU(VAR)3–9 homologous SUVH proteins are found and we studied the function of SUVH1, SUVH2 and SUVH4, representative members of three subgroups of Arabidopsis SU(VAR)3–9 homologues. Here, for the first time, we show the heterochromatin association of the Arabidopsis SU(VAR)3–9 homologues SUVH1 and SUVH2 and present evidence that SUVH proteins differentially control heterochromatic histone methylation marks. Our data define SUVH2 as a central function in heterochromatin formation and heterochromatic gene silencing in Arabidopsis. In suvh2 null mutations all heterochromatin-specific histone methylation marks are significantly reduced, which is connected with strong suppression of transcriptional gene silencing (TGS). After overexpression, SUVH2 shows ectopic nuclear distribution and causes extensive heterochromatization, accompanied with an increase of DNA and heterochromatic histone methylation. SUVH2-dependent suppression or enhancement of TGS shows partial epigenetic stability connected with heritable changes in symmetric and nonsymmetric DNA methylation over consecutive generations. Mutational dissection of SUVH2 revealed a role of its N-terminus in control of nuclear protein distribution and a function of its YDG domain in directing DNA methylation to the target sequences. We also show that this DNA methylation does not cause silencing but rather is a prerequisite for consecutive SUVH2-dependent histone H3 and H4 methylation to establish heterochromatic silencing. Results Heterochromatin association of SUVH proteins in Arabidopsis The heterochromatin-associated SU(VAR)3–9 proteins control repressive chromatin structures by histone H3K9 methylation (Rea et al, 2000; Nakayama et al, 2001; Schotta et al, 2002). In contrast to animals and fungi, which have one or two Su(var)3–9 homologues, in Arabidopsis 10 SUVH genes encode SU(VAR)3–9 homologous proteins (Baumbusch et al, 2001). Tree reconstruction of 21 SUVH protein sequences from different plant species results in four clearly distinct subgroups of SUVH proteins in angiosperms (Figure 1A). Branching of both moss and fern SUVH fragments at the root of the angiosperm SUVH subgroups argues for a phylogenetic split of SUVH genes during early evolution of seed plants. The SUVH4-like genes are more distant to all groups of SUVH genes. Only SUVH4-like genes contain introns and retrotransposition might have been involved in evolution of the different SUVH gene families. Branching of gymnosperm PtaSUV1f with the SUVH4 subgroup suggests that functional differentiation of SUVH proteins had already occurred before the angiosperm–gymnosperm split. Figure 1.SU(VAR)3–9 homologous proteins in plants. (A) Four conserved groups of SUVH genes are found in angiosperms. Phylogenetic analysis of 21 SUVH protein sequences from Arabidopsis (AtSUVH), rice (OsSUVH), Pinus taeda (Pta), Physcomitrella patens (Pp) and Ceratopteris richardii (Cri). (B) Immunostaining of plants expressing myc fusion protein of SUVH1, SUVH2 and Drosophila SU(VAR)3–9 and HP1 in Arabidopsis interphase nuclei with α-myc. Heterochromatin association is found for the SUVH1 and SUVH2 proteins as well as for Drosophila SU(VAR)3–9 and HP1. Download figure Download PowerPoint To identify chromatin targets of the Arabidopsis SU(VAR)3–9 homologues SUVH1 and SUVH2 proteins, we studied their nuclear distribution in transgenic Arabidopsis plants expressing myc or EGFP fusion proteins. The SUVH1 and SUVH2 fusion proteins associate with the DAPI bright regions of interphase nuclei (Figure 1B), which represent pericentromeric heterochromatin in Arabidopsis (Fransz et al, 2000; Soppe et al, 2002). We also tested nuclear distribution of the Drosophila SU(VAR)3–9 and HP1 heterochromatin proteins in Arabidopsis. The Drosophila proteins in Arabidopsis also associate with heterochromatin (Figure 1B), indicating conserved mechanisms of heterochromatin association of these proteins in plants and animals. SUVH1 and SUVH2 proteins are both heterochromatin associated and could comprise redundant functions. Alternatively, these proteins could control distinct heterochromatic histone methylation marks as well as their combinatorial interplay. Differential control of heterochromatic histone methylation by SUVH proteins We analysed the effect of SUVH1, SUVH2 and SUVH4 on heterochromatic histone H3 and H4 methylation marks using specific histone methylation antibodies (Peters et al, 2003). In these studies, we used loss-of-function T-DNA insertion mutations or antisense lines (cf. Supplementary Figures 1 and 2). None of the insertional mutations or the antisense lines show a significant phenotypic defect in homozygous constitution. Arabidopsis heterochromatin is enriched in methylated DNA and mono- and dimethylated histone H3K9 (Tariq et al, 2003; Figure 2A). Other heterochromatic histone methylation marks in Arabidopsis are mono- and dimethyl H3K27 and monomethyl H4K20 (Figure 2A). In contrast, trimethyl H3K9, trimethyl H3K27, and di- and trimethyl H4K20 are found together with methylated H3K4 and H3K36 in Arabidopsis euchromatin (Figure 2A and B). In our studies loss-of-function mutations of SUVH1 and SUVH4 appear to show only weak reduction of mono- and dimethyl H3K9 in pericentromeric heterochromatin (Figure 2A and Supplementary Figure 1). In contrast, SUVH2 loss-of-function mutations strongly reduce all heterochromatin-specific histone and DNA methylation marks (Figure 2A and C). The most dramatic effect of SUVH2 is on H4K20 monomethylation. Reduction of DNA methylation at heterochromatic sequences in SUVH2 null plants was further quantified by bisulphite sequence analysis of Athila transposons (Supplementary Figure 3). In vitro analysis shows that SUVH2 is a nucleosome-dependent HMTase and methylates histone H3 and H4 in recombinant nucleosomes (Figure 2D). Figure 2.Differential effects of suvh1 and suvh2 mutations on histone and DNA methylation. (A, B) Immunohistochemical staining of wild-type, suvh1 and suvh2 interphase nuclei with antibodies recognizing specific histone and DNA methylation marks. In suvh2, but not in suvh1, mono- and dimethyl H3K9, mono- and dimethyl H3K27, monomethyl H4K20 and 5-methylcytosine (heterochromatic marks) are significantly reduced (A). No effects are found on the trimethyl H3K9, trimethyl H3K27, di- and trimethyl H3K36 and di- and trimethyl H4K20 euchromatic marks (B). (C) Western analysis of nuclear extracts of wild-type, suvh1 and suvh2 mutant plants. Only in suvh2 significant reduction of mono- and dimethyl H3K9 and monomethyl H3K27 is found. (D) In vitro recombinant SUVH2 shows H3 and H4 HMTase activity in assays with reconstituted nucleosomes. Download figure Download PowerPoint Drosophila Su(var)3–9 is a haplo- and triplo-dosage-dependent modifier of heterochromatic gene silencing (Schotta et al, 2002). In order to study the dosage-dependent effects of Arabidopsis SUVH1 and SUVH2, we constructed transgenes with the 35S* promotor (Mindrinos et al, 1994) or used a glucocorticoid-mediated transcriptional system, which allows controlled induction of gene expression by dexamethasone treatment (Aoyama and Chua, 1997). Overexpression of SUVH1 has no significant effect on pericentromeric heterochromatin and the protein remains heterochromatin associated (Figure 3A). In contrast, after overexpression, SUVH2 shows dispersed nuclear distribution, resulting in ectopic heterochromatization. By electron microscopic analysis, additional blocks of heterochromatic material can be detected (Figure 3A). Immunocytological analysis of SUVH2 overexpression plants reveals a significant increase in mono- and dimethyl H3K9, mono- and dimethyl H3K27, and monomethyl H4K20, as well as cytosine methylation (Figure 3B). We have quantified the dosage-dependent effects of SUVH2 on five histone methylation marks by Western analysis of bulk histones. In suvh2 mutant plants, heterochromatic H3K9 mono- and dimethylation, H3K27 monomethylation and H4K20 monomethylation are significantly reduced (Figures 2C and 3C), whereas SUVH2 overexpression causes enhancement of H3K9 dimethylation and H4K20 monomethylation (Figure 3B and C). In contrast, trimethylation of H3K9, which is a euchromatic histone methylation mark in Arabidopsis, is significantly reduced after SUVH2 overexpression (Figure 3C and D). Significant reduction in immunostaining for other euchromatic histone modification marks like dimethyl H3K4, acetyl H3K9 and trimethyl H3K27 is also found in SUVH2 overexpression plants (Figure 3D). Figure 3.Ectopic heterochromatization in SUVH2 overexpression lines. (A) Immunostaining of nuclei from 35S*∷mycSUVH1, 35S*∷mycSUVH2 with α-myc and GFP fluorescence analysis in dexamethasone-treated GVG∷SUVH1EGFP and GVG∷SUVH2EGFP plants. Only SUVH2 shows ectopic distribution. Electron microscopic analysis of nuclei from 35S*∷mycSUVH2 plants shows ectopic heterochromatin (arrowheads). (B) Immunocytological analysis of heterochromatic histone and 5-methylcytosine methylation in SUVH2 overexpression plants. Enhanced staining for all heterochromatic marks (mono- and dimethyl H3K9, mono- and dimethyl H3K27, monomethyl H4K20 and 5-methylcytosine) is found. (C) Western analysis of suvh2 mutant and 35S*∷mycSUVH2 overexpression plants. In suvh2, dimethyl H3K9 and monomethyl H4K20 are reduced. In 35S*∷mycSUVH2 overexpression plants, heterochromatic H3K9 dimethyl and H4K20 monomethyl are enriched, whereas euchromatic H3K9 trimethyl is reduced. (D) Immunostaining for euchromatic histone modification marks in SUVH2 overexpression plants. Staining for dimethyl H3K4, acetyl H3K9, trimethyl H3K27, dimethyl H3K36 and dimethyl H4K20 is significantly reduced. Download figure Download PowerPoint SUVH2-dependent growth defects and the mutational dissection of its molecular function None of the 24 35S*∷mycSUVH1 overexpression lines nor the GVG∷SUVH1-EGFP lines after dexamethasone treatment show any phenotypic defects (not shown). In contrast, four of 24 independent 35S*∷mycSUVH2 transgenic lines show significant growth reduction (mini-plant phenotype) and a curled cotyledon phenotype (Figure 4A and B). Western blot analysis shows direct correlation between the extent of SUVH2 overexpression and the strength of mini-plant growth defects (Figure 4C). Line 35S*∷mycSUVH2#4 with lower amount of additional SUVH2 is only weakly affected, whereas the three lines 35S*∷mycSUVH2#5, #6 and #22 with higher amount of SUVH2 manifest a strong mini-plant phenotype. Similarly, the GVG∷SUVH2-EGFP lines manifest a mini-plant phenotype only after dexamethasone treatment (Figure 4A). Significant rescue of SUVH2 overexpression phenotypes is found in suvh2/+; 35S*∷mycSUVH2#6/+ plants by introducing a suvh2 null allele (Figure 4B). Figure 4.Growth and developmental defects in SUVH2 overexpression plants. (A) SUVH2 overexpression causes mini-plant phenotype in 35S*∷mycSUVH2#5, #6, #22 and dexamethasone-treated GVG∷SUVH2EGFP lines. (B) SUVH2 overexpression seedlings show a curled cotyledon phenotype (upper panel). By introducing a suvh2 null allele, the mini-plant and curled cotyledon phenotypes are significantly rescued in suvh2/+; 35S*∷mycSUVH2#6/+ plants (lower panel). (C) Western analysis of extracts from 35S*∷mycSUVH2 lines with α-myc. The amount of mycSUVH2 protein correlates with the strength of growth defects. 35S*∷mycSUVH2#4 plants with a weak growth reduction expresses a lower amount of additional SUVH2 as compared to lines with a mini-plant phenotype. Download figure Download PowerPoint For functional dissection of SUVH2, we isolated mutations within the 35S*∷mycSUVH2#5 transgene after EMS mutagenesis. These mutations were identified as dominant suppressors of the curled cotyledon phenotype in M1 seedlings. All confirmed dominant suppressors rescued the mini-plant growth defect. In M3 two classes of mutant lines could be established. In 224 lines the suppressor mutation and the SUVH2 transgene segregate independently, whereas in 96 lines complete linkage is found. The latter class represents 35S*∷mycSUVH2#5 transgene mutations and we studied the functional consequence of seven such mutations (Figure 5). All are missense mutations located either in the N-terminus (5–1), the YDG (5–2) or the SET domain (5–3 to 5–7) of SUVH2 (Figure 5A). Only the N-terminus mutation 5–1 interferes with ectopic nuclear distribution of SUVH2 (Figure 5B). Mutations 5–2 and 5–3 located in the YDG domain and the region between the YDG domain and the preSET region, respectively, cause loss of ectopic 5-methylcytosine, H3K9 dimethylation and H4K20 monomethylation, although ectopic distribution of SUVH2 is not affected (Figure 5B and C). Mutations in the SET domain (5–3 to 5–7) eliminate ectopic H3K9 and H4K20 methylation, but leave ectopic DNA methylation unaffected. Figure 5.Functional dissection of SUVH2 by transgene mutations. (A) Molecular nature of 35S*∷mycSUVH2#5 transgene mutations. Structure of SUVH2 with the conserved YDG, preSET, SET and postSET (p) domains. (B) Immunostaining with α-myc shows that only the N-terminus mutation 5-1 eliminates ectopic distribution of SUVH2. (C) All mutations eliminate ectopic H3K9 and H4K20 methylation. The N-terminus mutation 5-1, the YDG mutation 5-2 and mutation 5-3 eliminate ectopic DNA methylation, whereas in plants with the SET domain mutations 5-4, 5-5, 5-6 and 5-7 ectopic DNA methylation is observed. (D) Silencing of Athila transposons by SUVH2 overexpression is rescued by all transgene mutations independent of DNA hypermethylation (RT–PCR) (cf. Supplementary Figure 3 for bisulphite data). Download figure Download PowerPoint In order to study SUVH2 overexpression effects on heterochromatic silencing of endogenous sequences, we analysed expression of Athila transposons. Overexpression of SUVH2 causes strong repression of Athila, whereas all SUVH2 transgene mutations rescue this silencing effect (Figure 5D). Immunocytological and bisulphite sequence analysis shows that none of the SET domain mutations reduce DNA hypermethylation at Athila although its silencing is released (Figure 5D and Supplementary Figure 3). Taken together, our mutant analysis resolves a sequence of molecular events connected with SUVH2-induced heterochromatic gene silencing. Mutation 5–1 indicates a possible role of the SUVH2 N-terminus in target sequences recognition, whereas the YDG domain region appears to be involved in directing DNA methylation to these sequences. DNA methylation alone is not sufficient for silencing, but rather functions as a mark directing SUVH2-dependent histone methylation to sequences subjected to silencing. Moreover, our data suggest that SUVH2-mediated DNA methylation precedes histone methylation. Dosage dependence and epigenetic maintenance of SUVH2-induced transgene silencing To study dosage-dependent effects of SUVH proteins on repeat-induced TGS, we used a new type of transgene constructs that contain four tandem copies of the luciferase gene (cf. Materials and methods). The LUC2 transgene shows moderate luciferase silencing, which is significantly enhanced by SUVH2 (Figure 6) but not SUVH1 overexpression (data not shown). Significant enhancement of TGS is already found in crosses with the 35S*∷mycSUVH2#4 line showing only weak growth defects (Figure 5), indicating that TGS is already efficiently enhanced by low levels of SUVH2 overexpression. The loss-of-function effect of SUVH2 on TGS was studied using a specific antisense line. Out of 25 independent 35S*∷SUVH2as antisense lines, we selected line SUVH2as#11 causing complete elimination of SUVH2 transcript, whereas other tested SUVH genes are not affected (cf. Supplementary Figure 1). After a cross of SUVH2as#11/SUVH2as#11 with LUC2 homozygous plants, all SUVH2as#11/+; LUC2/+ offspring show strong suppression of TGS (Figure 6). Significant suppressor effects are also seen with antisense lines only partially eliminating the SUVH2 transcript (not shown). Our experiments reveal the dosage-dependent modifier effect of SUVH2 on TGS. Overexpression enhances whereas loss-of-function suppresses TGS. We further studied the epigenetic stability of SUVH2 modified LUC2 transgene silencing in progeny plants, lacking SUVH2 overexpression or antisense constructs which were generated by a backcross of the F1 progeny with wild-type plants. Partial epigenetic stability of modified LUC2 transgene silencing is found in the offspring plants for the enhanced as well as suppressed state of LUC2. Reciprocal crosses produced identical results (data not shown). Reversion to the level typical for control plants is found in offspring of the second backcross generation (Figure 6A and B). To study whether epigenetic maintenance of SUVH2-modified TGS is based on stable transmission of changes in DNA methylation, we performed bisulphite sequence analysis of the LUC2 transgene. In plants produced by the first backcross (BC1) to wild type, the LUC2 transgene inherited either from SUVH2 overexpression or SUVH2 antisense plants still shows increased and decreased DNA methylation, respectively (Figure 6A and C). The data suggest that DNA methylation is involved in maintenance of the SUVH2-induced epigenetic effects. Figure 6.Dosage-dependent modifier effects of SUVH2 on LUC2 transgene silencing. (A–C) Crosses of 35S*∷mycSUVH2#4/+ and 35S*∷SUVH2as#11/+ with LUC2 homozygous plants and backcrosses (BC) of F1 and F2 LUC2/+ plants to wild type (A). Structure and activity of the LUC2 repeated transgene in control plants (B). In 35S*∷mycSUVH2#4/+; LUC2/+ plants with SUVH2 overexpression, LUC2 silencing is enhanced, whereas in 35S*∷SUVH2as#11/+; LUC2/+ plants without SUVH2 LUC2 silencing is strongly released. The repressed or activated state of the LUC2 is maintained after a backcross of 35S*∷mycSUVH2#4/+;LUC2/+ and 35S*∷SUVH2as#11/+;LUC2/+ with wild-type plants, respectively. In LUC2/+ offspring from the second backcross generation, reversion of transgene silencing to the control level is found. Symmetric (red bars) and nonsymmetric (blue bars) DNA methylation at LUC2 transgenes was studied by bisulphite sequencing (A and B). Bisulphite sequence analysis of control LUC2/+, F1 35S*∷mycSUVH2#4/+; LUC2/+ and 35S*∷SUVH2as#11/+; LUC2/+ as well as LUC2/+ BC1 progeny plants. (C) Stars denote significantly changed symmetric CpG (red) and CpNpG (green) and nonsymmetric CpNpN (blue) cytosine residues (N, no G). Download figure Download PowerPoint Modification of TGS by SUVH2 is connected with complex DNA methylation pattern We analysed the effect of SUVH2 on DNA methylation at the LUC2 transgene, GUS transgene repeats and Athila transposon sequences (Figure 6 and Supplementary Figure 3). All these sequences show already in wild-type a consistent amount of DNA methylation. In our studies, we compared CpNpN (N, no G) nonsymmetric with CpG and CpNpG symmetric DNA methylation. Loss or overexpression of SUVH2 affects both symmetrical and nonsymmetrical DNA methylation at all studied sequences. In suvh2 mutant as well as SUVH2as#11 antisense plants, strongest reduction of nonsymmetrical methylation at CpC is observed (Supplementary Figure 3). Since DNA methylase MET1 is suggested to function in maintenance of cytosine methylation (Finnegan and Kovac, 2000), we studied at the completely silenced LUC7 transgene the effects of a newly isolated strong met1-h1 mutation on SUVH2-induced transgene silencing (Figure 7A and B). Furthermore, we also studied interaction between SUVH2 and a newly isolated strong cmt3-h1 CHROMOMETHYLASE mutation (Figure 7C). In LUC7 met1-h1 and LUC7 cmt3-h1 plants, silencing of the LUC7 transgene is significantly released (Figure 7B and C). LUC7 transgene silencing is differentially affected in 35S*∷mycSUVH2#5/+; LUC7/LUC7; met1-h1/met1-h1 and 35S*∷mycSUVH2#5/+; LUC7/LUC7; cmt3-h1/cmt3-h1 plants. The suppressor effect of met1-h1 dominates the enhancer effect of SUVH2 overexpression, whereas the enhancer effect of SUVH2 dominates the suppressor effect of cmt3-h1. These results demonstrate that transgene silencing by SUVH2 depends on MET1, but not significantly on CMT3. Figure 7.Genetic interaction of S