SET DOMAIN GROUP25 encodes a histone methyltransferase and is involved in FLOWERING LOCUS C activation and repression of flowering.

Covalent modifications of histone lysine residues by methylation play key roles in the regulation of chromatin structure and function. In contrast to H3K9 and H3K27 methylations that mark repressive states of transcription and are absent in some lower eukaryotes, H3K4 and H3K36 methylations are considered as active marks of transcription and are highly conserved in all eukaryotes from yeast (Saccharomyces cerevisiae) to Homo sapiens. Paradoxically, protein complexes catalyzing H3K4 and H3K36 methylations are less-extensively characterized in higher eukaryotes, particularly in plants. Arabidopsis (Arabidopsis thaliana) contains 12 SET DOMAIN GROUP (SDG) proteins phylogenetic classified to Trithorax Group (TrxG) and thus potentially involved in H3K4 and H3K36 methylations. So far only some genes of this family had been functionally characterized. Here we report on the genetic and molecular characterization of SDG25, a previously uncharacterized member of the Arabidopsis TrxG family. We show that the loss-of-function mutant sdg25-1 has an early flowering phenotype associated with suppression of FLOWERING LOCUS C (FLC) expression. Recombinant SDG25 proteins could methylate histone H3 from oligonucleosomes and mutant sdg25-1 plants showed weakly reduced levels of H3K36 dimethylation at FLC chromatin. Interestingly, sdg25-1 transcriptome shared a highly significant number of differentially expressed genes with that of sdg26-1, a previously characterized mutant exhibiting late-flowering phenotype and elevated FLC expression. Taken together, our results provide, to our knowledge, the first demonstration for a biological function of SDG25 and reveal additional layers of complexity of overlap and nonoverlap functions of the TrxG family genes in Arabidopsis.

As a sessile organism, plants adopted during evolution specific developmental processes to endure unfavorable conditions. Plants develop in three phases, which are embryonic, vegetative, and reproductive. Transition from one phase to another is precisely regulated, involving the perception and integration of a variety of endogenous signals and environmental cues. Flowering represents the transition from the vegetative to reproductive development, which in Arabidopsis (Arabidopsis thaliana) is mainly controlled by four pathways: the autonomous pathway that constitutively activates flowering, the gibberellin pathway that promotes flowering when plants are grown under noninductive conditions, the vernalization pathway that promotes flowering after prolonged cold exposure, and the photoperiod pathway that induces flowering in response to increasing day length (Komeda, 2004). FLOWERING LOCUS C (FLC), which encodes a transcription factor belonging to the MADSbox family, is a key component in the flowering time control. It represses flowering in a dosage-dependent way and its expression level is fine tuned by multiple regulators (for review, see He, 2009).
Chromatin carries genetic and epigenetic information in all eukaryotes. Chromatin remodeling through histone exchange and/or histone covalent modifications is necessary for the correct deployment of developmental programs, for the maintenance of cell fates, and for the adaptation to environmental changes. Despite the difference in sequence and structure, many plant MADS-box genes including FLC, like animal homeotic (HOX) genes, are controlled by fac-tors belonging to Polycomb group (PcG) and Trithorax group (TrxG; Pien and Grossniklaus, 2007;He, 2009;Shen and Xu, 2009). The SET [named after SuVar(3-9), E(z), and Trithorax (TRX)] domain conserved in components of both PcG and TrxG complexes has histone methyltransferase activity. While the SET-domain protein E(z) catalyzes methylation at H3K27 associated with gene silencing by PcG complexes, the TrxG proteins catalyze H3K4 and H3K36 methylations involved in transcriptional activation. Arabidopsis has three homologs of E(z): MEDEA (MEA), CURLY LEAF (CLF), and SWINGER (SWN). Loss-of-function mutants of these genes are described and it has been well documented that MEA, CLF, and SWN are key components of different complexes, which function similarly to the animal Polycomb Repressive Complex2 (PRC2) involved in transcription repression. The Arabidopsis PRC2 complexes play critical roles in many biological processes, including parental imprinting in seed formation, stem cell maintenance, stable repression of FLC upon vernalization in promoting flowering, and others (Huh et al., 2008;Kô hler and Villar, 2008).
Here we report on the molecular and functional characterization of SDG25/ATXR7. We show that loss of function of SDG25 promotes flowering through reduction of FLC expression. The SDG25 protein shows histone methyltransferase activity in vitro when oligonucleosomes were used as substrate. In the sdg25-1 mutant plants, weakly decreased levels of H3K36 dimethylation were detected at some regions of FLC. Transcriptomic analysis revealed 43 downregulated and 122 up-regulated genes in the sdg25-1 mutant. Despite the down-regulation of FLC in both sdg25-1 and sdg8 but up-regulation of FLC in sdg26, the number of perturbed genes found in sdg25-1 significantly overlapped with that previously reported in sdg26 but not in sdg8. To our knowledge, our results provide the first demonstration for biological function of SDG25 and reveal additional layers of complexity of the TrxG family members in the regulation of expression of common and specific genes.

Relationship of the Arabidopsis TrxG Family Members and Structure of SDG25
Initial phylogenetic analysis of SET-domain proteins in Arabidopsis identified 19 proteins as homologs of TRX and ASH1, both belonging to the TrxG family in animals (Baumbusch et al., 2001). Subsequent refined analysis narrowed the number to 12 proteins as members of the Arabidopsis TrxG family (Springer et al., 2003;Ng et al., 2007). We focused on this group and expanded the phylogenetic analysis by including proteins from Arabidopsis and rice (Oryza sativa) as well as proteins from nonplant species that have been demonstrated as H3K4 and/ or H3K36 methyltransferases. As shown in Figure 1A, the 12 proteins of Arabidopsis can be classified into two major groups. SDG25 is most closely related to the rice SDG717 and the SET1 proteins from fungi and animals. The highly conserved SET domain is found at the C terminus of this group of proteins (Fig. 1, B and C). Beside the C-terminal region, the rest of SDG25 sequences do not show significant homologies with other proteins of the group. The domains PWWP, FYR, and PHD conserved in ATX1 and ATX2 are not found in SDG25. Instead a GYF domain is found at the N-terminal part of SDG25 (Fig. 1B). GYF domains are also found in homologs of the chromatin remodelling SWIB complex, in transcription factors, and in other proteins (Kofler and Freund, 2006). It is proposed that GYF domains recognize Prorich sequences in protein-protein interactions (Kofler and Freund, 2006). To investigate the biological function of SDG25, we obtained a T-DNA insertion line of the SALK collection (Alonso et al., 2003), SALK_149692, from the Arabidopsis Biological Resource Center (ABRC). Genomic PCR analysis confirmed that this line contains a T-DNA insertion in the second exon of SDG25 ( Fig. 2A; data not shown). Homozygous (called hereinafter mutant sdg25-1) plants were obtained by self-pollination. Reverse transcription (RT)-PCR analysis showed that the full-length SDG25 transcripts were absent in the mutant plants (Fig. 2B), indicating that sdg25-1 is a loss-of-function mutant.
Compared to wild-type Columbia (Col), the sdg25-1 mutant showed an early flowering phenotype (Fig.  2C). Flowering time is temporally controlled by the developmental state (for example, plant size and age) and regulated by environmental cues, such as light and temperature. To further investigate the flowering phenotype of the mutant, we compared the flowering time of sdg25-1 with Col at different photoperiods and with or without vernalization treatment. As shown in Figure 2D, flowering of the sdg25-1 mutant plants, like that of wild type, was promoted by long-day photoperiods as well as by vernalization treatment. In all tested conditions, the sdg25-1 mutant plants flowered earlier than wild type. Flowering time measured from a developmental perspective, such as the total number of rosette leaves, further confirmed the early flowering phenotype of the sdg25-1 mutant (Fig. 2E). In the same experiments, the sdg25-1 mutant and wild-type Col were compared for the number of juvenile leaves, which are characterized by the absence of abaxial trichomes. The mutant plants showed a reduced number of juvenile leaves (Fig. 2F), indicating that the juvenile vegetative phase has been shortened. The adult phase also has been shortened since the mutant plants also showed a reduced number of adult leaves with abaxial trichomes (=total rosette leaf number in Fig. 2E minus juvenile leaf number in Fig. 2F).
To confirm that the mutant phenotype is indeed caused by the loss of function of SDG25, the sdg25-1 mutant was transformed with a genomic fragment containing the SDG25 gene (Fig. 3A). As shown in Figure 1. SDG25 is a member of the TrxG family. A, Phylogenic analysis of 12 Arabidopsis SDG proteins (At, Arabidopsis), nine rice SDG proteins (Os, rice), together with H3K4 and H3K36 methyltransferases so far found in fungi (Sp, Schizosaccharomyces pombe; Sc, yeast; Nc, Neurospora crassa), Drosophila melanogaster (Dm), and Homo sapiens (Hs). The plant proteins fall in two groups: Group 1 contains proteins of other organisms involved primarily in H3K36 methylation whereas group 2 contains proteins of other organisms involved primarily in H3K4 methylation. B, Comparison of structural domains of ATX1, ATX2, and SDG25. The SET domain is the most highly conserved, and the post-SET Cys-rich region C is also well conserved in all three proteins. The other domains (PWWP, FYR, and PHD) found in ATX1 and ATX2 are not conserved in SDG25. Instead, SDG25 contains a GYF domain. C, Sequence alignment of the C-terminal last 133 amino acids from SDG25, SpSET1, and HsSET1. This highly conserved region contains the SET domain and the post-SET Cys-rich region. Figure 3B, the transgene SDG25 rescued the sdg25-1 mutant, resulting in flowering time similar to wild type. Examination of two other T-DNA insertion lines within the SDG25 gene, GK-630A05-022799 and SAIL_446_F12 (http://www.arabidopsis.org), revealed that these mutants have an early flowering phenotype similar to sdg25-1. Taken together, our data establish SDG25 as an important gene involved in repression of flowering.

Expression of FLC Is Down-Regulated in the sdg25-1 Mutant
Both the autonomous and vernalization pathways act to repress FLC expression to induce flowering (Baurle and Dean, 2006). To investigate the molecular mechanism underlying the early flowering phenotype of sdg25-1, we compared the expression level of FLC between the mutant and wild-type plants. As shown in Figure 4, FLC expression was down-regulated in the sdg25-1 mutant compared to wild-type plants. The MADS AFFECTING FLOWERING1 (MAF1) to MAF5 genes encode MADS-box proteins closely related in sequence to FLC and could repress flowering when ectopically overexpressed (Ratcliffe et al., 2003). Expression of these MAF genes was, however, unchanged in the sdg25-1 mutant (Fig. 4), indicating that the early flowering phenotype of sdg25-1 is spe-cifically associated with down-regulation of FLC expression.
H3K36 Dimethylation Is Slightly Down-Regulated at Some Regions of FLC in the sdg25-1 Mutant Down-regulation of FLC expression in the sdg25-1 mutant is consistent with the expected role of SDG25 as an active TrxG member in transcriptional activation. To investigate histone methylation status in the mutant, we utilized antibodies specifically recognizing dimethyl-H3K4, trimethyl-H3K4, or dimethyl-H3K36. Western-blot analysis revealed similar levels of histone methylations between sdg25-1 and wild type (data not shown), indicating that SDG25 is not a major enzyme for global methylation on H3K4 or H3K36. We investigated H3K4 and H3K36 methylation at several regions of FLC (Fig. 5A) by chromatin immunoprecipitation (ChIP) analysis. While sdg25-1 barely affected levels of H3K4 dimethylation and trimethylation, reduced levels of H3K36 dimethylation were observed at some regions (c-g) of FLC in sdg25-1 compared to wild type (Fig. 5B). These reductions are weak but were reproducibly observed in two independent experiments. Real-time PCR analysis confirmed that compared with wild type the sdg25-1 mutant contained reduced levels of H3K36 dimethylation at FLC locus (see Supplemental Fig. S1). Besides early flowering, sdg25-1 did not show any additional visible phenotype. To better understand SDG25 function, we investigated its expression pattern and protein activities. We found that SDG25 is expressed ubiquitously in seedling, leaf, inflorescence, and root (Fig. 6A). The GFP-SDG25 fusion protein was localized in the cytoplasm and the nucleus (Fig. 6B). To investigate enzyme activity of SDG25, we expressed two fragments of 320 (SDG25S) and 699 (SDG25L) amino acids in length of the C terminus of the protein. Both SDG25S and SDG25L showed histone H3 methyltransferase activity when oligonucleosomes were used as substrates (Fig. 6C), whereas such activity was not detectable when free core histones (not shown) or mononucleosomes (Fig. 6C) were used as substrates. Similar substrate preference for oligonucleosomes with higher-order chromatin structure had also been previously observed for SDG8 and SDG26 proteins . Taken together, our data provide experimental demonstration of SDG25 as a histone H3 methyltransferase and suggest that SDG25 could have a broad function.
Transcriptome Analysis Reveals Overrepresentation of Common Genes Perturbed in the sdg25-1 and sdg26-1 Mutants We investigated the global genome expression pattern in sdg25-1 through transcriptome profiling analysis. As performed previously for the sdg8 and sdg26 mutants , 6-d-old seedlings of wild type and sdg25-1 mutant were used in this analysis. We found that 43 genes were down-regulated and 122 genes up-regulated to more than 1.5-fold in the sdg25-1 mutant seedlings compared to the wild type (see Supplemental Table S1). These differentially expressed genes are involved in several processes including transcription, phytohormone and external stimuli responses, metabolism, and transport. Comparison of these differentially expressed genes with those previously described in atx1-1 (Pien et al., 2008), atx2-2 (Saleh et al., 2008), sdg8-1, and sdg26-1  revealed that a significantly overrepresented number of genes are in common and are misregulated in the same direction between the sdg25-1 and sdg26-1 mutants whereas only a few genes are in common between the sdg25-1 and sdg8-1 mutants (Fig. 7A). We also did not detect a signifi-  cant overlap between sdg25-1 and atx1 or atx2 sets of differentially regulated genes.
Such a great similarity between the trancriptomes of sdg25-1 and sdg26-1 mutants is not expected because of the opposite flowering phenotypes of the mutants: sdg25-1 shows an early flowering phenotype whereas sdg26-1 shows a late-flowering phenotype. We further investigated the expression of some genes by RT-PCR analysis. Consistent with transcriptome data, we found that At3g45140 and At1g24020 are down-regulated, and At3g16720 and At3g11820 are up-regulated in both sdg25-1 and sdg26-1 mutants; whereas FLC is down-regulated in sdg25-1 mutant but up-regulated in sdg26-1 mutant (Fig. 7B). Taken together, our data indicate that SDG25 and SDG26 are closely related in regulation of genome transcription.

DISCUSSION
Our study identified SDG25 as a positive regulator of FLC expression, which inhibits flowering. Both the juvenile and adult phases of sdg25-1 mutant plants are shortened, resulting in an early transition to reproductive development. The early flowering phenotype of sdg25-1 compared to wild type was observed under different photoperiods and with or without vernalization. Nonetheless, the sdg25-1 mutant still responded to induction by long-day photoperiods and vernalization, indicating that SDG25 acts downstream of the latter two signaling pathways. A similar behavior of flowering phenotype was previously reported for the sdg8/efs mutants (Soppe et al., 1999;Kim et al., 2005;Zhao et al., 2005). The early flowering phenotype of sdg25-1 compared to sdg8/efs is weaker. This is likely associated, at least in part, with the down-regulation of MAF1, MAF4, and MAF5 expression specifically observed in the sdg8/efs mutants (Kim et al., 2005;Zhao et al., 2005), but not in the sdg25-1 mutant. Besides early flowering no other obvious phenotypes were observed in the sdg25-1 mutant. This is in contrast to the pleiotropic phenotype of the sdg8/efs mutants, which includes reduced organ size, increased shoot branching, altered carotenoid composition, and low fertility (Soppe et al., 1999;Zhao et al., 2005;Dong et al., 2008;Xu et al., 2008;Cazzonelli et al., 2009). Transcriptome analysis also indicated that few genes overlapped between the lists of genes differentially expressed in the sdg8/efs and sdg25 mutants. It appears clear that SDG25 and SDG8/EFS essentially regulate different target genes. While a global reduction of H3K36 di-and trimethylation had been observed in sdg8/efs mutants Dong et al., 2008;Xu et al., 2008), western-blot analysis did not detect significant changes of H3K36 dimethylation in the sdg25-1 mutant. Interestingly, the number of down-regulated genes found in the sdg25-1 mutant seedlings was only about two times less compared to that found in the sdg8/efs mutant seedlings. At FLC, the reduction of H3K36 dimethylation was much more drastic in sdg8/efs mutants than in the sdg25-1 mutant, whereas similar levels of reduction of FLC expression were observed in both sdg8/efs and sdg25-1 mutants. Together these observations indicate that there is no linear correlation between levels of H3K36 dimethylation and transcription rates. H3K36 methylation might affect transcription depending on gene context, on positional information of the nucleosome within a gene, on the context of other histone modifications, or others. In spite of a higher level of sequence homology of SDG25 with some H3K4 methyltransferases (Fig. 1), both western-blot analysis of nuclear extracts and ChIP analysis of FLC chromatin failed to detect a significant change on H3K4 di-or trimethylation in the sdg25-1 mutant plants. High levels of H3K4 trimethylation had been previously shown to be associated with FLC activation (He et al., 2004;Xu et al., 2008), and ATX1 is partially responsible for H3K4 trimethylation at FLC (Pien et al., 2008). Our study indicates that sdg25-1 may have affected FLC expression through defect on H3K36 dimethylation. At this stage, however, we cannot exclude the possibility that sdg25-1 may affect H3K4 methylation at some unknown genome regions. Recombinant SDG25 proteins only showed modest methyltransferase activity on oligonucleosomes, making it impractical to analyze enzyme specificity by in vitro biochemical approaches. H3K4 methylation was found in more than two-thirds of Arabidopsis genes (Zhang et al., 2009). From currently characterized mutants of the Arabidopsis TrxG family genes, only atx1 showed a mild reduction in the global level of H3K4 trimethylation (Alvarez-Venegas and Avramova, 2005) and sdg4 showed reduction of H3K4 di-and trimethylation in specific floral tissues (Cartagena et al., 2008). Locus-specific reduction of H3K4 methylation was observed in atx1 (Alvarez-Venegas and Avramova, 2005;Pien et al., 2008), atx2 (Saleh et al., 2008), and sdg8/efs (Kim et al., 2005;Cazzonelli et al., 2009). Our study here eliminates SDG25 being responsible for the genome-wide H3K4 methylation. Thus the major enzyme(s) involved in the H3K4 methylation in Arabidopsis remain to be identified and characterized.
The transcriptome of sdg25-1 showed little overlap with that of sdg8-1, atx1, or atx2, but showed a significant overlap with that of sdg26-1 (Fig. 7). Recombinant SDG26 protein could methylate oligonucleosomes in vitro but perturbation of histone methylation in sdg26-1 had not been detected . In contrast to sdg25-1, sdg26-1 showed elevated expres-sion of FLC and a late-flowering phenotype. Because histone methylation remained unchanged at FLC in mutant plants , sdg26-1 might have affected FLC expression indirectly. Our finding of a great similarity between the transcriptomes of sdg25-1 and sdg26-1 mutants indicates that perturbed expression of the shared genes in the two mutants are independent from modifications of FLC expression and flowering time. Molecular mechanisms underlying regulation of shared target genes by SDG25 and SDG26 will require more investigations; however, our finding of regulation of transcription of a common set of genes by SDG25 and SDG26 constitutes an important step toward understanding the relationship and biological functions of the TrxG family genes. Finally, it is interesting to note that SDG25 and SDG26 also have unique (nonoverlapping) target genes; an important number of genes are perturbed specifically in sdg25-1 or sdg26-1 mutant plants (Fig. 7). It thus seems evident that SDG25 and SDG26 have both redundant and nonredundant functions.
Although we cannot rule out the possibility that some of the genes identified in transcriptome analysis might represent secondary rather than primary targets of SDG25, the great number of up-regulated genes found in sdg25-1 mutant plants suggests that SDG25 may also act as a repressor of transcription. TrxG proteins are generally involved in transcriptional activation. Nevertheless, it is feasible that canonical positive regulators such as SDG25 may also play a role as negative regulators in certain contexts, and finding so many up-regulated genes in sdg25-1 mutant plants makes this idea attractive. Several possibilities Figure 6. Analysis of expression, subcellular localization, and enzyme activity of SDG25. A, RT-PCR analysis of SDG25 expression in different types of plant materials. ACTIN serves as an internal control. B, Localization of GFP:SDG25 in a cortex cell of the transgenic Arabidopsis root. Fluorescence confocal image is shown together with differential interference contrast image. Note that fluorescence is distributed in the nucleus (indicated by arrowhead) as well as in the cytoplasm. A similar subcellular localization in the nucleus and the cytoplasm was also observed in other types of cells in roots, stems, and leaves. C, In vitro methylation assay with recombinant SDG25 fragments. The top section shows Coomassie Blue-stained gel and the bottom section shows corresponding autoradiography. The tested recombinant proteins as enzymes are indicated on top of lanes and substrates used in the test are indicated at bottom of lanes. Positions of nucleosome core histones H2A, H2B, H3, and H4 are indicated.
exist to explain a repressor role of TrxG proteins in transcription. First, in yeast (Saccharomyces cerevisiae) H3K36 methylation is known to negatively influence transcription of some genes through, at least in part, recruitment of histone deacetylase activity (Carrozza et al., 2005;Joshi and Struhl, 2005;Keogh et al., 2005). Second, the genic patterns of distribution of histone methylations are important for transcriptional activity; H3K4 trimethylation occurring predominantly near the 5# transcription start site marks transcriptional engagement whereas H3K36 dimethylation occurring predominantly in the coding and 3# end of genes marks transcriptional elongation in all studied eukaryotes including Arabidopsis (Oh et al., 2008;Zhang et al., 2009). Enhancement of H3K36 dimethylation within the 5# transcription start site may interrupt initiation and is associated with reduction of gene expression (Biswas et al., 2006;Oh et al., 2008). Last but not least, some TrxG proteins are known to be involved in methylation of diverse sites of histones, e.g. ASH1 can methylate H3K4, H3K9, and H4K20 (Beisel et al., 2002); NSD1 can methylate both H3K36 and H4K20 in vitro (Rayasam et al., 2003); and sdg8 mutant plants exhibit reduced levels of H3K36 dimethylation, H3K36, and H3K9 trimethylation . H3K9 and H4K20 methylations are generally recognized as repressive marks of transcription. More investigations will be necessary to examine these above listed possibilities in SDG25-mediated repression of gene expression.

Sequence Analysis
For phylogenetic analysis, the entire protein sequences were aligned using ClustalW multiple sequence alignment program and then optimized manually by removing poorly aligning regions, which leaves only the SET domain. The finalized file was subjected to phylogenic analysis using MEGA3.0 with bootstrapping set at 500 replicates. The resulting consensus tree was displayed using the TreeView program. Protein structure analysis for conserved domains was performed using ScanProsite program (http://www.expasy.org/prosite/).

Plant Materials and Growth Conditions
The sdg25-1 mutant corresponds to SALK_149692 (Col) of T-DNA insertion strains from the ABRC. Mutant and wild-type Col plants were grown either on plates containing Murashige and Skoog media or in pots with soil. Growth conditions were as previously described . Flowering time was measured as days to bolting as well as from a developmental perspective, as the total number of rosette leaves (Michaels and Amasino, 1999).

Complementation of Mutant Plants
A fragment of about 10 kb in length that covers the whole SDG25 gene with promoter and terminator was isolated by restriction enzyme digestion (BamHI-ClaI) from bacterial artificial chromosome MDH9. This fragment was then cloned in the BamHI-ClaI sites of pBluescript II SK vector (Stratagene, www.stratagene.com), resulting in pBSK-SDG25G. Finally the SDG25containing KpnI-BamHI fragment from pBSK-SDG25G was cloned into the KpnI-BamHI sites of pCAMBIA1300 vector (CAMBIA; http://www.cambia. org), resulting in pCAMBIA1300-SDG25G. The pCAMBIA1300-SDG25G plasmid was introduced into Agrobacterium tumefaciens for plant transformation. The sdg25-1 mutant plants were transformed using the floral-dip method (Clough and Bent, 1998). Figure 7. Comparison of differentially expressed genes in sdg25-1, sdg26-1, and sdg8-1 mutants. A, The circles represent the genes that were down-regulated (left) or up-regulated (right) in sdg25-1 mutant upon transcriptome analysis. The total number of genes is shown in brackets. The percentages of these genes that were also down-or up-regulated in sdg26-1, in both sdg26-1 and sdg8-1, or in sdg8-1 are shown. *, Statistically significant overlap: P = 3.9 3 10 269 (left) and P = 7.1 3 10 281 (right). B, RT-PCR analysis of expression of some genes perturbed in both sdg25-1 and sdg26-1 mutants. ACTIN serves as an internal control.

Isolation of SDG25 cDNA
The entire coding region of SDG25 cDNA was PCR amplified from Col inflorescence cDNA using the forward primer 5#-cgggatccATGGTTGCGGTT-GATTCCAC-3# and the reverse primer 5#-ccgctcgagTCAGTTTCGTCTT-GAAAACCAC-3# (lowercase letters represent non-SDG25 nucleotides added to introduce restriction enzyme sites). The PCR product was digested and cloned into BamHI-XhoI sites of pCRII-TOPO vector (Invitrogen; http:// www.invitrogen.com), resulting in pCR-SDG25. Sequencing analysis revealed that the cDNA fragment was 4,167 bp in length and encoded a protein of 1,388 amino acids (database accession no. EU014691). This slightly differs from the predicted At5g42400.1 sequence in the Arabidopsis (Arabidopsis thaliana) database because of some alternative intron splicing. Alignment of amino acid sequences between SDG25 and At5g42400.1 was shown in Supplemental Figure  S2. We further confirmed the 3# end region of SDG25 by 3#-RACE analysis.

Protein Subcellular Localization
The SDG25 cDNA from pCR-SDG25 was amplified by PCR and cloned via a BP clonase reaction in the Gateway donor vector pDONR201 (Invitrogen, http://www.invitrogen.com). They were recombined by Gateway technology into pK7WGF2,0 vector (Karimi et al., 2002), resulting in GFP:SDG25 fusion gene under the control of the 35S promoter. The fusion gene was introduced into Arabidopsis plants by Agrobacterium tumefaciens-mediated transformation. The epifluorescence and differential interference contrast images were taken using a confocal laser-scanning microscope, Zeiss model LSM510 (Carl Zeiss). Essential results were reproducibly obtained from three independent transgenic lines.

Recombinant Protein Production and Enzyme Activity Assay
Two fragments SDG25L and SDG25S encoding the amino acid residues 690 to 1,388 and 1,069 to 1,388 of the SDG25 protein were PCR amplified from pCR-SDG25 using the forward primer 5#-atggatccGGATGTGAAAGCAAT-ATTAATATG-3# and 5#-atggatccTCACGTCGGAGGTTATCTCTC-3# with the reverse primer 5#-atctcgagTTAGTTTAGCGATCCACGG-3#, respectively. They were subsequently cloned into BamHI-XhoI sites of pGEX-4T-1 (Amersham-Pharmacia Biotech; http://www.amersham.com), resulting in pGEX-SDG25L and pGEX-SDG25S. Expression and purification of GST-fused proteins from bacteria transformed with pGEX-SDG25L or pGEX-SDG25S were performed according to the previously described procedure (Dong et al., 2005). Histone methyltransferase assays were performed as previously described . The recombinant SDG714 protein that methylates H3K9 (Ding et al., 2007) was used as a control in this study.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Real-time PCR on ChIP samples showing levels of H3K4 and H3K36 dimethylation at FLC in wild type and sdg25-1 mutant.
Supplemental Figure S2. Alignment of amino acid sequences of SDG25 and At5g42400.1.
Supplemental Table S1. List of differentially expressed genes found in the sdg25-1 mutant.