- © 2018 American Society of Plant Biologists. All Rights Reserved.
Abstract
DNA and histone methylation coregulate heterochromatin formation and gene silencing in animals and plants. To identify factors involved in maintaining gene silencing, we conducted a forward genetic screen for mutants that release the silenced transgene Pro35S::NEOMYCIN PHOSPHOTRANSFERASE II in the transgenic Arabidopsis (Arabidopsis thaliana) line L119. We identified MAT4/SAMS3/MTO3/AT3G17390, which encodes methionine (Met) adenosyltransferase 4 (MAT4)/S-adenosyl-Met synthetase 3 that catalyzes the synthesis of S-adenosyl-Met (SAM) in the one-carbon metabolism cycle. mat4 mostly decreases CHG and CHH DNA methylation and histone H3K9me2 and reactivates certain silenced transposons. The exogenous addition of SAM partially rescues the epigenetic defects of mat4. SAM content and DNA methylation were reduced more in mat4 than in three other mat mutants. MAT4 knockout mutations generated by CRISPR/Cas9 were lethal, indicating that MAT4 is an essential gene in Arabidopsis. MAT1, 2, and 4 proteins exhibited nearly equal activity in an in vitro assay, whereas MAT3 exhibited higher activity. The native MAT4 promoter driving MAT1, 2, and 3 cDNA complemented the mat4 mutant. However, most mat4 transgenic lines carrying native MAT1, 2, and 3 promoters driving MAT4 cDNA did not complement the mat4 mutant because of their lower expression in seedlings. Genetic analyses indicated that the mat1mat4 double mutant is dwarfed and the mat2mat4 double mutant was nonviable, while mat1mat2 showed normal growth and fertility. These results indicate that MAT4 plays a predominant role in SAM production, plant growth, and development. Our findings provide direct evidence of the cooperative actions between metabolism and epigenetic regulation.
DNA and histone methylation are important epigenetic modifications that regulate gene expression and genome stability, and can be inherited (Law and Jacobsen, 2010). Compared with animals, which largely display CG methylation, plants present symmetric CG and CHG methylation and asymmetric CHH methylation. In Arabidopsis (Arabidopsis thaliana), DNA methylation is mediated by the RNA-directed DNA methylation pathway (RdDM) and dicer-independent RdDM (Matzke et al., 2015; Yang et al., 2016; Ye et al., 2016); CG methylation is maintained by DNA METHYLTRANSFERASE1 (MET1; a functional equivalent protein of DNA methyltransferase 1 in mammals); CHH methylation is maintained by DOMAINS REARRANGED METHYLASE2 (DRM2) and CHROMOMETHYLASE2 (CMT2; Cao and Jacobsen, 2002a, 2002b; Du et al., 2014); and CHG methylation is maintained by CMT3. CHG methylation is recognized by the SET and RING Associated (SRA) domain histone methyltransferase KRYPTONITE/SU(VAR) homolog 4 (SUVH4) and its homologs SUVH5 and SUVH6 to establish the dimethylation of histone H3 at Lys 9 (H3K9me2) (Jackson et al., 2002; Ebbs et al., 2005; Ebbs and Bender, 2006). H3K9me2 is bound by CMT3 through its H3 tails (Johnson et al., 2007; Bernatavichute et al., 2008; Law and Jacobsen, 2010; Du et al., 2012), which form a reinforcing feedback loop that maintains CHG methylation and H3K9me2.
The one-carbon metabolism pathway plays an important role in epigenetic regulation because it provides methyl groups for most methylation reactions (Fig. 1). The initial methyl group donor is poly-Glu-5-methyl-tetrahydrofolate (5-CH3-THF-Glun), which is the most common form of folate and has a high affinity for folate-dependent methionine (Met) synthase as the methyl-group donor. Folate-dependent Met synthase catalyzes the methylation of homocysteine (Hcy) to Met using 5-CH3-THF-Glun as the methyl-group donor (Friso et al., 2002; Ravanel et al., 2004; Mehrshahi et al., 2010). S-adenosyl-Met (SAM), one of the most abundant cofactors in plant metabolism, is synthesized by Met adenosyltransferase (MAT; also known as S-adenosyl-Met synthetase [SAMS]) using Met and ATP as substrates. After transferring a methyl group to DNA, RNA, proteins or other metabolites by SAM-dependent methyltransferases (Sauter et al., 2013), SAM is changed into S-adenosyl-Hcy (SAH), which competes with SAM and is an inhibitor for many Met synthetases (Molloy, 2012). SAH is then converted to adenosine and Hcy by SAH hydrolase encoded by the HOMOLOGY-DEPENDENT GENE SILENCING1 (HOG1) in Arabidopsis, thus finishing a single cycle of one-carbon metabolism. The T-DNA (hog1-5) or transposon (hog1-4) insertion mutants are zygotic embryo lethal, whereas its weak mutation can cause delayed germination, poor growth, reduced seed viability, and reduced whole-genome DNA and histone methylation (Rocha et al., 2005; Mull et al., 2006; Baubec et al., 2010; Ouyang et al., 2012). A mutation of folylpolyglutamate synthetase 1 (FPGS1) that converts 5-CH3-THF-Glu1 to 5-CH3-THF-Glun in Arabidopsis can slow germination and reduce levels of whole-genome DNA methylation and H3K9me2 (Zhou et al., 2013). Treatment with sulfamethazine, which is a structural analog and competitor of p-aminobenzoic acid, the precursor of folate, causes the release of endogenous transposons and repeat elements and the reduction of DNA methylation levels and H3K9me2 (Zhang et al., 2012). A mutation in the cytoplasmic bifunctional methylenetetrahydrofolate dehydrogenase/methenyltetrahydrofolate cyclohydrolase (MTHFD1) leads to decreased levels of oxidized tetrahydrofolates, DNA hypomethylation, loss of H3K9me, and transposon reactivation (Groth et al., 2016).
Diagram of the methyl-group supply in one-carbon metabolism. Enzymes involved in one-carbon metabolism: MAT/SAMS, Met adenosyltransferase/S-adenosyl-Met synthetase; MT, methyltransferase; SAHH1/HOG1, S-adenosyl-homo-cysteine hydrolase/ homology-dependent gene silencing1; MS, Met synthase; and FPGS, folylpolyglutamate synthetase. FPGS catalyzes the synthesis of 5-CH3-THF-Glun, which provides active methyl group for Hcy for Met synthesis by MS. MAT/SAMS uses Met and ATP as substrates to synthesis SAM, which converts to SAH after methylation reaction of MT. SAHH1/HOG1 can hydrolyze SAH to Hcy.
The Arabidopsis genome has four MAT genes with different nomenclatures (MAT1/SAM1/AT1G02500, MAT2/SAM2/AT4G01850, MAT3/AT2G36880, and MAT4/MTO3/SAMS3/AT3G17390) in different publications (Peleman et al., 1989a, 1989b; Goto et al., 2002; Mao et al., 2015; Chen et al., 2016). A previous study indicates that SAMS RNAi transgenic rice (Oryza sativa) lines with down-regulation of OsSAMS1, 2, and 3 show reduced histone H3K4me3 and DNA methylation (Li et al., 2011). In Arabidopsis, the pollen expressed MAT3 is required for maintaining histone and tRNA methylation in pollen, pollen germination, and pollen tube growth (Chen et al., 2016). However, the biological roles of other MAT proteins in Arabidopsis epigenetic regulation are still unknown.
In this study, we screened a mutagenized population from the transgenic line L119, which harbors two silenced transgenes, Pro35S::NPTII (NEOMYCIN PHOSPHOTRANSFERASE II) and ProRD29A (RESPONSE TO DESSICATION 29A; a stress-inducible promoter)::LUC, and identified the mat4/sams3/mto3 (Met overaccumulation) mutant (Shen et al., 2002; Jin et al., 2017) that releases the silencing of both genes. We found that the mat4 mutant, harboring a missense point mutation, dramatically decreases SAM content and CHG and CHH methylation and H3K9me2, leading to the activation of some transposable elements. Exogenous additions of SAM to the medium partially restored histone methylation levels in mat4. The mat1, 2, or 3 mutants reduced SAM content and DNA methylation to a lesser extent than did mat4, indicating a predominant role of MAT4 among MATs. MAT3 showed the highest activity among the four MAT proteins in an in vitro assay. The expression of MAT4 in seedlings was much higher than MAT1, MAT2, and MAT3. The MAT4 promoter driving MAT1, MAT2, or MAT3 cDNA could complement the mat4 mutant, while most transgenic lines carrying the MAT1, MAT2, or MAT3 promoters driving MAT4 cDNA could not complement the mat4 mutant. The MAT4 loss-of-function mutation generated using the CRISPR/Cas9 technique was lethal. We also found that the MAT proteins in Arabidopsis interacted with each other and themselves both in vitro and in vivo, indicating that they may form homologous or heterogeneous oligomers in Arabidopsis.
RESULTS
Identification and Characterization of mat4
To further study the mechanisms regulating transcriptional gene silencing, we obtained the transgenic line L119 carrying ProRD29A::LUC and Pro35S::NPTII in the Columbia gl1 background. In L119, the transgene loci consist of at least two T-DNA insertions, each with two repeats (Supplemental Fig. S1, A–F). ProRD29A is a stress-inducible promoter that is induced by abscisic acid, low temperatures, and high NaCl concentrations (Yamaguchi-Shinozaki and Shinozaki, 1994). The L119 plants were very sensitive to kanamycin (Kan) and showed little luciferase activity; they grew poorly on media containing 25 mg/L Kan (Fig. 2A) and did not emit any fluorescence after NaCl treatment (Fig. 2D), indicating that both ProRD29A::LUC and Pro35S::NPTII are silenced in L119. However, after introducing the defective in meristem silencing 3 (dms3-1) mutation in the RdDM pathway (Kanno et al., 2008) into L119, ProRD29A::LUC, but not Pro35S::NPTII, was reactivated (Supplemental Fig. S1, G–J), suggesting that similar to the C24/RD29A::LUC line (He et al., 2009a), ProRD29A::LUC is regulated by the RdDM pathway, while Pro35S::NPTII is not.
Identification and characterization of MAT4. A, Kan resistance of mat4 mutants. Seeds were germinated on MS medium or MS supplemented with 25 mg/L Kan. L119 was the transgenic line harboring silenced Pro35S::NPTII and ProRD29A::LUC (proRD29A, an abiotic stress-inducible promoter). ddm1-18 (indicated as ddm1) was selected in the same genetic screening and reactivated both transgenic sites. B, Protein levels of NPTII in L119, mat4, and ddm1 detected by immunoblot. ACTIN was the loading control. C, Transcript levels of transgenic and endogenous loci by real-time RT-qPCR analysis. Transcript levels were normalized to ACTIN2 and relative to L119. Three independent experiments were conducted with similar results. Data are from one experiment with three technical replicates. Error bars are the means ± sd; asterisks indicate significant differences determined by Student’s t test (*P < 0.05, **P < 0.01, ***P < 0.001). D, Silenced ProRD29A::LUC reactivation in mat4 and ddm1. Seedlings of L119, mat4, and ddm1 were treated with 300 mm NaCl for 3 h before detecting the inflorescence signal with a CCD camera (Roper 1300D). E, Identification of MAT4 by map-based cloning. There was a G-to-A mutation, which changed Asp-246 to Asn-246 in AT3G17390. F, Complementation of Kan resistance and delayed germination in mat4 by MAT4. G, NPTII level restored to the basal level of L119 in MAT4-FLAG as determined by immunoblot analyses using anti-NPTII antibodies. ACTIN was the loading control. H, Subcellular localization of MAT4. a, Transgenic line carrying Pro35S::MAT4-GFP in L119; b, transient expression of ProMAT4::MAT4-GFP in a protoplast; c, transient expression of ProMAT4::MAT4-GFP in N. benthamiana leaf epidermal cells. I, Detection of the subcellular localization of MAT4-FLAG after isolating the cytosol and nuclei. PEPC was a marker protein in the cytosol and H3 was a marker protein in the nuclei.
The transgenic line L119 was mutagenized by EMS, and the F2 population was screened for Kan-resistant mutants. A mutant, named mat4-3, was isolated in this screen (hereafter referred to as mat4; Fig. 2A). mat4 seeds germinated later (Fig. 2A) and the seedlings were smaller compared with L119, although these seedlings had relatively normal fertility (Supplemental Fig. S2, A and B). Two alleles of ddm1, ddm1-18 and ddm1-19, were also identified in this system. ddm1-18 (a G-to-A change at position 2803 [counting from the first putative ATG in the coding frame], which causes a stop codon with a TGG-to-TGA transition; hereafter referred to as ddm1) and ddm1-19 (a G-to-A change at position 3125, which causes a stop codon with a TGG-to-TGA transition) exhibited more resistance to Kan than mat4, whereas L119 seedlings did not survive on the medium containing 25 mg/L Kan (Supplemental Fig. S2C). DDM1 is a nucleosome-remodeling protein involved in facilitating DNA methyltransferase access to heterochromatin to silence certain transposable elements and repeats in cooperation with the RdDM pathway (Singer et al., 2001; Zemach et al., 2013). Here, we used ddm1 as a positive control for reduced DNA methylation. Greater increases of NPTII expression and its protein levels were observed in mat4 than in L119, although the NPTII expression and protein levels were less in mat4 than in ddm1 (Fig. 2, B and C). Treatment with 300 mm NaCl reactivated the expression of ProRD29A::LUC in mat4 (Fig. 2D). Two silenced transgenic loci were reactivated in mat4, which was also observed in ddm1; thus, we detected endogenously silenced genes, DNA repeats, and transposable elements. SUPPRESSOR OF DRM1 DRM2 CMT3 (SDC) is regulated by non-CG DNA methylation (Henderson and Jacobsen, 2008); TSIs are endogenous transcriptionally silent information sites regulated by the DNA replication and repair pathway and DNA methylation independent of the RdDM pathway (Steimer et al., 2000; Xia et al., 2006); AtGP1 is a long terminal repeat-Gypsy transposon modulated by the RdDM pathway (He et al., 2009a, 2009b); 180-bp CEN is a centromeric satellite repeat regulated by DNA methylation independent of the RdDM pathway but not by the DNA replication and repair pathway (May et al., 2005; Xia et al., 2006). The transcript levels of all these loci were higher in mat4 than in L119 but lower than in ddm1 (Fig. 2C).
We then cloned the MAT4 gene by map-based cloning. We first crossed the mat4 mutant with the wild-type Ler. The 519 F2 plants that were Kan resistant were isolated and used for mapping. We narrowed mat4 to a region between bacterial artificial chromosome clones K14A17 and MPK6 on chromosome 3. We sequenced candidate genes in this region and observed that a G-to-A mutation in AT3G17390 changed 246D to 246N (Fig. 2E; Supplemental Fig. S2D). This mutation occurs in a conserved amino acid that is involved in binding Met during the reaction, according to the crystal structure of human (Homo sapiens) MAT2A, which is different from the mto3 mutation in the ATP binding site (Shen et al., 2002). This point mutation did not cause alteration in the transcript level (Supplemental Fig. S2E). To confirm whether the mutation in MAT4 is responsible for the Kan resistance of mat4, we transformed the full genomic length of MAT4, including the 2,221-bp promoter and the full genomic sequence fused with FLAG or a GFP tag into the mat4 mutant. A number of different transgenic lines were obtained and shown to be Kan sensitive, which was also observed in L119. We selected one MAT4-FLAG line (line 1) for further study (hereafter referred as MAT4-FLAG). Immunoblotting using anti-FLAG antibodies indicated that MAT4-FLAG expressed the MAT4-FLAG protein (Supplemental Fig. S2F). MAT4-FLAG was sensitive to Kan and had normal germination (Fig. 2F), and the NPTII protein level was restored to the basal level observed in L119 (Fig. 2G), suggesting that AT3G17390 could complement the mat4 mutant phenotype. Reverse transcription quantitative PCR (RT-qPCR) analyses indicated that the expression of NPTII and certain endogenous loci in the MAT4-FLAG also returned to the basal level observed in L119 (Supplemental Fig. S2G). We did not observe any enhanced severe phenotypes of the mat4 mutant after several generations. We used the egg cell-specific CRISPR/Cas9 system in L119 (Wang et al., 2015b) and created two MAT4 knockout lines, mat4-c19 and mat4-c32. mat4-c19 has a fragment deletion from 76 to 180 bp (counting from the first putative ATG) and mat4-c32 from 706 to 797 bp (Supplemental Fig. S2H). However, we were unable to obtain homozygous mat4 mutants, indicating that MAT4 is an essential gene in Arabidopsis.
The subcellular localization in transgenic L119 plants expressing Pro35S::MAT4-GFP indicated that MAT4-GFP was localized in the nucleus and cytosol (Fig. 2H, a). A similar localization of MAT4-GFP was observed in a transient transformation assay using Arabidopsis protoplasts and Nicotiana benthamiana leaves (Fig. 2H, b and c). To avoid the possibility that GFP translocates to the nucleus by itself, we isolated the cytosol and nuclei from L119 and the MAT4-FLAG transgenic line and performed an immunoblot assay. MAT4-FLAG protein was detected in both the cytosol and the nucleus (Fig. 2I). Previous studies also indicate that SAM1/MAT1, SAM2/MAT2, and MAT3 are all localized in both cytosol and nuclei (Mao et al., 2015; Chen et al., 2016). These results suggest that MAT proteins may function in both the cytosol and the nucleus in Arabidopsis.
DNA Methylation of Transgenic Loci Is Reduced in mat4
MAT4 catalyzes the biosynthesis of SAM, which is a universal methyl group donor for DNA and histone methylation; thus, mat4 may reactivate the silenced Pro-35S::NPTII and ProRD29A::LUC because of decreased DNA and/or histone methylation. Bisulfite sequencing analyses indicated that CHG and CHH methylation of transgenic and endogenous RD29A promoters largely decreased in mat4 compared with that in L119 (Fig. 3, B and C), which was consistent with the data from whole-genome bisulfite sequencing (Fig. 3D). However, CHG and CHH methylation decreased to a lesser extent at the 35S promoter in mat4 compared with that in L119 (Fig. 3, A and D). However, only limited changes were observed in the CG methylation with transgenic RD29A and 35S promoters in mat4. In contrast, ddm1 greatly reduced CG methylation and only moderately affected CHG and CHH methylation, except for the transgene RD29A promoter, in which ddm1 did not affect CHH methylation. In addition, the DNA methylation of the MAT4-FLAG line was consistently restored to the L119 level (Fig. 3, A–C). These results indicate that mat4 reduces DNA methylation in transgenes in L119.
DNA methylation of the transgenic and endogenous RD29A promoter in mat4. A, DNA methylation of the 35S promoter region by bisulfite sequencing in L119, mat4, ddm1, and MAT4-FLAG. B, DNA methylation of the transgenic RD29A promoter region by bisulfite sequencing in L119, mat4, ddm1, and MAT4-FLAG. C, DNA methylation of the endogenous RD29A promoter region by bisulfite sequencing in L119, mat4, ddm1, and MAT4-FLAG. D, DNA methylation of the T-DNA insertion region in L119 and mat4 as determined by whole-genome bisulfite sequencing as indicated by IGV software windows.
mat4 Reduces DNA Methylation at the Whole-Genome Level
We compared the DNA methylation level of mat4 with that of L119 at the whole-genome level by bisulfite sequencing. We obtained nearly 5G raw data including adapter and low-quality data for each sample, from which we obtained 4.2G clean data for our subsequent analyses. The total reads were mapped to the genome of TAIR10. We then obtained the methylation level of CG, CHG, and CHH by calculating the ratio of C to C+C/T using the tool of Bismark (Krueger and Andrews, 2011; Fig. 4A). We also included the previously published data of ddm1 (Zemach et al., 2013) for comparison. The methylation levels of CG (22.9%), CHG (4.2%), and CHH (1.4%) in mat4 were lower than the levels of CG (25.2%), CHG (8.2%), and CHH (2.4%) in L119 (Fig. 4A). CHG and CHH methylation in mat4 decreased by nearly half, whereas CG methylation decreased approximately by 9.2%, suggesting that mat4 has different effects on CG, CHG, and CHH DNA methylation (Fig. 4B). mat4 displayed a relatively smaller reduction of CG and CHG methylation and greater reduction of CHH methylation compared with that observed in ddm1 (Fig. 4B). Frequency distribution histograms of significant methylation differences between L119 and mat4 in CG, CHG, and CHH also indicated that the CHG and CHH methylation dramatically decreased in mat4 (Fig. 4C).
Whole-genome DNA methylation levels in mat4. A, Whole-genome DNA methylation levels of CG, CHG, and CHH in L119, ddm1, and mat4. Bisulfite sequencing data for L119 and mat4 were from this study. Data for ddm1 were from a previously published study (Zemach et al., 2013). B, Relative changes in the DNA methylation levels of CG, CHG, and CHH in L119, ddm1, and mat4. C, Frequency distribution histograms of significant methylation differences (P < 0.01) between L119 and mat4 in CG, CHG, and CHH. The histograms were made with 100-bp analyzable windows over the genome-wide scale and the methylation levels of L119 and mat4 in CG, CHG, and CHH context were calculated separately. D, CG, CHG, and CHH methylation of L119, ddm1, and mat4 at genes that do not contain TEs (including 2 kb upstream and downstream). TSS, Transcription start site; TTS, transcription termination site. E, CG, CHG, and CHH methylation of L119, ddm1, and mat4 at TEs that are shorter than 0.5 kb (S-TE), including 2 kb upstream and downstream, and at TE body regions. F, CG, CHG, and CHH methylation of L119, ddm1, and mat4 at TEs that are longer than 4 kb (L-TE), including 2 kb upstream and downstream, and at TE body regions.
To determine the distribution of changes in the DNA methylation patterns in detail, we calculated the DNA methylation 2 kb upstream and downstream of the genes and transposable elements (TEs), respectively. In genes that exclude TEs or repeats, CG methylation mainly occurred in the gene bodies. mat4 reduced the CG methylation in the gene body regions by ∼2.5% (Fig. 4D). However, ddm1 showed greater reductions in CG methylation compared with mat4 in these regions (Fig. 4D). CHG and CHH methylation did not show noticeable changes (Fig. 4D) because these regions exhibit limited CHG and CHH methylation. For TEs, we focused on two TE types: TEs shorter than 0.5 kb (S-TEs; usually regulated by the RdDM pathway) and TEs longer than 4 kb (L-TEs; usually regulated by the DDM1 pathway; Teixeira et al., 2009; Zemach et al., 2013). Generally, mat4 displayed less CG, CHG, and CHH methylation than the L119 in both S-TEs and L-TEs (Fig. 4, E and F). Compared with ddm1, mat4 displayed more CG methylation in both short and long TEs and more CHG methylation in long TEs. However, mat4 showed similar CHG methylation in short TEs and CHH methylation in long TEs as ddm1 (Fig. 4, E and F). mat4 displayed less CHH methylation in short TEs than ddm1, which is consistent with previous studies of DDM1 regulation of DNA methylation in long TEs, but not short TEs that are mostly regulated by the RdDM pathway (Teixeira et al., 2009; Zemach et al., 2013). These results suggest that mat4 mainly reduces CHG and CHH methylation and, to a lesser extent, CG methylation. When mapping these hypo-differentially methylated regions (hypo-DMRs) to the five chromosomes, we found that the distribution of these hypo-DMRs were concentrated around the five centromeres, which displayed a dramatic decrease of CHG and CHH methylation (Supplemental Fig. S3; Supplemental Data Sets S1–S3). These results indicate that mat4 reduces genomic-wide DNA methylation, especially CHG and CHH methylation at pericentromeric heterochromatin regions.
mat4 Decreases Histone Modifications in Heterochromatin Regions
Because DNA methylation, especially CHG and CHH methylation, is reduced throughout the genome in mat4, we sought to determine whether mat4 has an effect on histone methylation. We verified the histone modifications in H3K9me2, H3K9me1, and H3K27me1 because these modifications usually accompany DNA methylation in heterochromatin regions, and we also verified the modifications in H3K4me3 because this modification accompanies high gene expression (Tariq et al., 2003; Jacob et al., 2009). Using immunoblotting assays, we found that the H3K9me2 levels in mat4 were comparable to those in ddm1 and greatly reduced compared with that of L119. In addition, only a small decrease in H3K9me1 was exhibited in mat4, whereas a dramatic decrease was observed in ddm1 compared with L119 (Fig. 5, A and B). Both mat4 and ddm1 had a lower H3K27me1 level than L119. In the mat4 complementary line, H3K9me2, H3K9me1, and H3K27me1 were restored to the wild-type level, whereas in mat4, ddm1, L119, or MAT4-FLAG, H3K4me3 was not changed (Supplemental Fig. S4, A and B).
Histone H3K9me2 and H3K27me1 levels in mat4. A, Immunoblot assays with antibodies against H3K9me1, H3K9me2, and H3K27me1 in L119, ddm1, and mat4. H3 was the loading control. B, Statistical analyses of relative signal intensity in A. We set the signal intensity of L119 as 100 to calculate the relative signal intensity of other mutants. Error bars are the means ± sd (n = 3). Asterisks indicate significant differences determined by Student’s t test (*P < 0.05 and ** P < 0.01). C, Histone methylation patterns of H3K9me1 in the nuclei of L119, ddm1, and mat4 as detected by immunofluorescence assay. D, Histone methylation patterns of H3K9me2 in the nuclei of L119, ddm1, and mat4 as detected by immunofluorescence assay. E, Histone methylation patterns of H3K27me1 in the nuclei of L119, ddm1, and mat4 as detected by immunofluorescence assay. For C to E, on the right, the graphs show the percentage of nuclei with condensed or dispersed signal; gray represents a condensed, and white represents a dispersed signal. n = number of nuclei. DAPI stains the pericentromeric heterochromatin regions. F, Detection of H3K9me2 in L119, ddm1, and mat4 at several selected loci by ChIP combined with RT-qPCR. Three independent experiments were conducted with similar results. Data are from one experiment with three technical replicates. Error bars are the means ± sd (n = 3). Asterisks indicate significant differences determined by Student’s t test (*P < 0.05 and **P < 0.01).
We then compared the heterochromatin status in nuclei using an immunofluorescence assay with different antibodies. In the wild-type cells, more than 89% of the interphase nuclei showed H3K9me2, H3K9me1, and H3K27me1 immunofluorescence associated with the condensed pericentromeric heterochromatin regions stained with 4′,6-diamidino-2-phenylindole (DAPI). However, ∼78% of mat4 and 87% of ddm1 nuclei showed chromocenter decondensation and reduced H3K9me2 immunofluorescence. ddm1 showed strongly reduced H3K9me1 immunofluorescence, which was not observed in mat4, whereas ddm1 and mat4 mutants showed substantially reduced H3K27me1 immunofluorescence (Fig. 5, C–E). In MAT4-FLAG, H3K9me2 immunofluorescence was restored to the L119 level (Supplemental Fig. S4, C and D). However, we did not detect a difference in H3K4me3 immunofluorescence in L119, mat4, and MAT4-FLAG (Supplemental Fig. S4, E and F). We confirmed the decrease of H3K9me2 at certain loci using chromatin immunoprecipitation (ChIP)-PCR (Fig. 5F). These results suggest that mat4 reduces the histone methylation in heterochromatin regions, especially H3K9me2 and H3K27me1.
mat4 Reactivates Silenced TEs
To determine how mat4 modulates gene expression, we performed RNA sequencing (RNA-seq). Total RNA was extracted from 15-d-old seedlings and then subjected to RNA-seq with two biological replicates. We obtained 3G clean data with each replicate, mapped all of the obtained reads to TAIR10, and then compared the transcript levels between mat4 and L119 using edgeR (Robinson et al., 2010). We obtained 1284 protein coding genes and 364 TEs with transcript level changes of at least 2-fold and P < 0.0001. Among these protein-coding genes, 66% (842) were up-regulated and 34% (442) were down-regulated (Fig. 6, A and B; Supplemental Data Sets S4–S7). After mapping these genes on the five chromosomes, we found that they were evenly distributed along the chromosomes arms and rarely localized at the centromere regions (Fig. 6C). Approximately 92% (334) of the differentially expressed TEs were up-regulated and concentrated around the centromeres (Fig. 6C). The expression of certain genes and TEs was confirmed by RT-qPCR in L119, mat4, and MAT4-FLAG (Supplemental Fig. S5, A and B). Compared with previously published data in ddm1 and fpgs1, we found that 127 up-regulated TEs in mat4 were also up-regulated in ddm1 and fpgs1 with reduced DNA methylation and H3K9me2 (Fig. 6D; Zemach et al., 2013; Zhou et al., 2013). After dividing the up-regulated TEs according to their characteristics, two categories of TEs, long terminal repeat/Gypsy and Enhancer/Suppressor Mutator-like transposons (Fig. 6E), accounted for nearly 50% of all of the up-regulated TEs in mat4. Alterations to DNA methylation were not associated with the expression of protein coding genes; however, reduced DNA methylation in mat4 was closely associated with increased TE expression (Supplemental Fig. S5C). When comparing the CHG hypo-DMRs and up-regulated TEs in mat4 with the published data in suvh4/5/6 and cmt3 (Stroud et al., 2013), we found that these mutants had higher overlap than that by chance as viewed in VENNY diagram (Supplemental Fig. S5D), indicating that MAT4 affects a large number of targets shared with those methyltransferases. In conclusion, mat4 led to the activation of the silenced transposons as a result of the reduction in DNA methylation and histone methylation.
Gene expression changes in mat4 by RNA sequencing. A, Differentially expressed genes in mat4 compared with L119. Transcript levels of genes that changed more than 2-fold and had P < 0.0001 were selected. Gene up, up-regulated genes; Gene down, down-regulated genes. B, Differentially expressed TEs in mat4 compared with L119. Transcript levels of TEs that changed more than 2-fold and had a P < 0.0001 were selected. TE up, up-regulated TEs; TE down, down-regulated TEs. C, Distribution of the differentially expressed genes and TEs on the five chromosomes. The purple circle represents the differentially expressed genes, the blue circle represents the differentially expressed TEs, and the green circle represents the differentially methylated regions in mat4. The outer bars indicate the up-regulated genes, TEs, and hyper-DMRs, and the inner bars indicate the down-regulated genes, TEs, and hypo-DMRs; the length of the bars represents the fold change of the genes, TEs and DMRs. The black dots indicate the chromocenters. D, Overlap of up-regulated TEs among mat4, ddm1, and fpgs1. The overlap number was calculated using VENNY2.1. E. Categories of up-regulated TEs in mat4. The diagram shows the percentage of different TE types among the total up-regulated TEs.
Application of SAM Partially Rescues the Phenotype of mat4
To test whether the decreases of DNA and histone methylation were caused by the alteration of SAM content in mat4, we measured SAM contents in mat4, L119, and MAT4-FLAG by liquid chromatography-mass spectrometry (LC-MS). The SAM content in mat4 was decreased by nearly 35% compared to that in L119 (Fig. 7A). Interestingly, the content of SAH, which is a strong inhibitor of SAM-dependent methyltransferases, was significantly increased in mat4, leading to a large decrease in the ratio of SAM/SAH, which is an important index influencing the methylation status (Fig. 7B). Meanwhile, the contents of SAM and SAH in MAT4-FLAG were restored to L119 level (Fig. 7, A and B). Therefore, we sought to determine whether the exogenous application of SAM could rescue the phenotype of mat4. After adding 400 mg/L SAM to the medium, the Kan susceptibility of mat4 was partially rescued (Fig. 7, C and D). An analysis of the transcript levels of transgenic and endogenous genes indicated that the application of SAM inhibited the high expression of these genes in mat4 (Fig. 7E). An immunofluorescence assay indicated that H3K9me2 levels were restored by the application of SAM (Fig. 7, F and G). These results suggest that the decreased SAM in mat4 leads to the release of the silencing of these tested genes.
Application of SAM to rescue the release of silencing in mat4. A, SAM content in mat4 compared with L119 as determined by LC-MS. Three independent experiments were conducted with similar results. Data are from one experiment with three technical replicates. Error bars are the means ± sd, n = 3. Asterisks indicate significant differences determined by Student’s t test (**P < 0.01). B, SAH content in mat4 compared with L119 as determined by LC-MS. Three independent experiments were conducted with similar results. Data are from one experiment with three technical replicates. Error bars are the means ± sd, n = 3. Asterisks indicate significant differences determined by Student’s t test (***P < 0.001). C, Kan resistance of mat4 can be partially rescued by exogenously adding 400 mg/L SAM to medium supplemented with 25 mg/L Kan. D, Statistical results show the survival rate of seedlings grown on the indicated medium. Error bars are the means ± sd (n = 15). Asterisks indicate significant differences determined by Student’s t test (*P < 0.05 and ***P < 0.001). E, Transcript levels of NTPII and endogenous loci by real-time RT-qPCR analysis using the seedlings grown on medium supplemented with 400 mg/L SAM. Three independent experiments were conducted with similar results. Data are from one experiment with three technical replicates. Error bars are the means ± sd. Asterisks indicate significant differences determined by Student’s t test (*P < 0.05, **P < 0.01, and ***P < 0.001). F, Histone methylation patterns of H3K9me2 in L119 and mat4 seedlings grown on MS medium or MS medium supplemented with 400 mg/L SAM as determined by immunofluorescence assays with anti-H3K9me2 antibodies. DAPI staining (blue) was performed on the pericentromeric heterochromatin regions. G, The percentage of nuclei that showed a condensed or dispersed signal. n = number of nuclei.
MAT4 Plays a Predominant Role in SAM Production and DNA Methylation among Different MAT Homologs
In Arabidopsis, four MAT homologs present near 90% identity between each other in their amino acid sequences (Shen et al., 2002; Lindermayr et al., 2006). Because mat4 reduces SAM content and DNA methylation, we sought to determine whether other MAT mutants have similar roles. We obtained three T-DNA lines: SALK_059210 carrying a T-DNA insertion in the C terminus of AT1G02500 (mat1), SALK_052006 carrying a T-DNA insertion in the N terminus of AT4G01850 (mat2), and SALK_019375 carrying a T-DNA insertion after the putative stop codon of AT2G36880 (mat3), which was used in the previous study (Chen et al., 2016). All three T-DNA insertion lines greatly reduced the expression of each targeted gene (Supplemental Fig. S6A). We measured the contents of SAM in these mutants and found that SAM content in mat1 and mat2 decreased only about 6% compared to that in L119, but no clear change was observed in mat3 (Supplemental Fig. S6B), which is consistent with its main expression in pollen (Chen et al., 2016). We further measured the DNA methylation level in these mutants at the whole-genome level by bisulfite sequencing and found that the DNA methylation in CG, CHG, and CHH slightly decreased in these mutants (Supplemental Figure S7A), among which mat3 CHG and CHH methylation was reduced to a lesser extent than mat1 or 2 in L- and S-TEs (Supplemental Fig. S7, B and C), which was consistent with the SAM contents in these mutants. These results indicate that MAT1, 2, and 3 have less of an effect on DNA methylation than MAT4 in seedlings. However, MAT3 may play a major role in pollens (Chen et al., 2016).
In order to clarify the effect of these MAT proteins on expression of Pro35S::NPTII in L119, we created loss of function mutants of MAT1, 2, and 3 genes using the egg cell-specific CRISPR/Cas9 system in the L119 background (Supplemental Fig. S8A; Wang et al., 2015b). mat1-c3 was a single base insertion mutant, in which an A insertion after 421 bp (counting from the first putative ATG), led to a frame-shift mutation; mat1-c19 had a fragment deletion from 222 to 450 bp; mat2-c6 was a single base deletion mutant at 488 bp, leading to a frame-shift mutation; mat2-c13 had a fragment deletion from 490 to 569 bp; mat3-c8 had a fragment deletion from 366 to 420 bp. The expression of each target gene was greatly reduced compared to the wild type (Supplemental Fig. S8A). All these mutants were sensitive to Kan (Supplemental Fig. S8B) and the transcriptional levels of NPTII did not differ with that in L119 (Supplemental Fig. S8C), indicating that the mutations in MAT 1, 2, and 3 do not release the silencing of Pro35S::NPTII, which was consistent with the results of a smaller reduction in DNA methylation in their T-DNA lines. The mat3-c8 mutant produced one to two seeds per silique, which is similar with previous results (Chen et al., 2016).
The Expression Pattern of MAT4 Determines Its Predominant Biological Roles in Arabidopsis
Although the amino acid sequences of the four MATs shared a high percentage of identity, these proteins did not compensate for each other in plants. Whether their expression patterns or protein activities determined their specificity is not known. To address this question, first, we compared the catalytic activities of those four proteins in vitro. The MAT proteins were expressed in and purified from Escherichia coli. MAT1, MAT2, and MAT4 exhibited similar activity, while MAT3 had higher activity than other three (Fig. 8A). The amount of SAM produced increased with increasing MAT4 protein concentration as well (Fig. 8B). However, we could not detect any activity for the MAT4D246N mutant protein (Fig. 8B), indicating that the mutant protein largely loses its activity in vitro.
The catalytic activities of MAT proteins. A, Comparison of the catalytic activities of MAT proteins. The same amount of MAT proteins as indicated by Coomassie staining were added for individual reactions. The reaction that had no protein added was used as a negative control. Three independent experiments were conducted with similar results. Data are from one experiment with three technical replicates. Error bars are the means ± sd (n = 3). B, SAM production with increasing concentrations of MAT4. Protein amounts were indicated by Coomassie staining. The reaction that had no protein added was used as a negative control. Three independent experiments were conducted with similar results. Data are from one experiment with three technical replicates. Error bars are the means ± sd (n = 3). C and D, The GFP fluorescence of mat4 transgenic lines carrying the MAT4 promoter driving MAT1, MAT2, or MAT3 cDNA. The seedlings were grown on MS for 7 d, and the GFP fluorescence in root tips (C) or the whole seedlings (D) was visualized by a confocal microscope (Zeiss LSM 510 META) and a fluorescent microscope (Olympus SEX16), respectively. E, Kan resistance and delayed germination in mat4 was complemented by MAT1, MAT2, or MAT3 driven by the promoter of MAT4.
We used the MAT4 promoter driving MAT1, MAT2, or MAT3 cDNA to evaluate whether these cDNAs could complement mat4. Here, we fused MAT cDNA with GFP to observe its expression. We obtained 12 mat4 transgenic lines carrying ProMAT4::MAT1-GFP, 8 carrying ProMAT4::MAT2-GFP, and 19 carrying ProMAT4::MAT3-GFP. These transgenic plants had high GFP fluorescence. All these transgenic plants had high expression of transgenes and complemented mat4 mutant phenotypes (Fig. 8, C–E). These results indicated that MAT proteins have comparable biological functions in plants.
A previous study indicated that transforming Pro-35S::MAT2 into mto3-1 (the mat4 allele) failed to complement the mto3 phenotype (Shen et al., 2002). Given that MAT1 and MAT2 are expressed in most plant tissues (Peleman et al., 1989a; Mao et al., 2015) and MAT3 is mainly expressed in pollens (Chen et al., 2016), we used MAT1, 2, or 3 promoters driving MAT4 cDNA to examine whether they could complement mat4. We obtained 27 mat4 transgenic lines carrying ProMAT1::MAT4-GFP, 5 carrying ProMAT2::MAT4-GFP, and 6 carrying ProMAT3::MAT4-GFP. We found that all ProMAT2::MAT4-GFP or ProMAT3::MAT4-GFP and most ProMAT1::MAT4-GFP transgenic plants did not complement the mat4 Kan-resistant phenotype because they had lower GFP levels, as indicated by fluorescence imaging and immunoblotting using GFP antibodies (Fig. 9, A–D). In contrast, ProMAT4::MAT2-GFP transgenic lines had higher GFP levels (Fig. 9, A–D). In 27 ProMAT1::MAT4-GFP transgenic lines, 7 lines showed different Kan-sensitive phenotypes. We selected three lines and compared their Kan sensitivity with other lines. These three lines showed more Kan sensitivity than other lines (Supplemental Fig. S9A), indicating that they complemented or partially complemented the mat4 mutant. GFP protein levels were higher in these complemented lines than in noncomplemented lines as indicated by GFP fluorescence and protein immunoblotting (Supplemental Fig. S9, B–D). The expression levels of transgenes were also higher in these complemented lines than others (Supplemental Fig. S9E). The higher expression of ProMAT1::MAT4-GFP may be caused by the different genomic site in which the T-DNA was inserted. These results indicate that the expression level of MAT genes determined their biological roles in Arabidopsis.
MAT4 plays a predominant role in plant growth and development. A, Kan resistance and delayed germination in mat4 was not complemented by ProMAT1::MAT4-GFP, ProMAT2:: MAT4-GFP, or ProMAT3::MAT4-GFP but was complemented by ProMAT4::MAT2-GFP. B, The GFP fluorescence in seedlings of transgenic lines carrying ProMAT1::MAT4-GFP, ProMAT2::MAT4-GFP, ProMAT3::MAT4-GFP, or ProMAT4::MAT2-GFP grown on MS for 7 d. C, Statistical results of the fluorescence intensity of the transgenic lines in B in a fixed area in cotyledons by ImageJ. Error bars are the mean ± sd (n = 12). D, Detection of MAT4-GFP in transgenic lines by immunoblotting using anti-GFP antibodies. ACTIN was the loading control. E, mat1-c19 mat4 double mutant seedlings compared to the wild type (L119) grown in soil under long-day conditions. The mutant did not produce any seeds. F, The siliques of the wild type and mat2-c13(−/−) mat4(+/−). Asterisks indicate the wizened seeds of mat2-c13 mat4 homozygous double mutants. G, Wizened seed percentages in siliques of mat2-c13(−/−) mat4(+/−) heterozygous mutants compared to the wild type.
We performed further genetic analyses among these mutants. mat1mat2 double mutants did not show any growth or developmental differences from the wild-type plants (Supplemental Fig. S10A). mat1-c19 mat4 double mutant seedlings were much smaller than the wild type and did not produce any seeds (Fig. 9E). We could not obtain mat2-c13 mat4 homozygous double mutants because the double mutants had embryonic defects (Fig. 9, F and G; Supplemental Fig. S10B). These results indicate that the expression pattern of MAT4 gene determines its predominant biological roles in Arabidopsis.
MATs Form Homologous or Heterologous Oligomers in Cells
To explore the functions of MAT4, we tried to identify the MAT4-interacting proteins by immunoprecipitation followed by mass spectrometry analysis using the complementary transgenic line MAT4-FLAG. We precipitated MAT4-FLAG with anti-FLAG beads and used the L119 lines as negative controls. We identified MAT1, MAT2, and MAT3, each with unique peptides from MAT4-FLAG coimmunoprecipitation (co-IP) proteins (Supplemental Table S1). Then, we confirmed their interactions in vivo using coimmunoprecipitation assays in Arabidopsis protoplasts transiently expressing different proteins (Fig. 10A). In E. coli, when coexpressing MAT4-His with GST-MAT1, GST-MAT2, GST-MAT3, GST-MAT4, or only GST (as the negative control), respectively, we found that each of them, but not GST, could be purified together with MAT4-His (Fig. 10B), suggesting that MAT4 can interact with MAT1, MAT2, MAT3, and itself in vitro. We further confirmed that MAT1, MAT2, and MAT3 were able to interact with each other and themselves in both in vivo and in vitro assays (Supplemental Figs. S11 and S12). Next, we carried out gel filtration using the proteins isolated from the MAT4-FLAG transgenic complementary line. We observed three peaks from the eluted fractions (Fig. 10C). LC-MS analyses of each peak authenticated four MAT proteins (Fig. 10D), indicating that these MATs can form different sizes of homologous or heterologous oligomer complexes in vivo, which merits further examination.
MAT4 interacts with different MATs in plants. A, MAT4 interacted with MAT1, MAT2, MAT3, or MAT4 itself in a protein co-IP assay. Total proteins were extracted from Arabidopsis protoplasts transiently coexpressing the MAT4-FLAG with MAT1-, MAT2-, MAT3-, MAT4-GFP, or GFP (as a negative control) plasmids and immunoprecipitated with anti-GFP beads. The co-IP proteins were immunoblotted with anti-FLAG and anti-GFP antibodies. B, Protein pull-down assay for MAT4 interaction with MAT1, MAT2, MAT3, or MAT4 itself. Total proteins were isolated from E. coli coexpressing MAT4-His with GST-MAT1, -MAT2, -MAT3, -MAT4, or GST itself (as a negative control) and immunoprecipitated with Glutathione-Sepharose beads. The co-IP proteins were immunoblotted with anti-His and anti-GST antibodies. C, Gel filtration analyses. The 0.5 mg of total proteins extracted from ∼20 g of the 15-d-old seedlings of MAT4-FLAG was applied to an ANTI-FLAG M1 Agarose Affinity Gel. The proteins were eluted using 0.5 µg/µL FLAG Peptide. The elution at the peaks was used for LC-MS analysis. D, LC-MS/MS analyses of the proteins of the three peaks in C. Cov indicates the percentage of sequence coverage (%); Seq (sig) indicates number of significant sequences.
DISCUSSION
SAM provides methyl groups for numerous methyltransferases in transmethylation reactions, including DNA and histone methylations in all living cells. In this study, we identified MAT4 because its mutation reactivates the silenced Pro35S::NPTII and ProRD29A::LUC in L119. MAT is well conserved during evolution, and it usually has three domains: the N-terminal domain, the C-terminal domain, and the central M-domain (Takusagawa et al., 1996). In Arabidopsis, there are four homologs of MAT, MAT1 to MAT4; these homologs share nearly 90% amino acid sequence identity (Peleman et al., 1989a, 1989b; Shen et al., 2002). mto3, an allele of mat4, was isolated in a screen based on ethionine (a toxic analog of Met) sensitivity. The level of Met is increased more than 200-fold and the concentration of SAM is decreased by 35% compared with the wide type (Shen et al., 2002). In this study, we found that both DNA and histone methylation were largely reduced as a result of the decrease of SAM content in mat4. Our study provides direct evidence for the importance of SAM in providing methyl donors and modulating epigenetic status.
In theory, SAM is a general methyl group donor and the reduction of SAM should have an unbiased effect on DNA and histone methylations. However, we found that the reduction of DNA and histone methylation was uneven, with mat4 showing large decreases in CHG and CHH methylation as well as H3K9me2 and H3K27me1 (Figs. 4 and 5). Changes in H3K4me3 were not observed and CG methylation decreased to a lesser extent. In animals, the supplementation of Met, an essential amino acid, can modulate the SAM/SAH ratio and impact H3K4me3 (Mentch et al., 2015). Threonine (Thr), another essential amino acid, is the major fuel source for glycine, acetyl-CoA, and SAM. Restricted applications of Thr can reduce H3K4me3 levels, which results in slower growth and increased differentiation in mouse embryonic stem cells (Shyh-Chang et al., 2013). In addition, in SAMS RNAi rice, H3K4me3 is significantly reduced (Li et al., 2011). These results suggest that SAM limitation can result in different changes in DNA and histone modifications in different species. These differences can be explained by several factors. First, different methyltransferases might have different SAM concentration thresholds. In addition, MET1 and H3K4 methyltransferases might efficiently use low concentrations of SAM to complete the reactions in mat4, whereas the histone methyltransferases for H3K9me2 and DNA methyltransferases for CHG and CHH might have lower activity at such concentrations. Second, the reinforcing loop between CHG methylation and H3K9me2 cannot be maintained and is even disrupted in mat4, which would lead to a serious reduction in methylation for both. Third, the increased SAH in mat4 would compete with SAM and decrease SAM accessibility to methyltransferases, which mostly reduced the CHG and CHH methylation and H3K9me. Similar results have been observed in both fpgs1 and mthfd1 mutants (Zhou et al., 2013; Groth et al., 2016). Both fpgs1 and mthfd1 mutants accumulate relatively more SAH, which leads to a decreased ratio of SAM/SAH (Zhou et al., 2013; Groth et al., 2016). SAH is a strong inhibitor that competes with SAM for SAM-dependent transmethylation (De La Haba and Cantoni, 1959). These studies suggest that the CMT3 or the CMT2 pathway has a positive feedback circuit with SUVH4 (KRYPTONITE)/5/6 to maintain CHG or CHH methylation and H3K9me (Zhou et al., 2013; Stroud et al., 2014; Groth et al., 2016). Consistent with the reduced DNA methylation, our RNA-seq data indicated that a large number of TEs were activated in the mat4 mutant. These up-regulated TEs were enriched around the heterochromatin regions of the centromeres, which are largely shared with those found in ddm1 and fpgs1 mutants. Given that reduced DNA methylation is found only at certain sequence contexts, it is also possible that the inefficient histone methylation might indirectly affect DNA methylation. For example, the CHG DNA methylation in the Pro35S-NPTII transgene was only moderately reduced, while a more significant reduction in H3K9me2 was found in the 35S promoter of the mat4 mutant. However, this hypothesis is hard to test as SAM is a common substrate for both DNA and histone methylation. Reduced SAM must more or less affect both DNA and histone methylation.
In humans, three MAT genes encode MATα1, MATα2, and MATβ. MATα1 and MATα2 can form homodimers or tetramers that have different affinity for substrates, and MATα2 can interact with MATβ to strengthen the activity of (MATβ)4 (Murray et al., 2014). In Saccharomyces cerevisiae, when two MATs that share 92% identity in amino acid sequence are disrupted, the mutants display opposite phenotypes to the excess ethionine added in the growth medium (Thomas and Surdin-Kerjan, 1987, 1991; Thomas et al., 1988), indicating that different MAT isoforms act on their own rhythms. There are four close MAT homologs in Arabidopsis. However, we found that in the in vitro assays MAT3 has the highest activity, while MAT1, MAT2, and MAT4 have comparable but lower activities. Among them, MAT4 is predominant as its missense mutation reduces SAM and DNA methylation to greater extent than MAT1 and MAT2 loss-of-function mutations and can release the silencing of Pro35S::NPTII, while other mutations cannot. We used the CRISPR/Cas9 technique, but failed to get the loss-of-function homozygous mutant of MAT4. These results indicate that MAT4 is an essential gene for plant growth and development in Arabidopsis. We found that the MAT4 promoter driving different MAT cDNAs can complement mat4 mutants. However, only a few mat4 mutants can be complemented by the MAT1 promoter driving MAT4 cDNA, which might be caused by high expression of ProMAT1::MAT4-GFP in these transgenic lines, likely because the T-DNAs were inserted in environment-friendly sites in the genomic region. Nevertheless, mat4 mutants were not complemented in several ProMAT2::MAT4-GFP and ProMAT3::MAT4-GFP transgenic lines. These results indicate the expression pattern of MAT4, but not MAT4 protein itself, is important for its predominant biological roles in Arabidopsis. Using pull-down assays and co-IP assays, we found that MATs interacted with each other both in vitro and in vivo, suggesting that MATs can also form homo- and/or hetero-oligomers of different sizes in Arabidopsis. However, more attempts or even crystallographic structural analyses should be carried out to obtain more information about their precise composition in Arabidopsis.
MATERIALS AND METHODS
Plant Growth Conditions, Mutant Screening, and Identification
Arabidopsis (Arabidopsis thaliana) seeds were sterilized with 0.5% NaClO and then sown into Murashige and Skoog (MS) medium, which contained 2% (w/v) Suc and 0.8% (w/v) agar. After 3 d at 4°C, the plates were transferred to growth chambers with long-day conditions (23 h of light/1 h of dark) at 22°C. Generally, 10-d-old seedlings were transferred to soil and cultured in a greenhouse with long-day conditions (16 h of light/8 h of dark) at 20°C.
We used L119, which harbors two transgenes, ProRD29A::LUC and Pro35S::NPTII, as the wild-type line. The mutants were selected from an EMS-mutagenized population of L119 for resistance to 25 mg/L Kan, whereas the L119 lines are typically sensitive under this condition. Map-based cloning was conducted to identify the mutation. We crossed our mutants with Ler and obtained the F2 population, from which we selected Kan-resistant lines (519 total) for mapping.
For mutant complementation, the full genomic length of MAT4, including the 2,221-bp promoter and the overall genomic sequence, was cloned into pCAMBIA1307. The construct in Agrobacterium tumefaciens strain GV3101 was transformed into mat4 by the floral dip method (Clough and Bent, 1998). The homozygous transgenic lines were selected on MS medium supplemented with 30 mg/L hygromycin from the next T2 generation. All of the primers used in this study are listed in the Supplemental Table S2.
RNA Analysis
For real-time RT-qPCR, total RNA was isolated using TRIzol reagent (Invitrogen) from 15-d-old seedlings, and 4 µg RNA was reverse-transcribed into cDNA using the GOSCRIPT reverse transcription system (Promega A5001). Then, 2 µL of diluted (10×) cDNA mixture was used as the template for a PCR assay using 20 µL of SYBR Green Master Mix (TaKaRa) performed on a Step One Plus system (Applied Biosystems). The experiments were performed in three independent biological replicates with technical triplicate. All of the primers used in the real-time PCR assay are listed in Supplemental Table S2.
Subcellular Localization
The full-length cDNA of MAT4 fused with a GFP tag under the control of the super promoter was constructed in the pCAMBIA 1300 vector and the full genomic length of MAT4, including the 2,221-bp promoter and the overall genomic sequence, was cloned into pCAMBIA1300. The plasmids were extracted and purified with the Plasmid Maxprep Kit (VIGOROUS N001). Then, we introduced the plasmids into Arabidopsis protoplasts as previously described (Kong et al., 2015). After 14 to 16 h of incubation in light, the protoplasts were viewed using a confocal microscope (Zeiss LSM 510 META), and the GFP signal was detected with 488-nm excitation. Empty GFP plasmids were used as a control. A. tumefaciens strain GV3101 carrying the same constructs were also injected into Nicotiana benthamiana leaves. After 48 to 72 h of incubation, a section of the injected leaves was examined using the confocal microscope, and the GFP signal was detected using 488-nm excitation. An empty GFP construct was used as a control. We also obtained transgenic lines by transforming this construct into L119 using A. tumefaciens strain GV3101. The 7-d growth of the T2 homozygous transgenic plants was used to detect fluorescence signals using a confocal microscope (Zeiss LSM 510 META) with 488-nm excitation.
Cellular Distribution of MAT4-FLAG
Next, 0.1 g of 15-d-old seedlings was ground into powder in liquid nitrogen and suspended with 200 µL isolation buffer (0.4 m Suc, 10 mm Tris-HCl, pH 8.0, 10 mm MgCl2, 5 mm β-mercaptoethanol, and 1 mm PMSF), then filtered through microcloth (Calbiochem 475855-1R), and the flow-through was centrifuged at 2,800g for 10 min at 4°C. The supernatant was used for the cytosol, while the precipitate was used for the nuclei after four washes with isolation buffer. Then immunoblotting using antibodies (H3, Millipore, 07-690; FLAG, Sigma-Aldrich, F3165; PEPC, Agrisera, AS09458) was carried out. Here, phosphoenolpyruvate carboxylase (PEPC) was used as the cytosol marker, while H3 was used as the nucleus marker.
Bisulfite Sequencing
Genomic DNA was extracted using the DNeasy Plant Mini Kit (Qiagen 69104) from 15-d-old seedlings. The EZ Methylation-Gold Kit (Zymo Research D5005) was used to analyze DNA methylation. Five hundred nanograms of DNA was added to the reaction, and all steps followed the protocol supplied in the kit. Nearly 50 ng treated DNA was added to the PCR reaction using the specific primers listed in Supplemental Table S2. The PCR products were introduced into the pMD18-T simple vector (Takara 6011), and at least 15 clones were sequenced for each sample.
Histone Extraction and Immunoblotting
Histone proteins were extracted from 15-d-old seedlings following the protocol as previously described (Li et al., 2012). The antibodies used in immunoblotting were H3 (Millipore; 07-690), H3K9me1 (Millipore; 17-680), H3K9me2 (Abcam; ab1220), H3K27me1 (Millipore; 07-448), and H3K4me3 (Millipore; 07-473). H3 was used as the loading control. The experiments were performed in three independent biological replicates.
Histone Immunofluorescence Staining Assay
The assay mainly followed a previously described process (Soppe et al., 2002) with subtle modifications. The nuclei were isolated from 15-d-old seedlings. After resuspending with sorting buffer, the nuclei were dropped on the slides to air dry. The nuclei were then postfixed using 4% paraformaldehyde in PBS for 20 min, washed four times with PBS, and closed with signal enhancer (Cell Signaling; 11932) for 30 min at room temperature. After washing four times, the plates were incubated with primary antibody for 2 h at 37°C or overnight at 4°C covered with Parafilm. Then, the plates were washed four times in vats filled with PBST (PBS added with 0.1% Tween 20), each for 5 min, and incubated with secondary antibody in dark for 1.5 h at room temperature. Then, the plates were washed four times in vats filled with PBST. Eight microliters of DAPI (1 µg/mL) was added onto the slides to counterstain the nuclei. The slides were covered with cover glasses. The signal was observed with a confocal microscope (Leica sp5) and collected under the emission wavelength of 405 and 561 nm for DAPI and Rhodamine.
The primary antibodies used in this assay were H3K9me1 (Millipore; 17-680, 1:50, rabbit), H3K9me2 (Abcam; ab1220, 1:100, mouse), H3K27me1 (Millipore; 07-448, 1:100, rabbit), and H3K4me3 (Millipore; 07-473, 1:100, rabbit). Secondary antibodies used in this assay were Rhodamine Red conjugate-Goat anti-mouse (Invitrogen R6393) and Rhodamine Red conjugate-Goat anti-rabbit (Invitrogen R6394).
Whole-Genome Bisulfite Sequencing and Analyses
Genomic DNA was extracted from 15-d-old seedlings using DNeasy Plant Mini Kit (Qiagen 69104). Two micrograms of DNA was used for bisulfite treatment and library construction, and MethylC-seq was carried out using the HiSeq 2000 (Illumina).
Raw data were obtained from the whole-genome bisulfite sequencing using the Illumina HiSeq platform. Clean data were generated by trimming adaptor bases and removing low-quality reads. For data analysis, paired-end clean reads were mapped to the reference genome sequence of the Arabidopsis genome (TAIR10) with Bismark (Krueger and Andrews, 2011). The DMRs were determined and identified as previously published (Zhao et al., 2014).
For investigation of DMR enrichment, we followed the previously described analysis with some modifications (Zhao et al., 2014). The DNA methylation level in genes without TEs, S-TEs (shorter than 0.5 kb), and L-TEs (longer than 4 kb) were calculated.
RNA-Seq and Analysis
Total RNA was extracted from 15-d-old seedlings using the RNeasy Plant Mini Kit (Qiagen 74904). Two micrograms of RNA was used for library construction, each sample with two replicates. The transcriptome data set used in this study was obtained using the Illumina HiSeq platform, and 125-bp trimmed paired-end reads with high quality were generated. The trimmed reads were mapped to the reference genome sequence of the Arabidopsis genome (TAIR10) using bowtie2 (http://computing.bio.cam.ac.uk/local/doc/bowtie2.html) with default settings (Langmead et al., 2009). Differential gene expression analyses were performed using edgeR (http://bioinf.wehi.edu.au/edgeR/; Robinson et al., 2010). We selected genes with fold change > 2 and P < 0.0001 compared to the wild type as differential expression genes and TEs. The distribution of differentially expressed genes and TEs in the chromosomes was plotted by circos (Krzywinski et al., 2009). The categories of up-regulated TEs were divided as previously described (Wang et al., 2015a).
ChIP Assays
Nuclei were isolated from 15-d-old seedlings and fixed with 1% formaldehyde, following the protocol as described previously (Saleh et al., 2008). The pure nuclei were resuspended with 300 µL of cold nuclei lysis buffer, then the genomic DNA was sonicated into 250- to 500-bp fragments, and the supernatant was diluted with ChIP dilution buffer. Twenty microliters of protein A/G Magnetic beads (Millipore; 16-663) was added for 90 min at 4°C to decrease nonspecific combination with gentle rotation. All steps that needed to collect beads were carried out on a Magnetic rack on ice. The antibody H3K9me2 (Abcam; ab1220) was added with gentle rotation over night at 4°C to allow combination. The magnetic beads were washed five times: one time with low salt wash buffer, one time with high salt wash buffer, one time with LiCl wash buffer, and two times with TE buffer. Each wash was 5 min at 4°C with gentle rotation. The protein and DNA complex were eluted by elution buffer at 65°C and then incubated at 65°C for at least 6 h or overnight to reverse cross-linking. Next, the RNAs were digested using RNase A at 37°C for 2 h, and then the proteinase K was added to digest the protein at 65°C at least for 6 h. The QIAquick PCR purification kit (Qiagen; 28106) was used to obtain high quality DNA, then the concentrations of the DNA measured with Qubit Fluorometer 3.0 (Invitrogen; Q33216) were adjusted to 50 pg/µL. Finally, 1 µL DNA was used as the template in 20 µL of SYBR Green Master mix (TaKaRa) on a Step One Plus machine (Applied Biosystems). The experiments were performed in three independent biological replicates.
Measurement of SAM Contents by LC-MS
Sixteen-day-old seedlings were used for the subsequent measurements. The extraction and determination methods were followed as previously described with some modifications to extraction (Nikiforova et al., 2005). We added 300 µL methanol (v/v: 80% and precooled at −20°C) and 100 µL precooled methanol with 15 mg/mL DTT to 100 mg plant samples that had been ground into powder in liquid nitrogen, vortexed for 1 min, extracted for 30 min on ice and then centrifuged at 4°C for 10 min at 12,000g. We added 300 µL precooled isopropanol and 100 µL precooled methanol with 15 mg/mL DTT to resuspend the precipitation, extracted for 30 min on ice, and then centrifuged at 4°C for 10 min at 12,000g. The two supernatants were combined, filtered, and then analyzed via LC-MS. The experiments were performed in three independent biological replicates with technical triplicate.
Acquisition of Knockout Mutants Using the CRISPR/Cas9 Assay
The targets were selected according to the http://www.crisprscan.org/?page=sequence website. Two targets were chosen for each gene, and primers were designed. Fragments were amplified by PCR using pCBC-DT1T2 as a template (Wang et al., 2015b). After PCR products were purified, they were digested using the restriction enzyme BsaI and ligated using T4 ligase in one system for 5 h at 37°C, 5 min at 50°C, and 10 min at 80°C. They were then transformed into the competent cell of JM109. After incubation, the correct clone was identified. The construct in A. tumefaciens strain GV3101 was transformed into L119 by the floral dip method (Clough and Bent, 1998). The transgenic lines were selected on MS medium supplemented with 30 mg/L hygromycin from the T1 generation and sequenced to obtain knockout lines.
Determination of the Activities of MATs
The activities of MATs were determined by measuring the production amount of SAM after reactions. The purified proteins of MAT1-His, MAT2-His, MAT3-His, MAT4-His, and MAT4 (D246N)-His in Escherichia coli were desalted using a 10-kD Centrifugal Filter Unit (Millipore UFC501096). The reaction was carried out in a 200 µL mixture that included 40 µg MAT, 10 mm ATP, 5 mm Met, 0.1 m Tris-HCl (pH 8.0), 0.02 m MgCl2, and 0.2 m KCl at 37°C for 20 min, and then the reaction was terminated by adding 800 µL of 75% acetonitrile and 1.2% formic acid. The reaction solution was transferred to a 10-kD Centrifugal Filter Unit and centrifuged at full speed in 4°C for 10 min. The solution was used for LC-MS analysis. The experiments were performed in three independent biological replicates with technical triplicate.
Pull-Down Assays
The full-length cDNA of MAT1, MAT2, MAT3, and MAT4 fused with His and GST tags were constructed in PET30a and pGEX-4T-1. The relevant plasmids were cotransformed into the Rosetta (DE3) strain of E. coli. The Glutathione-Sepharose beads were used to purify the proteins. Then the proteins were eluted from the Glutathione-Sepharose beads using 10 mm reduced GSH in 50 mm Tris-HCl. The cell lysates before addition of Glutathione-Sepharose beads were used as input to detect whether two proteins were both expressed. Then products were detected by immunoblot using the antibodies of His and GST.
Co-IP Assays in Arabidopsis Protoplasts
The full-length cDNAs of MAT1, MAT2, MAT3, and MAT4 fused with FLAG and GFP tags under the control of the super promoter were constructed in the pCAMBIA 1300 vector. The plasmids were extracted and purified with the Plasmid Maxprep Kit (VIGOROUS N001). Then, the two relevant plasmids were cointroduced into Arabidopsis protoplasts as previously described (Kong et al., 2015). After 14 to 16 h of incubation in light, the protoplasts were collected and total protein was extracted using the immunoprecipitation buffer (50 mm Tris-HCl, pH 7.6, 5 mm MgCl2, 150 mm NaCl, 10% glycerol, 0.5% Nonidet P-40, 1 mm DTT, 1 mm PMSF, and protease inhibitor cocktail, 1:100, 1 plate per mL [Roche]) for 30 min on ice. The protein solution was centrifuged at 12,000g for 15 min at 4°C. Ten microliters of GFP-Trap_A (Chromotek; gta-20) was added to the supernatant, then gently rotated for 2.5 h at 4°C to allow combination. The GFP-Trap_A was washed five times using the immunoprecipitation buffer and then the immunoprecipitated products were detected by immunoblot using the antibodies of GFP and FLAG.
Gel Filtration Assay
MAT4-FLAG was used to prepare the protein for gel filtration. More than 20 g of MAT4-FLAG seedlings grown on MS medium for 15 d were collected and then ground to a powder in liquid nitrogen. The protein was extracted using IP buffer (10 mm HEPES, 1 mm EDTA, pH 8.0, 100 mm NaCl, 10% glycerol, 0.5%Triton X-100, 1 mm DTT, 1 mm PMSF, and protease inhibitor cocktail, 1 plate per mL [Roche]) for 30 min on ice. Then the protein solution was centrifuged at 12,000g for 15 min at 4°C. The supernatant was added to the ANTI-FLAG M1 Agarose Affinity Gel (Sigma-Aldrich; A4596), then gently rotated for 2.5 h at 4°C to allow combination. The FLAG Agarose was washed five times using the IP buffer, and the protein was eluted using 0.5 µg/µL FLAG Peptide (Sigma-Aldrich; F3290). We prepared 0.5 mg protein for gel filtration analysis. The effluent of the indicated peaks was sent for LC-MS analysis.
Accession Numbers
The gene accession numbers that were used in this study are as follows: AT3G17390 (MAT4/SAM3/MTO3), AT1G02500 (MAT1/SAM1), AT4G01850 (MAT2/SAM2), AT2G36880 (MAT3), DDM1 (AT5G66750), AT3G18780 (ACTIN), AT5G52310 (RD29A), AT2G17690 (SDC), AT4G03650 (AtGP1), and BD298459.1 (TSIs). RNA-seq, and BS-seq data were deposited in the National Center for Biotechnology Information Gene Expression Omnibus database under accession number GSE84014.
Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. T-DNA insertion positions in L119.
Supplemental Figure S2. Growth phenotypes of mat4 mutants and Kan-resistant phenotypes of two ddm1 alleles.
Supplemental Figure S3. Effects of mat4 on DNA methylation throughout the five chromosomes.
Supplemental Figure S4. Complementation of reduced histone modification in mat4 by MAT4-FLAG.
Supplemental Figure S5. Confirmation of RNA-seq data by RT-qPCR and association between DNA methylation and gene expression.
Supplemental Figure S6. T-DNA insertion positions, expression levels, and SAM contents in mat1, mat2, and mat3.
Supplemental Figure S7. Whole genomic DNA methylation changes in mat1, mat2, and mat3.
Supplemental Figure S8. Kanamycin sensitivity of mat1, mat2, and mat3 CRISPR/Cas9 mutants in L119.
Supplemental Figure S9. Complementation of mat4 by MAT4 driven by the native MAT1 promoter.
Supplemental Figure S10. Growth and development phenotypes of mat1mat2 double mutants and embryogenic defects of mat2-c19mat4 double mutants.
Supplemental Figure S11. The interaction of MAT1, MAT2, and MAT3 with different MATs as determined by coimmunoprecipitation assays.
Supplemental Figure S12. The interaction of MAT1, MAT2, and MAT3 with different MATs as determined by protein pull-down assays.
Supplemental Table S1. LC-MS/MS analyses of affinity copurified proteins from MAT4-FLAG seedlings.
Supplemental Table S2. Primers used in this study.
Supplemental Data Set S1. Hypo-differentially methylated regions of CG in mat4.
Supplemental Data Set S2. Hypo-differentially methylated regions of CHG in mat4.
Supplemental Data Set S3. Hypo-differentially methylated regions of CHH in mat4.
Supplemental Data Set S4. Differentially expressed genes up-regulated in mat4.
Supplemental Data Set S5. Differentially expressed genes down-regulated in mat4.
Supplemental Data Set S6. Differentially expressed TEs up-regulated in mat4.
Supplemental Data Set S7. Differentially expressed TEs down-regulated in mat4.
Acknowledgments
We thank Dr. Zhen Li and Dr. Zhongzhou Chen of China Agricultural University for assistance with LC-MS analyses and gel filtration, respectively.
Footnotes
The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (www.plantphysiol.org) is: Zhizhong Gong (gongzz{at}cau.edu.cn).
Z.G. conceived the original research plans; J.M. performed most of the experiments; LW. performed the mutant screening and gene cloning; J.C. provided bioinformatics analysis; X.Z., J.W., D.J., W.Y., and Q.L. assisted with some experiments. J.M. and Z.G. designed the project and wrote the article with contributions from all the authors.
↵1 This study was supported by the National Natural Science Foundation of China (31330041).
↵2 These authors contributed equally to the article.
↵[OPEN] Articles can be viewed without a subscription.
- Received February 13, 2018.
- Accepted March 16, 2018.
- Published March 23, 2018.