HDA6 directly interacts with DNA methyltransferase MET1 and maintains transposable elements silencing in Arabidopsis

The molecular mechanism of how histone deacetylase HDA6 participates in maintaining transposable elements (TEs) silencing in Arabidopsis is not yet defined. In the present study, we show that a subset of TEs was transcriptional reactivated, and TE reactivation was associated with elevated histone H3 and H4 acetylation as well as increased H3K4Me3 and H3K4Me2 in hda6 mutants. Decreased DNA methylation of the TEs was also detected in hda6 mutants, suggesting that HDA6 silences the TEs by regulating the histone acetylation and methylation as well as DNA methylation status of the TEs. Similarly, transcripts of some of these TEs were also increased in the met1 mutant, with decreased DNA methylation. Furthermore, H4 acetylation, H3K4Me3, H3K4Me2 and H3K36Me2 were enriched at the co-regulated TEs in the met1 and hda6 met1 double mutants. Protein-protein interaction analysis indicated that HDA6 physically interacts with MET1 in vitro and in vivo , and further deletion analysis demonstrated that the C-terminal region of HDA6 and the BAH domain of MET1 were responsible for the interaction. These results suggested that HDA6 and MET1 interact directly and act together to silence TEs by modulating DNA methylation, histone acetylation and histone methylation status. was reactivated in the hda6 axe1-5 and sil1 . In the histone acetylation, histone methylation and DNA methylation status of these TEs were affected in hda6 mutants. Direct protein-protein interaction between HDA6 and MET1 was detected by yeast two-hybrid, bimolecular fluorescence complementation (BiFC) and GST pull down assays. Furthermore, the landscapes of histone modification and DNA methylation at the TEs were also studied in met1-3 and axe1-5 met1-3 plants. Our results indicate that HDA6 and MET1 interact directly and act together to maintain


Introduction
Transposable elements (TEs) constitute a major part of the complex genomes of plants and animals. The genome sequence analysis of Arabidopsis has revealed that TEs are enriched in the centromeric region of chromosomes, which is highly methylated and packed into heterochromatin (Initiative, 2000). TEs are classical models for epigenetic inheritance and silent transposons can be activated and inherited in the active state (Lippman et al., 2003). Recent studies revealed that DNA methylation, histone deacetylation, histone methylation and RNA interference are involved in activation or silencing of TEs (Murfett et al., 2001;Johnson et al., 2002). DNA methylation in mammalian genome occurs predominantly in the context of CG sequence and is maintained by the DNMT1 methyltransferase. In Arabidopsis, the DNMT1 homolog, MET1, plays vital roles in maintaining cytosine methylation (Murfett et al., 2001). Antisense suppression or mutation of MET1 in Arabidopsis causes a global reduction in cytosine methylation, particularly at CG sites (Finnegan et al., 1996;Ronemus et al., 1996), and induces the release of TEs and transcriptional gene silencing (Kankel et al., 2003;Saze et al., 2003). In addition to DNA methylation, TEs are also subjected to the regulation mediated by histone deacetylation and methylation and RNA interference (RNAi) (Lippman et al., 2003).
DNA methylation and histone deacetylation are two major epigenetic marks that contribute to the stability of gene expression status (MacDonald and Roskams, 2009). In several cases, gene silencing has been reported to be relieved by treatment with either histone deacetylase (HDAC) inhibitors or an inhibitor of DNA methylation (Chen and Pikaard, 1997;Pikaart et al., 1998;Selker, 1998). Inhibiting cytosine methylation induces histone acetylation, whereas inhibiting histone deacetylation causes the loss of cytosine methylation (Lawrence et al., 2004). In mammals, HDACs and DNMTs were suggested to act in same protein complexes. direct recruitment by the DNA methyltransferase DNMT1 (Fuks et al., 2000), suggesting a tight interplay between histone deacetylation and DNA methylation.
In Arabidopsis, the Histone Deacetylase 6 (HDA6), a class I RPD3-like HDAC, is required for TE and rRNA gene silencing and cytosine methylation maintenance (Lippman et al., 2003;Earley et al., 2006;Earley et al., 2010). HDA6 was first identified to be involved in transgene silencing through an auxin-responsive-element mutant screening (Murfett et al., 2001). HDA6 mutant alleles axe1-1 to axe1-5 displayed increased expression of the auxin-responsive reporter genes in the absence of auxin treatment, suggesting a role of HDA6 in gene silencing (Murfett et al., 2001). Another hda6 mutant allele, sil1, was also identified in a screen for mutations releasing transgene silencing, indicating that HDA6 is required for maintenance of transcriptional gene silencing (Probst et al., 2004). Furthermore, HDA6 was also identified as an essential component in RNA-directed DNA methylation (RdDM) (Aufsatz et al., 2002;Aufsatz et al., 2007). The HDA6 mutant allele rts1 exhibits reactivation of RdDM-silenced promoters and results in reduced cytosine methylation in symmetric sequence contexts, highlighting a function for HDA6 in methylation maintenance (Aufsatz et al., 2002;Aufsatz et al., 2007). More recently, To et al. (2011) reported that HDA6 regulates locus-directed heterochromatin silencing in cooperation with MET1. In addition, HDA6 and MET1 co-target to the heterochromatin sites and maintain heterochromatin silencing (To et al., 2011). However, the molecular mechanism underlying the function of HDA6 in gene silencing is still unclear.
In the present study, we found that a subset of TEs was reactivated in the hda6 mutants, axe1-5 and sil1. In addition, the histone acetylation, histone methylation and DNA methylation status of these TEs were affected in hda6 mutants. Direct protein-protein interaction between HDA6 and MET1 was detected by yeast two-hybrid, bimolecular fluorescence complementation (BiFC) and GST pull down assays. Furthermore, the landscapes of histone modification and DNA methylation at the TEs were also studied in met1-3 and axe1-5 met1-3 plants. Our results indicate that HDA6 and MET1 interact directly and act together to maintain TEs silencing by modulating their histone acetylation, methylation and DNA methylation status.

Results
HDA6 maintains TEs silencing through modulating histone H3 and H4 acetylation as well as histone H3K4 methylation Our previous study revealed that a subset of TEs was up-regulated in axe1-5 and HDA6-RNAi plants (Yu et al., 2011). An additional hda6 mutant allele, sil1, in Col background (Probst et al., 2004) was also obtained and the expression of these transposons in the sil1 and axe1-5 were compared. As shown in Fig. 1, all eight transposons were remarkably activated in both axe1-5 and sil1 mutants, further supporting the notion that HDA6 is required for the silencing of TEs.
To investigate whether the reactivation of TEs was caused by histone acetylation, we measured the histone H3 and H4 acetylation levels of the TEs surrounding their transcription starting sites in axe1-5 and sil1 mutants by chromatin immunoprecipitation assay (ChIP) using antibodies specific for acetylated histone H3K9K14 and H4K5K8K12K16, respectively (Koch et al., 2008). As shown in Fig. 2A and B, the histone H3 and H4 acetylation levels of these TEs were elevated in axe1-5 and sil1 mutants compared to the wild-type, suggesting that HDA6 may silence TEs by histone deacetylation.
Since an increase in histone acetylation is often correlated with the methylation at lysine 4 of histone H3 (H3K4) (Strahl and Allis, 2000), we further analyzed H3K4Me3 and H3K4Me2 levels of the up-regulated TEs in axe1-5 and sil1 plants.
H3K4Me2 was accumulated in axe1-5 and sil1 plants. Therefore, the hyper-acetylation of the TEs in hda6 mutants is correlated with an increase in H3K4 methylation. Taken together, our results suggested that HDA6 silences the TEs by modulating the histone H3 and H4 acetylation as well as H3K4 methylation levels.
To investigate whether HDA6 directly regulates TEs in vivo, Arabidopsis plants over-expressing HDA6-MYC (35S:HDA6-TAP) were used to perform a ChIP assay with an anti-MYC antibody. Real-time PCR was used to analyze the relative abundance of HDA6, surrounding the transcription starting sites and coding regions of the representative TEs (AT2G04770 and AT2G09540). As seen in Fig. 3, HDA6 was recruited to the transcription starting sites and coding regions of AT2G04770 and AT2G09540, suggesting that these TEs are direct targets of HDA6.

HDA6 functions in maintaining CG, CHG and CHH methylation of TEs
To assess whether the reactivation of the TEs in hda6 mutant is associated with TEs are highly methylated. In contrast, a strong DNA band was detected in digested sample of AT5G59620, suggesting that this TE is relatively hypo-methylated in the wild-type. In the axe1-5 mutant, strong bands were scored in McrBC-digest samples of all the TEs, indicating loss of cytosine methylation of the TEs. Taken together, these data demonstrate that HDA6 affects the DNA methylation status of the TEs.
The DNA methylation status of three representative TEs, AT5G59620, AT2G04460 and AT2G04470, were further analyzed by bisulfite sequencing. As shown in Fig. 4B and Supplemental Fig. S1, the CG sites at these loci were heavily methylated in the wild-type. In comparison, methylation levels at CHG and CHH sites in these loci are much lower than those at CG sites in the wild-type ( Fig. 4B). An obvious decrease of CG methylation in AT2G04460 was detected in the axe1-5 mutant. Furthermore, substantial decreases of CHG and CHH were detected in AT2G04460 and AT5G59620, and the CHG methylation of AT2G04770 was also lost in the axe1-5 mutant. These data demonstrate that loss of function of HDA6 results in loss of cytosine methylation in CG, CHG and CHH sites in TEs, further suggesting that HDA6 plays an important role in maintaining CG, CHG and CHH methylation.
To explore the effect of HDA6 on genomic cytosine methylation, we employed methylation sensitive amplification polymorphism (MSAP) to analyze the landscapes of this DNA modification in axe1-5 mutants. MSAP assay is a modified version of AFLP analysis incorporating methylation-sensitive restriction enzymes in an efficient procedure to reveal genome-wide DNA methylation alterations in a locus-specific manner (Zhang et al., 2009). Hpa II and Msp I recognize the same restriction site (5'-CCGG) but have different sensitivities to methylation of the cytosines. Hpa II does not cut if either of the cytosine is fully (double strand) methylated, whereas Msp I does not cut if the external cytosine is fully or hemi (single) methylated. Thus, full methylation of the internal cytosine, or hemi-methylation of the external cytosine at the assayed CCGG sites can be unequivocally identified by MSAP. For clarity, we hereby refer to these two types of patterns as CG and CHG methylation, respectively (Dong et al., 2006 S2). 37.6% (38) of the bands were detected to be CG methylated in the wild-type.
In contrast, only 22.5% (18) CG methylation sites were found in the axe1-5 mutant, suggesting that the hda6 mutation induces a decrease in genomic cytosine methylation. This data further enforces the role of HDA6 in maintaining cytosine methylation.

Interactions of HDA6 with MET1 both in vitro and in vivo
In Arabidopsis, the DNMT1 homolog MET1 plays vital roles in maintaining cytosine methylation (Finnegan et al., 1996). The finding that HDA6 is required for CG, CHG and CHH methylation of TEs prompted us to investigate the interaction between HDA6 and MET1. Using the yeast two hybrid assay, we found that HDA6 can interact with MET1 in yeast cells ( were additively accumulated in axe1-5met1-3. These data indicate that HDA6 acts additively with MET1 to regulate the H4ac of TEs. Furthermore, H3K4Me3 and/or H3K4Me2 of these four TEs were also enriched in met1-3 mutant and axe1-5 met1-3 plants ( Fig. 7B and 7C). H3K4Me3 of AT2G20460 and AT2G26630 in axe1-5met1-3 were higher than that of axe1-5 and met1-3 single mutants. Moreover, H3K36Me2 of AT2G04470, AT5G19015 and AT2G20460 were elevated in axe1-5 and met1-3 plants and additively increased in axe1-5met1-3 plants. These data suggest that HDA6 and MET1 co-regulate the histone methylation profiles of TEs. Taken together, our results suggest that HDA6 and MET1 act together to regulate DNA methylation, histone acetylation and histone methylation of the TEs in Arabidopsis.

HDA6 silences TEs by modulating histone acetylation, histone methylation and DNA methylation
Plant genomes contain many transposable elements, most of which are inactivated or "silenced" by chromatin-remodeling and DNA methylation factors (Okamoto and Hirochika, 2001). A subset of TEs is transcriptional reactivated in hda6 mutants, suggesting that HDA6 is required for silencing of TEs. Activation marks, such as histone H3 and H4 acetylation, H3K4 tri-or di-methylation and H3K36 di-methylation are increased, whereas cytosine methylation is decreased at the TEs in hda6 plants, suggesting that HDA6 silences the TEs by regulating these modifications. HDA6 may therefore play important roles in the interplay among histone deacetylation, histone demethylation and DNA methylation in transcriptional regulation.
HDA6 is a Trichostain A sensitive HDAC capable of removing acetyl groups from multiple lysines of histone (Earley et al., 2006;Earley et al., 2010). HDA6 may therefore repress the TEs through direct deacetylating of their chromatin.
Accumulation of H3K4Me3, H3K4Me2 and H3K36Me2 marks at the TEs in hda6 mutants also suggests a role for HDA6 in repression of active histone methylation.
The crosstalk between histone deacetylation and demethylation has previously been implicated to modulate gene expression in mammalian cells (Shi et al., 2005;Lee et al., 2006;Shiekhattar et al., 2006 Cytosine methylation levels in CHG and CHH sites were reduced in axe1-5 plants, which is associated with the loss of the H3K9Me2 mark (Supplemental Fig.   S4). Previous studies revealed that methylation maintained by CMT3 is dependent on H3K9Me2 (Johnson et al., 2002;Lindroth et al., 2004), indicating that HDA6 may act indirectly to maintain CHG and CHH methylation through regulation of H3K9Me2 of TEs.

HDA6 directly interacts with MET1
The interplay between histone deacetylation and DNA methylation in gene silencing has been documented (Nan et al., 1998;Fuks et al., 2000;Lippman et al., 2003;To et al., 2011). In mammal cells, DNA methyltransferase DNMT1 represses gene expression via association with the methyl-CG-biding domain protein (MBD), which in turn recruits histone deacetylase activities (Nan et al., 1998). Based on this model, MBD proteins act as a bridge on the DNA methylation and histone deacetylation in gene silencing. A more direct connection between the DNA methylation and deacetylation has also been demonstrated by the finding that the mammalian DNMT1 binds to HDAC1 using the N-terminal non-catalytic domain to repress gene transcription through histone deacetylase activity (Fuks et al., 2000;Rountree et al., 2000). Furthermore, DNMT1 also interacts directly with HDAC2 to form a repression complex at replication foci during late S-phase (Rountree et al., 2000). Two de novo methyltransferases,  compared with wild-type, supporting a scenario that HDA6 and MET1 act cooperatively to silence TEs. The additive increase of these histone modifications of some loci in axe1-5met1-3 double mutant suggests that HDA6 and MET1 may collaborate with other chromatin remodeling factors to modulate the chromatin landscapes. In the present study, we demonstrated that HDA6 directly associates with MET1 to silence TEs by modulating DNA methylation, histone acetylation and histone methylation status. Our study provides a new prospective to understand the interplay between histone deacetylation and DNA methylation in gene silencing.

Chromatin immunoprecipitation assay
Chromatin immunoprecipitation assay was carried out as described (Gendrel et al., 2005). Chromatin was extracted from 18-day old plants growing in a long-day condition. After fixed with formaldehyde, the chromatin was sheared to an average length of 500 bp by sonication, and then immunoprecipitated with ACTIN2. Each of the immunoprecipitations was replicated three times with different sets of plants, and each sample was quantified at least in triplicates during real-time PCR analysis. The primers used for real-time PCR analysis in ChIP assays were listed in Supplemental Table S2.

Methylation sensitive amplification polymorphism (MSAP)
MSAP assay was carried out as described in (Zhang et al., 2009

Yeast two-hybrid assay
Yeast two-hybrid assays were performed according to the instructions for the Matchmaker GAL4-based two hybrid system 3 (Clontech). Constructs were generated by cloning full length or different regions of MET1 and HDA6 cDNA fragments into pGADT7 and pGBKT7 vectors. All constructs were transformed into yeast strain AH109 by the lithium acetate method and yeast cells were grown on minimal medium/-Leu-Trp according to the manufacture's instructions (Clonthech).

BiFC assay
Full length cDNA fragments of MET1 and HDA6 were subcloned into the