DRD1, a SWI/SNF-like chromatin remodeling protein, regulates a heat-activated transposon in Arabidopsis thaliana

Transposable elements (TEs) of various classes, including both DNA transposons and retrotransposons, are abundant in plant genomes and impact genome evolution and gene expression (Wessler, 1996; Kumar and Bennetzen, 1999; Kazazian, 2004; Makarevitch et al., 2015). Despite their potential to either disadvantage or benefit the host genome, most TEs are silent and rarely transpose due to genetic aberrations such as point mutations, deletions or recombination that disrupt their activities. Even though full-length autonomous transposons are intact and have the ability to transpose, host plants DRD1, a SWI/SNF-like chromatin remodeling protein, regulates a heat-activated transposon in Arabidopsis thaliana


INTRODUCTION
Transposable elements (TEs) of various classes, including both DNA transposons and retrotransposons, are abundant in plant genomes and impact genome evolution and gene expression (Wessler, 1996;Kumar and Bennetzen, 1999;Kazazian, 2004;Makarevitch et al., 2015). Despite their potential to either disadvantage or benefit the host genome, most TEs are silent and rarely transpose due to genetic aberrations such as point mutations, deletions or recombination that disrupt their activities. Even though full-length autonomous transposons are intact and have the ability to transpose, host plants ONSEN is a heat-activated LTR retrotransposon in Arabidopsis thaliana. Screens to identify transcriptional regulatory factors of ONSEN revealed a SWI/ SNF-like chromatin remodeling protein, DRD1, which cooperates with plant-specific RNA polymerase and is involved in RNA-directed DNA methylation. ONSEN transcript level was increased in the drd1 mutant relative to wild-type under heat stress, indicating that DRD1 plays a significant role in the silencing of activated ONSEN under the stress condition. The transcript level of HsfA2, which is directly involved in transcriptional activation of ONSEN, was not higher in the drd1 mutant than in the wild-type. Interestingly, no transgenerational transposition of ONSEN was observed in the drd1 mutant, even though DNA methylation levels were significantly reduced and expression levels were increased compared to the wild-type. These results suggest that other factors are involved in the regulation of ONSEN transposition in addition to the transcript level of ONSEN.
Key words: heat stress, DRD1, transposon, ONSEN, RNA-directed DNA methylation have evolved various types of epigenetic regulation, such as DNA methylation or histone modification, to defend their genome against such transposition.
One of the well-studied mechanisms for TE regulation is RNA-directed DNA methylation (RdDM), in which small interfering RNAs (siRNAs) direct the cytosine methylation of DNA sequences that are complementary to the siRNAs (Wierzbicki et al., 2008;Gao et al., 2010). RdDM requires transcriptional machinery that involves two plant-specific RNA polymerases, RNA polymerase IV (Pol IV) and RNA polymerase V (Pol V) (Kanno et al., 2005b;Onodera et al., 2005;Pontier et al., 2005). Pol IV generates primary RNA transcripts and Pol V generates non-coding transcripts to introduce siRNA-mediated DNA methylation on the target site. In Arabidopsis thaliana, RNA-dependent RNA polymerase 2 (RDR2) converts a transcript produced by Pol IV to double-stranded RNA, and DICER-LIKE 3 (DCL3) subsequently processes these precursor RNAs into 24-nt siRNAs (Zhang et al., 2007;Mosher et al., 2008). The siRNAs bind to an RNAinduced silencing complex, RISC, that contains ARGO-NAUTE 4 (AGO4). AGO4 interacts with Pol V and recruits the DNA methyltransferase DOMAINS REAR-RANGED METHYLTRANSFERASE 2 (DRM2), which directs de novo DNA methylation of target TEs (Cao and Jacobsen, 2002;Matzke and Birchler, 2005).
Although most TEs have their own modes of regulation, some are activated under stress conditions (Chandler and Walbot, 1986;Bennetzen, 1987;Hirochika, 1993;Grandbastien et al., 1997Grandbastien et al., , 2004Scortecci et al., 1997;Steward et al., 2000;Hashida et al., 2003;Henderson and Jacobsen, 2007;Hirayama et al., 2009;Lisch, 2009;Zeller et al., 2009). The stress might induce a change in the epigenetic state of the TE, generating stress-responsive elements in host genes in the proximity of TEs. Epigenomic diversity may allow phenotypic plasticity and the ability to cope with environmental variation.
Previously, we found heat stress-induced activation of a Ty1/copia-like retrotransposon named ONSEN in A. thaliana (Ito et al., 2011). The activation of ONSEN requires a heat stress transcription factor, HsfA2, which distinctively binds to a cis-regulatory sequence (heat response element, HRE) in the promoter of the ONSEN LTR (Cavrak et al., 2014). Furthermore, the activated ONSEN is transposed in stressed plants that are defective in the RdDM pathway (Ito et al., 2011;Matsunaga et al., 2012).
In this study we performed a genetic screening to investigate the regulation mechanism of ONSEN and revealed that an epigenetic regulator, DRD1 (defective in RNA-directed DNA methylation), regulates ONSEN expression. DRD1 is a member of the plant-specific subfamily of SWI2/SNF2-like proteins (Kanno et al., 2004(Kanno et al., , 2005b. It associates with many subunits of the Pol V complex and is required for the accumulation of Pol V-dependent transcripts to facilitate RdDM and gene silencing of homologous DNA sequences (Kanno et al., 2005a;Law et al., 2010). Although the activation of ONSEN has been studied in some mutants (Ito et al., 2011;Matsunaga et al., 2012), the precise mechanism of transcriptional regulation remains unknown. Here, we provide new insights into chromatin remodeling proteinmediated regulation of ONSEN.

MATERIALS AND METHODS
Plant material and stress treatments The A. thaliana plants used in the experiments included wildtype Columbia-0 (Col-0) and Landsberg erecta (Ler), mutants nrpd1-3 (Herr et al., 2005) and drd1-6 (Kanno et al., 2004), and transgenic plants that possessed a full-length LTR (genome position: 4212570-4213146) of ONSEN (AT5G13205) fused with a GFP gene (Matsunaga et al., 2015). The plants were grown on Murashige and Skoog medium under continuous light at 21 °C. For heat stress treatment to analyze ONSEN expression and DNA methylation, seven-day-old seedlings were subjected to a temperature shift from 21 °C to 37 °C for 24 h. The transcript level of HsfA2 was analyzed on seven-day-old seedlings that were subjected to a temperature shift from 21 °C to 37 °C for 1 h.
Real-time PCR Total RNA was extracted from seedlings using TRI Reagent (Sigma-Aldrich), according to the supplier's recommendations. Five individual plants were pooled prior to RNA extraction. Around 3 to 5 µg of total RNA was treated with RQ1 RNase-free DNase (Promega) and reverse-transcribed using the ReverTraAce qPCR RT Kit (Toyobo) with random primers. Real-time PCR was performed using the Applied Biosystems 7300 Real-Time PCR System with the Power SYBR Green PCR Master Mix (Thermo Fisher Scientific). Three biological repetitions were performed and standard deviation was determined.
Southern blot analysis Plant genomic DNA was isolated using the Nucleon PhytoPure DNA Extraction Kit (GE Healthcare). Southern blots were performed as described previously (Miura et al., 2004). We detected hybridization signals in a highly concentrated sodium dodecyl sulfate hybridization buffer (Church and Gilbert, 1984) using a radiolabeled ONSEN-specific probe (Supplementary Table S1) that was generated with the Amersham Megaprime DNA Labeling System (GE Healthcare).
Mapping of mutation by whole-genome sequencing To map the gene responsible for the boil mutant, the boil mutant (Col-0) was first crossed with wild-type (Ler), and then self-pollinated to produce a segregating population of boil lines. Seedlings of this population were heat-stressed at 37 °C for 24 h, and then grown for about two weeks in pots, one group with high expression of GFP signal (High population) as in the boil mutant, and the other with weak expression of GFP fluorescence (Low population) as in the GFP:wild-type (Col-0). The true leaves were then sampled together in three pieces. The number of individuals used in the analysis was 22 for the High sample and 23 for the Low sample in boil1, and 20 for the High sample and 21 for the Low sample in boil5. Genomic DNA was extracted using the Nucleon PhytoPure DNA Extraction Kit (GE Healthcare). The extracted DNA was fragmented by sonication (S220, Covaris) and then used to construct a library using TruSeq DNA Library Prep Kits (Illumina) according to the manufacturer's protocol. The library was sequenced by NextSeq500 (Illumina). The output bcl files were converted to fastq files by bcl2fastq (Illumina). Based on the read sequences, a search for the gene responsible was performed in Mitsucal (Suzuki et al., 2018). For each mutant, the areas where Col-0-type bias was observed were selected. The search conditions were as follows: [number of substitutions: 1-200; substitution rate: 80-100; QV substitution rate: 80-100; number of control substitutions: blank; only mutation positions affecting the amino acid sequence: checked; type of base substitution: blank].
DNA methylation analysis For bisulfite sequencing analysis, genomic DNA was extracted from seedlings using the Nucleon PhytoPure DNA Extraction Kit. Bisulfite treatment was performed using the MethylCode Bisulfite Conversion Kit (Thermo Fisher Scientific). Bisulfitetreated DNA was amplified by PCR with the EpiTaq HS (Takara Bio). Primers for the analysis are listed in Supplementary Table S1. PCR products were cloned into the pANT vector using the TA-Enhancer Cloning Kit (Nippon Gene), and fifteen clones were sequenced for the 5′ LTR region in each line. Methylated sites were analyzed using CyMATE (http://www.cymate.org).

RESULTS
Upregulation of ONSEN in EMS mutants subjected to heat stress To understand the molecular mechanism of ONSEN regulation, we tried to find a new regulatory factor by mutant screening. We used a transgenic Arabidopsis that possessed an intact LTR of ONSEN fused with a gene for green fluorescent protein (GFP). The transgenic plants with a single-copy insert were mutagenized by ethyl methanesulfonate and the resulting M2 progeny were screened for mutations. The GFP signals of 5,000 heat-stressed mutants were compared with that of the parental line subjected to heat stress. Twentyfour individual seedlings with stronger GFP signals were self-pollinated and endogenous ONSEN expression in their progeny was analyzed by quantitative reverse transcription PCR. More than twice the expression level of ONSEN relative to wild-type was observed in 16 of the lines. We named the mutant lines boils (burst of ONSEN induction lines). We focused on one of the boils, boil5, which has strong GFP signals in the seedlings ( heat-stressed boil5 was six times higher than that in the parental line (Fig. 1B). Next, we analyzed the transgenerational transposition of ONSEN in boils subjected to heat stress. To detect new inserted copies of ONSEN in the progeny of heat-stressed boil5, Southern blot analysis was conducted on the next generation in boil5 plants subjected to heat stress. As a control, ONSEN transposition in boil1 progeny was also analyzed. boil1 has a mutation in the NRPD1 gene, which encodes the largest subunit of Pol IV. The mutation causes a non-synonymous substitution in the NRPD1 protein and strong GFP signals in the seedlings ( Fig. 2A and 2C). The expression level of ONSEN in heat-stressed boil1 plants was six times higher than that in the parental line (Fig. 2B). Southern blot analysis detected transgenerational transposition of ONSEN in the next generation in boil1 but not in boil5 subjected to heat stress (Fig. 2D). This result indicated that transcriptional activation was necessary but not sufficient for ONSEN transposition.
Mapping of BOIL5 To identify the gene responsible for ONSEN regulation in boil5, the mutant was outcrossed with Ler. Twenty of 134 F2 seedlings showed a strong GFP signal, indicating that the mutation of the gene responsible is recessive. Approximately 20 F2 progeny each that showed high or low expression of GFP were collected, and DNA was extracted from each group in bulk. A DNA library was constructed for each, and sequenced using a high-throughput sequencer. To identify the mutation responsible for the phenotypes of boil5, we used the Mitsucal software (Suzuki et al., 2018), which aligned all reads for boil5 with reference genes by Bowtie with a parameter permitting multiple alignments. Within the 5-10-Mbp region of chromosome 2, 24 mutations with a > 90% ratio of mismatch were detected. Fourteen muta- tions had no effect on protein structure (mutation in intron, or synonymous substitution). Seven mutations of the remaining ten were a substitution from G to A (or C to T), the most typical mutation caused by EMS. Four mutations were observed only in the F2 progeny that showed high expression of GFP. One of the four was in a gene encoding a WD-40 repeat family protein; two others were in genes encoding proteins of unknown function. The remaining one of the four mutations was found in the gene encoding DRD1, an SNF2 domain-containing protein. This mutation was located at the 5′ end of the fourth exon of the gene labeled AT2G16390 and causes a non-synonymous substitution (Fig. 1C). To assess whether this mutation corresponded to the phenotype of boil5, boil5 was crossed with a disruptant for AT2G16390 (designated drd1-6) obtained from the Arabidopsis Biological Resource Center (numbered CS69758, http://abrc. osu.edu/). The transcript level of ONSEN in F1 plants was significantly higher than that in the wild-type (Fig.  1D). These results strongly suggest that the causative factor of boil5 is a mutation in the DRD1 gene.
DRD1 regulates the expression of a heat-activated ONSEN The heat shock transcription factor HsfA2 is an important factor among a subset of stress response genes and is required for ONSEN activation in plants subjected to heat stress (Cavrak et al., 2014). To reveal whether the amount of HsfA2 affected the expression level of ONSEN in the drd1-6 mutant, we compared the transcript level of HsfA2 between wild-type and drd1-6 mutant. There was a significant increase of ONSEN transcript in drd1-6 under heat stress conditions, although the transcript level of HsfA2 was not significantly different ( Fig. 3A and  3B). This observation revealed that the increase in transcript level of ONSEN in drd1-6 was not correlated with the transcript level of HsfA2.
DRD1 regulates ONSEN expression by DNA methylation RdDM leads to methylation of cytosines in all sequence contexts: CG, CHG and CHH (where H corre-sponds to A, T or C). To understand the role of DRD1 in the epigenetic regulation of ONSEN, DNA methylation level was analyzed in the LTR promoter of the ONSEN copy (At1g11265) that shows the highest expression level upon heat stress (Cavrak et al., 2014). DNA methylation levels of this ONSEN LTR sequence showed a significant difference for non-CG methylation between the wild-type and the drd1-6 mutant: 76% of CHG was methylated in the wild-type, compared to 35% in the drd1-6 mutant. The methylation level of CHH was 7.1% in drd1-6 compared to 60% in the wild-type. The drd1-6 mutant showed the same hypomethylation state of non-CG as the nrpd1-3 mutant ( Fig. 4A and 4B).
Transgenerational transposition of ONSEN is not detected in heat-stressed drd1-6 To detect transgenerational transposition of ONSEN in drd1-6, Southern blot analysis was performed in the offspring of drd1-6 plants subjected to heat stress. No individuals having ONSEN transposition were found. In several individuals of that generation of nrpd1-3, which was used as a control, new ONSEN insertions were detected, suggesting that there are other important factors for ONSEN transposition besides ONSEN expression levels and DNA methylation levels on the ONSEN sequences (Fig. 5).

DISCUSSION
Here we report that a plant-specific SWI/SNF-like chromatin remodeling protein, DRD1, regulates the expression of a heat-activated retrotransposon. DRD1 was first identified in a screen for mutants impaired in RdDM (Kanno et al., 2004), and a subsequent study demonstrated that the endogenous targets of DRD1 were trans- posons or sequences that encode short RNAs (Huettel et al., 2006). Compared with the well-studied SNF2-like protein DDM1, which regulates global DNA methylation, DRD1 acts locally to regulate levels of non-CG methylation (Jeddeloh et al., 1999;Kanno et al., 2004). Several heterochromatic repeats lose CG methylation in ddm1 mutants but non-CG methylation in drd1 mutants, suggesting that DRD1 is important for non-CG methylation of target sequences (Huettel et al., 2006). The sequence of the ONSEN LTR contains 78 cytosines in non-CG contexts that could be targets for DRD1-mediated RdDM. As expected, the levels of CHG and CHH methylation on the ONSEN sequence were significantly reduced in the drd1 mutant compared to the wild-type. Most transposons are silenced under non-stress conditions and are not activated in the drd1 mutant. We found that the expression of heat-activated ONSEN was upregulated in the drd1 mutant compared with the wild-type. The transcript level of ONSEN may be affected by chromatin state, which can be changed by heat stress. Although SWI2/SNF2 chromatin remodeling proteins play roles in stress responses (Shaked et al., 2006;Han et al., 2012;Gentry and Hennig, 2014), further research will be needed to determine the chromatin modification of ONSEN in the heat-stressed drd1 mutant.
One possible mechanism to explain the upregulation of ONSEN in the drd1 mutant subjected to heat stress is that the increase of ONSEN transcript is affected by the transcript level of HsfA2 in the drd1 mutant. Another possibility is that the physical association of HsfA2 with its targets is enhanced in the drd1 mutant under heat stress. The latter mechanism is supported by the fact that the level of HsfA2 transcription was not significantly increased in the drd1 mutant compared with the wildtype under heat stress conditions. However, we cannot exclude the possibility that a factor(s) other than the redundant heat-related transcription factor family is required for ONSEN activation under heat stress. The exact mechanisms underlying heat shock factor-mediated upregulation remain to be further elucidated.
The endogenous targets of DRD1 silencing machinery are short-RNA-encoding elements that are located in the 5′ flanking region of the target sequences (Huettel et al., 2006). In general, the target sequences that were upregulated in the drd1 mutant had a euchromatic character and reside in gene-rich regions (Huettel et al., 2006). In the Col-0 accession, seven full-length ONSEN copies exist in euchromatic regions and one copy is located in centromeric heterochromatin. Further analysis may reveal whether the expression of each copy can be equally regulated by DRD1.
It is necessary to separate transcriptional activity from transposable activity when discussing the activation of transposons. An increase in the level of transcription is not necessarily associated with an increase in the frequency of transposition. As shown in our study, ONSEN is upregulated by heat stress in the drd1 mutant to the same level as in the nrpd1 mutant, but transgenerational transposition of ONSEN was not observed in the drd1 mutant, whereas it was observed in the nrpd1 mutant. To explain the difference, the properties of the ONSEN insertion sites need to be investigated, but at present, no common motifs have been found in the primary sequences of ONSEN insertion sites, and it is not clear what targets are used to determine the insertion sites. Previous studies, however, have shown that ONSEN tends to be inserted into euchromatin genes (Ito et al., 2016). Therefore, it is possible that DNA demethylation is induced in the drd1 mutant, but chromatin condensation is maintained. This distinction may account for the lack of transgenerational transposition of ONSEN even though the element has similar methylation levels and increased expression in both the drd1 and nrpd1 mutants.
We thank Dr. Tatsuo Kanno for valuable information about DRD1 and Ms. Tomomi Shinagawa for helping with highthroughput sequence analysis. This work was supported by a Cooperative Research Grant of the PTradD from the Gene Research Center, Tsukuba-Plant Innovation Research Center, the University of Tsukuba; NIG-JOINT (30A2018); JSPS KAKENHI (JP18K06050); and a Grant-in-Aid for scientific Research on Innovative Areas (JP15H05960).