Publisher Correction: Exogenously overexpressed intronic long noncoding RNAs activate host gene expression by affecting histone modification in Arabidopsis

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Epigenetic modification of the FLC chromatin involves cis-acting lncRNAs, including COOLAIR, COLDAIR, and COLDWRAP 11,12,28 . COOLAIR is a set of alternative spliced and polyadenylated antisense lncRNAs transcribed from the 3′ UTR of FLC 11 . COOLAIR mediates the replacement of H3K36 methylation with H3K27me3 but work independently of the PcG complex during the early stage of vernalization 29 . COLDAIR, an intronic lncRNA from the first intron of FLC, cooperates with the FLC promoter-derived lncRNA COLDWRAP to facilitate the establishment of H3K27me3 and to thereby repress FLC expression during the late stage of vernalization 12,28 . Considering that COLDAIR is a functional lncRNA generated from the intron of FLC, we wonder how COLDAIR regulates FLC and whether there other intronic RNAs involved in regulating the expression of their corresponding host genes.
Intronic lncRNAs have been extensively identified and demonstrated to be functional in regulating the expression of their corresponding host genes in eukaryotes 30 . Generated from spliced introns, circular intronic RNAs (ciRNAs) are abundant in the nucleus and promote the transcription of their host genes by associating with RNA polymerase II in human cells 31 . Stable intronic RNAs were also found to play an important role in enhancing the expression of their host genes in Xenopus and Drosophila 32,33 . Although the function of intronic RNAs in the regulation of transcription has been extensively studied, it remains largely unknown how these intronic RNAs regulate transcription through affecting chromatin modification.
In this study, we demonstrate that the exogenously overexpressed intronic noncoding RNA COLDAIR is sufficient to enhance FLC expression in multiple independent COLDAIR transgenic lines. The enhancement depends on the recruitment of the H3K4me3 methyltransferase ATX1 and the removal of the H3K27 trimethyltransferase CLF at the FLC chromatin. Furthermore, we demonstrate that overexpression of intronic lncRNAs derived from several other H3K27me3-enriched MAD-box genes is also sufficient to enhance the expression of their corresponding host genes. These results show that intronic lncRNAs derived from H3K27me3-enriched MADS-box genes can enhance the expression of their corresponding host genes by suppressing H3K27me3 and promoting H3K4me3. The study suggests that ectopically overexpressed intronic RNAs may regulate the expression of their host genes by affecting the occupancy of histone methyltransferases.

Results ectopically overexpressed COLDAIR enhances FLC expression in vivo. FLC is a MADS-box-
containing transcriptional factor that functions as a critical flowering repressor in Arabidopsis 20 . The first intron of FLC is required for transcriptional regulation 22,34,35 . COLDAIR, an intronic lncRNA from the first intron of FLC, was known to promote H3K27me3 and thereby repress the transcription of FLC 12 , but how COLDAIR regulates H3K27me3 is largely unknown 36 . We developed a transgene system, in which the full-length or truncated sequences of the first FLC intron were overexpressed under the control of the Cauliflower Mosaic Virus (CaMV) 35S promoter (Fig. 1A). FLC functions in a dosage-dependent manner to control flowering time in Arabidopsis 20,37 . In this transgene system, we found that the 35S-COLDAIR T1 transgenic plants expressing the full-length COLDAIR showed significantly late flowering, whereas the transgenic plants harboring the COLDAIR sequence without the 35S promoter did not exhibit significant changes in flowering time (Fig. 1A,B). To confirm the effect of the COLDAIR overexpression, we randomly selected 20 individual 35S-COLDAIR T2 transgenic lines for determining FLC expression and flowering time. The results showed that 25% (5/20) of randomly selected 35S-COLDAIR lines displayed late flowering, which is accompanied by the increased expression of FLC (Fig. 1C,D). The results suggest that the ectopic overexpression of COLDAIR can enhance the expression of FLC in vivo. We assessed the transcript levels of FLC and COLDAIR in different ecotypes of Arabidopsis by RT-qPCR. The result indicated that the transcript levels of FLC were different in the indicated ecotypes (Fig. S1). The different expression levels of FLC in these ecotypes are primarily due to nature variants of FRIGIDA, a transcriptional activator of FLC 38 . Moreover, the levels of COLDAIR are positively correlated with the FLC transcript levels in the indicated ecotypes (Fig. S1), suggesting that COLDAIR is unlikely to play a critical role in the repression of FLC expression. the ectopically overexpressed COLDAIR specifically targets FLC. A high level of FLC expression is characteristic of the vernalization-responsive winter annual ecotypes of Arabidopsis, which are different from the rapid-cycling early-flowering ecotypes 20,38 . Vernalization reduces FLC expression and eliminates the late-flowering phenotype in the winter annual ecotypes. To elucidate the mechanism underlying the late-flowering phenotype of the 35S-COLDAIR lines, we determined whether the effect of the COLDAIR overexpression on flowering time is related to vernalization. Two representative late-flowering 35S-COLDAIR lines (35S-COLDAIR #3 and #17) with or without vernalization treatment were chosen for the analysis of flowering time. The results showed that the late-flowering phenotype of both transgenic lines was corrected by the vernalization treatment (Fig. S2), which supports the notion that the late-flowering phenotype of the COLDAIR transgenic lines is caused by increased expression of FLC.
To investigate the effect of the COLDAIR overexpression at the whole-genome level, we performed RNA deep sequencing (RNA-seq) in order to compare the expression of all the flowering time related genes between the representative 35S-COLDAIR line (35S-COLDAIR #3) and the wild type (Table S1). Our RNA-seq data showed that the expression of FLC was markedly increased in 35S-COLDAIR #3 ( Fig. 2A; Data set 1). Concomitantly, the expression of the key flowering promoter gene FLOWERING LOCUS T (FT) and of the critical floral meristem identity gene APETALA1 (AP1) was significantly reduced ( Fig. 2A; Data set 1). Because FLC reduces the expression of FT and AP1 39,40 , reduced expression of FT and AP1 is likely to be caused by the increased expression of FLC. To further verify whether a high level of FLC expression is required for the late-flowering phenotype, we crossed 35S-COLDAIR #3 with a loss-of function flc mutant (flc-8) to introduce the flc mutation into the 35S-COLDAIR plants. The results showed that the late-flowering phenotype of 35S-COLDAIR #3 was completely suppressed by the flc mutation (Fig. 2B,C). These data demonstrate that the late-flowering phenotype of the the exogenously expressed COLDAIR RNA may directly promote FLC expression without producing small RNAs or small peptides. Small RNAs derived from dsRNAs can not only suppress but also activate transcription 42 . To investigate whether small RNAs are generated from the COLDAIR dsRNA, we carried out small RNA deep sequencing in wild-type and 35S-COLDAIR plants. The data showed that small RNAs from the COLDAIR sequence were enriched in the 35S-COLDAIR plants but not in the wild-type plants (Fig. 4A). The sizes of these small RNAs were predominantly 21 nt, 22 nt, and to a lesser extent 24 nt (Fig. 4B). In Arabidopsis, there are four DCLs (Dicer-like proteins) responsible for generating distinct small RNAs. Whereas DCL1 is responsible for the cleavage of ssRNAs into 21-nt miRNAs, the other three DCLs are responsible for generating small RNAs from dsRNA precursors 43,44 . DCL2 plays a role in the formation of 22-nt natural antisense or viral siRNAs; DCL3 is involved in the biogenesis of 24-nt heterochromatic siRNAs; and DCL4 produces 21-nt siRNAs from inverted repeated (IR) genes and ta-siRNAs 45 .
Given that the exogenous COLDAIR produces small RNAs, we determined whether DCLs contribute to the regulation of FLC expression via generating the small RNAs. We introduced the 35S-COLDAIR transgene into the dcl2/3/4 triple mutant by genetic crossing, and then compared the effects of the exogenous COLDAIR on FLC expression between the wild-type and dcl2/3/4 mutant backgrounds. Our RT-qPCR results showed that, while   expression was enhanced by both the overexpressed COLDAIR and the dcl2/3/4 mutation, FLC expression was significantly enhanced by the dcl2/3/4 mutation in the 35S-COLDAIR line (Fig. 4C). The late-flowering phenotype of the 35S-COLDAIR line was slightly enhanced by the dcl2/3/4 mutation but the enhancement was not significant as determined by a statistical analysis (Fig. 4D). It is possible that the increase in FLC expression caused by the dcl2/3/4 mutation is not sufficient to further delay flowering in the 35S-COLDAIR background. These results indicated that, whether or not COLDAIR is overexpressed, DCL2/3/4 suppress rather than promote FLC expression, suggesting that the full-length exogenous COLDAIR RNAs, but not the small RNA byproducts, are responsible for the promotion of FLC expression.
Small peptides from lncRNAs were previously found to be involved in the regulation of gene expression and development 46,47 . Thus, the exogenous COLDAIR may enhance FLC expression by encoding small peptides. According to the putative open reading frames (ORFs) in COLDAIR, we truncated the COLDAIR sequences into three parts, which may translate different versions of peptides (Fig. S3A). Although the three truncated COLDAIR sequences were overexpressed under the control of the 35 S promoter in their corresponding transgenic plants, they failed to significantly affect FLC expression or flowering time (Fig. S3B,C). The results suggest that the exogenous COLDAIR transcripts do not promote FLC expression by expressing small peptides. Thus, the full-length exogenous COLDAIR transcripts may directly enhance FLC expression in vivo.

the exogenously overexpressed COLDAIR reduces H3K27me3 and increases H3K4me3. In
Arabidopsis, FLC chromatin is marked by both H3K27me3, H3K4me3, and H3K36me3 21,48,49 . We carried out chromatin immunoprecipitation followed by deep sequencing (ChIP-seq) to determine whether the exogenously overexpressed COLDAIR affects FLC expression by regulating H3K27me3 and H3K4me3. Consistent with previous studies 48,49 , our ChIP-seq results indicated that, in the wild type, H3K27me3 was enriched throughout the full length of the FLC genomic region and H3K4me3 was specifically enriched at the region shortly after the transcription start site (Fig. 5A). In the 35S-COLDAIR transgenic line, the H3K27me3 level was significantly lower than in the wild type, whereas the H3K4me3 level was significantly higher than in the wild type (Fig. 5A). The aberrant H3K27me3 and H3K4me3 signals of the COLDAIR region in the 35S-COLDAIR transgenic line were most likely due to the presence of the exogenous COLDAIR in the transgenic line, which could not be distinguished from the endogenous COLDAIR region by ChIP-seq.
To further validate the effect of the exogenous COLDAIR on H3K27me3 and H3K4me3, we performed ChIP followed by PCR (ChIP-PCR) in the wild-type Col-0 control and the 35S-COLDAIR transgenic line. In particular, we assessed the H3K27me3 and H3K4me3 levels of the endogenous COLDAIR region by using a pair of primers that can specifically amplify the fragment (the 5 th fragment) covering the edge of the endogenous COLDAIR region. The results indicated that the H3K27me3 level of the full-length FLC was lower in the www.nature.com/scientificreports www.nature.com/scientificreports/ 35S-COLDAIR transgenic line than in Col-0 even in the endogenous COLDAIR region (Fig. 5B). The H3K4me3 ChIP-PCR experiment also confirmed the increase of H3K4me3 at the region shortly after the transcription start site of FLC in the 35S-COLDAIR transgenic line (Fig. 5C). Moreover, the H3K4me3 enrichment in the endogenous COLDAIR region (the 5 th fragment) of the 35S-COLDAIR transgenic line was not detected by ChIP-PCR (Fig. 5C), confirming that the aberrant H3K27me3 and H3K4me3 signals identified by ChIP-seq in the COLDAIR region of the 35S-COLDAIR transgenic line was from the exogenous COLDAIR rather than from the endogenous COLDAIR. These results therefore demonstrate that the exogenous COLDAIR promotes FLC expression by reducing H3K27me3 levels and increasing H3K4me3 levels.
CLF and ATX1 are involved in the regulation of FLC expression by exogenous COLDAIR. CLF is a major histone H3K27 trimethyltransferase in the Arabidopsis PcG complex [24][25][26] . In the clf mutant, the H3K27me3 level of FLC is reduced, which results in an increase of FLC expression 23 . To investigate whether the PcG complex is related to the increase of FLC expression, we crossed the 35S-COLDAIR line with the clf mutant to determine the effect of clf on FLC expression. Our RT-qPCR result showed that the transcript level of 35S-COLDAIR is not significantly affected in the clf mutant compared to the wild type (Fig. S4). In the wild-type background, the clf mutation significantly (Student's t test; p < 0.01) enhanced FLC expression (Fig. 6A). In the 35S-COLDAIR line, however, the FLC expression level was markedly higher than in the wild type, and the clf mutation only slightly enhanced FLC expression (Fig. 6A). Considering that CLF is a major histone H3K27 trimethyltransferase in the PcG complex, we predicted that the overexpressed COLDAIR may promote FLC expression at least partially through suppressing the function of the PcG complex in H3K27me3.
Trithorax Group complexes (TrxG) are responsible for establishment of H3K4me3 at the FLC chromatin and thereby facilitate FLC expression 27 . ATX1 is a major histone H3K4 methyltransferase subunit of the TrxG complex 50,51 . In the atx1 mutant, the FLC expression was reduced, which led to an early-flowering phenotype 27 . To investigate whether the TrxG complex is involved in the regulation of FLC expression by the exogenous COLDAIR, we crossed the 35S-COLDAIR line with the atx1 mutant and determined how the atx1 mutation affects FLC expression in the 35S-COLDAIR line. The transcript level of COLDAIR in the 35S-COLDAIR line is comparable between the wild-type and atx1 mutant backgrounds as determined by RT-qPCR (Fig. S4). Consistent with the previous study 27 , our RT-qPCR experiment showed that, the atx1 mutation significantly reduced FLC expression in the wild-type background (Fig. 6A). FLC expression was reduced by the atx1 mutation in the 35S-COLDAIR line as well as in the wild type even though the basic FLC expression level was significantly higher in the 35S-COLDAIR line than in the wild type (Fig. 6A), suggesting that the increase of FLC expression www.nature.com/scientificreports www.nature.com/scientificreports/ caused by the COLDAIR overexpression is partially dependent on ATX1. Given that ATX1 is a major histone H3K4 methyltransferase in the TrxG complex, the results indicate that FLC expression depends, at least partially, on the TrxG complex.
To further investigate how the histone H3K27 methyltransferase CLF and the histone H3K4 methyltransferase ATX1 are coordinated to facilitate FLC expression in the 35S-COLDAIR lines, we generated a clf/atx1 double mutant in the 35S-COLDAIR and wild-type backgrounds by genetic crossing. RT-qPCR indicated that the transcript level of COLDAIR in the 35S-COLDAIR line is not significantly affected in the clf/atx1 mutant relative to the wild type (Fig. S4).
In the wild type background, the FLC expression level was reduced in the atx1 mutant, but the reduction of FLC expression was partially restored by the clf mutation in the clf/atx1 double mutant (Fig. 6A), suggesting that CLF and ATX1 cooperate to regulate FLC expression. In the 35S-COLDAIR line, however, the reduction of FLC expression caused by the atx1 mutation failed to be restored by the clf mutation (Fig. 6A). As indicated above (Fig. 6A), the clf mutation also had a very weak effect on FLC expression in the 35S-COLDAIR line even when ATX1 was not mutated. These results suggest that the promotion of FLC expression by the COLDAIR overexpression is related to CLF and ATX1.

COLDAIR suppresses CLF occupancy and increases ATX1 occupancy on FLC. Both CLF and ATX1
were previously shown to directly interact with the FLC chromatin in order to regulate transcription 23,27 . We performed ChIP-qPCR using Myc-tagged CLF and Flag-tagged ATX1 transgenic plants to determine whether the overexpressed COLDAIR affects the occupancy of CLF and ATX1 at the FLC chromatin loci. We found that, in the 35S-COLDAIR plants, the enrichment of CLF was significantly lower than in the wild type (Fig. 6B), whereas the enrichment of ATX1 was significantly higher than in the wild type (Fig. 6C). These data suggest that the overexpressed COLDAIR somehow modulates the occupancy of CLF and ATX1 on the FLC chromatin and thereby regulates H3K27me3 and H3K4me3, respectively.  www.nature.com/scientificreports www.nature.com/scientificreports/ expression through repression of H3K27me3, we asked whether the function of intronic lncRNAs is conserved in the other MIKC MADS-box genes.

Activation of host genes by intronic RNAs is conserved in MIKC MADS-box genes. In
We assessed the intron length of MIKC MADS-box genes and found that the MIKC MADS-box genes have larger intron sizes than the other protein-coding genes in Arabidopsis (Fig. 7A). Therefore, we predicted that overexpression of intronic lncRNAs from the other MIKC MADS-box genes may also enhance the expression of their corresponding host genes. To validate the function of intronic lncRNAs in the other MIKC MADS-box genes, we selected seven additional MIKC MADS-box genes including AGAMOUS (AG), AP1, AGAMOUS-LIKE 11 (AGL11), AGAMOUS-LIKE 24 (AGL24), SEPALLATA 1 (SEP1), FRUITFULL (FUL), and SEPALLATA 2 (SEP2) for analysis (Fig. 7B). By overexpressing the longest intronic RNAs of the seven MIKC MADS-box genes (Fig. 7B), we found that the overexpressed intronic RNAs were capable of activating the expression of AG, AP1, AGL11, AGL24, and SEP1 but not the expression of FUL and SEP2 in their corresponding T1 transgenic plants (Fig. 7C). The effect of these intronic RNAs on the expression of AG, AP1, AGL11, AGL24, and SEP1 can be inherited in the transgenic T2 plants (Fig. 7C). These results demonstrate that enhancement of the expression of host genes by overexpressing their corresponding intronic RNAs is conserved for MIKC MADS-box genes.

Discussion
In Arabidopsis, the endogenous COLDAIR was previously shown to associate with the major H3K27 methyltransferase CLF and thereby enhance H3K27me3 at the late stage of vernalization 12 . Unlike the previous study, the results in the current study strongly suggests that the exogenously overexpressed COLDAIR suppresses H3K27me3 at the FLC chromatin, indicating that the exogenous COLDAIR shows an antagonistic effect on H3K27me3 compared to the endogenous COLDAIR. We predict that the exogenously overexpressed COLDAIR may act as an inhibitor or a decoy for CLF and thereby make the cis-acting endogenous COLDAIR inaccessible to CLF for H3K27me3 establishment. Alternatively, it may also be possible that antagonistic effects of COLDAIR on H3K27me3 occur during different biological processes. The promotion of H3K27me3 by COLDAIR occurs during vernalization, whereas the suppression of H3K27me3 by COLDAIR may occur in the other biological processes such as early embryo development when FLC expression is reactivated.
In the exogenously overexpressed COLDAIR line, H3K27me3 is reduced, and H3K4me3 is increased (Fig. 5A,B). The dual effect on H3K27me3 and H3K4me3 is also observed in the mutant defective in the histone H3K27 tri-methyltransferase CLF 23 . Therefore, it is possible that H3K27me3 may be primarily reduced by the exogenous COLDAIR and the reduction of H3K27me3 may subsequently promote the establishment of H3K4me3. In humans, the lncRNA HOTTIP transcribed from the 5′ end of the HOXA locus directly associates with WDR5, a subunit of the H3K4 trimethyltransferase complex COMPASS and thereby promotes HOXA expression by increasing H3K4me3 8 . Components of the COMPASS complex in Arabidopsis, which are conserved with those in mammals, have been demonstrated to bind to FLC and its homologs and are required for promoting H3K4me3 54 . In Arabidopsis, an antisense lncRNA was shown to interact with WDR5a to recruit the COMPASS complex to MAF4, a flowering repressor gene closely related to FLC, leading to an increase in H3K4me3 55 . Our results indicate that the overexpressed intronic RNA derived from the intron 1 of FLC increases www.nature.com/scientificreports www.nature.com/scientificreports/ FLC expression and H3K4me3 by reinforcing the occupancy of ATX1 on the FLC chromatin. Because ATX1 is a major catalytic subunit of the Arabidopsis COMPASS complex, we propose that the intronic RNA may also directly associate with the COMPASS subunit WDR5a and thereby recruit the COMPASS complex to the FLC chromatin for H3K4me3. However, we cannot exclude the possibility that the mechanisms other than histone H3K4 and H3K27 methylation may also be involved in the promotion of FLC expression by the overexpressed COLDAIR. Considering that the coupled changes in H3K27me3 and H3ac were recently observed 56 , we predict that the histone acetylation may also be necessary for the promotion of FLC expression by the overexpressed COLDAIR.
By overexpressing intronic RNAs derived from randomly selected MIKC MADS-box genes, we demonstrated that the activation of gene expression by overexpressed intronic RNAs is conserved in the MIKC MADS-box genes AG, AP1, AGL11, AGL24, and SEP1, but not in FUL and SEP2 (Fig. 7B,C). Of note, FLC expression was enhanced in only a subset of the 35S-COLDAIR lines (Fig. 1A-D), suggesting that some uncharacterized mechanisms may be responsible for counteracting the activation of the host genes by overexpressed intronic RNAs. Our results showed that FLC expression is enhanced in the dcl2/3/4 mutant (Fig. 6A), which is consistent with the previous finding that DCL3 promotes flowering by reducing FLC expression 57,58 . Given that a large number of small RNAs are generated from the COLDAIR locus in the 35S-COLDAIR line, we predict that these small RNAs may be responsible for the reduction of FLC expression at either transcriptional or posttranscriptional levels, thereby counteracting the promotion of FLC expression by the overexpressed COLDAIR. We suspect that the function of overexpressed intronic RNAs in activating gene expression by reducing H3K27me3 may be conserved in H3K27me3-enriched genes in Arabidopsis. Because H3K27me3 occurs not only in MADS-box genes but also in thousands of other genes 48,53 , the function of overexpressed intronic RNAs may be much more common than we expected in H3K27me3-enriched genes. Considering that stable intronic RNAs have been extensively identified in eukaryotes 30,32,33 , we predict that these endogenous intronic RNAs may also play important roles in the regulation of gene expression through the same mechanism.
Repressed FLC expression needs to be reactivated during the early embryo development and therefore ensures the requirement for vernalization-induced H3K27me3 in every generation. The transcriptional activators LEAFY COTYLEDON 1 (LEC1), LEAFY COTYLEDON 2 (LEC2) and FUSCA 3 (FUS3) and the histone H3K27 demethylase EARLY FLOWERING 6 (ELF6) are required for the reactivation of FLC by reducing H3K27me3 levels [59][60][61] . Because the transcriptional activators required for FLC expression are exclusively expressed in the embryo 59,60 , it is unknown how FLC expression is maintained after the embryo-to-seedling transition. Moreover, loss of REF6 only partially affects H3K27 demethylation at FLC chromatin during reproductive development 61 , suggesting that an REF6-independent mechanism must be required for reducing H3K27me3 levels. The results from the current study indicate that the overexpressed intronic RNA from the intron 1 of FLC is sufficient for reducing H3K27me3 levels. We suspect that the embryo-specific transcriptional activators may be responsible for initiating the reactivation of FLC expression by reducing H3K27me3 during the early embryo development. With the reactivation of FLC expression, the intronic RNA excised from the intron 1 of FLC pre-mRNA may gradually accumulate, and the accumulated intronic RNA would further suppress H3K27me3 and increase H3K4me3, resulting in a full activation of FLC expression after the embryo-to-seedling transition. Based on all of these results, we suspect that overexpressed intronic RNAs may activate gene expression by counteracting H3K27me3 and/or by promoting H3K4me3 during important biological processes. RnA extraction and RnA level analysis. RNA was extracted from 14-day-old plants as described previously 63 . The first strand of cDNA was prepared with the PrimeScript RT Reagent Kit (TAKARA, RR037A). Quantitative PCR was performed on an ABI 7500 fast Real time PCR instrument with KAPA SYBR FAST Universal reaction regent, and the results were quantified by reference to a standard curve for each primer pair with at least two repeats. For assessment of noncoding RNA expression, sequence-specific primers were used for reverse transcription. For RNase-treated RT-qPCR, 10 µg of total RNA was digested with RNase III (NEB, M0245S) or RNase one (Promega, M4261) according to the instruction manual. After chloroform purification, 1 µg of RNA was treated with DNase I to remove genomic DNA, and the RNA was then subjected to RT-PCR. Oligonucleotides used in this study are indicated in Data set 2.

Materials and Methods
Transcriptome sequencing and data analysis. Total RNA was extracted from 12-day-old seedlings, and mRNA was isolated from the total RNA using poly-T oligo-attached magnetic beads. The Illumina ScriptSeq www.nature.com/scientificreports www.nature.com/scientificreports/ Complete Kit (Plant) was used for library construction. Small RNAs of 18-60 nt were gel-purified and subjected to library construction. Library construction and deep sequencing were performed by Vazyme Biotech (Nanjing, China). For analysis of transcriptome sequencing data, clean reads were mapped to the TAIR10 Arabidopsis genome by TopHat v2.1.0 64 , allowing up to two mismatches. The differentially expressed genes were calculated by cuffdiff v2.0.1 65 . For analysis of small RNA deep sequencing data, 18 to 30-nt clean reads were mapped to the TAIR10 Arabidopsis genome using Bowtie 66 , and only perfectly matched reads were retained for further analysis. chip assay. The ChIP experiment was performed according to a previous report with minor modification 67 . A 2-g quantity of 2-week-old seedlings was cross-linked in 1% formaldehyde for 20 min after ground. The cross-linking reaction was stopped by adding glycine to 125 mM and was incubated for 5 min. The chromatin was isolated with Nuclear Extraction Buffer [20 mM Tris-HCl (pH 7.5), 20 mM KCl, 2 mM EDTA (pH 8.0), 2.5 mM MgCl 2 , 25% glycerol, 250 mM sucrose, 5 mM DTT, 1 mM PMSF, and 1 × protease inhibitor mixture (Roche)]. After they were washed twice on two layers of Miracloth with Nuclear Extraction Buffer, and 3-5 times with Nuclear Resuspension Buffer [20 mM Tris-HCl (pH 7.5), 2.5 mM MgCl 2 , 25% glycerol, 0.2% Triton X-100], the nuclear pellets were resuspended in 800 μL of Nuclei Lysis Buffer [20 mM Tris-HCl (pH 8.0), 2 mM EDTA, 0.2% NP-40, 1 mM PMSF, and 1 × protease inhibitor mixture). The chromatin of each sample was sonicated by a Bioruptor Plus device at a high power level for 28-33 cycles (30 sec ON and 30 sec OFF for each cycle). A 800-μL volume of sonicated chromatin diluted with ChIP dilution buffer (1 mM PSMF, 2 mM EDTA pH 8.0, 20 mM Tris-HCl pH 8.0, 200 mM NaCl, and 1 × protease inhibitor mixture) to 1.5 mL was used for immunoprecipitation. Dynabeads Proein A (Thermo, 10001D) were used for conjugating the anti-bodies. The anti-bodies used in this study were anti-H3 (Abcam, ab1791), anti-H3K4me3 (Millipore, 07-473), anti-H3K27me3 (Millipore, 07-449). The conjugated antibodies were independently mixed with the chromatin by rotation at 4 °C overnight. Beads were washed five times with Wash Buffer (150 mM NaCl, 20 mM Tris-HCl pH 8.0, 2 mM EDTA pH 8.0, 0.1% Triton X-100, 1 mM PMSF). Finally, beads were washed twice for 5 min each time with TE buffer (10 mM Tris, pH 7.5, 1 mM EDTA). After they were washed, the immuno-complexes was eluted from the beads and reverse cross-linked by NaCl and 20 ug Proteinase K (Sigma, P4850). Phenol/Chloroform/Isoamyl alcohol was used to extract the ChIP-DNA and followed with qPCR or sequencing. Deep sequencing was performed by Novogene (Beijing, China).
Stranded mRNA sequencing. Total RNA was extracted from 14-day-old seedlings are subjected to the Vazyme VAHTS Total RNA-seq Library Prep Kit for deep sequencing (Vazyme, NR603). RNA were digested with RNase H and DNase I and purified with VAHTS RNA Clean Beads (Vazyme, N412). Ribosome-depleted RNA was used as a template for synthesis of both strands of cDNA. Double stranded cDNA was purified with VAHTS DNA Clean Beads (Vazyme, N411). The dA-tailing and adapter ligation were performed using VAHTS RNA Adapters Set 1 -Set 2 (Vazyme, N803, N804). 1 × VAHTS DNA Clean Beads (Vazyme, N411) was used for purification and size selection of adapter-ligated DNA followed by amplification. Agilent DNA 1000 chip (Agilent, 5067-1504) was used for determination of library qualitydetermination.

Data availability
The raw data of RNA-seq, small RNA-seq, and ChIP-seq were deposited in the Gene Expression Omnibus (GEO) database (accession number: GSE140140).