The histone H3K27 demethylase REF6/JMJ12 promotes thermomorphogenesis in Arabidopsis

Abstract Dynamic trimethylation of histone H3 at Lys27 (H3K27me3) affects gene expression and controls plant development and environmental responses. In Arabidopsis thaliana, RELATIVE OF EARLY FLOWERING 6 (REF6)/JUMONJI DOMAIN-CONTAINING PROTEIN 12 demethylates H3K27me3 by recognizing a specific DNA motif. However, little is known about how REF6 activates target gene expression after recognition, especially in environmental responses. In response to warm ambient temperature, plants undergo thermomorphogenesis, which involves accelerated growth, early flowering and changes in morphology. Here we show that REF6 regulates thermomorphogenesis and cooperates with the transcription factor PHYTOCHROME INTERACTING FACTOR 4 to synergistically activate thermoresponsive genes under warm ambient temperature. The ref6 loss-of-function mutants exhibited attenuated hypocotyl elongation at warm temperature, partially due to downregulation of GIBBERELLIN 20-OXIDASE 2 and BASIC HELIX-LOOP-HELIX 87. REF6 enzymatic activity is necessary for warm ambient temperature responses. Together, our results provide direct evidence of an epigenetic modifier and a transcription factor working together to respond to the environment.

Plants have evolved diverse strategies to respond to natural environmental fluctuations [23,24]. Ambient temperature, as a major environmental signal, affects plant growth and development, geographical distribution and seasonal adaptation [25,26]. Warm ambient temperature induces dramatic morphological changes in plants, such as hypocotyl growth, petiole and root elongation, leaf hyponasty and early flowering; together, these changes are termed thermomorphogenesis [26]. Various signaling regulators and mechanisms that respond to warm temperatures have been identified [27]. For instance, external environmental cues (warm temperature, light, photoperiod, etc.) are integrated to influence hormone levels, localization and signaling for hormones such as auxin, BR and gibberellic acid (GA) [27].
The bHLH transcription factor PHY-TOCHROME INTERACTING FACTOR 4 (PIF4) acts as a central regulatory hub in thermomorphogenetic responses [28]. PIF4 mediates different aspects of thermomorphogenesis by binding different targets. For instance, PIF4 binds genes involved in auxin biosynthesis (such as YUCCA 8, TRYPTOPHAN AMINOTRANSFERASE OF ARABIDOPSIS 1 and CYTOCHROME P450 FAM-ILY 79B) [29,30] and BR biosynthesis (such as DWARF4 and BRASSINOSTEROID-6-OXIDASE 2) [31] to mediate vegetative shoot thermomorphogenesis. PIF4 also binds the FT promoter to mediate flowering responses [32], and binds SPEECHLESS to restrict stomatal development under warm conditions [33]. In addition, warm temperature affects the expression, stability and activity of PIF4, and PIF4 integrates hormone signaling (auxin, BR and GA) in thermoresponsive regulation of growth behaviors, such as stem growth and flowering time [25,27].
Multiple regulatory mechanisms are involved in the extensive network of plant thermoresponsive growth, including transcriptional regulation, RNA metabolism and protein stability [27]. Among these, transcriptional regulation mediated by chromatin dynamics is one of the primary mechanisms [27]. Warm ambient temperatures lead to changes in occupancy of the histone variant H2A.Z [34][35][36] and in histone modifications (histone deacetylation [37] and histone methylation [36,38]) at genes that respond to temperature [27]. At warm temperature, the eviction of H2A.Z-containing nucleosomes at transcription start sites facilitates thermoresponsive gene expression [34,35,39].
Warm temperatures also induce H3K9 deacetylation at genes that respond to temperature, which is dependent on HISTONE DEACETYLASE 9 and POWERDRESS [37]. H3K36me3 is involved in regulating temperature-induced alternative splicing, co-transcriptional regulation and flowering time control in Arabidopsis [36,38]. The dynamic regulation of H3K27me3 is also essential for thermoresponsive flowering time control, which involves REF6, JMJ13 and JMJ30. REF6, together with HEAT SHOCK TRANSCRIPTION FAC-TOR A2, form a heritable feedback loop to induce transgenerational thermomemory for flowering [40]. JMJ13 acts as a temperature and photoperioddependent flowering repressor [9]. JMJ30 prevents precocious flowering caused by warm ambient temperature by removing H3K27me3 on the FLOWERING LOCUS C promoter [10], and functions in heat acclimation [11].
These reports revealed that histone methylation controls the balance between vegetative and reproductive growth at warm ambient temperature [41]. However, how histone methylation functions in early thermomorphogenesis and precisely regulates ambient temperature response remains elusive. Here, we demonstrate that REF6, together with the key thermomorphogenetic transcription factor PIF4, regulates plant responses to warm ambient temperature through its H3K27me3 demethylase activity. This study reveals the molecular mechanism by which REF6 participates in the response to warm ambient temperature, and demonstrates the importance of the cooperation of epigenetic factors and transcription factors in regulating gene expression and environmental responses.

REF6 promotes responses to warm ambient temperature
Epigenetic regulation is highly dynamic and multiple transcription factors cooperate with epigenetic regulators to respond to environmental changes. However, little is known about how the epigenetic regulators target and function at their target genes. Our previous study showed that REF6 demethylates H3K27me3 at its target loci by recognizing specific DNA motifs, and is implicated in plant development and responses to stimuli [8,14,15]. Moreover, hypocotyl elongation in ref6 mutants is insensitive to warm ambient temperature. We then used hypocotyl elongation as a system to investigate the function of REF6 in targeting, H3K27me3 demethylation and activation of target genes. Three-day-old seedlings grown at 22 • C were transferred to 22 • C or 28 • C for 3 days in long-day  conditions (Fig. 1A). Hypocotyl elongation of wild-type Col, ref6 mutants (ref6-1 weak allele and ref6-5 null allele) and REF6 complementation lines (pREF6::REF6-HA ref6-1, REF6-HA hereafter), were recorded after 3 days of treatment at 22 • C or 28 • C (Fig. 1A). We found strong hypocotyl elongation responses were observed in wild-type Col seedlings at warm ambient temperature (28 • C). However, although the ref6 mutants grew normally (like the wild-type Col) at 22 • C, the mutants exhibited markedly attenuated responses at 28 • C compared with the wild-type seedlings ( Fig. 1B and C). The attenuated hypocotyl length of ref6 mutant at 28 • C is mainly due to the reduced cell length, but not the cell number ( Supplementary Fig. 1). REF6-HA transgenic lines rescued the attenuated hypocotyl elongation of the ref6 mutants at 28 • C ( Fig. 1B and C). These results suggest that REF6 is involved in regulating thermal responses.

Targeting of REF6 is independent of temperature changes
We next examined REF6 expression and REF6 targeting activity at warm ambient temperature. The mRNA levels of REF6, as measured by reverse transcription-quantitative Polymerase Chain Reaction (PCR) (RT-qPCR), did not change at 22 • C and 28 • C ( Supplementary Fig. 2), suggesting that REF6 expression is not induced by warm ambient temperature. To test whether elevated temperature affects the recruitment of REF6 to its target genes, we profiled the genome-wide localization of REF6 in wild-type Col using chromatin immunoprecipitation coupled with high-throughput sequencing (ChIP-seq) with anti-REF6 antibody [16] at 22 • C and 28 • C (Supplementary Table  1). The ref6-5 mutant was included as a negative control. Two biological replicates of REF6 ChIPseq showed a high correlation (Pearson's correlation coefficient, r = 0.86 at 22 • C, r = 0.90 at 28 • C) with each other ( Supplementary Fig. 3). A total of 1954 and 2136 peaks covering 1834 and 1981 genes were bound by REF6 at 22 • C and 28 • C, respectively. These were highly correlated (r = 0.98) and overlapped (Fisher's exact test, P < 2.2 × 10 -16 ) with each other ( Fig. 2A), indicating that elevated ambient temperature did not influence REF6 genome-wide targeting.
We then checked the H3K27me3 levels in wildtype Col and ref6 mutants by H3K27me3 ChIPseq at 22 • C and 28 • C (Fig. 2B, and Supplementary  Fig. 3). This analysis detected 288 high-confidence hypermethylated H3K27me3 sites in ref6 mutants at both 22 • C and 28 • C, 193 of which were REF6 binding targets (Supplementary Table 2). These results demonstrated that REF6 transcript levels, REF6 targeting and REF6 enzymatic activity are temperature insensitive.

Temperature-dependent gene activation is dependent on REF6
Because REF6 is an H3K27me3 demethylase and H3K27me3 is associated with gene repression, we next investigated the relationship between REF6 targeting and gene activation, and whether the increased H3K27me3 levels in the ref6 mutant influence target gene expression at warm temperatures. We therefore carried out an RNA sequencing (RNAseq) analysis, using RNA from seedlings of wildtype Col and ref6-5 mutant treated at 22 • C and 28 • C (Supplementary Table 3). A total of 74 million (79%), 76 million (80%), 77 million (82%) and 76 million (82%) clean reads of highly correlated RNA-seq replicates were uniquely aligned to the Arabidopsis reference genome, and were mapped to 17 092, 16 981, 17 189 and 17 127 genes in the wildtype Col and ref6-5 mutants at 22 • C and 28 • C, respectively Reads Per Kilobase Million (RPKM > 1) (Supplementary Table 3). RNA-seq data from three Natl Sci Rev, 2022, Vol. 9, nwab213   replicate samples showed high correlation (Supplementary Fig. 4). Strikingly, although REF6 targeting activity is temperature insensitive, principal component analysis showed significant differences in gene expression among the wild-type Col and ref6-5 mutants at 22 • C and 28 • C (Supplementary Fig. 4). For this analysis, we first calculated the fold change of transcript levels comparing 28 • C to 22 • C in wild-type Col and ref6 mutants (Fig. 2C). Since REF6 is an H3K27me3 demethylase and increased H3K27me3 in ref6 mutants leads to decreased gene expression [8], we hypothesized that the activation of REF6 target genes at increased temperature may be inhibited because of H3K27me3 hypermethylation in the ref6 mutant. Therefore, we mainly focused on the genes that were upregulated (efficiently induced) in wild-type Col at 28 • C compared with 22 • C, and inefficiently induced in ref6 mutants (yellow dots in Fig. 2C). This identified 337 genes that were upregulated (fold change ≥2) only in wild-type Col at 28 • C compared with 22 • C, but not in ref6 mutants (Fig. 2D). Among these 337 genes, 44 genes were REF6 binding targets, 14 of which were H3K27me3hypermethylated (Fig. 2E, Supplementary Table 4 and Supplementary Fig. 5). Among these 14 genes, GIBBERELLIN 20-OXIDASE 2 (GA20ox2) encodes an enzyme that is involved in GA biosynthesis and temperature responses [42], and the basic helixloop-helix transcription factor gene BASIC HELIX-LOOP-HELIX 87 (bHLH87) was reported as a PIF4 downstream target in Arabidopsis [43]. GA20ox2 and bHLH87 were upregulated by elevated temperature in wild-type Col, but were inefficiently induced in ref6-5 mutants with high levels of H3K27me3 ( Fig. 3A and E). Therefore, we chose these two genes to dissect the regulatory role of REF6 in thermomorphogenesis.  Validation of REF6 binding and H3K27me3 hypermethylation in ref6 at GA20ox2 and bHLH87 by ChIP and quantitative PCR (ChIP-qPCR) with an independent batch of ChIP samples yielded consistent results (Fig. 3C, D, G and H). RT-qPCR analysis also confirmed that the upregulation of GA20ox2 and bHLH87 in wild-type Col at warm temperature was not fully activated in the ref6 mutant ( Fig. 3B and F). These results indicated that, although warm temperature does not influence REF6 targeting and enzymatic activity, it does affect the activation of thermoresponsive genes, which are direct targets of REF6.
The ref6 thermomorphogenesis phenotype is at least partly due to downregulated GA20ox2 and bHLH87 expression To further explore whether the reduced response to warm temperature of ref6 is associated with decreased GA20ox2 and bHLH87 expression, we generated ga20ox2 and bhlh87 Clustered Regularly Interspaced Short Palindromic Repeats/Cas9 knockout mutants in wild-type Col and the ref6-5 mutant background (Supplementary Fig. 6). Evaluation of hypocotyl elongation at warm temperature in the wild type, double mutants and the corresponding parental lines showed that both ga20ox2 and bhlh87 are thermo-insensitive, similar to the ref6-5 mutant (Fig. 4). Interestingly, in seedlings treated with 10 μM exogenous GA, the reduced hypocotyl elongation of ref6-5 and ga20ox2 mutants could be partially rescued at warm temperature ( Fig. 4A and B, and Supplementary Fig. 7). These results indicate that REF6 regulates thermomorphogenesis at least partly through demethylating H3K27me3 and upregulating expression of GA20ox2 and bHLH87.

REF6 enzymatic activity is necessary for warm ambient temperature responses
REF6 possesses a JmjC enzymatic domain at its N terminus, and a tandem array of C2H2 zinc finger domains at its C terminus [8]. REF6 function Three-day-old seedlings of the indicated genotypes grown at 22 • C were transferred to 22 • C or 28 • C for 3 days, after which representative plants were imaged (A, scale bars = 1 mm) and the hypocotyl length of each plant was measured (B). N is marked in the column, and error bars depict ± s.e.m. Student's t-test was used to calculate the P value between the indicated genotypes, and significant differences are shown by different letters (P < 0.01). The dots denote individual data points. n.s. not significant. (C and D) ChIP-qPCR of (C) REF6 binding and (D) H3K27me3 levels at bHLH87 locus using ChIP samples from the indicated plants. Three and four technical replicates were performed for each sample, respectively. (E) Transcript levels of bHLH87 were measured by RT-qPCR and normalized to ACTIN. mREF6 indicates REF6 H246A. Data are shown as means ± s.e.m. from three technical replicates. Significant differences are shown by different letters (P < 0.01), as determined by Student's t-test. and genome-wide targeting involve it recognizing specific DNA motifs via its tandem zinc finger domains [14,15]. Therefore, we wondered about the relationships among targeting, enzymatic activity and transcriptional activation of REF6, in the context of warm ambient temperature. To test this, we transformed the ref6 mutant with a construct encoding full-length REF6 with a mutation (H246A) that abolishes its enzymatic activity (pREF6::REF6 H246A-HA, referred to as REF6 H246A-HA hereafter) [8]. Transgenic lines that showed REF6 transcript levels comparable to those of wild-type Col were chosen for further analysis ( Supplementary  Fig. 8). Phenotypic analysis showed that REF6 H246A-HA ref6-1 transgenic lines did not rescue the attenuated hypocotyl elongation of the ref6 mutants at 28 • C (Fig. 5A and B), suggesting that the enzymatic activity of REF6 is essential for temperature responses.
To investigate whether REF6 enzymatic activity affects its targeting activity and transcription activation, REF6 binding, H3K27me3 status and gene activation were further validated by ChIP-qPCR and RT-qPCR of bHLH87 and GA20ox2 (Fig. 5C-E, and Supplementary Fig. 9). REF6 binding signals at bHLH87 and GA20ox2 were identical in REF6-HA and REF6 H246A-HA (Fig. 5C, and Supplementary  Fig. 9A). However, H3K27 residues at these loci in REF6 H246A-HA plants were hypermethylated as in ref6 (Fig. 5D, and Supplementary Fig. 9B), and the expression of bHLH87 and GA20ox2 was not activated in REF6 H246A-HA (Fig. 5E, and Supplementary Fig. 9C). Together, these results indicated that binding and enzymatic activity of REF6 are prerequisites for activation of thermoresponsive genes. Just binding of REF6 to its target loci, without enzymatic activity, cannot activate thermoresponsive gene expression, even at warm ambient temperature. Expression levels were normalized to ACTIN. Error bars ± s.e.m. (F) Model of how REF6 works with PIF4 to promote responses at warm ambient temperature. At 22 • C, REF6 binds some thermoresponsive genes and demethylates H3K27me3, placing these genes in a poised state where they are ready for expression. At 28 • C, thermo-induced transcription factors bind the thermoresponsive genes, activating their expression. However, H3K27me3 cannot be removed from these thermoresponsive genes in ref6 mutants, so they cannot be efficiently activated at warm ambient temperature.

REF6 and PIF4 are required for warm ambient temperature responses
The bHLH transcription factor PIF4 accelerates hypocotyl growth at warm temperatures and acts as a central regulatory hub in ambient-temperature signaling, in which various signaling pathways modulate PIF4 activity, and PIF4 integrates hormone signaling in thermoresponsive regulation of growth [25,26,28]. By analyzing public ChIP-seq data sets [43], we identified 656 common targets of REF6 and PIF4, and 144 genes that have high levels of H3K27me3 in the ref6 mutant (Fig. 6A). This set of overlapping genes includes bHLH87 (Fig. 6B).
By analyzing public transcriptome data sets [44], we compared the expression profiles of upregulated and downregulated genes between ref6 and pif4 mutants. We also constructed heat maps and conducted k-means clustering of 715 upregulated genes and 564 downregulated genes in 28 • C compared with 22 • C. Among these differentially expressed genes, 227 upregulated genes and 185 downregulated genes showed similar expression patterns in ref6 and pif4 mutants (Supplementary Fig. 10). These results indicate that REF6 and PIF4 cooperate with each other to regulate some thermoresponsive genes.
To determine whether PIF4 and REF6 directly bind to and activate bHLH87, we performed electrophoretic mobility shift assays (EMSAs) using DNA fragments from bHLH87 that contain the CATATG and CTCTGTTT motifs (Supplementary Fig. 11). The DNA probes containing the CATATG motif and CTCTGTTT motif were incubated with maltose-binding-protein (MBP)-tagged PIF4 (PIF4-MBP) and Glutathione-S-transferase (GST)-tagged C-terminal REF6 (GST-REF6C). Consistent with the ChIP-seq data, PIF4 and REF6 can directly bind to bHLH87 DNA probes (Supplementary Fig. 11). Genetic analysis showed that pif4-101 ref6-5 plants displayed similar phenotypes to pif4-101 plants (Fig. 6C and D). Overexpression of PIF4 in Col enhanced the hypocotyl elongation, while the hypocotyl elongation of plants overexpressing PIF4 in ref6-5 was not enhanced as much as in Col at warm ambient temperature (Fig. 6C and  D), indicating that REF6 and PIF4 may work together in the response to warm temperatures.

DISCUSSION
Epigenetic regulation is highly dynamic and diverse during development, in different cell types, and in response to changing environmental conditions. Here, we showed that the H3K27me3 demethylase REF6 regulates temperature responses in Arabidopsis and its targeting and enzymatic activity are essential for thermoresponsive gene activation. Loss-of-function ref6 mutants exhibited an attenuated hypocotyl elongation phenotype (Fig. 1). Genome-wide analysis showed that REF6 directly associates with and activates the thermoresponsive genes GA20ox2 and bHLH87 (Figs 2-4), and REF6 enzymatic activity is essential for gene activation (Fig. 5). Interestingly, activation of bHLH87 by warm temperature is regulated by both REF6 and the key thermoresponsive transcription factor PIF4 (Fig. 6). Together, the results of our study reveal the molecular mechanism by which REF6 participates in the response to warm ambient temperature, and demonstrate the importance of the cooperation of epigenetic factors and transcription factors in regulating gene expression and environmental responses.
Warm ambient temperatures lead to epigenetic changes, including H2A.Z nucleosomal dynamics, histone modifications and chromatin remodeling [27]. Here, we demonstrated that H3K27me3 demethylation mediated by REF6 controls plant thermal responses. At warm ambient temperatures, H2A.Z is evicted from nucleosomes, inducing thermoresponsive genes [32,34,35]. A recent study found that temperature-induced H2A.Z eviction at PIF4 targets is mediated by the histone deacetylase HDA9/PWR and the ATP-dependent chromatin remodeler INO80, and established the link between H2A.Z eviction and active transcription [37,39,45]. Since there is a strong correlation between H3K27me3 and H2A.Z enrichment [46], and the genomic targeting of INO80 requires transcription factors or chromatin regulators [45], H3K27me3 demethylation mediated by REF6 may work in concert with these factors to regulate their targets in response to thermal responses and other environmental stimuli. It will be of interest to further explore the detailed mechanism linking H3K27me3 demethylation, histone deacetylation, H2A.Z eviction, transcription factors and transcription activation.
Plants actively sense temperature and, under warm ambient temperatures, they promote the biosynthesis and localization of signaling molecules, such as auxin, BR and GA, to optimize plant morphology in a coordinated manner [27,47]. Our results found that the genes encoding key enzymes in auxin, BR and GA biosynthesis, such as YUC8, BR6ox and GA20ox2, respectively, are direct targets of REF6. However, only the expression of GA20ox2, not BR6ox1 and YUC8, was induced by warm ambient temperature, which was inefficiently induced in the ref6 mutant. This indicates that REF6 promotes thermomorphogenetic hypocotyl growth under warm temperatures by elevating thermoresponsive hormone levels (mainly GA). Thermoresponsive hormone signaling in different plant organs should be coordinated to facilitate synchronized growth and acclimation to elevated temperatures [47]. Whether and how REF6 affects hormone biosynthesis and thermomorphogenetic traits in different plant organs (such as shoots and roots), and whether they are linked with each other via interorgan communication, will be an interesting subject for further study.
Transcription factors bind DNA in a sequencespecific manner and regulate gene expression, ensuring the appropriate level of gene expression at the right time and right place. Long-standing debates have considered the relative importance of epigenetic modifications and transcription factors in cell fate determination and other cellular processes [48,49]. At present, transcription factors and epigenetic modifiers are believed to work together to maintain normal growth and development. For instance, the transcription factor BACH1 recruits H3K4 methyltransferase complexes to maintain pluripotency in mouse embryonic stem cells [50]. In another example, the plant pioneer transcription factor LEAFY [51] and the key thermoresponsive transcription factor PIF4 [45] work with chromatin remodeling factors to open chromatin and activate transcription. Here, our detailed examination of H3K27me3 and expression levels at the bHLH87 locus under elevated temperature suggest that bHLH87 activation is regulated by both H3K27me3 demethylation and the thermoresponsive transcription factor PIF4, but neither REF6 nor PIF4 alone can activate bHLH87 expression (Fig. 6). The activation of other thermoresponsive genes targeted by REF6, such as GA20ox2, likely requires other transcription factors that have yet to be identified. Our results provide direct evidence for the interdependent and indispensable relationship between epigenetic modifications and transcription factors with regard to synergistically promoting the expression of genes required for the response to warm temperature. This tight and indispensable control of gene activation at warm temperature is significant for the integration of temperature signals into plant morphogenesis.
How epigenetic modifiers recognize their target loci and modify the chromatin is a long-standing, fundamental question in epigenetic regulation. Our previous study showed that REF6 functions involve both self-targeting (i.e. targeting that does not require another DNA-binding protein) and DNA methylation [14,16]. Here, we demonstrated that, in addition to self-targeting, the enzymatic activity of REF6 is essential for gene activation. Expression of a mutant REF6 protein that lacks enzymatic activity (REF6 H246A) did not rescue the attenuated hypocotyl elongation of the ref6 mutants at 28 • C, but the mutation did not affect REF6 targeting to thermoresponsive target genes, suggesting that both catalytic activity and self-targeting are critical for REF6 function (Fig. 5). The targeting and enzymatic activity of a chromatin-modifying enzyme likely must be fine-tuned; therefore, additional factors, such as chromatin remodelers linking histone variants and nucleosome position, various epigenetic modifications, chromatin structure and chromatin context, may also influence its targeting and function. Further exploration of how these factors cooperatively regulate the function of a chromatin-modifying enzyme to maintain appropriate chromatin status in the genome remains an exciting topic for future research.

Plant growth conditions and hypocotyl length analysis
Surface-sterilized seeds were sown on half-strength Murashige and Skoog (MS) medium supplemented with 0.8% (w/v) agar and 1% (w/v) sucrose. The plates were stratified at 4 • C in darkness for 3 days, then transferred into a growth chamber (AR-22L, Percival Scientific) and germinated at 22 • C for 3 days, then transferred to 28 • C (treatment) or maintained at 22 • C (control) at ZT0 (Zeitgeber Time) for another 3 days under a 16-h light/8-h dark cycle. The plants were grown under white fluorescent light with a fluence rate of ∼75 μmol/m 2 /sec. Seedlings from each sample were scanned using ZEISS SteREO Discovery V20, and the hypocotyl length was measured using ImageJ software (National Institutes of Health).

Chromatin immunoprecipitation assays and library preparation
About 2.5 g of seedlings was used for ChIP assay, as previously described [14,16], using anti-REF6 antibody (custom made by Abmart) and anti-H3K27me3 antibody (Cell Signaling Technology, C36B11). The ChIP DNA was subjected to qPCR analysis or Illumina sequencing.

ChIP-seq analysis
All sequencing reads of ChIP-seq were aligned to the Arabidopsis thaliana TAIR10 reference genome using Bowtie2 (version 2.2.1) [52] with default parameters after removal of the adapter sequences and low-quality bases by Cutadapt (version 1.16) and PRINSEQ (version 0.20.4) [53]. After filtering out PCR duplicates with SAMtools (version 0.1.8) [54], only uniquely mapped reads were used for downstream analysis. The first read of paired-end reads was used to generate normalized genome coverage tracks by deepTools2 (version 2.5.0.1, https://deeptools.readthedocs.io/en/develop/) with the '-normalizeUsing RPKM -extendReads 200 -binSize 50' options. Then the output BigWig files were visualized using the Integrative Genomics Viewer [55]. MACS2 version 2.1.1 [56] was used to call peaks with the '-gsize 1.19e8 -keep-dup 1'. To identify the differentially modified regions, REF6 binding peaks and H3K27me3 peaks were merged using BEDtools (version 2.17.0) [57]. Then the signal intensity in merged peak regions was estimated by read counts per million mapped reads (RPM) after being normalized to library size. Only the regions with fold change greater than 4 between two different conditions were considered as differentially modified regions. All peaks were annotated with the priority order (promoter > exon > intron > downstream > intergenic) using ChIPseeker (version 1.17.1) [58] when a single peak spanned two genomic features.

RNA-seq analysis
For RNA-seq data, all reads were mapped to the TAIR10 reference genome using HISAT2 (version 2.0.5) [59] with default parameters. To count the uniquely mapped reads, featureCounts (version 1.6.2) [60] was used to calculate the read count per gene and RPKM value. Then the differentially expressed genes were identified by DESeq2 (version 1.20.0) [61] with these criteria: q-value < 0.05, log 2 (fold change) > 2.

DATA AVAILABILITY
ChIP-seq and RNA-seq data from this article can be found in the Gene Expression Omnibus data library under accession number GSE181292.

SUPPLEMENTARY DATA
Supplementary data are available at NSR online.