The Arabidopsis demethylase REF6 physically interacts with phyB to promote hypocotyl elongation under red light

Significance

Photomorphogenesis, a key stage in early seedling growth exposed to light, is mediated by the photoreceptors. The dynamic trimethylation of histone H3 at Lys27 (H3K27me3) modulates gene expression and governs plant development and responses to the environmental changes. However, how chromatin and histone modifications respond to light signals remains largely obscure. This study uncovers a critical role of the H3K27me3 demethylase RELATIVE OF EARLY FLOWERING 6 (REF6)/JMJ12 in regulating light-responsive hypocotyl elongation. REF6 is activated by light and interacts with phytochrome B (phyB) to modulate chromatin accessibility and expression of cell elongation-related genes. Our findings uncover a link between light signals and epigenetic regulation, offering valuable insights into understanding the complex interactions among phytochrome, epigenetic factors, and transcription factors.

Abstract

The plant photoreceptor phytochrome B (phyB) mediates the responses of plants to red (R) light. Trimethylation of histone H3 at Lys27 (H3K27me3) plays a crucial role in governing gene expression and controlling the response of plants to environmental changes. However, how dynamic H3K27me3 mediates plant response to R light is poorly understood. Here, we report that RELATIVE OF EARLY FLOWERING 6 (REF6), an H3K27me3 demethylase, promotes hypocotyl elongation under R light in Arabidopsis. Upon exposure to R light, REF6 preferentially interacts with the active Pfr form of phyB. Consequently, phyB enhances REF6 accumulation and its binding ability, which are necessary for inducing cell-elongation-related genes from open chromatin, ensuring normal plant growth under prolonged light exposure. Moreover, REF6 acts together with the phyB-PIF4 module to mediate light regulation of hypocotyl growth. These findings provide insights into the understanding of how phytochromes, epigenetic factors, and transcription factors coordinately control plant growth in response to changing light environment.

Light provides energy via photosynthesis for plant growth, and plants alter their morphology in response to its intensity and quality. Plants perceive light by photoreceptors, with the phytochromes primarily absorbing red (600 to 700 nm) and far-red (700 to 750 nm) light (1). Phytochrome B (phyB), the major red-light receptor, exists in two photo-interconvertible forms, Pfr (far-red-light-absorbing, biologically active) and Pr (red-light-absorbing, biologically inactive) (1–5). Upon red light (RL) exposure, the Pr form in the cytosol is converted to Pfr, which translocates into the nucleus (6, 7). As a major hub of light signal transduction, phyB regulates several downstream light-responsive factors, such as the transcription factors PHYTOCHROME-INTERACTING FACTORS (PIFs) and ELONGATED HYPOCOTYL 5 (HY5)—negative and positive regulators of photomorphogenesis. Proper photomorphogenesis is characterized by reduced hypocotyl length, vivid green pigmentation from chlorophyll biosynthesis, and fully expanded, photosynthetically competent cotyledons (2). Upon light exposure, phyB directly interacts with PIFs, rapidly inducing their phosphorylation, ubiquitination, and degradation (8–21). By contrast, phyB interacts with and stabilizes the transcription factor MYB DOMAIN PROTEIN 30 (MYB30). MYB30 positively regulates hypocotyl elongation by promoting PIF4 and PIF5 expression and protein accumulation after prolonged red-light exposure (22). Therefore, phyB has complex regulatory interactions with its downstream factors.

Epigenetic regulation has a prominent role in light responses and photomorphogenesis. Light induces deposition of the histone variant H2A.Z at hypocotyl-elongation-related genes. The Pfr form of phyB physically interacts with the SWI2/SNF2-RELATED 1 (SWR1) chromatin-remodeling complex and facilitates H2A.Z deposition at YUCCA9 (YUC9, encoding an enzyme for biosynthesis of the growth hormone auxin), inhibiting plant growth under RL (23–25). phyB also interacts with VIN3-LIKE 1 (VIL1, also known as VRN5), a component of the histone H3 at Lys27 (H3K27) methyltransferase Polycomb Repressive Complex 2 (PRC2). The PRC2–phyB interaction promotes trimethylation of H3K27 (H3K27me3) at ARABIDOPSIS THALIANA HOMEOBOX PROTEIN 2 (ATHB2) to inhibit its expression (26, 27). However, neither phyB nor VIL1 possess a DNA-recognition domain. Therefore, it is likely that phyB regulates chromatin status by interacting with trans-acting factors, such as transcription factors. The precise mechanisms by which phyB recognizes its target sites and modulates chromatin status warrant further investigation (26). Although phyB is involved in epigenetic regulation, the regulatory relationship between phyB and epigenetic factors, especially histone-modification regulators, is little known.

In addition to phyB, HY5 and PIFs are also involved in epigenetic regulation under light conditions (28–33). HY5 interacts with the histone deacetylase HDA15 to corepress auxin-signaling-related genes by decreasing the levels of histone H4 acetylation in a light-dependent manner, thus promoting photomorphogenesis (32). Furthermore, HY5 interacts with the chromatin remodeling factor PICKLE (PKL), and its activity is negatively regulated by PKL through the inhibition of H3K27me3 at target loci (28, 34). Moreover, PIF3 interacts with PKL and facilitates PKL’s binding to target genes, thereby coregulating skotomorphogenesis (35). PIF7, a major transcriptional regulator of the shade-avoidance syndrome in Arabidopsis thaliana (Arabidopsis), recruits the H3K4me3- and H3K36me3-binding proteins MORF-RELATED GENE 1 (MRG1) and MRG2, respectively, and brings histone acetylases to its target genes to induce histone acetylation and induce shade-responsive genes (31). PIF7 also recruits the histone chaperones ANTI-SILENCING FACTOR 1 (ASF1) and HISTONE REGULATOR HOMOLOG A (HIRA), which mediate deposition of the H3.3 variant into chromatin and promote shade-elicited gene activation (33).

The dynamic balance of H3K27 methylation in plant genomes is regulated by PRCs and the Jumonji-domain-containing histone demethylases (JMJs), namely EARLY FLOWERING 6 (ELF6/JMJ11) (36), RELATIVE OF EARLY FLOWERING 6 (REF6/JMJ12) (37), JMJ13 (38), and JMJ30 (39, 40). These demethylases spatially and temporally restrict H3K27me3 marks and activate gene expression via distinct targeting mechanisms (41–43). For instance, ELF6 physically interacts with SET DOMAIN GROUP 8 (SDG8, a H3K36 methyltransferase), and SDG8 colocalizes with RNA Pol II at the floral repressor locus, FLOWERING LOCUS C (FLC). ELF6 and SDG8 influence the accumulation of the associated histone modifications, H3K27me3 and H3K36me3, separately, ensuring the epigenetic reprogramming of FLC in the next generation (36, 44). JMJ13 acts as a temperature- and photoperiod-dependent flowering repressor through specifically recognizing H3K27me3 by hydrogen bonding and hydrophobic interactions (38).

Unlike other demethylases mentioned above, REF6 recognizes specific DNA motifs (CTCTGYTY) through its C₂H₂ zinc-finger (C₂H₂-ZnF) domains, while non-CG DNA methylation and closed-chromatin status affect REF6 targeting (45–47). REF6 is involved in seed dormancy, stem-cell development, lateral-root development, and leaf senescence (41, 48–52). It is also involved in responses to high temperature, in which REF6 cooperates with PIF4 and HEAT-SHOCK TRANSCRIPTION FACTOR A2 (HSFA2) to activate thermo-responsive genes (53, 54). However, the regulatory mechanism of REF6 in controlling plant growth and development under light conditions remains undiscovered. Here, we demonstrate that Arabidopsis REF6 is required for genome-wide H3K27me3 demethylation and chromatin accessibility to promote expression of cell-elongation-related genes by directly binding to CTCTGYTY motifs under RL. REF6 preferentially physically interacts with the Pfr form of phyB. Subsequently, phyB enhances the accumulation and binding ability of REF6 and coordinates the expression of DUF668 with REF6 and PIF4 to mediate hypocotyl elongation under light. This study reveals the molecular mechanism by which REF6 responds to light cues and illuminates the cooperation among phytochromes, epigenetic factors, and transcription factors in regulating gene expression and photomorphogenesis. This regulatory mechanism enables seedlings to maintain appropriate growth under variable light conditions.

Results

REF6 Is the Principal H3K27 Demethylase Positively Regulating Hypocotyl Elongation under Light.

Due to epigenetic modifications being dynamically regulated in response to light (26), we wondered whether H3K27me3 demethylases, such as REF6/JMJ12, ELF6/JMJ11, JMJ13, JMJ30, and JMJ32, are involved in photomorphogenesis. To test this, we grew wild-type Col-0 and ref6-3, ref6^C, elf6-3, jmj13^G, elf6-3 ref6^C, elf6-3 jmj13^G, ref6-3 jmj13^G, ref6-3 elf6-3 jmj13^G(41), and jmj30 jmj32 (39) mutant seedlings in darkness (DK) or continuous white light for 4 d. Changes in hypocotyl lengths (as a readout of photomorphogenesis) for mutant seedlings were similar to Col-0 in DK (SI Appendix, Fig. S1 A and B). In white light, however, ref6-3, ref6^C, elf6-3 ref6 ^C, elf6-3 jmj13^G, ref6-3 jmj13^G, and ref6-3 elf6-3 jmj13^Gmutant plants, but not the elf6-3, jmj13^G, and jmj30 jmj32 mutants, had shorter hypocotyls than Col-0 (SI Appendix, Fig. S1 A and B). Although elf6-3 and jmj13^G single mutants had no obvious hypocotyl defects, the elf6-3 ref6 ^C double mutant and the ref6-3 elf6-3 jmj13^G triple mutant had much shorter hypocotyls than ref6 (SI Appendix, Fig. S1 A and B), suggesting that ELF6 and JMJ13 are partially redundant with REF6 in regulating hypocotyl growth in light.

To further investigate the response of REF6 to monochromatic light, we grew Col-0, two ref6 mutants (ref6-1 and ref6-2), and two independent transgenic REF6–HA complementation lines (45) in DK or continuous red, far-red, or blue light for 4 d and measured hypocotyl lengths. Both ref6 mutant alleles had shorter hypocotyls than Col-0 in red, far-red, and blue light, but not in DK (Fig. 1 A and B). A tagged REF6 transgene under the control of the native REF6 promoter (designated REF6_pro: REF6–HA) successfully rescued the short-hypocotyl phenotype of ref6-1 mutants (Fig. 1 A and B), confirming this phenotype is due to the loss of REF6 function. These results indicate that REF6 is the principal H3K27me3 demethylase that positively regulates hypocotyl elongation under light in Arabidopsis seedlings.

Fig. 1.

REF6 accumulates under RL. (A) Phenotypes of wild-type Col-0, ref6 mutants, and two independent complementation lines expressing a REF6_pro: REF6–HA construct in the ref6-1 background. Seedlings were grown for 4 d in DK or in continuous red (25 μmol m^–2 s^–1), far-red (3 μmol m^–2 s^–1), or blue light (10 μmol m^–2 s^–1). (Scale bar, 1 mm.) (B) Hypocotyl lengths for Col-0, ref6 mutants, and two independent complementation lines expressing a REF6_pro: REF6–HA construct in the ref6-1 background grown for 4 d in the light conditions indicated in (A). Error bars represent SD of the mean from ≥12 seedlings. Different letters represent statistically significant differences between means determined by one-way ANOVA with Tukey’s HSD test (P < 0.05). (C) Anti-REF6 immunoblot of Col-0 and ref6-5 seedlings grown for 4 d in the light conditions indicated in (A). HSC70 was used as a loading control. Numbers below the immunoblots indicate the relative intensities of REF6 bands normalized to those of loading controls, and the ratio was set to 100 for Col-0 which grown in continuous RL. (D) Histological analysis of GUS staining in transgenic REF6_pro: gREF6–GUS 1# seedlings grown for 4 d in DK or continuous red, far-red, or blue light in the light intensity indicated in (A). (Scale bar, 200 µm.)

Next, we asked whether REF6 expression is responsive to light by growing wild-type Col-0 seedlings under different wavelengths. Compared with dark-grown controls, REF6 transcript had no significant changes in continuous red, far-red, and blue light (SI Appendix, Fig. S2). To detect if the abundance of REF6 protein is regulated by light, we grew wild-type Col-0 and ref6-5 mutant seedlings in DK or continuous red, far-red, and blue light for 4 d and then analyzed REF6 abundance by immunoblotting with anti-REF6 antibodies. REF6 accumulated to much higher levels in all light conditions than DK, particularly in RL (Fig. 1C).

To further characterize how light regulates the spatial accumulation of REF6, we generated transgenic REF6_pro: gREF6–GUS reporter lines in the ref6-1 background using REF6 genomic DNA and observed β-glucuronidase (GUS) staining by histology. REF6–GUS signal was mainly located in the cotyledons, shoot apex, and root tip (Fig. 1D). In red-light-grown seedlings, the REF6-GUS signal was stronger in the shoot apex, petiole, cotyledons, and hypocotyl compared to seedlings grown in DK (Fig. 1D and SI Appendix, Fig. S3), consistent with REF6 being enriched in RL (Fig. 1C). Taken together, these results demonstrate that REF6 is induced by light, which facilitates the response to light and promotes hypocotyl elongation.

REF6 Physically Interacts with the Pfr form of phyB.

To gain a deeper insight into REF6 function in light, we ran immunoprecipitation–mass spectrometry (IP–MS) to identify interacting partners using transgenic plants constitutively expressing REF6–YFP in the Col-0 background. REF6–YFP physically interacts with phyB, as well as other core light-responsive factors, including HMR (55–57) and COP1 (58, 59) (SI Appendix, Table S1). Therefore, to confirm whether REF6 directly interacts with phyB, we employed yeast two-hybrid assays using truncated REF6 and phyB variants and found that the region from 600 to 900 amino acids (aa) of REF6 could interact with the N-terminal domain of phyB (i.e., the phyB apoprotein) in yeast (Fig. 2 A and B). We then asked which form of phyB interacted most strongly with REF6 using a modified yeast two-hybrid system (60) containing phycocyanobilin isolated from Spirulina as the chromophore, which enables phyB to transform into Pfr and Pr forms in yeast after red and red + far-red light treatments, respectively. REF6 preferentially interacted with the Pfr form in yeast (Fig. 2C).

Fig. 2.

REF6 prefers to interact with the Pfr form of phyB. (A) Diagram of bait proteins used for yeast two-hybrid assays (PHYB–N and PHYB–C fused with LexA DNA-binding domains and REF6 fragments A–E fused with DNA-activation domains [AD]). (B) Yeast two-hybrid assay of the physical interactions between the truncated forms of PHYB and REF6. The N- and C-terminal domains of PHYB were fused to the DNA-binding domain and REF6 fragments A–E were fused to the AD. (C) GAL4 yeast two-hybrid assay showing that REF6 preferably interacts with the Pfr form of phyB in yeast. Error bars represent the SD of three independent yeast cultures. Different letters represent statistically significant differences between means as determined by one-way ANOVA with Tukey’s HSD test (P < 0.05). (D) Co-IP assay probing the interaction between REF6 and phyB in vivo. Proteins were extracted from 4-d-old Col-0 and transgenic 35S_pro: phyB–GFP seedlings grown in darkness (D) then exposed to either red light (R) or far-red light (FR) for 30 min. The numbers below the immunoblots represent the relative intensities of the REF6 bands, normalized to the loading controls, with the ratio set to 100 for the first lane of the IP sample. (E) Anti-REF6 immunoblot of 4-d-old wild-type Col-0 and phyB-9 mutant seedlings grown in DK or continuous RL. The numbers below the immunoblots represent the relative intensities of REF6 bands, normalized to the loading controls. The ratio was set to 100 for Col-0 grown under continuous RL. (F and G) Seedling phenotypes (F) and hypocotyl lengths (G) of wild-type Col-0, ref6-5, phyB-9, and phyB-9 ref6-5 mutant seedlings grown in RL or DK for 4 d. (Scale bar, 1 mm.) Error bars represent SD from >12 seedlings. Different letters represent statistically significant differences between means determined by one-way ANOVA with Tukey’s HSD test (P < 0.05).

To verify if the physical interaction between REF6 and phyB occurs in vivo, we performed coimmunoprecipitation (co-IP) assays using wild-type Col-0 and transgenic phyB–GFP (61). The seedlings were grown in darkness for 4 d and then exposed to red light or far-red light for 30 min. REF6 was coprecipitated with anti-GFP antibodies in the phyB–GFP plants, but not in Col-0 controls. Notably, larger amounts of REF6 coprecipitated with phyB–GFP after the D→R switch than those after the D-to-FR treatment (Fig. 2D), indicating that REF6 preferentially interacts with the Pfr form of phyB in a light-dependent manner in vivo.

Because REF6 is enriched in RL (Fig. 1 C and D) and physically interacts with the Pfr form of phyB (Fig. 2 C and D), we wondered whether REF6 enrichment is regulated by phyB. To test this, we grew wild-type Col-0 and phyB-9 (62) knockout mutant seedlings in continuous DK or RL for 4 d. While REF6 transcript was slightly decreased (SI Appendix, Fig. S4), REF6 was less abundant in the phyB-9 mutant in RL than Col-0 (Fig. 2E). Taken together, these results demonstrated that REF6 preferentially interacts with the active Pfr form of phyB, which promotes REF6 expression in RL.

To explore the genetic relationship between phyB and REF6 in light-responsive hypocotyl elongation, we generated ref6-5 phyB-9 double mutants by genetic crossing. Hypocotyl lengths of ref6-5 phyB-9 double-mutant seedlings were in between those of phyB-9 and ref6-5 single mutants in RL and were similar to Col-0 (Fig. 2 F and G). This indicated that the long hypocotyls of the phyB-9 mutant could be restored to the wild type by the ref6-5 mutation.

REF6’s DNA-Binding Ability and Enzymatic Activity Are Necessary for Red-Light Responses.

REF6 possesses a JmjC enzymatic domain at its N terminus and C₂H₂-ZnF domains at its C terminus (45). H3K27me3 is specifically demethylated by REF6 for gene activation in multiple plant developmental processes and responses to environmental stimuli by recognizing specific CTCTGYTY motifs via the C₂H₂-ZnF domain (37, 41, 45, 48–50, 52, 53). Here, we examined whether enzymatic activity and DNA-binding ability are required for proper REF6 function under RL. To test this, we used ref6-1 mutant lines harboring transgenes (45, 53) encoding full-length REF6 fused to a hemagglutinin tag (REF6_pro: REF6–HA, REF6–HA hereafter), full-length REF6 with the H246A mutation that abolishes its enzymatic activity (REF6_pro: REF6^H246A–HA, REF6^H246A–HA hereafter), and a truncated REF6 lacking the C₂H₂-ZnF domains (REF6_pro: REF6^ΔZnF–HA, REF6^ΔZnF–HA hereafter), all driven by the REF6 promoter. In RL, REF6–HA completely rescued the ref6-1 short-hypocotyl phenotype. This was partially rescued by REF6^H246A–HA plants, but not by REF6^ΔZnF–HA plants (Fig. 3 A and B). This suggests that REF6 DNA-binding ability and enzymatic activity are essential for full red-light responses.

Fig. 3.

REF6 enzymatic activity and DNA-binding ability are essential for proper red-light responses. (A) Phenotypes of Col-0, ref6 mutants, and transgenic REF6–HA, REF6^H246A–HA, and REF6^ΔZnF–HA complementation lines in the ref6-1 mutant background grown for 4 d in DK or continuous RL. (Scale bar, 1 mm.) (B) Hypocotyl lengths of Col-0, ref6 mutants, and transgenic REF6–HA, REF6^H246A–HA, and REF6^ΔZnF–HA complementation lines in the ref6-1 mutant background grown for 4 d in DK or continuous RL. Error bars represent SD from ≥12 seedlings. Different letters represent statistically significant differences in means determined by one-way ANOVA with Tukey’s HSD test (P < 0.05).

REF6 Is Required for Genome-Wide H3K27me3 Demethylation and Chromatin Accessibility to Induce Cell-Elongation-Related Genes under RL.

The DNA-binding ability and enzymatic activity of REF6 are critical for red-light-induced hypocotyl elongation (Fig. 3). This prompted us to survey enrichment of REF6 targets via chromatin immunoprecipitation with anti-REF6 antibody followed by sequencing (ChIP-seq) of wild-type Col-0 and ref6-5 mutant seedlings grown under RL for 4 d. We found 3,735 REF6 target genes under RL conditions (SI Appendix, Fig. S5E and Datasets S1 and S2). Next, we performed H3K27me3 ChIP-seq using Col-0 and ref6-5 seedlings grown in DK and RL and found 1,125 H3K27me3-hypermethylated regions covering 918 genes (>twofold) in ref6-5 under RL (SI Appendix, Figs. S5 A–D and S6 and Datasets S1 and S3). H3K27me3 levels in ref6-5 under RL were higher than that under DK. This indicates that many more genes are demethylated by REF6 in RL (SI Appendix, Fig. S6). Among these, 688 genes were REF6 targets (Fig. 4A and SI Appendix, Fig. S7). Gene-ontology (GO) analysis of these targets revealed involvement in cell growth and morphogenetic processes (SI Appendix, Fig. S8), suggesting that REF6 acts in photomorphogenesis by demethylating H3K27me3 of growth-related target genes.

Fig. 4.

REF6 is required for H3K27 demethylation and chromatin opening for cell-elongation-related genes in RL. (A and B) The heat map of dynamic H3K27me3 (A) and ATAC-seq analysis of chromatin accessibility (B) were generated for wild-type Col-0 and ref6-5 mutants grown in DK and RL. This analysis focused on 688 overlapping genes that exhibited hypermethylated H3K27me3 in ref6-5 mutants under RL, as well as REF6 binding targets in the same conditions. (C) Scatter plot of differentially expressed genes in wild-type Col-0 and ref6-5 grown under RL. Red dots represent genes up-regulated in ref6-5 under RL and blue dots represent down-regulated genes. The number of genes is marked on the top. The x axis represents log-transformed fold change (ref6-5/Col-0), and the y axis represents log-transformed P-values. (D) Venn diagram of overlap among 3,735 REF6 binding targets, 918 H3K27me3-hypermethylated genes, and 958 genes down-regulated in ref6-5 mutants versus Col-0 under RL. (E) Heatmap of transcriptome changes for REF6 binding targets, H3K27me3-hypermethylated genes, and genes down-regulated in ref6-5 mutants versus Col-0 under RL. Blue depicts downregulation, red depicts upregulation, and white depicts unchanged. (F) Genome browser views of REF6 ChIP-seq, H3K27me3 ChIP-seq, ATAC-seq, and RNA-seq signals at XTH22, DWF4, and DUF668 in Col-0 and ref6-5 under RL and DK. Black arrows represent the direction of transcription. The red vertical bar above the gene marks the position of the CTCTGYTY motif. (G) REF6 ChIP–qPCR (Top), H3K27me3 ChIP–qPCR (Middle), and RT–qPCR of XTH22, DWF4, and DUF668 relative expression (Bottom). ChIP–qPCR enrichment was normalized to an intergenic region not bound by REF6 and used as a negative control and transcript levels were normalized to TUB3. Different letters represent statistically significant differences between means determined by ANOVA with Tukey’s HSD test (P < 0.05).

H3K27me3 acts as a facultative repressive chromatin mark mostly distributed in the transcribed regions within euchromatin (63). Therefore, we wondered whether REF6-dependent H3K27 demethylation alters chromatin accessibility around the transcribed region in RL. We ran an assay for transposase-accessible chromatin sequencing (ATAC-seq) using wild-type Col-0 and ref6-5 mutant seedlings grown in DK or RL for 4 d (SI Appendix, Fig. S9 and Dataset S1). At gene-transcribed regions, the majority of the open-chromatin regions localized at 5′ transcription start sites in Col-0. Chromatin accessibility was however statistically significantly decreased around the 5′ and 3′ regions in ref6-5 in RL (Fig. 4B), indicating that loss of REF6 results in a global increase of H3K27me3 and decreased chromatin accessibility in RL. Overall, these results indicate that REF6 is required for chromatin accessibility at its target genes in RL. Chromatin dynamics are often accompanied by changes in gene expression (25, 43). To capture the transcriptional dynamics and identify the downstream factors that are regulated by REF6 during this process, we performed RNA sequencing (RNA-seq) using Col-0 and ref6-5 seedlings also grown under DK or RL for 4 d. The data with highly correlated RNA-seq replicates were uniquely aligned to version 10 (TAIR10) of the Arabidopsis reference genome (Fragments Per Kilobase Million, FPKM > 1, SI Appendix, Fig. S10 and Dataset S1). 958 genes were down-regulated (fold change ≥ 1.5) in ref6-5 versus Col-0 under RL (Fig. 4C and Dataset S4). Among these, 171 were bound by REF6 and are H3K27me3-hypermethylated genes (Fig. 4 D and E). This group comprised several growth-associated genes, including XYLOGLUCAN ENDOTRANSGLUCOSYLASE/HYDROLASE 22 (XTH22, encoding a cell-wall-modifying enzyme rapidly up-regulated in response to environmental stimuli) (64), DWARF 4 (DWF4; encoding a 22α hydroxylase catalyzing in a rate-limiting step in brassinosteroid biosynthesis), (65) and DUF668 (annotated as a positive growth regulator) (Fig. 4F). Validation of REF6 binding and H3K27me3 hypermethylation in ref6-5 at XTH22, DWF4 and DUF668 by ChIP and qPCR yielded results that were consistent with those obtained from ChIP-seq and RNA-seq analyses (Fig. 4G). Real-time qPCR (RT–qPCR) confirmed that XTH22, DWF4, and DUF668 induction in RL cannot be fully achieved to Col-0 levels in ref6-5 mutants (Fig. 4G). These results indicated that REF6 is responsible for genome-wide demethylation of H3K27me3, opening chromatin and activating light-responsive downstream targets.

Phytochrome B Facilitates the Association of REF6 with Chromatin.

The DNA-binding capacity of REF6 plays a critical role in light responses, and we are curious about whether and how its physical interaction with phyB influences this. To investigate this, we ran ChIP-seq using an anti-REF6 antibody on wild-type Col-0, ref6-5, and phyB-9 mutant plants grown under RL. Intensities of 525 peaks covering 566 genes were statistically significantly reduced in phyB-9 (fold change ≥ 1.5), suggesting phyB facilitates REF6 chromatin association under RL (Fig. 5 A–C and SI Appendix, Fig. S5 E and F). Among the identified candidates, we selected ROPGEF2 (acted as a negative regulator in phyB-mediated RL-induced stomatal opening) (66), MYB7 (acts as a repressor of flavonol biosynthesis and plays a role in the regulation of UV-sunscreen production) (67) and DUF668 as representative examples (Fig. 5C). We then validated these three genes using REF6 ChIP–qPCR and confirmed that REF6 enrichment was statistically significantly reduced in phyB-9 mutants (SI Appendix, Fig. S11). To identify the coregulated downstream factors by REF6 and phyB in this process, we carried out RNA-seq on Col-0, ref6-5, phyB-9, and ref6-5 phyB-9 seedlings grown under RL and DK (Dataset S1). We identified 4,857 DEGs (differentially expressed genes, fold change ≥ 1.5) in phyB-9 mutants compared to Col-0 (Fig. 5D and Dataset S4). Among these, 42 genes were positively regulated by phyB in their targeting of REF6, and their expression levels were notably decreased in the ref6-5 mutants (Fig. 5 D and E). These results suggest that phyB facilitates REF6 targeting and partly coregulates the expression of several REF6 target genes.

Fig. 5.

Phytochrome B facilitates the association of REF6 with chromatin and regulates DUF668 expression via REF6 and PIF4. (A) Metaplots and heatmaps of REF6 ChIP-seq reads over REF6 binding peaks. Wild-type Col-0 and phyB-9 mutants grown in 25 μmol m⁻² s⁻¹ of RL were used for ChIP-seq assays. The x axis represents the distance from the REF6 peak summit (in kilobases). (B) Ratio of REF6 binding-peak intensities in phyB-9 over Col-0 as in (A). Only peaks with the ratio phyB-9/Col-0 > 1.5 kb (n = 525) or <–1.5 kb (n = 18) were shown in the plot. (C) Genome-browser views of REF6 ChIP-seq signals at ROPGEF2, MYB7, and DUF668 in wild-type Col-0, phyB-9, and ref6-5 mutants under RL. Black arrows represent the direction of transcription. The red vertical bar above the gene indicates the position of the CTCTGYTY motif. (D) Venn diagram of overlap of REF6 binding targets reduced in phyB-9, DEGs in phyB-9, and down-regulated genes in ref6-5 mutants under RL. (E) Heatmap of the overlap of REF6-binding genes reduced in phyB-9, DEGs in phyB-9, and down-regulated genes in ref6-5 under RL. Top: Genes up-regulated in phyB-9. Bottom: Genes down-regulated in phyB-9. Blue depicts downregulation, yellow depicts upregulation, and white depicts unchanged. (F) RT–qPCR of relative DUF668 expression in wild-type Col-0, ref6-5, phyB-9, and ref6-5 phyB-9 mutants grown under DK and 25 μmol m⁻² s⁻¹ RL. Transcript levels were normalized to TUB3. Different letters represent statistically significant differences between means determined by ANOVA with Tukey’s HSD test (P < 0.05). (G and H) Phenotypes (G) and hypocotyl lengths (H) of wild-type Col-0, ref6-5, pif4-101, and ref6-5 pif4-101 mutant seedlings grown in RL (red; 100 μmol m⁻² s⁻¹) or DK for 4 d. (Scale bar, 1 mm.) Error bars represent SD from >12 seedlings. Different letters represent statistically significant differences between means determined by one-way ANOVA with Tukey’s HSD test (P < 0.05). (I) RT–qPCR of relative DUF668 expression in wild-type Col-0, ref6-5, pif4-101, and ref6-5 pif4-101 mutant seedlings grown in 25 μmol m⁻² s⁻¹ RL for 4 d. Transcript levels were normalized to TUB3. Different letters represent statistically significant differences between means determined by one-way ANOVA with Tukey’s HSD test (P < 0.05).

REF6, phyB, and PIF4 Coregulated DUF668 Gene Expression.

The DUF668 gene is a REF6 binding target, and its REF6-ChIP enrichment was reduced in phyB-9 mutants (Fig. 5C and SI Appendix, Fig. S11), indicating that DUF668 is coregulated by both phyB and REF6. Although DUF668 expression is reduced in ref6-5, it is strongly induced in phyB-9 mutant under RL (Fig. 5 E and F), hinting that other factors may also be involved in this process. REF6 cooperates with PIF4 to activate thermo-responsive genes under high ambient temperature (53). We asked whether REF6 and PIF4 synergistically regulate light-responsive hypocotyl elongation. In RL, ref6-5 pif4-101 double-mutant seedlings had much shorter hypocotyls than either ref6-5 or pif4-101 single mutants and Col-0 controls (Fig. 5 G and H), indicating that REF6 and PIF4 coregulate hypocotyl growth under RL conditions. DUF668 induction was impaired in ref6-5, pif4-101, and ref6-5 pif4-101 mutants in RL (Fig. 5I). This suggests that REF6 coregulates DUF668 expression with PIF4 and is essential for PIF-mediated gene activation under red-light conditions. Thus, REF6 acts together with the phyB-PIF4 module to mediate plant growth in response to changing light environment.

Discussion

Epigenetic regulation is highly dynamic, controlling gene expression spatially and temporally during plant development and environmental responses. Here, we showed that the H3K27me3 demethylase REF6, together with phyB and PIF4, regulate Arabidopsis photomorphogenesis under RL. Loss of function of REF6 resulted in an attenuated hypocotyl-elongation phenotype in RL (Fig. 1). REF6 promoted genome-wide H3K27 demethylation and chromatin accessibility on light-responsive genes. Further, the major red-light receptor phyB physically and genetically interacts with REF6 and enhances the accumulation and binding ability of REF6, thereby coordinating DUF668 expression in conjunction with REF6 and PIF4 (Figs. 2–5). Taken together, our study reveals how REF6 participates in the response to RL and provides insight into how the coordination of light sensors (phytochromes), epigenetic factors (REF6), and transcription factors (PIF4) fine-tune hypocotyl growth in response to environmental stimuli.

The dynamic balance of repressive H3K27me3 marks is regulated by Polycomb-group proteins, and the major H3K27me3 demethylases ELF6/JMJ11 (36), REF6 (37), JMJ13 (38), and JmjC-domain-only JMJ30 and JMJ32 proteins (39, 40). These demethylases remove H3K27me3 marks during plant development and environmental responses (43). Unlike other H3K27me3 demethylases, REF6 recognizes a specific DNA motif through its zinc finger domains (45, 47). REF6 is essential for the binding of the SWI/SNF ATPase BRAHMA at target loci enriched in stimulus-responsive genes, promoting chromatin accessibility (47). Our findings indicate that REF6 is necessary for H3K27 demethylation and chromatin opening in cell-elongation-related genes under RL. Both REF6’s DNA-binding and enzymatic activities are crucial for effective red-light responses. These results demonstrate that REF6 promotes chromatin accessibility mainly through its binding and enzymatic functions and then appears to facilitate binding of transcription factors. Further exploration is needed to identify chromatin-remodeling factors that collaborate with REF6 in light signaling. Additionally, REF6’s response to light is nonlinear, with distinct targets requiring various transcription factors. For example, XTH22 is targeted by BES1 and REF6, while DUF668 is coregulated by REF6 and PIF4, indicating that REF6 regulates different target genes in response to light alongside specific transcription factors.

The phyB elongated-hypocotyl phenotype is rescued to varying extents by mutations in pif3 (68), pif4, pif5 (19, 69), pifq (70), pif7 (71), and myb30 (22). Our genetic data indicated that phyB-9 ref6-5 hypocotyls were of intermediate length compared with ref6-5 and phyB mutants and were similar with those of wild-type Col-0 in light (Fig. 2 F and G). Interestingly, while PIFs, MYB30, and REF6 all act as negative regulators of photomorphogenesis, phyB inhibits PIFs but promotes MYB30 and REF6. As the main photoreceptor in RL, phyB promotes PIFs degradation to ensure photomorphogenesis since PIFs promote growth. Conversely, phyB increases MYB30 levels, which in turn promotes PIF4 and PIF5 reaccumulation under extended RL exposure (22). In this study, our results reveal that phyB also promotes the accumulation of the epigenetic factor REF6 and facilitates the association of REF6 with chromatin. It seems likely that phyB-induced REF6 and MYB30 accumulation may prevent an exaggerated inhibition of plant growth under prolonged light exposure. This suggested that phyB plays a dual regulatory role in the light, affecting different factors that regulate suitable growth of plants under the variable light conditions.

The Arabidopsis U-box proteins MOS4-ASSOCIATED COMPLEX 3A (MAC3A) and MAC3B physically interact with cryptochromes and compete for shared targets with HY5 to fine-tune light-responsive hypocotyl elongation (72). The MAC3A-binding motif largely overlaps with the REF6 binding motif (72), and ref6 mutants also showed photoinhibition of hypocotyl elongation in blue and far-red light (Fig. 1A), akin to mac3a mac3b double mutants. This indicated that MAC3A, MAC3B, and REF6 may collaboratively regulate chromatin modification and status in light responses. It would be beneficial to investigate whether and how REF6, MAC3A, and MAC3B cooperatively mediate light responses and photomorphogenesis under natural light conditions.

In conclusion, we have uncovered the molecular mechanism that REF6 responds to light cues, which sheds light on the collaboration among phytochromes, epigenetic factors, and transcription factors in regulating gene expression during photomorphogenesis. This regulatory mechanism enables seedlings to adapt to variable light conditions, thereby maintaining normal plant growth and development.

Materials and Methods

Plant Materials and Growth Conditions.

All plants used in this study were in the Arabidopsis thaliana Col-0 ecotype, unless otherwise indicated. The mutants ref6-1, ref6-3, ref6-5 (37, 47, 53), ref6^C, elf6-3, jmj13^G, ref6^C elf6-3 jmj13^G (41), jmj30 jmj32 (39), phyB-9 (62), and ref6-5 pif4-101 (53), along with transgenic lines 35S_pro: phyB–GFP (61) and 35S_pro: REF6-YFP (37) have been previously described. The ref6-5 phyB-9 mutant was generated by genetic crossing. Growth conditions and light sources followed previous descriptions (22), with fluence rates of the growth chambers detailed in the respective figure legends.

Plasmid Construction and Generation of Transgenic Arabidopsis Plants.

To generate the REF6_pro: REF6–GFP–GUS:3′ UTR and REF6_pro: REF6–GUS:3′ UTR vectors, a 593-bp 3′ UTR fragment following the REF6 stop codon was and subcloned into the pCAMBIA1300 with GFP-GUS or GUS tags. The REF6 promoter (1.2 kb upstream of its start codon) and genomic fragment were cloned into the pCAMBIA1300-GFP-GUS-3′ UTR or pCAMBIA1300-GUS-3′ UTR binary vector. Primers are listed in SI Appendix, Table S2. REF6_pro::REF6–GFP–GUS:3′ UTR was transformed into ref6-1 plants (REF6_pro: REF6: GUS 1#), while REF6_pro: REF6–GUS:3′ UTR was transformed into ref6-5 plants (REF6_pro: REF6: GUS 2# and 3#) via floral dip method (73). The detailed method was described previously (45).

For yeast two-hybrid assays, the LexA–PHYB-N and LexA–PHYB-C constructs were previously described (74). AD–REF6 variants (A-E) were created by cloning their coding sequences with primers in SI Appendix, Table S2, into the pB42AD vector. The PHYB–BD (PHYB–D153) construct for the GAL4 system was also described before (60, 74). The AD–REF6 construct, containing the REF6 genomic sequence, was amplified and cloned into the pGADT7 vector using specified primers.

Yeast Assays.

Yeast two-hybrid (LexA and GAL4 system) assays were performed following protocols described previously (22, 75). For GAL4-based assays, yeast cultures were grown in SD/-Trp-Leu-His medium containing 10 µM phycocyanobilin (PCB). Cultures were irradiated with 8 min of RL (25 μmol m⁻² s⁻¹), either alone or followed by 8 min of FRL (20 μmol m⁻² s⁻¹), and incubated for 2 h. After this, cultures received an 8-min pulse of red or red + far-red light and were incubated for another 1 h (22).

Nuclear Protein Extraction.

For nuclear-protein extracts, 0.5 g of 4-d-old seedlings were ground in liquid nitrogen and thawed in ChIP extraction buffer 1 (10 mM Tris-HCl pH 8.0, 0.4 M sucrose, 10 mM MgCl₂, and 1% Triton X-100) at 4 °C and then filtered through a 70-µm membrane. The flow-through was centrifuged at 4,000 g for 20 min at 4 °C. The nuclear pellet was washed three times with ChIP extraction buffer 2 (10 mM Tris-HCl pH 8.0, 0.25 M sucrose, 10 mM MgCl₂, and 1% Triton X-100) and centrifuged at 4,000 g for 10 min. The final pellet was resuspended in 2× SDS loading buffer, heated at 100 °C for 15 min, and used for immunoblotting.

Immunoblotting.

Immunoblotting followed a previous method (75) using anti-REF6 polyclonal antibodies (1:1,000) produced from a truncated REF6 (670 to 950 aa). Anti-HSC70 (1:2,000; cat. no. SPA-818; Assay Designs) was used as a negative control.

Co-IP Assays.

For the co-IP assay, 4-d-old wild-type Col-0 and homozygous 35S_pro: phyB–GFP seedlings grown in darkness were ground to a fine powder in liquid nitrogen and homogenized in extraction buffer (22). After light treatment with red/far-red light pulses for 30 min, samples were incubated with GFP-trap agarose beads at 4 °C for 3 h. Beads were washed three times with extraction buffer, and proteins were eluted in 2× SDS loading buffer by heating at 100 °C for 15 min. The eluted proteins were then analyzed by immunoblotting.

IP–MS.

REF6-YFP fusion protein was purified from 20-d-old seedlings overexpressing REF6-YFP (35S:REF6-YFP-HA, pEG101) using a GFP antibody (Roche, 11814460001) and protein G Dynabeads (Invitrogen, 10003D), with wild-type as control. Seedlings were ground in Honda buffer (25 mM Tris-HCl pH 7.4, 2.5% Ficoll 400, 5% Dextran T40, 0.4 M sucrose, 10 mM MgCl₂, 1 µM DTT, 0.5% Triton X-100, and 1 tablet/50 mL complete Protease Inhibitor Cocktail), filtered, and centrifuged at 1,000 g for 10 min at 4 °C. The pellet was resuspended in high-salt lysis buffer (25 mM HEPES pH 7.5, 0.5 M NaCl, 0.1% CA-630, and 1 tablet/50 mL complete Protease Inhibitor Cocktail) and rotated at 4 °C for 2 h. Postcentrifugation, the supernatant was diluted with low-salt buffer (25 mM HEPES pH 7.5, 0.1% CA-630, and 1 tablet/50 mL complete Protease Inhibitor Cocktail) and incubated with GFP antibody and Dynabeads overnight. Beads were washed four times with ice-cold wash buffer (PBS, 0.05% CA-630), eluted in 2× SDS loading buffer by heating at 100 °C for 15 min, separated by SDS-PAGE, stained with Coomassie Brilliant Blue R-250, divided based on molecular weight, digested with trypsin, and analyzed by mass spectrometry. The proteomic analysis was previously described (76).

GUS Staining.

Transgenic REF6pro:gREF6–GUS reporter lines were analyzed for GUS activity following the method described previously (77).

RNA Extraction, cDNA Synthesis, and RT–qPCR.

Total RNA from 4-d-old seedlings was extracted using TRIzol (Invitrogen, 15596026) and quantified. cDNA synthesis was performed with a FastKing RT Kit (TIANGEN). RT-qPCR was done in triplicate using ChamQ Universal SYBR qPCR Mastermix, normalized to TUBULIN 3 Primer details are in SI Appendix, Table S2.

Transcriptome Analysis by RNA sequencing.

Three sequencing libraries were independently constructed from separate samples and sequenced for each line and light condition. Trimmed, paired-end, 150-bp reads were generated using Trim Galore (version 0.6.6, https://github.com/FelixKrueger/TrimGalore) and aligned to the TAIR10 reference genome using HISAT2 (version 2.2.1, https://daehwankimlab.github.io/hisat2). Reads mapped to genes were counted with featureCounts (version 2.0.0) (78). Differential gene-expression analysis was performed using the DESeq2 (https://github.com/thelovelab/DESeq2) R package with parameters of greater than 1.5-fold change in expression and P-value < 0.05. Gene-ontology analysis was performed with the clusterProfiler R package (https://yulab-smu.top/biomedical-knowledge-mining-book). Heatmaps of expression data were generated by the ggplot2 R package (https://ggplot2.tidyverse.org) using Z-normalized FPKM.

Chromatin-Immunoprecipitation Assays and Library Preparation.

For ChIP-seq, antibodies against H3K27me3 (Cell Signal Technology, 9733S) and REF6 [custom mouse monoclonal (46)] were used at 5 µg and 0.5 µg per immunoprecipitation, respectively. ChIP was performed as previously described (46). Briefly, 2.5 g of 4-d-old seedlings grown under continuous RL or DK were ground in liquid nitrogen, cross-linked with formaldehyde in ChIP extraction buffer 1, and nuclei isolated through buffers 2 and 3. Pellets resuspended in high-salt buffer (20 mM Tris-HCl pH 8.0, 0.5 M NaCl, 0.1% SDS, 1% Triton X-100, and 2 mM EDTA) were sonicated, and DNA fragments (~0.5 kb) were obtained by centrifugation before incubation with antibody-bound Dynabeads overnight at 4 °C. The washing, elution, reverse cross-linking, and DNA purification steps were all performed as described previously (79). DNA was analyzed by qPCR or sequenced. Libraries for sequencing were prepared using the NEXTflex Rapid DNA-seq Prep Kit. NEGATIVE CONTROL 4 (NC4), an intergenic region not bound by REF6, was used as a negative control for fold enrichment normalization. Primer sequences are provided in SI Appendix, Table S2.

ChIP-seq Analysis.

Paired-end ChIP-seq reads were aligned to the TAIR10 reference genome using Bowtie2 (version 2.3.5.1) (80) in local-alignment mode. Unmapped reads, multimapped reads, and unpaired reads were filtered by Samtools (version 1.10, https://github.com/samtools/samtools), and PCR-duplicated reads were removed using Sambamba (version 0.7.1) (81). BigWig files for visualization were created by deepTools (version 3.3.2) (82) with a 10 bp bin size, viewed via Integrative Genomics Viewer (83). Peaks were identified by MACS2 (version 2.2.6) (84), and annotated to their nearest TSS using Bedtools and custom Perl scripts. MAnorm merged peak regions to identify differential H3K27me3 methylation and REF6 binding areas. R scripts generated scatterplots and Pearson correlation coefficients of normalized read counts. Binding intensity, H3K27me3 enrichment, and chromatin accessibility were quantified as RP10M and normalized for library size. Common peaks were merged using MAnorm (version 1.3.0) (85), and raw counts calculated with multiBamCov. DESeq2 identified differential peaks requiring >twofold change and P-value < 0.05.

ATAC-seq Library Preparation.

ATAC-seq was conducted with modifications as previously described (86). Briefly, 4-d-old seedlings (0.2 to 0.5 g) were macerated in lysis buffer (15 mM Tris-HCl pH 7.5, 20 mM NaCl, 80 mM KCl, 5 mM 2-mercaptoethanol, 0.5 mM spermine, protease inhibitors, and 0.2% Triton X-100), filtered, and nuclei collected by centrifugation at 600 g for 5 min at 4 °C. Nuclei were washed twice with wash buffer (10 mM Tris-HCl pH 8.0, 5 mM MgCl₂, and protease inhibitors) and counted after DAPI staining. Using 100,000 nuclei, ATAC-seq libraries were prepared with the TruePrep DNA Library Prep Kit V2 (Vazyme, TD501). DNA was fragmented with Tn5 at 37 °C for 30 min, purified using VATHS beads, and sequenced on the Illumina platform to generate 150-bp paired-end reads, including two biological replicates. All steps were performed under green light on ice or at 4 °C, except for cell counting.

ATAC-seq Analysis.

Paired-end sequencing reads were trimmed of adaptors, filtered of low-quality reads by Trim-Galore, and mapped to the TAIR10 reference genome. PCR-duplicated reads were removed by Sambamba. Open-chromatin peaks were determined using MACS2 with default parameters. Peaks identified from MACS2 were merged using MAnorm. For data visualization, BigWig coverage files were generated by using bamCoverage with 10-bp bin size and normalized with RP10M.

Accession Numbers.

Sequence data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: REF6 (AT3G48430), PHYB (AT2G18790), XTH22 (AT5G57650), DWF4 (AT3G50660), DUF668 (AT5G51670), ROPGEF2 (AT1G01700), MYB7 (AT2G16720), and TUB3 (AT5G62700).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Data, Materials, and Software Availability

The datasets from this study (REF6 ChIP-seq, H3K27me3 ChIP-seq, ATAC-seq and bulk RNA-seq) are available at the China National Center for Bioinformation [accession PRJCA024521 (87)]. All other data are included in the manuscript and/or supporting information.