DELLA-mediated gene repression is maintained by chromatin modification in rice

The rice DELLA protein SLR1 interacts with PRC2 proteins

To identify in vivo DELLA interacting proteins, we produced rice plants expressing FLAG-tagged SLR1. The plants showed a GA-deficient semi-dwarf phenotype (Appendix Fig S1A). Analysis of nuclear proteins isolated from the SLR1-FLAG plants by anti-FLAG immunoprecipitation coupled with mass spectrometry (IP-MS) detected several PRC2 core components (EMF2b, FIE2, and MSI1, FVE/MSI4) and PRC2-associated proteins VAL1 (VP1/ABI3-LIKE1) and VIL3 (Vernalization Insensitive 3-like protein) (Wang et al, 2013), in addition to several other chromatin proteins (Fig 1A). The GA-signaling protein GID2 that targets SLR1 for ubiquitination (Gomi et al, 2004) was also detected by IP-MS. Immunoblot analysis with an EMF2b antibody (Tan et al, 2022) detected the presence of EMF2b in the anti-FLAG IP products (Fig 1B). Immunoblotting analysis with a SLR1 antibody produced in this study (Appendix Fig S1B) revealed the presence of the endogenous SLR1 protein in the anti-GFP IP products of plants expressing FIE2-GFP (Appendix Fig S1C, Fig 1C). SLR1 interactions with EMF2b and FIE2 were confirmed by in vitro pull-down assays using E. coli-produced SLR1-His, FIE2-MBP and EMF2b-GST fusion proteins (Fig 1D). In Y2H and luciferase complementation assays, SLR1 could interact with the PRC2 core components SDG711 or SDG718, FIE2, MSI1, EMF2b, but not with EMF2a (Fig 1E and F), indicating that the rice DELLA protein specifically associated with the EMF2b-PRC2 complex. Deletion analysis indicated that the SDG711 C5 domain (Tan et al, 2022) and the EMF2b C-terminal part interacted with SLR1 (Appendix Fig S2A and B) and that the SLR1 GRAS domain was involved in the interaction with SDG711 (Appendix Fig S2C), suggesting that other GRAS family proteins could potentially interact with PRC2.

SLR1 is required to maintain high levels of H3K27me3 at a set of genomic loci

To study whether SLR1-mediated gene repression involved PRC2-mediated H3K27me3 deposition, we analyzed levels of H3K27me3 along with several other histone methylation marks (H3K4me3, H3K9me2, and H3K36me3) in 12-day-old seedlings by immunoblots. In the slr1 mutant (Ikeda et al, 2001), we detected a clear decrease of H3K27me3 (Fig 2A). The other tested histone marks were unchanged (Fig 2A). Next, we performed H3K27me3 ChIP-seq analysis of the slr1 mutant and its wild-type (Nipponbare) seedlings (Table EV1). In total, we detected 12,684 (10,301) peaks (genes) marked by H3K27me3 in the genome (Fig 2B), consistent with previous results in rice (Lu et al, 2020; Wang et al, 2021; Tan et al, 2022). In slr1 relative to wild-type plants, 3,615 peaks (3,435 genes) showed decreased (FC > 1.5, P < 0.05), while only 73 peaks (57 genes) showed increased (FC > 1.5, P < 0.05) H3K27me3 (Fig 2B), confirming a net loss of the mark in the mutant (Fig 2A). Thus, the slr1 mutation resulted in H3K27me3 decreases in about 28.5% (3,615/12,684) of the total marked peaks. Metaplots of the peaks indicated that H3K27me3 was clearly decreased in the genic region in the mutant (Fig 2C). Heatmap analysis revealed that the 3,435 genes with decreased H3K27me3 in slr1 plants were marked by much higher levels of H3K27me3 but lower levels of the active histone marks (H3K9ac, H3K27ac, H3K4me3, and H3K36me3) than the 3,000 randomly selected genes in the wild-type background (Fig 2D). These genes were particularly enriched for transcriptional functions (Fig 2E), and transcription factor genes (such as Dof, ERF, WOX, bHLH, and NAC) were over represented (Appendix Fig S3), suggesting that SLR1 was required to establish high levels of H3K27me3 preferentially in transcription factor genes in the rice genome.

SLR1-repressed genes display high H3K27me3

To study whether SLR1-dependent H3K27me3 was involved in SLR1-mediated gene repression, we compared the transcriptomes of slr1 and WT seedlings by RNA-seq. Three biological replicates were analyzed (Appendix Fig S4A). Compared with the wild-type 2,155 genes were downregulated and 2,014 genes were upregulated (FC > 2, FDR < 0.05) in the mutant (Fig 3A). The upregulated genes were enriched for auxin-signaling and growth promoting function (consistent with the role of SLR1 in auxin-signaling, the agonistic functions between GA and auxin and excessive stem elongation phenotype of the mutant), while the downregulated genes were enriched for photosynthesis (Appendix Fig S4B and C). Several GA biosynthetic genes (CPS, GA20ox) were downregulated whereas GA catabolic genes (e.g., GA2ox) were upregulated (Fig 3B), consistent with the constitutive GA-signaling in the slr1 mutants (Ikeda et al, 2001), and DELLAs functions required for expression of genes involved in GA biosynthesis (Zentella et al, 2007; Locascio et al, 2013). The slr1 upregulated genes displayed higher H3K27me3 (about twice of the controls) and lower H3K9ac levels than the genome-wide averages or the downregulated genes (Fig 3C), eliciting the hypothesis that the slr1 mutation mostly de-repressed genes marked by H3K27me3 in the wild-type background. Scattering plots detected, however, no general correlation between the changes of H3K27me3 and transcript levels, suggesting additional chromatin modification or mechanism was involved. Still, there were 397 upregulated genes that showed decreased H3K27me3 in the mutant (Fig 3D, Table EV2). The 397 genes were marked by much higher H3K27me3 levels (about 4 times) than the average levels of 400 randomly selected genes (Fig 3E) and most of them were GA-inducible genes according to the RiceXPro website (https://ricexpro.dna.affrc.go.jp/) (Sato et al, 2013) (Table EV2).

Histone deacetylase HDA702 interacts with both SLR1 and PRC2 to maintain low H3K9ac levels at target genes

Immunoprecipitation coupled with mass spectrometry data suggested that SLR1 might interact with additional chromatin factors including WDR5a, IBM2, and HDA702 (Fig 1A). WDR5 is involved in deposition of H3K4me3 (Jiang et al, 2018), however, the overall H3K4me3 level seemed unchanged in slr1 plants (Fig 2A). The IBM2 is a BAH domain-containing protein, which was recently shown to recognize H3K27me3 (Zhang et al, 2023). HDA702 is orthologous to the Arabidopsis HDA19 (Hu et al, 2009), a primary plant HDAC with pleiotropic functions (Zheng et al, 2023). Their interaction with SLR1 was confirmed by Co-immunoprecipitation (Co-IP) of transiently transfected protoplasts and/or luciferase complementation assays in tobacco leaves (Appendix Fig S5A and B). The SLR1–HDA702 interaction was further confirmed by in vitro pull-down and in vivo Co-IP assays in rice plants (Fig 4A and B). Tests with several HDACs revealed that only HDA702 interacted with SLR1 in Y2H and luciferase complementation assays (Fig 4C and D). Immunoblots revealed higher acetylation levels of H3 and H3K9 (but not H3K27) in slr1 than wild-type plants (Fig 4E). Analysis of HDA702 RNAi and overexpression (OE) plants indicated that HDA702 had an activity to deacetylate H3K9 in the genome (Fig 4F). The data suggested that SLR1 might recruit HDA702 to remove H3K9ac from its genomic targets. Previous data showed that Arabidopsis HDA19 (orthologous to the rice HDA702) could interact with the MSI1 subunit of PRC2 (Questa et al, 2016; Ning et al, 2019). Tests with Co-IP and luciferase complementation assays confirmed that the HDA702 did interact with the rice MSI1 (Appendix Fig S5C and D). These results suggested that SLR1, PRC2, and HDA702 could form a tripartite complex to control histone modifications for gene repression.

SLR1-mediated gene expression involves recruitment of both PRC2 and HDA702

To study the function of PRC2 and HDA702 in SLR1-mediated gene repression, we selected 8 of the 397 upregulated genes that showed decreased H3K27me3 in the slr1 (Table EV2), to test their expression and histone modifications in SLR1, PRC2 (SDG711 and EMF2b) (Tan et al, 2022), and HDA702 (Hou et al, 2022) genes RNAi, mutant, or overexpression (OE) plants. The eight genes or family members have been reported to be induced by GA (Appendix Fig S6A), including KCS (Xiao et al, 2016), PLA1 (Mimura et al, 2012), EXPB8 (Lee et al, 2006), GA2OX7 (Chen et al, 2016), OSH15 (Su et al, 2021), PME25 (Ogawa et al, 2003), SPL7 (Jung et al, 2012), and TB1 (Lo et al, 2008). The tests revealed that all of the genes were upregulated to about 2–4 folds in the RNAi or mutant plants of SLR1, SDG711, EMF2b, and HDA702 but were repressed in the HDA702 OE lines (Appendix Fig S6B). ChIP-PCR analysis with H3K27me3 antibody confirmed the decreases of the histone mark in the analyzed genes in the slr1 mutant (Fig 5A and B). ChIP-PCR analysis with SLR1 and EMF2b antibodies detected that both proteins physically associated with the genes and that the slr1 mutation reduced the levels of EMF2b-binding (Fig 5C and D). The H3K9ac levels in the genes increased in slr1 and HDA702 RNAi plants, but decreased in the HDA702-HA OE lines (Fig 5E). Anti-HA ChIP-PCR analysis of HDA702-HA OE plants (with HDA702 RNAi plants as control) indicated that the genes were also directly targeted by HDA702 (Fig 5F). Together, these results indicated that SLR1 recruited both HDA702 and PRC2 to target genes to deacetylate H3K9 and maintain high H3K27me3 levels.

To test whether SLR1 was involved in regulation of other PRC2 targets, we tested MULTIPASS (MPS), an R2R3-type MYB transcription factor gene (Schmidt et al, 2013), which is modified by H3K27me3 (Appendix Fig S6C). GA treatment or the slr1 mutation did not affect H3K27me3 level in the MPS locus (Appendix Fig S6C). ChIP-PCR indicated that SLR1 could not bind to MPS (Fig 5C). EMF2b could bind to MPS, but the slr1 mutation had no effect on the EMF2b-binding (Fig 5D). The analysis confirmed that SLR1 regulates only a subset of the PRC2 targets.

To study whether SLR1-PRC2 were recruited by DNA-binding transcription factors to downstream targets, we analyzed by bioinformatics (http://plantregmap.gao-lab.org/) (Tian et al, 2020), promoters of the 397 upregulated genes that showed decreased H3K27me3 in slr1 plants, and found that several transcription factor family (ZnF, MYB, NAC, WRKY, and bZIP) members potentially binding to these genes (Appendix Fig S7A). Among the ZnF factors, YABBY4 (YAB4) was previously shown to interact with SLR1 to repress GA-responsive genes in rice (Yang et al, 2016). YAB4-binding sites could be identified in several tested SLR1 downstream genes (Appendix Fig S7B). We obtained yab4 mutant and YAB4-Flag overexpression plants (Appendix Fig S1D and E). Analysis of the YAB4-Flag plants with anti-Flag ChIP-PCR indicated that the two tested SLR1 downstream genes (PLA1 and EXPB8) were bound by YAB4 (Appendix Fig S7C). ChIP-PCR analysis with SLR1 and EMF2b antibodies showed that the yab4 mutation reduced the levels of SLR1 and EMF2b binding to the targets (Appendix Fig S7D). These results suggest that the YAB4 is involved in recruiting SLR1-PRC2 to the downstream genes.

GA signaling increases H3K9ac and decreases H3K27me3

To study whether the slr1 mutation effects on histone modifications could be mimicked by GA-signaling, we tested the H3K9ac and H3K27me3 levels in wild-type plants treated with 100 μM GA and harvested the plant at different time points (Appendix Fig S8A and B). We found that the overall H3K9ac level started to increase 1 h after the GA treatment and H3K27me3 to decline about 3 h after the treatment (Fig 6A). To confirm the GA-induced changes in histone modifications, we tested H3K27me3 and H3K9ac levels in gid1 (GA-insensitive dwarf1) (Ueguchi-Tanaka et al, 2005) and slr1-d5 (a gain-of-function mutation of SLR1) (Zhang et al, 2016) plants treated with or without GA. Without GA treatment, there was no clear change in the H3K27me3 and H3K9ac levels in the mutants compared to wild-type plants (Appendix Fig S8C). GA treatment of gid1 plants had no clear effect on H3K9ac or H3K27me3, confirming that GA-induced changes of the histones modifications depended on the GA perception and signaling. GA treatment induced about 20% increase of H3K9ac and decrease of H3K27me3 in slr1-d5 plants 6 h after the treatment (Appendix Fig S7D), which appeared less strong and/or later than in wild-type plants (Fig 6A), consistent with the delayed response of the slr1-d5 plants to gibberellin treatment (Zhang et al, 2016). The results confirmed that GA perception and signaling lead to alteration of the histone modifications. Conversely, testing plant response (second leaf sheath growth) of PRC2 mutants (i.e., EMF2b RNAi plants; Tan et al, 2022) to GA at different concentrations (0, 10⁻⁸, 10⁻⁷, 10⁻⁶, or 10⁻⁵ M), revealed that compared to wild-type, EMF2b RNAi plants were hypersensitive to GA at 10⁻⁶ and 10⁻⁵ M (Appendix Fig S8E), supporting that PRC2 represses GA signaling.

To further confirm the results, we monitored the expression and histone modification levels in slr1-derepressed genes in wild-type plants treated with GA. The expression of the genes was induced 1 h after GA treatment (Fig 6B). The genes displayed increased H3K9ac and decreased H3K27me3 (Fig 6C). As observed in the immunoblots, H3K9ac was found to increase 1 h after the GA treatment, while H3K27me3 was detected to clearly decrease 3–6 h after the treatment (Fig 6C), suggesting that H3K9ac was likely deposited prior to the removal of H3K27me3 during the GA-induced gene activation.

GA signaling dissociates PRC2 and HDA702 from target genes

To study whether GA-induced SLR1 degradation affected the abundance of PRC2 and HDA702 proteins, we treated wild-type plants with GA at different time points and tested the protein levels of SLR1, EMF2b, and HDA702 by immunoblotting with their antibodies. The analysis indicated that SLR1 almost disappeared 30 min after the treatment, while the levels of PRC2 (EMF2b) and HDA702 proteins were unaffected 3 h after the treatment (Fig 7A and B). In addition, the EMF2b and HDA702 protein level was unaffected in the slr1 plants (Fig 7A and B). Likely, GA signaling did not affect the stability of the chromatin regulators. To study whether GA signaling-mediated SLR1 degradation dissociated PRC2 (EMF2b) and HDA702 from the target genes, using ChIP-PCR we analyzed wild-type and HDA702-HA OE plants treated with GA and found that the treatment reduced the binding levels of both EMF2b and HDA702 in the tested genes (Fig 7C), corroborating the observation that SLR1 was required for EMF2b binding to target genes (Fig 5D). The data indicated that GA-induced SLR1 destruction dissociated the chromatin regulators from the target genes, leading to H3K9ac deposition and subsequently H3K27me3 removal from the targets.

Because GA induced a rapid increase of H3K9ac that associates tightly with gene activation, we analyzed the transcriptomes of rice seedlings treated with or without GA for 1 h to better capture genes that are directly induced by GA signaling. Three replicates were sequenced (Table EV3). Principal component analysis (PCA) and correlation confidents indicated the good reproducibility of the replicates (Appendix Fig S9A). The GA treatment led to upregulation (FC > 2, P < 0.05) of 245 genes and downregulation of 207 genes (Appendix Fig S9B) (Dataset EV1). The upregulated genes were enriched mainly in the gibberellin oxidation and DNA replication pathways (Appendix Fig S9C). The upregulated genes showed high levels of H3K27me3 and low levels of the active histone marks (Appendix Fig S9D), consistent with chromatin signature of SLR1-repressed genes. Among the 245 genes, 62 (> 25%) were unregulated in the slr1 seedlings (Table EV4).

To study the GA-induced genome-wide histone modification changes, we performed H3K27me3 ChIP-seq analysis of rice seedlings treated with or without GA for 6 h. Two replicates were tested (Appendix Fig S10A, Table EV1). H3K27me3 ChIP-seq analysis revealed 3,163 (2,920) hypo peaks (genes) but only 26 (21) hyper (P < 0.05) peaks (genes) in GA-treated versus untreated plants (Appendix Fig S10B), consistent with the immunoblots (Fig 6A). The genome-wide decreases of H3K27me3 in GA-treated plants showed a similar trend (R = 0.31) as in the slr1 mutant, with more than 50% (1,618/2,920) of the genes with decreased H3K27me3 in GA-treated plants overlapped with those in the slr1 mutant (Appendix Fig S10C). The overall methylation levels of the H3K27me3 peaks in GA-treated plants were lower than untreated plants but higher than the slr1 plants (Appendix Fig S10D). The larger decreases of H3K27me3 in slr1 might be due to the developmental defects of the mutant plants. Among the 245 upregulated genes in 1 h GA-treated plants 58 were found to have a significant decrease of H3K27me3 after 6 h GA treatment (Table EV5), which included the 5 analyzed marks genes (Fig 7D). The analysis supported the notion that H3K27me3 deposition/removal was involved in SLR1-mediated gene repression and GA-induced gene activation.

Chromatin mechanism of DELLA-mediated gene repression

DELLA proteins can repress GA-responsive genes by sequestering transcriptional activators or forming repression complexes with other proteins (Phokas & Coates, 2021). In this work, we provided evidence that DELLA protein SLR1 has a function to establish or maintain silent chromatin signatures to repress gene expression by recruiting PRC2 and HDA702 and other PRC2-associated proteins. This is consistent with previous results that Arabidopsis DELLAs interact with the chromatin remodeling factor PICKLE that promotes H3K27me3 (Zhang et al, 2014). The observation that SLR1 was required to establish high levels of H3K27me3 preferentially in transcription factor genes supported the notion that DELLAs are master transcriptional repressors. SLR1 may target PRC2 and HDA702, potentially other PRC2-associated or H3K27me3-binding proteins (VAL1, VIL3, and IBM2), to downstream targets by interacting with specific DNA-binding transcription factors, as previous data showed that DELLA proteins could interact with a large number of transcription factors belonging to different families (La Rosa et al, 2014; Lantzouni et al, 2020). The observation that the slr1 mutation or GA signaling partially reduced PRC2 association with GA-responsive genes indicated that, in addition to interacting with SLR1, PRC2 may also be recruited to SLR1 targets by interacting with DNA-binding transcription factors, which was suggested to be primarily involved in PRC2 recruitment in plants (Godwin & Farrona, 2022). Thus, SLR1 might directly or indirectly stabilize the PRC2 association with the targets for maintaining gene repression. This is reminiscent of the finding that the nitrogen-dependent recruitment of PRC2 is facilitated by the rice transcription factor NGR5 to repress expression of shoot branching-inhibitory genes, thus promoting tillering in response to increasing nitrogen supply (Wu et al, 2020). DELLA interacts with NGR5 leading to the stabilization of NGR5, thus indirectly of PRC2, by reducing gibberellin-promoted proteasomal destruction of NGR5. However, the present data rather establish a direct link between DELLA and PRC2 for repression of GA-induced genes in rice, which may represent a general chromatin mechanism of DELLA-mediated gene repression.

Chromatin mechanism of GA signaling-induced gene activation

Gibberellin signaling induces rapid degradation of DELLA proteins to de-repress downstream genes. The present data reveal that the chromatin change caused by the rapid destruction of SLR1 is a mechanism of GA signaling-induced gene activation in rice. Although the SLR1 destruction resulted in dissociation of both PRC2 and HDA702 from the downstream genes, H3K9ac increased prior to H3K27me3 decline during the gene activation process. This suggests that the HDA702 dissociation allows rapid acetylation by HATs that might readily associate with the nearby nucleosomes at the loci. This is consistent with the notion that histone acetylation enzymes are part of the gene transcription machinery and that histone acetylation change is a first chromatin event occurred during gene transcriptional activation (Kouzarides, 2007). However, the RNAi of the rice GCN5 gene (encoding a H3K9 acetyltransferase) (Zhou et al, 2017) seemed not affecting the activation course of the tested genes induced by GA signaling (Appendix Fig S11A), suggesting either the GCN5 expression is not sufficiently reduced in the plants or other HATs might be involved in the process. That the rapid destruction of SLR1 or gene activation induced by GA signaling did not lead to immediate removal of H3K27me3 indicates that the presence of H3K27me3 in the locus is insufficient to inhibit gene transcription. Likely, additional chromatin modification (e.g., histone deacetylation) and/or presence of PRC2 and PRC2-associated proteins (such as VIL3, VAL1, and IBM2) might be required to maintain the repression state of SLR1 targets. This corroborates previous and present observations that changes of H3K27me3 do not correlate with the overall changes of gene expression (Fig 3D) (Arthur et al, 2014; Wang et al, 2021; Tan et al, 2022). That the decline of H3K27me3 occurred 3–6 h after the GA treatment suggests the removal of H3K27me3 is required to promote or stabilize GA-induced gene activation course. The GA signaling-triggered H3K27me3 decreases seemed not due to dilution by cell division, as 6 h after the treatment the plant height was unchanged (Appendix Fig S7), suggesting that an active demethylation process is involved. However, the mutation of the JMJ705 gene (encoding an H3K27me3 demethylase) (Li et al, 2013) did not alter the course of gene induction by GA signaling (Appendix Fig S11B).

In summary, the results established that HDA702-dependent histone deacetylation and PRC2-mediated H3K2me3 are the chromatin basis of DELLA-mediated gene repression and that GA-induced gene activation involves first increases of H3K9ac and subsequently decreases of H3K27me3, as a consequence of destruction of DELLA and dissociation of PRC2 and HDA702 from the target genes (Fig 8).

Plant materials and growth conditions

Rice (Oryza sativa ssp japonica) Zhonghua11 (ZH11) variety was used to produce tagged SLR1 and FIE2 overexpression plants. The slr1 mutant, the SDG711 RNAi, and the HDA702 RNAi/OE-HDA702 plants were described previously (Ikeda et al, 2001; Hou et al, 2022; Tan et al, 2022). The slr1 mutant and HDA702 RNAi as well as the OE-HDA702 were in the Nipponbare (Oryza sativa ssp japonica) background. Two independent knockout lines for YABBY4 (yabby4-1 and yabby4-2) were generated in the ZH11 accession using the CRISPR–Cas9 system described in (Gao & Zhao, 2014). The OE-YABBY4 line was also produced in the ZH11 background: the YABBY4 coding sequence (CDS) was cloned in-frame and upstream of that of Flag and driven by the maize Ubiquitin promoter in the vector pCAMBIA1301. For germination, seeds of wild-type and mutant plants were sterilized with 0.15% HgCl₂. After washing with sterilized distilled water, the seeds were germinated in 1/2 MS medium with a cycle of 14 h light (28°C)/10 h dark (24°C). Seedling ages were calculated as the days after germination and were used for phenotypic examination. To grow in the field, rice seedlings were transplanted in the Wuhan area in early May, and the seeds were harvested in September. In chemical treatment experiments, 100 μM GA (Sigma, G1025) was supplied to the culture media of seedlings.

Polyclonal antibody preparation

The recombinant SLR1 (amino acids [aa] 1–133) protein was used as an antigen to produce polyclonal antibodies in rabbits. Recombinant antigenic protein was expressed and purified from E. coli DE3 cells (Yeasen Biotech, 11804ES80). After three cycles of injection, the antibody was affinity purified with antigen-coupled magnetic beads and mixed with 20% glycerol and stored at −80°C.

Affinity purification and mass spectrometry

To identify SLR1 interaction proteins, we performed immunoaffinity purification as previously described by (Tan et al, 2022) with minor modifications. Briefly, OE-SLR1 (3xFLAG tagged SLR1) and wild-type (Nipponbare, as a negative control) plants were harvested and ground in liquid nitrogen and quickly placed in 30 ml pre-cooled IP buffer (50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 10% glycerol, 0.1% NP-40, 0.5 mM DTT, 1 mM PMSF, 1× protease inhibitor cocktail from Roche, Basel, Switzerland). Samples were further homogenized to be free of clumps and centrifuged at 5,000 g and 4°C for 15 min. The lysate was filtered through two layers of Miracloth (Millipore, 475855) and the supernatant was incubated with 200 μl of anti-FLAG M2 magnetic beads (Sigma, M8823) for 3 h at 4°C. Protein-bound beads were washed twice with 10 ml IP buffer, then six times with 1.5 ml IP buffer (5 min spin at 4°C) and finally with IP buffer without NP-40. The beads were sent to PTM-BIOLABS (Hangzhou, China) for mass spectrometry analysis. Two independent replicates were performed.

Co-immunoprecipitation

Co-immunoprecipitation assays were conducted as previously described with minor modifications (Wang et al, 2021). In brief, total proteins were extracted from 12-day-old seedlings or isolated rice protoplasts expressing respective tagged proteins and immunoprecipitated with anti-FLAG affinity gel (Bimake, B26102), anti-GFP affinity gel (AlpaLife, KTSM1334), or anti-HA affinity gel (Pierce, 88836), followed by immunoblotting using anti-SLR1 (1:1,000 dilution), anti-EMF2b (Tan et al, 2022) (1:1,000 dilution), anti-FLAG (Sigma, F1804; 1:3,000 dilution), anti-HA (Sigma, H3663; 1:3,000 dilution), or anti-GFP (Abcam, ab290; 1:3,000 dilution).

Y2H assays

The cDNAs of the investigated genes were obtained by PCR. SDG711, SDG718, EMF2a, EMF2b, FIE2, HDA702, HDA703, HDA704, HDA710, and SRT1 cDNAs were cloned into the Bam HI site of pGBK-T7, and SLR1 cDNAs were cloned into the Bam HI site of pGADT7 (Clontech). The truncated primers for SDG711, EMF2b were described in (Tan et al, 2022). The truncated primers for SLR1 are shown in the Table EV7. Y2H assays were performed using the Matchmaker GAL4 two-hybrid system (Clontech) according to the manufacturer's instructions. Synthetic media free of tryptophan, leucine, histidine, and adenine (–WLHA) are used for yeast growth.

In vitro pull-down assay

Recombinant SLR1-SUMO-His was incubated with FIE2-MBP in MBP beads or HDA702-GST and EMF2b-4-GST in GST beads and pulled down with MBP beads (Smart-Lifesciences, SA026100) or GST beads (GE Healthcare, 45-000-139). The pulled down proteins were subsequently analyzed by immunoblotting. SLR1-His was detected with anti-His antibody (Abcam, Ab9108; 1:5,000 dilution).

Split-luciferase complementation assays

SDG711, SDG718, EMF2a, EMF2b, FIE2, MSI1, VAL1, WDR5a, IBM2, RP1, HDA702, HDA703, HDA704, HDA710, and SRT1 were separately fused with cLUC in the pCAMBIA-cLUC vector. SLR1 and MSI1 were fused with nLUC in the pCAMBIA-nLUC vector. Agrobacterium-mediated transformation with the indicated nLUC and cLUC constructs were introduced into Nicotiana benthamiana leaves. After 72 h of incubation under normal growth conditions, leaves were injected with 0.15 mg/ml of D-fluorescein potassium salt as a substrate and placed in a dark room for about 10 min and harvested for the determination of the luminescence signal by a CCD camera (TANAON 5200, Shanghai).

Chromatin immunoprecipitation (ChIP)

The ChIP experiment was performed as described (Cheng et al, 2018). Approximately 3 g of 12-day-old seedlings were cross-linked with 1% formaldehyde under vacuum. To investigate changes in H3K27me3 at the genome-wide level in rice after GA treatment, 12-day-old wild-type seedlings were treated with 100 μM GA for 1 and 6 h, respectively, followed by 3 g of GA-treated seedlings cross-linked with 1% formaldehyde under vacuum. Chromatin was extracted and fragmented to 200–500 bp by sonication, and ChIP was performed using the following antibodies: anti-H3K9ac (Sigma, 07-352), anti-H3K27me3 (Diagenode, C15410195), anti-SLR1, anti-EMF2b, anti-HA (Sigma, H3663), anti-IgG (Abcam, ab37415).

ChIP-seq and data analysis

Approximately 3 g of seedlings were used per sample and different antibodies were used for the ChIP assay as described above. The precipitated and input DNA were used to construct sequencing libraries and sequenced by Illumina NovaSeq 6000. TrimGalore (version 0.6.6) was used to trim the adaptors and remove low quality reads from the raw reads. Clean reads were mapped to the rice genome (MSU7.0, http://rice.plantbiology.msu.edu/) by Bowtie2 (version 2.4.4). SAMtools (version 1.9) was used to convert between SAM and BAM formats and to sort the BAM file. Duplicate reads were removed with picard tools (version 2.1.1-Java-1.8.0_92) using default parameters. MACS2 (version 2.2.7.1) was used to call histone modification peaks with the parameters (-g 4.01e8 --broad --broad-cutoff 0.1), and input sample was used as control. Differential peaks were identified with DiffBind (version 3.8.4) using DESeq2 method to test the differential peaks. P-value < 0.05 and FoldChange > 1.5 were used to define differential peaks. The bigwig files were generated using bamCoverage with the parameters “-bam -binSize 10 –normalize Using RPKM” in deepTools (version 3.5.0), which were visualized in the Integrated Genome Browser.

ChIP-qPCR

After extensive washing and de-crosslinking of the DNA obtained by the ChIP method described above, the precipitated and input DNA samples were analyzed by qPCR. Primers for ChIP-qPCR are listed in Table EV7. Three biological replicates were performed using 12-day-old seedlings harvested from three separate cultures. Each biological replicate was tested in three technical repeats.

RNA-seq and analysis

To identify GA early responsive genes, 12-day-old wild-type seedlings treated with 100 μM GA or without treatment for 1 h and slr1 mutant seedlings were performed for RNA-seq sequencing. High quality total RNA was isolated with Trizol reagent (Invitrogen). RNA-seq libraries were prepared using the Illumina TruSeq RNA Sample Preparation Kit and sequenced (paired-end, 2 × 150 bp) on an Illumina HiSeq 2000 instrument. RNA-seq data were filtered by TrimGalore (version 0.6.6) to remove low-quality reads. The clean data were mapped against the rice reference genome (MSU 7.0, http://rice.plantbiology.msu.edu/) with the corresponding annotation by Hisat2 (version 2.2.1) using default parameters. FeatureCounts (version 2.0.0) and DESeq2 (version 1.38.3) were used to calculate DEGs with an adjusted P < 0.05 and Foldchange > 2. GO enrichment analysis of DEGs was performed on the ShinyGO website (version 0.77, http://bioinformatics.sdstate.edu/go/).

qRT-PCR

Approximately 1 μg of total RNA was reverse-transcribed in a reaction volume of 20 μl using DNase and reverse transcriptase (Vazyme, Nanjing, China; R233-01) according to the manufacturer's instructions to obtain cDNA. RT-qPCR was performed as follows: 95°C for 10 s, 45 cycles of 95°C for 5 s and 60°C for 40 s. Three independent biological replicates were carried out, with three technical replicates of three identical samples from one experiment.

Tests of GA induction of shoot elongation

To investigate plant response to GA, the second leaf sheath lengths of WT and EMF2b RNAi plants (n = 30) were measured at day 7 after germination. The seeds were germinated on 1% agar plates supplemented with GA3 at different concentrations (0, 10⁻⁸, 10⁻⁷, 10⁻⁶, or 10⁻⁵ M). The plants were grown at 28°C (12 h light/12 h dark).

Statistical analysis

Statistical analysis was carried out by using GraphPad Prism V9 (https://www.graphpad.com/scientific-software/prism/). The significance of the differences between the two groups was analyzed by means of unpaired and two-tailed t-tests for students and Welch's correction. For multiple comparisons, a one-way ANOVA (nonparametric) test with Tukey's multiple comparisons was performed where standard deviations were equal. One-way ANOVA (nonparametric) with Games–Howell's multiple comparison test was performed with unequal variances (different numbers).

Accession numbers

The sequence data from this article can be found in the MSU database (http://rice.plantbiology.msu.edu/) under the following gene locus identifiers: SLR1 (LOC_Os03g49990), FIE2 (LOC_Os08g04270), SDG711 (LOC_Os06g16390), SDG718 (LOC_Os03g19480), EMF2b (LOC_Os09g13630), EMF2a (LOC_Os04g08034), VAL1 (LOC_Os07g37610), WDR5a (LOC_Os03g51550), IBM2 (LOC_Os01g42460), RP1 (LOC_Os05g07700), MSI1 (LOC_Os03g43890), HDA702 (LOC_Os06g38470), HDA703 (LOC_Os02g12350), HDA704 (LOC_Os07g06980), HDA710 (LOC_Os02g12380), SRT1 (LOC_Os04g20270), KCS (LOC_Os02g56860), PLA1 (LOC_Os10g26340), EXPB8 (LOC_Os03g01260), GA2OX7 (LOC_Os01g11150), OSH15 (LOC_Os07g03770), PME25 (LOC_Os08g34900), SPL7 (LOC_Os04g46580), TB1 (LOC_Os03g49880), PIL1 (LOC_Os03g56950), XTH8 (LOC_Os08g13920), YABBY4 (LOC_Os02g42950), MPS (LOC_Os02g40530).