OsABF1 Represses Gibberellin Biosynthesis to Regulate Plant Height and Seed Germination in Rice (Oryza sativa L.)

Gibberellins (GAs) are diterpenoid phytohormones regulating various aspects of plant growth and development, such as internode elongation and seed germination. Although the GA biosynthesis pathways have been identified, the transcriptional regulatory network of GA homeostasis still remains elusive. Here, we report the functional characterization of a GA-inducible OsABF1 in GA biosynthesis underpinning plant height and seed germination. Overexpression of OsABF1 produced a typical GA-deficient phenotype with semi-dwarf and retarded seed germination. Meanwhile, the phenotypes could be rescued by exogenous GA3, suggesting that OsABF1 is a key regulator of GA homeostasis. OsABF1 could directly suppress the transcription of green revolution gene SD1, thus reducing the endogenous GA level in rice. Moreover, OsABF1 interacts with and transcriptionally antagonizes to the polycomb repression complex component OsEMF2b, whose mutant showed as similar but more severe phenotype to OsABF1 overexpression lines. It is suggested that OsABF1 recruits RRC2-mediated H3K27me3 deposition on the SD1 promoter, thus epigenetically silencing SD1 to maintain the GA homeostasis for growth and seed germination. These findings shed new insight into the functions of OsABF1 and regulatory mechanism underlying GA homeostasis in rice.


Introduction
Increasing crop production has been an eternal theme of the world. Previous findings implied that gibberellins (GAs) are major regulators for crop growth and yield through optimizing the harvest index and lodging resistance [1]. In addition, the dormancy and germination process of plant seeds are closely related to GA content, which is a key development stage in crop life cycle and an important factor determining grain yield and quality [2]. As a class of diterpenoids hormones, GAs participate in nearly all important plant growth and developmental events, such as stem and hypocotyl elongation, leaf expansion, seed dormancy and flowering [3,4]. According to the number of carbon atoms in the GA molecule, GAs may be divided into two types: Carbon-20 GAs and Carbon-19 GAs [5]. Thus far, GA biosynthesis pathways have been well documented and have dwarf phenotype and with retarded seed germination, could be used in rice breeding with lodging resistance and pre-harvest sprouting resistance.

OsABF1 Overexpression Lines Exhibited GA-Deficient Phenotypes
Two representative overexpression lines, OxABF1-6 and OxABF1-9, with over 300 folds upregulation were used for phenotypical characterization ( Figure S1a). Compared with wild-type (WT), the OxABF1-6 and OxABF1-9 lines showed obvious dwarfism, which was a typical GA-deficient phenotype (Figure 1a). By comparing the length of the internodes of OxABF1-6, OxABF1-9, and WT, we found that all the internodes of OsABF1 overexpression lines were shortened by 20% to 30%, which finally made the height of OsABF1 overexpression lines only 73% of that of the WT (Figure 1b,c). To figure out the reason for the semi-dwarf phenotype of OxABF1-6 plants, we employed paraffin sections to observe the cell size in the second internodes of the OxABF1-6 and the wild-type (Figure 1d,g). The cell length and width of OxABF1-6 was significantly decreased as compared with the WT (Figure 1i,j). Moreover, the number of large vascular bundles (LVB) and small vascular bundles (SVB) were also reduced in OxABF1-6 (Figure 1e,f), indicating that the dwarfism of OxABF1-6 could be attributed to the reduced cell size and proliferation ( Figure 1h). Additionally, we performed seed germination assays on OsABF1 overexpression seeds and WT. OsABF1 overexpression lines showed obviously lower germination rate and post-germination growth than the WT (Figure 2a). More importantly, exogenous GA 3 fully rescued the OxABF1-6 and OxABF1-9 germination rates to the WT level, and also partially recovered its post-germination growth (Figure 2c,d,f,g). Meanwhile, application of paclobutrazol (PAC, an endogenous GA biosynthesis inhibitor) significantly inhibited the seed germination rate and post-germination growth of WT seeds to the OxABF1-6 level but showed no significant effects on those of the OxABF1-6 ( Figure 2b,f,g). Furthermore, hormone quantification assay confirmed that the GA 3 in 15 days seedlings of OxABF1-6 reduced over 50% compared with the WT (Figure 2e). Meanwhile, the other endogenous GAs were also quantified, and the results showed that the level of GA 5 , GA 8 , GA 13 , GA 20 , GA 24 , and GA 34 were significantly decreased in OxABF1-6 ( Figure S1b). Thus, we concluded that the observed retarded seed germination and seedling growth of OxABF1 lines were due to GA deficiency.
Employing the CRISPR/Cas9 technique, we successfully knocked out OsABF1 and simultaneously knocked out OsABF1 and its closest homolog bZIP40 (LOC_Os05g36160). The obtained mutants harbored different types of insertions or deletions in the coding sequence of the targeted genes, some of which shifted the open reading frames and reduced the transcriptional levels ( Figure S1c,d). However, both abf1 and abf1 bzip40 mutants showed no substantial difference with the WT in plant height and seed germination, suggesting that OsABF1 might be functionally redundant with bZIP40 and other bZIP sibling genes (Figures 1c,d and S2).  Employing the CRISPR/Cas9 technique, we successfully knocked out OsABF1 and simultaneously knocked out OsABF1 and its closest homolog bZIP40 (LOC_Os05g36160). The obtained mutants harbored different types of insertions or deletions in the coding sequence of the targeted genes, some of which shifted the open reading frames and reduced the transcriptional levels ( Figure S1c,d). However, both abf1 and abf1 bzip40 mutants showed no substantial difference with the WT in plant height and seed germination, suggesting that OsABF1 might be functionally redundant with bZIP40 and other bZIP sibling genes (Figures 1c,d and S2).
As a crucial gene in GA biosynthesis and the green revolution in rice breeding, SD1 attracted our particular attention due to its downregulation pattern and the existence of a conserved bZIP binding G-box cis element in its promoter [45] (Figure 3c). In order to check whether SD1 is a direct target of OsABF1, we performed yeast one-hybrid (Y1H) experiment. In the yeast, OsABF1 successfully activated the expression of proSD1::LacZ reporter, thus OsABF1 may directly bind to the SD1 promoter ( Figure 3e). Moreover, we conducted electrophoresis mobility shift assay (EMSA) to test the binding in vitro. The GST-OsABF1 protein could physically bind to the SD1a probe containing the G-box, and the excess unlabeled competitive probes effectively alleviated the biotin probe bound by GST-ABF1 ( Figure 3d). ChIP-qPCR result further demonstrated that OsABF1 was significantly enriched in the regions P3 of SD1 promoter where the G-box cis element was located ( Figure 3f). Finally, we performed luciferase (LUC) transient transcriptional activity assay to detect the effect of OsABF1 on SD1 transcription in rice protoplasts. The result showed that the luciferase activity of the proSD1::LUC reporter was significantly repressed when OsABF1 was introduced. The result indicated that OsABF1 acted as a transcription repressor on SD1, which consistent with the reduced SD1 expression in OsABF1 overexpression lines showed OsABF1 directly binds to the promoter of SD1. The 1×, 10×, 25×, 50×, and 100× fold excess non-labeled probes were used for competition. The red letter represented G-box. (e) Yeast one-hybrid assay showed that OsABF1 could bind to the SD1 promoter (1500 bps upstream of the SD1 transcription start site); UDT1-pB42AD and RMS2-pLacZi were used as CK + . (f) ChIP-qPCR analysis of the OsABF1 enrichment on SD1 promoter in OxABF1-6. (g,h) Luciferase (LUC) transient transcriptional activity assay in rice protoplast. Error bars represent means ± SD (n = 3). ** p < 0.01 by Student's t-test analysis.
As a crucial gene in GA biosynthesis and the green revolution in rice breeding, SD1 attracted our particular attention due to its downregulation pattern and the existence of a conserved bZIP binding G-box cis element in its promoter [45] (Figure 3c). In order to check whether SD1 is a direct target of OsABF1, we performed yeast one-hybrid (Y1H) experiment. In the yeast, OsABF1 successfully activated the expression of proSD1::LacZ reporter, thus OsABF1 may directly bind to the SD1 promoter ( Figure 3e). Moreover, we conducted electrophoresis mobility shift assay (EMSA) to test the binding in vitro. The . Bar 1 to 6 indicate the regions checked by ChIP-qPCR. The red asterisk (SD1a) represents the position of the G-box motif. (d) EMSA assay showed OsABF1 directly binds to the promoter of SD1. The 1×, 10×, 25×, 50×, and 100× fold excess non-labeled probes were used for competition. The red letter represented G-box. (e) Yeast one-hybrid assay showed that OsABF1 could bind to the SD1 promoter (1500 bps upstream of the SD1 transcription start site); UDT1-pB42AD and RMS2-pLacZi were used as CK + . (f) ChIP-qPCR analysis of the OsABF1 enrichment on SD1 promoter in OxABF1-6. (g,h) Luciferase (LUC) transient transcriptional activity assay in rice protoplast. Error bars represent means ± SD (n = 3). * p < 0.05, ** p < 0.01 by Student's t-test analysis.

OsABF1 Bound to OsEMF2b, Causing the Deposition of H3K27me3 on the SD1 Promoter
Polycomb group (PcG) proteins have been shown to regulate growth and development in plants via mediating H3K27me3-suppressing histone modification on the target genes, which has been recognized as a significant epigenetic mechanism in gene regulation [46,47]. The PcG proteins in rice include two E(z)-like homologues (OsCLF and OsiEZ1), two Su(z)12 homologues (OsEMF2a and OsEMF2b), and two ESC homologues (OsFIE1 and OsFIE2), among which OsEMF2b has been reported to control plant height and seed germination by regulating the expression of OsVP1 [48,49]. As mentioned above, the overexpression of OsABF1 reduced the plant height and seed germination; therefore, we speculated that OsABF1 may interact with OsEMF2b to recruit PcG-mediated repression on SD1. The hypothesis was tested by yeast two-hybrid assays, which showed that OsABF1-BD physically binds to OsEMF2b-AD in yeast (Figure 4a). In the GST pull-down assay, the purified His-OsEMF2b was pulled down with GST-OsABF1, but not with GST tag, indicating that OsABF1 interacts with OsEMF2b in vitro (Figure 4c). The BiFC experiment in plant also revealed that the nYFP-OsEMF2b and OsABF1-cYFP pair produced strong fluorescent signals in the nucleus of tobacco leaf cells, whereas no fluorescence signals were observed when nYFP-OsEMF2b or OsABF1-cYFP was expressed alone (Figure 4b).
Therefore, these results strongly demonstrated that OsABF1 binds to OsEMF2b. Finally, we performed ChIP-qPCR to test the deposition patterns of OsEMF2b and its mediated histone modification H3K27me3 on the SD1 region. The results showed that both OsEMF2b and H3K27me3 were significantly enriched in the SD1 region, particularly the P3 region, which is largely overlapped with the OsABF1 deposition patterns (Figure 4d,e). Hence, OsABF1 represses the SD1 transcription via recruiting the PcG-mediated H3K27me3.

OsABF1 Bound to OsEMF2b, Causing the Deposition of H3K27me3 on the SD1 Promoter
Polycomb group (PcG) proteins have been shown to regulate growth and development in plants via mediating H3K27me3-suppressing histone modification on the target genes, which has been recognized as a significant epigenetic mechanism in gene regulation [46,47]. The PcG proteins in rice include two E(z)-like homologues (OsCLF and OsiEZ1), two Su(z)12 homologues (OsEMF2a and OsEMF2b), and two ESC homologues (OsFIE1 and OsFIE2), among which OsEMF2b has been reported to control plant height and seed germination by regulating the expression of OsVP1 [48,49]. As mentioned above, the overexpression of OsABF1 reduced the plant height and seed germination; therefore, we speculated that OsABF1 may interact with OsEMF2b to recruit PcG-mediated repression on SD1. The hypothesis was tested by yeast two-hybrid assays, which showed that OsABF1-BD physically binds to OsEMF2b-AD in yeast (Figure 4a). In the GST pull-down assay, the purified His-OsEMF2b was pulled down with GST-OsABF1, but not with GST tag, indicating that OsABF1 interacts with OsEMF2b in vitro (Figure 4c). The BiFC experiment in plant also revealed that the nYFP-OsEMF2b and OsABF1-cYFP pair produced strong fluorescent signals in the nucleus of tobacco leaf cells, whereas no fluorescence signals were observed when nYFP-OsEMF2b or OsABF1-cYFP was expressed alone (Figure 4b). Therefore, these results strongly demonstrated that OsABF1 binds to OsEMF2b. Finally, we performed ChIP-qPCR to test the deposition patterns of OsEMF2b and its mediated histone modification H3K27me3 on the SD1 region. The results showed that both OsEMF2b and H3K27me3 were significantly enriched in the SD1 region, particularly the P3 region, which is largely overlapped with the OsABF1 deposition patterns (Figure 4d,e). Hence, OsABF1 represses the SD1 transcription via recruiting the PcG-mediated H3K27me3.

OsEMF2b and OsABF1 Had Antagonistic Roles in GA Biosynthesis Regulation
In contrast to the GA-induced expression pattern of OsABF1, the transcription of OsEMF2b was significantly downregulated by exogenous GA in young seedlings at 1 HAT, suggesting the antagonistic roles of OsABF1 and OsEMF2b in GA biosynthesis regulation (Figure 3a). We adopted an OsEMF2b T-DNA insertional mutant from another lab for the characterization of its biological function [47]. The emf2b contains a T-DNA insertion in the fourth exon, which largely disrupted the expression of the gene ( Figure S1e). The homozygous emf2b mutants displayed a similar but more severe dwarf phenotype than OsABF1 overexpression lines (Figure 5a-c). Unfortunately, we were not able to test the seed germination of emf2b mutants due to the complete sterility of the homozygous emf2b lines ( Figure S1f). In emf2b, OsABF1 was significantly elevated (Figure 5d). Meanwhile, OsEMF2b expression was drastically reduced in OsABF1 overexpression lines (Figure 5e). We also performed qRT-PCR to examine the expression of GA-related genes in emf2b and the WT and found that all the tested genes showed the same differential expression pattern in emf2b as in OsABF1 overexpression lines (Figure 5f). In particular, GA20ox2/SD1 expression in emf2b was decreased to only 1% of that in the WT, which well explained the extremely dwarf nature of emf2b.
In contrast to the GA-induced expression pattern of OsABF1, the transcription of OsEMF2b was significantly downregulated by exogenous GA in young seedlings at 1 HAT, suggesting the antagonistic roles of OsABF1 and OsEMF2b in GA biosynthesis regulation (Figure 3a). We adopted an OsEMF2b T-DNA insertional mutant from another lab for the characterization of its biological function [47]. The emf2b contains a T-DNA insertion in the fourth exon, which largely disrupted the expression of the gene ( Figure  S1e). The homozygous emf2b mutants displayed a similar but more severe dwarf phenotype than OsABF1 overexpression lines (Figure 5a-c). Unfortunately, we were not able to test the seed germination of emf2b mutants due to the complete sterility of the homozygous emf2b lines ( Figure S1f). In emf2b, OsABF1 was significantly elevated (Figure 5d). Meanwhile, OsEMF2b expression was drastically reduced in OsABF1 overexpression lines (Figure 5e). We also performed qRT-PCR to examine the expression of GA-related genes in emf2b and the WT and found that all the tested genes showed the same differential expression pattern in emf2b as in OsABF1 overexpression lines (Figure 5f). In particular, GA20ox2/SD1 expression in emf2b was decreased to only 1% of that in the WT, which well explained the extremely dwarf nature of emf2b .

OsABF1 Targets SD1 to Regulate GA Biosynthesis
The growth and development of plants are regulated by many factors in vivo and vitro, including light, temperature, water, and genetic factors. The plant hormones are of the most importance in regulating the growth and development, for which the biosynthetic pathways are adjusted by many transcription factors. In this study, we revealed that a bZIP transcription factor OsABF1 regulates GA biosynthesis via targeting the green revolution gene SD1. Although the knock-out mutants of OsABF1 and the double mutants abf1bzip40 grew as normal as the WT (Figure S2), the overexpression of OsABF1 gave rise to typical GAdeficient phenotypes such as semi-dwarf and later seed germination (Figures 1a-c and 2a). Meanwhile, overexpression of OsABF1 exhibited significantly lower GA 3 concentration than the WT (Figure 2e). More importantly, the application of exogenous GA 3 successfully restored the phenotype of OsABF1 overexpression lines to the WT level, suggesting that GA deficiency is the major reason for the observed phenotypes (Figure 2c,d). Actually, OsABF1 has also been reported as a suppressor of rice flowering [42]. Given that GA is a major hormone promoting plant flowering, the later flowering phenotype should be attributed to the GA deficiency in OsABF1 overexpression lines as well. Moreover, Y1H, EMSA, and LUC transient transactivation and ChIP-qPCR assays confirmed that OsABF1 could directly bind to the G-box cis elements in the promoter regions of SD1 to repress its transcription, which strongly suggested that SD1 is a key node of the OsABF1-governed GA homeostasis (Figure 3c-h).
SD1 has been identified as an oxidase enzyme in GA biosynthesis. Mutations in the SD1 led to the "green revolution" in rice breeding by reducing the plant height with stronger lodging resistance and higher harvest index [12]. Meanwhile, the qSD1-2, another allele of SD1, was found involved in the reduction of GA synthesis, seed dormancy enhancement, and plant height reduction [8]. In breeding practice, double-stranded RNAi on two SD1 fragments influenced the plant height in different extents [50]. Gene editing on SD1 enabled the rapid improvement of landraces with shorter plant height and enhanced lodging resistance without sacrificing many other excellent agronomic traits such as low phosphorus tolerance and broad-spectrum resistance to a variety of diseases and insect pests [51].

OsABF1 May Mediate the Antagonistic Relationship of ABA and GA
Like indispensable lubrication in plants lifecycle, various plant hormones regulate plant growth and development synergistically or antagonistically, their relationship woven into a series of complex networks in plants [32,[52][53][54]. Abscisic acid (ABA) and GA work antagonistically in regulating various plant growth and development events such as seed dormancy and germination, root growth, plant height, flowering initiation, and abiotic stress [11,55,56]. It was found that exogenous ABA usually reduces GA accumulation, while ABA-deficient or ABA-signaling mutants were able to bypass the GA requirement for seed germination [57]. Thus far, a large number of mediators of ABA-GA antagonism have been identified. In Arabidopsis, ABF4 significantly increased the level of ABA and promoted the transcriptional regulation of GA metabolism genes, which suggested that ABF4 was a hub of ABA and GA signaling pathway [58]. It was revealed that SnRK2s-APC/CTE regulatory module mediated the antagonism between GA and ABA signal transduction in higher plants [59]. A very recent study from our lab also reported that rice WRKY72 inhibits GA accumulation and seed germination via the "WRKY72-LRK1-OsKO2" pathway [60]. Given that WRKY72 is a phosphorylation substrate protein of the ABA signaling core component SAPK10 [61], WRKY72 may serve as a key node in the ABA-GA interaction. Actually, several pieces of evidence have shown that OsABF1 is functionally implicated in ABA signaling. Transcription of OsABF1 is significantly induced by ABA and is able to bind to the ABA-responsive cis elements [44]. Its mutants became more sensitive to drought and salinity treatments, which are closely related to the functions of ABA. Additionally, its closest Arabidopsis orthologs, OsABF1, 2, 3, and 4, are also ABA-inducible and functionally involved in abiotic stress responses [62]. Combining the roles of OsABF1 in ABA signaling and GA biosynthesis suppression, it is rational to propose that OsABF1 is a key player in mediating the antagonistic relationship of ABA and GA.

OsABF1 Interacts with OsEMF2b-PRC2 Complex to Suppress SD1 Transcription through H3K27me3
Higher eukaryotes have evolved profound epigenetic regulatory mechanisms to finetune gene transcription. Initially identified in Drosophila, PcG proteins have been reported to inactivate gene expression and maintain the silencing state of the target chromatin through covalent histone modifications [63,64]. PcG proteins contain polycomb repressive complex 1 (PRC1) and PRC2. PRC1 complexes are E3 ubiquitin ligases depositing mono ubiquitylation of lysine 119 mark on histone 2A (H2AK119ub), while PRC2 majorly catalyzes tri-methylation on histone H3 at lysine 27 (H3K27me3) to repress target transcription [48,49,64]. PRC2 is not able to bind to the target inherent DNA, due to lack of DNA binding domains. Thus, its localization to target sites largely relies on other transcription factors that can link PRC2 and the DNA motifs [65]. Recently, recruitment of PRC2 by transcription factors has been demonstrated as a significant mechanism in gene transcriptional regulation. For example, in 2016, two independent labs reported that VAL1 and VAL2 bound to the RY motifs in the FLC promoter, wherein they recruit PRC2-mediated H3K27me3 for FLC repression and flowering promotion after vernalization [66,67]. Similarly, OsLF recruits PRC2 to the CANNTG (bHLH cis-element) motif of Hd1 by interacting with PRC2 component OsVIL2 and represses the Hd1 expression via H3K27 tri-methylation modification [68]. In the current study, we demonstrated that OsABF1 recruits EMF2b-mediated H3K27me3 on the promoter of SD1, thus epigenetically regulating GA biosynthesis. This conclusion has been supported by several lines of evidence including (1) OsABF1 being a direct repressor of SD1 (Figure 3c (Figures 3f and 4d,e). Together with other reports mentioned above, our study suggests that plant growth and development may subject to a very general regulatory mechanism that transcription factors repress target gene transcription via recruiting PRC2-mediated H3K27me3 modification.
In addition, there might be a regulation mechanism that overexpression of OsABF1 could repress the expression of SD1. In the meantime, enrichment of OsEMF2b would mediate more H3K27me3 in the promoter of SD1, and also repress its expression. Further, in plants, there must be a regulation mechanism of maintaining the GA homeostasis, namely, the moderate and stable expression of SD1. Thus, when OsABF1 was overexpressed, then the expression of OsEMF2b was reduced (Figure 5e), whereas when OsEMF2b mutant to emf2b, then the expression of OsABF1 was significantly increased to repress the expression of SD1 (Figure 5d) and then maintain the moderate and stable expression and the GA homeostasis in plant.

Plant Materials and Growth Conditions
The complete coding sequence of OsABF1 was amplified using the cDNA of Nipponbare (Oryza sativa ssp. japonica) leaf and inserted into the fusion vector pCAM-BIA1390 driven by the ubiquitin promoter with restriction digestion site of KpnI and BamHI. To produce OsABF1 mutant lines, we used a reported CRISPR/Cas9 technique to generate abf1 and abf1bzip40 double mutants [69]. All the binary vectors were transformed into the WT (japonica cv. Nipponbare) by Agrobacterium tumefaciens-mediated method.
The emf2b (A T-DNA insertion mutant of OsEMF2b) lines in japonica cv. Dongjin were adopted from other labs [50]. To measure plant height and test seed germination rate, we cultivated all plants during the normal rice-growing season in the experimental field of China National Rice Research Institute, Hangzhou (119 • 57 E, 30 • 05 N). The primers and sgRNA sequences are provided in Table S1.

Seed Germination and Seedling Growth Assays
The experiment of seed germination was conducted on the basis of the previous study with minor revision [32]. Briefly, all of the dehulled seeds were sterilized in 70% ethanol for 2 min, then in 50% NaClO for 40 min, and washed in sterilized water until clean. Then, the seeds were placed on 1/2 strength Murashige and Skoog medium (pH 5.8) and incubated in a growth cabinet (28 ± 2 • C, 12/12 h photoperiod and 70% relative humidity) for 7 days. As the seed coleoptile reached 5 mm in length, this phenotype was defined as complete germination. Seed germination rate was recorded every 12 h, and three biological replications (each replicate including 50 seeds) were used for each sample. Seedling height was measured after 7 days of germination. For the hormone treatment, different concentrations of GA 3 and PAC were added into 1/2 strength Murashige and Skoog medium treatment.

Hormone Quantification
For GAs quantification in Nipponbare and OxABF1-6 plants, 3 g of 15-days old fresh seedlings of Nipponbare and OxABF1-6 were harvested and then frozen in liquid nitrogen. Quantification of endogenous GAs levels was measured by liquid chromatography-tandem mass spectrometry system. This method was conducted as described in previous studies, with a few modifications [70,71]. All samples were extracted with 1.5 mL methanol with concentration of 80% at 4 • C for 16 h, then centrifuged at 13,000× g at 4 • C for 10 min. Further, the supernatant was collected and evaporated to dryness under a nitrogen gas stream. Finally, it was reconstituted in 100 mL of 95% acetonitrile for the liquid chromatography-tandem mass spectrometry analysis.

Paraffin Sections
Internodes of WT and OxABF1-6 collected at post-heading stage were soaked in FAA (70% ethanol 90 mL, 5 mL acetic acid, and 37-40% formaldehyde (5 mL)) solution for about 24 h, then dehydrated through gradient ethanol concentrations followed by different graded chloroform concentrations. Finally, all the treated samples were soaked in wax, embedded and sliced according to the previous study [72]. All tissues were stained with Safranine O-Fast Green Stain Kit (YESEN Biotech, Shanghai, China) and sliced into 10-mm sections using a microtome (Leica, Wetzlar, Germany). Images were photographed using a Leica DM2500 microscope (Leica, Wetzlar, Germany). The cell size was measured by ImageJ software (ImageJ V.1.8.0).

Quantitative Reverse Transcription (qRT)-PCR
RNA isolation kit (YESEN Biotech, Shanghai, China) was used to extract total RNA. cDNAs were synthesized by Prime Script RT Master Mix (Takara, Dalian, China) and quantitative real-time PCR of various genes was conducted by CFX 96 real-time PCR instrument (Bio-Rad, California, USA). Each sample was performed with three technical repetitions, and UBQ (LOC_Os03g13170) was used as internal control. The transcription levels were calculated by the 2 − CT method. Primers used in this study were listed in Table S1.

Yeast One-Hybrid Assay
Yeast one-hybrid system (Clontech, CA, USA) was used to check the interaction between OsABF1 and the SD1 promoter. The complete coding sequence of OsABF1 was fused to the pB42AD vector with EcoRI and XhoI sites. The promoter of SD1 (1500 bp upstream of the transcriptional start site) was amplified and inserted into the pLacZi vector using EcoRI and KpnI sites. All fused vectors were introduced into EGY48 (yeast strain), then grown on SD/-Ura/-Trp plates.

Electrophoresis Mobility Shift Assay (EMSA)
The probes of sd1 were synthesized and labeled with Cy5.5 by SUNYA biotech (Hangzhou, China). Non-labeled probes were used as competitors. For the Cy5.5-labeled probe, all the process of the EMSA was performed under darkness. Odyssey CLx infrared fluorescence imaging system (LI-COR, Lincoln, NE, USA) was used to detect the fluorescence signal of the EMSA gel. Probes used were listed in Table S1.

Luciferase Transient Transcriptional Activity Assay
For the luciferase activity assay, 1500bp upstream of the SD1 transcription start site was amplified and inserted into 190LUC vector with HindIII and BglII to generate the reporter. The CDS of OsABF1 was fused into "None" with BamHI and EcoRI as an effector. Both fusion vector plasmids were co-transformed into rice protoplast, and the luciferase activity was analyzed by the Luciferase Reporter Assay System (Promega, Madison, WI, USA). Rice protoplast extraction and transformation were conducted according to a predecessor study [73]. Primers used were listed in Table S1.

ChIP-qPCR
The ChIP-qPCR assay was conducted following the previously published protocol [48]. OsABF1 and OsEMF2b antibodies were commercially synthesized by Gene script company (Shanghai, China), and H3K27me3 antibody was purchased (Cat No: Ab6002. Abcam, Cambridge, MA, USA). Chromatin was extracted from 3 g of 15 DAG (days after germination) OsABF1 overexpression seedlings and then cross-linked with 1% formaldehyde under vacuum. Then, the extracted chromatin was fragmented to about 300 bps in size by sonication. Finally, DNA/protein complex was immuno-precipitated with polyclonal OsABF1, OsEMF2b, and H3K27me3 antibodies. Triple biological replicates of each sample were performed, and the data analysis was conducted according to previously reported methods [46,49]. Primers used were listed in Table S1.

Yeast Two-Hybrid Analysis and Bimolecular Fluorescence Complementation
The yeast two-hybrid assay was conducted on the basis of the manufacturer's protocol (Clontech, Palo Alto, CA, USA). The full-length CDS of OsABF1 and OsEMF2b were amplified and separately fused into bait vector pGBKT7 and prey vector pGADT7. pGBKT7-OsABF1 and pGADT7-OsEMF2b were co-transformed into yeast strain Y2H Gold. pGBKT7-53 and pGADT7-T were used as positive controls, pGBKT7 and pGADT7 were used as negative controls. For the bimolecular fluorescence complementation (BiFC) experiment, OsABF1 CDS without stop codon was fused into pDOE vector at the BspeI site, and the complete coding sequence of OsEMF2b was inserted into pDOE vector at the BamHI site. Then, the fusional expression vectors were transformed into Agrobacterium strain EHA105 by electroporation.

In Vitro Pull-Down Assay
For the vitro pull-down assay, the full CDS of the OsABF1 and OsEMF2b were amplified and ligated into the pGEX-4T-1 and pET28a vectors, respectively. The fused constructs and empty pGEX-4T-1 were introduced into E. coli strain Transetta (Transgen, Beijing, China). The primers were provided in Table S1.