Japonica cultivars Zhonghua 11 (ZH11) or Dongjin (DJ) was used as the wild type (WT) for the transgenic analyses. Plants were grown on soil in field or in greenhouse or on 0.5x Murashige and Skoog (1/2MS) medium in a conditioned growth chamber at 30°C for 10 h (day) and 24°C for 14 h (night). For BR induction experiments, young seedlings were transferred to 1/2MS medium supplemented with various concentrations of brassinolide (BL, one of the active BRs) for 48 h. For lamina inclination assay, ethanol (1 μL) containing various amount of BL was spotted onto the top of lamina after 2-day germination and 3-day growth at 30°C (Tong and Chu, 2017). Images were taken after 3-day incubation, and the angles of lamina joint bending were measured using IMAGEJ software¹. For bikinin treatment, 1-week-old seedlings were grown in 1/2MS supplemented with or without 30 μM chemical for 3 days before sampling.

The DNA sequence containing the promoter and coding sequence of OFP1 was cloned into pCAMBIA2300 vector for OFP1 overexpression (OFP1-OE or 1o). A pCAMBIA2300-35S-GFP vector was used for OFP1-GFP construction. A pCAMBIA1300-35S-Flag vector was used for OFP1-Flag overexpression (OFP1-Flag-OE or 1Fo). OFP1 promoter (2-kb) was introduced into pCAMBIA2391Z vector for OFP1p:GUS construction. CRISPR/Cas9 knock-out vectors were constructed following previous reports (Ma and Liu, 2016). Primer sequences used for vector constructions and additional plasmid information are listed in the Supplementary Table 1. Sequences were introduced into vectors by either in-fusion cloning strategy (Clontech) or traditional cut-ligation cloning method. These constructs were used to transform rice plant or to infiltrate tobacco leaf epidermis cells by Agrobacterium-mediated method (Sparkes et al., 2006).

Yeast two-hybrid tests were performed following standard procedures descried by manufacturer (Clontech). BiFC (bimolecular fluorescence complementation) analyses were performed following the method described previously (Waadt et al., 2008). Rice protoplast preparation and plasmid transfection were performed according to the previous method (Bart et al., 2006). Fluorescence was observed on a confocal fluorescence microscope (Leica TCS SP6). Split-luciferase complementation assays were performed in tobacco leaves as described (Chen et al., 2008). Chemiluminiscence was photographed using a imaging system equipped with a cold CCD (NightOWL II LB983 with indigo software). See Supplementary Table 1 for the information of the primer sequences and plasmids used for the vector constructions prepared for the above analyses.

RNA was isolated using Trizol reagent (Invitrogen) according to manufacturer’s instruction. The first-strand cDNA was synthesized using a Revertase Transcription kit (Toyoba). Quantitative RT-PCR (qRT-PCR) was performed on a real-time PCR detection system following manufacturer’s instructions (Bio-Rad CFX96). Rice Ubiquitin2 gene (UBQ) was used as internal control for all analyses. Primer sequences are listed in the Supplementary Table 2. GUS staining was performed according to previous description (Tong et al., 2009).

Commercialized anti-Flag (Sigma) and anti-ACTIN (Abmart) antibodies were used for immunoblotting analyses. Commercialized anti-OsBZR1 antibody (BPI) was used for ChIP (chromatin immunoprecipitation) analysis. ChIP-qPCR (ChIP followed by quantitative PCR analysis) was performed according to previous description (Tong et al., 2014). Primer sequences are listed in the Supplementary Table 2.

About 4 g of shoots was harvested from the rice seedlings for GA measurements. Quantification of endogenous GAs was performed as described (Chen et al., 2012).

We previously showed that GSK2 kinase, the central negative regulator of BR signaling, targets DLT, the GRAS family transcription factor, to regulate downstream BR responses (Tong et al., 2012). To further explore the functional mechanism of the proteins, we performed extensive yeast two-hybrid screenings of the interacting proteins using DLT as bait. By this approach, we identified OFP1 (Ovate Family Protein 1) as one of DLT-interacting proteins from the yeast library prepared using mixed rice tissues. Full length OFP1 prey vector (OFP1-AD) was further constructed and the interaction was confirmed in yeast (Figure 1A). In addition, we found that when OFP1 was used as both bait and prey, the interaction also occurred (Figure 1A), suggesting that OFP1 may form homo-dimer or oligomer to function.

OFP1 contains both the putative NLS (nuclear localization signal) and NES (nuclear export signal) (Figure 1B). To analyze the protein subcellular localization, we fused the protein with GFP and introduced the corresponding vector into rice protoplasts. Consistently, microscopic observation revealed that OFP1-GFP fusion proteins were localized in both cytoplasm and nucleus, and can be co-localized with DLT-RFP distributed in nucleus (Figure 1C), supporting the possibility that they may interact in plant cells.

We then further verified their interactions by BiFC analysis in rice protoplasts. Different combinations of the proteins fused with either C-terminal or N-terminal of the split Venus YFP (VC or VN) were introduced into the rice protoplast cells, and the reconstituted fluorescence signals were detected using a laser confocal microscope. While as negative controls no obvious signal was detected between any of the fusion proteins and the corresponding empty split YFP partial protein, we observed the strong interaction between VC-OFP1 and VN-DLT in nucleus (Figure 1D). Interestingly, the interacting signals of both protein pairs were mostly spotted in punctate-structures, reminiscent of nuclear bodies which could be formed due to the assembling and spatial-specific distribution of protein complex (Figure 1D). This distinct pattern also indicated that the fluorescence was not false-positive signal and the interaction was indeed occurred. In addition, we also detected the clear interaction between OFP1 itself (Figure 1E). Notably, in most observed cells, the interacting signals were diffusely distributed in part of cells, which appeared to be mainly in cytoplasm, with unclear reason (Figure 1E).

According to rice genome annotation project², OFP1 (LOC_Os01g12690) is intronless and contains 1185-bp coding sequence, 44-bp 5′-UTR and 244-bp 3′-UTR (Figure 2A). We analyzed the gene expression pattern by qRT-PCR, and found OFP1 was ubiquitously expressed in various rice tissues, preferentially higher in young panicles (Figure 2B). We also evaluated the gene promoter activity using GUS reporter and successfully detected the signals in various tissues of OFP1p:GUS transgenic plants, but mainly at young stages of the tissues (Figure 2C).

We produced the loss-of-function mutants of OFP1 by CRISPR/Cas9 gene-editing technology. A segment located at the 5′-end of the gene coding sequence was targeted for the editing, and a number of independent homozygous mutant lines were obtained. Two distinct alleles, ofp1-1 and ofp1-2, were selected for the further analysis. These alleles contained 35- and 65-bp deletions around the initial codon respectively, leading to the frameshift afterward of the sequence, suggesting that both are knockout mutants (Figure 2A). However, we failed to observe the morphological difference between ofp1 mutants and WT at both seedling stage and mature stage (Figure 2D). Both the mutants also exhibited indistinguishable BR sensitivity compared with the WT in lamina bending assay (Figure 2E). We further generated ofp1 ofp2 double mutant by similar strategy, yet the homozygous mutant plants still showed no obviously phenotypes under normal growth conditions as well as unaltered BR sensitivity in lamina bending assay (Figures 2F,G).

There are 33 OFPs in rice, and different OFP members could function with high redundancy, as strongly suggested in Arabidopsis (Wang et al., 2011; Liu et al., 2014). In this case, transgenic analysis was generally used to analyze the protein function. To this end, the whole OFP1 gene containing 2-kb upstream promoter and coding region was introduced into either Dongjing (DJ) or Zhonghua 11 (ZH11), two different Japonica WTs. Strikingly, although with different extents due to the various gene expression levels, most of the transgenic plants (OFP1-OE, designated as 1o for short in Figures) showed obvious phenotypes in both WT backgrounds (Figure 3). Unless specified, we majorly used transgenic lines in DJ background for our analysis, since we obtained the stable homozygous lines at the first place.

OFP1-OE had narrow leaves, enlarged leaf angles, decreased plant height, reduced tiller number, and long and narrow seeds (Figures 3A–F,H). The grain morphology was remarkably changed due to the greatly increase of the length/width ratio (Figures 3C,E,F). The severe lines are sterile, and have roll leaves and significantly decreased tiller numbers (Figure 3H). Most of these phenotypes, including the typical BR-enhanced phenotypes, largely resembled DLT-overexpression plants (Do), featured with the significantly narrowed leaves and seeds which could be, at least partially, BR-independent (Tong et al., 2012). Severities of the phenotypes were obviously consistent with the gene expression levels in the transgenic plants (Figures 3A–I), suggesting the involvement of OFP1 in controlling these phenotypes. However, we surprisingly found that the gene was remarkably over-expressed in the positive transgenic plants (Figures 3G,I). While the reason is unclear, one possibility is that there exists additional suppressor elements outside the selected promoter region.

We further tested the BR sensitivity of the plants by lamina inclination assay. OFP1-OE showed increased leaf angles than WT either with or without BR (Figures 4A,B). To exclude the possibility that the enhanced leaf angles were caused by increased BR levels, we analyzed the expression of BR synthetic genes including D2, D11, and DWARF and found that all of them have markedly decreased expression in OFP1-OE (Figure 4C), similar to that in Do plants (Tong et al., 2012). This result also implied that the BR synthesis was inhibited by enhanced BR responses in a feedback manner in OFP1-OE. Combined with the phenotypes, these results demonstrated that OFP1-OE plants have activated BR responses.

To further analyze the OFP1 functions, we introduced OFP1 into d2, a BR synthesis mutant (Hong et al., 2003), by transgenic method, and found that d2 OFP1-OE showed obviously enlarged leaf angles and decreased grain width compared with d2 (Figures 4D–F). We also introduced OFP1 into dlt mutant, and found dlt OFP1-OE showed similar morphology as dlt, suggesting that OFP1 required DLT to regulate plant architecture (Figure 4G). It should be noted that OFP1-OE can obviously suppress the grain width of dlt mutant (Figures 4G,H), demonstrating that the protein is functional in the transgenic plant. Considering BR has relatively minor roles in regulating grain width (Tanabe et al., 2005), this result indicated that OFP1 and DLT have differential functional relationship in different tissues or biological processes. Taken together, these results strongly suggested that OFP1 positively regulates BR responses.

We have performed RNA-seq analysis to compare the differentially expressed genes between high BL and low BL treated rice plants. In our available data, we found OFP1 was differentially over-expressed in 10^-6 M BL treated rice seedlings, but not in 10^-9 M BL treated materials. To confirm this result, we treated WT plants using various concentrations of BL, and then analyzed the expression of OFP1 by qRT-PCR. Consistently, the results showed that OFP1 was markedly induced by application of BR above 10^-7 M, but not induced under lower BR concentrations (Figure 5A). In addition, our available ChIP-seq data using OsBZR1 antibody enriched OsBZR1 protein on the promoter region of OFP1 with 3.28-fold increase in WT rice seedlings. Accordingly, we found OFP1 promoter contains multiple cis-elements that have been suggested to be bound by BZR1 (Figure 5B) (Sun et al., 2010; Yu et al., 2011). The association between OsBZR1 and the OFP1 promoter was further confirmed by ChIP-qPCR (Figure 5C). Compared with the samples with no antibody supplemented, OsBZR1 antibody can significantly pull down the DNA segments containing the putative BZR1 binding elements (Figure 5C).

In Arabidopsis, it was shown that AtOFP1 directly suppresses GA synthetic gene to repress plant growth (Wang et al., 2007). Our previous study also suggested that high BR levels or enhanced BR signaling will repress GA level to inhibit plant growth in rice (Tong et al., 2014). The specific induction of OFP1 by high BR (Figure 5A) and the strongly repressed growth of OFP1-OE plants (Figures 3A,B) prompted us to test whether OFP1 was involved in this process. Indeed, we found GA3ox-2 and GA20ox-2, two GA synthetic genes, had decreased expression, whereas GA2ox-3, the GA inactivation gene, had increased expression in OFP1-OE (Figure 6A). Accordingly, GA₁, the major bioactive GA form in young seedlings, was significantly decreased in the plants (Figure 6B), consistent with their decreased seedling height (Figure 3B). Actually, all the GA forms quantified were decreased in OFP1-OE plants (Figure 6B), reminiscent of the hormonal profiles obtained from high-BL-treated plants as shown previously (Tong et al., 2014). These results indicated that OFP1 is involved in the BL-induced GA repression and growth inhibition, depending on BR concentrations or plant tissues.

To test whether OFP1 was regulated at protein level, we fused the coding sequence with Flag tag and introduced into the WT under 35S promoter (OFP1-Flag-OE, designated as 1Fo for short in Figures). We obtained a collection of the overexpression plants which showed similar phenotypes as above described OFP1-OE plants, including enlarged leaf angle, reduced plant height, and altered grain shape (Figure 7A). Immunoblotting analysis using Flag antibody can detect the protein expression in the corresponding plants, demonstrating the fusion proteins are functional (Figure 7B). We treated the plants with different concentrations of BL, and found the protein was gradually accumulated (Figure 7C). We also tested the OFP1 protein in different leaves representing the different developmental stages, and revealed that the protein was highly accumulated in young leaves, but gradually decreased in old leaves (Figure 7D). These results suggested that OFP1 was dynamically regulated by hormone levels and development stages.

GSK3-like kinases, as the central negative regulators of BR signaling, can interact with many transcriptional factors to modulate their stability and activity (Youn and Kim, 2015). The promotion of OFP1 stability by BR prompted us to test whether GSK2 can directly interact with OFP1 to regulate this process. Indeed, yeast two-hybrid analysis revealed the obviously interaction between GSK2 bait and OFP1 prey (Figure 8A). Surprisingly, we can rarely or only weakly observe the interaction between GSK2 and OFP1 by BiFC analysis in rice protoplasts. One possibility is that the interaction is unstable in rice, considering GSK2 might promote OFP1 degradation. Alternatively, we adopted split-luciferase complementation assay to analyze their interaction in tobacco leaves (Figure 8B). When the GSK2 fused with N-terminal of luciferase protein (GSK2-NLuc) was co-expressed with the OFP1 fused with C-terminal of luciferase protein (CLuc-OFP1), the interaction signals were evidently detected. No signal was detected when the GSK2-NLuc was co-expressed with the empty CLuc protein (Figure 8B). Thus, GSK2 indeed can interact with OFP1 in plant.

To explore whether GSK2 regulates OFP1 stability, we treated OFP1-Flag-OE plants using bikinin, a specific chemical inhibitor of GSK3-like kinase activity, and analyzed the OFP1 protein level. Compared with the plants without treatment, we found OFP1 was significantly accumulated (Figure 8C). Taken together, these results suggested that BR enhances OFP1 protein stability by inhibiting GSK2.