The activation tagging line of OsBDG1, PFG_1B_05536 is in Dongjin background (Jeong et al., 2006). Transgenic rice plants were produced with Tainung67 (TNG67) japonica rice cultivar. Rice plants (Oryza sativa) were grown in the field under natural long days or in the greenhouse with 28°C day/25°C night cycles. To assess the leaf angles of mature rice plants in the paddy field, leaves were numbered from top to bottom. Lamina angles were measured between a stem and a leaf blade at the second or the third leaf from eight to twelve individual plants per line when panicles come out of flag-leaf sheaths. Transgenic plants used for analyses in this work are all T3 or T4 independent lines.

Sterilized seeds were germinated and grown for 10 days in a dark chamber. Lamina joint inclination bioassays were performed as previously described (Jeong et al., 2007). Seedlings were sampled by excising approximately 2 cm segments that contained lamina joints at the same position from each plant under dim light condition. They were floated on distilled water containing various concentrations of BL. After incubation in a dark chamber at 28°C for 2 days, the angle induced between the lamina and the sheath was measured.

Each open reading frame (ORF) of OsBDG1, OsAP2, and OsWRKY24 was cloned into pGA3426 (Kim et al., 2009) or its derivatives for overexpression and/or dsRNAi purposes in rice. For expression by OsBUL1 promoter, the ubiquitin promoter of pGA3426 was replaced with the 2.2 kb-OsBUL1 promoter (Jang et al., 2017). Constructed plasmids were individually transformed into embryonic calli of TNG67 rice cultivars by Agrobacterium-LBA4404 mediation as described previously (Jeon et al., 2000).

Ten-day-old rice (O. sativa cv. TNG67) seedlings grown in the growth chamber were treated with brassinolide (1 μM, BL from Sigma–Aldrich) or GA (100 μM GA3 from Sigma–Aldrich). Whole parts above the roots were harvested for RNA extraction at 24 h time point after treatment.

Total RNAs of all the materials harvested were isolated using RNeasy plant mini kit (Qiagen) or Trizol solution (Invitrogen) according to the manufacturer’s instructions. RNAs after DNase treatment were subjected to reverse transcriptase reactions using oligo(dT) primer and Superscript III reverse transcriptase (Invitrogen) according to the manufacturer’s protocol. Subsequent PCR was conducted with the first-strand cDNA mixture and EX-Taq polymerase (Takara Bio). Quantitative PCR (qPCR) was carried out by a CFX96TM real-time system (Bio-Rad) using Maxima SYBR Green qPCR Master Mix (Thermo). For PCR, each sample was analyzed in triplicate. The run protocol was: denaturation at 95°C for 10 min and annealing/extension repeated 45 times (95°C for 15 s and 60°C for 30 s, data acquisition was performed). Housekeeping genes such as OsUBQ (Komiya et al., 2008) and OsAct (Caldana et al., 2007) were included in the reactions as internal controls for normalizing the variations in the amount of cDNA used (Guénin et al., 2009). The threshold cycle (C_T) was automatically determined for each reaction by the system set with default parameters. The specificity of the qRT-PCR was determined by curve analysis of the amplified products using the standard method installed in the system. Information on primers used is presented in Supplementary Table S1.

Collars of 80-day-old rice plants and young panicles less than 10 cm in length were harvested for RNA extraction for microarray. We used the Rice Whole Genome OneArray v1.1 (Phalanx Biotech Group, Taiwan) containing 22,003 DNA oligonucleotide probes. Each probe is a 60-mer designed in the sense direction. Among the probes, 21,179 probes corresponded to the annotated genes in the RGAP v.6.1 and BGI database. Additionally, we included 824 control probes. The detailed descriptions of the gene array list are available from http://www.phalanx.com.tw/products/RiOA_Probe.php. Fluorescent antisense RNA targets were prepared from 1 μg total RNA samples with the OneArray Amino Allyl aRNA Amplification Kit (Phalanx Biotech Group, Taiwan) and Cy5 dyes (Amersham Pharmacia, United States). Fluorescent targets were hybridized to the Rice OneArray with Phalanx hybridization buffer using Phalanx Hybridization System. After hybridization for 16 h at 50°C, non-specific binding targets were washed away by three washing steps (Washing I, 42°C 5 min; Washing II, 42°C, 5 min; 25°C, 5 min; Washing III, rinse 20 times), and the slides were dried by centrifugation and scanned by an Agilent G2505C scanner (Agilent Technologies, United States). The Cy5 fluorescent intensities of each spot were analyzed by GenePix 4.1 software (Molecular Devices). The signal intensity of each spot was introduced into Rosetta Resolver System (Rosetta Bio-software) to process data analysis. The error model of the Rosetta Resolver System could remove both systematic and random errors from the data. We filtered out the spots whose flag was less than 0. Spots that passed the criteria were normalized by 50% media scaling normalization method. The technical repeat data was tested by Pearson correlation coefficient calculation to check the reproducibility (R-value > 0.975). Normalized spot intensities were transformed to gene expression log₂ ratios between the control and treatment groups. The interest spots which show significant differences were selected by log₂ ratio ≥ 1 or log₂ ratio ≤-1 and P < 0.05. Two independent biological replicates of hybridizations were performed.

Lamina joint samples were fixed by 4% paraformaldehyde in 0.1 M sodium phosphate buffer, dehydrated through a series of graded ethanol baths, replaced with xylene, and embedded in Paraplast plus (Sigma–Aldrich). Paraffin sections (12 μm) were cut and stained with filtered 1% toluidine blue. The sections were photographed under a light microscope (Olympus BX51).

For cellular localization of OsBC1, OsBDG1, OsBUL1, OsAP2.2, and OsWRKY24 in rice, either yellow florescence protein (YFP):gateway (GW) or cyan fluorescent protein (CFP):GW vector was used for the florescence fusion as described previously (Jang et al., 2015). Isolation and transfection of rice protoplasts were followed as described by Zhang et al. (2011) and images of cells with fluorescence were taken by confocal microscopy (LSM 510 META NLO DuoScan, Carl Zeiss).

To observe epidermal cells of lemma in rice grains by scanning electron microscopy (SEM), whole grains were coated with gold by the ion sputter machine (E-1010, Hitachi) and examined with SEM (Quanta 250, FEI, Hillsboro, OR, United States).

To identify the downstream genes of OsBUL1 in the expressional hierarchy, comparison of transcriptomes of collars and panicles between WT and osbul1 plants was conducted using rice 22k-oligo microarrays¹. The osbul1 is a mRNA null mutant caused by a T-DNA insertion in the first exon of OsBUL1 gene (Jang et al., 2017).

Fourteen genes were upregulated, while 221 genes were downregulated in both collars and panicles of osbul1 plants (Figure 1A). Among candidate genes, OsBUL1 DOWNSTREAM GENE (OsBDG1) was selected for further study since expression level of OsBDG1 was reduced both in collars and panicles of osbul1 plants (Figure 1B), and OsBDG1 encodes a novel small protein containing a conserved LRR N-terminal domain (LRRNT) at the N-terminal part and three leucin-rich-repeat (LRR) domains toward its carboxyl terminus. In addition, its expression is known to be induced in transgenic rice plants overexpressing Arabidopsis CYP90B1 which encodes for a sterol C-22 hydroxylase active in BR biosynthesis (Wu et al., 2008). To verify the microarray results, we performed expressional analyses on OsBDG1 in collars and panicles of osbul1 and WT plants, and found similar expression patterns to the microarray results (Figure 1C). On the contrary, the level of OsBDG1 expression was higher in OsBUL1-overexpressing rice plants compared to the WT control (Figure 1D) indicating OsBUL1 may act upstream of OsBDG1. Transgenic rice plants harboring pUbi:OsBDG1 show increased lamina inclination with elongated cells in the lamina joint and grain size (Figures 2A–C) whereas reduced expression of OsBDG1 results in erect leaves with reduced cell length in lamina joint and smaller grains (Figures 2D–F). Phenotypes of overexpressors and dsRNAi lines for OsBDG1 are similar to those of rice plants with increased OsBUL1 expression and osbul1, respectively, supporting the expressional relationship between OsBUL1 and OsBDG1.

A putative activation tagging line for OsBDG1, PFG_1B_05536 has been identified through the rice T-DNA database² (An et al., 2003). Phenotypic analysis of the heterozygous population displayed a 3:1 ratio of OsBDG1D mutant phenotype to wild-type indicating OsBDG1D is a dominant mutant. T-DNA flanking sequence of the mutant was confirmed by PCR and comparison of the sequence through blastn search showed that the T-DNA was inserted in the 4.1 Kb downstream of Os11g31530 near Os11g31520 on chromosome 11 (Figure 3A). The expression of Os11g31530 was obviously influenced whereas the expression level of other genes located near the insertion position was not significantly affected (Figure 3B). The lamina angle increased more than two-fold in the activation tagging line with an increase of grain size (Figures 3C–E), which is similar to OsBDG1 overexpressors.

To confirm the expression cascade of OsBUL1 and OsBDG1, we made a construct for expression of OsBDG1 under the control of the 2.2 kb-OsBUL1 promoter (Figure 4A) preferentially active in lamina joints and panicles of rice (Jang et al., 2017). Transgenic rice containing the construct exhibited a dramatic increase in lamina inclination and grain size (Figures 4B–E), which was similar to the lamina inclination and grain phenotypes of the OsBDG1 activation tagging line, PFG_1B_05536 and OsBDG1-overexpressing lines driven by ubiquitin promoter. Thus, expression of OsBDG1 under the control of OsBUL1 promoter phenocopies OsBUL1 overexpression by ubiquitin promoter or activation tagging systems (Jang et al., 2017) suggesting that OsBDG1 may act downstream of OsBUL1.

To identify candidate genes affected by OsBDG1 in the lamina joint and also likely responsible for the increased lamina inclination, we compared transcriptomes of collars among two independent homozygous transgenic lines of pOsBUL1:OsBDG1, #13-2-2 and #20-1-5, and wild-type plants by microarray hybridization³. Since the expression level of OsBDG1 in the collar of a #13-2-2 plant is lower than that of a #20-1-5 plant (Figure 4C), we selected genes that showed the same increasing or decreasing patterns in the order of WT, #13-2-2 and #20-1-5. Five genes were identified as being upregulated by OsBDG1 in the lamina joint (Figures 5A,B) but no candidate was found to be downregulated in the comparison. Also, no significant difference in OsBUL1 expression was detected among WT, #13-2-2 and #20-1-5 lines (GEO accession no. GSE93817 for microarray results). We confirmed the expression patterns of each gene by qRT-PCR (Figure 5C). The Os10g41330 (OsAP2) produced two transcripts, Os10g41330.1 and Os10g41330.2 by alternative splicing (Supplementary Figure S1) and both transcripts showed similar accumulation patterns. Based on the increased expression level of each gene in the transgenic plants, we focused on two genes, Os10g41330 (OsAP2) and Os01g61080 (OsWRKY24) encoding putative transcription factors containing an AP2 domain and WRKY domain, respectively. Conversely, in collars of OsBDG1 dsRNAi lines, the expression of OsAP2.2 and OsWRKY24 is reduced without altered expression level of OsBUL1 supporting the notion that the two genes are downstream of OsBDG1 in expression (Supplementary Figure S2). However, the expression of OsBUL1 and OsBDG1 was positively affected by pOsBUL1:OsAP2 and/or pOsBUL1:OsWRKY24 (Supplementary Figure S3).

Spatiotemporal expression of OsBDG1, OsAP2, and OsWRKY24 also overlapped in collars and growing panicles (Supplementary Figure S4A). Interestingly, the expression of the three genes was upregulated by phytohormones such as GA3 and BL, which affect cell elongation (Supplementary Figure S4B). Indeed, OsBDG1 dsRNAi lines exhibited reduced, but transgenic rice containing pOsBUL1:OsBDG1 showed increased sensitivity in BR response through lamina bending assays (Supplementary Figure S4C).

The OsBDG1 protein containing a short LRR motif was localized in the cytoplasm as well as the nucleus similar to OsBUL1 (Jang et al., 2017; Supplementary Figure S5). However, OsAP2.2 and OsWRKY24 proteins are localized in the nucleus and each protein shows a transcriptional activation activity in the yeast system (Supplementary Figures S5, S6C).

Transgenic rice plants containing pOsBUL1:OsAP2.2, a short transcript of Os10g41330 and pOsBUL1:genomic OsAP2 were generated for phenotypic analyses of lamina angles and grain size (Figure 6 and Supplementary Figure S1). Compared with the wild-type control, transgenic rice plants showed increased lamina inclination with increased amounts of OsAP2.2 transcripts (Figures 6A,B). However, we could not detect the long form of the transcript, OsAP2.1 in the transgenic plants containing pOsBUL1:genomic OsAP2. Transgenic rice for pOsBUL1:OsWRKY24 also exhibited a significant increase in lamina angles (Figure 6C) indicating both OsAP2.2 and OsWRKY24 are likely affected by OsBDG1 in the lamina joint. Of note, reduced expression level of both OsAP2.2 and OsWRKY24 was observed in panicles and lamina joints of osbul1 plants (Supplementary Figure S2). Additionally, panicle morphology of transgenic rice was affected by the two transgenes: panicle branches were spread and grain size was also increased (Supplementary Figure S3C). Elongated epidermal cells of lemma were also observed with higher expression level of genes involved in cell elongation such as OsExpansin (OsEXP) genes, OsXyloglucan endotransglucosylase/hydrolase1 (OsXTH1) and Osxyloglucan endotransglycosylase related1 (OsXTR1) in spikelets of transgenic rice plants (Supplementary Figure S6; Cho and Kende, 1997; Uozu et al., 2000; Hara et al., 2014).