Wild type rice ZH11 (Oryza sativa L. subsp. japonica cv. Zhonghua No.11), the lax1, A989 mutants, and all the other transgenic plants used in this study were grown in a paddy field or in pots in a greenhouse under standard growth conditions.

Homolog sequences of RA2 from Schizachyrium sanguineum (SsRA2), Sorghum bicolor (SbRA2), Andropogon hallii (AhRA2), Cymbopogon flexuosus (CfRA2), Andropterum stolzii (AsRA2), Phacelurus digitatus (PdRA2), Zea mays (ZmRA2), Chrysopogon gryllus (CgRA2), Hordeum vulgare (HvRA2), Oryza sativa (OsRA2), Loudetia sp. MCE (LsRA2), AtASL4 and AtLBD25 were obtained from Uniprot¹ (Supplementary Table S1), homologous sequences of AtASL4 and AtLBD25 were obtained from the National Centre for Biotechnology Information². Multiple sequence alignments of proteins were conducted using ClustalX Version 1.83 (Thompson et al., 1997). A phylogenetic tree of the sequences was reconstructed by the neighbor-joining (NJ) method by MEGA3 (Kumar et al., 2004).

To generate OsRA2 RNAi plasmid pdsRNAiOsRA2, a gene-specific fragment was amplified by KOD-plus DNA polymerase (TOYOBO) using primers OsRA2-RNAiF/OsRA2-RNAiR and cloned into pCAMBIA1301RNAi vector. To obtain the pUbi::OsRA2 plasmid, full-length cDNA of OsRA2 was amplified using primers OsRA2-BamHIA/OsRA2-KpnI and cloned into pCAMBIA1301UbiNos vector using BamHI and KpnI.

The embryogenic calli induced from immature ZH11 seeds were transformed by Agrobacterium tumefaciens infection and transgenic plants were regenerated (Hiei et al., 1994). Primer pairs Ubi 90+/OsRA2-KpnI and ocs-160/OsRA2-RNAiF were used to check pUbi::OsRA2 and pdsRNAiOsRA2 transgenic plants respectively.

For qRT-PCR assays, total RNAs were extracted from various tissues, panicles and seeds at different developmental stages using TRIzol reagent (Invitrogen), and reverse transcribed using oligo(dT) primer and ReverAce (TOYOBO). cDNA was synthesized from 2 μg of total RNA treated with DnaseI (TOYOBO) and used as templates for RT-PCR or qRT-PCR. Rice actin gene was used for normalization. Single and double asterisks represent significant difference determined by the Student’s t-test at ^∗P < 0.05 and ^∗∗P < 0.01 respectively.

For OsRA2 expression analysis, the roots, leaves, sheaths, culms, shoot apical meristem (SAM) were collected from around 30-day-old plants; panicles meristems less than 1 cm were collected under microscope, and flowers were collected from heading stage.

In situ hybridization was performed according to previous description (Coen et al., 1990). Young panicles at different developmental stages were fixed in 4% paraformaldehyde PBS solution overnight at 4°C, dehydrated through a concentration grade of ethanol, cleared through a dimethylbenzene series, then infiltrated through a series of paraffin (Sigma–Aldrich) melted at 60°C, and finally embedded in 100% paraffin. The longitudinal sections of young panicle were sectioned into 7 μm strips and mounted on RNase-free glass slides (Sigma). Gene-specific fragments of OsRA2 and OSH1 were amplified using primer pairs OsRA2-insitu F/OsRA2-insitu R and OSH1F/OSH1R respectively, and cloned into pBSK(-) vector, then linearized and used as templates for digoxigenin-labeled sense and antisense RNA probes, which were transcribed in vitro using a DIG RNA labeling kit (Promega).

Fifty grains from each sample were used for measurement of grain traits, including grain length (GL), grain width (GW) and grain thickness (GT). Length-to-width ratio (LWR) was obtained by dividing GL by GW. The 1,000-grain weight was got by measuring 100 grains in 10 biological repeats. Data was shown as mean ± SD.

Pedicels of ZH11, dsRNAiOsRA2 and pUbi::OsRA2 plants were cut transversely and fixed in 50% FAA at 4°C overnight after vacuuming. After dehydration in several concentrations of ethanol, samples were embedded in epoxide resin and cut into 2–3 μm slices and spread at 42°C on a hot platform overnight, stained using 0.5% toluidine Blue O and sealed for observation under the microscope (Wang et al., 2010). For the thickness measurement, at least 10 pedicels were made into histological sections and measured under microscope.

Full-length cDNA of OsRA2 was amplified and cloned in-frame with the LexA DNA-binding domain using primers Osa_yRA2EcoRI and Osa_ yRA2BamHI in PEG202 (His3, 2 μm, Amp^r, ADH constitutive promoter, LexA DNA-binding domain) (DupLEX-A^TM system) (OriGene Technologies). The pEG202-OsRA2 and pEG202 plasmids were introduced into yeast strain EGY48 (MATtrp1his3ura3leu2::6lexAop-LEU2) harboring the reporter plasmid pSH18-34 (URA3, 2 mm, Amp^r, LexAops-lacZ). Yeast clones were grown on SD medium without histidine or uracil in the presence of X-gal (5-bromo-4-chloro-3-indolyl-β-galactopyranoside) for 2 days at 30°C.

Full length cDNA of OsRA2 was cloned in frame with GFP into p1301-GFP using primers OsRA2-BamHIB/OsRA2-PstI and genetically transformed into ZH11. Root apices from the transgenic plants were observed through a confocal laser scanning microscopy (OLYMPUS FV1000).

Primer sequences used in this study were listed in Supplementary Table S2.

A search of the public databases^{3,4 ,5} revealed that only one gene, OsRA2, of the rice genome was homologues to maize RA2. OsRA2 contains four exons and three introns, with an open reading frame (ORF) in the third exon encoding a protein of 251 amino acids (Figure 1A). The deduced OsRA2 protein is composed of a characteristic N-terminal LOB domain and a variable C-terminal RA2 domain. The LOB domain is characteristic of the LBD family proteins (Bortiri et al., 2006), and in OsRA2 consists of a C block that functions in DNA-binding, a leucine zipper coiled coil motif for protein dimerization, and a Gly-Ala-Ser block (Figure 1B). The C block is represented by a CX₂CX₆CX₃C motif containing four conserved cysteine (C) residues and other non-conserved residues (X), and the leucine-zipper-like motif includes five hydrophobic amino acids separated by six variable amino acid residues (Shuai et al., 2002; Matsumura et al., 2009; Majer and Hochholdinger, 2011). Phylogenetic alignment of protein sequences revealed that RA2 is highly conserved among monocotyledons, while some diversification has occurred between grass RA2 proteins and Arabidopsis LBD proteins (Figures 1B,C).

To investigate the function of OsRA2, we carried out genetic analysis, using gene-specific doubled–stranded RNA interference (dsRNAi) and ectopic expression technologies.

We obtained a total of 27 T0 generation plants from the dsRNAiOsRA2 transformation, of which 24 showed elongated pedicels (Figures 2A–C). Successful down-regulation of OsRA2 in these lines was verified (Supplementary Figure S1). The pedicels of the panicles in dsRNAiOsRA2 plants were longer than those in wild type (WT) ZH11 (>1.4 cm vs. 1 cm; Figures 2B,C,F), but there was little change in the number and length of the PBs and SBs (Supplementary Figures S2A–D).

By contrast, pUbi::OsRA2 plants in which OsRA2 was over expressed showed a compact panicle and shortened pedicel (Figure 2A and Supplementary Figure S1), with an average pedicel length of 0.8 cm (Figure 2F), which was obviously shorter than WT (Figures 2B,C,F). Furthermore, the PBs, SBs, and the entire panicle were shortened (Figure 2G and Supplementary Figures S2A,B). There was little change in the number of the PBs (Supplementary Figure S2C), but the number of SBs decreased (Supplementary Figure S2D), such that the entire panicle was shortened and condensed (Figure 2A).

Based on the phenotype of dsRNAiOsRA2 and pUbi::OsRA2 transgenic plants, OsRA2 appears to regulate the pedicel length in rice.

Besides controlling pedicel length, OsRA2 also regulated the thickness of the pedicel, which was decreased in dsRNAiOsRA2 plants and increased in pUbi::OsRA2 plants (Figure 2C and Supplementary Figure S3); the thicknesses of the branches and the culms in their panicles were altered accordingly (Figures 2A,B). We further examined the transverse structure of the pedicels, and identified fewer vascular bundles in dsRNAiOsRA2 plants (Figure 2M), and more in pUbi::OsRA2 plants (Figure 2N), as compared with ZH11 (Figure 2L). So that, dsRNAiOsRA2 plants had thinner pedicels, while pUbi::OsRA2 plants had thicker ones.

OsRA2 also influenced the shape of the grains in transgenic plants. In dsRNAiOsRA2 plants, the grains were elongated (Figures 2D,H), but less wide (Figure 2I), so that the GL to width ratio was increased (Figure 2J). In pUbi::OsRA2 plants, the GL did not change (Figure 2H), but the GW increased (Figure 2I), so that the length to width ratio was decreased (Figure 2J). As a result, the 1,000-grain weight of dsRNAiOsRA2 plants decreased, while that of pUbi::OsRA2 plants increased (Figure 2K).

At the whole plant level, the height of pUbi::OsRA2 plants decreased, while that of dsRNAiOsRA2 plants remained unchanged (Figure 2E and Supplementary Figure S2E).

Taken together, OsRA2 appears to regulate pedicel development and seed development, and its over-expression also influences plant height in rice.

Gene expression is tightly linked to its function. To detect OsRA2 expression, we carried out quantitative real-time RT-PCR (qRT-PCR), semi-quantitative RT-PCR and in situ hybridization analyses. OsRA2 was shown to mainly express in young panicles and in shoot apical meristems (SAMs) at the vegetative stage. Weaker expression was also detected in the culms and flowers, and basal expression was revealed in the roots, leaves, and sheaths (Figure 3A). Based on our observation that OsRA2 regulates pedicel development, and that OsRA2 over-expression results in shortened PBs, SBs, and panicles, we next examined the expression of OsRA2 at different stages of panicle development (indicated by panicle length). Expression of OsRA2 gradually increased and then decreased during panicle development, with the highest transcription level observed in 0.2–0.6 mm and 0.6–0.9 mm panicle meristems (Figure 3B). The 0.2–0.6 mm panicle meristem corresponds to the stage of PBs formation and development, while the 0.6–0.9 mm panicle meristem corresponds to the formation and development of SBs, tertiary branches and pedicels. This suggests that levels of OsRA2 expression are in accordance with its function in pedicel development and panicle development.

We further used in situ hybridization to locate OsRA2 mRNA in the young panicles. OsRA2 mRNA was highly enriched in the anlagen of PB meristems (Figure 3C), SB meristems (Figure 3D), spikelet meristems (Figure 3E), and floral organ meristems (Figure 3F) as compared with the sense probe (Figure 3G). ORYZA SATIVA HOMEOBOX1 (OSH1), an ortholog of KNOTTED1 (KN1) in rice, is highly expressed in meristematic cells and is required for meristem maintenance (Sentoku et al., 1999; Oikawa and Kyozuka, 2009; Tsuda et al., 2011). Similarly, OSH1 transcripts were detected in the anlagen of PB meristems (Figure 3H), SB meristems (Figure 3I), and floral organ meristems (Figure 3J), as compared with the sense probe (Figure 3K), indicating the meristematic character of the region enriched in OsRA2 mRNA transcripts.

Most LBD proteins are transcription factors (Husbands et al., 2007). OsRA2 was shown to have a putative nuclear localization signal in its N-terminal LOB domain (Figure 1B)⁶, indicating that it is also likely to be a transcription factor.

To study the cellular localization of OsRA2, we constructed transgenic plants containing OsRA2 fused in-frame to Green Fluorescent Protein (GFP) at the C-terminal. The fluorescence signal from the fusion protein was observed in the nuclei of p35S::GFP-OsRA2 root tips (Figure 4A); but was ubiquitously distributed in the cytoplasm of control plants carrying an empty GFP vector (p35S::GFP) (Figure 4A), suggesting that OsRA2 is a nuclear-localized protein. Examination of the phenotype of transgenic plants suggested that OsRA2 functions normally when fused with GFP, because p35S::GFP-OsRA2 plants showed a dwarf phenotype similar to pUbi::OsRA2 plants (Supplementary Figure S4A), and the panicles of p35S::GFP-OsRA2 plants were shortened (Supplementary Figure S4B), similar to pUbi::OsRA2 plants.

We next detected if OsRA2 contained transcriptional activity. We used two plasmids (effector pEG202 and reporter pSH18-34) selected from a DupLEX-A^TM yeast two-hybrid system (Zhu et al., 2003). Full-length OsRA2 cDNA was fused in-frame with LexA protein and transformed into recipient yeast containing LacZ, whose promoter elements can be recognized by LexA. Yeast clones transformed with OsRA2-lexA showed a blue color similar to the positive control when supplemented with the β-galactosidase substrate X-gal, indicating that the OsRA2 protein has trans-activation activity (Figure 4B).

Together, these experiments strongly indicate that OsRA2 protein is a transcription factor.

To explore the pathway by which OsRA2 regulates pedicel development, we detected the expression of panicle-regulating genes in dsRNAiOsRA2 and pUbi::OsRA2 plants (Supplementary Figure S5). Among them, LAX1, MOC1 and the RCNs are involved in meristem initiation and development (Nakagawa et al., 2002; Komatsu K. et al., 2003; Li et al., 2003), SP1 regulates the panicle length (Li et al., 2009), DEP1 regulates number of branches and panicle length (Huang et al., 2009), and OSH1 is a meristem marker gene (Tsuda et al., 2011). LAX1 was shown to be up-regulated in pUbi::OsRA2 plants but down-regulated in dsRNAiOsRA2 plants (Figure 5A). lax1 mutants contain longer pedicel because of defects in initiation or maintenance of the lateral and terminal spikelets (Komatsu K. et al., 2003; Tabuchi et al., 2011), so we wondered if LAX1 is transcriptionally regulated by OsRA2. However,we found no OsRA2 binding motifs in the LAX1 promoter⁷. We also performed a yeast two-hybrid assay to check an interaction between OsRA2 and LAX1, but negative result turned out. Therefore, it remains unclear if any direct regulation or interaction between OsRA2 and LAX1exists.

OsRA2 expression was down-regulated in the lax1 mutant (Figure 5B), so to investigate the genetic relationship between OsRA2 and LAX1, we crossed the lax1 mutant with dsRNAiOsRA2 plants. And the number of SBs in the hybrid increased as compared with that in the lax1 mutant (Figures 5C,F), suggesting that LAX1 functions upstream of OsRA2 in determining the number of SBs. F2 lax1/dsRNAiOsRA2 plants showed longer pedicels compared with WT, which were more similar to those of dsRNAiOsRA2 plants (Figures 5D,E). Thus, LAX1 might also function upstream of OsRA2 in regulating pedicel length.

The over expression of RCN2 showed indeterminacy with increased PBs and SBs (Ikeda-Kawakatsu et al., 2012). To study the possible genetic relation between RCN2 and OsRA2, we crossed the pUbi::OsRA2 plants to the T-DNA insertion mutant of RCN2 in our lab (A989, Figure 6A), in the F2 generation, the characteristics of pUbi::OsRA2 plants and those of A989 were effectively additive, with the hybrid showing the shortened branches and panicles similar to those of pUbi::OsRA2 plants (Figures 6A,B,D,E), and an increased number of SBs similar to A989 (Figures 6A,C). This indicates that OsRA2 acts downstream of RCN2 in terms of pedicel and branch lengths, but upstream of RCN2 in terms of the number of SBs.