SLG controls grain size and leaf angle by modulating brassinosteroid homeostasis in rice

Highlight The rice SLG gene, functioning as homomers, plays essential roles in regulating grain size and leaf angle via modulation of brassinosteroid homeostasis.


Introduction
Rice (Oryza sativa) feeds more than half of the world's population as one of the world's most important cereal crops. Given the rapid increase in the world's population and decrease in cultivated land area, improving rice production remains a great challenge for rice breeding programs. Grain size and leaf angle are two important traits determining rice grain yield and have always been a consideration in breeding programs (Sinclair and Sheehy, 1999;Ikeda et al., 2013).
Grain size, determined by grain length, grain width, and grain thickness, not only contributes to grain yield, but also influences the appearance, processing, cooking, and eating quality of rice. For example, people in Japan, Korea, and Northern China favor medium length round grains, whereas people in the USA, Southeast Asian countries, and Southern China prefer long and slender grains (Unnevehr et al., 1992). The organ size is largely determined by the cell number and cell size during organogenesis (Potter and Xu, 2001;Sugimoto-Shirasu and Roberts, 2003). In recent years, several genes and quantitative trait loci (QTLs) that affect grain size by influencing cell number have been identified in rice, including GS3, GW2, GW5, GS5, GW8, qGL3, TGW6, GW6a, andBG1 (Song et al., 2007, 2015;Weng et al., 2008;Mao et al., 2010;Li et al., 2011;Wang et al., 2012;Zhang et al., 2012;Ishimaru et al., 2013;Liu et al., 2015). Some other genes and QTLs that control grain size by influencing cell size have also been isolated in rice, including PGL1, GL7, and GS2/GL2 (Heang and Sassa, 2012;Che et al., 2015;Duan et al., 2015;Hu et al., 2015;Wang et al., 2015). It has been documented that at least some of those genes participate in seed size control by regulating biosynthesis and signaling of plant hormones, including brassinosteroids (BRs), cytokinins, gibberellins, and auxin (Ashikari et al., 1999Hong et al., 2003;Tanabe et al., 2005;Ishimaru et al., 2013).
Leaf angle, the inclination between the leaf blade and vertical culm, is a key factor determining the plant architecture (Hoshikawa, 1989;Sinclair and Sheehy, 1999). A compact plant type with erect leaves is preferred since it increases photosynthetic efficiency and nitrogen storage for grain filling, and facilitates dense planting (Sinclair and Sheehy, 1999;Sakamoto et al., 2006). A number of genes or QTLs have been reported to have a role in controlling leaf angle, including Ta1, OsDWARF4, D2, OsBRI1, OsBU1, ILI1, LC2, and ILA1 (Li et al., 1998(Li et al., , 1999Sakamoto et al., 2006;Tanaka et al., 2009;Zhang et al., 2009;Zhao et al., 2010;Ning et al., 2011). The leaf lamina joint that connects the leaf blade and sheath is considered the most important tissue governing the leaf angle. The degree of the leaf angle largely depends on cell division and expansion as well as cell wall composition at the joint (Nakamura et al., 2009;Zhang et al., 2009;Zhao et al., 2010;Ning et al., 2011). Nevertheless, it is well known that BR treatment stimulates leaf inclination in rice (Wada et al., 1981).
BRs are a group of steroidal phytohormones that regulate diverse plant growth and developmental processes, including cell expansion and division, vasculature differentiation, root and leaf development, stem elongation, skotomorphogenesis, and grain filling (Clouse and Sasse, 1998;Fujioka and Yokota, 2003;Wu et al., 2008). In recent decades, researchers have clarified many genes and the main pathway of BR biosynthesis utilizing genetic studies, chemical feeding, and enzymatic analysis. Most of the enzymes known to catalyze BR biosynthesis belong to the cytochrome P450 protein family (Choe, 2006). The BR biosynthesis pathway mainly consists of the early and late C-22 oxidation pathway, and the early and late C-6 oxidation pathway (Choe, 2006). Similarly, research on BR signaling has also developed rapidly, and most of the main participants in BR signaling have been determined in Arabidopsis (Belkhadir and Chory, 2006). BRs are perceived by the receptor kinase BRI1 to transmit signaling (Li and Chory, 1997). In rice, it has been reported that BR plays important roles in the regulation of grain size, leaf angle, and yield potential. For example, most loss-of-function mutants in BR biosynthesis or signaling pathways, such as d2, d11, and d61, display short grains, erect leaves, and dwarf phenotypes (Yamamuro et al., 2000;Hong et al., 2003;Tanabe et al., 2005), while some other mutants or transgenic plants with enhanced BR signaling or increased BR levels, such as GSK2 knockdown lines, the DLT overexpresser, and the D11 activation mutant m107, show longer grains and larger leaf angles (Tanabe et al., 2005;Wan et al., 2009;Tong et al., 2012). More importantly, modulating the expression level of BR-related genes such as OsDWARF4 and OsBRI1 has been proven to improve rice grain yield at higher planting densities Sakamoto et al., 2006). BR homeostasis is vital for normal growth and development of plants. BRs are synthesized in most plant tissues, and their level is the highest in young developing organs but low in mature organs (Shimada et al., 2003). Unlike the other plant hormones such as auxin that can be transported from the site of synthesis to a distant target site (Berleth and Sachs, 2001), BRs do not undergo long-distance transport and have the same site of synthesis and action (Symons and Reid, 2004). Therefore, there exist mechanisms that cells or tissues use to modulate levels of endogenous BRs precisely to keep cell expansion in balance and ensure normal plant growth and development (Symons and Reid, 2004). Negative feedback regulation is a common mechanism that also regulates BR homeostasis. It is reported that expression of many BR biosynthesis and signaling genes is inhibited by BR treatment, such as D2, D11, OsDWARF4, BRD1, OsBRI1, and DLT in rice (Yamamuro et al., 2000;Hong et al., 2002Hong et al., , 2003Tanabe et al., 2005;Sakamoto et al., 2006;Tong et al., 2009). However, BR homeostasis is still poorly understood.
In this study, we characterized a rice semi-dominant mutant, slender grain Dominant (slg-D), with slender grains and enlarged leaf angles, which are caused by enhanced expression of SLG, a BAHD acyltransferase-like protein gene. We provide genetic evidence that the BR contents are associated with the expression level of SLG. In addition, the plants expressing an RNAi or a truncated version of SLG showed a semi-dwarf architecture with smaller leaf angles, which may be useful for rice yield improvement.

Plant materials
The slg-D mutant (3A-10513) was isolated from a collection of activation-tagging T-DNA insertion rice lines (Jeon et al., 2000;Jeong et al., 2006), and kindly provided by Professor Gynheung An. The wild type (WT) of slg-D was Dongjin, a japonica cultivar. d61-1, d11-2, and m107 were kindly provided by Professor Chencai Chu (Tong et al., 2012). The WT of d61-1 and d11-2 was a japonica cultivar, Zhonghua11, and the WT of m107 was a japonica cultivar, Nipponbare. Rice plants were cultivated in an experimental field under natural long-day conditions in Nanjing, China.
Scanning electron microsocpy (SEM) and light microscopy For SEM, lemmas were harvested from florets after flowering and fixed in 2.5% (v/v) glutaraldehyde. Fixed samples were soaked in 2% (w/v) OsO4 for 2 h, dehydrated in a graded ethanol series, infiltrated and embedded in butyl methyl methacrylate, treated with critical point drying, and then sputter coated with platinum. The outer and inner epidermal cells of lemmas were observed using a HITACHI S-3400N scanning electron microscope. For light microscopy, lamina joints of the second leaves were harvested 10 d after flowering and fixed with FAA solution, followed by a graded series of dehydration and infiltration steps. Fixed tissues were embedded in paraplast. After sectioning, 10 μm thick sections were dewaxed with xylene, rehydrated, stained with 1% toluidine blue, and observed with a Leica DM5000B microscope. Cell lengths and widths of each organ were measured with IMAGEJ software.

Isolation, cloning, and RNAi suppression of the SLG gene
To identify the T-DNA insertion locus in slg-D, we searched the flanking sequence database (Jeong et al., 2006; http://orygenesdb. cirad.fr/). The T-DNA loci were confirmed by PCR genotyping, using the primers P1, P2, and P3 (see Supplementary Table S1 at JXB online). To recapitulate the phenotype of slg-D, full-length cDNAs of Loc_Os08g44830 and Loc_Os08g44840 were amplified by PCR and cloned into the binary vector pCUbi1390 under the control of the maize Ubi promoter to create p1390-Ubi-830, and p1390-Ubi-840 constructs, respectively. These constructs were then transformed into the rice variety Dongjin according to a published Agrobacterium-mediated method (Hiei et al., 1994).
To obtain SLG RNAi plants, the construct pCUbi1390-ΔFAD2 (an FAD2 intron and ubiquitin promoter inserted into pCUbi1390) was used as an RNAi vector (Wu et al., 2007). Both sense and antisense versions of a specific 305 bp fragment from the cDNA of SLG were amplified with primer pairs SLG-RNAiL and SLG-RNAiR (Supplementary Table S1), and cloned into pCUbi1390-ΔFAD2 to create the pUbi-dsRNAiSLG construct, which was then transformed into the rice variety Dongjin by the Agrobacterium-mediated method described above.

RNA extraction and quantitative RT-PCR
Total RNA from roots, leaves, leaf sheaths, lamina joints, shoot apices, culms, and different stages of panicles were isolated using the RNAprep Pure Plant Kit (TIANGEN, Beijing, China). First-strand cDNA was reverse transcribed from 1 μg of total RNA using the PrimeScript 1st Strand cDNA Synthesis Kit (TaKaRa). Quantitative RT-PCR was performed using a SYBR Premix Ex TaqTM kit (TaKaRa) on an ABI prism 7500 Real-Time PCR System according to the manufacturer's instructions, and the ACTIN1 gene was used as an internal control. The primers for quantitative RT-PCR analysis are listed in Supplementary Table S1.

GUS staining
To analyze the expression pattern of SLG, an ~2.5 kb promoter fragment was cloned into the pCAMBIA1381Z vector to create the PRO SLG :GUS (β-glucuronidase) reporter gene construct, which was then transformed into the rice variety Dongjin by the Agrobacteriummediated method. GUS staining was performed on PRO SLG :GUS T 1 generation transgenic plants according to a method described previously (Jefferson et al., 1987). Images were taken using a Nikon CD5Ri1P camera. Primers used to clone the promoter fragment are listed in Supplementary Table S1.
In situ hybridization RNA in situ hybridization was performed as described previously (Bradley et al., 1993). A 305 bp gene-specific region of SLG amplified with primers SLG-PF and SLG-PR (see Supplementary Table S1) was cloned into the pGEM-T Easy vector (Promega). The linearized templates were amplified from the pGEM-T plasmid containing the gene-specific region of SLG using primers Yt7 and Ysp6. Digoxigenin-labeled RNA probes were transcribed in vitro using T7 and SP6 RNA polymerases, respectively, using a DIG Northern Starter Kit (Cat. No. 2039672, Roche) following the manufacturer's instructions. Images were taken using a Leica DM5000B microscope.

Subcellular localization of SLG
To determine the subcellular localization of SLG, green fluorescent protein (GFP) was fused to the C-terminus of SLG under the control of the 35S promoter in the PAN580 vector. In addition, the nuclear marker D53-mCherry was constructed. The SLG-GFP fusion construct was transiently co-transferred into rice protoplasts with the D53-mCherry constructs according to the method described previously (Chen et al., 2006). Next, GFP was fused to the C-terminus of SLG under the control of the Cauliflower mosiac virus (CaMV) 35S promoter in the pCAMBIA1305.1 vector. The pCAMBIA1305-SLG-GFP construct was transformed into the rice variety Dongjin by the Agrobacterium-mediated method. GFP fluorescence was examined in the young roots of 2-week-old T 1 transgenic plants. Fluorescence images were observed using a Zeiss LSM510 confocal laser microscope. Primers used to make these constructs are listed in Supplementary Table S1.

BR and BRZ treatment
The lamina joint bending assay using excised leaf segments was performed as described by Wada et al. (1981). Seeds were germinated for 2 d and then grown in the dark for 8 d at 30 °C. Segments of 2 cm comprising the second leaf blade, lamina joint, and leaf sheath were floated on distilled water for 24 h and then incubated in 2.5 mM maleic acid potassium solution containing various concentrations of brassinolide (BL; Sigma, http://www.sigmaaldrich.com/) for 48 h in the dark. Lamina joint angles were measured using IMAGEJ software. The coleoptile and root elongation tests were performed using a previously described method (Yamamuro et al., 2000). Seeds were germinated on agar plates containing various concentrations of BL, and then the coleoptile and root lengths were measured 1 d after germination.
To measure the effect of brassinazole (BRZ; TCI) treatment on lamina joint bending, the leaf tips of 8-day-old seedlings of slg-D and the WT were spotted with 1 μl of DMSO containing 0 or 10 μM BRZ daily for 3 d, followed by 7 d growth in a controlled growth chamber under long-day conditions (16 h light at 28 °C/8 h darkness at 24 °C). The angles of the third lamina joints were measured using IMAGEJ software.

Yeast two-hybrid assay
The full-length cDNA of SLG was cloned into pGBKT7 (Clontech, http://www.clontech.com). Full-length SLG as well as its N-and C-terminal truncated deletions were then subcloned into pGADT7 (Clontech) and all vectors were transformed into yeast strain AH109. A yeast two-hybrid library was constructed from the mRNA of young rice panicles 0.1-5 cm long. Yeast transformation and screening procedures were performed according to the manufacturer's instructions (Clontech). Primers used to make these constructs are listed in Supplementary Table S1.

Bimolecular fluorescence complementation (BiFC) assay
The full-length SLG cDNA was cloned into the vector pSPYCE(M), and the SLG cDNA and its truncated deletions were then subcloned into the vector pSPYNE173. The plasmids were transiently expressed in Nicotiana benthamiana leaves as described previously (Waadt and Kudla, 2008). Yellow fluorescent protein (YFP) fluorescent signals were observed under a Zeiss LSM510 confocal laser microscope between 48 h and 72 h post-transfection. Primers used to make these constructs are listed in Supplementary Table S1.

Pull-down assay
The SLG cDNA was cloned into the vectors pMAL-c2x and pGEX4T-2 to generate fusions with maltose-binding protein (MBP) and glutathione S-transferase (GST), respectively. Expression of MBP-SLG, GST-SLG, and GST in BL21 Rosetta cells was induced with 0.5 mM isopropyl-β-d-thiogalactoside at 16 °C for 20 h. The total protein concentration was quantified using the Bio-Rad protein assay reagent. The pull-down assay was performed as reported previously (Miernyk and Thelen, 2008). The proteins were separated on a 10% SDS-PAGE gel and immunoblotted with anti-GST or anti-MBP antibodies (Abmart, http://www.ab-mart.com). The primers used to make these constructs are listed in Supplementary Table S1.

Phenotypic characterization of the semi-dominant mutant slg-D
To identify new components involved in regulating rice grain size, we screened a collection of activation-tagging T-DNA insertion rice lines (Jeon et al., 2000;Jeong et al., 2006). As a result, we isolated a mutant (3A-10513) with a slendergrain phenotype, and named it slender grain Dominant (slg-D). slg-D showed less compact plant architecture than the WT at both the vegetative and mature stages (Fig. 1A, B). In slg-D, the grain length was significantly increased while the grain width decreased, and the 1000-grain weight was slightly decreased (Fig. 1C-F). The lamina joint bending angles of slg-D were larger than those of the WT, especially for the flag leaves (Fig. 1G, H). Together, these results indicate that slg-D displays slender grain and an enlarged leaf angle.
The F 1 plants from the cross slg-D×WT exhibited an intermediate phenotype in grain shape and leaf angle, indicating a semi-dominant nature of the mutation (Supplementary Fig. S1A-F). Genetic analyses of an F 2 population derived from the same cross showed a segregation ratio of 1:2:1 (64 normal:120 intermediate:56 mutant; χ 2 =0.09, P>0.05), suggesting that slg-D is a single locus mutation ( Supplementary  Fig. S1G). This observation provides a hint that the semidominant nature of slg-D might be associated with an insertion of the activation-tagging T-DNA.

Cell length change in slg-D determines the mutant phenotypes
To investigate the mutant phenotypes in slg-D at a cell level, we performed SEM and light microscopy on slg-D plants along with the WT. The observations on the outer and inner epidermal cells of lemmas, which determine the shape and size of grains, showed that those cells were stretched longitudinally in slg-D, such that the slender-grain phenotype was developed ( Fig. 2A-G). The histological analysis on the second leaf lamina joints indicated that there was no significant alteration in cell size in the abaxial sides between slg-D and the WT (Fig. 2H-K). In contrast, the cell length of the adaxial sides was increased in slg-D (Fig. 2H, L-N). Therefore, it is the asymmetric cell expansion at the opposite sides of the lamina joint that causes a larger leaf bending in slg-D. Overall, these results indicate that the changes in cell length are responsible for the phenotype in slg-D.

Enhanced expression of Loc_Os08g44840 in slg-D leads to the mutant phenotypes
To isolate the gene for slg-D, we searched the T-DNA insertion database and obtained a genomic flanking sequence (Jeong et al., 2006; http://orygenesdb.cirad.fr/). Based on this information, we designed three PCR primers (P1, P2, and P3) and confirmed the site of the T-DNA insertion in slg-D (Fig. 3A, B). To understand whether the mutation in slg-D is related to an insertion of the activation-tagging T-DNA, we PCR-genotyped the mutant F 2 plants derived from the cross slg-D×WT and found that the mutant phenotypes were always associated with the presence of the T-DNA ( Supplementary Fig. S2). A BLAST search (http://www.ncbi. nlm.nih.gov/) showed that the T-DNA was inserted in an intergenic region, with Loc_Os08g44830 2000 bp upstream and Loc_Os08g44840 4500 bp downstream (Fig. 3A). Another nearby gene is Loc_Os08g44820, upstream of Loc_ Os08g44830 (Fig. 3A). Next we examined if the four enhancer repeats in the T-DNA had an influence on the expression level of the three nearby genes, and found that two of them (Loc_ Os08g44830 and Loc_Os08g44840) had elevated expression (Fig. 3C). This result suggests that the changed expression level of the two genes might be responsible for the mutant phenotypes in slg-D.
We assumed that overexpression of Loc_Os08g44830, Loc_Os08g44840, or both in the WT might recapitulate the phenotypes observed in slg-D. To test this assumption, Loc_Os08g44840 and Loc_Os08g44830, under control of the maize Ubi promoter, were individually overexpressed in WT plants. Interestingly, only plants overexpressing Loc_ Os08g44840, not Loc_Os08g44830, showed varying degrees of enlarged leaf angle and slender grain, thus phenocopying slg-D (Fig. 3D-I; Supplementary Fig. S3). The phenotypic variation in the transgenic plants was well correlated with the expression level of Loc_Os08g44840 (Fig. 3D-I). We concluded that it is the enhanced expression of Loc_Os08g44840 that causes the phenotypic changes in slg-D. We designated Loc_Os08g44840 as a SLENDER GRAIN (SLG) gene.

SLG encodes a putative BAHD acyltransferase-like protein
SLG encodes a protein of 445 amino acids. SLG belongs to the putative BAHD family of acyltransferases, which catalyze formation of a diverse group of plant metabolites using CoA thioesters as substrates (D'Auria, 2006). A phylogenetic analysis revealed that SLG-like proteins largely fall into two groups: dicot and monocot, and SLG is a member of the monocot group ( Supplementary Fig. S4). However, to date, none of the genes in this group has been functionally characterized.

Expression pattern and subcellular localization of SLG
Our quantitative RT-PCR analysis showed that expression of SLG is strong in young panicles, relatively high in lamina joints, low in shoot apices, culms and leaves, and very little in roots and leaf sheaths (Fig. 4A). Further, SLG has the highest expression level during early panicle development, but drops dramatically as the spikelets reach their final size (Fig. 4B). To investigate further the expression pattern of SLG, a genomic sequence ~2.5 kb upstream of the translation start site was cloned and introduced into the pCAMBIA1381Z vector, resulting in the PRO SLG :GUS reporter construct. Analysis of GUS activity in transgenic lines showed that strong GUS staining was observed in young panicles, lamina joints, and young stem nodes, with faint staining in leaf veins, but not visible in roots and leaf sheaths (Fig. 4C, parts 1-8). Cross-sectioning of the GUS-stained leaf lamina joint and the bottom portion of the stained internode further showed that the GUS signals were mainly restricted to vasculature regions (Fig. 4C, parts 9-12). We also performed RNA in situ hybridization to localize SLG expression during early panicle development more precisely. Strong SLG expression was detected in spikelet meristem primordia, floral meristem primordia, lemma and palea primordia, and vasculature regions (Fig. 4D). The predominant expression of SLG in young panicles and lamina joints implies its role in controlling grain shape and leaf angle.
To determine the subcellular localization of SLG, we fused the green fluorescent protein (GFP) to the C-terminus of SLG. Transient expression of this fusion protein in rice protoplasts revealed that the GFP signals were found in both the cytoplasm and the nucleus (Fig. 4E). Transgenic rice plants harboring the same fusion construct also showed the cytoplasmic and nuclear localization pattern of SLG (Fig. 4F).

SLG positively regulates endogenous BR levels
The slg-D phenotype resembles that of an activation mutant or transgenic plants with elevated BR accumulation (Wu et al., 2008;Wan et al., 2009), and that of transgenic plants with enhanced BR signaling (Tanaka et al., 2009;Tong et al., 2012), leading us to hypothesize that SLG may be involved in regulating the BR pathway. To determine whether slg-D responds differently to BR treatment, we first performed lamina joint bending assays using excised leaf segments (Wada et al., 1981). We measured the effects of a range of 24-epibrassinolide (BL; a type of active BR) concentrations on the angle of the lamina joints, and found that lamina joint bending was increased in a dose-dependent manner and the sensitivity to BL treatment was similar in slg-D and the WT (Fig. 5A,  B). Next we performed another BR response assay involving coleoptile and root elongation using a previously described method (Yamamuro et al., 2000). Comparison of coleoptile and root lengths also showed that slg-D had a similar response to BL as the WT (see Supplementary  Fig. S6). These results indicate that BR signaling is not altered in slg-D.
To investigate whether SLG functions in regulating endogenous BR levels, we first tested the effect of brassinazole (BRZ; a specific BR biosynthesis inhibitor; Asami et al., 2000) on slg-D in a lamina joint bending experiment. The leaf tips of 8-day-old seedlings of slg-D and the WT were spotted with 10 μM BRZ daily for 3 d, followed by 7 d growth in a chamber. We observed that the leaf angle of slg-D was restored to the WT level by BRZ treatment, whereas the WT seedlings had a milder response to the same treatment (Fig. 5C, D), indicating that slg-D is more sensitive to BRZ. A similar result was also seen when the D11 activation line m107, a BR overproduction mutant, was treated with BRZ (see Supplementary Fig. S7). Next we introduced the SLG-overexpressing construct into d61-1 (a loss-of-function mutant of the BR receptor gene OsBRI1; Yamamuro et al., 2000) and d11-2 (a mutant deficient in BR biosynthesis; Tanabe et al., 2005), and found that the d61-1 SLG:OE and d11-2 SLG:OE plants still retained the dwarfism, smaller and round grains, and erect leaves (Fig. 5E-G). Those results suggest a role for SLG in regulating BR levels. Consistent with this, our chemical analysis indeed showed a higher content of 6-deoxocastasterone (6-deoxoCS), typhasterol (TY), and castasterone (CS) in slg-D than in the WT ( Supplementary  Fig. S8).
It is known that excessive BRs down-regulate the BR-related genes D2, D11, OsDWARF4, BRD1, OsBRI1, and DLT, but up-regulate OsBZR1 as a feedback mechanism (Yamamuro et al., 2000;Hong et al., 2002Hong et al., , 2003Wang et al., 2002;He et al., 2005;Tanabe et al., 2005;Sakamoto et al., 2006;Tong et al., 2009). We analyzed the expression level of those genes in slg-D and found that all except BRD1 had the expected transcription level change as a response to the elevated BR levels (Fig. 5H).
As an alternative control, we also measured expression of those BR genes in the mutant m107, where D11 was dramatically enhanced and BR levels increased, and found similar expression changes (see Supplementary Fig. S9). Those results further confirm higher BR contents in slg-D. However, SLG itself did not respond to the exogenous BL treatment (Supplementary Fig. S10). Taken together, these results suggested that SLG positively regulates endogenous BR levels and is a new regulator of BR homeostasis in rice.

Suppression of SLG leads to BR-deficient phenotypes
To explore further the function of SLG, a SLG RNAi vector was constructed and introduced into WT plants. The SLG RNAi plants displayed a more compact architecture, reduced plant height, smaller leaf angle, and shorter and rounder grain (Fig. 6A-G). These phenotypes are similar to those of BR-deficient mutants, such as d61 and d11 (Hong et al., 2003;Tanabe et al., 2005). In addition, we investigated expression changes of the genes involved in BR synthesis or signaling in R7, a typical SLG RNAi line with greatly reduced leaf angle and SLG expression (Fig. 6F, G). Six of the genes detected, D2, D11, OsDWARF4, BRD1, OsBRI1, and DLT were up-regulated, but OsBZR1 was down-regulated in R7 compared with the WT (Fig. 6H). The feedback regulation of those BR-related genes caused by knock-down of SLG further supports involvement of SLG in regulating the BR level. These results further highlight a role for SLG in regulating BR homeostasis. This observation also suggests that an optimized expression level of SLG may help create a compact and semi-dwarf ideal plant type.

SLG functions as homomers
It has been reported that enzyme proteins often function as homomers or heteromers (Ali and Imperiali, 2005). To investigate the functional forms of SLG, the full-length SLG protein was used as a bait to screen a yeast two-hybrid library prepared from young rice panicles. We identified four positive clones that contain different SLG cDNA fragments from ~1 million yeast transformants. To confirm the self-interaction of SLG, different truncated SLG proteins were used for interaction analysis. As shown in Fig. 7A, a 30 amino acid region in the N-terminus of SLG (SLG∆C4), rather than the two conserved motifs, was required for the self-interaction of SLG. An in vitro GST pull-down assay also confirmed the self-interaction (Fig. 7B). In addition, BiFC analysis also showed that SLG physically interacted with itself and this interaction required the N-terminal 30 amino acid region (Fig. 7C).
To study the importance of SLG self-interaction, two truncated SLG CDS, SLG∆C1 with only the 190 N-terminal amino acids, and SLG∆N3 without the 30 N-terminal amino acids, were individually overexpressed in slg-D and the WT. We found that overexpression of SLG∆C1 but not SLG∆N3 in both the WT and slg-D resulted in shorter and rounder grains, smaller leaf angles, and dwarf phenotypes, similar to those of SLG RNAi plants ( Fig. 7; Supplementary Fig.  S11). The truncated SLG∆C1 protein might interfere with formation of functional homomers between the intact SLG proteins, thus leading to a dominant negative mutant phenotype. When the interaction region was removed in SLG∆N3, however, the truncated protein did not exert any effect on SLG, thus providing genetic evidence that SLG indeed functions as homomers in vivo.

Discussion
In this study, we have provided evidence that SLG is involved in BR homeostasis by positively regulating endogenous BR levels to control grain size and leaf angle in rice. First, the activation-tagging mutant slg-D and transgenic plants overexpressing SLG displayed longer and narrower grains and larger leaf angles that are similar to the mutants or transgenic plants with enhanced BR signaling or increased BR levels. Secondly, slg-D and the WT had similar sensitivity to BL treatment in lamina joint bending, coleoptile elongation, and root elongation assays. Thirdly, the BRZ treatment restored slg-D to the WT. Fourthly, overexpression of SLG in the BR-related mutants, d61-1 and d11-2, did not lead to slender grains and enlarged leaf angles. Fifthly, the major BRs were increased in slg-D. Sixthly, feedback regulation on expression of the known BR genes was seen in slg-D. Lastly, knockdown of SLG resembled mild BR-deficient mutants.
The size of an organ is determined by cell proliferation and cell expansion (Potter and Xu, 2001;Sugimoto-Shirasu and Roberts, 2003). Our results showed that SLG is required for cell expansion in grains. To investigate the possible regulatory relationship between SLG and other previously identified genes that control grain size by influencing cell expansion, such as PGL1, GL7, and GS2/GL2 (Heang and Sassa, 2012;Che et al., 2015;Duan et al., 2015;Hu et al., 2015;Wang et al., 2015), we examined the transcript level of these genes and found no obvious difference between slg-D and the WT (see Supplementary  Fig. S12). This result suggests that SLG may regulate grain size in a pathway independent of PGL1, GL7, and GS2/GL2. SLG is predicted to encode a BAHD acyltransferase-like protein. Previous studies of BAHD acyltransferase family members have shown that this family is capable of using CoA thioesters and catalyzing the formation of a wide variety of plant metabolites by generating ester or amide bonds (D'Auria, 2006). In Arabidopsis, two BAHD acyltransferases, BIA1 and BAT1, are involved in BR homeostasis, probably by conversion of active BR intermediates into inactive acylated BR conjugates (Roh et al., 2012;Choi et al., 2013). Overexpression of BIA1 or BAT1 results in decreased levels of active BRs and typical BR-deficient phenotypes (Roh et al., 2012;Choi et al., 2013). In our study, SLG, as a BAHD acyltransferase, probably converts an as yet unidentified substrate to the corresponding acyl conjugate to affect endogenous BR levels in an opposite way. Overexpression of SLG induced increased levels of active BRs and BR-overproduction phenotypes. The difference in Arabidopsis and rice implies that the function of BAHD acyltransferases in BR homeostasis has been differentiated. On the other hand, most of the enzymes known to catalyze BR biosynthesis belong to the cytochrome P450 protein family (Choe, 2006), implying that SLG, as a BAHD acyltransferase, may not work directly on the known BR intermediates, or that it may represent a different class of enzymes mediating BR synthesis. Further studies are needed to clarify how SLG participates in BR homeostasis.
In many cases, homomer formation is an essential biochemical process as it forms the complex quaternary structures of proteins to regulate selectivity against different substrates, enzyme activity, or stability (Dayhoff et al., 2010). Here, we showed that SLG interacted with itself, and its N-terminal 30 amino acid region was required for the interaction. It is likely that the self-interaction of SLG forms a functional enzyme complex with a special quaternary structure that binds the target substrates effectively. A truncated protein SLG∆C1 lacking a 255 amino acid C-terminus is still able to interact with the intact version but may form a complex unable to function properly due to a change in the quaternary structure, thus leading to a dominant negative phenotype. This finding also suggests that the substrate recognition and/or catalyzing domain may be located in the C-terminus. The version without the N-terminal interaction region failed to create a dominant negative phenotype, further confirming that the homomer formation of SLG indeed exists in vivo. It will be interesting to investigate further the number of SLG proteins required to form a functional enzyme complex and its structural organization.
The plant architecture determines planting density, and thus yield. In rice, BR-deficient or -insensitive mutants show the erect leaf phenotype, such as d2, d11, and d61 (Yamamuro et al., 2000;Hong et al., 2003;Tanabe et al., 2005). More erect leaves that increase light capture and thus enhance photosynthetic efficiency and nitrogen storage for grain filling can be combined with high planting densities to improve grain yield and biomass in rice (Sinclair and Sheehy, 1999;Morinaka et al., 2006;Sakamoto et al., 2006). For example, modulating the expression levels of OsDWARF4 and OsBRI1 led to the erect leaf phenotype and efficiently improves rice grain yield and biomass in dense planting conditions Sakamoto et al., 2006). SLG, when knocked-down by RNAi or interfered with by a truncated version, can create a compact semi-dwarf plant type with smaller leaf angles. Therefore, SLG can be used as an alternative to manipulate plant height for lodging resistance and leaf angle for planting density by optimizing its expression level, offering the potential for improving rice production.

Supplementary data
Supplementary data are available at JXB online. Figure S1. The slg-D mutation behaves in a semi-dominant manner. Figure S2. Co-segregation analysis of phenotypes and genotypes in F 2 progeny. Figure S3. Overexpression of Loc_Os08g44830 does not phenocopy the phenotypes of slg-D. Figure S4. Phylogenetic tree of SLG homologs. Figure S5. Alignment of the monocot group of SLG homologs. Figure S6. Sensitivities of roots and coleoptiles to BL are not altered in slg-D. Figure S7. Responses of WT and m107 leaf lamina joint angles to BRZ. Figure S8. Measurements of endogenous BR intermediates. Figure S9. Quantitative RT-PCR analysis of BR-related genes in young m107 and WT panicles. Figure S10. Quantitative RT-PCR analysis of SLG expression in WT seedlings treated with BL. Figure S11. Overexpression of SLG∆N3 does not change the phenotypes of the WT and slg-D. Figure S12. Quantitative RT-PCR analysis of several genes that control grain size by influencing cell expansion in slg-D and the WT. Table S1. Primers used in this study.