The Bsister MADS Gene FST Determines Ovule Patterning and Development of the Zygotic Embryo and Endosperm

Many homeotic MADS-box genes have been identified as controllers of the floral transition and floral development. However, information regarding Bsister (Bs)-function genes in monocots is still limited. Here, we describe the functional characterization of a Bs-group MADS-box gene FEMALE-STERILE (FST), whose frame-shift mutation (fst) results in abnormal ovules and the complete abortion of zygotic embryos and endosperms in rice. Anatomical analysis showed that the defective development in the fst mutant exclusively occurred in sporophytic tissues including integuments, fertilized proembryos and endosperms. Analyses of the spatio-temporal expression pattern revealed that the prominent FST gene products accumulated in the inner integument, nucellar cell of the micropylar side, apical and base of the proembryos and free endosperm nuclei. Microarray and gene ontology analysis unraveled substantial changes in the expression level of many genes in the fst mutant ovules and seeds, with a subset of genes involved in several developmental and hormonal pathways appearing to be down-regulated. Using both forward and reverse genetics approaches, we demonstrated that rice FST plays indispensable roles and multiple functions during ovule and early seed development. These findings support a novel function for the Bs-group MADS-box genes in plants.


Introduction
Sexual reproduction in higher plants includes a key phase to produce male and female gametes, ensuring pollination and fertilization. The female gametophyte is essential for the sexual reproduction of plants [1][2][3]. During the last few years, significant progress has been made in determining the molecular components that control ovule identity, embryo sac polarity, gametophytic cell specification, female gametic cell fate determination, embryogenesis and endosperm development [4][5][6][7][8][9][10][11].
Members of the homeotic MADS-box genes encode a family of transcription factors that fulfill the important functions of regulating vegetative growth and flowering time, controlling meristem and floral organ identity, and determining fruit and seed development [12][13]. Many MADS-box genes have been identified to constitute an intricate network controlling the orchestration of the floral transition and floral development [14][15][16][17][18][19][20][21][22][23][24][25]. Important key developmental biology questions that remain unanswered include: how is the pattern of formation accomplished, and how does the genetic interaction of floral homeotic genes occurs at the molecular level? Although extensive knowledge on these MADS domain transcription factors that regulate the floral transition and floral organ development is available, little is known regarding the molecular mechanisms they employ to act as the developmental switches for specifying the female reproductive unit in flowering plants. In addition, it is unclear how the homeotic transcription factors organize the spatial patterns of cell differentiation during diverse or specific developmental processes such as embryonic shoot/root initiation and endosperm formation in a developing seed.
The ABC/DE model [13,[18][19][20] of floral development describes the genetic interaction of the five major classes of floral selector genes, and each class determines the identity of different floral organs: sepals, petals, stamens and carpels. The B s -group MADS-box genes are close relatives of the B class of the MADSbox gene, a family only described in a few plant species [24,[26][27][28][29][30][31][32][33]. In eudicot plants, the B s -group MADS-box genes (e.g. FBP24 in petunia and ABS, STK and GOA in Arabidopsis) are essential for regulating integument, seed coat, fertilization and endosperm development, and fruit growth [28][29][30][31][32]. However, information on B s -function genes (e.g. rice MADS29, MADS30 and MADS31) in monocots still remains limited [24,27,33]. In previous studies, the FEMALE-STERILE (FST) gene, a B s -group MADS-box transcription factor, was reported to putatively control the identity of ovule, pericarp and seed development, as well as the translocation of reserve nutrients during the entire reproductive stage [34,35]. In addition, the fst phenotype is genetically sporophytic and controlled by a single recessive locus in the nucleus [34]. OsMADS29 (whose ORF is identical to that of FST) plays an important role in seed germination [24] and affects the degradation of the nucellus and nucellar projections during the development of a rice seed [33] although this function appears to be restricted to the post-fertilization stage.
In this study, we report that FST (DQ004266) acts as positional determinants regulating chalaza formation, integument morphogenesis and the early development of the zygotic embryo and endosperm. FST is expressed in reproductive organs throughout all developmental processes. Importantly, the spatio-temporal expression pattern revealed the stepwise formation of a prominent FST gene product. This product accumulated in sporophytic organs/tissues in the apical-basal region of ovules and proembryos, as well as the center of endosperm nuclei tissue, pericarp, and seed coat. Microarray and gene ontology (GO) analyses unraveled substantial changes in the expression level of many genes in fst mutant ovules and seeds, with a subset of genes involved in developmental and hormonal pathways appearing to be down-regulated. This finding supports a novel function for the B s -group MADS-box genes in plants. The discovery of FST offers novel insights into developmental biology processes and contributes to a better understanding of the mechanisms regulating the female reproductive unit and seed development mediated by floral homeotic genes in flowering plants.

The fst Mutation Leads to Sporophytic Female Sterility
The spontaneous fst mutant plants exhibited similar morphology to that of the wild-type (WT) plants but are completely female sterile. The size, shape and structural composition of floral organs in the fst mutant were identical to those of the WT (Figure S1A-S1D). The fst mutation did not affect anther development and pollen viability (Figure S1B, S1E & S1F). Pollen from both the fst mutant and WT germinated normally in the fst mutant style (Figure S1E & S1F), but unlike WT ( Figure 1A-1C) the fst mutant plants did not produce any visible embryo and endosperm when pollinated with their own pollen or with WT pollen ( Figure 1D-1F). All the megagametophytes derived from heterozygous FST/fst (+/-) plants were completely fertile. The ratio between fertile and sterile plants was 3:1 in progeny derived from the heterozygous FST/fst (+/-) plants (n = 2036). These results confirmed that fst is responsible for sporophytic female-specific sterility in rice, which is completely different from other reported female-sterile mutations in rice that are controlled by quantitative traits or two recessive genes [36][37].

The fst Mutation Disrupts Development of the Ovule, Zygotic Embryo and Endosperm
To dissect the cyto-embryological mechanism of sterility in the fst mutant, morphological and anatomical analyses were conducted with hundreds of flowers sampled at different stages, ranging from the panicle and floral initiation to ripening during pre-and post-fertilization. Integument development in the fst mutant was severely affected and consequently the transformed integuments formed additional chalaza-like ( Figure  We investigated megasporogenesis and megagametogenesis (data not shown), and did not observe any distinct abnormalities in these developmental processes in the fst mutant. After cellularization, however, the shape of the fst embryo sac was altered and the antipodal tissues did not mature and position correctly ( Figure 2B-2D) when compared to those of the WT (Figure 2A).
To examine whether the observed morphological defects in ovule development influence fertilization, we examined mature fst ovules with and without pollination. Notably, the defective development of integuments and micropyle did not affect fertilizations when fst ovules were pollinated with either their own pollen or the WT's. The early development of both the zygotic embryo and endosperm did not show distinct differences when compared to that of the WT 1-3 days after pollination (DAP) ( Figure 3A, 3B & 3E). Nevertheless, the proembryo and endosperm nuclei tissue were arrested at 3-5 DAP and degenerated tissues were detected 5-7 DAP in the fst ovules ( Figure 3C, 3D & 3E). In contrast, no embryogenesis and endosperm development events were observed in the mutant ovules without pollination (data not shown). Consequently, both morphological and functional defects in the fst mutant eventually caused complete failure of embryo and endosperm development after double fertilization. We mapped the fst locus using an F 2 population derived from a hybrid between the rice cultivar Samgangbyeo and the fst mutant (spontaneous mutation identified from the rice cultivar Junambyeo; Figure S2A).
Sequence comparison revealed that the genomic DNA of the WT allele FST was approximately 4 kb long with 8 exons and 7 introns. Exon 2 carries the ATG initiation codon and contains a highly conserved MADS domain ( Figure S2B). The mutant gene fst produced a non-functional transcript due to an 8-bp nucleotide deletion, causing a translational frame shift in the MADS-box domain at exon 2 ( Figure S2C). Homology analysis of the MADS domain protein revealed that the FST protein is closely related to FBP24 in petunia and ABS in Arabidopsis in the B s -group MADSbox genes [26][27] (Figure S2D & S2E).
To verify the function of the fst gene, we performed complementation analysis. Two binary plasmids, carrying either a 1.3-kb full-length cDNA of the WT FST and a 2.6-kb upstream region (pFST::FSTc) or a 6.9-kb WT genomic DNA fragment containing the entire ORF plus a 2.6-kb upstream region (pFST::FSTg), were incorporated into the genome of the fst mutant by an Agrobacterium-mediated transgenic approach. Both transgenic plants carrying the individual constructs were able to restore spikelet fertility. Noticeably, pFST::FSTg fully restored both embryo and endosperm development ( Figure 4D). Most seeds derived from T 0 and T 1 transgenic plants germinated with normal shoots and roots ( Figure 4E). However, in some of seed recovered by pFST::FSTc, the shoot and root germinated on the inner side of the endosperm, causing a reversed hypocotyl axis ( Figure S1I & 1J). The T 0 and T 1 plants recovered by both pFST::FSTg and pFST::FSTc showed a high frequency of seed set (.85%) ( Figure  4D).
To confirm the function of FST, we generated 61 independent transgenic lines with down-regulated FST expression by using RNA interference targeting (RNAi) constructs (35S::RNAi-FST) in the Nipponbare and Liyu-B backgrounds. Compared to the blank vector transgenic plants, 35S::RNAi-FST repression plants primarily displayed partial female sterility, defective endosperm development, incomplete grain filling, reduced germination, and increased of total protein and fat contents in rice grain ( Figure 4A-4H). In addition, we also created transgenic plants over-expressing FST driven by the 35S promoter in Nipponbare and Liyu-B backgrounds. Compared to WT plants, the fst and FST-RNAi plants flowered 8.1 days and 2.3 days earlier, whereas the FST over-expressing plants flowered about 43.6 days later ( Figure 4F). The phenotypes resulting from the directed disruption of FST expression were similar to the defective developmental phenotypes observed in the fst mutant.

Abundant FST Gene Products Accumulate in the Apical-Basal Region of Reproductive Units
A fusion protein of FST-GFP driven by the CaMV-35S promoter was expressed exclusively in nuclei of onion cells ( Figure  S3B). To determine the cellular function of FST, we investigated the expression patterns during ovule and seed development using a b-glucuronidase (GUS) reporter transgene (pFST::GUS). Generally, the pFST::GUS construct was expressed in reproductive structures throughout all developmental processes ( Figure 5A, 5C-5F & 5M-5O). From the panicle initiation to seed development stages, GUS activity was detected throughout emerging floral primordia and was later confined to the sexual reproductive organs inside the ovary, stamen primordium, tips of lemma, palea, glumes, anther wall, grain apiculus, seed coat and shoot apex of the embryo ( Figure 5, Figure S3). In addition, in early pistil development, GUS accumulated at the base of the stigma ( Figure 5C & 5D). GUS expression was also observed in the nucellus and inner integument ( Figure 5E & 5F). Consistently, both the Northern blot and RT- PCR analyses in samples ranging from seeding to ripening showed no expression of FST in vegetative organs such as roots, leaf blades, leaf sheaths, node and internodes of stems. But transcrip-tion was detected in reproductive organs such as flowers and seeds during inflorescence and spikelet development ( Figure S3F & 3G).
We examined the expression patterns of the FST gene in WT plants using in situ hybridization. Consistently, the FST transcripts were detected in pistil primordia ( Figure 5B), nucellar cells on the micropylar side of the mature ovary ( Figure 5G), proembryos 3 DAP ( Figure 5H) and free endosperm nuclei 3 DAP ( Figure 5I). Interestingly, strong hybridization signals were observed in both the apical and basal parts of the proembryos 3 ( Figure 5H) to 4 DAP ( Figure 5J), respectively, but these signals were detected exclusively in the shoot apex and coleorhizae in embryos 5 ( Figure  5K) and 7 DAP ( Figure 5L).

Comparative Transcriptome Analysis of WT and fst Mutant Ovules
We compared the expression profiles of the fst mutant and WT at different stages of ovule development using microarray analysis (Affymetrix Rice GeneChip). This allowed us to assess the role of FST in integument morphogenesis, embryogenesis and endosperm development. Three sample replicates were included for both the fst mutant and WT to guarantee the reproducibility of microarray analysis.
Before flowering (BF), 59 genes showed at least 4-fold expression level changes in fst panicles compared to that of the WT (Table S1) and the 23 main down-regulated genes are displayed at Table 1. GO classification of the down-regulated genes revealed that most genes played important roles in auxin efflux and polarity, gibberellin acid, localization, flower and ovule development, embryonic development, signal transduction, nutrient reservoir, programmed cell death (PCD) and regulation of transcription.
Three to five DAP, 406 genes showed at least 4-fold expression changes; moreover, 171 genes displayed at least 10-fold downregulation in fst panicles compared to that of the WT (Table S2). The main representative genes are displayed in Table 2. Genes were functionally grouped into categories involved in auxin efflux and polarity, cell differentiation, embryonic development, seed maturation and endosperm development, morphogenesis, cell-tocell signaling, signal transduction, nutrient reservoir, starch biosynthesis, sugar and glucose transporter, protein and lipid transport and PCD.

A Novel Function Identified in B s MADS-box Genes in Plants
Our results demonstrated that the complete female-specific sterility (fst) identified in a japonica rice cultivar Junambyeo was caused by a spontaneous mutation in a B s MADS-box gene, although this fst mutation does not affect male gamete development and viability (Figure S1A-S1F). Nevertheless, the mutant phenotype of female sterility only occurred in homozygous (-/-) plant, while the female gamete carrying fst mutant allele derived from F 1 heterozygous (+/-) plant was completely fertile, suggesting that the fst mutant phenotype of female sterility depended on the sporophytic, instead of gametophytic genotypes. Consistently, morphological and anatomical analyses revealed that, in the fst mutant plants, defective developments exclusively occurred in Our data demonstrated that the loss-of-function in the B s -group MADS-box gene FST influences the patterning and growth of ovules before fertilization but also disrupts early development of the zygotic proembryo and endosperm after the fertilization, leading to subsequent sterility. The severe defect in integument development and transformed ovule shape observed in the fst mutant is similar to that observed in the eudicot loss-of-function mutants [28,29,30,32]. Nevertheless, the failure in proembryo development and differentiation leading to aborted seed develop-ment is unique in rice fst mutant plants. To date, the B s -group MADS-box genes were known for regulating integument, seed coat, fertilization, endosperm development, fruit growth, seed germination and degradation of the nucellus and the nucellar projection during rice seed development [24,[26][27][28][29][30][31][32][33]. Undoubtedly, rice FST plays indispensable roles and multiple functions during ovule and early seed development.

FST Acts as a Negative Regulator of Flowering Time
Our data based on phenotypic analyses demonstrated that the flowering time in plants of the fst mutant (loss-of-function in FST) and suppressed FST (FST-RNAi) was significantly earlier than that of WT (FST) and over-expressed FST. Microarray analysis revealed that expression of FIE2 (AK242200) and OsMADS22 (AK070121) was prominently reduced in the fst mutant compared to the WT (Table 2). FIE2, a polycomb group gene, and OsMADS22, a SVP group MADS-box gene, are well-known negative regulators of flowering time [41][42]. Thus, we conclude that FST may act as a direct or indirect negative regulator of flowering time.

FST Is a Key Regulator Required for Female Organ/Tissue and Seed Development
Our in situ hybridization data showed strong expression of FST at the base of the stigma, nucellus and inner integument at early stage of pistil development ( Figure 5A-5F). Consistently, morphological and anatomical analyses provided evidence that fst mutant plants displayed transformed shape and defective development at the chalaza and integument (Figure 2). This suggests that FST may play a pivotal role in regulating development of pistil components. Interestingly, high expression level of FST was also detected in apical and basal parts of the 3-4 DAP proembryos, and in the shoot apex and coleorhizae 5-7 DAP embryos ( Figure 5H-5L). We therefore postulate the arrest of proembryo development observed in fst mutant plants probably resulted from the loss of correctly coding FST protein. In addition, we found that FST was expressed in endosperm nuclei tissue and seed coat at an early stage of seed development ( Figure 5M-5O). These findings imply that FST may participate in regulating endosperm development and nutrient metabolism.
Microarray analyses allowed us to evaluate the impact of fst mutation on gene expression, where the level of a large number of genes was significantly altered (Table 1 and 2). Consistent with the results from molecular and morphological analyses, some changes of gene expression appeared to be associated with the developmental defects visualized in fst mutant and FST-RNAi repression lines. For instance, remarkably down-regulated expression of genes observed in auxin efflux, GA signaling, or other cellular processes, such as PCD, polarity and cell fate determination [38][39][40][47][48][49][50], may influence chalaza positioning and integument formation, and subsequently result in defective zygotic embryo and endosperm development.
Based on the observations and analyses, we propose a model to explain how the FST works throughout pistil and seed development ( Figure 5P-5R) as follows. The expression of FST at the base of stigma may define the chalaza formation at correct position by the induction of a signal factor (SF) activated via signal transduction. Failure of this function can cause abnormal shape and positioning of the chalaza in the fst mutant. Likewise, the expression of FST in the nucellus and inner integument may lead to the development of the nucellus and differentiation of the innerouter integument by the induction of a SF ( Figure 5P2 & 5P3). The high expression of FST activates the expression of a subset of genes involved in developmental and hormone signaling pathways that induce cell division in the top section of the proembryo ( Figure 5Q1), thus constantly influence morphogenesis and axis determination ( Figure 5Q2). In contrast, a low expression level of FST may suppress cell division in this process. During differentiation of the primordium into the shoot and root, a high level of FST induces differentiation of the shoot apex and determine the axis of shoot and root growth during germination ( Figure 5Q3 &  5Q4), respectively. In addition, FST may guide the development of ENT (endosperm nuclei tissues) predating ENC (primary cellular endosperm cells) formation and accelerate the building-up of the nutrient reservoir and starch biosynthesis [43][44][45][46]51] in the endosperm by the induction of a SF. The phenotype of suppressed FST and OsMADS29 [33] transgenic plants strongly supports this hypothesis.
In conclusion, we demonstrated the multifunctional roles of the B s -group MADS-box gene FST in rice. Our study provided a complete overview of biological, genetic, and molecular mechanisms, as well as expression profiling data underlying the ovule and seed development. These findings will likely lead to a better understanding of the evolution and molecular mechanisms of reproductive processes in higher plants.

Mutant Plant History
In 1999, we discovered the spontaneous fst mutant G39 (O. sativa L. ssp. japonica) from Junambyeo, a newly developed rice breeding line in the experimental field of Yeungnam University at Gyeongsan (35.9uN, 128.6uE, 58 masl.) in Korea.

Experimental Materials
An F 2 mapping population was derived from a cross between Samgangbyeo (O. sativa L. ssp. indica) and G39, which contained fst. The fst allele was introduced into Chinese Dongxiang wild rice (O. rufipogon Griff.) and several elite varieties including Junambyeo, Nan34 and Ansanbyeo (O. sativa L. ssp. japonica) to study the morphology, cytology, inheritance and molecular mechanisms through a backcrossing method.

Genetic Mapping and Cloning of FST
FST was mapped with SSR and STS markers using 1286 F 2 plants. The candidate gene was amplified using gene-specific primers in both fst and WT plants (Table S3). The linkage map was constructed with the Map Manager program QTXb17 [52]. ORF search and homology analysis of nucleotide and amino acid sequences were performed using NCBI databases (http://www. ncbi.nlm.nih.gov). The sequence was aligned using MEGA 5.0 [53].

RNA Isolation and Expression Analysis
Total RNA was prepared using TRI Reagent (Ambion:http:// www.ambion. com). First-strand cDNA was synthesized from 2 mg of total RNA. RT-PCR was performed using gene-specific primers (Table S3), 25-30 reaction cycles and three biological replicates for each reaction.

Microscopic Observations
Pollen fertility. Anthers of the fst mutant and its WT were collected from 30 spikelets from ten plants (three panicles each) at the flowering stage. Pollen grains from anthers were suspended in a 1% potassium iodide solution (I 2 /KI). Pollen fertility was calculated by determining the percentage of normal pollen grains against total pollen grains per spikelet.
Pollination process. Artificial pollination was carried out one day before flowering using previously emasculated mature flowers. Around 3000-3500 perfect flowers from both fst and FST were emasculated prior to anthesis and enclosed with paper bags. Pollen grains from fst were pollinated on its own stigmas and on WT stigmas. Approximately 150-200 pistils (3 replications of 50-70 pistils each) were sampled every 10 min for up to 60 min between 1 and 10 DAP. The pistils were placed in fixative FAA (80% ethanol: 37% formaldehyde: 100% acetic acid, in a ratio of 8:1:1) for 24 h, rinsed in distilled water for 4 h, softened for 1 h at 60uC in 1N NaOH and rinsed again for 4 h in distilled water. The pistils were then stained in 0.1% water-soluble aniline blue for 10 min. Samples were immediately prepared in a droplet. For each Table 2. Genes involved in seed development with expression changes of at least 2-fold (fst/ WT) in fst seeds and ovaries 5 DAP. genotype, the ovary was separated from the base of the style during the sample preparation procedure. The samples were then covered with cover slips and squashed gently. The ovules was observed under a microscope (Olympus BX51, UV filter set) and photographed. Ovary development. To analyze ovary development of both the fst mutant and its WT, several hundreds of flowers were prepared at different stages ranging from panicle and floral organ initiation to ripening during pre-and post-fertilization. The lemma, palea and anthers were dissected out from the flowers in order to reach the ovaries. The ovaries were immersed in glutaraldehyde fixative solution, which contained 1.4% glutaraldehyde, 2% paraformaldehyde and 50 mM PIPES (pH 7.2), at 4uC overnight as described previously [54]. After rinsing again in PIPES buffer, the samples were dehydrated in an ethanol series of 10 to 100% and then embedded in paraffin (Paraplast Plus, Sigma). The paraffin-embedded ovary samples were further sliced into 4 mm sections with a microtome (Leica DMR) and stained with 0.05% toluidine blue containing 0.1% sodium carbonate or propidium iodide (PI) (5 mm/ml). The tissue sections were observed under a light microscope or in fluorescent mode (Leica DM 2500, fluorescent set).
Scanning electron microscopy (KYKY-EM3200) of whole mature ovules of both fst and FST was performed as described previously [55]. Development of ovules and embryo sacs of the transgenic plants were detected with mature ovaries; 20 ovaries were prepared from each of the 50 RNAi plants sampled. To investigate starch formation and endosperm shape in the mutant and RNAi lines, spikelets at 10 and 20 DAP were sampled. The endosperms were cut with a sharp knife and stained with 1% potassium iodine solution (I 2 /KI).
The morphological terminologies of rice organs or tissues are described in Table S4. Statistical analysis was performed by Statistix for Windows version 2.0 by Analytical Software (http:// www.statistix.com).

Microarray Analysis
Three independent biological replicates of the fst mutant and WT panicle mRNA at different stages of ovule development were used for microarray experiments. Pools of panicles were used to evaluate genes predominantly expressed in the ovary and seed. For microarray analysis, the standard protocol of the Affymetrix GeneChip service was used when setting up the experiment (CapitalBio, http://www.capitalbio.com). To identify differentially expressed genes, the signal ratio of each gene between the WT and the mutant was calculated. The array data sets were named after the genotype (W, WT; M, fst or cynosure, a previous designation) and ovule stage pool (BF, before flowering stage; 5D, 5 DAP stage). The definition and function of the genes were referenced using SWISS-PROT (by BLASTX) and GO classification (GeneBank) in the KOMME (http://cdna01.dna.affrc.go.jp/ cDNA/) and Rice Genome Browser (http://rice.plantbiology. msu.edu/cgi-bin/ gbrowse/rice/) databases.

Northern Blot Analysis and In Situ Hybridization
For Northern blot analysis, 15 mg of purified total RNA from tissues of vegetative and reproductive organs was run on a 1.2% agarose gel containing formaldehyde and transferred onto a Hybond-N+ membrane (Amersham: http://www. gelifesciences.com /Amersham). The gene-specific probe was amplified using the primers FSTRT-F and FSTRT-R (Table S3) to amplify a region of ,200 bp on the 3' end of the ORF of FST. PCR fragments were inserted into the pGEM-T easy vector (Promega: http:/www. promega.com) and transcribed in vitro by either a T7 or SP6 promoter for sense or antisense strand synthesis using the Digoxigenin RNA labeling kit (Roche: http://www.roche-applied-science.com). The blot was performed as described previously [56].
For in situ hybridization, hybridization and immunological detection were performed as described by Jackson [56] using the same probe generated for the Northern blot analysis.

Binary Vector Constructs
To determine the complementation of the FST phenotype to an 8 bp deletion, pFST::FSTg (harboring a 6.9-kb genomic DNA fragment containing the entire FST coding region, a 2.6 kb promoter region and a 0.7 kb 3' region) and pFST::FSTc (containing a 2.6 kb promoter region and cDNA sequence of FST) were constructed using the binary vector pCAMBIA1300. To construct the FST RNAi vector (35SX2::RNAi-FST), an intron fragment containing 155 bp was used as a linker between a 396 bp gene-specific fragment in the antisense and sense orientations; these reconstructed fragments were inserted into the pHB binary vector containing a double 35S promoter. A transgenic 35S::FST over-expressing line of plants was also generated. To study promoter activity, a 2.6 kb genomic DNA fragment upstream of the FST coding region was fused to the GUS reporter gene with the nopaline synthase terminator and cloned into the binary vector pCAMBIA1305 to generate the pFST::GUS plasmid.

Plant Transformation
The binary plasmids were introduced into Agrobacterium tumefaciens EHA105 and the calli induced from the anther of the fst mutant plants were transformed. The binary vectors were also transformed into rice calli, which were induced from mature WT embryos of Nipponbare and Liyu B (O. sativa L.). Transgenic plants were selected by hygromycin resistance and subsequently transferred to soil. In total, 17, 12, 61 and 30 independent transgenic lines were obtained for the pFST::FSTc, pFST::FSTg, 35SX2::RNAi-FST and pFST::GUS constructs, respectively. All transgenic materials were assayed in the T 0 and/or T 1 , T 2 generations using 10 to 30 independent or sibling plants.
For the nuclear localization analysis, the binary plasmid 35S::FST-GFP was transformed into onion epidermal cells. The expression of the fusion protein was observed with a fluorescent microscope.

Biological Trait and Physiochemical Property Analysis
For all samples of the mutant, WT and transgenic lines, the degree of grain filling was measured by 100 seed-weight from 15 individual plants (100 seeds each) while the WT was used as the control. In addition, the germination ability was evaluated using 30 individual plants (100 seeds each) and the seeds were germinated for 3 days at 28uC. The total protein and fat contents of grain were detected by using NIR spectra (Bruker FT-NIR, Vector 22/N-I) with three biological replicates from three independent RNAi lines (three plants per line) and WT plants.

Promoter Activity Detection
For the GUS assay, transgenic plants were harvested from various tissues at different developmental stages and fixed in a solution of 2% paraformaldehyde, 1 mM EDTA, 100 mM sodium phosphate buffer pH 7.0. GUS activity was analyzed by staining overnight at 37uC in a staining solution (0.5% Triton X-100, 2 mM X-GluA, 50 mM sodium phosphate buffer, pH 7.0).

Sequencing Data
Sequence data from this article can be found in the NCBI/ GenBank data libraries under the accession number DQ004266. The microarray datasets have been deposited in GEO database with the accession number GSE33441.