Polycomb Protein OsFIE2 Affects Plant Height and Grain Yield in Rice

Polycomb group (PcG) proteins have been shown to affect growth and development in plants. To further elucidate their role in these processes in rice, we isolated and characterized a rice mutant which exhibits dwarfism, reduced seed setting rate, defective floral organ, and small grains. Map-based cloning revealed that abnormal phenotypes were attributed to a mutation of the Fertilization Independent Endosperm 2 (OsFIE2) protein, which belongs to the PcG protein family. So we named the mutant as osfie2-1. Histological analysis revealed that the number of longitudinal cells in the internodes decreased in osfie2-1, and that lateral cell layer of the internodes was markedly thinner than wild-type. In addition, compared to wild-type, the number of large and small vascular bundles decreased in osfie2-1, as well as cell number and cell size in spikelet hulls. OsFIE2 is expressed in most tissues and the coded protein localizes in both nucleus and cytoplasm. Yeast two-hybrid and bimolecular fluorescence complementation assays demonstrated that OsFIE2 interacts with OsiEZ1 which encodes an enhancer of zeste protein previously identified as a histone methylation enzyme. RNA sequencing-based transcriptome profiling and qRT-PCR analysis revealed that some homeotic genes and genes involved in endosperm starch synthesis, cell division/expansion and hormone synthesis and signaling are differentially expressed between osfie2-1 and wild-type. In addition, the contents of IAA, GA3, ABA, JA and SA in osfie2-1 are significantly different from those in wild-type. Taken together, these results indicate that OsFIE2 plays an important role in the regulation of plant height and grain yield in rice.


Introduction
Rice (Oryza sativa) has been a hot spot in plant science research because it is considered a main staple food for more than half of the world's population. Its yield is mainly determined by grain weight, spike number and number of grains per panicle, but also can be affected by plant height and flowering time [1]. To date, numerous genes associated with spike number and number of grains per panicle, such as GN1A, TAW1, DEP1, OsSPL14, Ghd7, DTH8/ Ghd8 have been isolated and characterized [1][2][3][4]. On the other hand, grain weight is determined by grain size and ratio of grain filling. Recently, many grain size quantitative trait loci (QTLs) have been identified and characterized. GS3 and GS5 regulate grain size [5,6], GW2, GW5/qSW5 and GW8 control grain width [7][8][9], qGL3/qGL3.1, GL7 and GLW7 affect grain length [10][11][12][13]. These genes are involved mainly in signaling pathways mediated by proteasome degradation, phytohormones and G proteins to regulate cell proliferation and cell elongation [14]. Others such as GIF1 and TGW6 were considered to regulate the degree of grain filling [15,16]. Because of the complexity of the genetic mechanism underlying grain weight, regulation pathways have not yet been fully clarify consequently the identification and characterization of other genes related to grain weight will be useful to generating high yield cultivars.
Plant height is also a crucial trait for grain yield in modern agriculture [17]. Rice 'green revolution' had a positive impact in increasing the yield potential of rice, which has been represented by breeding of dwarf cultivars [18,19]. To date, a large number of dwarf mutants have been identified and characterized in rice, with most being related to the biosynthesis and responsiveness to phytohormones [20]. Dwarf mutants such as sd1, d18, d35, slr1, gid1 and gid2 are affected in the biosynthesis of or responsiveness to gibberellic acids (GAs) [21][22][23][24][25][26]. The genes D2, D11, BRD1, BRD2, OsDWARF4, RAVL1, BZR1, LIC1, BU1, TUD1, DLT, BAK1 and OsBRI1 are involved in biosynthesis or signaling pathways of brassinolide (BL) [27]. While other rice genes such as D10, D17/HTD1, D27, D3, D14/HTD2, D53 are implicated in the biosynthesis or signaling of strigolactones (SLs), a recently discovered group of plant regulators that control shoot branching [28,29]. Despite this, exploring for new genetic mechanisms controlling rice plant height is still research focus in rice genetics and genomics.
In this study, we isolated a rice OsFIE2 mutant, osfie2-1, which exhibited a dwarf phenotype as well as reduced seed setting rate, abnormal floral organs and small grains. We further demonstrate that OsFIE2 protein plays an important role in regulating plant height and grain yield.

Plant materials and field experiments
The dwarf and small grain mutant osfie2-1 was obtained from an EMS-induced mutant population of japonica rice cv. Zhonghua 11. The F 1 plants and F 2 populations derived from the cross between osfie2-1 and cv. Dular, reciprocal crosses between osfie2-1 and its wild-type were used for genetic analysis of the mutant gene. The F 2 generated from the cross between osfie2-1 and cv. Dular also used for gene fine-mapping. All rice plants were cultivated in paddy fields under natural conditions (China National Rice Research Institute, Hangzhou). A total of 12 plants for each genotype, osfie2-1 and wild-type, were used to evaluate phenotypic data, which included plant height, grain size (length, width and thickness), 1000-grain weight, seed-setting rate and panicle length.

Histological analyses
The second internodes of the osfie2-1 and wild-type were sampled and fixed in formalin-acetic acid-alcohol (FAA) overnight at 4°C and then dehydrated in a graded alcohol series (70%, 80%, 95%, 100%). The samples were embedded in a resin based on Technovit 7100 semi-thin section kit (http://www.emsdiasum.com/), followed by sectioning with an ultramicrotome (Leica, http://www.leica.com/). The 2-μm thick sections were stained with 0.1% toluidine blue and observed using a light microscope. For scanning electron microscopy analysis, spikelet hulls were fixed in a 2.5% glutaraldehyde solution for more than 2 h and then dehydrated in a graded alcohol series. The samples were critical-point-dried, mounted, gold-sputter-coated, and then observed and photographed under a KYKY-EM3200 scanning electron microscope.
Positional cloning of osfie2-1 A 32-plants F 2 progeny which displayed a similar osfie2-1 mutant phenotype derived from the crosses between osfie2-1 and Dular and were used for rough linkage analysis, and subsequently a F 2 progeny of 603 individuals was used for fine mapping. The positional cloning strategy was described by Zhang et al. (1994) [49]. PCR conditions for amplification of genomic DNA were as follows: 10μl reactions contained 25 ng of template DNA, 1.0 μl 10 X PCR buffer, 0.1 mM dNTPs, 0.1 μM of primer pairs and 0.1 U Taq DNA polymerase. The amplification protocol included an initial denaturation at 95°C for 3 min, followed by 35 cycles at 95°C for 30 s, 30 s annealing at 55°C, and 72°C for 30 s, and a final extension step at 72°C for 5 min. PCR products were separated on a 6% polyacrylamide gel, and silver-stained for visualization. The molecular markers including SSR and Indel markers used for mapping of osfie2-1 are listed in S2 Table. Vector construction and rice transformation A 1.1kb cDNA corresponding to the full-length ORF of OsFIE2 was PCR-amplified using two primers (p1300-OsFIE2F/R) and then inserted into the binary vector pCUbi1390 (Ubi promoter inserted into the pCAMBIA1390 vector). The resulting vector was introduced into the rice osfie2-1 mutant background via Agrobacterium-mediated transformation. The primers for vector constructions are described in S4 Table. The co-transformations with prey and bait were examined on control media, and the interactions between bait and prey were performed on selective media. The primers for vector constructions are described in S4 Table. Bimolecular fluorescence complementation Full-length OsFIE2 and OsiEZ1 were amplified by PCR, and inserted into the binary vectors pSPYNE and pSPYCE to obtain the OsFIE2-CY and OsiEZ1-NY constructions. Both OsFIE2-CY and OsiEZ1-NY were co-expressed in rice protoplasts. The yellow fluorescent signals were observed and photographed under laser confocal scanning microscope. The wavelengths for eYFP detection were 488 nm (excitation) and 527 nm (emission). The primers for vector constructions are described in S4 Table. RNA extraction and qRT-PCR analysis Total RNA from seedling stage and heading stage of various tissues was isolated using the Trizol method according to manufacturer's instructions (Invitrogen, www.invitrogen.com). 2μg of DNaseI-treated RNA was used for cDNA synthesis with First Strand cDNA Synthesis kit (Toyobo, http://www.bio-toyobo.cn/). Real time RT-PCR was performed using Thunderbird SYBR qPCR Mix (Toyobo, http://www.bio-toyobo.cn/) and LightCycler 480 Real-time PCR System (Roche, http://www.roche-applied-science.com/). The rice Ubiquitin gene (GenBank accession AF184280) was used as the internal control. Oligonucleotide primers are listed in S5 Table. RNA-sequencing analysis Total RNA was extracted using the Trizol reagent (Invitrogen) following the manufacturer's procedure. RNA quantity and purity were analyzed with the Bioanalyzer 2100 and RNA 6000 Nano LabChip Kit (Agilent, CA, USA) with RIN number >7.0. 10 ug of total RNA extracted from osfie2-1 and wild-type were used to isolate Poly-(A) mRNA using poly-T oligo-attached magnetic beads (Thermo-Fisher). Following purification, the mRNA was cleaved into small pieces using divalent cations and under elevated temperature and used to construct a final cDNA library according to the protocol for the Illumina RNA ligation method (Illumina, San Diego, USA). Briefly, purified RNA with the RNeasy MinElute Kit (Qiagen) was ligated with a pre-adenylated 3' adapter which enables the subsequent ligation of a 5' adapter. Based on the adapter sequence, a reverse transcription reaction followed by PCR was used to create the cDNA constructs. The average size of the paired-end libraries was 300 ± 50 bp. Single end sequencing was performed on an Illumina Hiseq2500 instrument at LC Sciences (Hangzhou, China) following the vendor's recommended protocol. Significant differentially expressed genes were identified considering a p value 0.05 and a log2 fold-change (log2_FC) ! 1. The functional information of these genes were carried out by using Gene Ontology (GO) analysis tool at http://www.geneontology.org/.

Hormone content analysis
Contents of IAA, GA3, ABA, JA and SA contents were measured in leaves of wild-type and osfie2-1 at flowering stage. Each sample was composed of at least three plants, and each measurement was repeated three times. The extractions and determination of plant hormones were performed using HPLC-MS as described by Trapp et al (2014) [50].

Results
The osfie2-1 mutant exhibited dwarf phenotype, small grains and floret defects The phenotype of the osfie2-1 mutant showed no obvious differences before the tillering stage (Fig 1A and 1B). osfie2-1 presented a markedly dwarf stem, reduced panicle length, narrow leaf, decreased number and size of grains, and suffered from seed sterility problem at maturity compared to wild-type (Fig 1C, 1D, 1J and 1Q; S1A Fig). The stem diameter, leaf length and leaf width were distinctly reduced in osfie2-1 compared to wild-type, resulting in an overall shorter plant with a final plant height of only about half of that of wild-type (Fig 1O). The decreased grain size of osfie2-1 was due to the reduction in grain length and grain width, which resulted in significantly lower 1000-grain weight than wild-type (Fig 1R-1T). Taken together, these abnormalities in osfie2-1 were responsible for a considerably low grain yield per plant, although the number of panicles per plant in osfie2-1 was not significantly different from wildtype (Fig 1P and 1V). Most of the osfie2-1 florets were normal however some florets presented abnormal floral morphology owing to the defective development of their palea and lemma. The set of developmental defects in mutant flowers, included elongated lodicules and altered numbers of stamens and pistils, an opened palea or lemma, formation of extra palea or lemma, and sometimes lemma malformation (Fig 1E-1I). Inspection of a hull section from osfie2-1 and wild-type showed details of these change (Fig 1K-1N). In addition, the heading stage of osfie2-1 was noted one week earlier than wild-type (data are not shown). The pollen was normal in osfie2-1 compared to wild-type (S2 Fig). Histological analysis of the internode and spikelet hull of osfie2-1 Compared to wild-type, the length of the panicles and internodes of osfie2-1 was significantly reduced (Fig 2A and S1B and S1C Fig). Cytology analysis revealed no obvious difference in longitudinal cell size between wild-type and osfie2-1 (Fig 2D, 2E, 2M and 2N), indicating that the dwarfism of osfie2-1 could be attributed to the decrease cell number. Moreover, the number of large vascular bundles (LVB) and small vascular bundles (SVB) also decreased in osfie2-1 ( Fig  2K-2I), and the lateral cell layer of osfie2-1 was markedly thinner than in wild-type, while no difference was observed in their lateral cell number (Fig 2B and 2C). These observations indicated that the thinner culm in osfie2-1 was due to a reduced VB number and cell layer. The spikelet hulls of osfie2-1 were much smaller than those of wild-type ( Fig 2F). The cell length and width in outer and inner spikelet hulls of osfie2-1 were significantly reduced in both cell size and cell number (Fig 2G-2J and 2O-2R).

Genetic analysis and map-based cloning of osfie2-1
The F 1 plants derived from the reciprocal cross between the osfie2-1 and its wild-type, between osfie2-1 and cv. Dular all showed normal phenotype, resembled the wild-type (S3 Fig). And the trait segregated ratio of 3:1 for the normal to mutant plants in each F 2 population (S1 Table), showing that the mutant trait of osfie2-1 is inherited as a monogenic recessive nuclear gene. The F 2 populations derived from the cross between osfie2-1 and Dular was subsequently used for fine gene mapping. The osfie2-1 gene was preliminarily located on the short arm of chromosome 8, between the markers M1 and M2. The advanced mapping localized osfie2-1 at an interval of 131-kb region between the markers M7 and M8, where it co-segregated with the markers M9 and M10. Within this region, 31 open reading frames (ORFs) were predicted, of which ORF11 (Os08g0137100, OsFIE2) and ORF14 (Os08g0137250, OsFIE1) have been related to plant height [47,48]. Sequencing analysis showed that the sequence of OsFIE1 no difference between wild type and mutant, while sequencing comparison of ORF11 between osfie2-1 and wild-type as well as with other eight varieties showed a single nucleotide change (C to T) in its first exon, resulting in a conversion of Leu (wild-type) into Phe (in osfie2-1) at the 40th amino acid (Fig 3A). The expression level of OsFIE2 was also obviously down regulated in osfie2-1 (S5 Fig). Thus, we suggest that OsFIE2 is the candidate gene for dwarf and small grain mutant osfie2-1. To confirm this, OsFIE2 was overexpressed in the osfie2-1 mutant. The resultant T 0 transgenic lines did not show any of the mutant phenotypes, including plant height, panicle length and grain size (Fig 3B-3G and S4 Fig), suggesting that OsFIE2 rescued the mutation.

Sub-cellular localization and expression profile of OsFIE2
When the OsFIE2:GFP fusion protein was transformed into rice protoplasts, the green fluorescence clearly appeared in both the nucleus and cytoplasm (Fig 4A and S6 Fig). We also evaluated the expression pattern of OsFIE2 by qRT-PCR. The ubiquitous expression of OsFIE2 was found in all the plant organs examined, including leaves, leaf sheaths, culms, roots, and panicles throughout vegetative and reproductive stages (Fig 4B and 4C). After pollination, the expression level of OsFIE2 gradually increased in the developing endosperm (Fig 4D).

OsFIE2 physically interacts with OsiEZ1
In Arabidopsis thaliana, FIE and E(Z) form the PRC2 complex core. E(Z) functions as a histone methylation enzyme. OsiEZ1 and OsCLF are two rice homologues for E(Z) [46]. To determine whether OsFIE2 interacts with either OsiEZ1 or OsCLF, we conducted protein-protein interaction assays. A yeast two-hybrid (Y2H) assay revealed that OsFIE2 physically interacts with OsiEZ1 but not with OsCLF ( Fig 5A). This result was further confirmed in vivo via a bimolecular fluorescence complementation (BiFC) experiment using the N and C termini of YFP to reconstitute a functional fluorescent protein (Fig 5B).

The osfie2-1 mutation compromised the expression of genes involved in various metabolic pathways
An RNA-sequencing-based transcriptome analysis was performed between wild-type and osfie2-1 to further clarify the function of OsFIE2. A total of 1,747 genes were differentially expressed, and among them, 1,078 genes were up-regulated in oosfie2-1, while the remaining 669 were down-regulated. The functional information of these genes was carried out by using Gene Ontology (GO) analysis (http://www.geneontology.org/). Considering their molecular function annotation, most of the differentially expressed genes were involved in catalytic and binding activity. At the same time, the majority of these genes were associated with many biological processes, including metabolic process, cellular process, and response to stimulus ( Fig  6A and S6 Table).
Further scrutiny of the differentially expressed genes revealed that the expression of a set of genes with known function were up-or down-regulated. These were homeobox MADS-box genes involved in floral development [51], FON1 that regulates floral organ number [52], Hd3a and RFT1 which are related to heading stage [53], OsHUS1 and OsDMC1 associated with meiosis [54,55], CycD5;1, OsRAN2 and OsEXP3 that affect cell cycle and cell expansion [56,57], OsSSI, OsAGPL1 and OsAGPL4 implicated in starch synthesis in the endosperm [48], the osfie2-1. Data are given as mean ± SD. Student's t-test was used to generate the P values; * and ** indicate P<0.05 and P<0.01, respectively. Scale bars: 5 cm (A, B); 10 cm (C, D); 0.25 cm (E-I); 0.5 cm (J); 0.05 cm (K-N).

Discussion
Polycomb proteins regulate extensive developmental processes in plants. In this study, we identify a dwarf and small grain mutant osfie2-1, and found that a fertilization independent endosperm protein OsFIE2 was responsible for the abnormal phenotype of osfie2-1 (Fig 3). We found that osfie2-1 mutant produced pleiotropic and effects on agronomic traits, exhibited markedly shorter and thinner culms, characterized by smaller panicles and gain size when compared to the wild-type (Fig 1A-1D and 1J), and also displayed abnormal floral organs ( Fig  1E-1I). Finally, osfie2-1 also presented a considerably low grain yield per plant (Fig 1V). In previous research, the characteristics of OsFIE2 gene knock down line by RNAi technique have been already reported. The RNAi lines also shown pleiotropic phenotypes in vegetative and reproductive organ generation, such as dwarf, abnormal enlarge lemma and small gain. In osfie2-1 mutant, only OsFIE2 was down regulated (S5 Fig), whereas, expression levels of both OsFIE1 and OsFIE2 were reduced in OsFIE2 RNAi lines [33]. So, osfie2-1 mutant is more helpful in the study of the function of OsFIE2 by eliminated the interference of the OsFIE1. Histological analysis found that that the longitudinal cell number in internodes decreased in osfie2-1, and the lateral cell layer of the internodes was markedly thinner than in wild-type, whereas there was no difference in the number of lateral cells between wild-type and osfie2-1. In addition, the number of LVB and SVB decreased in osfie2-1, as well as the cell number and size of spikelet hulls (Fig 2). All these observations may explain the dwarfism of these mutants and the smaller size of their grains.
In Arabidopsis, FIE protein regulates endosperm and embryo development and represses flowering during embryo and seedling development [67]. Moreover, FIE has also been shown to be essential for controlling shoot and leaf development [41]. Rice orthologs of FIE, OsFIE1 and OsFIE2, have also been identified. A recent study showed that mutation in OsFIE1 causes a dwarf stature and various floral defects [47]. It was established that OsFIE1 is regulated by DNA methylation and histone H3K9me2 and is involved in H3K27me3-mediated gene repression [47]. Another research group associated the function of OsFIE2 to endosperm development [48], however the mechanism of how OsFIE2 controls grain filling and yield in was not fully described nor was any other effect on plant growth. In our study, we distinguished and described multiple atypical morphologies in the a mutant of OsFIE2, osfie2-1, such as dwarfism, small grain size, narrow leaf and abnormal floral organs. OsFIE2 expresses in all tissues, which differs from OsFIE1 which is expressed only in the endosperm [46,47]. We also detect OsFIE2 localized in both the nucleus and cytoplasm (Fig 4A and S6 Fig), which agrees with a previous report [33]. In Arabidopsis, as two-core component of PRC2-like complexes, FIE and E(Z) parenchyma cell (PC) length and width, (n = 12). O-R, Cell length and cell width in outer and inner spikelet hulls, (n = 12). Data are given as mean ± SD. Student's t-test was used to generate the P values; * and ** indicate P<0.05 and P<0.01, respectively. Scale bars: 15 cm (A); 0.05 mm (B-E); 2 mm (F); 40 μm (G-J).
doi:10.1371/journal.pone.0164748.g002 proteins physically interact [41,68]. In our case, the physical interaction between OsFIE2 and OsiEZ1 (the rice E(Z) ortholog) was also demonstrated (Fig 5). Thus, we postulate that OsFIE2 may function through PRC2-like complexes to control growth and development by posttranslational modifications similarly to what was described in Arabidopsis.
Down-regulation of the Arabidopsis FIE gene produced dramatic morphological aberrations resulting from de-repression of KNOTTED-like homeobox and MADS-box genes [41]. Significantly reduced expression levels of OsFIE2 lead to pleiotropic aberrant phenotypes in OsFIE2 osfie2-1 and OsFIE2 overexperssion transgenic lines, their panicles and grains. E, Relative expression level of OsFIE2 in wild-type, osfie2-1 and OsFIE2 overexperssion transgenic lines. F and G, Grain length and 1000-grain weight in wild type, osfie2-1 and OsFIE2 overexperssion transgenic lines. Data are given as mean ± SD. Scale bars: 25 cm (B); 3.5 cm (C); 2.5 mm (D).
doi:10.1371/journal.pone.0164748.g003  RNAi lines and ectopic expression of some key development regulators, such as the MADS-box (OsMADS3) and KNOX (Oskn3) genes [33]. Morphological changes were also evident in osfie2-1, thus we performed an RNA-sequencing-based transcriptome analysis and found that there are many genes up or down-regulated in osfie2-1 (Fig 6). Among them, the expression levels of homeobox genes (MADS-box and Oskn3) also changed significantly in osfie2-1 suggesting a probable cause for the abnormalities in floral and seed morphology as well as plant height. Moreover, the different expression levels of cell division/expansion-related genes in osfie2-1 compared to wild-type may explain the reduction in the cell number and cell size in the internodes and spikelet hulls in the osfie2-1 mutant. In addition, there were several hormone-related genes differently expressed between the wild-type and osfie2-1, which may lead to a different content of ABA, IAA, GA 3 , SA, and JA in osfie2-1 compared to wild-type. In fact, it has been established that PcG proteins can target genes involved in the biosynthesis, transport, perception and signaling of phytohormones in Arabidopsis, specially those implicated in promoting growth [69,70]. In this regard, an Arabidopsis PcG gene, CLF, was shown to control both cell division and elongation during leaf expansion [71]. Because of OsFIE2 function through PRC2-like complexes to control growth and development by posttranslational modifications [48], some the up-regulated genes may be directly controlled by OsFIE2 via epigenetic repressive marks, however most differential expression genes may be indirectly regulated by OsFIE2. Of course all of these need further experiments to prove. In summary, our data demonstrates that OsFIE2 plays a role in determining plant height and grain yield in rice and support the idea that PcG genes such as OsFIE2, OsFIE1 and OsiEZ1 are essential for growth and development.