FGW1,a protein containing DUF630 and DUF632 domains,regulates grain size and filling in Oryza sativa L.

Yngyng Li,Peilong He,Xiowen Wng,Hongyn Chen,Jile Ni,Weijing Tin,Xiobo Zhng,Zhibo Cui,Gunghu He,,Xinchun Sng,

a Rice Research Institute,Key Laboratory of Application and Safety Control of Genetically Modified Crops,Academy of Agricultural Sciences,Southwest University,Chongqing 400715,China

b Guizhou University of Traditional Chinese Medicine School of Pharmacy,Guiyang 550025,Guizhou,China

Keywords: Seed size Grain filling DUF630/DUF632 Starch synthesis Rice

ABSTRACT Grain filling influences grain size and quality in cereal crops.The molecular mechanisms that regulate grain endosperm development remain elusive.In this study,we characterized a filling-defective and grain width mutant,fgw1,whose mutation increased rice seed width mainly via cell division and expansion in grains.Sucrose contents were higher but starch contents lower in the fgw1 mutant during the grainfilling stage,resulting in inferior endosperm of opaque,white appearance with loosely packed starch granules.Map-based cloning revealed that FGW1 encoded a protein containing DUF630/DUF632 domains,localized in the plasma membrane with preferential expression in the panicle.RNA interference in FGW1 resulted in increased grain width and weight,whereas overexpression of FGW1 led to slightly narrower kernels and better grain filling.In a yeast two-hybrid assay,FGW1 interacted directly with the 14-3-3 protein GF14f,bimolecular fluorescence complementation verified that the site of interaction was the membrane,and the mutated FGW1 protein failed to interact with GF14f.The expression of GF14f was down-regulated in fgw1,and the activities of AGPase,StSase,and SuSase in the endosperm of fgw1 increased similarly to those of a reported GF14f-RNAi.Transcriptome analysis indicated that FGW1 also regulates cellular processes and carbohydrate metabolism.Thus,FGW1 regulated grain formation via the GF14f pathway.

1.Introduction

Effective panicle number,grain number per panicle,and 1000-grain weight are the main phenotypic traits that contribute to rice yield,the main trait under selection in rice breeding[1].Grain size,determined by grain length,width,and thickness,not only affects grain weight but determines grain quality [2].In recent years,many genes or quantitative trait loci (QTL) associated with grain size have been isolated,and generally control seed size by regulation of the growth of maternal tissues [3].Numerous biological pathways associated with grain size have been elucidated and include ubiquitination,phytohormone,G-protein,photosynthesis,epigenetic modification,and microRNA pathways [4].

Organ size is determined largely by cell number and cell size during organogenesis.Most cloned genes that regulate seed size are associated with cell division and cell expansion.GW7/GL7[5,6],TGW6[7],SG1[8],qSW5[9],WCR1[10],andBG2[11] are all associated with cell division.GW7/GL7encodes a TONNEAU1-recruiting motif protein.Upregulation ofGW7expression is associated with the production of more slender grains,as a result of increased cell division in the longitudinal plane and decreased cell division in the transverse plane [5,6].TGW6encodes a novel protein with indole-3-acetic acid (IAA)-glucose hydrolase activity.The mutated genetgw6affects the timing of the transition from the syncytial to the cellular phase by controlling the supply of IAA,thereby limiting cell number and grain length [7].Other rice genes and QTL that control grain size by influencing cell expansion have been isolated,includingSLG[12],PGL1[13],andPGL2[14,15].SLGencodes a BAHD acyltransferase-like protein and increased cell length is responsible for the mutant phenotypes ofslg-D[12].qTGW2[16],OsSPL16[17],GS5[18],andBG1[19]control grain size via both cell division and cell expansion.qTGW2,a semi-dominant QTL encoding CELL NUMBER REGULATOR 1 (OsCNR1),negatively regulates grain width and weight by interacting with KRP1 to influence cell proliferation and expansion in the lemma [16].OsSPL16encodes GRAIN-WIDTH 8,a SQUAMOSA promoterbinding-like protein that belongs to the SBP domain family of transcription factors and regulates grain width by binding to theGW7promoter and repressing its expression[5,17].GS5encodes a putative serine carboxypeptidase and regulates grain size by increasing cell number and cell size [18].In summary,the majority of genes regulates grain size by influencing cell division and cell expansion in the glumes.Traits describing the appearance and quality of rice grain include grain shape and grain chalkiness[20].In general,rice quality is negatively associated with grain size.Large kernels show inferior quality and various degrees of chalkiness,originating mainly in abnormal starch synthesis during the grain-filling stage[21-23].GS2,which encodes GROWTH-REGULATING FACTOR 4(OsGRF4),not only regulates grain size by promoting cell division and cell expansion[24],but also promotes and integrates nitrogen assimilation,carbon fixation,and growth [25].GW2,encoding a RING-type E3 ubiquitin ligase,controls rice grain width and weight by increasing cell number in the spikelet hull and acceleration of the grain milk filling rate [26].Recent studies [27] suggest thatGW2ubiquitinates and targets WG1 for degradation;WG1encodes a glutaredoxin protein and promotes grain growth by increasing cell proliferation.GIF1encodes a cell-wall invertase that is required for carbon partitioning during an early stage of grain filling [28].FLO2influences the regulation of rice grain size and starch quality by affecting the accumulation of stored substances in the endosperm;loss of function ofFLO2results in a reduction in these traits[29].FLO10encodes a P-type pentatricopeptide repeat (PPR) protein and functions in the maintenance of mitochondrial function;loss of function ofFLO10leads to smaller,opaque grains with defective starch synthesis in the endosperm during the grainfilling stage [30].Additional PPR proteins such as OsNPPR1 [31]and FLO18 [32] support mitochondrial function and endosperm development in rice.OsBT1encodes an ADP-glucose transporter involved in starch synthesis by controlling ADP-glucose flux into starch in the endosperm and regulates seed dormancy via the glycometabolism pathway [33,34].Theosbt1mutant shows a white-core endosperm and reduced grain weight due to defective development of starch granules [35].Despite progress in research on grain size and endosperm development,the molecular mechanism is still unknown.In this study,we identified a novel rice mutant,filling-defectiveandgrain width(fgw1).Compared to the wild type,fgw1exhibited wider seeds and abnormal grain filling.We further analyzed the molecular mechanism of grain size and filling by studying this mutant.

2.Materials and methods

2.1.Plant materials and growth conditions

Thefgw1mutant was identified in an M2population grown fromOryza sativaL.subsp.indica‘Jinhui 10’ seeds treated with ethyl methanesulfonate.An F2mapping population was produced from a cross between thefgw1mutant and‘Xida 1A’,a cytoplasmic male-sterile line bred by the Rice Research Institute of Southwest University,Chongqing,China.All plants were grown in paddy fields in Chongqing under field conditions with 20 cm spacing within and 30 cm spacing between rows.At the maturity stage,10 plants were selected for recording plant traits.

2.2.Map-based cloning of FGW1

Thefgw1mutant was crossed separately with Xida 1A and Jinhui 10 to obtain F1populations,which were selfed to yield F2populations.The F2population offgw1×Jinhui 10 was used for genetic analysis,and F2plants of Xida 1A ×fgw1with the mutant phenotype were selected and used to mapFGW1.Initial mapping was conducted using simple sequence repeat (SSR) markers based on 12 rice linkage maps (https://rice.uga.edu/annotation_pseudo_putativessr.shtml).Fine mapping was performed using Indels and new SSR markers after the chromosome region was determined by linkage analysis.The sequences of primers used in the mapping are listed in Table S2.

2.3.Histological analysis

Spikelet hulls of the wild type andfgw1mutant were collected at the booting stage and fixed in FAA solution (70% ethanol,5%formaldehyde,and 5%acetic acid)for 48 h.The samples were infiltrated and embedded in paraffin after dehydration through an ethanol series.The hulls were cut into 10-μm-thick sections with a rotary microtome(Leica RM2245;Leica Microsystems,Hamburg,Germany) and transferred onto poly-L-Lys-coated glass slides,deparaffinized in xylene,and dehydrated through an ethanol series.The sections were stained with Fast Green and counterstained with safranin,dehydrated through an ethanol series,infiltrated with xylene,and finally mounted in neutral balsam and observed under a light microscope (Eclipse Ci-L;Nikon,Tokyo,Japan).

To measure the sizes of epidermal cells in spikelet hull,samples were excised from spikelet with scissors and observed with a Hitachi SU3500 scanning electron microscope under a strong vacuum[72].Cell length and width were measured with ImageJ (National Institutes of Health,https://imagej.net/ij/index.html).

2.4.Vector construction and plant transformation

For a complementation test,a 6157-bp genomic fragment containingFGW1(including 1828 bp upstream of the start codon,the entire coding region,and 733 bp downstream of the terminal codon)was amplified from wild-type genomic DNA using PrimeSTAR Max DNA Polymerase (Takara,https://www.clontech.com/takara),before being cloned into the binary vector pCAMBIA 1301 using T4 DNA Ligase(Takara,https://www.takarabio.com/products/cloning/ligation-enzymes/t4-dna-ligase).The recombinant plasmids were introduced into thefgw1mutant byAgrobacterium tumefaciensmediated transformation.To generate aFGW1-RNAi construct,a 322 bp gene-specific fragment of theFGW1coding sequence was amplified and cloned into the pTCK303 vector as described previously[36].The recombinant plasmids were transformed into Jinhui 10 plants.FGW1full-length coding sequences were cloned into the pTCK303 vector to generate a pTCK303-FGW1OEoverexpression construct.For a promoter-GUS assay,a 2923-bp genomic fragment upstream of theFGW1translation start codon was amplified and cloned into the pCAMBIA1301 vector.Both overexpression and promoter-GUS recombinant plasmids were transformed into Jinhui 10 plants.The primer sequences used are listed in Table S3.

2.5.RNA isolation and qRT-PCR

Total RNA was extracted from the root,culm,leaf,sheath,and young panicles of varying lengths using an Eastep Total RNA Extraction Kit(Promega,Shanghai,China).The concentration,purity,and integrity of extracted RNA were determined with a Nano-Drop 2000 spectrophotometer(Thermo Fisher Scientific,Waltham,MA,USA) and 1% agarose gel electrophoresis.First-strand cDNA was synthesized using the PrimeScript RT Reagent Kit with gDNA Eraser (Takara,https://www.takarabio.com/products/real-time-pcr/reverse-transcription-prior-to-qpcr/primescript-rt-reagent-kit).PCR was performed using gene-specific primers andACTINwas used as an endogenous control.qRT-PCR was performed using a TB Green Advantage qPCR Premix kit (Takara,https://www.takarabio.com/products/real-time-pcr/real-time-pcr-kits/dye-based-qpcrmixes/tb-green-advantage-qpcr-premixes) on an ABI Prism 7500 Real-Time PCR System (Thermo Fisher Scientific).At least three replicates were performed to obtain a mean expression level.The primers used for qRT-PCR are listed in Table S4.

2.6.In situ hybridization

To prepare theFGW1probe,a 227 bp fragment ofFGW1cDNA was amplified using the primersFGW1H-F andFGW1H-R.The probe was synthesized using a DIG RNA labeling kit (SP6?T7;Roche).Pretreatment of sections,hybridization,and immunological detection were performed as described previously[37].The primer sequences used are listed in Table S5.

2.7.Subcellular localization

The coding region ofFGW1without the stop codon was fused to the N-terminus of the GFP gene under the control of the CaMV 35S enhanced promoter in theSpeI andBamHI sites of the pAN580 vector to generatepFGW1-GFP.The FGW1-GFP fusion protein and the plasma membrane marker were transiently co-expressed in rice protoplasts following a previously described method[38].A confocal laser scanning microscope(LSM800;Zeiss,Jena,Germany)was used for fluorescence detection.The primer sequences used are listed in Table S3.

2.8.Yeast two-hybrid assay

The DUAL membrane system (Oebiotech,Chongqing,China),was used to perform a yeast two-hybrid screen or assay.Young panicles of Jinhui 10 shorter than 5 cm were used to construct an expression cDNA library for yeast two-hybrid screening.The yeast strain NMY51 was used for transformation.The coding sequences(CDSs) ofFGW1,DUF630,DUF632,andBZIPwere inserted into the linearized bait vector pBT-SUC to generate pBT-SUC-FGW1,pBTSUC-DUF630 (amino acids [aa]: 1-59),pBT-SUC-DUF632 (aa:324-633) and pBT-SUC-BZIP (aa: 482-767),respectively.The CDS ofGF14fwas inserted into the prey vector pPR3-N to generate pPR3-N-GF14f.The primer sequences used are listed in Table S3.

2.9.Bimolecular fluorescence complementation assay

The CDSs ofFGW1andGF14fwere cloned into the linearized pMDC43-nYFP and pMDC43-cYFP vectors (digested withXbal1 andBamH1) to generate the nYFP-FGW1 and cYFP-GF14f constructs,respectively.The plasmids were transformed intoAgrobacteriumstrain GV3101,configured in different combinations for co-transformation intoNicotiana benthamianaleaves.After 48 h,the confocal laser scanning microscope was used for fluorescence detection.The primer sequences used are listed in Table S3.

2.10.Enzyme activity assays

The enzyme activities of AGPase,StSase and SuSase were measured using enzyme kits according to the product manual(Cominbio,https://www.cominbio.com).Activities of enzymes were expressed in mol g-1FW min-1for AGPase and StSase and μg-1FW min-1for SuSase [39].

2.11.Starch and amylose content measurement

The content of starch and amylose was measured using starch and amylose kits according to the product manual (Solarbio,Beijing,China,https://www.solarbio.com).Starch and amylose content were expressed as percentages.

2.12.Multiple sequence alignment and phylogenetic tree construction

Phylogenetic analysis was performed with MEGA 5.0 (https://www.megasoftware.net/).Protein sequences for multiple sequence alignment and phylogenetic tree construction were retrieved from the National Center for Biotechnology Information(NCBI,https://www.ncbi.nlm.nih.gov/) database,using the FGW1 sequence as search query,and used to construct a maximumlikelihood phylogenetic tree.A bootstrap analysis with 500 replicates was performed to estimate the statistical support for each node.Multiple sequence alignment was performed using ClustalX(https://www.clustal.org/) and DNAMAN (https://www.lynnon.com/download/) software.

2.13.Transcriptome analysis

At the booting stage,1-cm-long panicles were sampled for RNA extraction and transcriptome sequencing with three biological replicates.Total RNA of each sample was extracted from panicles with TRlzol Reagent (Life Technologies,Carlsbad,CA,USA).RNA integrity and concentration were checked with an Agilent 2100 Bioanalyzer (Agilent Technologies,Inc.,Santa Clara,CA,USA).Library construction and RNA sequencing were performed by Biomarker Technologies (Beijing,China) on a HiSeq X-ten platform(Illumina,San Diego,CA,USA).Identification of differentially expressed genes (DEGs)was performed with the DESeq R package[73].The DEGs were used for gene ontology(GO)enrichment analysis with the agriGO online resource (http://systemsbiology.cau.edu.cn/agriGOv2/).Gene abundance differences between the samples were calculated based on the ratio of the fragments per kilobase of transcript per million mapped reads.The false discovery rate (FDR) control method was used to identify the threshold of theP-value in multiple tests to for use in identifying differences.Only genes with an absolute value of log2ratio ≥2 and FDR significance score ＜0.01 were used for subsequent analysis [74].Gene sequences were searched various protein databases using the BLASTX tool,including the NCBI non-redundant protein(Nr) database (https://www.ncbi.nlm.nih.gov/refseq/about/nonredundantproteins/) and the Swiss-Prot database [75] with a cut-off E-value of 10e-5.In addition,genes were searched against the NCBI non-redundant nucleotide sequence (Nt) database using the BLASTn tool with a cutoff E-value of 10e-5.Genes were retrieved based on the best BLAST hit(highest score)together with their protein functional annotation.

3.Results

3.1.The fgw1 mutant produced wider grains by affecting cell proliferation and expansion

To understand the genetic and molecular mechanisms that regulate rice grain size and filling,we initiated a genetic screen for mutants with altered grain appearance.Thefgw1mutant was isolated from the M2generation of the rice cultivar ‘Jinhui 10’ subjected to mutagenesis with ethyl methanesulfonate (EMS).It showed semi-dwarfism,reduced seed setting rate,fewer tillers and branches,shortened roots,and slightly curled leaves (Fig.S1;Table S1).It also displayed deep green leaves with higher chlorophyll content,containing more mesophyll cells infgw1than the wild type,increasing net photosynthetic rate (Pn),stomatal conductance (Gs) and transpiration rate (Tr) (Fig.S2).

The grain width of thefgw1mutant was increased by 28.4%over that of the wild type,whereas the grain length of thefgw1mutant was slightly but not significantly decreased in comparison with the wild type.These changes were reflected in an increase in 1000-grain weight relative to the wild type(Fig.1A-D).Anatomical observation of the lemma revealed that the outer parenchyma cell layer of thefgw1mutant was longer(by 10.4%)and contained more cells (1.8% more),and that the cell length was increased (by 8.4%)compared with the wild type (Fig.1E-H).We also measured the width and length of cells in the lemma epidermal tissue of wildtype andfgw1by scanning electron microscopy.(Fig.1I).Compared with those of the wild type,in thefgw1mutant the number of outer epidermal cells in the transverse and longitudinal planes increased by 25.6%and 8.8%,respectively(Fig.1J,K).The length of the inner epidermal cells was significantly (P＜0.001) decreased by 12.2%,whereas the cell width increased by 23.8%,compared with the wild type(Fig.1L,M).Similar results were obtained for the cell width of the outer epidermis (Fig.S1).Consistent with these results,genes associated with cell division and expansion were upregulated to varying degrees in the panicle of thefgw1mutant(Fig.S3A).All of these findings showed thatFGW1regulated the development of both cell proliferation and expansion.

Fig.1.Phenotypic characterization of the filling-defective and grain width(fgw1)mutant.(A)Grain morphology of the wild type(WT)and fgw1 mutant.Scale bars,4 mm.(BD) Grain length (B),grain width (C),and 1000-grain weight (D) in WT and fgw1 mutant.(E) Transverse sections of the spikelet of WT and fgw1 mutant.(a,b) Image of the spikelet hull of WT and fgw1 mutant.Scale bars,500 μm.(c,d)Magnified views in the boxes.Scale bars,50 μm.(F-H)Total length(F),cell number(G),and cell length(H)in the outer parenchyma layer of the spikelet hull of WT and fgw1 mutant.(I)Scanning electron micrographs of WT and fgw1 lemma.Scale bars,1 mm in(a,b),50 μm in(c,d),and 25 μm in (e,f).(J,K) Number of outer epidermal cells in the transverse plane (J) and longitudinal plane (K).(L,M) Average length (L) and average width (M) of inner epidermal cells of lemmas.Means ± SD are shown in (B,C,H,L,M) (n=20) and (D,F,G,J,K) (n=10).**, P ＜0.01 (Student’s t-test).

3.2.FGW1 regulated opaque and white endosperm via its effects on grain filling

Because the finding that thefgw1mutant produced a larger hull and the grain is wrinkled and opaque at maturity suggested thatFGW1might be associated with incomplete grain filling,the grain milk filling rates in the wild type andfgw1mutant were studied in detail.In the first six days after pollination(DAP),the endosperm fresh and dry weights of the wild type were higher than those of thefgw1mutant.Those of thefgw1mutant were significantly higher than those of the wild type from 9 days after fertilization,with the differences peaking 21 days after fertilization (Fig.2AC),when they were respectively 9.0% and 9.8% greater than those of the wild type.These results indicated that the larger endosperm and heavier grain of thefgw1mutant resulted from accelerated accumulation of dry matter and suggested thatFGW1might negatively regulate the rate of dry matter accumulation.

Fig.2.FGW1 regulates seed size and grain filling.(A)Endosperm development from 3 to 21 days after fertilization(DAP).Scale bar,1 mm.(B,C)Time-course of increase in endosperm fresh and dry weights.Data given as mean±SD(n=20).(D,E)Transverse sections of mature seeds of the wild type(WT)(D)and fgw1 mutant(E).Scale bar,1 mm.(F,G)Magnified views of the boxes in(D)and(E),respectively.Starch granules developed abnormally and were packed loosely in grains of fgw1 mutant.Scale bar,10 μm.(HK) Starch(H),amylose (I),glucose(J),and sucrose(K) contents of grains of the WT and fgw1 mutant.Data in (H-K)are shown as means ±SD(n=4).*, P ＜0.05;**, P ＜0.01(Student’s t-test).

When transverse sections of mature kernels from the wild type and mutant were compared,the endosperm of thefgw1mutant was opaque and uniformly white (Fig.2D,E).The starch grains of the wild-type endosperm cells were polyhedral and densely packed(Fig.2F),whereas thefgw1mutant showed markedly more grain chalkiness as a result of abnormally developed and loosely packed starch granules (Fig.2G).The starch content was dramatically lower,whereas the glucose and sucrose contents were higher,in kernels of thefgw1mutant,despite the presence of defective starch granules (Fig.2H-K).Further,we performed qRT-PCR analysis of genes involved in cell division and cell expansion.Two genes,D11andGL7,regulating both grain size and quality showed changes in transcript level.D11is a BR-synthesis gene regulating sugar accumulation and seed size[40],whose expression increased by 3.7 folds infgw1compared with the wild type.GL7[5,6],regulating both seed size and grain quality,increased its expression by 2.4 times (Fig.S3B).We inferred that theFGW1involves endosperm filling by affecting the conversion of sucrose to starch.

3.3.FGW1 encodes a protein with DUF630 and DUF632 domains

For gene mapping ofFGW1,all F1plants displayed a normal phenotype.The F2plants showed a phenotypic segregation ratio of 3:1(normal:wider grains),indicating thatfgw1was a recessive mutation.

The candidate genomic region offgw1was narrowed to a 123.7 kb interval between the markers Ind 10-2 and Ind 10-3 on the long arm of chromosome 10(Fig.3A).Twenty annotated genes were located within the candidate region in the Gramene database(http://www.gramene.org/,Chromosome 10: 22,090,451-22,213,869).To define the mutation site of thefgw1mutant,genomic DNA sequences from thefgw1mutant and wild type were amplified by PCR and sequenced.The only mutation detected was inLOC_Os10g41310infgw1,where a single-nucleotide deletion caused a frame shift that led to premature termination of the predicted protein(Fig.3A).LOC_Os10g41310was accordingly assigned as the candidate gene for thefgw1mutation.The identity of thefgw1gene was further confirmed by genetic complementation analysis.A 6157-bp wild-type genomic fragment ofLOC_Os10g41310,including an 1828-bp sequence upstream of the start codon and a 733-bp sequence downstream of the terminal codon,was transformed into thefgw1mutant.The mutant phenotype was completely rescued in transgenic plants (Figs.3B-E,S4).FGW1showed higher expression in developing grain at 9 DAP than other tissues (Fig.3F).These results confirm thatLOC_Os10g41310was theFGW1gene.FGW1 contains a DUF630 domain at the Nterminus and a DUF632 domain at the C-terminus (Fig.S5),and has been reported to regulate leaf rolling (REL2) and development of fewer tillers(DLT10),probably by monitoring auxin homeostasis[41,42].By qPCR,genes (PINs) involved in auxin transports also showed changed transcription infgw1panicles.The expression levels ofPIN1aandPIN1bexceeded those of the wild type(Fig.S6).These results further suggested thatFGW1is a novel allele ofREL2andDLT10.

Fig.3.Map-based cloning of FGW1.(A) Fine mapping of FGW1.The mutated locus is shown in the candidate gene LOC_Os10g41310.(B) Grain size was rescued in fgw1-complement plants(C-fgw1).(C)Grain width in wild type(WT),fgw1,and C-fgw1.(D)Defective starch was restored to normal in C-fgw1.(E)Starch contents in WT,fgw1,and C-fgw1.(F)Quantitative RT-PCR analysis of FGW1 expression in grain on days 3,9,and 15 of development.Means±SD are shown in(C)(n=20)and(E)(n=4).**,P ＜0.01;**,P ＜0.001 (Student’s t-test).

3.4.FGW1 negatively regulates grain width and positively regulates grain quality

To verify whether the expression ofFGW1is responsible for grain development in rice,we overexpressedFGW1cDNA under the control of the ubiquitin promoter in the wild type.FGW1-overexpressing plants (OE) had narrower seeds,consistent with FGW1 expression,and higher expression had narrower seeds,along with lower thousand grain weight (Fig.4A-E).Suppression ofFGW1by RNA interference(RNAi)led to clearly opposite phenotypes,including broader grain and increased 1000-grain weight(Fig.4F-J).The phenotype severity was consistent with theFGW1expression level in transgenic plants (Figs.4,S4),suggesting that the expression level ofFGW1was also responsible for the rice grain appearance and thatFGW1was a negative regulator of grain size formation and a positive regulator of grain filling.

Phylogenetic analysis indicated thatFGW1was highly conserved and showed a relatively close genetic relationship with homologs from other graminaceous species such asZea mays,Sorghum bicolor,andBrachypodium distachyon,and was not as closely related to the homolog fromArabidopsis thaliana(Fig.S5)[41].

3.5.FGW1 is preferentially expressed in young panicles and involved in development of vascular tissues

To confirm the expression pattern ofFGW1,relative transcript levels were quantified by qRT-PCR in various organs during vegetative growth and reproductive development.ThoughFGW1was expressed in multiple tissues,the transcript level was higher in culms and young panicles and gradually decreased as panicles matured (Fig.5A).FGW1promoter activity in various tissues ofPROFGW1:GUStransgenic plants was also checked.In agreement with the qRT-PCR results,β-glucuronidase (GUS)staining revealed thatFGW1was expressed in all examined rice tissues,with the highest expression level observed in culms and developing panicles(Fig.4B-H).Cross-sectioning of the GUS-stained hull and culm further showed that the GUS signal was mainly restricted to vasculature regions (Fig.4I-L).Further,in situhybridization of developing panicles demonstrated preferential expression ofFGW1in the primary and secondary branch meristem,floret meristem,and hull(Fig.5M-O).The predominant expression ofFGW1in different organs suggests its role in the control of grain size and development of vascular tissues.

3.6.FGW1 interacted with the 14-3-3 protein GF14f

To examine the subcellular localization of FGW1,the full-length coding sequence ofFGW1was fused to the N-terminus of the green fluorescent protein (GFP) and transformed into rice protoplasts.The FGW1-GFP fusion protein co-localized exclusively with the Tracker Red plasma membrane marker (Fig.6A).Similarly,the FGW1-GFP fusion protein was localized exclusively to the plasma membrane in root cells of transgenic rice plants (Fig.S8).Tus,FGW1 was localized to the plasma membrane.

Fig.6.FGW1 interaction with the 14-3-3 protein GF14f.(A)Subcellular localization of FGW1 in rice protoplasts,and the FGW1-GFP fusion protein was co-expressed with the plasma membrane marker Tracker Red.(B)FGW1 interaction with GF14f in yeast cells.DUF630(amino acids[aa]:1-59),DUF632(aa:324-633)and bZIP(aa:482-767).(C)FGW1 interaction with GF14f in rice protoplasts.(D)FGW1 interaction with GF14f in N.benthamiana leaves.(E)Subcellular localization of GF14f in rice protoplasts.Scale bars,10 μm.

To understand the molecular mechanisms by which FGW1 affected rice grain development,a yeast two-hybrid assay based on the DUAL membrane system was used to screen the interacted proteins.The yeast exhibited normal growth on SD/-Leu/-Trp/-His/-Ade medium for FGW1/GF14f,but no growth for DUF630/GF14f,DUF632/GF14f,and BZIP/GF14f.These results suggest that the unabridged protein of FGW1 is essential for the direct interaction of FGW1 and GF14fin vitro(Fig.6B).Biomolecular fluorescence complementation (BiFC) assay further confirmed this interactionin plantaand showed that the location of the interaction was specific to the membrane but not in the nucleusin vivo(Fig.6C,D),although GF14f was localized in both cytoplasm and nucleus(Fig.6E).We accordingly hypothesized that GF14f is precluded from interacting with the mutated FGW1 protein on the membrane.

3.7.FGW1 regulated cellular processes and carbohydrate metabolism

To further explore the function ofFGW1,we performed RNA sequencing (RNA-seq) analysis with panicles from wild-type andfgw1plants.Of 989 differentially expressed genes,72.8% (720)were upregulated and 27.2% (269) were downregulated infgw1plants in comparison with wild-type plants(Table S6).Gene ontology enrichment analysis and Clusters of Orthologous Groups function classification (https://www.ncbi.nlm.nih.gov/research/cog)showed that the expression of genes involved in carbohydrate transport and metabolism,cell division,and cell expansion was changed dramatically (Fig.7A,B).The relative transcript levels of six genes involved in cell division and expansion were compared by qPCR:GW5(BGIOSGA019303),GS5(BGIOSGA018771),CycA3;1(BGIOSGA013106),CycB1;5(BGIOSGA020095),CYCU4;1(BGIOSGA033453),andFSM(BGIOSGA004976).The transcript levels increased to varying degrees infgw1panicles,a finding consistent with the RNA-seq transcript-abundance changes (Fig.S9A;Table S6).The expression of five genes of carbohydrate metabolism:OsRBCS1(BGIOSGA007558),OsRBCS2(BGIOSGA018498),OsRBCS3(BGIOSGA037257),OsRBCS4(BGIOSGA037260),andSUS1(BGIOSGA010570),showed similar relative increases (Fig.S9B;Table S6).

Fig.7.RNA-sequencing and enzyme activity analysis.(A) Gene ontology enrichment analysis of differentially expressed genes (DEGs).(B) Cluster of orthologous groups function classification of differentially expressed genes.(C)Relative expression levels of GF14f in fgw1 and wild type(WT).(D-F)Activity of AGPase(D),StSase(E),and SuSase(F) in fgw1 and wild-type (WT) grain.Vertical bars represent ± SEM (n=3),and ** represents a difference at P ＜0.01 (Student’s t-test) between fgw1 and WT plant.

Inhibiting the expression ofGF14fincreased the activities of AGPase,StSase,and SuSase,and produced larger seeds [39].The expression ofGF14fdecreased by 55.0%and the activity of AGPase,StSase and SuSase increased by 37.3%,17.8% and 44.8%,respectively,infgw1compared with those in the wild type,with all increases significant(Fig.7C-F).Collectively,these results support the conclusion thatFGW1functions as a regulator of cell division,cell expansion,and carbohydrate metabolism,at least partially,by controlling the 14-3-3 protein GF14f,which has been reported[39] to regulate grain development and filling by interacting with enzymes involved in sucrose breakdown,starch synthesis,and glycolysis.

4.Discussion

FGW1contained two domains of unknown function:DUF630,a putative proline-rich domain,located at the N-terminus,and DUF632 located at the C-terminus (Fig.S5A).To date,only a few plant genes containing DUF630/DUF632 domains have been studied:APSR1,NRG2,andREL2/OsDLT10.APSR1is negatively regulated by low phosphate availability and is required for meristem maintenance[43],andNRG2mediates nitrate signaling and regulates cadmium tolerance inArabidopsis[44,45].REL2/OsDLT10is the only identified rice gene that contains the DUF630/DUF632 domains.Therel2mutant shows leaf rolling [41] and plants carrying the allele ofdlt10produce fewer tillers [42].Although proteins with DUF630/DUF632 domains function in plant development,their molecular mechanisms remain unclear.We characterized another allelic mutant,fgw1,that developed larger and filling-defective grain and differed from the reportedrel2anddlt10.FGW1 may perform complex functions and at least leads to broad grain with opaque and white endosperm via a GF14f-specific pathway.

4.1.FGW1 regulates starch synthesis in rice endosperm

Starch synthesis in rice assumes two forms: transient starch synthesis and reserve starch synthesis [46].Starch synthesis in the endosperm belongs to the reserve starch synthesis type.Identification of relevant mutants is essential for elucidating its molecular mechanism.In the present study,thefgw1mutant showed a much faster grain-filling rate than the wild type,but the endosperm offgw1was opaque and uniformly white,containing abnormal and loosely packed starch granules (Fig.2).

Sucrose from leaf photosynthesis or starch degradation is transported to the endosperm in the phloem and then transported to the cytosol via a sucrose transporter[47,48].The sucrose contents of thefgw1mutant were slightly higher than those of the wild type at the early grain-filling stage and increased to three times higher at the maturity stage (Fig.2).These results suggest that sucrose may be transported continuously to the endosperm from the delayed senescent leaves of thefgw1mutant,in which the net photosynthetic rate(Pn)was much higher than in the wild type owing to the greater number of mesophyll cell layers (Fig.S2).

Sucrose hydrolysis is the first step in the reserve starch synthesis pathway,and reserve starch biosynthesis uses sucrose as a substrate [49].Sucrose in the cytosol is converted into fructose and uridine diphosphate glucose by sucrose synthase,after which they are transported to the endosperm for starch synthesis by a monosaccharide transporter [49,50].The starch content of thefgw1mutant was much lower,but the glucose content was slightly higher at the early grain-filling stage and moderately lower at the maturity stage,compared with the wild type (Fig.2).These findings suggest that abnormal endosperm and starch grain development in thefgw1mutant may be caused by defective starch synthesis.And surplus sucrose from delayed-senescence leaves(Fig.S2)was unavailable to increase grain yield and quality where the grain filling efficiency had not been improved.

4.2.FGW1 influences grain size by regulation of cell division and expansion

Sugar metabolism plays a vital role in plant development.However,only a small number of relevant genes have been cloned and the mechanism by which sugar metabolism affects rice grain size remains elusive[51].In rice,grain size is limited by the hull,which consists of the lemma and the palea [52].OsGPRP3 regulates rice grain size and shape by influencing the cell width of spikelet hulls and the accumulation of storage protein and lipids[53].In the present study,we isolated thefgw1mutant,which produces broader,heavier kernels than the wild type.Several genes that influenced grain size by moderating cell division and expansion were upregulated in the panicle of thefgw1mutant:GS3,SRS5,andAFD1(Fig.S3).GS3 is a major QTL that negatively regulates grain length and grain weight,and a minor QTL that affects grain width and grain filling [54-56].SRS5 encodes an alpha-tubulin protein that promotes small,round grain by regulating the cell number and cell length [57].AFD1 determines grain shape and size by modifying cell division and expansion in the grain hull[58,59].The grain size phenotypes of thefgw1mutant were caused mainly by the broader cells and increased number of cells in the hulls (Figs.1,S1).Thus,FGW1negatively determines grain size by influencing the cell size and number.

Cell division in plants is controlled by the activity of cyclindependent kinase complexes.In the present study,transcriptome analysis indicated thatFGW1regulated cellular processes and carbohydrate metabolism (Fig.7D).Four cell cyclin-related genes,namelyCYCU4;1,CycB1;5,CycA3;1,andFSM,were upregulated in thefgw1mutant(Fig.S9;Table S6).CycU4;1participates in the cell cycle at the early G1/S stage and regulates leaf angle [60].CycB1;5is an A-type cyclin gene,which participates in the cell cycle at the late S stage and is expressed in the development of mature tissues in many species and in specific developmental processes,such as endosperm development [61,62].CycA3;1is a B-type cyclin gene and participates in the cell cycle at the G2/M transition [63].FSMis required for proper maintenance of the shoot apical meristem and participates in the cell cycle at the G1 stage[64].These results further suggest thatFGW1affects grain size by regulating cell division and cell expansion.

4.3.FGW1 regulates rice filling by the GF14f pathway

We identified no protein interaction using a classical yeast twohybrid assay,probably because FGW1 is localized to the plasma membrane.We accordingly performed a yeast two-hybrid assay based on the DUAL membrane system to screen for protein interactions and identified eight proteins,one of which was GF14f,a 14-3-3 protein.The BiFC assay confirmed the interaction,which was localized specifically on the membrane and not in the nucleus or cytoplasmin vivo(Fig.6C,D).The 14-3-3 proteins are known[65-68] to affect the biological function of their client proteins by alteration of their subcellular locations.GF14c,another 14-3-3 protein,interacted with rice florigen in Hd3a (FT homolog) in stem apical cells,and the complex was translocated to the nucleus and interacted with transcription factor OsFD1,influencing rice flowering and lateral branching [67,69].Corresponding to that,thefgw1also showed multiple phenotypes,developing larger seeds,poor grain quality,shortened roots,fewer branches,and semi-curled and deep green leaves.The finding that the transcript levels ofGF14fdiffered between interfering and overexpressing strains ofFGW1suggested that the expression ofGF14fis influenced by the amount ofFGW1expressed (Fig.S10).This evidence also suggests that there is a negative feedback regulation betweenFGW1andGF14fexpression when the function ofFGW1is not impaired.In this model,the change inGF14ftranscript levels is a negative feedback regulation of its own protein content.

Although theGF14f-RNAi mutant affected only grain length in a previous study[39],it also plays a negative regulatory role in grain size and filling.In a recent report[70],a model of grain regulation ofGF14fwas proposed:OsCEN2-GF14f-OsFD2.In this model,OsCEN2andOsFD2have the same negative regulatory effect on grain size and weight asFGW1.UnlikeGF14f,bothOsCEN2andOsFD2can also negatively regulate grain width.Considering that grain size is determined by a complex regulatory network,we propose thatGF14faffects some functions ofFGW1once the effect of the interaction is lost.That is,FGW1is a target gene ofGF14f,and the down-regulated expression ofGF14fdid not result in shorter grain in thefgw1mutant.

GF14fis reportedly expressed in all tissues of rice,with the highest expression detected in the panicle [71],a finding consistent with the expression pattern ofFGW1(Fig.5).GF14f accumulates in the nucleus and cytoplasm (Fig.6E),FGW1 accumulates in the membrane (Figs.6A,S8),and fragments of FGW1 fail to interact with GF14f (Fig.6B).GF14fplays a role in carbohydrate metabolism and starch accumulation during the grain filling stage by interacting with identified genes related to carbohydrate metabolism [40].In thefgw1mutant,the starch content was dramatically lower but the glucose and sucrose contents higher than those of the wild type (Fig.2H-J).Correspondingly,some genes associated with carbohydrate metabolism were up-regulated.The activities of AGPase,StSase and SuSase increased infgw1compared with those in wild type(Fig.7D-F).All of these characteristics were similar to those of the reportedGF14f-RNAi[39].The expression ofGF14fdecreased in the developed seed offgw1,to a level 45% of that in the wild type (Fig.7C).These results further confirm thatFGW1influences rice grain development by regulation of carbohydrate metabolism and starch accumulation via a novelGF14fpathway.

5.Accession numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries:FGW1(Os10g0562700),GW5(Os05g0187500),GS5(Os05g0158500),CycA3;1(Os03g0607600),CycB1;5(Os05g0493500),CYCU4;1(Os10g0563900),FSM(Os01g 0896300),OsRBCS1(Os02g0152400),OsRBCS2(Os12g0274700),OsRBCS3(Os12g0291100),OsRBCS4(Os12g0292400),SUS1(Os03g 0401300),GW8(Os08g0531600),GS3(Os03g0407400),DSG1(Os09g0434200),SRS5(Os11g0247300),PGL2(Os02g0747900),AFD1(Os02g0811000),GS2(Os02g0701300).

CRediT authorship contribution statement

Yangyang Li:Project administration,Investigation,Formal analysis,Writing-Original Draft,Writing-Review&Editing.Peilong He:Project administration,Investigation,Formal analysis,Writing-Original Draft.Xiaowen Wang:Project administration,Formal analysis,Writing -Original Draft.Hongyan Chen:Investigation,Formal analysis.Jile Ni:Investigation,Formal analysis.Weijiang Tian:Investigation,Resources.Xiaobo Zhang:Resources.Zhibo Cui:Resources.Guanghua He:Conceptualization,Supervision,Resources and Funding acquisition.Xianchun Sang:Conceptualization,Supervision,Resources,Writing -Original Draft,Writing -Review &Editing,Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

We thank Dr.Xiaoping Lian for assistance with theArabidopsistransgenic experiments,Dr.Ling Ma for help with vector construction,Dr.Wuzhong Yin and Dr.Ting Zhang for help with thein situhybridization experiments,Tianquan Zhang and Dr.Dan Du for assistance with the subcellular localization experiments,and Han Chen for help with RNA-seq.This work was sponsored by the National Key Research and Development Program of China(2022YFD1201600,2016YFD0100501),Natural Science Foundation of Chongqing of China (cstc2020jcyj-msxm0539),the National Natural Science Foundation of China (32171964),and Chongqing Natural Science Foundation Innovation Group (cstc2021jcyjcxttX0004).

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2023.03.015.