Weidan Zhang,Huanjun Li,Feiyang Xue,Wanqi Liang*

Joint International Research Laboratory of Metabolic & Developmental Sciences,State Key Laboratory of Hybrid Rice,Shanghai Jiao Tong University-University of Adelaide Joint Centre for Agriculture and Health,School of Life Sciences and Biotechnology,Shanghai Jiao Tong University,Shanghai 200240,China

Keywords:

A B S T R A C T In plants,non-green plastids in heterotrophic tissues are sites for starch and fatty acids biosynthesis,which are essential for plant development and reproduction.Distinct from chloroplasts,the metabolites for these processes in non-green plastids have to be imported through specific transporters.Glucose 6-Phosphate/Phosphate Translocator 1 is required for the uptake of cytosolic Glucose 6-Phosphate into non-green plastids.In Arabidopsis,GPT1 has been demonstrated to play essential roles in male,female gametophyte and embryo development.However,the roles of GPTs in other species are yet largely unknown.Here,we reported that rice OsGPT1 is indispensable for normal tapetal degeneration and pollen exine formation during anther and pollen development.OsGPT1 is localized in the plastid and distributed in the anther wall layers and late-stage pollen grains.Different from the gametic defects caused by mutation in AtGPT1,disruption of OsGPT1 does not affect male and female gamete transmission as well as embryo development.On the contrary,osgpt1 mutant exhibits delayed tapetum degeneration,decreased Ubisch bodies formation and thinner pollen exine,leading to pollen abortion at the mature stage.Furthermore,the expression of several genes involved in tapetal programmed cell death(PCD)and sporopollenin formation is decreased in osgpt1.Our study suggests that OsGPT1 coordinates the development of anther sporophytic tissues and the male gametophyte by integrating carbohydrate and fatty acid metabolism in the plastid.

1.Introduction

Development and maturation of pollen,the male gamete of flowering plants,occurs in the anther and requires the coordination between somatic tissues and the gametophyte.The anther is composed of four cell layers surrounding the central reproductive cells,from outside to inside,the epidermis,endothecium,middle layer and tapetum[1].The innermost cell layer tapetum plays essential roles in providing enzymes,nutrients and pollen wall components to the developing microspores.Accompanied with the microspore development,tapetal cells undergo programmed cell death(PCD)and disintegrate at mature stage.Defects in the tapetal function or the timely degradation of tapetum usually lead to pollen abortion[2–6].It is well documented that the development and degeneration of the tapetum are controlled by a complicated transcriptional regulatory network[7].Emerging evidence indicates that abnormalities in lipid metabolism will also affect tapetum degradation and function[8].It has been shown that disruption of several enzymes involved in pollen exine formation causes delayed or premature tapetal degradation,such as cytochrome P450 CYP704B2[9],glucose-methanol-choline oxidoreductase OsNP1[10],atypical strictosidine synthase OsSTRL2[11]and tetraketide α-pyrone reductase TKPR1[12]in rice,a putative glucose-methanol-choline oxidoreductase IPE1[13]in maize and Tic40 protein BnMs3 inBrassica napus[14].In plant cells,de novo fatty acid biosynthesis takes place in plastids.However,compared with the chloroplast in green tissue,the functions of plastid in heterotrophic tissues are much less understood.

Anthers and pollen are heterotrophic sink tissues,whose growth rely on the photosynthates produced by the photosynthetic organs.In green tissues,solar energy is transformed into C3 compounds in the chloroplast,which are then generated hexose phosphate(such as glucose 6-phosphate(Glc6P)and glucose 1-phosphate(Glc1P))via Calvin cycle.In heterotrophic tissues,non-green plastids function as carbohydrate-importing organelles and are sites for starch synthesis in storage tissues,such as the endosperm and pollen.The heterotrophic plastids are usually unable to produce hexose phosphate from C3 compounds due to lacking fructose 1,6-bisphosphatase activity[15].Instead,nongreen plastids employ Glucose 6-Phosphate/Phosphate translocators(GPTs)to import Glc6P generated in the cytosol as carbon source[16].Alternatively,amyloplasts in storage tissues also take in Glc1P or ADP-Glc as the precursor for starch synthesis.Assimilates imported into these plastids are further converted into storage compounds and metabolic intermediates,or acting as a substrate for the oxidative pentose phosphate pathway(OPPP)[17].

GPTs have been demonstrated to play critical roles in the plant growth and stress response,especially in reproductive growth.InViciaembryos with repressed expression ofVnGPT1,the specific transport rate of Glc6P is significantly reduced,accompanied with a reduction in starch biosynthesis and production of amyloplasts with smaller size,which provides a strong evidence that the function of VnGPT1 in starch synthesis[18].In Arabidopsis,AtGPT1is ubiquitously expressed while its paralogAtGPT2is mainly expressed in heterotrophic organs.The T-DNA insertion lines ofAtGPT1show a reduction in both male and female transmission efficiency and distorted segregation ratio,indicating thatAtGPT1is required for male and female gametophyte development.Abortion of mutant pollen is associated with reduced lipid body and dispersed vacuole formation,together with disintegration of the inner membrane system[19].A recent report demonstrates that the fatty acid biosynthesis mediated by AtGPT1 in pollen is regulated by the MKK4/MKK5-MPK3/MPK6 cascade and downstream transcription factors WRKY2/WRKY34[20].Very recently,AtGPT1 has been shown to have dual subcellular localization,not only in the plastid but also in peroxisomes,and the first 155 amino acids of AtGPT1 are crucial for peroxisomes targeting[21].Furthermore,this study indicates that AtGPT1 is able to exchange Glc6P for ribulose-5-phosphate(Ru5P),which is one product of the OPPP and an important precursor of nucleotide biosynthesis.These results suggested thatAtGPT1is involved in translocating both the substrate and product of the OPPP.The role ofAtGPT1in sporophytic tissues is obscure due to the lethality of homozygous mutant embryos[19].On the contrary,mutation inAtGPT2does not affect pollen and embryo sac development[19].The role of GPTs in other plant species is yet largely unknown.

In this study,we identifiedOsGPT1in rice,which plays essential roles in tapetal degeneration and pollen development.OsGPT1is ubiquitously expressed and localized to the plastid.Different from the gametophytic defects caused by mutation inAtGPT1,disruption ofOsGPT1does not affect male and female gametophyte transmission.On the contrary,osgpt1mutant display delayed tapetal degeneration,defective Ubisch bodies and pollen exine formation,which leads to pollen abortion at mature stage.These results suggest that OsGPT1 is essential for the tapetum function and involved in the biosynthesis of aliphatic components of the pollen exine.The functional divergence between Arabidopsis and rice may be the consequence of functional diversification of different GPT members during evolution.

2.Materials and methods

2.1.Plant materials,growth condition and map-based cloning of OsGPT1 gene

Rice plants(cv.9522,japonica)used in this study were grown in the paddy field in Shanghai Jiao Tong University.In order to identifyOsGPT1,we performed the map-based cloning analysis using the male sterile plants selected from the F2mapping population generated from a cross between theosgpt1mutant(japonica)and Guangluai 4(indica).Theosgpt1mutated site was narrowed to a genomic region(genetic distance of 35.7 to 75.7 cM)on chromosome 8.The molecular markers used for map-based cloning were described previously[22].High-throughput sequencing analysis indicated that a single-base substitution in the fourth exon of a putative TPT domain of LOC_Os08g08840,which caused the pretermination of protein translation.

2.2.Characterization of the mutant phenotype

The preparation and observation of samples for DAPI staining,semi-thin sectioning,TEM(transmission electron microscopy),SEM(scanning electron microscopy),and TUNEL(transferasemediated dUTP nick-end labeling)performed as previously described[23].The developmental stages of anther and microspore are as defined by Zhang et al.[1].

2.3.GUS staining

Anthers collected fromOsGPT1pro:OsGPT1gDNA-GUScomplementedosgpt1plants were stained by GUS solution(924 mL 50 mmol sodium phosphate buffer,1 mL triton X-100,25 mL methanol,20 mL 0.5 mol EDTA,10 mL 50 mmol K3Fe(CN)6,10 mL 50 mmol K4Fe(CN)6,10 mL 100 mg mL-1X-Gluc)at 37 °C 5 h in the dark,then de-stained by 75% ethanol under room temperature for 10 days.Fresh solution was changed once a day.Images were captured by a Leica MZ8 dissecting microscope.

2.4.Complementation of the osgpt1 mutant

For the complementation experiment,a 6014 bp genomic DNA fragment ofOsGPT1,which contains 3005 bp promoter sequence,2632 bp genic region and 377 bp 3′UTR ofOsGPT1,was cloned into the binary vector pCAMBIA1301(CAMBIA).To analyze the localization of OsGPT1 protein,we generated theOsGPT1pro:OsGPT1gDNAGUS and OsGPT1pro:OsGPT1gDNA-GFPvectors,containing the 3005 bp promoter and 2629 bp genic region without the stop codon fused with GUS or GFP,and then cloned into the binary vector pCAMBIA1301 respectively.These vectors were independently introduced intoAgrobacterium tumefaciens(EHA105)and transformed into the calli induced from young panicles of theosgpt1mutant.Primers used for vector construction are listed in Table S1.

2.5.RT-qPCR analysis

Total RNA was respectively isolated from wild-type plant(root,stem,leaf,palea,lemma,panicle of different length,and anthers at different stages)andosgpt1(anthers at different stages)using Trizol reagent(Generay).RT-PCR and RT-qPCR procedures and data calculation were performed as described by Uzair et al.[24].

2.6.Subcellular localization

To determine the subcellular localization,we constructed the 2×35Spro:OsGPT1cDNA-GFPtransgenic vector.1161 bp full-length cDNA ofOsGPT1was cloned into the BamH I site of 2×35Spro:GFPvector.Protoplasts were isolated from etiolated rice hypocotyls.Transformation was performed as previously described[5].The GFP fluorescence signal was analyzed by a laser scanning confocal microscope(Leica TCS SP5),with 488 nm excitation wavelength and 505–530 nm emission wavelength.

2.7.Phylogenetic analysis

The full-length protein sequence of OsGPT1 was used to search for homologous sequences in public database(Phytozome-BLASTP).39 homologous sequences from 14 plant species were obtained.Sequence alignment and Neighbor-Jointing(N-J)phylogenetic tree construction were performed by using MEGA4 software(version 4.0)with the following parameters:Poisson correction,pairwise deletion,and bootstrap(1000 replicates)[25].

2.8.Generation CRISPR knockout mutant

The sgRNA-Cas9 plant expression vectors were constructed as previously described in Xie et al.[26].The primers for constructing the sgRNA vectors are listed in Table S1.

3.Results

3.1.Isolation and genetic analysis of osgpt1 mutant

To identify genes vital for rice anther development,we isolated a male sterile mutant from our rice mutant library which was generated through60Co γ-ray radiation[27].We named this mutantglucose 6-phosphate/phosphate translocator 1(osgpt1),because the mutation occurred inOsGPT1,which is highly homologous toAtGPT1/2.osgpt1exhibited significantly reduced tiller number(average 13-tillers in wild-type,n=11;4 tillers inosgpt1,n=9),slightly dwarf plant stature(wild-type plants are average 82.6 cm in height,n=9;osgpt1plants are average 62.4 cm in height,n=9)and smaller panicle at the heading stage(Fig.1A,B).Meanwhile,osgpt1showed morphologically normal spikelets but with smaller and pale yellow anthers that produced a small amount of viable pollen grains compared with wild-type(Fig.1C–E).After anthesis,the anther fromosgpt1displayed unsuccessfully dehiscence(Fig.S1).At maturity,no seed was borne onosgpt1panicles(Fig.1B).All F1plants from the backcross ofosgpt1with wild-type were fertile,which indicated that female development was normal in the mutant.In addition,theosgpt1+/-heterozygous plants produced completely fertile pollen,which was similar to wild-type(Fig.S2).And the F2progeny showed an approximate 3:1(245 fertile vs.62 male sterile)segregation(χ2=4.36;P＞0.5),suggesting thatosgpt1was a monofactorial recessive mutant caused by sporophytic defects,which is different from the reported gametic defect ofatgpt1.

3.2.OsGPT1 is essential for anther and pollen development

To characterize the defects ofosgpt1during male reproductive development,we examined the semi-thin section of wild-type andosgpt1anthers.No detectable morphological differences were observed between wild-type(Fig.2A–C)andosgpt1(Fig.2D–F)until young microspore stage(stage 9).osgpt1appeared to undergo normal meiosis and produced normal tetrads and young microspores(Fig.2E,F).Consistent with this result,4′,6-diamidino-2-phe nylindole(DAPI)staining assay showed that male meiocytes inosgpt1displayed normal chromosome behavior during meiosis(Fig.S3).At vacuolated microspore stage(stage 10),the wildtype tapetal cells become thinner due to condensation and degradation,microspores were spherical with large vacuole(Fig.2G).By contrast,tapetal cells inosgpt1were less condensed(Fig.2J).Subsequently,in the wild-type anther,the tapetal cells continued to degenerate and pollen grains became filled with starch granules after two rounds of mitosis(Fig.2H,I).Inosgpt1,tapetal cells did not degenerate but were slightly expanded and most microspores were collapsed or vacuolated without accumulation of the reserve substances at mature stage(Fig.2K,L).These results indicated that OsGPT1 is indispensable for both tapetum and postmeiotic haploid microspore development.

3.3.osgpt1 is defective in tapetum degeneration

Fig.1.The phenotypic comparison between wild-type and osgpt1.(A)The wild-type and osgpt1 plants after heading.(B)The panicles showing fertile seeds in wild-type and infertile ones in osgpt1.(C)The wild-type and osgpt1 spikelets at stage 13 after removal part of the palea and lemma.(D)and(E)I2-KI staining of wild-type(D)and osgpt1(E)pollen grains at stage 13.gl,glume;pi,pistil;st,stamen.Scale bars,15 cm in(A),5 cm in(B),1 mm in(C),and 100 μm in(D,E).

Fig.2.Transverse section analysis of anther development in wild-type and osgpt1.Representative images from the cross-section of an individual locule of anthers in wildtype(A–C,G–I)and osgpt1(D–F,J–L)at different developmental stages.Bp,bicellular pollen;DBp,defective bicellular pollen;DMp,defective mature pollen;Dy,dyad cell;E,epidermis;En,endothecium;ML,middle layer;Mp,mature pollen;Msp,microspore;St8a to St12,stage 8a to stage 12;T,tapetum;Tds,tetrads.Scale bars,25 μm.

We performed TEM to examine the developmental defects in theosgpt1anther more closely.At stage 9,in wild-type tapetal cells,the large central vacuole became dispersed and the cytoplasm became condensed,which were signs of cell degeneration(Figs.3A,S4A).Higher number of mitochondria were observed inosgpt1tapetal cells at this stage(Figs.3E,S4E).At stage 10,in wild-type anther,the tapetal layer continued to condense and were deeply stained,accompanied with nuclei starting to disintegrate(Figs.3B,S4B).By contrast,inosgpt1,tapetal cells became vacuolated and expanded with intact double nuclei present in the cytoplasm(Figs.3F,S4F).At stage 11,the tapetum in wild-type became much thinner with disappearance of the nuclear envelope and rupture of the nucleolus(Figs.3C,S4C),while in vacuolated tapetal cells ofosgpt1,intact nuclei were still detected(Figs.3G,S4G).At stage 12,only residues of the tapetum were remained in the wild-type anther(Figs.3D,S4D).However,tapetal cells with unbroken nuclei were persistent in theosgpt1anther(Figs.3H,S4H).Furthermore,mature pollen grains in wild-type were in round shape,full of reserve substances(especially starch granules)at stage 12(Fig.3I).However,pollen grains inosgpt1anthers were collapsed and vacuolated,with no or only a few starch granules formation in the cytoplasm(Fig.3J).The irregular and collapsed pollen grains were also detected inosgpt1from SEM assay(Fig.3K,L).

It has been demonstrated that the tapetum degeneration is a process of PCD[2].We performed the terminal deoxynucleotidyl TUNEL assay to further analyze the tapetal PCD in wild-type andosgpt1.In the wild-type anther,TUNEL signals firstly appeared in tapetal cells at stage 7(Fig.4A)when male meiocytes entered meiosis,then continued to increase and reached the strongest at stage 8a(Fig.4B).While inosgpt1,although tapetal PCD also initiated at stage 7,TUNEL-positive signals were weaker than those in wild-type(Fig.4D),intensified during meiosis and were persistent after the tetrad formed(Fig.4E,F,J).In wild-type,the TUNEL fluorescence signals decreased when tetrads formed(Fig.4C,G)and disappeared after the vacuolated microspore stage(Fig.4H,I).In contrast,inosgpt1,TUNEL signals were still quite strong at the vacuolated microspore stage and residual signals were still detectable after degeneration of microspores(Fig.4K,L).Together with the TEM observations(Fig.3E–H),these results confirmed thatosgpt1was defective in tapetal PCD.

3.4.osgpt1 exhibited smaller Ubisch body and thinner pollen exine

Fig.3.Transmission electron microscopy(TEM)and scanning electron microscopy(SEM)observation of the tapetum and the pollen grain in wild-type and osgpt1 anthers.(A)–(H)TEM images showing the tapetal cells of wild-type(A–D)and osgpt1(E–H)at stage 9(A,E),stage 10(B,F),stage 11(C,G)and stage 12(D,H)respectively.(I)and(J)TEM images displaying the pollen grain of wild-type(I)and osgpt1(J)at stage 12.(K)and(L)SEM images of the pollen grain of wild-type(K)and osgpt1(L)at stage 12.E,epidermis;En,endothecium;M,mitochondrion;N,nucleus;Neo,nucleolus;SG,starch granules;St9 to St12,stage 9 to stage 12;T,tapetum.Scale bars,1 μm in(A,E,C,G),2 μm in(B,F,D,H),5 μm in(I,J),and 10 μm in(K,L).

It has been reported that rice mutants defective in tapetal PCD,such aseat1[5],ostkpr1[12]andptc2[24],usually exhibit abnormalities in the formation of pollen exine which is composed of sporopollenin,an extremely chemically inert aliphatic biopolymer.TEM observation showed that a two-layered pollen exine composed of the tectum,bacula and foot layer was formed on the surface of wild-type microspores at the vacuolated stage(Fig.5C).Meanwhile U-shaped Ubisch bodies,a special structure observed in the anther of many species which is believed to carry sporopollenin,were produced on the tapetal inner surface in the wild-type anther(Fig.5A).At later stages,the exine layers became thicker and the intine developed underneath the foot layer(Fig.5G,K).Inosgpt1,Ubisch bodies were much smaller than those in the wild-type,due to less sporopollenin like materials accumulation on the surface(Fig.5A,B,E,F,I,J).In addition,the exine layers were significantly thinner at all developmental stages(Fig.5C,D,G,H,K,L).Consistently,SEM observations found that less and smaller Ubisch bodies were distributed on the inner surface of the anther(Fig.5M,N),and the density of sporopollenin granules on the pollen surface was decreased(Fig.5O,P)inosgpt1.These observations indicated that OsGPT1 plays a crucial role in formation of Ubisch body and pollen exine.

3.5.Map-based cloning and sequence analysis of OsGPT1

We used a map-based cloning approach to identifyOsGPT1.Based on initial mapping results,osgpt1mutated gene was located to a region on chromosome 8 between markers Os804 and Os808(Fig.6A).Then the High-throughput sequencing revealed a single-base substitution in the fourth exon of LOC_Os08g08840,which generated a premature stop codon at 339th amino acids(Fig.6A,B).Sequence analysis indicated that LOC_Os08g08840 encodes a TPT domain containing protein that is the homolog ofArabidopsisGPT1.In order to confirm the mutation was responsible for theosgpt1phenotypes,we introduced the wild-type genomic DNA ofOsGPT1into theosgpt1mutant,and the fertility of transgenic plants were completely restored(Fig.S5).

We created anotherosgpt1allelic mutant(c-osgpt1)using CRISPR-cas9 technology,which introduced a 37-bp deletion within the first exon ofOsGPT1,causing frameshift and premature translational termination and loss of the functional TPT domain(Fig.S6A).c-osgpt1also showed smaller and pale yellow anthers that produced few fertile pollen grains(Fig.S6B-D).Semi-thin section and SEM assays showed thatc-osgpt1exhibited similar defects in tapetal degeneration(Fig.S7),Ubisch body and pollen exine formation(Fig.S8).These results demonstrated that the phenotypes ofosgpt1were caused by the mutation inOsGPT1.

To identify the homologs ofOsGPT1in rice and other plants, we performed phylogenetic analysis. We use OsGPT1 full-length protein sequence to search its homologous sequences in public database (Phytozome-BLASTP), totally 39 homologous sequences retrieved from 14 plant species were obtained. These genes are divided into two different branches according to dicots and monocots and the genes in dicots/monocots clade appear to be further divided into two subclades (Fig. S9A). The rice GPTs shared higher homology with others GPTs in monocot (such asSetariaviridis,Panicumhallii,Zeamays,Sorghumbicolorand so on), and theArabidopsisGPTs are more closely related with dicot (Glycinemax,CucumissativusandMedicagotruncatula) homologs (Fig. S9A). In rice genome, we identified two paralogs ofOsGPT1,LOC_Os07g33954/OsGPT2-1and LOC_Os07g34006/OsGPT2-2,whose protein sequences share 71.32% and 69.9% identity with OsGPT1 respectively.OsGPT2-1andOsGPT2-2have same gDNA sequence and coding sequence, while differ in the promoter region.There is another GPT like gene,LOC_Os07g33910/OsGPT2-3in rice genome.OsGPT2-3also shares highly similar gDNA sequence withOsGPT2-1andOsGPT2-2.ButOsGPT2-3cDNA lacks 444 bp(148 aa)C-terminal coding sequence,indicating that it might be a pseudogene(Fig.S9B).Three OsGPT2 genes are tandemly arranged on the chromosome 7,suggesting that they may be derived from recent gene duplication events.In addition,the TPT domain is highly conserved in rice andArabidopsis,while the Nterminal coding sequence of GPTs are obviously different between rice andArabidopsis(Fig.S9B),indicating that GPTs may have distinct functions in two species.

Fig.4.Defective tapetal PCD in osgpt1.(A–C,G–I)transferase-mediated dUTP nick-end labeling(TUNEL)assay of wild-type anthers from stage 7 to stage 11;(D–F,J–L)TUNEL-assay of osgpt1 anthers from stage 7 to stage 11.Red fluorescence indicates nuclei stained with propidium iodide(PI),and yellow fluorescence is the overlap of TUNEL and PI signals.Bp,bicellular pollen;DBp,defective bicellular pollen;Dy,dyad cell;Mc,meiotic cell;Msp,microspore;St7 to St11,stage 7 to stage 11;T,tapetum;Tds,tetrads.Scale bars,25 μm.

3.6.Expression pattern and subcellular localization of OsGPT1

Fig.5.Transmission electron microscopy(TEM)and scanning electron microscopy(SEM)assay of the Ubisch body and pollen exine in wild-type and osgpt1 mutant.(A-L)TEM images showing Ubisch body(A,B,E,F,I,J)and pollen exine(C,D,G,H,K,L)of wild-type(A,C,E,G,I,K)and osgpt1(B,D,F,H,J,L)at different developmental stages.(M–P)SEM images showing Ubisch body(M,N)and pollen exine(O,P)at stage 12 in the wild-type(M,O)and osgpt1(N,P)anthers.Ba,bacula;F,foot layer;In,intine;St10 to St12,stage 10 to stage 12;Te,tectum;Ub,Ubisch body.Scale bars,0.5 μm in(A,B,E–L),2 μm in(M,N),and 1 μm in(C,D,O,P).

To understand its role in male reproduction,we conducted the reverse transcription-quantitative PCR(RT-qPCR)to compare the spatiotemporal expression patterns ofOsGPT1andOsGPT2during rice development.The results showed that bothOsGPT1andOsGPT2were widely expressed in vegetative and reproductive tissues,including anthers from stage 7 to stage 13,but the expression levels ofOsGPT1were obviously higher thanOsGPT2(Fig.7A).During anther development,the transcription level ofOsGPT1reached the maximum in the stage 9 anther when microspores were released from the tetrad(Fig.7A).Consistently,inosgpt1lines complemented by theOsGPT1pro:OsGPT1gDNA-GUSconstruct,GUS signals were observed in the anther wall from stage 7 to stage 13(Fig.7B).OsGPT1 expressions were also detected in pollen grains at stage 10 to stage 13,and reached the highest level at stage 13(Fig.7C-E).To more precisely analyze the spatial distribution of OsGPT1,we detected the GFP signals inosgpt1lines complemented byOsGPT1pro:OsGPT1gDNA-GFPthrough confocal microscopy.During early developmental stages,GFP fluorescence was detected in all anther wall layers,including epidermis,endothecium,middle layers and tapetum,as well as the vascular bundle in the anther,but not in meiotic cells or microspores at stage 9(Fig.8A).At later stages,the fluorescence appeared in microspores and mature pollen grains(Fig.9).To determine the subcellular localization of OsGPT1,a translational fusion of the full-lengthOsGPT1coding region and GFP under the control of the double 35S promoter was constructed and introduced into rice protoplasts isolated from etiolated rice hypocotyls.As expected,the OsGPT1-GFP fusion protein was distributed exclusively in the plastids(Fig.8B).In the endothecium layer ofosgpt1lines carryingOsGPT1pro:OsGPT1gDNA-GFP,we found the GFP fluorescence overlapped with the auto-fluorescence of plastids(Fig.8A).In microspores at stage 10,GFP signals appeared in a small number of dots and gradual increased until became filled with the cytoplasm of mature pollen grains,similar to starch granules shown by I2-KI staining(Fig.9).All the above results suggested that OsGPT1 was localized in the plastid,consistent with thatOsGPT1encodes a putative plastid envelope translocator.

Fig.6.Map-based cloning of OsGPT1.(A)Initial mapping of the OsGPT1 gene on chromosome 8.Positions of the molecular marker are noted.(B)A schematic representation of the OsGPT1.The mutation site in osgpt1 and putative TPT domain of the OsGPT1 protein are indicated.

Fig.7.Expression analysis of OsGPT1.Analysis of OsGPT1 and OsGPT2s transcripts in different tissues by RT-qPCR.*,P＜0.1;**,P＜0.01;***,P＜0.001.Error bars indicate the standard deviations of three biological replicates.(B–E)GUS staining anthers(B),microspores(C,D)and pollen grains(E)at different developmental stages in osgpt1 lines complemented by OsGPT1pro:OsGPT1gDNA-GUS.St7 to St13,stage 7 to stage 13.Scale bars,200 μm in(B)and 50 μm in(C–E).

Fig.8.OsGPT1 is localized in the plastid.(A)Representative confocal images of anthers at stage 7 to stage 10 in the osgpt1 lines complemented by OsGPT1pro:OsGPT1gDNAGFP.(B)Representative confocal images of rice protoplasts expressing 2×35Spro:OsGPT1CDS-GFP or 2×35Spro:GFP.Showing images from GFP fluorescence(green),chlorophyll fluorescence(red),bright field and the merged image,respectively.E,epidermis;En,endothecium;ML,middle layer;St7 to St10,stage 7 to stage 10;T,tapetum.Scale bars,25 μm in(A)and 5 μm in(B).

Fig.9.OsGPT1 is localized in pollen grains at later stages.(A)Representative confocal images of pollens at stage 10 to stage 13 in the osgpt1 lines complemented by OsGPT1pro:OsGPT1gDNA-GFP.(B)I2-KI staining of wild-type pollen grains at starch accumulation stages.Scale bars,10 μm.

3.7.Mutation in OsGPT1 affects the transcript levels of the genes associated with tapetum PCD and pollen wall formation

To further understand the impact of OsGPT1 on anther and pollen development,we compared the transcript levels of certain known rice genes associated with tapetal PCD and required for pollen exine formation in wild-type andosgpt1developmental anthers through RT-qPCR.The results showed that the expression of three key pollen wall formation genes,OsTKPR1,OsSTRL2andOsNP1which also necessary for the tapetal PCD,were significantly downregulated inosgpt1(Fig.10).In addition,the expression of three tapetal PCD regulators,OsAP37,OsTIP2andOsPTC2also decreased inosgpt1(Fig.10).These data suggested that disruption of OsGPT1 also affects the expression of genes associated with lipid biosynthesis and tapetum development.

4.Discussion

4.1.OsGPT1 regulates pollen development and maturation in a sporophytic controlled manner

In Arabidopsis,atgpt1shows severe defects in male and female gamete transmission,suggestingAtGPT1has indispensable roles in gametophyte development[16].However,in rice the absence ofOsGPT1does not interfere with male or female gamete transmission.Genetic analysis indicates that the mutant phenotype is controlled by sporophytic factors.The segregation ofosgpt1mutant in the progeny of selfing or backcrossing heterozygous plants is approximately 1/4,in line with the Mendelian inheritance of a single nuclear recessive gene mutation.osgpt1mutant plants pollinated with wild-type pollen are able to set seeds normally,suggesting the female gametes is properly developed.Meanwhile,pollen viability and seed setting ofosgpt1+/-heterozygotes are indistinguishable from wild-type(Fig.S2).Taken together,all these observations indicate that OsGPT1 controls microspore development and pollen maturation in a sporophytic regulation manner,although it was expressed in both the sporophytic anther layers and developing pollen grains.It is likely that the paralogs ofOsGPT1,OsGPT2shave compensate or different roles in the haploid microspores,albeit at a much lower expression level.

Fig.10.Analysis of alteration in expression of key regulatory genes involved in the tapetal PCD and pollen exine formation in osgpt1 developmental anthers by RT-qPCR.Error bars indicate the standard deviations of three biological replicates.

4.2.OsGPT1 and OsGPT2s may have redundant functions in pollen starch biosynthesis

Amyloplast,a kind of non-green plastids perform indispensable function in starch synthesis in storage tissues.Developing pollen grains accumulate storage compounds that are essential for pollen maturation,germination,pollen tube growth and successful fertilization[28,29].ADP-Glc,Glc1P or Glc6P can be taken up by the plastid as the precursor for starch synthesis[17].Glc6P is converted to Glc1P via the plastidic phosphoglucomutase(pPGM)and Glc1P is then converted to ADP-Glc by the ADP-glucose pyrophosphorylase(AGP).It has been proposed that in rice Glc6P is preferentially imported into the amyloplasts of pollen grains as a precursor of starch biosynthesis[30].OspPGMandOsAGPL4are highly expressed in anthers at late pollen developing stages.Mutations inOspPGMabolish starch biosynthesis and amyloplast formation in the pollen,leading to male sterility;while disruption ofOsAGPL4results in reduced starch accumulation in the pollen and the seed setting rate,suggesting that starch synthesis in the rice pollen may depend on Glc6P imported via GPTs[30].Previous studies indicate that those genes involved in the rice pollen starch biosynthesis and utilization are under the control of the gametophytic genome,includingOspPGM,OsAGPL4and the hexokinaseOsHXK5.Mutants of these genes are gametophytic mutants and segregate in 1:1:0(wild-type:heterozygote:homozygote)in the heterozygote selfing population.In addition,half of the pollen grains in heterozygotes of these mutants show defects in pollen starch accumulation and/or utilization[30,31].OsGPT1 is not detectable in the meiotic cells but appears in microspores and mature pollen grains(Figs.7C–E,9),indicating that the expression ofOsGPT1in the pollen is controlled by the haploid genome.Therefore,half of pollen grains inosgpt1+/-heterozygotes do not contain functional OsGPT1.In our study,the absence ofOsGPT1in the pollen does not affect starch granule formation and pollen vitality,which is evidenced by the fully fertile pollen and normal seed setting in theosgpt1+/-heterozygotes.The pollen abortion and failure in amyloplast genesis inosgpt1homozygotes are more likely the consequence of deficiencies in tapetal degeneration and pollen exine formation.Consistently,previous reports indicate that rice mutants defective in these aspects are normally arrested before or around the binucleate pollen stage when starch begin to accumulate[12,24,32].Additionally,pollen grains in mutants that are specifically affected in starch accumulation could develop normal but fail to fertilize,supporting that pollen abortion inosgpt1is not simply caused by lack of starch biosynthesis.However,OsGPT1 is localized in pollen amyloplasts and its protein level elevated accompanied with the increase of starch content during pollen development.The possibility that OsGPT1 is involved in the starch synthesis in pollen cannot be ruled out.It is likely that the defects of starch accumulation inosgpt1+/-pollen grains are rescued by its paralogs OsGPT2s.In Arabidopsis,AtGPT1 plays a special role in NADPH supply in plastids,whereas AtGPT2 contributes to starch biosynthesis and stress responses[21].It remains to be determined in rice whether OsGPT2s are responsible for the uptake of Glc6P for starch biosynthesis,and whether OsGPT1 and OsGPT2s have redundant functions in the gametophyte development.

4.3.OsGPT1 is required for proper tapetal function and degeneration

In Arabidopsis,atgpt1homozygotes are embryo lethal.Therefore,the function ofAtGPT1in sporophytic tissues of the anther is unclear.Our study shows that disruption ofOsGPT1leads to severe tapetal abnormalities,including reduced sporopollenin production and delayed tapetal PCD.In many species,a special structure called Ubisch body is produced on the inner face of the tapetal layer and is believed to carry sporopollenin.Inosgpt1,the number and density of Ubisch bodies are significantly decreased(Fig.5).Consistently,the pollen exine showed much thinner inosgpt1mutant than wild-type(Fig.5),suggesting the sporopollenin precursors generation could be reduced and-OsGPT1 might play a critical role in lipid biosynthesis.In the tapetum,the de novo biosynthesis of acyl-lipid derived components of sporopollenin initiates in plastids,using carbons transported from photosynthetic tissues[33].Plastidial acetyl-CoA is the building block of fatty acids,which could be synthesized from three metabolites:acetate,pyruvate and citrate.In non-green plastids,Glc6P can be converted into pyruvate via glycolytic reactions,which then serves as the precursor for fatty acid production.In addition,Glc6P is fed into the OPPP pathway to supply the redox equivalent NADPH for assimilation and biosynthesis in the whole cell.Our results suggest that in the tapetum,carbons imported by OsGPT1 are used for fatty acids and subsequent sporopollenin precursors biosynthesis,which is similar to the role of AtGPT1 in pollen lipid body formation.

Except for pollen exine formation,osgpt1exhibits an obvious defect in tapetum degeneration(Figs.2,3,4).It is yet unknown how OsGPT1 affect tapetal PCD.It is likely that the delayed degradation of the tapetal layer is caused by the abnormalities in the lipid metabolism.Several rice mutants defective in sporopollenin biosynthesis also exhibit defective tapetal PCD[11,12,24],suggesting that there may have an intrinsic link between the lipid metabolism and tapetal PCD.In non-green plastids,the reducing powers NADH and NADPH are produced by the pyruvate dehydrogenase complex(PDC)and the OPPP pathway respectively[33].And NADP/NADPH is important for fatty acid synthesis and detoxification of oxidative stresses in chloroplasts in the light[34,35].We propose that OsGPT1 may be mainly involved in the production of NADPH via the OPPP pathway in tapetum.Mutations in OsGPT1 may affect the reducing equivalents generation,therefore interferes with fatty acid and energy metabolism as well as the redox status in the tapetal cell,leading to abnormal tapetal PCD.

In conclusion,our study suggests that in rice OsGPT1 acts as a critical mediator to integrate carbohydrate and fatty acid metabolism in the tapetum,which is essential to coordinate the development of sporophytic tissues and male gametophytes.Our study together with a recent report highlight the importance of plastid metabolism in regulating the tapetal PCD.Zhu et al.indicate that deficiency in ZmGPAT6,a tapetum localized glycerol-3-phosphate acyltransferase(GPAT),disrupts the normal structure and function of endothecium chloroplasts[8].Mutation in ZmGPAT6 affects membrane lipid biosynthesis and aberrancy in the structure of endothecium chloroplasts,leading to metabolism reprogramming and excess ROS production which induces premature tapetal PCD[8].Future investigations will further clarify how tapetal plastid sugar and fatty acid metabolism control the progression of tapetal PCD.

CRediT authorship contribution statement

Wanqi Liangdesigned the research and supervised the project and experiments;Weidan Zhang,Huanjun Li,and Feiyang Xueperformed the experiments in this study.Wanqi Liang and Weidan Zhangwrote the manuscript.

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

The authors thank Zhijin Luo,Mingjiao Chen,and Xiaofei Chen for mutant screening,allelic test,and generation of F2populations for mapping.This work was supported the National Natural Science Foundation of China(U19A2031)and the National Key Research and Development Program of China(2016YFD0100903).

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2021.03.010.