GL9 from Oryza glumaepatula controls grain size and chalkiness in rice

Shojun Lin,Zupei Liu,Kui Zhng,Weifeng Yng,Penglin Zhn,Quny Tn,Yjun Gou,Shuipeng M,Xin Lun,Chubing Hung,Zhili Xio,Yunyun Liu,Bihung Zhu,Ruiqing Ling,Wenqi Zhou,Hito Zhu,Suhong Bu,Guifu Liu,Guiqun Zhng,*,Shokui Wng,*

a Guangdong Provincial Key Laboratory of Plant Molecular Breeding,State Key Laboratory for Conservation and Utilization of Subtropical Agro-Bioresources,South China Agricultural University,Guangzhou 510642,Guangdong,China

b Guangdong Laboratory for Lingnan Modern Agriculture,Guangzhou 510630,Guangdong,China

Keywords:Oryza glumaepatula GL9 Grain size Grain chalkiness Single-segment substitution line

ABSTRACT Grain size is a key factor influencing grain yield and appearance quality in rice.We identified twelve quantitative trait loci(QTL)for grain length(GL),nine for grain width(GW),and nine for 1000-kernel weight(TKW)using GLU-SSSLs,which are single-segment substitution lines with Oryza glumaepatula as donor parent and Huajingxian 74(HJX74)as recipient parent.Among the QTL,qGL1-2,qGL1-4,qGL9-2,qGW2-2,qGW9-1 and qTKW9-2 contributed to high grain yield.GL9 was identified as a candidate gene for qGL9-2 by map-based cloning and sequencing,and is a novel allele of GS9.The kernel of NIL-gl9 was slenderer and longer than that of HJX74,and the TKW and grain yield per plant of NIL-gl9 were higher than those of HJX74.The proportion of grain chalkiness of NIL-gl9 was much lower than that of HJX74.Thus,gl9 increased grain yield and appearance quality simultaneously.Three pyramid lines,NIL-gs3/gl9,NIL-GW7/gl9 and NIL-gw8/gl9,were developed and the kernel of each was longer than that of the corresponding recipient parent lines.The gl9 allele may be beneficial for breeding rice varieties with high grain yield and good appearance quality.

1.Introduction

Rice(Oryza sativa L.)plays an indispensable part in world food security[1,2].Grain yield of rice is a complex agronomic trait controlled by multiple genes.The three key factors of grain yield are grain weight,grain number per panicle,and number of panicles per plant[3,4].Grain size is a major determinant of grain weight and is determined by GL,GW and grain thickness(GT)[4,5].

During the past twenty years,a number of QTL for rice grain size and weight were detected using F2population,recombinant inbred line(RIL),backcross inbred line(BIL),chromosome-segment substitution line(CSSL),introgression line(IL),near isogenic line(NIL)and single-segment substitution line(SSSL)[3,6,7].Only a small number of relevant genes have been cloned in rice as major grain-size regulators,such as the GRAIN WIDTH 2(GW2),GRAIN SIZE 3(GS3),GRAIN LENGTH 3.1(GL3.1),GRAIN WIDTH 5(GW5),GRAIN SIZE 5(GS5),GRAIN WIDTH 6(GW6),GRAIN WEIGHT 6a(GW6a),GRAIN LENGTH AND WEIGHT 7(GLW7),GRAIN WIDTH 8(GW8),GRAIN WIDTH 7/GRAIN LENGTH 7(GW7/GL7),GRAIN SHAPE GENE ON CHROMOSOME 9(GS9),GRAIN WIDTH 10(GW10)and GRAIN LENGTH 10(GL10)[4,8–20].Among them,the GS3,GW8,GW7/GL7 and GS9 could regulate grain size and appearance quality in rice simultaneously and have great potential to be widely applied in rice breeding by design[9,16–19].The GS9 acts as a transcriptional activator to regulate grain shape and appearance quality in rice[19].The gs9 from N138 has one 7-kb insertion in the second exon of GS9,the insertion results in loss-of-function of GS9 and hence the slender kernels in N138[19].The NIL gs9/qpe9-1 could improve grain appearance significantly with no alteration in rice yield and panicle architecture compared with their recurrent parental lines[21].

Many QTL/genes regulating grain size have been cloned in cultivated rice just as mentioned above.Wild rice carries alleles that are valuable genome resources for rice breeding[22–27].In recent years,some researchers have focused on exploiting the germplasm resources from wild rice in rice breeding[24,28–32].They have found that O.glumaepatula has potential for high yield and has strong outcrossing ability[2,4,28,30–33].Several QTL controlling grain size and weight have been identified,but a limited number of natural variation genes for grain size have been cloned[3].And few yield-related QTL from wild rice species O.glumaepatula have been reported[24,28,30–32].In addition,the huge genetic distance between rice and wild species limits the application of genes in wild rice which controlled excellent characteristics.The SSSLs will provide a solution for the detection and application of the desirable QTL/genes from wild species.We have constructed a SSSL library of 2360 SSSLs using an elite indica variety HJX74 as recipient parent,and 43 accessions from seven rice species of AA genome as donor parents.Among them,more than 600 SSSLs are derived from wild species[6,7]and a set of 73 SSSLs derived from the cross between O.glumaepatula(the donor parent)and HJX74 are referred to as GLU-SSSLs.In this study,we mapped the QTL for grain size and grain weight using the GLU-SSSLs.More addition,a new allele of grain size gene was cloned and introgressed into indica rice cultivars and the relationships between gl9 and other grain size genes were tested.The results show that the gl9 is the perfect germplasms for breeding rice varieties high yield and superior grain appearance.

2.Materials and methods

2.1.Plant materials and field experiment

73 GLU-SSSLs were employed for QTL identification.All the GLU-SSSLs with substitution segment length of 431.90 Mb on 12 chromosomes were selected to perform QTL analysis of grain size and grain weight(Table S1).All the materials were planted in the paddy fields at South China Agricultural University in Guangzhou,Guangdong,China(23°07′N,113°15′E).All materials were planted in the first cropping season(FCS)and the second cropping season(SCS)in each year.Seedling transplanting,standardized cultivation,and pest and disease control were performed as previously described[2,4].

2.2.Substitution mapping of QTL for grain size and grain weight

The GL,GW,and TKW of 73 GLU-SSSLs were measured in the second cropping season of 2016(2016 SCS).Then 18 GLU-SSSLs(Table S2)with significantly different phenotypes from HJX74 were selected and observed in 2017 FCS and 2017 SCS.QTL that were detected in at least two cropping seasons were considered reliable[34].Only GLU-SSSLs displaying QTL for grain size and grain weight for more than three cropping seasons were used.

Substituted segment lengths in GLU-SSSLs including the minimum length(Lmin),the maximum length(Lmax)and the estimated length(Lest)were calculated as previously described[2,6,28,29].QTL for grain size and grain weight were mapped by substitution mapping[35,36].

2.3.Genetic analysis and fine mapping of GL9

GLU-SSSL S62 which with greater grain length than that of HJX74 was chosen to develop an F2population by crossing with HJX74.A set of 120 F2plants were used for genetic and linkage analysis.An additional 1200 F2:3plants and 3200 F3:4plants were used for fine mapping,using selfed progeny to select recombinants.Among the location-based DNA markers used in this study,SSR markers prefixed‘‘RM”were retrieved from a database(https://archive.gramene.org/markers/)and markers named‘‘PSM”and‘‘InDel”were designed using Primer Premier 5.0(Carnegie Institution of Washington,Stanford,CA,USA)based on genomic sequence variations between HJX74 and NIL-gl9.All primer sequences are listed in Table S3.

2.4.Quantitative real-time PCR,gene cloning,and sequence analysis

Total RNA was isolated using TRIzol(Invitrogen Corporation,Carlsbad,CA,USA).First-strand cDNA was synthesized by reverse transcription and the transcript levels of candidate genes were measured by quantitative real-time PCR(qRT-PCR)as described previously[4].Nucleotide and protein sequences of candidate genes based on the Rice Genome Annotation Project(https://rice.plantbiology.msu.edu/)were retrieved from NCBI(https://www.ncbi.nlm.nih.gov/).Based on the genome sequences of HJX74[37],specific primers were designed to clone candidate genes using KOD FX DNA polymerase(TOYOBO Co.,ltd.,Osaka,Japan).The software DNAMAN(https://www.lynnon.com)was used to align the coding region of the candidate genes between HJX74 and NIL-gl9.Plant tissues at booting stage of HJX74 and NIL-gl9 were used as materials for expression analysis and cloning of genes.All the primer sequences for the experiment were listed in Table S4.

2.5.Vector construction and genetic transformation analysis

To produce CRISPR/Cas9 plasmids,we designed two targets with specific recognition on the first exon of the gl9 candidate gene.Two fragments,target 1 and target 2,were cloned into the vector SK-gRNA separately.Then Kpn I/Nhe I,Xba I/Bgl II and Kpn I/BamH I were used to digest target 1,target 2,and the pC1300-Cas9 vector,respectively.In addition,the two targets were incorporated into the pC1300-Cas9 vector,and the plasmid was introduced into rice cultivar HJX74 by Agrobacterium-mediated transformation.All the primer sequences for the experiment are listed in Table S4.

2.6.Phenotype investigation and data processing

GL,GW,TKW,number of panicles per plant(NPPP),panicle length(PL),grain number per panicle(GNPP),seed setting rate(SSR),grain yield per plant(GYPP),and plant height(PH)were recorded at rice maturity stage,and heading date(HD)was recorded when the first spikelet emerged about 2 cm long for a plant.Grain from the upper end of panicles of each plant were measured for GL and GW with the software Microtek ScanWizard EZ scanner V-2.140 and Wan Shen grain analyzer(Hangzhou Wanshen Testing Technology Co.,Ltd.,Hangzhou,Zhejiang,China).One hundred kernels were weighed and the weight was converted to TKW,using at least 3 repeats.The mean weight of filled grain per plant for 20 individual plants was used to represent grain yield per plant.

The arcsine square root transformation was applied to phenotypic values and one-way ANOVA.Dunnett t-test,Student’s t-test and Duncan test were employed to analyze phenotypic data.The phenotypic data analysis was performed with IBM SPSS Statistics 26(https://www.ibm.com/products/spss-statistics),and figures were drawn with OriginPro 9.0(https://www.originlab.com).Additive effect was calculated following Tan et al.[38].

3.Results

3.1.Detection of QTL for rice grain size and weight

To identify QTL for rice agronomic traits,73 GLU-SSSLs were employed.All the substituted segments in the GLU-SSSLs covered 54.45% of the rice genome(Table S1).The mean values of GL,GW and TKW were respectively 8.34 mm,2.65 mm and 21.92 g for HJX74,and 7.92–8.82 mm,2.31–2.80 mm and 18.46–23.60 g for the 73 GLU-SSSLs(Fig.S1A–C).

Only 18 GLU-SSSLs with trait values significantly different from those of HJX74 in more than two environments were considered to carry the corresponding QTL(Tables S2,S5–S7).Thirty QTL for grain size and grain weight were detected on seven different chromosomes of the 18 SSSLs:twelve for GL,nine for GW,and nine for TKW,and the interval variation of the QTL ranged from 0.50 to 17.26 Mb(Fig.1;Table 1).The regions of RM403–RM543 and PSM423–end on chromosome 1,RM413–RM161 on chromosome 5,and PSM398–PSM161 on chromosome 9 were shared by QTL for GL,GW,and TKW in the same condition(Fig.1).Among the 30 QTL,only qGL1-2,qGL1-4,qGL9-2,qGW2-2,qGW9-1,and qTKW9-

2 showed positive additive effects,compared with the corresponding traits of HJX74(Table 1).qGL9-2 and qTKW9-2 were located in the same region and both showed very high positive additive effects.This finding strongly implied that the QTL in the RM434–RM242 region on chromosome 9 carried by GLU-SSSL S62 and S64 could improve rice appearance quality and yield simultaneously.

Fig.1.Chromosomal distribution of 30 QTL for grain size in GLU-SSSLs.The bars in black,red,and blue represent QTL for grain length,grain width,and 1000-kernel weight,respectively.Each bar to the right of each chromosome represents the estimated interval of the QTL.Chr.,chromosome.

Table 1 QTL for GL,GW,and TKW and their effects in GLU-SSSLs.

3.2.Map-based cloning of GL9

To identify genes underlying qGL9-2 and qTKW9-2 on GLU-SSSL S62,an F2population was constructed by crossing HJX74 with GLU-SSSL S62.In the 120 F2plants,the segregation of genotypes followed a 1:2:1 distribution(χ2=1.11＜χ20.01,2=9.21)and grain length was linked with the genotype(Fig.2A).It was also indicated that a semidominant qGL9-2 allele from O.glumaepatula controlled grain length,and the gene controlling grain length in this location was named GL9.

The near-isogenic line gl9(NIL-gl9)was developed from GLUSSSL S62,and NIL-gl9 plants formed kernel that was substantially longer and narrower than that of HJX74(NIL-GL9)(Fig.2B–E).The TKW and the grain yield per plant of NIL-gl9 were greater than those of HJX74(Fig.2F,G).NIL-gl9 had more panicles per plant and less grain per panicle than HJX74.There was no significant difference between HJX74 and NIL-gl9 in panicle length,seed setting rate,heading date,or plant height(Fig.S2).These findings indicated that GL9 controls grain size and weight in rice.

A localized high-resolution mapping based on 1200 F2:3segregants allowed the localization of GL9 between markers RM6235 and RM7175(Fig.3B).The recombinant plants selected from the 3200 F3:4population allowed this region to be narrowed to a 12.15-kb interval flanked by markers ID09L3 and ID09L6 that contained only one predicted open reading frame(ORF),LOC_Os09g27590(Fig.3C).LOC_Os09g27590 encodes GS9 protein[23].Sequence comparison of GL9 between HJX74 and NIL-gl9 revealed a 12-bp deletion in the first exon(InDel1),one single nucleotide polymorphism(SNP)in the third exon(named SNP1)and three SNPs in the fourth exon(named SNP2,SNP3,SNP4 respectively).The sequence variations resulted in a 4-amino-acid residue deletion and a 2-residue substitution in NIL-gl9(Fig.3D).The expression profiles of GL9 in various tissues did not differ between HJX74 and NIL-gl9.Expression of GL9 gradually increased during early panicle development(＜P2)and gradually decreased during later panicle development(＞P2)(Fig.3E).

3.3.GL9 controls grain size,weight,and chalkiness in rice

GL9 knockout transgenic plants were generated by the CRISPR/Cas9 genome editing system in HJX74.Two homozygous mutants,KO-GL9-1 and KO-GL9-2 with a 51-bp deletion and a 2-bp deletion in the first exon of GL9 respectively,were selected for further study(Fig.4A).KO-GL9-1 and the KO-GL9-2 produced narrower and longer rice grain than that of the HJX74(Fig.4C–E).The 1000-kernel weights of KO-GL9-1 and KO-GL9-2 were greater than that of HJX74(Fig.4D).Both showed a significant decrease in grain number per panicle,heading date,and plant height,and a significant increase in number of panicles per plant in comparison with HJX74(Fig.S3).All of these results indicated that LOC_Os09g27590 is identical to GL9,which controls grain size and grain weight in rice.NIL-gl9,KO-GL9-1 and KO-GL9-2 produced longer and slenderer milled grain than did HJX74,and milled grain of KO-GL9-1 and KO-GL9-2 was longer and slenderer than that of NIL-gl9(Fig.5A–C).The proportions of chalky grain and grain chalkiness of NIL-gl9,KO-GL9-1 and KO-GL9-2 were significantly lower than those of HJX74,and the KO-GL9-2 showed the lowest proportions of chalky grain and grain chalkiness among HJX74,NIL-gl9,and KO-GL9-1(Fig.5D–E).All these results showed that GL9 controls grain size,weight,and chalkiness in rice.

Fig.2.Morphological characteristics of HJX74 and NIL-gl9.(A)Frequency distribution of grain length in F2 population.(B)Morphology of HJX74 and NIL-gl9 plants.Scale bar,15 cm.(C)Rice grain of HJX74 and NIL-gl9.Scale bars,1 cm.(D–G)Statistics of grain length(D),grain width(E),1000-kernel weight(F),and grain yield per plant(G).Values are means±SE of two cropping season.Two-tailed,two-sample Student’s t-test.

Fig.3.Map-based cloning of GL9.(A)The initial substituted segment of GL9.Scale bar,1 Mb.(B,C)Substitution mapping of GL9.Values are means±SE,one-way ANOVA,Duncan,α=0.01.(D)Schematic of the variations in GL9 between HJX74 and NIL-gl9.(E)Relative expression of GL9 in tissues of HJX74 and NIL-gl9.P0.5–P11,young panicles with mean lengths of about 0.5,1,2,5,and 11 cm.R,root;ST,stem;L,leaf;LS,leaf sheath.

3.4.Breeding by design for grain size and appearance quality using GL9

NIL-gl9(donor)was crossed with NIL-gs3,NIL-GW7,and NILgw8(recipients)to generate the respective two-gene pyramid lines NIL-gs3/gl9,NIL-GW7/gl9 and NIL-gw8/gl9 in the HJX74 background(Fig.6A;Table S8).The GW of NIL-gs3/gl9 was comparable to that of NIL-gs3,whereas the GL of NIL-gs3/gl9 was longer than that of NIL-gs3,leading to a higher RLW and TKW(Fig.6B–E).Although the GW of NIL-GW7 was lower than that of NIL-GW7/gl9,the GL of NIL-GW7/gl9 was longer than that of NIL-GW7,leading to a higher TKW(Fig.6F–I).Although the GW of NIL-gw8/gl9 was greater than that of NIL-gw8,the GL of NIL-gw8/gl9 was greater than that of NIL-gw8,leading to a higher RLW and TKW(Fig.6J–M).The increased GL of the three pyramid lines improved their appearance quality.These results show that GL9 might be used in breeding by design for good appearance quality and high yield in rice.

4.Discussion

Several QTL controlling grain size and weight have been identified from wild rice,but a limited number of genes for grain size have been cloned[3].Few yield-related QTL from wild rice species O.glumaepatula have been reported[24,28,30–32].We detected 30 QTL for grain size and grain weight using GLU-SSSLs,with only a few co-localizing with known genes(Fig.1).They will provide us with the perfect materials to explore the mechanism of grain size and grain weight regulation in the future.Most of the 30 QTL were identified on chromosome 1 and none were detected on chromosomes 7,8,10,or 11.The chromosomal coverage of chromosome 1 was 89.78%,whereas the segment coverages in most of the 12 chromosomes were less than 60.00%(Table S1).We infer that QTL identification in SSSLs was affected by the coverage of the substituted segments.

Fig.4.GL9 controls grain size and weight in rice.(A)Variation in two GL9-knockout mutants.(B)Morphology of HJX74,KO-GL9-1 and KO-GL9-2 plants.Scale bar,15 cm.(C)Grain of HJX74,KO-GL9-1 and KO-GL9-2.Scale bar,1 cm.(D–F)Statistics of grain length(D),grain width(E),and 1000-kernel weight(F)in HJX74,KO-GL9-1,and KO-GL9-2 plants.Values are means±SE over two cropping seasons.Dunnett t-test,***,P≤0.001.

Rice grain weight is determined mainly by grain size[39].There is a pleiotropism in the inheritance of grain size in rice[3].Several QTL controlling grain-size traits were identified on the same substitution segment(Table 1;Fig.1).Ten QTL for grain size and grain weight were identified within an interval of less than 2.00 Mb(Table 1).We found that 6 QTL,qGL1-2,qGL1-4,qGL9-2,qGW2-2,qGW9-1 and qTKW9-2 acted as positive regulators of grain yield(Table 1).qGL5-1,qGW5-1 and qTKW5-1 were detected on the same region of chromosome 5 between markers RM413 and RM161,which contains two grain-size genes,GS5 and GW5[19,21].qGL9-2 and qTKW9-2 on chromosome 9 shared the same region with GS9[19]and then gl9 was cloned from O.glumaepatula contributing to narrower and longer grain and higher grain weight.GL9 is a novel allele of GS9 and is a key regulatory factor for grain size and grain weight.

NIL-gl9,KO-GL9-1,and KO-GL9-2 produce slenderer and longer grain than HJX74,but the GW of KO-GL9-1 and KO-GL9-2 is lower than that of NIL-gl9(Fig.S4).The proportions of chalky grain and grain chalkiness of NIL-gl9,KO-GL9-1 and KO-GL9-2 were lower than those of HJX74,but the proportions of chalky grain and grain chalkiness of NIL-gl9 were higher than those of KO-GL9-1 and of KO-GL9-2(Fig.5D–E).These results indicate that GL9 is a partial-loss-of-function protein in NIL-gl9 but a full-loss-offunction protein in KO-GL9-1 and KO-GL9-2.Slender and longer grain is commonly accompanied by greater plant height in rice.The greater plant height is one of the major reasons for lodging,which is the main limiting factor for high yield[40].KO-GL9-1 and KO-GL9-2 had slenderer and longer grain than HJX74 and their heights were lower than that of HJX74(Figs.4B–E,S3A).It may be a good comprehensive agronomic trait for molecular breeding by design in rice.

Fig.5.Chalkiness traits of milled grain in HJX74,NIL-gl9,KO-GL9-1,and KO-GL9-2.(A)Milled grain of HJX74,NIL-gl9,KO-GL9-1,and KO-GL9-2.Scale bar,1 cm.(B–E)Comparison of milled grain length(B),milled grain width(C),proportion of chalky grain(D)and proportion of grain chalkiness(E)in HJX74,NIL-gl9,KO-GL9-1,and KO-GL9-2.Values are means±SE,one-way ANOVA,Duncan,α=0.01.

CSSLs,ILs,NILs and SSSLs allow more precise analysis of target QTL and the positional cloning of target genes[7].In the past two decades,we have developed an SSSL library to investigate rice breeding by design[7].We have detected QTL for complex traits[2,38,41–49],cloned target genes[4,13,16,20,50],analyzed the allelic variations[51–56],and performed breeding by design by pyramiding target genes using the SSSL library[2,53,57–59].In the present study,we identified 30 QTL for grain size and weight(Table 1;Fig.1),and cloned GL9(Figs.2,3 and 4).Based on the SSSL library,we generated NIL-gs3/gl9,NIL-GW7/gl9 and NIL-gw8/gl9 double-gene pyramid lines by crossing NIL-gl9 with respectively NIL-gs3,NIL-GW7,and NIL-gw8 in the HJX74 background(Fig.6A).Compared with NIL-gs3 and NIL-gw8,the corresponding pyramid lines have long grain and a high RLW(Fig.6).Simiao Rice,a rice cultivar,is popular in the south of China and has high quality[60].The key phenotype indicators of Simiao Rice are a RLW of at least 3.50 and a TKW of at most 21.00 g,or a RLW of at least 4.00 and a TKW of at most 23.00 g[60].The NIL-gw8/gl9 pyramid line meets these indicators and NIL-gs3/gl9 and NIL-GW7/gl9 pyramid lines have the potential to meet them.Thus,our work provides a target gene gl9 for good quality and high-yield rice breeding by design.

Fig.6.Increasing grain size and grain weight by QTL pyramiding.(A)Grain size of parents and pyramid lines.Scale bar,1 cm.(B–E)Statistics of grain length(B),grain width(C),ratio of grain length to width(D),and 1000-kernel weight(E)in NIL-gs3/gl9 and its parents.(F–I)Statistics of grain length(F),grain width(G),ratio of grain length to width(H),and 1000-kernel weight(I)in NIL-GW7/gl9 and its parents.(J–M)Statistics of grain length(J),grain width(K),ratio of grain length to width(L),and 1000-kernel weight(M)in NIL-gw8/gl9 and its parents.Values are means±SE over two cropping seasons.One-way ANOVA,Duncan,α=0.01.GL,grain length;GW,grain width;RLW,ratio of grain length to width;TKW,1000-kernel weight.

5.Conclusions

Twelve QTL for GL,nine for GW,and nine for TKW were identified on 7 chromosomes using GLU-SSSLs.GL9 as a candidate gene of qGL9-2 was cloned by substitution mapping.The gl9 from O.glumaepatula positively controls GL and GW,and negatively regulates chalkiness in rice.Pyramid lines,which were generated by crossing the NIL-gl9 separately with NIL-gs3,NIL-GW7,and NILgw8,show longer grain with a high RLW than the corresponding recipients.These results provide a target gene gl9 for molecular breeding by design in rice.

CRediT authorship contribution statement

Shaojun Lin:Conceptualization,Data curation,Methodology,Investigation,Validation,Writing–original draft.Zupei Liu:Conceptualization,Methodology,Validation,Writing–original draft.Kui Zhang:Investigation,Validation.Weifeng Yang:Methodology,Investigation.Penglin Zhan:Methodology,Investigation.Quanya Tan:Methodology,Investigation.Yajun Gou:Investigation,Validation.Shuaipeng Ma:Methodology.Xin Luan:Methodology.Chubing Huang:Investigation.Zhili Xiao:Investigation.Yuanyuan Liu:Investigation.Bihuang Zhu:Investigation.Ruiqing Liang:Investigation.Wenqi Zhou:Investigation.Haitao Zhu:Writing–review&editing.Suhong Bu:Writing–review&editing.Guifu Liu:Writing–review&editing.Guiquan Zhang:Conceptualization,Project administration,Resources,Supervision.Shaokui Wang:Conceptualization,Project administration,Resources,Supervision,Writing–original draft,Writing–review & editing.

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

This study was supported by the major science and technology research projects of Guangdong Laboratory for Lingnan Modern Agriculture(NT2021001),the Key Projects of Basic Research and Applied Basic Research of Guangdong Province(2019B030302006),the National Natural Science Foundation of China(32072040,31622041),and the National Innovation and Entrepreneurship Training Program for College Students(202110564045).

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2022.06.006.