Genome-wide association study identifies novel candidate loci or genes affecting stalk strength in maize

Shuhui Xu,Xio Tng,Xiomin Zhng,Houmio Wng,Weidong Ji,Chenwu Xu,Zefeng Yng,*,Pengcheng Li,*

a Jiangsu Key Laboratory of Crop Genetics and Physiology/Key Laboratory of Plant Functional Genomics of the Ministry of Education/Jiangsu Key Laboratory of Crop Genomics and Molecular Breeding,Yangzhou University,Yangzhou 225009,Jiangsu,China

b Jiangsu Co-Innovation Center for Modern Production Technology of Grain Crops,Yangzhou University,Yangzhou 225009,Jiangsu,China

c National Maize Improvement Center of China,China Agricultural University,Beijing 100193,China

d State Key Laboratory of Crop Stress Adaptation and Improvement,Key Laboratory of Plant Stress Biology,School of Life Sciences,Henan University,Kaifeng 475001,Henan,China

Keywords:Stalk strength Maize GWAS Candidate genes Lodging resistance

ABSTRACT Stalk strength increases resistance to stalk lodging,which causes maize(Zea mays L.)production losses worldwide.The genetic mechanisms regulating stalk strength remain unclear.In this study,three stalk strength-related traits(rind penetrometer resistance,stalk crushing strength,and stalk bending strength)and four plant architecture traits(plant height,ear height,stem diameter,stem length)were measured in three field trials.Substantial phenotypic variation was detected for these traits.A genome-wide association study(GWAS)was conducted using general and mixed linear models and 372,331 single-nucleotide polymorphisms(SNPs).A total of 94 quantitative trait loci including 241 SNPs were detected.By combining the GWAS data with public gene expression data,56 candidate genes within 50 kb of the significant SNPs were identified,including genes encoding flavonol synthase(GRMZM2G069298,ZmFLS2),nitrate reductase(GRMZM5G878558,ZmNR2),glucose-1-phosphate adenylyltransferase(GRMZM2G027955),and laccase(GRMZM2G447271).Resequencing GRMZM2G069298 and GRMZM5G878558 in all tested lines revealed respectively 47 and 2 variants associated with RPR.Comparison of the RPR of the zmnr2 EMS mutant and the wild-type plant under high-and low-nitrogen conditions verified the GRMZM5G878558 function.These findings may be useful for clarifying the genetic basis of stalk strength.The identified candidate genes and variants may be useful for the genetic improvement of maize lodging resistance.

1.Introduction

Lodging disrupts the spatial distribution of plants,resulting in inhibited water and nutrient transport as well as relatively inefficient mechanical harvesting and decreased yields[1].Root lodging and stalk lodging are two major types of lodging,and stalk lodging involves the breakage of the stalk at or below the ear,which is estimated to decrease the total annual global maize yield by 5%–20%[2,3].Lodging is caused primarily by adverse weather conditions(excessive rainfall and strong winds)[4],but the overuse of nitrogen fertilizers and high plant densities can increase the likelihood of lodging[5–7].Identifying the genetic basis of stalk strength is desirable for improving stalk lodging resistance to enhance crop performance.

Stalk strength,one of the main determinants of stalk lodging,is influenced by stalk morphology,including stalk diameter and rind thickness,mechanical strength,and chemical composition including cellulose and lignin contents[8].With respect to the morphology of plants and internodes,stalk lodging is correlated with plant height,ear height,number of internodes above the ear,stalk diameter,and length of internodes below the ear[9].Maize(Zea mays L.)provides＞30% of the calories in the human diet[10].Since the Green Revolution,breeding semi-dwarf varieties with increased lodging resistance has increased grain yields.Reducing plant and ear height and increasing lodging resistance is a continuous goal for maize breeders.In maize SPL12 acts upstream of D1 to regulate GA biosynthesis and thereby plant and ear height.Overexpression lines of ZmSPL12 reduced plant and ear height and increased stalk strength,facilitating high-density planting and increased yield in maize[11].Earlier research[12,13]showed that stem basal internode diameter(SD)and length(SL)influence maize stem lodging resistance.Shortened basal internodes resulting from SBI expression increased the lodging resistance of rice[12].Rind penetrometer resistance(RPR),stalk bending strength(SBS),and stalk crushing strength(SCS)are closely associated with stalk strength and stalk lodging under field conditions[14–16].Stalk rind strength is believed to be responsible for 50%–80% of stalk strength,making it a determinant of stalk lodging resistance[14,17].Because RPR-based methods can efficiently and nondestructively assess stalk strength,they have been widely used for estimating stalk lodging resistance in large-scale genetic analyses of maize[18–20].

Several genetic studies[3,18,21]have identified numerous quantitative trait loci(QTL)or single-nucleotide polymorphisms(SNPs)associated with stalk strength traits.A total of 26 QTL associated with RPR were detected by linkage mapping in four maize F2:3populations[3].Most of these QTL explained＜15% of phenotypic variation.Nine and seven QTL for RPR were identified in two additional sets of recombinant inbred lines(RILs)[18,21].The major QTL,qRPR3-1,was localized to a 3.1-Mb interval that contained four potential candidate genes contributing to the biosynthesis of cell wall components[18].In another study,evaluation of a maize nested association mapping panel and an intermated B73×Mo17(IBM)family detected 18 family-nested QTL and 141 significant associations with RPR[20].A multi-locus GWAS was conducted for SD,SBS,and RPR in 257 inbred lines,and 29,34,and 48 loci were commonly revealed by multiple methods or across multiple environments to be associated with SD,SBS,and RPR,respectively[22].Respectively 14,14,and 16 QTL were identified for SD,SBS,and RPR by single-environment mapping in an IBM Syn10 DH population[23].A major QTL for RPR was cloned by combining map-based cloning with association mapping,and stiff1 was identified as a novel gene increasing stalk strength by its effects on cell wall cellulose and lignin contents[19].Another study[6]showed that miRNA528 affects maize lodging resistance by regulating lignin biosynthesis under nitrogen-luxury conditions.Transcriptomic and metabolomic analyses of lodgingresistant and lodging-sensitive lines revealed differentially expressed genes associated with cell wall lignin and polysaccharide biosynthesis[14,24].These studies provided researchers with a wealth of information about genomic regions and candidate genes controlling stalk strength.However,the molecular genetic mechanism governing stalk strength in maize remains largely unknown,and more genetic markers need to be mined to improve stalk quality by molecular breeding.

In this study,three stalk strength-related traits and four plant architecture traits were evaluated in three field trials.The objectives were to(i)identify SNPs associated with lodging resistancerelated traits;(ii)combine GWAS with gene expression profiling to identify candidate genes;and(iii)identify natural sequence variations in these candidate genes.

2.Materials and methods

2.1.Plant materials and phenotyping

The panel used in this study comprised 345 inbred lines collected from five heterotic groups in China:Reid,Lancaster,Tangsipingtou,Zi330,and mixed group[25–27].The inbred lines were grown in three field trials in Sanya from 2015 to 2017.Each field trial followed a randomized complete block design,with one-row plots and two replications.Each row was 3.5 m long and 0.5 m wide and contained 13 plants.At the flowering stage,nine plants per line from each replication were randomly selected for phenotyping.The means of seven traits,SD,SL,PH,EH,SBS,SCS,and RPR,were calculated.The flowering times of the inbred lines ranged from 50 to 68 days(Fig.S1),and the stalk strength traits were determined by testing the middle of the flat side of the third internode two weeks after flowering.SBS,SCS,and RPR were measured using different pressure sensors of the YYD-1 instrument(Zhejiang Top Cloud-Agri Technology Co.,Ltd.,Hangzhou,Zhejiang,China;Fig.S2).Values were recorded in N mm-2(RPR)and N(SCS and SBS).Phenotypeswerefittedbythefollowinglinearmodel:yijk=μ+Gi+Ej+GEij+-Bk(Ej)+eilk,where yijkrepresents each observation of phenotypic value,μ represents the population mean,Girepresents genotype effect,Ejrepresents environment effect,GEijrepresents genotypeby-environment interaction,Bk(Ej)represents block effect,and eilkrepresents random error.Analysis of variance(ANOVA)as implemented in R[28]was used to test for differences between environment,lines,and interactions(G×E).The BLUP value of each trait was calculated using the Mixed linear model(MLM)in the R package lme4.The lme4 package was used to estimate the genotypic(σ2g),environment(σ2evn),genotype-by-environment interaction(σ2ge),and error(σ2e)variances.Likelihood ratio tests were used in R package lmerTest to determine the significance of variance components.Broad-sense heritability was calculated using the following equation:h2=σ2g/（σ2g+σ2ge/e+σ2e/re）,where e and r represent respectively the number of environments and the number of blocks in each environment[29].BLUP values were used for calculating pairwise correlation between traits.The coefficient of variation(CV)was calculated for each trait as the ratio of the standard deviation to the mean.

2.2.Genotypic analysis and genome-wide association study

The panel of 345 inbred lines was genotyped by applying the genotyping-by-sequencing strategy[25–27].After a quality control step(missing rate≤20%;Minimum allele frequency[MAF]≥0.05),372,331 SNPs were retained for the GWAS.A principal component analysis(PCA)was also performed with TASSEL[30],and the top five principal components were used to create a population structure matrix to control for population structure.Population structure was characterized with ADMIXTURE[31]using k=3.The panel into three main groups,referred to as Lancaster,Reid,and Tangsipingtou according to their known pedigrees and germplasm(Fig.S3).Kinship matrices were calculated according to the centered IBS method of TASSEL to estimate the genetic relatedness among individuals.A GWAS was conducted using the general linear model(GLM)with PC and the MLM with PC and kinship in TASSEL.The GWAS analysis was performed with a modified Bonferroni correction(P＜1/372,331),but this was found to be too strict for less significant trait associations[32].A less stringent criterion(P-valu e＜1×10-4)used in previous studies[33,34]was used to detect trait-associated SNPs.Linkage disequilibrium(LD)decay was determined with PopLDdecay software[35],and the mean LD across all chromosomes decayed to r2=0.20 within approximately 50 kb[26].Significant SNPs were grouped into one QTL if the physical distance between neighboring SNPs was＜50 kb.SNPs with the minimum Pvalue in the quantitative trait locus(QTL)were designated as lead SNPs,and the QTL were named with Q for 1 to 94.

2.3.Candidate gene analysis and gene-based association mapping

All potential candidate genes within 50 kb of the detected loci were identified.Gene annotation information was obtained from MaizeGDB(https://www.maizegdb.org).The physical locations of the genes and SNPs were determined using the maize B73 RefGen_V3 genome(version 5b.60).Published gene expression data[14]were used to identify candidate genes.The MoSCSSS C0 population was derived from 14 inbred lines belong to the Iowa Stiff Stalk Synthetic heterotic group,and high(C15-H)and low(C15-L)populations were derived after 15 cycles of divergent selection for RPR on the C0 population.Inbred lines derived by the ear-to-row method for six cycles from C15-H and C15-L populations were designated as Hrpr1 and Lrpr1,respectively.Internodes were collected at 0,3,6,9,12,15,18,24,27 days after sowing(DAS)with three biological replicates for RNA extraction and sequencing.Differential expression was calculated using DESeq2[36].Genes with log2fold change＞2 and false discovery rate(FDR)-corrected P＜0.05 were selected for subsequent analyses.Transcriptome comparison of Hrpr1 and Lrpr1 revealed that 8327 genes were differentially expressed during one or more stages.

For gene-based association mapping,genomic DNA was extracted from fresh young leaves of the 345 inbred lines.Candidate genes from the tested inbred lines were sequenced using the targeted sequence capture technology of the NimbleGen platform[37]by BGI Life Tech Co.,ltd.,Shenzhen,Guangdong,China.A multiple-sequence alignment was constructed with MAFFT 7.313[38].Gene-based polymorphisms(MAF≥0.05)were identified with TASSEL.The significance of associations between SNPs and traits was assessed based on the MLM model(MLM+PCA+kin ship)of TASSEL.The P value threshold for controlling the genomewide type I error rate was 0.05/n(where n is the number of markers from the candidate gene)[39].

By screening the Maize EMS-induced Mutant Database(https://www.elabcaas.cn/memd/)for the B73 inbred line[40],we identified zmnr2(EMS3-062433)as an EMS mutant for ZmNR2 harboring STOP_GAINED mutation sites.The mutation sites were validated by sequencing PCR products using specific primers(Table S1).Three replications of B73(wild-type)and zmnr2 plants were grown under high nitrogen(HN)and low nitrogen(LN)conditions in Zhenjiang in 2021.Before sowing,the fields were supplied with 135 kg ha-1KCl and 750 kg ha-1CaH4O8P2.LN and HN conditions were provided by treating the fields with respectively 0 kg ha-1and 391 kg ha-1urea.At two weeks after flowering,RPR was evaluated as described above.

3.Results

3.1.Phenotypic variation

Seven traits:three stalk strength traits(RPR,SCS,and SBS)and four plant architecture traits(PH,EH,SD,and SL),were evaluated.Trait distribution analysis revealed a slightly left-skewed continuous distribution(Fig.1).There was high variation in all traits(Table 1;Fig.1).The coefficients of variation ranged from 8.02%(SD)to 22.81%(SBS)over the three environments.All traits showed＞1.5-fold differences,ranging from 1.59 to 3.07 among the lines(Table 1).The effects of genotype,environment,and genotype×environment interaction were significant for most of the traits(Table 1).The significances of the variance components were similar to the results of ANOVA(Table S2).

Table 1 Descriptive statistics and ANOVA for stalk strength traits.

The three stalk strength traits were highly positively correlated with one another(r=0.422–0.764).The plant architecture-related traits were also highly positively correlated(r=0.326–0.570),except for SD and SL(Fig.1).All four plant architecture traits were positively correlated with RPR(r=0.134–0.368).Among these four traits,PH,EH,and SD were significantly positively correlated with SCS and SBS(Fig.1).

Fig.1.Phenotypic correlations between all measured traits in 345 inbred lines.*,P＜0.05;**,P＜0.01;***,P＜0.001.RPR,rind penetrometer resistance;SCS,stalk crushing strength;SBS,stalk bending strength;PH,plant height;EH,ear height;SD,stem diameter;SL,stem length.

3.2.Genome-wide association study

Respectively 907 and 3 marker–trait associations were identified by GLM and MLM(P＜1/372,331)(Table S3).The Q–Q plot did not show a straight line and tail for all observed traits in GLM,but rather a divergence upward,indicating that there were false positives for GLM(Fig.S4).In contrast the line diverged downward for MLM,indicating that there are false negatives or no true positive.Under the relaxed significance threshold,241 marker–trait associations were detected for the seven traits.Multiple SNPs in specific genomic regions were tightly associated with the traits in the Manhattan plots(Fig.2).After clustering,94 QTL were detected for the seven traits,with 11–20 QTL per trait.On chromosome 1,Q14 contained 33 significant SNPs associated with SD,the most significantly associated SNP being S1_288149975,which explained 6.9%of the phenotypic variance.On chromosome 5,Q56 contained 14 significant SNPs associated with RPR,the most significantly associated SNP being S5_210155797,which explained 9.7%of the phenotypic variance.Five pleiotropic QTL involving four traits were identified.Q15 was associated with SCS and RPR,Q35,Q37,and Q49 were associated with SCS and SBS,and Q66 was associated with SCS and SD(Fig.2;Tables S3,S4).

3.3.Identification of candidate genes

A set of 256 candidate genes within 50 kb of all significant SNPs were identified on the basis of the GWAS results and the filtered predicted gene set from the annotated B73 maize reference genome(Table S4).Among genes differentially expressed between the high-RPR and low-RPR lines,56 were within the GWASidentified loci(Table S4).For example,GRMZM2G079768,which encodes a LOB domain-containing protein,was detected in Q35 and was associated with SCS and SBS.This gene was differentially expressed between the high-RPR and low-RPR lines at 6,12,18,and 27 days after sowing(DAS)(Table 2;Fig.3).Another gene,GRMZM2G028980,which encodes an auxin response factor,was localized to Q47 and was associated with RPR.This gene was differentially expressed at 12,15,18,and 24 DAS.

3.4.Verification of the candidate genes ZmFLS2 and ZmNR2

GRMZM2G069298(ZmFLS2)and GRMZM5G878558(ZmNR2)were selected for resequencing to investigate the association between allelic and phenotypic variation.GWAS identified S5_210155797(Tables 2,S4;Fig.4)as a significant SNP associated with RPR in Q56.The GRMZM2G069298 gene,encoding a flavonol synthase,was located 91 base pairs(bp)upstream of S5_210155797.The GRMZM2G069298 genomic region comprising 3641 bp,including the 1799-bp upstream region,the 1003-bp coding region,and the 503-bp downstream region(after the translation termination site).In total,234 sequence variations with MAF≥0.05 were identified,including 181 SNPs and 53 InDels.The MLM-based marker–trait association analysis detected 47 variants significantly associated with RPR(Fig.4D;Table S5),with the most significant SNP(SNP-397)located in the upstream region.RPR was higher for the lines carrying the G allele than for the lines with the A allele(Fig.4F).Five major haplotypes,present in more than five lines,emerged from these significant sites across the 345 inbred lines.A significant phenotypic difference was revealed via ANOVA between the haplotypes for RPR(Fig.S5).

Fig.2.GWAS associations for all measured traits by mixed linear model.The Manhattan plot combined all 7 traits,and only SNPs with P-value＜0.01[–log10(P)＞2]are shown.RPR,rind penetrometer resistance;SCS,stalk crushing strength;SBS,stalk bending strength;PH,plant height;EH,ear height;SD,stem diameter;SL,stem length.

Table 2 Candidate genes identified by combining GWAS and expression profile information for RPR,SCS,and SBS.

Fig.3.The expression profiles of candidate genes in nine development stages in high-and low-RPR lines.‘‘FC”indicates the fold change of gene expression level between high-and low-RPR lines.DAS,days after sowing.

Another SNP associated with RPR,S5_210786313,was located 1.6 kb downstream of GRMZM5G878558,which encodes a nitrate reductase.Analysis of the GRMZM5G878558 genomic region comprising 5906 bp,including the 1879-bp upstream region,the 2682-bp coding region,and the 525-bp downstream region,revealed 218 sequence variations(MAF≥0.05),of which 121 and 97 were respectively SNPs and InDels.The MLM-based marker–trait association analysis detected two variants significantly associated with RPR(Fig.4E;Table S5),with the most significant SNP(SNP623)located in the first exon.RPR was higher for lines carrying the A allele than for lines with the G allele(Fig.4G).To verify the function of the candidate gene ZmNR2,one B73 EMS mutant(Mut_Sample:EMS4-0a0498)with a termination mutation in the first exon of ZmNR2 was obtained from the maize EMS mutant library(https://www.elabcaas.cn/memd/)(Fig.5A,B).The zmnr2 and wild-type plants were compared with respect to their RPR under HN and LN conditions.The zmnr2 mutant showed significantly lower RPR and SD and higher EH than the wild-type control under both conditions(Figs.5C,S6–S8).

4.Discussion

A strong stalk will increase stalk lodging resistance.Breeding crops with high stalk strength is a promising strategy for increasing yields under stress conditions[19].Maize stalk strength is determined by several factors,including stalk architecture,SBS,SCS,and rind strength,which are complex quantitative traits controlled by numerous QTL[18,22].The most common way to improve quantitative traits in breeding involves phenotypic selection,but improving maize stalk strength through breeding is a time-consuming process.It took＞50 years of recurrent selection to reduce the stalk lodging rate of the Iowa Stiff Stalk Synthetic maize population from 19.6% to 13.6%[41].Identifying genes or functional markers associated with stalk strength useful for gene editing or marker-assisted selection is a promising way to improve stalk strength.In this study,a GWAS was conducted for three stalk strength traits and four plant architecture traits,resulting in the identification of two candidate genes and their variants potentially associated with stalk rind strength.

Accurate high-throughput phenotyping is crucial for a trait–marker association analysis[42].Several previous studies[18,19,22,43]showed that RPR,SCS,and SBS can easily and inexpensively be measured for large-scale assessment of maize stalk strength.Although stalk strength is a complex quantitative trait,the broad-sense heritability of RPR,SCS,and SBS was high in this study.The h2ranged from 0.72 to 0.92 in maize populations in previous studies[18,21,22].These findings suggest that stalk strengthrelated traits are highly heritable.There are several reports describing the relationships among stalk strength,lodging resistance,cell wall structural components,and morphological traits.In this study,the three stalk strength-related traits were correlated,and PH,EH,and SD were positively correlated with the stalk strength traits.This finding is consistent with Wang’s finding[16]that plant height and ear height were positively correlated with stalk breaking force at silking stage.The possible explanation is that plant height requires a stronger stalk for the same cultivars.In contrast,Helms and Compton(1984)reported no correlation between ear height,ear weight,and stalk lodging[44].Five pleiotropic QTL involving four traits(SCS vs RPR,SCS vs SBS,and SCS vs SD)were identified.In an earlier investigation[22],SD,SBS,and RPR were correlated and four quantitative trait nucleotides associated with both SBS and RPR were detected.The gene(stiff1)associated with a major quantitative trait locus for stalk strength in maize was confirmed to influence both RPR and SBS[19].These results are suggestive of a similarity in the genetic mechanisms underlying stalk strength-related traits and plant architecturerelated traits.

Fig.4.Associated loci and candidate genes for RPR on chromosome 5.(A)Manhattan plot for RPR on chromosome 5(150–218 Mb).(B,C)Genes surrounding the peak signals(±50 kb).(D,E)Gene-based association mapping for GRMZM2G069298(ZmFLS2)and GRMZM5G878558(ZmNR2).(F,G)Phenotypic difference of RPR between different alleles of GRMZM2G069298(ZmFLS2)and GRMZM5G878558(ZmNR2).The significance test was conducted with two-sample t-test.RPR,rind penetrometer resistance.

To elucidate the genetic basis of stalk strength,241 significant SNPs were detected for the seven examined traits via GWAS.The phenotypic variation explained(PVE)by each SNP ranged from 4.5% to 9.8%,and most of the SNPs explained＜6% of the phenotypic variation.Previous studies identified 69 RPR-associated QTL in 33 segregating populations,with a PVE range of 5.6%–20.2%[18].A major QTL associated with stalk strength reportedly[19]accounted for respectively 23.8% and 15.6% of the variation in SBS and RPR,in a RIL population derived from a cross between typical stiff-stalk and non-stiff-stalk lines.These results,combined with the findings of other studies[18,21]suggest that stalk strength is controlled by a few large-effect and many small-effect QTL.Among 69 RPR-associated QTL detected earlier,only about 10 were common to at least two populations or two studies,reflecting the complex nature of stalk lodging.In this study,24 SNPs from seven QTL for RPR were also detected in previous investigations.One QTL associated with RPR,Q56,was located on chromosome 5(210.15 Mb)and overlapped with a pleiotropic QTL(pQTL5-2)previously detected at four different stages(days to silking,10 days after silking,20 days after silking,40 days after silking)in a lodging-resistant population[45].Three SNPs in Q87 were located in the confidence interval of qBhf8,which was detected at 40 days after silking in a high-oil population[45].Q1,Q16,and Q47 were identified in a RIL population derived from B73×Ce03005 hybridization[21].Of these QTL,Q47 is located close to a SNP for RPR detected earlier by a multi-locus GWAS[22].

Fig.5.Phenotypic difference of ZmNR2 EMS mutant and the wild type(B73)under high(HN)and low(LN)nitrogen conditions.(A)Gene structure of ZmNR2 and the mutant site.(B)Sequence chromatograms of ZmNR2 in the wild type and the EMS mutant covering the mutation site.(C)Phenotypic difference of RPR between ZmNR2 EMS mutant and the wild type(B73)under HN and LN conditions.The significance test was conducted with a two-sample t-test.RPR,rind penetrometer resistance.

An integrated analysis of GWAS and gene expression data is useful for detecting potential causal genes for complex traits[46–48].Previous transcriptome studies generated information about differentially expressed genes that control stalk strength.A total of 8327 differentially expressed genes were identified between Lrpr1 and Hrpr1 during one or more stages[14].In this study,combining gene expression and GWAS data led to the detection of 56 candidate genes for all examined traits.Stalk strength is affected mainly by cell wall development.Previous research[1,19,20]indicated that stalk strength is closely associated with cell wall components,including soluble sugars,glucose,cellulose,hemicellulose,and lignin.One of the pleiotropic QTL(Q66)detected in the present study consisted of seven significant SNPs associated with SCS and SD,of which two(S6_166788742 and S6_166792172)were located in GRMZM2G027955,which encodes a glucose-1-phosphate adenylyltransferase.This gene was differentially expressed between the high-RPR and low-RPR lines at 9,12,15,18,and 27 DAS.GRMZM5G814718 in Q5 is a laccase gene(lac12)that was differentially expressed in the high-RPR and low-RPR lines at 24 DAS.This gene was associated with EH,but its homolog(lac21;GRMZM2G447271)was previously[24]identified as a gene differentially expressed between lodging-resistant and lodging-sensitive lines.In the present study,we detected some genes involved in nitrogen and amino acid metabolism.GRMZM5G878558(ZmNR2),which encodes a nitrate reductase,was located 1.6 kb upstream of a SNP(S5_210786313)significantly associated with RPR.A phenotypic analysis of the zmnr2 mutant revealed that its RPR was significantly lower than that of the wild-type control under both HN and LN conditions.Another gene,GRMZM2G053669(ZmASN3),which encodes an asparagine synthetase,was located 1.0 kb upstream of a SNP(S1_45123835)significantly associated with SL.The enhanced metabolism of amino sugars may increase plant mechanical strength in the absence of optimal lignification(11).In this study,we determined that variants around GRMZM2G069298(ZmFLS2),which encodes a flavonol synthase,were associated with RPR.Flavonol synthase is the key enzyme responsible for the biosynthesis of flavonols,which are the most abundant flavonoids with critical roles in reinforcing the cell wall and increasing plant mechanical strength[14,49].Plant hormones,such as auxin and cytokinins,also modulate stalk strength by inducing the expression of genes involved in lignin biosynthesis[50,51].Some of the candidate genes identified in the present study are involved in auxin and cytokinin biosynthesis and signal transduction.GRMZM2G028980 is an auxin response factor gene,GRMZM2G399644 is an auxin-responsive SAUR gene family member,and GRMZM2G151332 is a cytokinin-Oglucosyltransferase gene.The associated loci and candidate genes described here provide a foundation for future attempts to identify the molecular mechanism underlying stalk strength and for the genetic improvement of maize lodging resistance.

CRediT authorship contribution statement

Shuhui Xu:Investigation,Visualization.Xiao Tang:Investigation,Writing–original draft.Xiaomin Zhang:Investigation,Funding acquisition.Houmiao Wang:Visualization,Writing–review&editing,Funding acquisition.Weidong Ji:Investigation,Writing–original draft.Chenwu Xu:Conceptualization,Funding acquisition.Zefeng Yang:Conceptualization,Writing–review&editing.Pengcheng Li:Conceptualization,Writing–review & editing,Project administration,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

This work was supported by the National Natural Science Foundation of China(31972487,31902101,32172009 and 32061143030),the Innovative Research Team of Universities in Jiangsu Province,the Science and Technology Development Plan Project of Henan Province(212102110152),the High-end Talent Project of Yangzhou University,and the Priority Academic Program Development of Jiangsu Higher Education Institutions(PAPD).

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2022.04.016.