ldentification and validation of novel loci associated with wheat quality through a genome-wide association study

PU Zhi-en ,YE Xue-ling ,Ll Yang ,SHl Bing-xin ,GUO Zhu ,DAl Shou-fen,MA Jian,LlU Ze-hou,JlANG Yun-feng,Ll Wei,JlANG Qian-tao,CHEN Guo-yue,WEl Yu-ming,ZHENG You-liang

1 College of Agronomy,Sichuan Agricultural University,Chengdu 611130,P.R.China

2 Triticeae Research Institute,Sichuan Agricultural University,Wenjiang 611130,P.R.China

3 Key Laboratory of Coarse Cereal Processing,Ministry of Agriculture and Rural Affairs/College of Food and Biological Engineering,Chengdu University,Chengdu 610106,P.R.China

4 Crop Research Institute,Sichuan Academy of Agricultural Sciences,Chengdu 610066,P.R.China

Abstract Understanding the genetic basis of quality-related traits contributes to the improvement of grain protein concentration(GPC),grain starch concentration (GSC),and wet gluten concentration (WGC) in wheat. In this study,a genome-wide association study (GWAS) based on a mixed linear model (MLM) was performed on 236 wheat accessions,including 160 cultivars and 76 landraces,using a 55K single nucleotide polymorphism (SNP) array in multiple environments. A total of 12 stable QTL/SNPs that control different quality traits in this populations in at least two environments under stripe rust stress were identified. Among these 12,three,seven and two QTLs associated with GPC,GSC and WGC were characterized,respectively,and they were located on chromosomes (chr) 1B,1D,2A,2B,2D,3B,3D,5D,and 7D with the phenotypic variation explained (PVE) ranging from 4.2 to 10.7%. Compared with the previously reported QTLs/genes,five QTLs (QGsc.sicau-1BL,QGsc.sicau-1DS,QGsc.sicau-2DL.1,QGsc.sicau-2DL.2,and QWgc.sicau-5DL) were potentially novel. KASP markers for the SNPs AX-108770574 and AX-108791420 on chr5D associated with wet gluten concentration were successfully developed. The phenotypes of the cultivars containing the A-allele in AX-108770574 and the T-allele in AX-108791420 were extremely significantly (P＜0.01) higher than those of the landraces containing the G-or C-allele with respect to the wet gluten concentration in each of the environments. The KASP markers developed and validated in this study could be utilized in molecular breeding aimed at improving the quality of wheat.

Keywords: cultivars,landraces,grain protein concentration (GPC),grain starch concentration (GSC),wet gluten concentration (WGC),55K SNP,validation

1.lntroduction

Wheat (Triticum aestivumL.) is a culturally and historically important staple food consumed by one-third of the world’s population for the daily dietary intake of protein. It is the most important source of protein,starch,and energy globally. The grain protein concentration (GPC),grain starch concentration (GSC),and wet gluten concentration (WGC)are the important measures of nutritional quality,and they play a vital role in many aspects. For example,GPC is one of the major pricing factors for wheat trading,while GPC and GSC are the major traits that determine yield,nutritional,and processing quality. GPC has a strong relationship with WGC,and they determine the end-use of grains. With the increase in the atmospheric carbon dioxide concentration,a decline in the GPC has been observed (Fernandoet al.2012). This poses a threat to human nutrition,and given that an increase in the development of people’s living standard has increased the demand for wheat quality,there is a need to increase nutritional qualityviabreeding.Therefore,the focus of this study is primarily centered on the important nutritional traits of GPC,GSC and WGC.

In wheat,GPC,GSC,and WGC are complex quality traits that are strongly influenced by genetic factors,environmental factors,and many other factors. Some reports have shown that the GPC and WGC exhibit lower heritability (Liet al.2009;Jerniganet al.2018;Kumaret al.2018). Several studies have reported that various environmental factors exert significant effects on GPC,GSC,and WGC (Labuschagneet al.2007;?ugowskaet al.2012;Balyanet al.2013;Konget al.2013;Kumaret al.2018).Moreover,stripe rust is one of the most destructive wheat diseases in the world. It is caused by the fungusPuccinia striiformisWestend.f.sp.triticiErikss.(Pst) and is a serious threat to wheat quality (Sunderman and Wise 1964;O’Brienet al.1990;Devadaset al.2014). Particularly in Sichuan Province,China,wheat has been plagued by stripe rust for years,which has resulted in the low-quality of grains.

The quality traits related to GPC,GSC and WGC are important. GPC affects the baking properties and varies among the different wheat cultivars from 10 to 20% (Gillieset al.2012). Generally,wheat can be separated into bread wheat,noodle wheat and cookie wheat according to the end-use based on GPC to some extent. The results of previous studies have shown (1) a significant influence of the environment on GPC,(2) a negative correlation between GPC and grain yield,and (3) the presence of many loci with small genetic effects on GPC and low heritability of the trait (Simmonds 1995;Grooset al.2003;Balyanet al.2013). Therefore,independent and stable quantitative trait loci (QTL) or genes with significant effects on GPC are better for breeding. Starch,which is composed of amylopectin and amylose (Preiss 1991),is the major component in the wheat endosperm that accounts for almost three-quarters of grain composition.The GSC influences grain processing and the end-use quality of traditional flour products,such as noodles and steamed bread (Panozzo and McCormick 1993;Chiotelliet al.2002). It is positively correlated with grain yield in wheat (Surmaet al.2012;Rakszegiet al.2016;Krystkowiaket al.2017). WGC is a special form of protein that plays a pivotal role in determining the bread-baking quality and pasta-making technological properties. It is positively correlated with GPC and negatively correlated with GSC (Surmaet al.2012;Denget al.2013;Tian Jet al.2015a). Grains with high WGC levels are suitable for bread-type products,while those with low WGC levels are suitable for cake-type products (Chenet al.2019).

Recent molecular technologies have helped researchers to identify many QTLs associated with quality traits. More than 500 QTLs spread over 21 chromosomes have been associated with GPC through linkage mapping based on biparental populations (Krystkowiaket al.2017;Marcotuliet al.2017;Nedelkouet al.2017;Zouet al.2017;Kumaret al.2018;Liuet al.2018,2019;Mir Drikvandet al.2018;Rappet al.2018;Rosellóet al.2018;Chenet al.2019;Goelet al.2019;Nigroet al.2019;Thorwarthet al.2019;Fatiukhaet al.2020;Suet al.2020).Gpc-B1is the most critical gene that has resulted in a maximum increase in GPC,and it has been widely used in many cultivars (Tabbitaet al.2017). Approximately 250 QTLs spread over 21 chromosomes were identified for GSC (McCartneyet al.2006;Reifet al.2011;Hu 2013;Panget al.2014;Denget al.2015a,2018;Tian Bet al.2015;Tian Jet al.2015b;Zhang 2016,2019;Krystkowiaket al.2017;Guan 2018).They include various starch synthesis-related enzymes,such as starch synthases,soluble starch synthase,starch branching enzyme,starch-debranching enzyme,and granule-bound starch synthase. In addition,130 QTLs distributed over 21 chromosomes were identified for WGC(Zhanget al.2008;Liet al.2009,2012,2013;Denget al.2015b;Tian Jet al.2015a,b;Cuiet al.2016;Krystkowiaket al.2017;Liuet al.2017;Chenet al.2019;Johnsonet al.2019;Suet al.2020).

GPC,GSC and WGC are complex traits regulated by several loci with small genetic effects,so genomewide association studies (GWAS) can identify the associations between phenotypic variations and nucleotide polymorphisms using a diverse population panel to elucidate the genetic effects on specific traits(Bazakoset al.2017). Large numbers of molecular markers that facilitate the progress of more efficient mapping techniques have been developed (Julianaet al.2019;Sheret al.2019),and they have been used to detect numerous natural allelic variations and historical chromosomal recombination events that occur over multiple generations of natural populations. Recently,complex agronomic traits and their potential causal genes were detected by GWAS (Fiedleret al.2017;Liuet al.2022). In this study,GWAS was used to detect the QTLs associated with variations in GPC,GSC,and WGC,with the aim of identifying more stable QTLs associated with quality traits and broadening the genetic basis of wheat varieties. In this study,we analyzed the three nutritional quality traits (GPC,GSC and WGC) of 236 Sichuan wheat accessions with the following objectives: (1) evaluate GPC,GSC and WGC of Sichuan wheat accessions in multiple environments;(2) select the elite germplasms for wheat quality breeding;and (3) identify the novel QTLs for these three quality traits using a genome-wide association study.

2.Materials and methods

2.1.Plant materials

A total of 236 Sichuan wheat accessions,including 76 landraces and 160 cultivars,were used in this study(Appendix A). All these accessions were homozygous lines provided by the Triticeae Research Institute,Sichuan Agricultural University (germplasms abbreviated as AS)and the Chinese Crop Germplasm Resources Bank of China (germplasms abbreviated as ZM). The cultivars had different genetic backgrounds and were released from 1997 to 2016 by different breeding organizations,such as Sichuan Academy of Agricultural Sciences,Mianyang Academy of Agricultural Sciences,Neijiang Academy of Agricultural Sciences,Chengdu Institute of Biology of Chinese Academy of Sciences,Sichuan Agricultural University,and Southwest University of Science Technology of China.

2.2.Field trials

The accessions were planted in four different environments across two locations in Sichuan Province,China,under stripe rust stress: Chongzhou (30°33′N,103°39′E) and Mianyang (31°23′N,104°49′E). Field trials were conducted according to a randomized block design with three replications over three growing seasons in Chongzhou (2017,CZ17;2018,CZ18;2019,CZ19)and during one season in Mianyang (2017,MY17). In all field trials,20 seeds of each accession were planted 10 cm apart in a 2-m row,with 30 cm between rows,and three replicates were maintained per accession.

2.3.Measurement and analysis of quality traits

GPC and GSC in the whole-grain were determined by Near Infrared Reflectance Spectroscopy (NIR,Foss,Sweden) (AACC 2000). Gluten extraction was carried out by adopting the procedure described in AACC (2000). All these traits were expressed on a grain dry weight basis.The three quality traits were adjusted to 14% moisture concentration for further analysis (Hayneset al.2009).Stripe rust infection type (IT) was evaluated at the adult stage following the method of Yeet al.(2019b).

Best linear unbiased prediction (BLUP) values for each accession across the different locations were calculated by fitting the linear mixed model in R package ‘lme4’(Bateset al.2014) to eliminate the environmental impacts on the quality traits. The genotypic and environmental variances (VGandVE) were also computed using the ‘lme4’package (Bateset al.2014). Broad-sense heritability (H2)for each of the quality traits was calculated across all test environments using the formulaH2=VG/(VG+VE) (Smithet al.1998). The phenotypic diversity was confirmed using the Shannon-Weaver diversity index (H′) (Liet al.2015)calculated based on the BLUP values for each trait. SPSS 20.0 Software (IBM Corp.,Armonk,NY,USA) was used to calculate the Pearson’s correlation coefficient among the four environments or three traits,to perform thet-test for determining the significant differences in GPC,GSC and WSC between landraces and cultivars,and to test the normal distribution of the quality traits based on a quantilequantile (Q-Q) plot. Differences and interactions of the effects on the variances of wheat accessions,years,and plant locations were tested by two-way ANOVA.

2.4.Genotyping and molecular analysis

Genomic DNA was extracted using the plant genomic DNA Kit (Biofit Co.,China) from a mixed sample of leaves that were collected from five one-week-old seedlings of each accession. All 236 DNA samples were genotyped using a 55K single nucleotide polymorphism (SNP)array (Affymetrix Axiom Wheat55K) at China Golden Marker Biotechnology Co.Ltd.(Beijing,China). The SNP markers with missing values of ≤10% and minor allele frequency (MAF) of ≥5% were selected for further analysis. The polymorphic information concentration (PIC)of 144 326 SNP markers was analyzed to evaluate the genetic diversity using Software POWERMARKER v3.25(Liu and Muse 2005).

2.5.Population structure,neighbor-joining (NJ)phylogeny,and linkage disequilibrium (LD) decay

The population structure (Q-matrix) of the wheat accessions was analyzed based on a Bayesian model using STRUCTURE Software (v2.3.4) (Pritchardet al.2000). Five independent runs for K (from 1 to 10) were performed using the admixture model with 10 000 burnin and 10 000 Markov Chain Monte Carlo (MCMC)iterations. The results from STRUCUTRE were summarized to obtain the optimum population structure(optimum K) using the delta K (ΔK) method in the webbased Software STRUCTURE HARVESTER (Earl 2012). To understand the genetic relationships among all accessions,a phylogenetic tree was constructed by applying the neighbor-joining (NJ) method based on shared allele distance (Chakraborty and Jin 1993) using TASSEL Software v5.2.38 (Bradburyet al.2007). The NJ phylogenetic tree was collapsed and formatted using iTOL v4 (Letunic and Bork 2019). The pairwise measure of linkage disequilibrium (LD) was estimated as the squared allele frequency correlation (r2) between pairs of intrachromosomal markers with known chromosomal positions using TASSEL v5.2.38 (Bradburyet al.2007). Significant pairwise markers were chosen with the thresholds ofP＜0.001 andr2＞0.1,and the LD decay plot and half decay distance were generated withr2using the ggplot2 package in the R Program (Wickham 2016).

2.6.GWAS

To identify the loci associated with GPC,GSC and WGC,GWAS was performed on the 236 wheat accessions with the 44 326 effective markers using TASSEL v5.2.38 (Bradburyet al.2007) based on a mixed linear model (MLM) with Q and K as covariates. To identify significantly associated loci,we set a thresholdP-value of(-log10P≥2.5) and the criterion of being detected in at least two test environments. Manhattan plots withP-values were generated using the ggplot2 package in the R Program (Wickham 2016) to visualize the loci associated with the quality traits. All the associated loci with-log10P≥2.5 in the half decay distance region on the same chromosome were assigned to the same QTL block.

2.7.Candidate genes analysis

We further analyzed the putative candidate genes to identify the novel QTLs. The genes included in the potentially novel QTLs were selected based on their LD decay distances using the Chinese Spring reference genome (IWGSC RefSeq v1.0,RefSeq Annotation v1.1)(Appelset al.2018). The BLAST tool was used to detect the homologous genes at the EnsemblPlants website(https://plants.ensembl.org/Multi/Tools/Blast?db=core)with default parameters. The candidate genes were identified based on their functional annotations.

2.8.Validation of the KASP markers for novel quality-related loci

To verify the validity of the QTLs identified in this study,we developed some KASP molecular markers for significant loci to screen the population and natural population in order to validate the loci. The sequences of the primer pairs for PCR amplification are listed in Appendix A. At-test was applied to detect the significant differences in the traits between the allele types based on the SNPs.

3.Results

3.1.Evaluation of GPC,GSC and WGC

Samples were collected from fields under four different environments (CZ17,CZ18,CZ19,and MY17) and the quality traits of GPC,GSC and WGC were measured.Pearson’s correlation analysis showed significant positive correlations in GPC,GSC and WGC among all test environments (Appendix B). The BLUP values were calculated to eliminate the impacts of environmental factors and to facilitate the further analysis as follows.The statistical analysis based on the Q-Q plot revealed that the number of accessions for quality traits all followed a normal distribution (Fig.1;Appendix C). GPC ranged from 12.02 to 14.13% (mean,12.90%),GSC ranged from 51.24 to 59.89% (mean,55.87%),and WGC ranged from 22.08 to 26.67% (mean,24.10%) (Table 1).The ANOVA performed by MLM to determine the effects of environment and genotype revealed highly significant variation (P＜0.001) among the different environments for each quality trait,and the variation among the genotypes was significant at the level ofP＜0.001. The genotype×environment interactions were significant for both GPC and WGC. TheH2values of GPC,GSC and WGC were 0.644,0.841 and 0.656,respectively(Table 1).

In addition to the environment,stripe rust is another significant factor influencing quality traits as it is a major epidemic disease for wheat in Sichuan Province. The correlation analysis between infection type and the quality traits revealed that stripe rust had significant negative correlations with both GPC (P＜0.01) and WGC (P＜0.05),and a significant positive correlation with GSC (P＜0.01).Among the three quality traits,GPC was negatively correlated with GSC (P＜0.01) but positively correlated with WGC (P＜0.01). Meanwhile,WGC and GSC showed a significant negative correlation (P＜0.01) (Table 2).

Table 1 The statistical analysis of quality traits,and the estimates of variance components and heritability1)

Table 2 The Pearson correlation coefficients among quality traits and infection type (IT)1)

3.2.Differences in GPC,GSC,and WGC between landraces and cultivars

Thet-test was performed to compare the significant differences (P＜0.05) of quality traits between all landraces and cultivars. Landraces had significantly higher GPC(P＜0.05) and WGC (P＜0.01) than the cultivars,while the cultivars had significantly higher GSC (P＜0.01) than the landraces (Table 3). Similar results were revealed in the phenotypic diversity analysis,as the landraces had higherH′in GPC and WGC,while the cultivars had higherH′in GSC (Table 3).

3.3.Analysis of SNP markers

A total of 44 326 effective SNP markers were selected for further analysis from the accessions genotyped using a 55K SNP array (with missing values ≤10%and MAF ≥5%). Of these,16 330,16 831 and 11 165 SNP markers were mapped on sub-genomes A,B and D,respectively,and covered map lengths of 4 930,5 176,and 3 946 million base pairs (Mb),respectively.The average number of markers per chromosome was 2 111 markers,and the average distance between the markers was 0.32 Mb (average marker density was 3.0 markers per Mb). Chromosome (chr) 6A had the highest marker density (4.0 markers per Mb),while chr4D had the lowest marker density (1.6 markers per Mb). The analysis of PIC showed that chr5B had the highest PIC value (0.330),and chr4A had the lowest(0.247) (Table 4). We also compared the PIC values between landraces and cultivars. The PIC values for cultivars were significantly higher than for landraces among the three sub-genomes and 21 chromosomes(Table 4).

Table 3 The analysis of phenotypic variations for landraces and cultivars based on the best linear unbiased prediction (BLUP) values1)

Table 4 The analysis of SNP markers and genetic diversity

3.4.Population structure and linkage disequilibrium(LD) decay

The population structure (Q-matrix) analysis using 44 326 SNP markers based on the Bayesian Model showed the optimal ΔK to be 2. The 236 accessions were classified into two subpopulations based on the three traits:subpopulation 1 (SP1) with 74 landraces and one cultivar(Xifu 14) and subpopulation 2 (SP2) with 159 cultivars and two landraces (Kaixianluohanmai and Yupi). Furthermore,we constructed the neighbor-joining phylogenetic tree based on the shared allele distances. An obvious genetic difference was observed when the tree interval was 0.25.The accessions were divided into two clusters,cluster 1 and cluster 2. A total of 74 landraces and 2 cultivars(Xifu 14 and Yumai 1) belonged to cluster 1,while 158 cultivars and 2 landraces (Kaixianluohanmai and Yupi)belonged to cluster 2 (Appendix D).

LD values across the three different sub-genomes and the whole genome were estimated using ther2between the significant pairs of intra-chromosomal SNP markers with physical distance,respectively. There were 553 495,546 714,and 326 580 significant (P＜0.001 andr2＞0.1)pairwise SNPs on sub-genomes A,B and D,respectively.The LD decay distances for sub-genomes A,B,and D and the whole genome were around 2.31,4.83,1.40,and 1.92 Mb,respectively,based on the best fitting curve (Fig.2).

3.5.Genome-wide association study (GWAS)

GWAS was used to analyze the 44 326 SNP markers and the associations with the quality traits in all test environments based on the MLM using Q and K as covariates. The loci with high confidence levels of association in the four environments were displayed as Manhattan plots withP-values across the 21 wheat chromosomes (Fig.3). A total of 15 SNP markers were significantly associated with GPC,GSC and WGC. Based on the LD decay distances for the corresponding subgenomes,12 QTLs located on chrs1B,1D,2A,2B,2D,3B,3D,5D,and 7D were named as follows:QGpc.sicau-1BS,QGpc.sicau-2AS,QGpc.sicau-2BS,QGsc.sicau-1BL,QGsc.sicau-1DS,QGsc.sicau-2DL.1,QGsc.sicau-2DL.2,QGsc.sicau-3BS,QGsc.sicau-3DS,QGsc.sicau-5DS,QWgc.sicau-5DL,andQWgc.sicau-7DL(Table 5).The phenotypic variation explained (PVE) by the markers ranged from 4.2 to 10.75%. Among the QTLs,three were significantly associated with GPC,seven with GSC,and two with WGC. In total,five QTLs were defined as potentially novel through the physical distances of the reported QTLs/genes based on the reference RefSeq v1.0 (Table 5). After considering the site contributions,QWgc.sicau-5DLwas identified as a significant QTL.

Table 5 The characteristics of the QTLs associated with quality traits

3.6.Candidate genes for potentially novel QTLs

A total of 24 putative candidate genes were predicted for the potentially novel QTLs,which may be associated with the quality traits. Among them,18 candidate genes were identified in the three QTLs (QGsc.sicau-1BL,QGsc.sicau-1DS,andQGsc.sicau-2DL.2) associated with GSC. Six candidate genes,all inQWgc.sicau-5DL,were predicted to be associated with WGC (Appendix D). They are genes involved in catalytic enzyme,carbohydrate metabolism or transportation,photosynthesis,programmed cell death,and abscisic acid-ethylene balance,which all affect quality traits directly or indirectly.

3.7.The effects of favorable alleles of QWgc.sicau-5DL on the wet gluten concentration

As mentioned above,QWgc.sicau-5DLwas identified in this study as a novel and stable QTL which includes four candidate genes,with PVE values varying from 5.0 to 7.9%. In order to verify the effect of the SNP loci on the WGC variation,the relationships between the SNP loci and phenotype were analyzed. There are no significant differences in WGC among the lines carrying from one to four favorable alleles or particular alleles. Lines with the favorable allele atQWgc.sicau-5DLshowed increased WGC by 9.1% in CZ17,by 12.0% in MY17,by 4.7% in CZ18,by 11.5% in CZ19,and by 5.2% with the use of the BLUP values (Fig.4).

Consequently,we then successfully developed two KASP markers from the QTLs on chr5DL for validating the association between the genetic variation and phenotype in this natural population. A total of 76 cultivars or landraces from Sichuan wheat were used to amplify the DNA fragments harboring the significant WGC-related SNPsAX-108770574andAX-108791420,respectively.GWAS revealed that the variation type of theAX-108770574locus was “A/G” in the association pool,and “A”and “G” showed respective positive and negative effects on WGC. T/C variation existed inAX-108791420,which showed that “T” exhibits a positive effect on WGC while the “C” locus is negative. At-test was then conducted on the phenotype of WGC between the two groups in each of the environments of the two KASP markers. The results showed that the phenotypes of the cultivars containing the A-allele inAX-108770574and the T-allele inAX-108791420were extremely significantly (P＜0.01) higher than those of the landraces containing the G-or the C-allele in each of the environments (Fig.5).

4.Discussion

4.1.lmpacts of environment on GPC,GSC and WGC

We analyzed GPC,GSC,and WGC of 236 wheat accessions,including 76 landraces and 160 cultivars,in four different field trials in Sichuan Province under stripe rust stress. We found that the quality traits were heavily influenced by the environments. The relatively smaller correlation coefficients and lower broadsense variability (H2) values among the different trials revealed the significant influence of the environment on the quality traits (Table 1;Appendix B). Furthermore,ANOVA intuitively proved the considerable differences in GPC,GSC and WGC among the environments. These findings are consistent with many previous reports on wheat in different environments (Turneret al.2004;Labuschagneet al.2007;Krystkowiaket al.2017;Kumaret al.2018). Meanwhile,theH2of GSC was higher than those of GPC and WGC.H2is an important genetic parameter that is used for phenotypic prediction and indicates all of the genetic contributions to phenotypic variance. The highH2value combined with relatively higher correlation coefficients among the four environments indicated a major impact of genotypes on GSC,and the possibility to improve traitsviaselective breeding. Den?i?et al.(2012) reported lowH2values of GPC and WGC in wheat. All these findings together indicate that it would be difficult to improve GPC and WGC by traditional breeding,since such traits should be selected at a later generation. Thus,traditional breeding combined with maker-assisted selection is a good choice for improving the breeding efficiency and shortening the breeding period.

4.2.lmpact of stripe rust on GPC,GSC,and WGC

The correlation coefficient analysis in this study showed that infection type was negatively correlated with GPC and WGC,but positively correlated with GSC (Table 2).The impacts of environmental factors (including stripe rust) on GPC and WGC were greater than on GSC,and the relatively higherH2(0.841) indicated that GSC was stable and not easily influenced by the environment(Table 1). We considered that the decreases in GPC and WGC were greater than in GSC. Meanwhile,the present study and several previous reports (Surmaet al.2012;Denget al.2013;Tian Jetal.2015a) demonstrated a positive correlation between GPC and WGC,but negative correlations between GSC and GPC and between GSC and WGC. Therefore,the analysis of these three quality traits and IT under certain conditions revealed a relatively positive correlation between IT and GSC. If we compare the quality traits obtained under the control and treatment settings,there must be a negative relationship between IT and the three quality traits. There is no doubt about the obvious negative impacts of stripe rust on grain quality,and many reports can support this speculation(Sunderman and Wise 1964;O’Brienet al.1990;Devadaset al.2014). The environment is complicated,involving many factors like weather,abiotic or biotic stresses that affect the growth of the plants and further influence the expression of QTLs/genes. Based on the complicated environment,more and more conditional QTLs/genes have been identified (Kodamaet al.2018;Sukumaranet al.2018;Fanet al.2019;Yeet al.2019a). It is also meaningful to detect the stable loci that confer quality traits even under stripe rust conditions like in Sichuan Province.

4.3.Differences in phenotypes and genotypes between landraces and cultivars

In the present study,obvious differences were found between landraces and cultivars. Bayesian classification based on a Q-matrix and neighbor-joining (NJ)phylogeny based on shared allele distances clearly grouped the accessions into landraces and cultivars.The genotypic diversity (PIC) of cultivars was higher than that of landraces,and this result is also supported by Maet al.(2020). The phenotypic diversity analysis showed better performance of landraces with respect to GPC and WGC,and better performance of cultivars with respect to WGC (Table 3). The landraces and cultivars have their own unique advantages in genotypes or phenotypes,so combining their advantages can produce elite lines in breeding. Moreover,regardless of whether the differences in genotype or phenotype were obviously identified between landraces and cultivars,the results indicate that the use of landraces in breeding can broaden the genetic background and improve the phenotypic diversity.

4.4.QTL associated with the quality traits

In this study,236 accessions were genotyped with a 55K SNP array,and a total of 44 326 effective SNP markers were analyzed. Based on the thresholdP-value (-log10P≥2.5) and the criterion of being detected in at least two test environments,GWAS for the quality traits in the accessions identified 12 QTLs associated with GPC,GSC,and WGC (Table 5).

Three of the QTL,QGpc.sicau-1BS,QGpc.sicau-2AS,andQGpc.sicau-2BS,were associated with GPC.QGpc.sicau-1BSwas located around 51.99 Mb on the short arm of chr1B with PVE ranging from 5.7 to 6.2%. It was covered byQGPC.ndsu.1B,which was linked to markerswPt-1684andwPt-5899(Echeverry-Solarteet al.2015).QGpc.sicau-2ASwas located in the distal region of the short arm of chr2A,which was similar to MQTL2A2 linked towPt-4197andwPt-5245in wheat (Bogardet al.2013).QTKW.sicau-2AS.1(Yeet al.2019a) was associated with thousand-kernel weight and located on the same block asQGpc.sicau-2AS.QGpc.sicau-2BSwas mapped on the short arm of chr2B at around 25.42 Mb. It was located in the same region asQGpc.crc-2B,which is flanked by the SSR markersXgwm210andXwmc25(McCartneyet al.2006).QSlC.sicau-2BS(Yeet al.2019a) was associated with spikelet compactness and located in the same region asQGpc.sicau-2BS. Thousand-kernel weight and spikelet compactness are important factors that determine grain yield in wheat. Moreover,several reports have demonstrated a close association between GPC and grain yield (Kumaret al.2018). Thus,in the present study,theQGpc.sicau-2BSidentified was found to be associated with both GPC and thousand-kernel weight/spikelet compactness.

We also identified seven QTLs associated with GSC.QGsc.sicau-3BSwas associated with markerAX-110012661,located on the short arm of chr3B,and was covered byQTsc-3B.12linked to SSR markersXwmc612andXbarc068(Tian Bet al.2015).QGsc.sicau-3DSwas mapped on the short arm of chr3D at around 17.18 Mb.It was very close to the reportedQftsc3D,which was flanked bywPt-2313andwPt-6965(Denget al.2018).QGsc.sicau-5DSwas close to the reported QTL linked to the SSR markerXbarc130(Krystkowiaket al.2017) at around 3.61 Mb,and explained a phenotypic variation of 10.75%. The other four QTLs (QGsc.sicau-1BL,QGsc.sicau-1DS,QGsc.sicau-2DL.1,andQGsc.sicau-2DL.2)were located at greater distances from the previously identified GSC genes or QTL regions,so they may be novel loci.

Two QTLs associated with WGC were identified.QWgc.sicau-5DLwas flanked by markersAX-111139947andAX-108791420and located in the interval between 555.91 and 556.72 Mb on the long arm of chr5D.This QTL is different from the previously reported genes/QTLs associated with WGC. Therefore,we considerQWgc.sicau-5DLto be a potentially novel QTL. Moreover,lines carrying the favorable allele atQWgc.sicau-5DLshowed significantly increased WGC (Fig.4). Another QTL,QWgc.sicau-7DL,was located on the long arm of chr7D around 619.11 Mb,and its PVE was from 4.19 to 5.05%. It was close toQGlu7D,which was flanked by the SSR markersXwmc634andXwmc273.2(Tian Bet al.2015).

In this study,the positioning of the 12 QTLs revealed that four of them were covered by previously reported QTL,three were close to reported QTLs,and five were identified as potentially novel QTLs. The position comparison for QTLs was based on the high-quality physical map of the reference genome (RefSeq v 1.0),comparisons with multiple linked markers,and keeping the strict thresholds,which lays the foundation for the allelism tests and fine mapping of the potentially novel QTLs.

4.5.Candidate genes for potentially novel QTLs

The GWAS and post-GWAS analyses were used to confirm the previously reported candidate genes and identify the new candidates that appear to be functionally linked to the analyzed quality traits in this study. Seven candidate genes are believed to exist inQGsc.sicau-1BL.TraesCS1B02G335900is homologous toArabidopsisgeneUGT91C1(UDP-glycosyltransferase 91C1),which is involved in the UDP-glycosyltransferase activity and affects the direct precursors of starch synthesis(Liet al.2018).TraesCS1B02G338600,aligned withArabidopsisgeneJAL3(Jacalin related lectin 3),is a carbohydrate binding protein that may play a role in determining GSC (Chiaet al.2020). Meanwhile,bothTraesCS1B02G340000andTraesCS1B02G340200are orthologous to wheat protein Agglutinin isolectin 3 and have the function of carbohydrate binding.TraesCS1B02G339200is orthologous toArabidopsisgeneTRAF1A(TNF receptor-associated factor homolog 1a),which is associated with autophagosomes and takes part in the regulation of autophagy dynamics (Qiet al.2017). Young and Gallie (1999) have reported programmed cell death (PCD) in the starchy endosperm cells of wheat and maize during the final stages of seed development. Based on this report,we presume thatTraesCS1B02G339200may influence GSCviaautophagy(type II PCD) in the starchy endosperm cells of wheat.TraesCS1B02G339500andTraesCS1B02G339700were found to be homologous toPSBW(Photosystem II reaction center W protein),which is involved in photosynthesis and photosystem II stabilization (García-Cerdánet al.2011) and further influences GSC in wheat through the photosynthetic products.

Another eight candidate genes were predicted forQGsc.sicau-1DSand associated with starch concentration.TraesCS1D02G022300has sequences homologous to rice glycosyltransferaseBC10. Glycosyltransferase catalyzes the transfer of sugars during starch biosynthesis (Sadoet al.2009;Zhouet al.2009).TraesCS1D02G023300is aligned with theArabidopsisgenePP2A1(Phloem protein 2-like A1) (Dinantet al.2003),which plays a role in carbohydrate-binding similar toJAL3.TraesCS1D02G024100is homologous to the geneEXGA(Probable glucan 1,3-beta-glucosidase A) ofEmericella nidulans. GO annotation indicated its role in polysaccharide catabolism.TraesCS1D02G024200,orthologous to the beta-galactosidase 1Os01g0533400in rice,might be involved in carbohydrate metabolism(GO annotation),and in turn,influence GSC.TraesCS1D02G026200has sequences homologous to the rice geneWNK2,a probable cytoplasmic serine/threonine kinase involved in protein phosphorylation (Gaudetet al.2011). Grimaudet al.(2008) reported the significance of protein phosphorylation in starch synthesis. Meanwhile,TraesCS1D02G026500andTraesCS1D02G032500are aligned withArabidopsiscysteine proteaseXCP1,which is related to PCD (Avciet al.2008). Similar toTraesCS1B02G339200,TraesCS1D02G026500andTraesCS1D02G032500may also influence the GSC of wheatviaPCD.TraesCS1D02G031700is orthologous to theArabidopsisgenePGRL1A(PGR5-like protein 1A),which is involved in photosynthesis and photosynthetic electron transport in photosystem I (DalCorsoet al.2008;Hertleet al.2013).

We identified three candidate genes inQGsc.sicau-2DL.2. The orthologous gene ofTraesCS2D02G277300isArabidopsisgeneEMB1674,which is involved in the abscisic acid (ABA)-activated signaling pathway. The balance between ABA and ethylene is crucial in PCD regulation in the starchy endosperm (Young and Gallie 2000).TraesCS2D02G278100showed homology withCHIT5B(Class V chitinase) ofMedicago truncatula. GO annotation analysis revealed its role in polysaccharide catabolism,similar toEXGA.TraesCS2D02G278200is aligned with theArabidopsisgeneTIF3H1(Translation initiation factor 3 subunit H1),which responds to ABA,glucose,and sucrose levels (Kimet al.2004). ABA can regulate PCD and also assist in the conversion of glucose and sucrose into starch.

Six candidate genes identified inQWgc.sicau-5DLmay affect WGC. WGC is a special form of GPC,and has a strong relationship with GPC.TraesCS5D02G546400is orthologous to theArabidopsisgeneERECTA(LRR receptor-like serine/threonine kinase) that regulates plant organ morphogenesis (Toriiet al.1996). The grain morphogenesis in wheat is related to protein concentration and further affects the WGC in the grain.TraesCS5D02G550500is aligned with theArabidopsisgeneASIL2,which is the trihelix transcription factor that represses the seed maturation program during early embryogenesis (Willmannet al.2011). The seed maturation program involves protein accumulation,so repression of the seed maturation program affects WGC indirectly.TraesCS5D02G551900andTraesCS5D02G553000are orthologous to rice geneCIN4(beta-fructofuranosidase,insoluble isoenzyme 4).TraesCS5D02G552000andTraesCS5D02G552900showed homology with1-FEHw3(fructan 1-exohydrolase w3) andCIN3(beta-fructofuranosidase,insoluble isoenzyme 3),respectively. These genes are similar toOs01g0533400and take part in carbohydrate metabolism(GO annotation),which might influence the GPC and WGC of wheat. Although KASP markers were not located within predicted candidate genes,there may be a genetic linkage between them.

5.Conclusion

We have shown that a genome-wide association study effectively detected both stable and environment-specific QTLs for GPC,GSC,and WGC. Multi-trait chromosomal regions have been detected,and the region on chr2DL associated with GPC may be particularly useful in MAS following proper validation. In the context of nutritional quality,five QTL regions were potentially novel and control GSC or WGC,implying the possibility of using vegetation indices for the indirect assessment of certain nutritional quality traits. Although some objective limitations for position comparisons existed,we tried to improve the analysis threshold and made the comparisons with multiple linked markers to obtain the maximum information.Two KASP markers were successfully developed for validating the WGC phenotype and are expected to be used in wheat breeding programs.

Acknowledgements

The authors thank Profs.Li Lihui and Li Xiuquan(Chinese Academy of Agricultural Sciences) for providing plant materials. This work was supported by the National Key Research and Development Program of China (2017YFD0100900,2016YFD0102000 and 2016YFD0100100),the International Science and Technology Cooperation and Exchanges Programs of Science and Technology Department of Sichuan Province,China (2019YFH0063),and the Sichuan Science and Technology Program,China (2022ZDZX0014).

Declaration of competing interest

The authors declare that they have no conflict of interest.

Appendicesassociated with this paper are available on http://www.ChinaAgriSci.com/V2/En/appendix.htm