Genome-wide association and linkage mapping strategies reveal the genetic loci and candidate genes of important agronomic traits in Sichuan wheat

ZHANG Zhi-peng, Ll Zhen, HE Fang, Lü Ji-juan, XlE Bin, Yl Xiao-yu, Ll Jia-min, Ll

Jing5, SONG Jing-han6, PU Zhi-en1,2, MA Jian1,3, PENG Yuan-ying1,3, CHEN Guo-yue1,3, WEl Yu-ming1,3,ZHENG You-liang1,3, Ll Wei1,2,3#

1 State Key Laboratory of Crop Gene Exploration and Utilization in Southwest China, Sichuan Agricultural University, Chengdu 611130, P.R.China

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

3 Triticeae Research Institute, Sichuan Agricultural University, Chengdu 611130, P.R.China

4 Sichuan Provincial Seed Station, Chengdu 610044, P.R.China

5 Huaiyin Institute of Agricultural Sciences of Xuhuai Area in Jiangsu, Huai’an 223001, P.R.China

6 Beijing Foreign Studies University, Beijing 100081, P.R.China

Abstract Increasing wheat yield is a long-term goal for wheat breeders around the world.Exploiting elite genetic resources and dissecting the genetic basis of important agronomic traits in wheat are the necessary methods for high-yield wheat breeding.This study evaluated nine crucial agronomic traits found in a natural population of 156 wheat varieties and 77 landraces from Sichuan, China in seven environments over two years.The results of this investigation of agronomic traits showed that the landraces had more tillers and higher kernel numbers per spike (KNS), while the breeding varieties had higher thousand-kernel weight (TKW) and kernel weight per spike (KWS).The generalized heritability (H2) values of the nine agronomic traits varied from 0.74 to 0.95.Structure analysis suggested that the natural population could be divided into three groups using 43 198 single nucleotide polymorphism (SNP) markers from the wheat 55K SNP chip.A total of 67 quantitative trait loci (QTLs) were identified by the genome-wide association study (GWAS) analysis based on the Q+K method of a mixed linear model.Three important QTLs were analyzed in this study.Four haplotypes of QFTN.sicau-7BL.1 for fertile tillers number (FTN), three haplotypes of QKNS.sicau-1AL.2 for KNS, and four haplotypes of QTKW.sicau-3BS.1 for TKW were detected.FTN-Hap2, KNS-Hap1, and TKW-Hap2 were excellent haplotypes for each QTL based on the yield performance of 42 varieties in regional trials from 2002 to 2013.The varieties with all three haplotypes showed the highest yield compared to those with either two haplotypes or one haplotype.In addition,the KASP-AX-108866053 marker of QTL QKNS.sicau-1AL.2 was successfully distinguished between three haplotypes(or alleles) in 63 varieties based on the number of kernels per spike in regional trials between 2018 and 2021.These genetic loci and reliable makers can be applied in marker-assisted selection or map-based gene cloning for the genetic improvement of wheat yield.

Keywords: Sichuan wheat, GWAS, yield traits, haplotype analysis, KASP

1.lntroduction

Wheat is one of the most important food crops in the world (Dwivediet al.2007).The genome of bread wheat is composed of three subgenomes, namely A, B, and D(Walkowiaket al.2020).Improving the yield per unit area is one of the main goals in wheat breeding, as well as an essential method to meet the growing food demand in China.Several yield traits influence the wheat yield.Spike number per unit area, kernel number per spike(KNS), and thousand-kernel weight (TKW) are the main factors that determine wheat yield per unit area (Slaferet al.2014; Philippet al.2018).Effective tillering impacts the spike number per unit area.Spike length, density,total spikelet number per spike (TSN), and KNS also affect yield.Plant height (PH) affects reproductive growth through the vegetative growth process as well as kernel weight (Denget al.2011; Zhang Jet al.2018; Brinton and Uauy 2019; Maet al.2019).The length of the awn also impacts kernel weight, thereby affecting the spike grain weight and TKW.Thus, increasing wheat yield is a continuous coordination and optimization process among different agronomic traits (Fernandeset al.2022).

For the three factors that directly affect wheat yield, i.e.,the number of ears per unit area, the number of kernels per ear, and TKW, breeders have performed many studies related to identifying and characterizing the key genes.At present, a series of validated and recognized genes that directly affect wheat yield, includingTaGW2,TaGW7,TaD27,TaPIN1s,KAT-2B,TaCol-B5, andWAPO-A1, and a large number of associated quantitative trait loci (QTLs)have been discovered (Zhaiet al.2018; Wanget al.2019;Chenet al.2020; Xinet al.2020; Yanget al.2020; Zhaoet al.2020; Gaoet al.2021; Quet al.2021; Kuzayet al.2022; Liuet al.2022; Zhanget al.2022).Although PH,ear length, ear density, and awn length are not the main factors that directly affect yield, they indirectly influence the final yield of wheat.Hence, breeders have also focused on these traits, e.g.,Rht14,ALI-1,TaHST1L, andQand many related QTLs (Xieet al.2018; Vikheet al.2019; Wanget al.2020; Zhaoet al.2020; Zhouet al.2020; Jiet al.2021; Dinget al.2022).Despite all the studies on a large number of genes and QTLs that affect wheat yield, the genetic basis of wheat yield traits remains unclear.Therefore, further exploration of yield-related genes and QTLs is required to promote wheat molecular marker breeding and to lay the foundation for further improvements in wheat yield.

The wheat planting area in Sichuan is one of the six major wheat-producing areas in China.Good growth and high yield of wheat are inseparable from adaptation to the environment and the development and utilization of ecological resources.After years of natural and artificial selection, Sichuan wheat landraces have developed the characteristics of early maturation, hybrid compatibility,and multiple tillers.By exploiting local wheat resources in Sichuan, a few new loci for yield-related traits have been discovered.The discovery of the remote hybrid incompatibility geneKr4in Sichuan wheat indicated that there may be other excellent genetic resources yet to be discovered (Luoet al.1992).Based on the original local varieties, 19 quality traits of 67 Sichuan wheat local varieties were clustered, and five high-quality wheat varieties were selected for wheat variety improvement(Chenet al.2006).Cluster analysis was performed on eight varieties and eight Sichuan wheat landraces with simple sequence repeat (SSR) markers, which proved that the landrace ‘Zhongchun’ was the breeding line of‘Chengduguangtou’ (Chenet al.2006).In general, as an important germplasm resource, Sichuan local wheat shows significant differences in its varied yield-related traits.Although a few important genes have been unearthed,there are still many other genes yet to be discovered.

In this study, nine agronomic traits in seven environments, including the best linear unbiased prediction (BLUP) values, were investigated.The genetic diversity, population structure, and linkage imbalance level of Sichuan wheat germplasms were evaluated by a highthroughput 55K single nucleotide polymorphism (SNP)chip covering the whole genome of wheat.A genomewide association study (GWAS) of phenotype and molecular markers was carried out to identify the QTLs closely related to the target traits, with haplotype analysis and optimal combination screening being performed on the QTLs of TKW, KNS, and fertile tiller number (FTN)that directly affect wheat yield.Effective kompetitive allele-specific polymerase chain reaction (KASP) markers were then developed for the important QTLs.The overall goal of this study was to provide valuable insights into the genetic structure of wheat yield traits and the use of genetic resources to promote wheat breeding.

2.Materials and methods

2.1.Plant materials

The plant varieties comprised 233 wheat germplasms from Sichuan Province, China including 77 landraces and 156 breeding lines (Appendix A).In addition, 42 wheat varieties that participated in the 2002–2013 regional trials in Sichuan Province (Appendix B), and 63 wheat varieties in the 2018–2021 regional trials in Sichuan Province(Appendix C) were used.

2.2.Field trials for trait evaluation

A total of 233 wheat accessions in the association panel were planted in six environments, namely Chongzhou in 2019 and 2020 (CZ19 and CZ20), Wenjiang in 2019 and 2020 (WJ19 and WJ20), and Ya’an in 2019 and 2020(YA19 and YA20).All tested accessions were planted in three non-replicated rows in each environment.The rows were 2 m in length and 0.3 m apart.A total of 20 seeds of each accession were sown with 0.1 m inter-plant spacing.In each cropping season, fields were managed following the standard practices of local wheat production.

Regional trials of new wheat varieties in Sichuan Province were conducted to identify the high yield, stress resistance,and adaptability of the varieties (species) bred or introduced in Sichuan Province.Excellent wheat varieties with better comprehensive traits and meeting the new variety approval standards were selected and recommended for approval.For example, the 2021 experiment consisted of five groups,each with 14 varieties (species) including the control,and was conducted in 10 pilot studies.The control was Mianmai 367, and the experiment used a randomized block design with three replicates.The area of each plot was 13.33 m2, the walkway between the cells was 0.40 m,and the walkway between the repetitions was 0.5 m, with protective rows around the test.This study was selected from the data from the regional trials of 42 varieties from 2002 to 2013 to screen and label the haplotypes.In addition, the genotyping results ofKASP-AX-108866053were verified using the data from the 63 varieties used in regional trials between 2018 and 2021.The Seed Station of Sichuan Province provided the data.

Surveys and counts of wheat yield-related traits at maturity and after harvest included FTN, PH, spike length(SL), awn length (AL), TSN, spikelet density (SD), KNS,KWS, and TKW (Eltaheret al.2021).FTN, PH, SL, AL,TSN, SD, and KNS represent the measured average values of five plants of a single variety.TKW is the average weight of three measurements from 1 000 kernels randomly harvested from each wheat variety.KWS is the average weight measured after threshing five different spikes of a single variety harvested in the field (Maet al.2012).

2.3.Phenotypic data analysis

Variance and correlation analyses of all traits were performed with the “cor” package in R (version 1.4.1106,2021, https://github.com/rstudio/rstudio) using the Pearson correlation method (Pandeyet al.2015).The BLUP value of each trait was calculated using the R Software package “l(fā)me4”.The generalized heritability(H2) of each trait was calculated asH2=VG/(VG+VE),whereVGrepresents the genotype variance andVEis the environmental variance (Smithet al.1998).Genotype and environmental variances were calculated by BLUP.The highest and lowest values, mean values, standard deviation (STDEV), and coefficient of variation (CV) of each trait were also calculated.

2.4.Genotyping and genetic analysis

Genomic DNA was extracted from the mixed leaves of 1-wkold seedlings through the hexadecyltrimethylammonium bromide (CTAB) method (Murray and Thompson 1980).Genotyping of the 233 accessions was conducted by CapitalBio Technology (Beijing, China) using the wheat 55K SNP array.The original data were controlled according to the screening criteria with a minimum allele frequency (MAF) greater than 5% and a missing rate less than 10%, to eliminate invalid SNP markers.The Software POWERMARKER v.3.2 (2006, https://brcwebportal.cos.ncsu.edu/powermarker/index.html) was used to analyze the gene diversity index, while the polymorphism information content index (PIC) was used to evaluate the genetic diversity of the Sichuan wheat germplasms (Liu 2005).

2.5.Population genetics and linkage disequilibrium(LD)

The population structure (Q-matrix) was analyzed with STRUCTURE v.2.3.4 (2012, https://web.stanford.edu/group/pritchardlab/structure.html) using the Bayesian Clustering Model (Hubiszet al.2009).Based on the admixture model,five independent runs were performed with K-values from 1 to 10.Moreover, 10 000 replicates of the burn-in cycles and Monte Carlo Markov Chain (MCMC) were set at the same time.The web-based informatics tool STRUCTURE HARVESTER (2014, http://taylor0.biology.ucla.edu/struct_harvest/) was employed using the ΔKmethod to confirm the optimal K value (Earl and Vonholdt 2012).The kinship matrix (K-matrix) of Sichuan wheat germplasms was analyzed using the Software Tassel v.5.2.38 (2017, https://tassel.bitbucket.io/) (Bradburyet al.2007).PopLDdecay was used to calculate the LD coefficient (r2) between pairwise SNPs (Zhanget al.2019).The parameters were set as “./PopLDdecay -InVCF./snp.vcf -MaxDist 1000-OutType 3 -OutStat out” and the results were used to estimate the LD decay.The LD attenuation map was then depicted with the R package “ggplot2” using ther2of all significant site combinations selected with the screening criteria of pDiseq<0.001 andr2>0.1 (Bradburyet al.2007;Ginestet 2011).All the high-confidence associated loci included in the half-decay distance regions on the same chromosome were defined as the same QTL block.

2.6.GWAS

The mixed linear model (Q+K) in Tassel v.5.2.38 Software was used to analyze the correlations between the germplasm characters and genotypes in 233 Sichuan wheat samples.After quality control, effective SNP molecular markers were obtained for GWAS of the traits, including FTN, PH, SL, AL, TSN, KNS, SD, KWS,and TKW (Bradburyet al.2007; Zhanget al.2010).A Manhattan plot was generated by the R package “CMplot”to reveal the association results (https://github.com/YinLiLin/R-CMplot).

Significant SNPs detected in the same LD block formed a QTL.The distance between two SNPs on each side of a QTL region was considered the QTL interval.The annotated genes within the physical range of identified QTLs were extracted from the IWGSC Reference Genome v.1.1.

2.7.Haplotype analysis

Haploview v.4.2 was used to divide haplotype linkage regions and calculate genotypes of the related QTLs(Barrettet al.2005).Haplotype (allelic or allelic locus combination) analysis was performed on the new potential QTLs associated with FTN, KNS, and TKW, and the genotype-missing varieties were excluded.

2.8.Marker development and evaluation

KASP markers were developed based on the different sites in the haplotype linkage region using the PolyMarker website (http://www.polymarker.info/).The forward primers were 5′-TTTTGCTGACTTCGCATGGT-3′ and 5′-TTTTGCTGACTTCGCATGGC-3′, and the reverse primer was 5′-TTATTACCATTGAGTTGTGGGGA-3′.Genotyping of the markers was performed using a CFX96 Real-Time PCR System (Bio-Rad, USA).The PCR reaction system was as follows: template DNA (50 ng μL–1)1 μL, KASP 2× Master Mix 5 μL, primer Mix (10 μmol L–1)0.24 μL, reverse primer 0.6 μL, and ddH2O 2.9 μL.The PCR procedure consisted of Stage I, 94°C for 15 min, one cycle; Stage II, 94°C for 15 s and 65–55°C for 1 min; and Stage III, 94°C for 20 s and 57°C for 1 min, 42 cycles.This test population located by GWAS was used to verify the validity of the markers and analyze the trait association specificity.In addition, the test area variety population(2018–2021) was used to verify the validity of the KASP markers for molecular-assisted selection breeding.

3.Results

3.1.Phenotypic variations and correlations

The phenotypic variations of nine yield traits in seven different environments (six virtual environments and the BLUP value) were evaluated (Fig.1; Table 1).The variation coefficients of wheat FTN, PH, SL, AL, TSN, KNS, SD,KWS, and TKW were all greater than 10%, except for TSN,in the seven environments (Appendix D).The average FTN ranged from 4.14 (20CZ) to 8.03 (19CZ) in the seven environments.The CV of FTN was the highest at 64.00%in 20CZ and the lowest at 31.00% using the BLUP value.The mean values of KNS were 46.41 in 20WJ and 54.61 in 19WJ, and the CV value of KNS was the highest in 19CZ(31.00%).The average TKW ranged from 34.22 to 37.22 in the different environments, and the CV in 20YA was the highest (27.00%).However, the BLUP value was the lowest (17.00%).These results showed that the genetic variability of these traits was high among the varieties.

The average values of FTN and KNS of landraces(7.69 and 54.59) were higher compared to the breeding varieties (5.21 and 48.39), respectively.In contrast,TKW of the breeding varieties (38.11) was significantly higher than the landraces (30.71).Variation coefficient analysis of TKW showed that the landraces were equal to the breeding varieties.Variation coefficient analyses of FTN and KNS also showed that the variation coefficients of landraces were more significant compared to the breeding varieties.A total of 156 breeding varieties in this population were developed from 1997 to 2016 at an average interval of five years, with the breeding varieties being divided into four stages (Table 2).The average FTN in the first stage was the highest (5.78), while those in the second and fourth stages were the lowest.In the second stage, KNS and TKW were the lowest (KNS=29.98,TKW=29.27), whereas, in the fourth stage, KNS and TKW were the highest (KNS=51.78, TKW=37.60).

Fig.1 Box plots of phenotypic statistics for nine correlated traits across multiple environments.19 and 20, 2019 and 2020,respectively.CZ, Chongzhou; WJ, Wenjiang; YA, Ya’an, Sichuan Province, China; BLUP, best linear unbiased prediction.AL,awn length; FTN, fertile tillers number; KNS, kernel number per spike; PH, plant height; SD, spike density; KWS, kernel weight per spike; SL, spike length; TKW, thousand-kernel weight; TSN, total spikelet number per spike.

A significant positive correlation was observed between TKW, KWS, and KNS, and between PH, FTN, and SD.However, TKW, KWS, and KNS, which directly affect wheat yield, were significantly negatively correlated with the general agronomic traits of wheat, including PH, FTN,and SD.In addition, AL and SL were significantly positively correlated with TKW and KWS, but significantly negatively correlated with PH, FTN, and SD.TSN was unique among the nine traits, as it was significantly positively correlated with SL and KWS, but significantly negatively correlated with FTN, PH, AL, SD, KWS, and TKW (Fig.2).In addition, the correlations of FTN, KNS, and TKWwith different environments showed that the correlations between different environments were significant or extremely significant and positive (Fig.3-A–C).

Table 1 Analysis of variance of the agronomic traits in Sichuan landraces and breeding varieties of wheat1)

Table 2 The changes in fertile tillers number (FTN), spike grain number (KNS), and thousand-kernel weight (TKW) at different stages1)

TheH2values of FTN, PH, SL, AL, TSN, KNS, SD,KWS, and TKW were 0.74, 0.95, 0.83, 0.92, 0.88, 0.86,0.88, 0.84, and 0.80, respectively.In general,H2was higher than 70% (Appendix D), indicating that the genetic factors less affected by environments mainly regulated the phenotypes of these traits.

Fig.2 Correlation matrix for nine yield-related agronomic traits.AL, awn length; SL, spike length; TKW, thousand-kernel weight;KWS, kernel weight per spike; KNS, kernel number per spike;TSN, total spikelet number per spike; PH, plant height; FTN,fertile tillers number; SD, spike density.

3.2.Genomic variation, LD, and population structure

A total of 53 063 SNP markers were used to scan the whole genome of 233 accessions.A total of 43 198 effective molecular markers were identified on the 21 wheat chromosomes.The numbers of SNPs from the A and B subgenomes were more significant compared to that of subgenome D.The numbers of SNPs per chromosome ranged from 916 on 4D to 2 504 on 2A.The averages of the index of genome-wide gene diversity and the polymorphism index of all SNPs were 0.381 and 0.302, respectively.Among the 21 chromosomes, the 5B chromosome showed the highest gene diversity (0.422), and the 4A chromosome was the lowest (0.314).Also, the 5B chromosome showed the highest PIC (0.329), and the 4A chromosome was the lowest (0.258) (Appendices E and F).The structure results simulated by STRUCTURE HARVESTER showed that the structure was most stable when K=3, suggesting that the population of Sichuan wheat germplasms could be divided into three groups (Appendix G).

The whole-genome LD decay distance of Sichuan wheat germplasm was evaluated using 43 198 effective SNP markers and the attenuation map was drawn according to the linkage disequilibrium sites of the whole genome.According to the fitting curve analysis, the LD decay distance was 1.642 Mb (Appendix H) when the labeledr2was attenuated to half (r2=0.474), and this value was used as the maximum range of the QTL region.

3.3.GWAS

Based on theQ+Kmethod of a mixed linear model, GWAS was carried out for FTN, PH, SL, AL, TSN, KNS, SD,KWS, and TKW in seven environments (six environments and one BLUP value) using 43 198 SNP markers.The 77 marker loci obtained were significantly associated with yield-related traits, the results of which were visualized using Manhattan plots and theQ–Qdiagram (Appendices I and J).The number of related SNP markers for each trait varied from 1 of TKW to 24 of PH.These tags comprised 67 QTLs based on the previously defined LD decay distance.The numbers of associated QTLs for each trait ranged from 1 of TKW to 22 of PH.The QTLs associated with phenotypic variations (R2) ranged from 5.1 to 20.1%.The numbers of QTLs associated with FTN, PH, SL, AL,TSN, KNS, SD, KWS, and TKW were 5, 22, 3, 8, 12, 7, 6,3, and 1, respectively.Of them,QAL.Sicau-6BL.1showed the largest explicable phenotypic variation (R2=0.20).The markerAX-108866053was associated with KNS and PH,AX-109913257with TSN and PH, andAX-109933220with KNS and TKW.The loci validated in multiple natural environments are multiple effector loci.A total of 67 QTLs were significant in at least five environments, of which 43 QTLs were significant in all environments (1 QTL of FTN, 24 of PH, 11 of AL, and 7 of KNS) (Appendix K).Sixty-seven significant QTLs were unevenly distributed on the three subgenomes (20 on the D subgenome,24 on the A subgenome, and 23 on the B subgenome).Their distribution showed the most QTLs on the 2D chromosome (16), with 7 QTLs each on chromosomes 1A and 1B, and one QTL each on the 2B, 3D, 4A, 5B, and 6A chromosomes, respectively.No QTLs of agronomic traits were identified on chromosomes 4D and 5D.

3.4.Haplotype analysis

Fig.4 Haplotype analysis of fertile tillers number (FTN), kernel number per spike (KNS), and thousand-kernel weight (TKW) in 233 Sichuan wheat germplasm populations.A, D, and G, Manhattan maps of FTN, KNS and TKW, respectively.B, E, and H,linkage disequilibrium plots of QFTN.sicau-7BL.1, QKNS.sicau-1AL.2, and QTKW.sicau-3BS.1, respectively (in which the values of linkage disequilibrium (r2) between SNP pairs are presented according to the grey level, and the darker backgrounds show higher r2 values).C, F, and I, Boxplot displays the mean FTN, KNS, and TKW of the accessions carrying the different haplotypes,respectively.ns, no significance at P=0.05; ＊＊ and ＊＊＊＊, significances at the P=0.01 and P=0.0001 probability levels, respectively.

Among the QTLs that were located in all environments and the GWAS results of the BLUP value, the present study selected the one with the highest contribution rate to each trait for marker development.The potential novel QTLs with the highest contribution rates, namelyQFTN.sicau-7BL.1(R2=6.4–17.6%, detected in seven environments),QKNS.sicau-1AL.2(R2=6.5–10.5%,detected in seven environments), andQTKW.sicau-3BS.1(R2=6.5–11.2%, detected in six environments), were the haplotypes analyzed using HaploView v.4.2 (2006,https://www.broadinstitute.org/haploview/haploview).The linkage regions of the three QTLs were 1 055, 2 155, and 2 035 kb, respectively (Fig.4-B, E, and H).A statistical analysis was carried out on the average yield per hectare of 42 breeding experiments used in the regional trials of Sichuan Province from 2002 to 2013 among the 156 breeding experiments (Appendix B).

QFTN.sicau-7BL.1had three haplotypes, of whichFTN-hap1was displayed in 169 varieties, and bothFTNhap1andFTN-hap3were displayed in 23 varieties,respectively.The mean FTN ofFTN-hap1was 5.70,which was significantly lower than the mean FTN of 6.47 forFTN-hap2andFTN-hap3(Table 3).FTN-hap1was significantly different fromFTN-hap2andFTN-hap3by theT-test.At the same time, a comparison of the average yields of the 42 regional trial varieties tested from 2002 to 2013 also followed the same trend.The average FTN of the varieties containingFTN-Hap2was higher than those withoutFTN-Hap2.Similarly, in the regional trial varieties,the average yield of the varieties withFTN-Hap2(5 445.68 kg ha–1) was higher than those withoutFTNHap2(5 429.97 kg ha–1).The FTN values of the varieties withFTN-Hap2were significantly higher than those withoutFTN-Hap2(Fig.4-C).This indicated thatFTNHap2could significantly increase the FTN value and improve the final yield to a certain extent (Appendix L).

The haplotype ofQKNS.sicau-1AL.2was also divided into three parts, with average KNS values of 58.81, 52.27,and 52.99, respectively (Table 3).Moreover,KNS-Hap1was found in 140 varieties, whileKNS-Hap2andKNSHap3were found in 38 and 6 varieties, respectively.TheT-test results showed thatKNS-Hap1was significantly better thanKNS-Hap2andKNS-Hap3(Fig.4-F), and this was verified in the regional trial varieties.The average yield withKNS-Hap1was 5 431.96 kg ha–1, which was higher than the 5 382.81 kg ha–1average of those withoutKNS-Hap1.This indicated thatKNS-Hap1could effectively improve KNS with a promoting effect on the final yield(Appendix L).

Similarly,QTKW.sicau-3BS.1had three haplotypes.The average TKW values of the three haplotypes were 35.61, 38.01, and 38.01 g, respectively, which were significantly different by theT-test (Table 3).TKW-Hap2andTKW-Hap3were found in 40 varieties, respectively,with the average TKW values of these two haplotypes being significantly higher than the average TKW values of the 139 varieties containingTKW-Hap1(Fig.4-I).In the regional trial varieties, the average yield and TKW ofTKW-Hap1(5 400.94 kg ha–1and 43.56 g) were lower thanTKW-Hap2(5 548.29 kg ha–1and 44.91 g) andTKW-Hap3(5 548.29 kg ha–1and 44.91 g).This indicated that bothTKW-Hap2andTKW-Hap3were excellent haplotypes ofQTKW.sicau-3BS.1, which significantly increased TKW(Appendix L).

3.5.Haplotype combination analysis

The excellent haplotype analysis of FTN, KNS, and TKW showed that the numbers of varieties containing haplotypesTKW-Hap2,KNS-Hap1,TKW-Hap2/KNSHap1,KNS-Hap1/FTN-Hap2, andTKW-Hap2/KNS-Hap1/FTN-Hap2, were 6, 82, 24, 21, and 6, respectively.The phenotypes of FTN, KNS, and TKW were further improved by increasing the number of combinations of excellent haplotypes among the varieties (Table 4).

A statistical analysis of haplotype combinations and phenotypes of the 42 varieties in the wheat cultivars tested in the regional trials in Sichuan Province showed that only one regional trial variety containingTKW-Hap2had a yield of 5 437.65 kg ha–1.The average yields and numbers of the regional trial varieties containingKNSHap1,TKW-Hap2/KNS-Hap1,KNS-Hap1/FTN-Hap2,andTKW-Hap2/KNS-Hap1/FTN-Hap2were 5 532.78,5 536.05, 5 568.11, 5 673 kg ha–1and 28, 3, 4, and 2,respectively.In summary, the yields of these varieties with a combination of two haplotypes were better than those with one haplotype.Also, the yields of the varieties with a combination of three excellent haplotypes were better than those with two haplotypes.As the number of haplotypes in the varieties increased, the yields of the varieties also increased gradually, with a reduction in the number of varieties (Table 4).

3.6.Development and validation of KASP markers for KNS marker-assisted selection

KASP markerKASP-AX-108866053was developed based on the excellent haplotypeKNS-Hap1ofQKNS.sicau-1AL.2, which was associated with the kernel number per spike.The allelic variation of this locus was A/G.This marker divided the population into three types (Fig.5;Table 5), where type 1 (KNS-Hap1) significantly correlated with KNS in six environments by phenotypic verification(Table 5).Thus,KNS-Hap1was found to be closely linked toQKNS.sicau-1AL.2.In addition, the KASP marker was successfully typed in the 63 varieties in the 2018–2021 regional trials, but the difference in the number of grains per panicle was not significant.

4.Discussion

4.1.Sichuan local wheat varieties have many excellent traits that can be applied to breeding

Germplasm resources are important gene sources for wheat breeding, and using new gene resources helps in the breeding of novel varieties with high yield, good quality,disease resistance, and wide adaptability.Germplasm resources of wheat include ancient local varieties, newly developed and promoted varieties, synthetic genetic varieties, and wild relatives.Wheat landraces are generally formed using traditional planting methods to suit the local environment.In the current study, significant differences were observed in phenotypic characteristics between the landraces and the breeding varieties (Table 1).Usually, the outstanding features of landraces include FTN, TSN, and SD, while the breeding varieties show good AL, KWS, and TKW.Moreover, big spikes and high TKW germplasms are obtained mainly from breeding varieties by screening, while more tillers and dense-eared germplasms are obtained from landraces.Hence, in different environments, landraces show better adaptability,diversity, and heritability (Jaradat 2011).In this study, the average values of tiller number, panicle number, panicle length, and panicle density of the Sichuan local varieties were higher compared to those of the breeding varieties,whereas TKW was lower in Sichuan local varieties than the breeding varieties.This difference may be due to local farmers choosing varieties with higher seed yields to sustain future plantings (Yeet al.2019).Although the existing breeding varieties in Sichuan had the advantages of high yield and stability, the Sichuan local varieties possessed many excellent characteristics which could be used to further improve the breeding varieties.Thus,Sichuan local wheat varieties can be considered important germplasm resources to further improve multiple traits in the breeding varieties in the future.

Table 4 Phenotype statistics of different haplotypes or combinations of haplotypes1)

Fig.5 Real-time PCR amplification results of representative KASP marker KASP-AX-108866053 in some Sichuan wheat germplasms.A, Typing results in the population.B, typing results in the area test population.Blue, homozygous type 1, allele AA; orange,homozygous type 2, allele GG; black, heterozygous type, allele missing.RUF, relative fluorescence unit; FAM, carboxyfluorescein;HEX, hexachloro fluorescein.

Table 5 Typing results and correlation analysis of KASP-AX-108866053 designed for kernel number per spike (KNS) in the population1)

4.2.Analysis of the QTL results using mapped landrace and breeding varieties

Landrace and breeding varieties are also strictly separated at the molecular level.In this study, landraces and breeding varieties were classified into three categories according to the Bayesian model classification method.The first group was dominated by local varieties,while the second and third groups were dominated by the breeding varieties.Significant differences were observed in the affinities between landraces and breeding varieties, including phenotypic differences and molecular components.Hence, using Sichuan landraces in breeding can effectively broaden the genetic background of wheat.

A total of 107 significant markers for yield-related traits were obtained on 19 chromosomes, except for 4D and 5D.Based on the previously obtained decay distance of 1.642 Mb, they were divided into 67 QTLs.Studies on the landraces have shown that they contain genes related to yield control, including theTaGW2gene for the regulation of kernel width, genesTaCKX6-D1aandTaCKX6-D1bfor kernel weight control, geneTaGS5-A1for the regulation of TKW, and geneTaGS5-3A-Tfor kernel size (Maet al.2016;Wanget al.2016; Zhang Yet al.2018; Kabir and Nonhebel 2021).These findings suggested that landraces may still have the ability to affect yield, a trait that is considered important for genes yet to be identified.On the other hand, research on local wheat varieties also demonstrated this possibility.For example, an association analysis was performed on 23 agronomic traits of 723 landrace wheat varieties in China, and 149 new loci significantly associated with 21 traits were obtained (Liuet al.2017).Yield traits,including KNS and TKW, were mapped using GWAS in 260 durum wheat samples (including 163 landraces) from around the world (Salsmanet al.2021).

Xuet al.(2017) mapped the recombinant inbred lines(RIL) population constructed from wheat high-tiller dwarf mutants.Three QTLs for PH and four QTLs for FTN were obtained, which were located on the 3B chromosome.In this study,QT.nau-3B(Xcfb3376-Xgwm66) overlapping withQFTN.sicau-3BS.1was localized, in whichQFTN.sicau-3BS.1had a higher explained variance.Twelve QTLs, includingQFTN.sicau-3BS.1, that overlapped or were associated with previously reported yield-related traits were obtained, and the remaining 55 QTLs were completely new, which to some extent proves the reliability of the results of this study.An analysis of a common wheat population using genome-wide mapping identified 60 QTLs associated with panicles and 85 QTLs associated with other agronomic traits on 18 chromosomes, except for 5D,6D, and 7D (Echeverry-Solarteet al.2015).A mapping analysis of 244 populations of landrace and breeding wheat varieties using GWAS identified 13 QTLs that were highly associated with yield-related traits under disease stress (Yeet al.2019).The 67 QTLs for yield-related traits identified in this study were also widely distributed on the chromosomes, which indicated that wheat yield-related traits are complex traits driven by the joint regulation of multiple genes.Overall, the wheat yield was established as a comprehensive and complex trait in this study.By mapping local wheat varieties, many reliable QTLs could be obtained and used for the fine mapping of subsequent yield-related traits and future molecular markers.

4.3.Excellent haplotype analysis

A set of chromosomal molecular markers that are closely linked on the same chromosome, called haplotypes, tend to be inherited together during the genetic process (Andersen and Lübberstedt 2003).Haplotype associations based on decay distances are often more robust compared to individual markers (Wanget al.2015).Haplotype analysis has been widely used in wheat.For example, three differentTaPYL4shaplotypes were identified by developing three derived Cleaved Amplified Polymorphic Sequences(dCAPS) markers.Plants containingTaPYL4-2A-Hap2showed greater height and those withTaPYL4-2B-Hap1showed more kernels per spike (Wuet al.2022).TheTaD11gene affecting grain size in wheat was discovered using the protein sequence alignment ofD11(OsD11) in rice, and an insertion and deletion marker (DA-InDel2) was developed to distinguish between two differentTaD11-2Ahaplotypes (Xuet al.2022).In addition,TaD11-2A-HapIwas found to be an excellent haplotype for significantly improving wheat kernel-related traits, including kernel width and TKW, and it has undergone positive selection in the process of wheat breeding in China.In this study, the respective excellent haplotypes, namelyTKW-Hap2,KNSHap1, andFTN-Hap2, which were significantly correlated with yield, were obtained for TKW, KNS, and SPN.The accuracy of the haplotype analysis results was proven by verification in regional trials conducted from 2002 to 2013.

In wheat breeding, multiple excellent haplotypes with different traits could be selected for molecular-assisted selective breeding (Yuet al.2020; Scottet al.2021;Yuanet al.2021).In this study, different haplotypes corresponding to FTN, kernel number per panicle, and TKW were combined and analyzed, and theFTN-Hap2/KNS-Hap1/TKW-Hap2type was obtained and showed the highest yield.This suggested that wheat yield can ultimately be further improved by integrating the excellent haplotypes of different traits (Nyineet al.2021).The yield performance of a single haplotype and that of the combined haplotypes mutually verified the accuracy of the obtained excellent haplotype and the combined haplotypes to a certain extent.Thus, the yield of wheat can be further improved by increasing the number of excellent haplotype combinations in the varieties during breeding (Liuet al.2022).In addition, the number of varieties decreased with an increase in the number of excellent haplotypes,indicating that the excellent haplotypes in this study were not fully applied to breeding work.In short, the various excellent haplotypes and combinations of excellent haplotypes obtained in this study can provide a reference for the future breeding of new high-yielding wheat varieties and the accurate and stable selection of breeding parents.

4.4.Future targets and the emphasis of wheat variety breeding in Sichuan

In this study, the genetic diversity of breeding varieties was higher than the landraces at three subgenomic levels.This suggested that the molecular genetic diversity of breeding varieties developed in Sichuan was high, which was similar to the reports of other researchers (Whiteet al.2008).The breeding varieties were divided into four stages at intervals of five years.The genetic diversity and quantity of wheat in the last three stages were significantly higher than in the first stage, of which the second and third stages were the highest with the fourth stage being intermediate.With the passage of breeding time, the number of breeding varieties and their molecular genetic diversities increased significantly, which is strong evidence for the progress of breeding work in Sichuan (Sthapitet al.2022).Meanwhile,phenotypic analyses of breeding and landrace varieties in different periods showed that the FTN of the breeding varieties decreased gradually.In contrast, KNS and TKW increased with time (Table 2).The typing results from the usage of the KASP marker,KASP-AX-108866053, in the regional trial varieties from 2018 to 2021 further illustrated that the number of kernels per ear was improved by artificial selection in the breeding varieties.This not only shows the results of the breeding work but also proves that FTN can be the focus of future wheat breeding.This study investigated the past history of Sichuan wheat germplasms from multiple aspects in order to provide directions for future breeding.

5.Conclusion

A total of 63 QTLs for yield-related traits were mapped.Through haplotype analysis of the key QTLs for the three elements of wheat yield, includingQFTN.sicau-7BL.1,QKNS.sicau-1AL.2, andQTKW.sicau-3BS.1, excellent haplotypes for each trait were obtained.The results indicated that the excellent haplotype combinationTKWHap2/KNS-Hap1/FTN-Hap2could significantly improve the actual yield of wheat.A tightly linked KASP marker,which could be used for molecular marker breeding, was developed forQKNS.sicau-1AL.2from among the three key QTLs, and it was further verified by the typing of new varieties.Our results provide a framework for the careful selection of parents for future crosses.

Acknowledgements

This work was supported by the Sichuan Science and Technology Program, China (2022ZDZX0014 and 2021YFYZ0002) and the Plan of Tianfu Qingcheng of Sichuan Province, China.

Declaration of competing interest

The authors declare that they have no conflict of interest.Appendicesassociated with this paper are available on https://doi.org/10.1016/j.jia.2023.02.030