Evaluation and association analysis of quality traits in Chinese winter wheat landraces on the Qinghai-Tibet Plateau Authors: Xingxin Han, Yuqing Zhang, Wanwei Hou, Miaomiao Huang Published: 2026 ABSTRACT BackgroundWheat grain quality is a complex traits governed by both genetic and environmental factors. Chinese winter wheat landraces on the Qinghai–Tibet Plateau harbor valuable genetic diversity, yet the loci underlying their grain quality remain largely unexplored. This study aimed to identify stable SNP loci, favorable haplotypes, and candidate genes for grain quality traits.MethodsEight quality traits—moisture content (MC), flour water absorption (WA), grain protein content (GPC), wet gluten content (WGC), dough stability time (ST), dough development time (FT), grain hardness (HD), and test weight (TW)—were assessed in 279 winter wheat landraces across four environments on the Qinghai–Tibet Plateau. A genome-wide association study (GWAS) was conducted using a 120 K SNP array.ResultsAll eight traits showed wide phenotypic variation. In total, 62 stable SNP loci were detected across environments (P < 0.001), explaining 3.86%–15.64% of the phenotypic variance; 19 of these showed pleiotropic effects. A pleiotropic locus on chromosome 1B (1B:610760001) was significantly associated with WA, GPC, ST and FT in all four environments. Haplotype analysis of a 488 kb linkage disequilibrium (LD) block identified four haplotypes (Hap1–Hap4). Hap2 and Hap3 had significantly higher WA than Hap1, and Hap2 had significantly higher GPC and longer ST than Hap1 and Hap4 (P < 0.05); FT was significantly longer in Hap2 than in Hap1 and Hap4. Ten candidate genes were predicted, annotated with functions involving signal transduction, metabolic regulation, cell development and stress response.ConclusionsThis study identified 62 stable SNP loci, 19 pleiotropic loci, and two favorable haplotypes (Hap2 and Hap3) relevant to wheat quality improvement. These findings highlight the breeding potential of Chinese winter wheat landraces on the Qinghai–Tibet Plateau and provide targets for marker-assisted selection.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09821-y. Background Wheat (Triticum aestivum L.) is one of the most widely cultivated cereal crops, providing approximately 20% of global food calories and a substantial proportion of dietary protein [1]. Its unique gluten properties make it an ideal raw material for foods such as bread, noodles and steamed bread [2]. Rising living standards have increased the demand for high-quality wheat, making quality improvement a major breeding objective [3]. Wheat quality encompasses interrelated morphological, nutritional and processing components. Grain protein content is a key nutritional component of wheat grain and influences dough development time and dough stability time, both of which are closely linked to nutritional and processing quality [4]. Moisture content, dough development time and dough stability time also affect flour processing quality [5]. Wet gluten content indirectly reflects flour protein concentration and mainly affects processing properties such as elasticity and viscosity [6]. Grain hardness influences processing quality through its effects on flour yield, flour particle size and damaged starch content [7, 8]. As quantitative traits, these characters are controlled by multiple genes and are strongly influenced by the environment and by genotype-by-environment interactions. Identifying the underlying loci is therefore essential [9]. Genome-wide association studies (GWAS) have become an efficient approach for dissecting complex quantitative traits [10, 11]. GWAS have been widely applied in major crops, including wheat, rice [12], maize, cotton [13], and soybean [14], as well as other species [15–17]. Previous studies have demonstrated the utility of GWAS in dissecting the genetic architecture of wheat quality traits. In a panel of 126 wheat genotypes evaluated in 12 environments with a 35 K SNP array, six multi-locus GWAS models identified 67 quantitative trait nucleotides (QTNs) explaining 3%–44% of the phenotypic variation, including 17 stable and 14 novel loci [18]. Similarly, a 90 K SNP array study of 192 common wheat cultivars from southwest China detected 184 significant SNPs, 102 of which were associated with grain protein content. In 236 wheat varieties genotyped with a 55 K array, 15 markers on chromosomes 1B, 1D, 2 A, 2B, 2D, 3B, 3D, 5D and 7D were significantly associated with grain protein content and wet gluten content, explaining 4.2%–10.75% of the phenotypic variation. A further GWAS of 486 accessions with 19,552 SNP markers detected 35 stable quantitative trait loci (QTLs) for grain protein content, wet gluten content and grain hardness, explaining 2.60%–12.68% of the phenotypic variation. Among these, QNgh.cau-5D and QNGPC.cau-6 A harboured the hardness gene Pinb and the grain protein content gene NAM-a1, respectively [19–21]. Grain hardness is controlled by two major genes together with several minor modifier genes [22]. The major genes, puroindoline a and puroindoline b (Pina and Pinb), are located on chromosome 5D and constitute the molecular basis of grain hardness. The minor genes are located mainly on chromosomes 1 A, 2 A, 2D, 3 A, 5 A, 5B and 6D [23, 24]. However, most of these studies focused on modern cultivars or lowland germplasm, and winter wheat landraces adapted to high-altitude environments such as the Qinghai–Tibet Plateau have received little attention. The Qinghai–Tibet Plateau is a major agro-pastoral region with a distinctive high-altitude climate. Winter wheat landraces from this region have undergone long-term natural and artificial selection and may carry quality-related alleles absent from modern cultivars. Nevertheless, the genetic basis of quality traits in Chinese landraces grown at high altitude remains poorly understood. Multi-environment stability of these traits has also rarely been assessed. To address these gaps, eight quality traits—MC, WA, GPC, WGC, ST, FT, HD and TW—were measured by near-infrared spectroscopy in 279 Chinese winter wheat landraces grown in four environments on the Qinghai–Tibet Plateau. A GWAS was then conducted with a 120 K SNP array to identify stable and pleiotropic loci associated with these traits. Ten candidate genes were identified by subsequent screening. A major pleiotropic locus on chromosome 1B was further characterized by haplotype analysis. These results provide a basis for marker-assisted selection in wheat quality breeding. Materials and methods Plant materials The 279 winter wheat accessions used in this study were provided by the Qinghai Academy of Agriculture and Forestry Sciences (Supplementary Table S1) and comprised 112 accessions from the middle and lower Yangtze River winter wheat region (Hubei), 131 from the southwest winter wheat region (78 from Sichuan and 53 from Yunnan) and 36 from the Qinghai–Tibet winter wheat region (Tibet). The experimental sites were in Shanping Village, Hexi Town, Guide County, Hainan Tibetan Autonomous Prefecture, Qinghai Province (36.04°N, 101.38°E) and Ahetan Village, Chahandusi Township, Salar Autonomous County, Xunhua County, Haidong City, Qinghai Province (35.87°N, 102.38°E), at altitudes of 2235 m and 1780 m, respectively. Guide County has an annual precipitation of 250–559 mm, a frost-free period of 166 days and a mean annual temperature of 7.2–9.7 °C. Xunhua County has a mean annual precipitation of 265.9 mm, a frost-free period of 220 days and a mean annual temperature of 8.7 °C. The four site-year combinations were designated 2023 Guide (E1), 2023 Xunhua (E2), 2024 Guide (E3) and 2024 Xunhua (E4). Each trial was sown in a sequential arrangement with two replicates. Each accession was sown in a single 1 m row with 0.3 m row spacing. Border rows surrounded each trial, and crop management followed local agronomic practice. Phenotyping Plants were harvested and threshed manually at maturity. Grain of each accession was scanned with a DA7200 near-infrared analyser (Botong Ruihua Scientific Instrument Co., Ltd.) to determine MC, WA, GPC, WGC, ST, FT, HD and TW. The near-infrared calibration model was developed and validated by the Grain Quality Supervision and Inspection Center of the Ministry of Agriculture and Rural Affairs of China (Tai’an, Shandong Province). GWAS Young leaves were collected from the 279 landraces at the seedling stage. DNA extraction and genotyping with a wheat 120 K SNP array were performed by Chengdu Tiancheng Future Company (https://www.tcuni.com). Quality control of the raw genotype data was performed in PLINK. SNPs with > 20% missing data, accessions with > 10% missing genotypes and SNPs with a minor allele frequency (MAF) < 5% were removed. After filtering, 82,308 polymorphic SNPs were retained. The screened SNP data were analyzed using ADMIXTURE software. The number of subpopulations (K) was tested from 1 to 18, with five independent runs per K value, and results were averaged. The K value with the highest ΔK was taken as the optimal number of subpopulations. GWAS was performed with the Q + K mixed linear model (MLM) implemented in TASSEL 5.0, and SNPs with P < 0.001 were considered significantly associated with the target traits. For candidate gene identification, a physical interval of ± 5 Mb (total window 10 Mb) around each significant SNP was defined based on the genome-wide LD decay distance. All annotated genes within these intervals were retrieved from the Chinese Spring reference genome (IWGSC RefSeq v1.0, https://www.wheatgenome.org/). Functional annotations were obtained by BLASTx searches against the wheat TGACv1 (https://plants.ensembl.org/), NCBI (https://www.ncbi.nlm.nih.gov/) and WheatOmics (https://wheatomics.sdau.edu.cn/) databases. Expression patterns of the retrieved genes were examined using RNA-seq data in the WheatOmics database. Data processing Descriptive statistics and correlation analyses of the phenotypic data were performed in Microsoft Excel 2021 and IBM SPSS Statistics 26. SNP density along chromosomes was plotted with the R package CMplot. Population structure analysis was performed using STRUCTURE 2.3.4 software. LD decay was analyzed in TASSEL 5.0, and haplotypes were identified and visualized in Haploview 4.1. Results Phenotype and correlation analysis The eight quality traits (MC, WA, GPC, WGC, ST, FT, HD and TW) differed among the four environments (E1–E4) in the 279 winter wheat landraces (Table 1). The coefficients of variation across environments were 3.21%–5.07%, 3.09%–6.45%, 9.51%–14.60%, 9.95%–15.80%, 12.59%–24.13%, 13.44%–26.19%, 6.38%–10.41% and 1.56%–2.56%, respectively. TW showed the lowest variation (1.56%–2.56%) and FT the highest (13.44%–26.19%). All traits were approximately normally distributed (Fig. 1), consistent with polygenic quantitative inheritance and suitable for GWAS. The strongest correlation was between GPC and ST (r = 0.982), followed by ST and HD (r = 0.804) (Fig. 2). Correlation patterns were highly consistent across the four environments, indicating coordinated variation among these traits. Table 1Quality traits of the 279 winter wheat landracesTraitEnvironmentMinimum valueMaximum valueMean ± SDCV(%)H2MC(%)E18.2310.889.51 ± 0.474.94%0.74E27.5110.039.2 ± 0.44.35%E39.9113.0211.63 ± 0.595.07%E49.8811.8110.9 ± 0.353.21%WA(%)E131.267.459.26 ± 3.826.45%0.63E236.266.358.69 ± 3.666.24%E353.469.263.36 ± 2.093.30%E453.968.0361.88 ± 1.913.09%GPC(%)E19.2620.3614.01 ± 2.0414.60%0.71E211.8920.2715.87 ± 1.519.51%E39.3524.1913.83 ± 1.9514.10%E412.4424.1916.93 ± 1.8410.87%WGC(%)E117.3241.0228.49 ± 4.3215.16%0.72E221.2939.7731.77 ± 3.169.95%E314.6449.0427.84 ± 4.415.80%E423.9549.0734.49 ± 3.7710.93%ST(min)E12.413.58.53 ± 2.023.45%0.83E22.812.67.5 ± 1.8124.13%E35.6316.6711.42 ± 2.0818.21%E48.1315.9512.15 ± 1.5312.59%FT(min)E11.5352.94 ± 0.7726.19%0.76E21.84.63.14 ± 0.6520.70%E31.85.73.42 ± 0.8324.27%E42.535.934.39 ± 0.5913.44%HD(%)E1418462.25 ± 6.4810.41%0.66E24371.559.35 ± 5.379.04%E346.3377.3364.48 ± 5.879.13%E449.6775.3365.19 ± 4.166.38%TW(g/l)E1737837789.93 ± 16.822.13%0.79E2717883776.68 ± 19.92.56%E3750829.67791.53 ± 12.311.56%E4749.67828.33792.67 ± 13.121.66%E1, 2023 Guide; E2, 2023 Xunhua; E3, 2024 Guide; E4, 2024 Xunhua. The same abbreviations apply throughout Table 1Quality traits of the 279 winter wheat landracesTraitEnvironmentMinimum valueMaximum valueMean ± SDCV(%)H2MC(%)E18.2310.889.51 ± 0.474.94%0.74E27.5110.039.2 ± 0.44.35%E39.9113.0211.63 ± 0.595.07%E49.8811.8110.9 ± 0.353.21%WA(%)E131.267.459.26 ± 3.826.45%0.63E236.266.358.69 ± 3.666.24%E353.469.263.36 ± 2.093.30%E453.968.0361.88 ± 1.913.09%GPC(%)E19.2620.3614.01 ± 2.0414.60%0.71E211.8920.2715.87 ± 1.519.51%E39.3524.1913.83 ± 1.9514.10%E412.4424.1916.93 ± 1.8410.87%WGC(%)E117.3241.0228.49 ± 4.3215.16%0.72E221.2939.7731.77 ± 3.169.95%E314.6449.0427.84 ± 4.415.80%E423.9549.0734.49 ± 3.7710.93%ST(min)E12.413.58.53 ± 2.023.45%0.83E22.812.67.5 ± 1.8124.13%E35.6316.6711.42 ± 2.0818.21%E48.1315.9512.15 ± 1.5312.59%FT(min)E11.5352.94 ± 0.7726.19%0.76E21.84.63.14 ± 0.6520.70%E31.85.73.42 ± 0.8324.27%E42.535.934.39 ± 0.5913.44%HD(%)E1418462.25 ± 6.4810.41%0.66E24371.559.35 ± 5.379.04%E346.3377.3364.48 ± 5.879.13%E449.6775.3365.19 ± 4.166.38%TW(g/l)E1737837789.93 ± 16.822.13%0.79E2717883776.68 ± 19.92.56%E3750829.67791.53 ± 12.311.56%E4749.67828.33792.67 ± 13.121.66%E1, 2023 Guide; E2, 2023 Xunhua; E3, 2024 Guide; E4, 2024 Xunhua. The same abbreviations apply throughout Quality traits of the 279 winter wheat landraces E1, 2023 Guide; E2, 2023 Xunhua; E3, 2024 Guide; E4, 2024 Xunhua. The same abbreviations apply throughout Fig. 1Normal distribution of wheat grain quality traits Normal distribution of wheat grain quality traits Fig. 2Correlations among grain quality traits in the 279 winter wheat landraces Correlations among grain quality traits in the 279 winter wheat landraces Linkage disequilibrium and population structure analysis The SNPs covered all 21 chromosomes but were unevenly distributed within them (Fig. 3). Quality filtering of the raw data yielded 82,308 genome-wide SNPs. The A genome contained 32,888 SNPs, averaging 4,698 per chromosome. The B genome contained 34,575 SNPs, averaging 4,939 per chromosome. The D genome contained 14,845 SNPs, averaging 2,120 per chromosome. Fig. 3Genome-wide SNP marker density Genome-wide SNP marker density Genome-wide LD decay was analyzed using TASSEL 5.0, with the physical distance at r² = 0.2 was used as the LD decay distance. The LD decay distances for the A and B genomes exceeded 30 Mb (Fig. 4A). Because large intervals contain many unrelated genes and hinder functional validation, a narrower window commonly used in wheat GWAS was adopted [25]. Candidate genes were screened within ± 5 Mb of each significant SNP (10 Mb window). The population structure was analyzed using STRUCTURE V2.3.4 software. When the number of subpopulations (K) was set to 10, the ΔK value reached its maximum, indicating that an optimal number (K = 10) was identified for this population (Fig. 4B). The composition of these subpopulations is shown in Fig. 4D. Subpopulations 1–10 contained 41, 12, 39, 20, 35, 25, 44, 22, 19 and 22 accessions, representing 14.70%, 4.31%, 13.98%, 7.17%, 12.54%, 8.96%, 15.77%, 7.89%, 6.81% and 7.89% of the panel, respectively. Principal component analysis (PCA) supported the same grouping (Fig. 4C). Association analyses were therefore conducted with an MLM incorporating population structure (Q, K = 10) and kinship (K). Fig. 4Population structure and linkage disequilibrium analysis of the 279 wheat varieties. a: Genome-wide LD decay distance; b: Population structure; c: Principal component analysis; d: Subpopulation structure distribution Population structure and linkage disequilibrium analysis of the 279 wheat varieties. a: Genome-wide LD decay distance; b: Population structure; c: Principal component analysis; d: Subpopulation structure distribution Association analysis of quality-related traits GWAS was performed for each of the four environments using the 82,308 filtered SNPs from the 120 K array. In total, 145 SNPs were significantly associated with the eight quality traits (P < 0.001). Of these, 62 were detected in two or more environments and were considered stable; these were located on chromosomes 1 A (5), 1B (3), 1D (1), 2 A (3), 2B (4), 2D (1), 3 A (2), 3B (5), 3D (3), 4 A (3),4B (2), 4D (1), 5 A (9), 5B (3), 5D (1), 6 A (2),6B (4), 6D (1), 7 A (3), 7B (2), and 7D (4). These loci explained 3.86%–15.64% of the phenotypic variation (Fig. 5; Tables 2 and 3). Fig. 5Manhattan plots for the eight grain quality traits Manhattan plots for the eight grain quality traits Table 2Representative SNP loci significantly associated with each quality traitTraitMarkerChr.Position/MbP vauleEnvironmentR2/%MC3B:779,661,9913B779.664.73E− 11 ~ 2.27E− 04E3/E45.20 ~ 15.646D:109,087,5966D109.091.05E− 10 ~ 2.15E− 04E2/E3/E44.87 ~ 14.117B:591,580,2167B591.586.71E− 11 ~ 8.86E− 06E2/E3/E46.91 ~ 14.33WA3B:276,114,8843B276.111.59E− 04 ~ 8.23E− 04E1/E33.98 ~ 5.033B:429,107,1633B429.116.02E− 05 ~ 2.83E− 04E1/E34.65 ~ 5.685 A:548,092,4975 A548.092.31E− 04 ~ 2.43E− 04E1/E34.76 ~ 4.80GPC3B:205,243,2473B205.244.01E− 09 ~ 2.42E− 04E2/E34.77 ~ 11.815D:419,910,9865D419.915.38E− 09 ~ 2.15E− 05E1/E36.34 ~ 11.596B:37,013,3836B37.011.77E− 08 ~ 3.73E− 04E2/E34.49 ~ 10.88WGC4B:612,010,3064B612.011.64E− 08 ~ 6.48E− 04E1/E3/E44.12 ~ 10.935B:711,893,1345B711.891.24E− 08 ~ 1.31E− 05E3/E46.66 ~ 11.117 A:197,943,6657 A197.946.86E− 08 ~ 1.83E− 04E3/E44.95 ~ 10.03ST2 A:731,613,9222 A731.611.16E− 04 ~ 6.60E− 04E1/E44.12 ~ 5.237 A:33,932,4327 A33.938.58E− 06 ~ 7.15E− 05E1/E25.54 ~ 6.937D:55,212,8927D55.211.25E− 04 ~ 2.39E− 04E1/E44.82 ~ 5.26FT5 A:598,262,0875 A598.261.92E− 04 ~ 5.90E− 04E2/E44.19 ~ 4.926B:132,847,9316B132.851.13E− 04 ~ 5.22E− 04E1/E44.27 ~ 5.257D:28,939,7997D28.941.76E− 05 ~ 1.19E− 04E2/E45.24 ~ 6.47HD3 A:568,594,3883 A568.592.23E− 06 ~ 6.07E− 04E1/E44.18 ~ 7.786 A:17,582,7436 A17.581.17E− 07 ~ 2.98E− 04E1/E24.64 ~ 9.656B:106,445,5316B106.458.20E− 07 ~ 3.03E− 04E1/E24.63 ~ 8.42TW3D:66,362,6503D66.361.08E− 04 ~ 7.21E− 04E2/E3/E44.09 ~ 5.334 A:2,783,1584 A2.788.65E− 05 ~ 9.83E− 04E3/E43.86 ~ 5.445 A:664,960,4835 A664.963.38E− 06 ~ 7.51E− 04E3/E4/E24.04 ~ 7.54 Table 2Representative SNP loci significantly associated with each quality traitTraitMarkerChr.Position/MbP vauleEnvironmentR2/%MC3B:779,661,9913B779.664.73E− 11 ~ 2.27E− 04E3/E45.20 ~ 15.646D:109,087,5966D109.091.05E− 10 ~ 2.15E− 04E2/E3/E44.87 ~ 14.117B:591,580,2167B591.586.71E− 11 ~ 8.86E− 06E2/E3/E46.91 ~ 14.33WA3B:276,114,8843B276.111.59E− 04 ~ 8.23E− 04E1/E33.98 ~ 5.033B:429,107,1633B429.116.02E− 05 ~ 2.83E− 04E1/E34.65 ~ 5.685 A:548,092,4975 A548.092.31E− 04 ~ 2.43E− 04E1/E34.76 ~ 4.80GPC3B:205,243,2473B205.244.01E− 09 ~ 2.42E− 04E2/E34.77 ~ 11.815D:419,910,9865D419.915.38E− 09 ~ 2.15E− 05E1/E36.34 ~ 11.596B:37,013,3836B37.011.77E− 08 ~ 3.73E− 04E2/E34.49 ~ 10.88WGC4B:612,010,3064B612.011.64E− 08 ~ 6.48E− 04E1/E3/E44.12 ~ 10.935B:711,893,1345B711.891.24E− 08 ~ 1.31E− 05E3/E46.66 ~ 11.117 A:197,943,6657 A197.946.86E− 08 ~ 1.83E− 04E3/E44.95 ~ 10.03ST2 A:731,613,9222 A731.611.16E− 04 ~ 6.60E− 04E1/E44.12 ~ 5.237 A:33,932,4327 A33.938.58E− 06 ~ 7.15E− 05E1/E25.54 ~ 6.937D:55,212,8927D55.211.25E− 04 ~ 2.39E− 04E1/E44.82 ~ 5.26FT5 A:598,262,0875 A598.261.92E− 04 ~ 5.90E− 04E2/E44.19 ~ 4.926B:132,847,9316B132.851.13E− 04 ~ 5.22E− 04E1/E44.27 ~ 5.257D:28,939,7997D28.941.76E− 05 ~ 1.19E− 04E2/E45.24 ~ 6.47HD3 A:568,594,3883 A568.592.23E− 06 ~ 6.07E− 04E1/E44.18 ~ 7.786 A:17,582,7436 A17.581.17E− 07 ~ 2.98E− 04E1/E24.64 ~ 9.656B:106,445,5316B106.458.20E− 07 ~ 3.03E− 04E1/E24.63 ~ 8.42TW3D:66,362,6503D66.361.08E− 04 ~ 7.21E− 04E2/E3/E44.09 ~ 5.334 A:2,783,1584 A2.788.65E− 05 ~ 9.83E− 04E3/E43.86 ~ 5.445 A:664,960,4835 A664.963.38E− 06 ~ 7.51E− 04E3/E4/E24.04 ~ 7.54 Representative SNP loci significantly associated with each quality trait Table 3Stable loci significantly associated with two or more grain quality traitsTraitMarkerChr.Position/MbP vauleEnvironmentR2/%WA/GPC/HD1B:175,681,1541B175.68 ~ 176.97.47E− 08 ~ 7.34E− 04E3/E4/E24.11 ~ 10.11GPC/HD1B:340,043,0471B340.04 ~ 346.595.38E− 09 ~ 8.40E− 04E2/E3/E13.97 ~ 11.62GPC/HD1B:354,204,6881B354.29.97E− 08 ~ 3.81E− 04E2/E34.48 ~ 9.79WA/GPC/ST/FT1B:610,760,0011B610.76 ~ 611.293.57E− 08 ~ 9.20E− 04E2/E3/E1/E43.91 ~ 10.44GPC/WGC1D:350,306,1841D350.313.54E− 10 ~ 9.32E− 04E1/E2/E33.90 ~ 13.27GPC/WGC2B:531,265,2412B531.27 ~ 534.411.15E− 05 ~ 7.42E− 04E1/E34.05 ~ 6.72MC/WGC2D:64,667,0212D64.67 ~ 64.761.98E− 08 ~ 1.25E− 04E3/E45.20 ~ 10.81GPC/HD3 A:359,918,6993 A359.924.15E− 06 ~ 8.43E− 04E1/E3/E23.97 ~ 7.40GPC/HD3 A:569,854,5963 A569.85 ~ 569.871.12E− 06 ~ 7.37E− 04E1/E2/E3/E44.04 ~ 8.24GPC/HD3B:847,832,2393B847.83 ~ 847.882.54E− 07 ~ 2.82E− 04E1/E2/E34.67 ~ 9.19GPC/HD3D:615,275,8783D615.282.14E− 06 ~ 5.71E− 04E2/E3/E14.22 ~ 7.83WA/GPC/WGC4D:479,734,9724D479.731.79E− 08 ~ 9.14E− 04E1/E2/E3/E43.91 ~ 10.87GPC/HD4D:7,813,8034D7.812.24E− 09 ~ 8.69E− 04E2/E3/E13.93 ~ 12.17GPC/HD5 A:162,362,9685 A162.365.30E− 07 ~ 4.93E− 04E2/E3/E14.31 ~ 8.73WA/GPC5 A:427,530,5045 A427.532.90E− 05 ~ 6.96E− 04E2/E34.09 ~ 6.15MC/WGC5 A:689,016,2445 A689.024.74E− 06 ~ 3.64E− 04E1/E2/E3/E44.51 ~ 7.29WA/GPC5B:117,007,7035B117.019.67E− 07 ~ 8.03E− 04E2/E34.01 ~ 8.37GPC/WGC6D:403,951,2996D403.951.81E− 06 ~ 8.08E− 04E1/E33.99 ~ 7.91MC/WGC7 A:50,772,5067 A50.778.30E− 08 ~ 7.31E− 04E2/E3/E44.06 ~ 9.90 Table 3Stable loci significantly associated with two or more grain quality traitsTraitMarkerChr.Position/MbP vauleEnvironmentR2/%WA/GPC/HD1B:175,681,1541B175.68 ~ 176.97.47E− 08 ~ 7.34E− 04E3/E4/E24.11 ~ 10.11GPC/HD1B:340,043,0471B340.04 ~ 346.595.38E− 09 ~ 8.40E− 04E2/E3/E13.97 ~ 11.62GPC/HD1B:354,204,6881B354.29.97E− 08 ~ 3.81E− 04E2/E34.48 ~ 9.79WA/GPC/ST/FT1B:610,760,0011B610.76 ~ 611.293.57E− 08 ~ 9.20E− 04E2/E3/E1/E43.91 ~ 10.44GPC/WGC1D:350,306,1841D350.313.54E− 10 ~ 9.32E− 04E1/E2/E33.90 ~ 13.27GPC/WGC2B:531,265,2412B531.27 ~ 534.411.15E− 05 ~ 7.42E− 04E1/E34.05 ~ 6.72MC/WGC2D:64,667,0212D64.67 ~ 64.761.98E− 08 ~ 1.25E− 04E3/E45.20 ~ 10.81GPC/HD3 A:359,918,6993 A359.924.15E− 06 ~ 8.43E− 04E1/E3/E23.97 ~ 7.40GPC/HD3 A:569,854,5963 A569.85 ~ 569.871.12E− 06 ~ 7.37E− 04E1/E2/E3/E44.04 ~ 8.24GPC/HD3B:847,832,2393B847.83 ~ 847.882.54E− 07 ~ 2.82E− 04E1/E2/E34.67 ~ 9.19GPC/HD3D:615,275,8783D615.282.14E− 06 ~ 5.71E− 04E2/E3/E14.22 ~ 7.83WA/GPC/WGC4D:479,734,9724D479.731.79E− 08 ~ 9.14E− 04E1/E2/E3/E43.91 ~ 10.87GPC/HD4D:7,813,8034D7.812.24E− 09 ~ 8.69E− 04E2/E3/E13.93 ~ 12.17GPC/HD5 A:162,362,9685 A162.365.30E− 07 ~ 4.93E− 04E2/E3/E14.31 ~ 8.73WA/GPC5 A:427,530,5045 A427.532.90E− 05 ~ 6.96E− 04E2/E34.09 ~ 6.15MC/WGC5 A:689,016,2445 A689.024.74E− 06 ~ 3.64E− 04E1/E2/E3/E44.51 ~ 7.29WA/GPC5B:117,007,7035B117.019.67E− 07 ~ 8.03E− 04E2/E34.01 ~ 8.37GPC/WGC6D:403,951,2996D403.951.81E− 06 ~ 8.08E− 04E1/E33.99 ~ 7.91MC/WGC7 A:50,772,5067 A50.778.30E− 08 ~ 7.31E− 04E2/E3/E44.06 ~ 9.90 Stable loci significantly associated with two or more grain quality traits Thirteen SNP loci significantly associated with MC were distributed on chromosomes 1B, 1D, 2B, 2D, 3B, 4 A, 4B, 5 A, 6 A, 6D, and 7B, explaining 4.04%–15.64% of the phenotypic variation. Six loci significantly associated with WA were located on chromosomes 1 A, 3B, 3D, 5 A, and 7D, explaining 3.88%–5.68% of the phenotypic variation. Nine SNP loci significantly associated with GPC were distributed on chromosomes 1 A, 2 A, 2B, 3B, 3D, 5B, 5D, 6B, and 7D. Nine loci associated with WGC were located on chromosomes 1 A, 2 A, 4B, 5 A, 5B, 6B, and 7 A, explaining 3.92%–11.11% of the phenotypic variation. Six loci associated with ST were located on chromosomes 1 A, 2 A, 7 A, and 7D, explaining 3.93%–6.93% of the phenotypic variation. Six loci associated with FT were located on chromosomes 2B, 5 A, 5B, 6B, and 7D, explaining 3.86%–6.47% of the phenotypic variation. Ten SNP loci associated with HD were located on chromosomes 1 A, 1B, 3 A, 3B, 4 A, 5 A, 6 A, 6B, and 7B, explaining 3.86%–9.65% of the phenotypic variation. Three loci associated with TW were located on chromosomes 3D, 4 A, and 5 A, explaining 3.86%–7.54% of the phenotypic variation. Among the stable loci, 19 exhibited associations with two or more grain quality traits. The loci 1B:610,760,001, 3 A:569,854,596, 4D:479,734,972, and 5 A:689,016,244 were detected across all four environments. The locus 1B:610,760,001 was significantly associated with WA, GPC, ST, and FT. The locus 4D:479,734,972 was significantly associated with WA, GPC, and WGC, explaining 3.91%–10.87% of the phenotypic variation. The loci 1B:340,043,047, 1B:354,204,688, 3 A:359,918,699, 3 A:569,854,596, 3B:847,832,239, 3D:615,275,878, 4D:7,813,803, and 5 A:162,362,968 were significantly associated with GPC and HD, explaining 3.93%–12.17% of the phenotypic variation. The locus 1B:175,681,154 was significantly associated with WA, GPC, and HD, explaining 4.11%–10.11% of the phenotypic variation. The loci 1D:350,306,184, 2B:531,265,241, and 6D:403,951,299 were significantly associated with GPC and WGC, explaining 3.90%–13.27% of the phenotypic variation. The loci 2D:64,667,021, 5 A:689,016,244, and 7 A:50,772,506 were significantly associated with MC and WGC, explaining 4.06%–10.81% of the phenotypic variation. The loci 5 A:427,530,504 and 5B:117,007,703 were significantly associated with WA and GPC, explaining 4.01%–8.37% of the phenotypic variation (Table 3). Haplotype analysis Haplotype analysis of the pleiotropic loci explaining a high proportion of phenotypic variation revealed strong linkage disequilibrium (LD) in the regions harboring these SNPs, forming distinct LD blocks. Locus 1B:610,760,001, located at 610.76–611.29 Mb on chromosome 1B, together with four downstream SNP markers, formed a 488 kb LD block (Fig. 6A). This block contained four haplotypes (Hap1–Hap4), carried by 156, 72, 11, and 10 accessions, respectively (Fig. 6B). Mean values of WA, GPC, ST, and FT were 59.63%, 14.98%, 9.72 min, and 3.35 min for Hap1; 60.80%, 15.45%, 10.24 min, and 3.66 min for Hap2; 61.13%, 15.22%, 10.09 min, and 3.61 min for Hap3; and 60.17%, 14.87%, 10.03 min, and 3.48 min for Hap4. WA was significantly higher in Hap2 (60.80%) and Hap3 (61.13%) than in Hap1 (59.63%) (P < 0.05). GPC was significantly higher in Hap2 (15.45%) than in Hap1 (14.98%) and Hap4 (14.87%) (P < 0.05). ST was significantly longer in Hap2 (10.24 min) than in Hap1 (9.72 min) (P < 0.05). FT was significantly longer in Hap2 (3.66 min) than in Hap1 (3.35 min) and Hap4 (3.48 min) (P < 0.05). GPC and ST did not differ significantly between Hap2 and Hap3 (Fig. 6C). The haplotype distribution frequencies were ranked as follows: Hap1 (58.4%) > Hap2 (29.6%) > Hap3 (7.3%) > Hap4 (3.4%) (Fig. 6B). Hap1 was most prevalent in the Qinghai–Tibet winter wheat region (85.7%), followed by the middle and lower Yangtze River region (66.0%) and the southwest region (54.5%). Hap2 showed a distinct regional pattern: southwest region (44.6%) > middle and lower Yangtze River region (15.0%) > Qinghai–Tibet region (10.7%). Hap3 was more frequent in the middle and lower Yangtze River region (10.0%) than in the Qinghai–Tibet region (3.6%). Hap4 was largely confined to the middle and lower Yangtze River region (9.0%), with minimal representation in the southwest region (0.8%). Fig. 6Haplotype analysis of the 1B:610,760,001 locus. a. LD block analysis of the 1B:610,760,001 locus; b, the four haplotypes and their allele combinations c, differences in WA, GPC, ST, and FT among the four haplotypes Haplotype analysis of the 1B:610,760,001 locus. a. LD block analysis of the 1B:610,760,001 locus; b, the four haplotypes and their allele combinations c, differences in WA, GPC, ST, and FT among the four haplotypes Prediction of candidate genes for wheat grain quality A total of 1,247 annotated genes were retrieved from the genomic intervals of the 62 stable SNP loci. Candidate genes were then prioritized using a stepwise filtering strategy. First, genes within pleiotropic loci associated with two or more quality traits (Table 3) were given precedence, as such regions are more likely to harbor regulators with broad effects on grain quality. Second, expression profiles were examined in the WheatOmics database, and only genes highly expressed in grain or spike tissues were retained. Third, genes whose Arabidopsis or rice orthologs have established functions in grain development or quality regulation were preferentially selected. Finally, genes previously reported to be associated with wheat quality traits were given additional consideration. This process yielded ten candidate genes potentially involved in grain quality formation (Table 4). TraesCS1B02G138100 encodes an auxin-responsive protein involved in auxin signaling, a phytohormone pathway regulating plant growth and development. TraesCS1B02G377900 encodes an uncharacterized protein homologous to Arabidopsis At4g29660. TraesCS2B02G371700 encodes 6,7-dimethyl-8-ribityllumazine synthase. TraesCS3A02G204900 encodes a homolog of ROOT HAIR DEFECTIVE 3 (RHD3). TraesCS3D02G543100 encodes a DnaJ protein homologous to P58IPK. TraesCS4D02G022000 encodes nucleolar GTP-binding protein 1. TraesCS4D02G312300 encodes mitogen-activated protein kinase kinase 6 (MKK6). TraesCS4D02G316500 encodes a non-specific serine/threonine protein kinase. TraesCS5A02G212800 encodes a jasmonate ZIM-domain (JAZ) protein that represses jasmonic acid signaling and functions in jasmonate-induced resistance to bacterial blight. TraesCS6D02G294500 encodes an SNW/SKI-interacting protein. Table 4Screening for candidate gene informationTraitMarkerChromosomePosition/MbGene IDGene annotation or coding proteinWA/GPC/HD1B:175,681,1541B175.68 ~ 176.9TraesCS1B02G138100Auxin-responsive proteinWA/GPC/ST/FT1B:610,760,0011B610.76 ~ 611.29TraesCS1B02G377900Uncharacterized protein At4g29660GPC/WGC2B:531,265,2412B531.27 ~ 534.41TraesCS2B02G3717006,7-dimethyl-8-ribityllumazine synthaseGPC/HD3 A:569,854,5963 A569.85 ~ 569.87TraesCS3A02G204900Protein ROOT HAIR DEFECTIVE 3 homologGPC/HD3D:615,275,8783D615.28TraesCS3D02G543100DnaJ protein P58IPK homologGPC/HD4D:7,813,8034D7.81TraesCS4D02G022000Nucleolar GTP-binding protein 1WA/GPC/WGC4D:479,734,9724D479.73TraesCS4D02G312300Mitogen-activated protein kinase kinase 6WA/GPC/WGC4D:479,734,9724D479.73TraesCS4D02G316500Threonine protein kinaseWA/GPC5 A:427,530,5045 A427.53TraesCS5A02G212800Repressor of jasmonic acid signalingGPC/WGC6D:403,951,2996D403.95TraesCS6D02G294500SNW/SKI-interacting protein Table 4Screening for candidate gene informationTraitMarkerChromosomePosition/MbGene IDGene annotation or coding proteinWA/GPC/HD1B:175,681,1541B175.68 ~ 176.9TraesCS1B02G138100Auxin-responsive proteinWA/GPC/ST/FT1B:610,760,0011B610.76 ~ 611.29TraesCS1B02G377900Uncharacterized protein At4g29660GPC/WGC2B:531,265,2412B531.27 ~ 534.41TraesCS2B02G3717006,7-dimethyl-8-ribityllumazine synthaseGPC/HD3 A:569,854,5963 A569.85 ~ 569.87TraesCS3A02G204900Protein ROOT HAIR DEFECTIVE 3 homologGPC/HD3D:615,275,8783D615.28TraesCS3D02G543100DnaJ protein P58IPK homologGPC/HD4D:7,813,8034D7.81TraesCS4D02G022000Nucleolar GTP-binding protein 1WA/GPC/WGC4D:479,734,9724D479.73TraesCS4D02G312300Mitogen-activated protein kinase kinase 6WA/GPC/WGC4D:479,734,9724D479.73TraesCS4D02G316500Threonine protein kinaseWA/GPC5 A:427,530,5045 A427.53TraesCS5A02G212800Repressor of jasmonic acid signalingGPC/WGC6D:403,951,2996D403.95TraesCS6D02G294500SNW/SKI-interacting protein Screening for candidate gene information Discussion Phenotypic variation of wheat grain traits Wheat grain quality comprises multiple traits, and systematic investigation of high-quality germplasm resources is critical for advancing genetic improvement of wheat quality [26]. In the study, eight quality traits were measured in 279 winter wheat landraces from different wheat regions and grown across four environments. All traits showed extensive phenotypic variation. ST and FT exhibited the highest coefficients of variation, indicating substantial genetic diversity suitable for GWAS [27]. Correlation analysis revealed strong positive associations between GPC and both ST and HD, consistent with previous findings [28]. WA was positively correlated with TW, HD, and ST, suggesting that varieties with higher WA possess greater water-binding capacity, which is closely linked to gluten network formation and overall dough processing quality. A weak negative correlation was observed between GPC and WGC, differing from previous findings [29]. This may reflect the unique genetic background of the population or asynchronous accumulation of protein and gluten under varying environmental conditions. Similar patterns have been reported in other wheat germplasm, highlighting the importance of non-synergistic trait variation in specific populations [30]. MC was positively correlated with FT and HD. The strong positive association between FT and both HD and TW indicate that these traits may share common genetic regulatory mechanisms. These results reveal a strongly correlated trait cluster comprising GPC, ST, HD, WA, and TW, suggesting that these traits may be regulated by shared genetic factors or coordinated developmental processes. Association analysis of wheat grain quality traits Genome-wide association studies (GWAS) provide an effective approach for dissecting the genetic basis of complex quantitative traits and have been widely applied in wheat quality research [31]. A 120 K SNP array was used to genotype 279 wheat accessions. A total of 62 stable SNP loci were detected across the four environments (P < 0.001), with individual loci explaining 3.86%–15.64% of the phenotypic variation, consistent with previous GWAS [32]. These stable loci were distributed across all 21 wheat chromosomes. Comparative analysis revealed that several loci colocalized with previously reported QTLs or resided within adjacent genomic regions. A locus significantly associated with HD was previously detected at 179.5–184.5 Mb on chromosome 1B [33], 3.82 Mb from the 1B:175,681,154 locus (harboring candidate gene TraesCS1B02G138100) identified here Based on the LD decay distance (≈ 10 Mb genome-wide), these may represent the same locus. The pleiotropic locus 1B:610,760,001 (harboring candidate gene TraesCS1B02G377900) showed significant associations with WA, GPC, ST, and FT. A locus (S1B_620015835) was previously reported at 615.02 Mb on chromosome 1B [34], 3.73 Mb from the 1B:610,760,001 locus identified here, suggesting they may represent the same QTL. The pleiotropic locus 2B:531,265,241 is located approximately 5.04 Mb from previously reported GPC- and WGC-associated loci on chromosome 2B [35], suggesting they may represent different markers within the same linkage block. The candidate gene TraesCS3D02G543100 located near loci associated with GPC and HD on chromosome 3D, aligns with previous findings [36, 37]. The locus at 403.95 Mb on chromosome 6D, associated with GPC and WGC, is located 2.25 Mb away from a previously reported locus at 406.20 Mb [38]. This proximity, suggest the presence of a potential gene clusters in this region that regulates controlling multiple quality traits. Several novel loci associated with grain quality traits were identified. The locus 3 A:569,854,596 (harboring TraesCS3A02G204900) on chromosome 3 A showed significant associations with GPC and HD. This locus is 52.81–52.83 Mb from a GPC-associated locus previously reported at 517.04 Mb on chromosome 3 A [39], indicating that they are distinct. Therefore, 3 A:569,854,596 represents a novel locus. The locus 4D:7,813,803 at 7.81 Mb demonstrated significant associations with GPC and HD across three environments, differing from a pleiotropic QTL previously reported at 75.15 Mb on chromosome 4D [40]. A SNP locus associated with moisture content was reported at 208.52–208.57 Mb on chromosome 4D by Song et al. [41]. This locus is physically distant (> 200 Mb) from the 4D:479,734,972 locus identified here, which was associated with WA, GPC, and WGC across four environments. This suggests that 4D:479,734,972 represents a novel locus. The locus 5 A:427,530,504 at 427.53 Mb explained 6.15% of the phenotypic variation. This locus has not been previously reported and represents a novel candidate for improving wheat grain quality traits. These colocalization findings validate the reliability of the present results and provide important clues for subsequent candidate gene identification. More than 30.6% of the stable loci showed pleiotropic effects, consistent with the strong phenotypic correlations among quality traits and reflecting the complex genetic architecture underlying grain quality. The pleiotropic locus 1B:610,760,001 was stably associated with four traits (WA, GPC, ST, and FT), suggesting that it may influence multiple quality parameters through shared regulatory pathways. The identification of such pleiotropic loci provides valuable targets for simultaneous improvement of multiple quality traits through marker-assisted selection. Haplotype diversity and its geographical distribution Haplotype analysis reveals allele combination patterns and provides a direct basis for molecular marker-assisted selection [42]. Haplotype analysis was conducted on a 488 kb LD block harboring a pleiotropic locus on chromosome 1B. Four haplotypes (Hap1–Hap4) were identified with frequencies of 58.4%, 29.6%, 7.3%, and 3.4%, respectively. Phenotypic comparisons showed that Hap2 and Hap3 outperformed the predominant Hap1 for multiple quality traits. Hap2 exhibited significantly higher GPC and longer ST and FT than Hap1 and Hap4 (P < 0.05) and, together with Hap3, showed significantly higher WA than Hap1. Hap3 performed comparably to Hap2 for WA, GPC, and ST but did not differ significantly from Hap1 for GPC or ST, likely because of its limited sample size (n = 11). This result is similar to high-GPC haplotypes (Hap2 and Hap3) previously identified on chromosome 5D [43], indicating that favorable haplotypes on different chromosomes can serve as candidate resources for quality improvement. Haplotype distribution frequencies differed significantly across geographical sources. The high-GPC haplotype Hap2 exhibited the highest frequency in the southwest region (44.6%), dropping to 15.0% in the middle and lower Yangtze River region and 10.7% in the Qinghai–Tibet region. The high-WA haplotype Hap3 was predominantly distributed in the middle and lower Yangtze River region (10.0%) and the Qinghai–Tibet region (3.6%). This geographical differentiation likely reflects regional variations in breeding history and environmental adaptation. The small sample size from the Qinghai–Tibet region (n = 28) limits interpretation of haplotype frequencies in this area. Future studies should prioritize increased sampling in this region to better define haplotype distribution patterns. Hap1 (58.4%), the most frequent haplotype, showed moderate performance across all four quality traits, whereas the superior haplotypes Hap2 and Hap3 together accounted for only 36.9% of the population, indicating that high-quality allelic variation remains underexploited [44]. Hap2 and Hap3 could be introrse into high-yielding genetic backgrounds through marker-assisted selection and hybridization to achieve simultaneous improvements in quality and yield. Candidate gene function prediction and potential mechanism of action Candidate genes were screened within a ± 5 Mb physical interval around each significant SNP, based on the genome-wide LD decay distance of approximately 10 Mb. From the 62 stable SNP loci, 1,247 genes were retrieved from the Chinese Spring reference genome (IWGSC RefSeq v1.0). Following filtration based on functional annotation, expression patterns (prioritizing genes highly expressed in grain or spike tissues in the WheatOmics database), and location within pleiotropic loci (Table 3), ten candidate genes potentially involved in grain quality formation were identified (Table 4). These genes encode proteins involved in signal transduction, metabolic regulation, cell development, and stress response, collectively forming a complex regulatory network influences grain quality. The TraesCS1B02G138100 encodes an auxin-responsive protein involved in plant growth, photomorphogenesis, and auxin signal transduction [45]. This gene may regulate cell elongation and tissue differentiation, potentially influencing grain hardness through effects on cell expansion and cell wall composition. TraesCS4D02G312300 encodes mitogen-activated protein kinase kinase 6 (MKK6). TraesCS4D02G316500 encodes a non-specific serine/threonine protein kinase. TraesCS5A02G212800 encodes a jasmonate ZIM-domain (JAZ) protein, which acts as a repressor of jasmonic acid signaling, and TraesCS6D02G294500 encodes an SNW/SKI-interacting protein. These genes play important roles in signal transduction and transcriptional regulation. The MAPK signaling pathway plays a central role in plant development and stress responses [46]. For example, constitutively active OsMKK6 has been shown to enhance salt tolerance and influence grain weight in rice [47]. Non-specific serine/threonine protein kinases play central roles in various biological processes, including mitosis, cell division, and photosynthesis [48, 49], as well as in wheat growth and stress resistance [50]. TraesCS5A02G212800 on chromosome 5 A encodes a repressor of jasmonic acid signaling involved in plant immune responses and the regulation of secondary metabolic pathways [51]. JAZ proteins may influence grain quality by modulating storage protein synthesis during grain development. TraesCS2B02G371700 encodes 6,7-dimethyl-8-ribityllumazine synthase. As a riboflavin is, a precursor for the coenzymes FAD and FMN, participates in key cellular energy metabolism reactions [52]. Expression of this synthase may influence energy supply during grain filling and thus regulate the rate of reserve accumulation. TraesCS3A02G204900 encodes a homolog of ROOT HAIR DEFECTIVE 3 (RHD3), which regulates cell pattern formation, differentiation, and growth in Arabidopsis by modulating transcriptional regulators [53]. TraesCS3D02G543100 encodes a DnaJ protein (P58IPK homolog) that functions as a co-chaperone in the heat shock protein Hsp40 complex and participates in protein folding and stress response [54]. Given the environmental stresses encountered during grain development, this gene likely ensures normal grain filling and quality by maintaining protein homeostasis. TraesCS4D02G022000 encodes nucleolar GTP-binding protein 1 (NOG1), which functions in plant immunity against bacterial pathogens [55] and regulates non-host resistance and stomatal defense. Among the ten candidate genes, two were prioritized based on their localization within stable pleiotropic loci and their functional annotations. The first, TraesCS1B02G377900, resides within the LD block of the pleiotropic locus 1B:610,760,001, which was stably associated with WA, GPC, ST, and FT across all four environments and linked to two favorable haplotypes (Hap2 and Hap3). These haplotypes offer direct targets for marker-assisted selection. Functionally, TraesCS1B02G377900 encodes an uncharacterized protein whose Arabidopsis ortholog (EMB2752) is mitochondrial-localized and essential for embryo development [56], suggesting a potential role in energy metabolism during grain filling. As this gene has not previously been associated with wheat quality, it represents a novel candidate for functional validation. The second candidate, TraesCS4D02G312300, is located within the pleiotropic locus 4D:479,734,972, which showed stable associated with WA, GPC, and WGC across all four environments. This gene encodes MKK6, whose rice ortholog (OsMKK6) influences grain weight when constitutively activated [47], indicating a conserved role in grain development. Given the stability and pleiotropic effects of this locus, it constitutes a key target for future functional studies. Functional validation of these candidate genes was beyond the scope of the present study. Future work will prioritize the developing of KASP markers for the favorable haplotypes and the validation of these candidate genes functions through transgenic or gene-editing approaches. Conclusion This study employed a 120 K SNP array to conduct genome-wide association analysis of eight grain quality traits in 279 winter wheat landraces grown on the Qinghai–Tibet Plateau. A total of 62 stable SNP loci and 19 pleiotropic loci were identified. Haplotype analysis identified two favorable haplotypes (Hap2 and Hap3) exhibiting superior WA, GPC, ST, and FT. Ten candidate genes potentially involved in grain quality formation were identified, with annotated functions in signal transduction, metabolic regulation, cell development, and stress response. These findings provide a basis for wheat quality improvement and marker-assisted breeding of new varieties. Supplementary Information Supplementary Material 1. Supplementary Material 1.